Anda di halaman 1dari 652

Metallurgical

Modelling
of Welding
SECOND EDITION
0YSTEIN GRONG
Norwegian University of
Science and Technology,
Department of Metallurgy,
N-7034 Trondheim, Norway

MATERIALS MODELLING SERIES

Editor: H. K. D. H. Bhadeshia
The University of Cambridge
Department of Materials Science
and Metallurgy

T H E INSTITUTE OF MATERIALS

Book 677
First published in 1997 by
The Institute of Materials
1 Carlton House Terrace
London SWlY 5DB
First edition (Book 557)
Published in 1994
The Institute of Materials 1997
All rights reserved
ISBNl 86125 036 3

Originally typeset by
PicA Publishing Services
Additional typesetting and corrections by
Fakenham Photosetting Ltd
Printed and bound in the UK at
The University Press, Cambridge

TO TORHILD, TORBJ0RN AND HAVARD


without your support, this book would never have been finished.

Preface to the second edition

Besides correcting some minor linguistic and print errors, I have in the second edition included a collection of different exercise problems which have been used in the training of students at NTNU. They illustrate how the models described in the previous chapters can be used
to solve practical problems of more interdisciplinary nature. Each of them contains a 'problem description' and some background information on materials and welding conditions. The
exercises are designed to illuminate the microstructural connections throughout the weld
thermal cycle and show how the properties achieved depend on the operating conditions applied. Solutions to the problems are also presented. These are not complete or exhaustive, but
are just meant as an aid to the reader to develop the ideas further.
Trondheim, 28 October, 1996
0ystein Grong

Preface to the first edition

The purpose of this textbook is to present a broad overview on the fundamentals of welding
metallurgy to graduate students, investigators and engineers who already have a good background in physical metallurgy and materials science. However, in contrast to previous textbooks covering the same field, the present book takes a more direct theoretical approach to
welding metallurgy based on a synthesis of knowledge from diverse disciplines. The motivation for this work has largely been provided by the need for improved physical models for
process optimalisation and microstructure control in the light of the recent advances that have
taken place within the field of materials processing and alloy design.
The present textbook describes a novel approach to the modelling of dynamic processes in
welding metallurgy, not previously dealt with. In particular, attempts have been made to rationalise chemical, structural and mechanical changes in weldments in terms of models based
on well established concepts from ladle refining, casting, rolling and heat treatment of steels
and aluminium alloys. The judicious construction of the constitutive equations makes full use
of both dimensionless parameters and calibration techniques to eliminate poorly known kinetic constants. Many of the models presented are thus generic in the sense that they can be
generalised to a wide range of materials and processing. To help the reader understand and
apply the subjects and models treated, numerous example problems, exercise problems and
case studies have been worked out and integrated in the text. These are meant to illustrate the
basic physical principles that underline the experimental observations and to provide a way of
developing the ideas further.
Over the years, I have benefited from interaction and collaboration with numerous people
within the scientific community. In particular, I would like to acknowledge the contribution
from my father Professor Tor Grong who is partly responsible for my professional upbringing
and development as a metallurgist through his positive influence on and interest in my research work. Secondly, I am very grateful to the late Professor Nils Christensen who first
introduced me to the fascinating field of welding metallurgy and later taught me the basic
principles of scientific work and reasoning. I will also take this opportunity to thank all my
friends and colleagues at the Norwegian Institute of Technology (Norway), The Colorado School
of Mines (USA), the University of Cambridge (England), and the Universitat der Bundeswehr
Hamburg (Germany) whom I have worked with over the past decade. Of this group of people,
I would particularly like to mention two names, i.e. our department secretary Mrs. Reidun
0stbye who has helped me to convert my original manuscript into a readable text and Mr.
Roald Skjaerv0 who is responsible for all line-drawings in this textbook. Their contributions
are gratefully acknowledged.
Trondheim, 1 December, 1993
0ystein Grong

Contents

Preface to the Second Edition ........................................................

xiii

Preface to the First Edition .............................................................

xiv

1. Heat Flow and Temperature Distribution in Welding ...........

1.1

Introduction ...............................................................................

1.2

Non-steady Heat Conduction ....................................................

1.3

Thermal Properties of Some Metals and Alloys ........................

1.4

Instantaneous Heat Sources .....................................................

1.5

Local Fusion in Arc Strikes ........................................................

1.6

Spot Welding .............................................................................

10

1.7

Thermit Welding ........................................................................

14

1.8

Friction Welding ........................................................................

18

1.9

Moving Heat Sources and Pseudo-steady State ......................

24

1.10 Arc Welding ...............................................................................

24

1.10.1 Arc Efficiency Factors ..................................................

26

1.10.2 Thick Plate Solutions ...................................................


1.10.2.1 Transient Heating Period .............................
1.10.2.2 Pseudo-steady State Temperature
Distribution ...................................................
1.10.2.3 Simplified Solution for a Fast-moving High
Power Source ..............................................

26
28

1.10.3 Thin Plate Solutions .....................................................


1.10.3.1 Transient Heating Period .............................
1.10.3.2 Pseudo-steady State Temperature
Distribution ...................................................
This page has been reformatted by Knovel to provide easier navigation.

31
41
45
48
49

vi

Contents

vii

1.10.3.3 Simplified Solution for a Fast Moving High


Power Source ..............................................

56

1.10.4 Medium Thick Plate Solution .......................................


1.10.4.1 Dimensionless Maps for Heat Flow
Analyses ......................................................
1.10.4.2 Experimental Verification of the Medium
Thick Plate Solution .....................................
1.10.4.3 Practical Implications ...................................

59

1.10.5 Distributed Heat Sources .............................................


1.10.5.1 General Solution ..........................................
1.10.5.2 Simplified Solution .......................................

77
77
80

1.10.6 Thermal Conditions during Interrupted Welding ..........

91

1.10.7 Thermal Conditions during Root Pass Welding ...........

95

61
72
75

1.10.8 Semi-empirical Methods for Assessment of Bead


Morphology ..................................................................
1.10.8.1 Amounts of Deposit and Fused Parent
Metal ............................................................
1.10.8.2 Bead Penetration .........................................

96
99

1.10.9 Local Preheating ..........................................................

100

References .........................................................................................

103

Appendix 1.1: Nomenclature ............................................................

105

Appendix 1.2: Refined Heat Flow Model for Spot Welding ..............

110

Appendix 1.3: The Gaussian Error Function ....................................

111

Appendix 1.4: Gaussian Heat Distribution .......................................

112

96

2. Chemical Reactions in Arc Welding ...................................... 116


2.1

Introduction ...............................................................................

116

2.2

Overall Reaction Model .............................................................

116

2.3

Dissociation of Gases in the Arc Column ..................................

117

2.4

Kinetics of Gas Absorption ........................................................

120

2.4.1

Thin Film Model ...........................................................

120

2.4.2

Rate of Element Absorption .........................................

121

This page has been reformatted by Knovel to provide easier navigation.

viii

Contents
2.5

The Concept of Pseudo-equilibrium ..........................................

122

2.6

Kinetics of Gas Desorption ........................................................

123

2.6.1

Rate of Element Desorption .........................................

123

2.6.2

Sieverts Law ...............................................................

124

Overall Kinetic Model for Mass Transfer during Cooling in


the Weld Pool ............................................................................

124

Absorption of Hydrogen ............................................................

128

2.8.1

Sources of Hydrogen ...................................................

128

2.8.2

Methods of Hydrogen Determination in Steel


Welds ...........................................................................

128

2.8.3

Reaction Model ............................................................

130

2.8.4

Comparison between Measured and Predicted


Hydrogen Contents ......................................................
2.8.4.1 Gas-shielded Welding ..................................
2.8.4.2 Covered Electrodes .....................................
2.8.4.3 Submerged Arc Welding ..............................
2.8.4.4 Implications of Sieverts Law .......................
2.8.4.5 Hydrogen in Multi-run Weldments ...............
2.8.4.6 Hydrogen in Non-ferrous Weldments ..........

131
131
134
138
140
140
141

Absorption of Nitrogen ..............................................................

141

2.9.1

Sources of Nitrogen .....................................................

142

2.9.2

Gas-shielded Welding ..................................................

142

2.9.3

Covered Electrodes .....................................................

143

2.9.4

Submerged Arc Welding ..............................................

146

2.10 Absorption of Oxygen ................................................................

148

2.10.1 Gas Metal Arc Welding ................................................


2.10.1.1 Sampling of Metal Concentrations at
Elevated Temperatures ...............................
2.10.1.2 Oxidation of Carbon .....................................
2.10.1.3 Oxidation of Silicon ......................................
2.10.1.4 Evaporation of Manganese ..........................
2.10.1.5 Transient Concentrations of Oxygen ...........

148

2.7
2.8

2.9

This page has been reformatted by Knovel to provide easier navigation.

149
149
152
156
160

Contents

ix

2.10.1.6 Classification of Shielding Gases ................


2.10.1.7 Overall Oxygen Balance ..............................
2.10.1.8 Effects of Welding Parameters ....................

166
166
169

2.10.2 Submerged Arc Welding ..............................................


2.10.2.1 Flux Basicity Index .......................................
2.10.2.2 Transient Oxygen Concentrations ...............

170
171
172

2.10.3 Covered Electrodes .....................................................


2.10.3.1 Reaction Model ............................................
2.10.3.2 Absorption of Carbon and Oxygen ..............
2.10.3.3 Losses of Silicon and Manganese ...............
2.10.3.4 The Product [%C] [%O] ...............................

173
174
176
177
179

2.11 Weld Pool Deoxidation Reactions .............................................

180

2.11.1 Nucleation of Oxide Inclusions .....................................

182

2.11.2 Growth and Separation of Oxide Inclusions .................


2.11.2.1 Buoyancy (Stokes Flotation) ........................
2.11.2.2 Fluid Flow Pattern ........................................
2.11.2.3 Separation Model .........................................

184
185
186
188

2.11.3 Predictions of Retained Oxygen in the Weld Metal ......


2.11.3.1 Thermodynamic Model ................................
2.11.3.2 Implications of Model ...................................

190
190
192

2.12 Non-metallic Inclusions in Steel Weld Metals ...........................

192

2.12.1 Volume Fraction of Inclusions ......................................

193

2.12.2 Size Distribution of Inclusions ......................................


2.12.2.1 Effect of Heat Input ......................................
2.12.2.2 Coarsening Mechanism ...............................
2.12.2.3 Proposed Deoxidation Model .......................

195
196
196
201

2.12.3 Constituent Elements and Phases in Inclusions ..........


2.12.3.1 Aluminium, Silicon and Manganese
Contents ......................................................
2.12.3.2 Copper and Sulphur Contents .....................
2.12.3.3 Titanium and Nitrogen Contents ..................
2.12.3.4 Constituent Phases ......................................

202

This page has been reformatted by Knovel to provide easier navigation.

202
202
203
204

Contents
2.12.4 Prediction of Inclusion Composition .............................
2.12.4.1 C-Mn Steel Weld Metals ..............................
2.12.4.2 Low-alloy Steel Weld Metals ........................

204
204
206

References .........................................................................................

212

Appendix 2.1: Nomenclature ............................................................

215

Appendix 2.2: Derivation of Equation (2-60) ....................................

219

3. Solidification Behaviour of Fusion Welds ............................ 221


3.1

Introduction ...............................................................................

221

3.2

Structural Zones in Castings and Welds ...................................

221

3.3

Epitaxial Solidification ...............................................................

222

3.3.1

Energy Barrier to Nucleation ........................................

225

3.3.2

Implications of Epitaxial Solidification ..........................

226

Weld Pool Shape and Columnar Grain Structures ....................

228

3.4.1

Weld Pool Geometry ....................................................

228

3.4.2

Columnar Grain Morphology ........................................

229

3.4.3

Growth Rate of Columnar Grains .................................


3.4.3.1 Nominal Crystal Growth Rate ......................
3.4.3.2 Local Crystal Growth Rate ...........................

230
230
234

3.4.4

Reorientation of Columnar Grains ...............................


3.4.4.1 Bowing of Crystals .......................................
3.4.4.2 Renucleation of Crystals ..............................

239
240
242

Solidification Microstructures ....................................................

251

3.5.1

Substructure Characteristics ........................................

251

3.5.2

Stability of the Solidification Front ................................


3.5.2.1 Interface Stability Criterion ...........................
3.5.2.2 Factors Affecting the Interface Stability .......

254
254
256

3.5.3

Dendrite Morphology ...................................................


3.5.3.1 Dendrite Tip Radius .....................................
3.5.3.2 Primary Dendrite Arm Spacing ....................
3.5.3.3 Secondary Dendrite Arm Spacing ...............

260
260
261
264

3.4

3.5

This page has been reformatted by Knovel to provide easier navigation.

Contents
3.6

3.7

3.8

xi

Equiaxed Dendritic Growth .......................................................

268

3.6.1

Columnar to Equiaxed Transition .................................

268

3.6.2

Nucleation Mechanisms ...............................................

272

Solute Redistribution .................................................................

272

3.7.1

Microsegregation .........................................................

272

3.7.2

Macrosegregation ........................................................

278

3.7.3

Gas Porosity ................................................................


3.7.3.1 Nucleation of Gas Bubbles ..........................
3.7.3.2 Growth and Detachment of Gas Bubbles ....
3.7.3.3 Separation of Gas Bubbles ..........................

279
279
281
284

3.7.4

Removal of Microsegregations during Cooling ............


3.7.4.1 Diffusion Model ............................................
3.7.4.2 Application to Continuous Cooling ...............

286
286
286

Peritectic Solidification ..............................................................

290

3.8.1

Primary Precipitation of the p-phase ...........................

290

3.8.2

Transformation Behaviour of Low-alloy Steel Weld


Metals ..........................................................................
3.8.2.1 Primary Precipitation of Delta Ferrite ...........
3.8.2.2 Primary Precipitation of Austenite ................
3.8.2.3 Primary Precipitation of Both Delta
Ferrite and Austenite ...................................

290
290
292
292

References .........................................................................................

293

Appendix 3.1: Nomenclature ............................................................

296

4. Precipitate Stability in Welds ................................................. 301


4.1

Introduction ...............................................................................

301

4.2

The Solubility Product ...............................................................

301

4.2.1

Thermodynamic Background .......................................

301

4.2.2

Equilibrium Dissolution Temperature ...........................

303

4.2.3

Stable and Metastable Solvus Boundaries ..................


4.2.3.1 Equilibrium Precipitates ...............................
4.2.3.2 Metastable Precipitates ...............................

304
304
308

This page has been reformatted by Knovel to provide easier navigation.

xii

Contents
4.3

Particle Coarsening ...................................................................

314

4.3.1

Coarsening Kinetics .....................................................

314

4.3.2

Application to Continuous Heating and Cooling ...........


4.3.2.1 Kinetic Strength of Thermal Cycle ...............
4.3.2.2 Model Limitations .........................................

314
315
315

Particle Dissolution ....................................................................

316

4.4.1

Analytical Solutions ......................................................


4.4.1.1 The Invariant Size Approximation ................
4.4.1.2 Application to Continuous Heating and
Cooling ........................................................

316
319

Numerical Solution .......................................................


4.4.2.1 Two-dimensional Diffusion Model ................
4.4.2.2 Generic Model .............................................
4.4.2.3 Application to Continuous Heating and
Cooling ........................................................
4.4.2.4 Process Diagrams for Single Pass 6082T6 Butt Welds ..............................................

325
326
328

References .........................................................................................

334

Appendix 4.1: Nomenclature ............................................................

334

4.4

4.4.2

322

329
332

5. Grain Growth in Welds ........................................................... 337


5.1

Introduction ...............................................................................

337

5.2

Factors Affecting the Grain Boundary Mobility ..........................

337

5.2.1

Characterisation of Grain Structures ............................

337

5.2.2

Driving Pressure for Grain Growth ...............................

339

5.2.3

Drag from Impurity Elements in Solid Solution ............

340

5.2.4

Drag from a Random Particle Distribution ...................

341

5.2.5

Combined Effect of Impurities and Particles ................

342

Analytical Modelling of Normal Grain Growth ...........................

343

5.3.1

Limiting Grain Size .......................................................

343

5.3.2

Grain Boundary Mobility ...............................................

345

5.3

This page has been reformatted by Knovel to provide easier navigation.

Contents

xiii

Grain Growth Mechanisms ..........................................


5.3.3.1 Generic Grain Growth Model .......................
5.3.3.2 Grain Growth in the Absence of Pinning
Precipitates ..................................................
5.3.3.3 Grain Growth in the Presence of Stable
Precipitates ..................................................
5.3.3.4 Grain Growth in the Presence of Growing
Precipitates ..................................................
5.3.3.5 Grain Growth in the Presence of
Dissolving Precipitates .................................

345
345

Grain Growth Diagrams for Steel Welding ................................

360

5.4.1

Construction of Diagrams ............................................


5.4.1.1 Heat Flow Models ........................................
5.4.1.2 Grain Growth Model .....................................
5.4.1.3 Calibration Procedure ..................................
5.4.1.4 Axes and Features of Diagrams ..................

360
360
361
361
363

5.4.2

Case Studies ...............................................................


5.4.2.1 Titanium-microalloyed Steels .......................
5.4.2.2 Niobium-microalloyed Steels .......................
5.4.2.3 C-Mn Steel Weld Metals ..............................
5.4.2.4 Cr-Mo Low-alloy Steels ................................
5.4.2.5 Type 316 Austenitic Stainless Steels ...........

364
364
367
370
372
375

Computer Simulation of Grain Growth ......................................

380

5.3.3

5.4

5.5

5.5.1

347
348
351
356

Grain Growth in the Presence of a Temperature


Gradient .......................................................................

380

Free Surface Effects ....................................................

382

References .........................................................................................

382

Appendix 5.1: Nomenclature ............................................................

384

5.5.2

6. Solid State Transformations in Welds ................................... 387


6.1

Introduction ...............................................................................

387

6.2

Transformation Kinetics ............................................................

387

6.2.1

387

Driving Force for Transformation Reactions ................

This page has been reformatted by Knovel to provide easier navigation.

xiv

Contents

6.3

6.2.2

Heterogeneous Nucleation in Solids ............................


6.2.2.1 Rate of Heterogeneous Nucleation ..............
6.2.2.2 Determination of Ghet.* and Qd ...................
6.2.2.3 Mathematical Description of the C-curve .....

389
389
390
392

6.2.3

Growth of Precipitates ..................................................


6.2.3.1 Interface-controlled Growth .........................
6.2.3.2 Diffusion-controlled Growth .........................

396
396
397

6.2.4

Overall Transformation Kinetics ...................................


6.2.4.1 Constant Nucleation and Growth Rates ......
6.2.4.2 Site Saturation .............................................

400
400
402

6.2.5

Non-isothermal Transformations ..................................


6.2.5.1 The Principles of Additivity ...........................
6.2.5.2 Isokinetic Reactions .....................................
6.2.5.3 Additivity in Relation to the Avrami
Equation ......................................................
6.2.5.4 Non-additive Reactions ................................

402
403
404

High Strength Low-alloy Steels .................................................

406

6.3.1

Classification of Microstructures ..................................

406

6.3.2

Currently Used Nomenclature ......................................

406

6.3.3

Grain Boundary Ferrite ................................................


6.3.3.1 Crystallography of Grain Boundary
Ferrite ..........................................................
6.3.3.2 Nucleation of Grain Boundary Ferrite ..........
6.3.3.3 Growth of Grain Boundary Ferrite ................

408
408
408
422

6.3.4

Widmansttten Ferrite ..................................................

427

6.3.5

Acicular Ferrite in Steel Weld Deposits ........................


6.3.5.1 Crystallography of Acicular Ferrite ...............
6.3.5.2 Texture Components of Acicular Ferrite ......
6.3.5.3 Nature of Acicular Ferrite .............................
6.3.5.4 Nucleation and Growth of Acicular
Ferrite ..........................................................

428
428
429
430

Acicular Ferrite in Wrought Steels ...............................

444

6.3.6

This page has been reformatted by Knovel to provide easier navigation.

404
405

432

6.4

6.5

Contents

xv

6.3.7

Bainite ..........................................................................
6.3.7.1 Upper Bainite ...............................................
6.3.7.2 Lower Bainite ...............................................

444
444
447

6.3.8

Martensite ....................................................................
6.3.8.1 Lath Martensite ............................................
6.3.8.2 Plate (Twinned) Martensite ..........................

448
448
448

Austenitic Stainless Steels ........................................................

453

6.4.1

Kinetics of Chromium Carbide Formation ....................

456

6.4.2

Area of Weld Decay .....................................................

456

Al-Mg-Si Alloys ..........................................................................

458

6.5.1

459

6.5.2

Quench-sensitivity in Relation to Welding ....................


6.5.1.1 Conditions for (Mg2Si) Precipitation
during Cooling ..............................................
6.5.1.2 Strength Recovery during Natural
Ageing .........................................................

459
461

Subgrain Evolution during Continuous Drive


Friction Welding ...........................................................

464

References .........................................................................................

467

Appendix 6.1: Nomenclature ............................................................

471

Appendix 6.2: Additivity in Relation to the Avrami Equation ............

475

7. Properties of Weldments ........................................................ 477


7.1

Introduction ...............................................................................

477

7.2

Low-alloy Steel Weldments .......................................................

477

7.2.1

477
478

Weld Metal Mechanical Properties ..............................


7.2.1.1 Weld Metal Strength Level ...........................
7.2.1.2 Weld Metal Resistance to Ductile
Fracture .......................................................
7.2.1.3 Weld Metal Resistance to Cleavage
Fracture .......................................................
7.2.1.4 The Weld Metal Ductile to Brittle
Transition .....................................................

This page has been reformatted by Knovel to provide easier navigation.

480
485
486

xvi

Contents
7.2.1.5

Effects of Reheating on Weld Metal


Toughness ...................................................

491

7.2.2

HAZ Mechanical Properties .........................................


7.2.2.1 HAZ Hardness and Strength Level ..............
7.2.2.2 Tempering of the Heat Affected Zone ..........
7.2.2.3 HAZ Toughness ...........................................

494
495
500
502

7.2.3

Hydrogen Cracking ......................................................


7.2.3.1 Mechanisms of Hydrogen Cracking .............
7.2.3.2 Solubility of Hydrogen in Steel .....................
7.2.3.3 Diffusivity of Hydrogen in Steel ....................
7.2.3.4 Diffusion of Hydrogen in Welds ...................
7.2.3.5 Factors Affecting the HAZ Cracking
Resistance ...................................................

509
509
513
514
514

H2S Stress Corrosion Cracking ....................................


7.2.4.1 Threshold Stress for Cracking .....................
7.2.4.2 Prediction of HAZ Cracking Resistance .......

524
524
525

Stainless Steel Weldments .......................................................

527

7.3.1

HAZ Corrosion Resistance ..........................................

527

7.3.2

HAZ Strength Level .....................................................

529

7.3.3

HAZ Toughness ...........................................................

530

7.3.4

Solidification Cracking ..................................................

532

Aluminium Weldments ..............................................................

536

7.4.1

Solidification Cracking ..................................................

536

7.4.2

Hot Cracking ................................................................


7.4.2.1 Constitutional Liquation in Binary Al-Si
Alloys ...........................................................
7.4.2.2 Constitutional Liquation in Ternary Al-MgSi Alloys .......................................................
7.4.2.3 Factors Affecting the Hot Cracking
Susceptibility ................................................

540

7.2.4

7.3

7.4

7.4.3

HAZ Microstructure and Strength Evolution during


Fusion Welding ............................................................

This page has been reformatted by Knovel to provide easier navigation.

518

541
542
544
547

Contents
7.4.3.1
7.4.3.2
7.4.3.3
7.4.3.4
7.4.4

Effects of Reheating on Weld Properties .....


Strengthening Mechanisms in Al-Mg-Si
Alloys ...........................................................
Constitutive Equations .................................
Predictions of HAZ Hardness and
Strength Distribution ....................................

HAZ Microstructure and Strength Evolution during


Friction Welding ...........................................................
7.4.4.1 Heat Generation in Friction Welding ............
7.4.4.2 Response of Al-Mg-Si Alloys and Al-SiC
MMCs to Friction Welding ............................
7.4.4.3 Constitutive Equations .................................
7.4.4.4 Coupling of Models ......................................
7.4.4.5 Prediction of the HAZ Hardness
Distribution ...................................................

xvii
547
548
548
550
556
556
557
558
558
560

References .........................................................................................

564

Appendix 7.1: Nomenclature ............................................................

567

8. Exercise Problems with Solutions ......................................... 571


8.1

Introduction ...............................................................................

571

8.2

Exercise Problem I: Welding of Low Alloy Steels ......................

571

8.3

Exercise Problem II: Welding of Austenitic Stainless Steels .....

583

8.4

Exercise Problem III: Welding of Al-Mg-Si Alloys ......................

587

Index .............................................................................................. 595


Author Index ................................................................................. 602

This page has been reformatted by Knovel to provide easier navigation.

1
Heat Flow and Temperature Distribution in
Welding

1.1 Introduction
Welding metallurgy is concerned with the application of well-known metallurgical principles
for assessment of chemical and physical reactions occurring during welding. On purely practical grounds it is nevertheless convenient to consider welding metallurgy as a profession of its
own because of the characteristic non-isothermal nature of the process. In welding the reactions are forced to take place within seconds in a small volume of metal where the thermal
conditions are highly different from those prevailing in production, refining and fabrication of
metals and alloys. For example, steel welding is characterised by:
High peak temperatures, up to several thousand 0 C.
High temperature gradients, locally of the order of 103 0C mm"1.
Rapid temperature fluctuations, locally of the order of 103 0C s 1 .
It follows that a quantitative analysis of metallurgical reactions in welding requires detailed
information about the weld thermal history. From a practical point of view the analytical
approach to the solution of heat flow problems in welding is preferable, since this makes it
possible to derive relatively simple equations which provide the required background for an
understanding of the temperature-time pattern. However, because of the complexity of the
heat flow phenomena, it is always necessary to check the validity of such predictions against
more reliable data obtained from numerical calculations and in situ thermocouple measurements. Although the analytical models suffer from a number of simplifying assumptions, it is
obvious that these solutions in many cases are sufficiently accurate to provide at least a qualitative description of the weld thermal programme.
An important aspect of the present treatment is the use of different dimensionless groups
for a general outline of the temperature distribution in welding. Although this practice involves several problems, it is a convenient way to reduce the total number of variables to an
acceptable level and hence, condense general information about the weld thermal programme
into two-dimensional (2-D) maps or diagrams. Consequently, readers who are unfamiliar with
the concept should accept the challenge and try to overcome the barrier associated with the use
of such dimensionless groups in heat flow analyses.

1.2 Non-Steady Heat Conduction


The symbols and units used throughout this chapter are defined in Appendix 1.1.

Since heat losses from free surfaces by radiation and convection are usually negligible in
welding, the temperature distribution can generally be obtained from the fundamental differential equations for heat conduction in solids. For uniaxial heat conduction, the governing
equation can be written as:1

(i-D
where T is the temperature, t is the time, x is the heat flow direction, and a is the thermal
diffusivity. The thermal diffusivity is related to the thermal conductivity X and the volume
heat capacity pc through the following equation:

(1-2)
For biaxial and triaxial heat conduction we may write by analogy:1

d-3)
and

(1-4)
The above equations must clearly be satisfied by all solutions of heat conduction problems,
but for a given set of initial and boundary conditions there will be one and only one solution.

1.3 Thermal Properties of Some Metals and Alloys


A pre-condition for obtaining simple analytical solutions to the differential heat flow equations is that the thermal properties of the base material are constant and independent of temperature. For most metals and alloys this is a rather unrealistic assumption, since both X, a,
and pc may vary significantly with temperature as illustrated in Fig. 1.1. In addition, the
thermal properties are also dependent upon the chemical composition and the thermal history
of the base material (see Fig. 1.2), which further complicates the situation.
By neglecting such effects in the heat flow models, we impose several limitations on the
application of the analytical solutions. Nevertheless, experience has shown that these problems to some extent can be overcome by the choice of reasonable average values for X, a and
pc within a specific temperature range. Table 1.1 contains a summary of relevant thermal
properties for different metals and alloys, based on a critical review of literature data. It should
be noted that the thermal data in Table 1.1 do not include a correction for heat consumed in
melting of the parent materials. Although the latent heat of melting is temporarily removed
during fusion welding, experience has shown this effect can be accounted for by calibrating
the equations against a known isotherm (e.g. the fusion boundary). In practice, such corrections are done by adjusting the arc efficiency factor Tq until a good correlation is achieved
between theory and experiments.

Hx-H0 = PC(T-T0 ),J/mm3

Carbon steel

Temperature, 0C
Fig. 1.1. Enthalpy increment H7-H0
2-4.

referred to an initial temperature T0 = 200C. Data from Refs.

Table 1.1 Physical properties for some metals and alloys. Data from Refs 2 - 6 .

Material

(WrTIm-10C-1)

(mm2 s"1)

(Jmnr 3 0C"1)

(0C)

(J mnr 3 )

(J mnr 3 )

Carbon
Steels

0.040

0.005

1520

7.50

2.0

Low Alloy
Steels

0.025

0.005

1520

7.50

2.0

High Alloy
Steels

0.020

0.005

1500

7.40

2.0

Titanium
Alloys

0.030

10

0.003

1650

4.89

1.4

Aluminium
(> 99% Al)

0.230

85

0.0027

660

1.73

0.8

Al-Mg-Si
Alloys

0.167

62

0.0027

652

1.71

0.8

Al-Mg
Alloys

0.149

55

0.0027

650

1.70

0.8

Does not include the latent heat of melting (AH1n).

X9 W/mm 0C

(a)

Temperature, 0C

(b)

X, W/mm 0C

High alloy steel

Temperature, 0C
Fig. 1.2. Factors affecting the thermal conductivity X of steels; (a) Temperature level and chemical
composition, (b) Heat treatment procedure. Data from Refs. 2-4.

1.4 Instantaneous Heat Sources


The concept of instantaneous heat sources is widely used in the theory of heat conduction.1 It
is seen from Fig. 1.3 that these solutions are based on the assumption that the heat is released
instantaneously at time t - 0 in an infinite medium of initial temperature T0, either across a
plane (uniaxial conduction), along a line (biaxial conduction), or in a point (triaxial conduction). The material outside the heat source is assumed to extend to x = + for a plane source
in a long rod, to r = for a line source in a wide plate, or to R = for a point source in a heavy
slab. The initial and boundary conditions can be summarised as follows:

T-T0 = oo for t = O and x = O (alternatively r = O or R = O)


T-J 0 = O for t = O and x * O (alternatively r > O or 7? > O)
7-T 0 = O for O < t < oo when x = oo (alternatively r = oo or R = oo).
It is easy to verify that the following solutions satisfy both the basic differential heat flow
equations (1-1), (1-3) and (1-4) and the initial and boundary conditions listed above:
(i)

Plane source in a long rod (Fig. 1.3a):


d-5)

where Q is the net heat input (energy) released at time t = O, and A is the cross section of the
rod.
(ii)

Line source in a wide plate (Fig. 1.3b):


(1-6)

where d is the plate thickness.


(iii)

Point source in a heavy slab (Fig. 1.3c):


(1-7)

Equations (1-5), (1-6) and (1-7) provide the required basis for a comprehensive theoretical
treatment of heat flow phenomena in welding. These solutions can either be applied directly
or be used in an integral or differential form. In the next sections a few examples will be given
to illustrate the direct application of the instantaneous heat source concept to problems related
to welding.

(a)

Fig. 1.3. Schematic representation of instantaneous heat source models; (a) Plane source in a long rod.

(b)

R
(C)

Fig. 1.3.Schematic representation of instantaneous heat source models (continued); (b) Line source in a
wide plate, (c) Point source in a heavy slab.

1.5 Local Fusion in Arc Strikes


The series of fused metal spots formed on arc ignition make a good case for application of
equation (1-7).
Model

The model considers a point source on a heavy slab as illustrated in Fig. 1.4. The heat is
assumed to be released instantaneously at time t = 0 on the surface of the slab. This causes a
temperature rise in the material which is exactly twice as large as that calculated from equation
(1-7):
(1-8)
In order to obtain a general survey of the thermal programme, it is convenient to write
equation (1-8) in a dimensionless form. The following parameters are defined for this purpose:
Dimensionless temperature:
(1-9)
where Tc is the chosen reference temperature.
Dimensionless time:
d-10)
where tt is the arc ignition time.
Dimensionless operating parameter:

(1-11)
where qo is the net arc power (equal to Qlt(), and (Hc-Ho) is the heat content per unit volume at
the reference temperature.
Dimensionless radius vector:
(1-12)
By substituting these parameters into equation (1-8), we obtain:

(1-13)

Heat source

Isotherms

3-D heat flow


Fig. 1.4. Instantaneous point source model for assessment of temperatures in arc strikes.

0Zn1

e/n

Linear time scale

T1

^i
Fig. 1.5. Calculated temperatures in arc strikes.
Equation (1-13) has been solved numerically for different values ofCT1and T1. The results
are presented graphically in Fig. 1.5. Due to the inherent assumption of instantaneous release
of heat in a point, it is not possible to use equation (1-13) down to very small values OfCT1 and
T1. However, at some distance from the heat source and after a time not much shorter than the
real (assumed) time of heating, the calculated temperature-time pattern will be reasonably
correct. Note that the heavy broken line in Fig. 1.5 represents the locus of the peak temperatures. This locus is obtained by setting 3In(OAi1VdT1 = 0:

from which

Substituting this into equation (1-13) gives:

(1-14)
where Qp is the peak temperature, and e is the natural logarithm base number.
Example (1.1)

Consider a small weld crater formed in an arc strike on a thick plate of low alloy steel. Calculate the cooling time from 800 to 5000C (Af875), and the total width of the fully transformed
region adjacent to the fusion boundary. The operational conditions are as follows:

where r| is the arc efficiency factor. Relevant thermal data for low alloy steel are given in
Table 1.1.
Solution

In the present case it is convenient to use the melting point of the steel as a reference temperature (i.e. 0 = 0m = 1 when Tc = TJ. The corresponding values OfZi1 and 9 (at 800 and 5000C,
respectively) are:

Cooling time At8/5

Since the cooling curves in Fig. 1.5 are virtually parallel at temperatures below 800 0 C, Af875
will be independent of Cr1 and similar to that calculated for the centre-line ((J1 = 0). By rearranging equation (1-13) we get:

and

Total width offully transformed region


Zone widths can generally be calculated from equation (1-14), as illustrated in Fig. 1.6. Taking the Ac3-temperature equal to 8900C for this particular steel, we obtain:

and

Alternatively, the same information could have been read from Fig. 1.5. Although it is
difficult to check the accuracy of these predictions, the calculated values for Ats/5 and ARlm are
considered reasonably correct. Thus, because the cooling rate is very large, in arc strikes a
hard martensitic microstructure would be expected to form within the transformed parts of the
HAZ, in agreement with general experience.

1.6 Spot Welding


Equation (1-6) can be used for an assessment of the temperature-time pattern in spot welding
of plates.
Model

The model considers a line source which penetrates two overlapping plates of similar thermal
properties, as illustrated in Fig. 1.7. The heat is assumed to be released instantaneously at time
Heat source

Fig. 1.6. Definition of isothermal zone width in Example (1.1).

Electrode

Heat source

Fig. 1.7. Idealised heat flow model for spot welding of plates.

t = 0. If transfer of heat into the electrodes is neglected, the temperature distribution is given
by equation (1-6).
This equation can be written in a dimensionless form by introducing the following group of
parameters:
Dimensionless time:
(1-15)
where th is the heating time (i.e. the duration of the pulse).
Dimensionless operating parameter:
(1-16)
where dt is the total thickness of the joint.
Dimensionless radius vector:
(1-17)
By substituting these parameters into equation (1-6), we get:
(1-18)
where 6 denotes the dimensionless temperature (previously defined in equation (1-9)).

6/n2

e/n2

Linear time scale

T2
Fig. 1.8. Calculated temperature-time pattern in spot welding.
Figure 1.8 shows a graphical representation of equation (1-18) for a limited range of a 2 and
T2. A closer inspection of the graph reveals that the temperature-time pattern in spot welding
is similar to that observed during arc ignition (see Fig. 1.5). The locus of the peak temperatures in Fig. 1.8 is obtained by setting d\n{^ln7}ldx2 - 0.

which gives
and
(1-19)

Example (1.2)

Consider spot welding of 2 mm plates of low alloy steel under the following operational conditions:

Calculate the cooling time from 800 to 5000C (Af8/5) in the centre of the weld, and the cooling
rate (CR.) at the onset of the austenite to ferrite transformation. Assume in these calculations
that the total voltage drop between the electrodes is 1.6 V. The M^-temperature of the steel is
taken equal to 475C.
Solution

If we use the melting point of the steel as a reference temperature, the parameters n2 and 6 (at
800 and 5000C, respectively) become:

Cooling time Atg/5

The parameter A%5 can be calculated from equation (1-18). For the weld centre-line (CT2 = 0),
we get:

and

Cooling rate at 475 0C

The cooling rate at a specific temperature is obtained by differentiation of equation (1-18) with
respect to time. When (J2 = 0 the cooling rate at 9 = 0.3 (475C) becomes:

and

Since the cooling curves in Fig. 1.8 are virtually parallel at temperatures below 8000C (i.e.
for QZn2 < 0.15), the computed values of Ar8/5 and CR. are also valid for positions outside the
weld centre-line. In the present example the centre-line solutions can be applied down to
("2m)2 ~ 2. According to equation (1-19), this corresponds to a lower peak temperature of:

which is equivalent with:

It should be emphasised that the present heat flow model represents a crude oversimplification of the spot welding process. In a real welding situation, most of the heat is generated at the
interface between the two plates because of the large contact resistance. This gives rise to the
development of an elliptical weld nugget inside the joint as shown in Fig. 1.9. Moreover, since
the model neglects transfer of heat into the electrodes, the mode of heat flow will be mixed and
not truly two-dimensional as assumed above. Consequently, equation (1-18) cannot be applied for reliable predictions of isothermal contours and zone widths. Nevertheless, the model
may provide useful information about the cooling conditions during spot welding if the efficiency factor if] and the voltage drop between the electrodes can be estimated with a reasonable
degree of accuracy.
A more refined heat flow model for spot welding is presented in Appendix 1.2.

1.7 Thermit Welding


Thermit welding is a process that uses heat from exothermic chemical reactions to produce
coalescence between metals and alloys. The thermit mixture consists of two components, i.e.
a metal oxide and a strong reducing agent. The excess heat of formation of the reaction product provides the energy source required to form the weld.
Model

In thermit welding the time interval between the ignition of the powder mixture and the completion of the reduction process will be short because of the high reaction rates involved.
Assume that a groove of width 2L1 is filled instantaneously at time t = 0 by liquid metal of an
initial temperature Tt (see Fig. 1.10). The metal temperature outside the fusion zone is T0. If
heat losses to the surroundings are neglected, the problem can be treated as uniaxial conduction where the heat source (extending from -L 1 to +L1) is represented by a series of elementary
sources, each with a heat content of:
(1-20)
At time t this source produces a small rise of temperature at position JC, given by equation (1 -5):

(1-21)

The final temperature distribution is obtained by substituting u = (x-xy(4at)m (i.e. dx'- du(4at)m) into equation (1-21) and integrating between the limits JC'= -L 1 and x'- +L1. This
gives (after some manipulation):
(1-22)

Isl'srau*'*'=]

Fusion

zone

Fig. 1.9. Calculated peak temperature contours in spot welding of steel plates (numerical solution). Operational conditions: / = 23kA, 64 cycles. Data from Bently et al1

Fusion

zone

Fig. 1.10. Idealised heat flow model for thermit welding of rails.
where erf(u) is the Gaussian error function. The error function is defined in Appendix 1.3*.
Because of the complex nature of equation (1-22), it is convenient to present the different
solutions in a dimensionless form by introducing the following groups of parameters:
*The error function is available in tables. However, in numerical calculations it is more convenient to use the
Fortran subroutine given in Appendix 1.3.

Dimensionless temperature:
(1-23)
Dimensionless time:
(1-24)
Dimensionless jc-coordinate:
(1-25)
Substituting these parameters into equation (1-22) gives:
(1-26)
Equation (1-26) has been solved numerically for different values of Q and T3. The results
are presented graphically in Fig. 1.11. As would be expected, the fusion zone itself (Q < 1)
cools in a monotonic manner, while the temperature in positions outside the fusion boundary
(Q > 1) will pass through a maximum before cooling. The locus of the HAZ peak temperatures
in Fig. 1.11 is defined by 3673T3 = 0. Referring to Appendix 1.3, we may write:

which gives
(1-27)
The peak temperature distribution is obtained by solving equation (1-27) for different combinations of Qm and T3m and inserting the roots into equation (1-26).

Example (1.3)

Consider thermit welding of steel rails (i.e. reduction of Fe2O3 with Al powder) under the
following operational conditions:

Calculate the cooling time from 800 to 5000C in the centre of the weld, and the total width
of the fully transformed region adjacent to the fusion boundary. The Ac3-temperature of the
steel is taken equal to 8900C.

91

Definition of parameters:

Fig. 1.11. Calculated temperature-time pattern in thermit welding.


Solution

For positions along the weld centre-line (Q. = 0) equation (1-26) reduces to:

Cooling time At 8/5

From the above relation it is possible to calculate the cooling time from Tt = 22000C to 800 and
5000C, respectively:

and

By rearranging equation (1-24), we obtain the following expression for Ar875:

The computed value for A/8/5 is also valid for positions outside the weld centre-line, since
the cooling curves at such low temperatures are reasonably parallel within the fusion zone.
Total width of fully transformed region
The fusion boundary is defined by:

The locus of the 8900C isotherm in temperature-time space can be read from Fig. 1.11.
Taking the ordinate equal to 0.40, we get:

By inserting this value into equation (1-27), we obtain the corresponding coordinate of the
isotherm:

The total width of the fully transformed HAZ is thus:

Unfortunately, measurements are not available to check the accuracy of these predictions.
Systematic errors would be expected, however, because of the assumption of instantaneous
release of heat immediately after powder ignition and the neglect of heat losses to the surroundings. Nevertheless, the present example is a good illustration of the versatility of the
concept of instantaneous heat sources, since these solutions can easily be added in space as
shown here or in time for continuous heat sources (to be discussed below).

1.8 Friction Welding


Friction welding is a solid state joining process that produces a weld under the compressive
force contact of one rotating and one stationary workpiece. The heat is generated at the weld
interface because of the continuous rubbing of the contact surfaces, which, in turn, causes a
temperature rise and subsequent softening of the material. Eventually, the material at the
interface starts to flow plastically and forms an up-set collar. When a certain amount of upsetting has occurred, the rotation is stopped and the compressive force is maintained or slightly
increased to consolidate the weld.
Model (after Rykalin et al.5j

The model considers a continuous (plane) heat source in a long rod as shown in Fig. 1.12(a).
The heat is liberated at a constant rate q'o in the plane x = 0 starting at time / = 0. If we
subdivide the time t during which the source operates into a series of infinitesimal elements dt/
(Fig. 1.12b), each element will have a heat content of:
(1-28)

(a)
Continuous heat source

(b) q

t
Fig. 1.12. Idealised heatflowmodel for friction welding of rods; (a) Sketch of model, (b) Subdivision of
time into a series of infinitesimal elements dt'.
At time / this heat will cause a small rise of temperature in the material, in correspondance
with equation (1-5):
(1-29)

If we substitute t"=t-1'into equation (1 -29), the total temperature rise at time t is obtained
by integrating from t"= t (t'= 0) to /"= 0 (t'= t):

(1-30)
In order to evaluate this integral, we will make use of the following mathematical transformation:

where

and

Hence, we may write:

The latter integral can be expressed in terms of the complementary error function* erfc{u)
by substituting:

and integrating between the limits u = x I (4at)l/2 and w = <*>.


This gives (after some manipulation):

(1-31)
If the temperature of the contact section at the end of the heating period is taken equal to Th,
equation (1-31) can be rewritten as:
(1-32)

where t'h denotes the duration of the heating period (t < t'h). Measured contact section temperatures for different metal/alloy combinations are given in Table 1.2.
Equation (1-32) may be presented in a dimensionless form by the use of the following
groups of parameters:
Dimensionless temperature:
(1-33)
Dimensionless time:
(1-34)
The complementary error function is defined in Appendix 1.3.

Table 1.2 Measured contact section temperatures during friction welding of some metals and alloys.
Data from Tensi et al.10
Metal/Alloy
Combination

Measuring
Method

Temperature Level
[0C]

Partial
Melting

Steel

Thermocouples

1080-1340

No

1260-1400

No/Yes

1080

No

Direct readings

548

Yes

Copper-Nickel

Direct readings

1083

Yes

Al-Cu-2Mg

Thermocouples

506

Yes

Al-4.3Cu

Thermocouples

562

Yes

Al-12Si

Thermocouples

575

Yes

Al-5Mg

Thermocouples

582

Yes

Steel-Nickel

Direct readings

Steel-Titanium

Direct readings1

Copper-Al

Based on direct readings of the voltage drop between the two work-pieces.

Dimensionless .^-coordinate:
(1-35)

By substituting these parameters into equation (1-32), we obtain:


(1-36)

Equation (1-36) describes the temperature in different positions from the weld contact section during the heating period. However, when the rotation stops, the weld will be subjected to
free cooling, since there is no generation of heat at the interface. As shown in Fig. 1.13(a) this
can be accounted for by introducing an imaginary heat source of power +qo at time t = t'h
which acts simultaneously with an imaginary heat sink of negative power -q o. It follows from
the principles of superposition (see Fig. 1.13b) that the temperature during the cooling period
is given by:9
(1-37)
where 6"(x4) and 6"(T 4 - 1) are the temperatures calculated for the heat source and the heat
sink, respectively, using equation (1-36).
Equations (1-36) and (1-37) have been solved numerically for different values of Q'and T4.
The results are presented graphically in Fig. 1.14. Considering the contact section (Q'= 0), the
temperature increases monotonically with time during the heating period, in correspondance
with the relationship:
(1-38)

q
(a)
Imaginary heat source
Real heat
source
t
Imaginary heat sink

e"
(b)

Heating
period

$ffl9
\

Fig. 1.13. Method for calculation of transient temperatures during friction welding; (a) Sketch of
imaginary heat source/heat sink model, (b) Principles of superposition.

Similarly, for the cooling period we get:


(1-39)
Outside the contact section (Q / > 0), the temperature rise will be smaller and the cooling
rate lower than that calculated from equations (1-38) and (1-39).

Heating

e"

Cooling

\
Fig. 1.14. Calculated temperature-time pattern in friction welding.

Example (1.4)

Consider friction welding of 026mm aluminium rods (Al-Cu-2Mg) under the following conditions:

Calculate the peak temperature distribution across the joint. Assume in these calculations
that the thermal diffusivity of the Al-Cu-2Mg alloy is 70mm2 s"1.
Solution
Readings from Fig. 1.14 give:

Next Page

In this particular case, it is possible to check the accuracy of the calculations against in situ
thermocouple measurements carried out on friction welded components made under similar
conditions. A comparison with the data in Fig. 1.15 shows that the model is quite successful in
predicting the HAZ peak temperature distribution. In contrast, the weld heating and cooling
cycles cannot be reproduced with the same degree of precision. This has to do with the fact
that the present analytical solution omits a consideration of the plastic straining occurring
during friction welding, which displaces the coordinates and alters the heat balance for the
system.

1.9 Moving Heat Sources and Pseudo-Steady State


In most fusion welding processes the heat source does not remain stationary. In the following
we shall assume that the source moves at a constant speed along a straight line, and that the net
power supply from the source is constant. Experience shows that such conditions lead to a
fused zone of constant width. This is easily verified by moving a tungsten arc across a sheet of
steel or aluminium, or by moving a soldering iron across a piece of lead or tin. Moreover,
zones of temperatures below the melting point also remain at constant width, as indicated by
the pattern of temper colours developed on welding ground or polished sheet.
It follows from the definition of pseudo-steady state that the temperature will not vary with
time when observed from a point located in the heat source. Under such conditions the temperature field around the source can be described as a temperature 'mountain' moving in the
direction of welding (e.g. see Fig. 15 in Ref. 11). For points along the weld centre-line, the
temperature at different positions away from the heat source (which for a constant welding
speed becomes a time axis) may be presented in a two-dimensional plot as indicated in Fig.
1.16. Specifically, this figure shows a schematic representation of the temperature in steel
welding from the base plate ahead of the arc to well into the solidified weld metal trailing the
arc. If we consider a fixed point on the weld centre-line, the temperature will increase very
rapidly during the initial period, reaching a maximum of about 2000-22000C for positions
immediately beneath the root of the arc.11 When the arc has passed, the temperature will start
to fall, and eventually (after long times) approach that of the base plate. In contrast, an observer moving along with the heat source will always see the same temperature landscape,
since this will not change with time according to the presuppositions.
It will be shown below that the assumption of pseudo-steady state largely simplifies the
mathematical treatment of heat flow during fusion welding, although it imposes certain restrictions on the options of the models.

1.10 Arc Welding


Arc welding is a collective term which includes the following processes*:
- Shielded metal arc (SMA) welding.
- Gas tungsten arc (GTA) welding.
- Gas metal arc (GMA) welding.
*The terminology used here is in accordance with the American Welding Society's recommendations. 12

Previous Page

In this particular case, it is possible to check the accuracy of the calculations against in situ
thermocouple measurements carried out on friction welded components made under similar
conditions. A comparison with the data in Fig. 1.15 shows that the model is quite successful in
predicting the HAZ peak temperature distribution. In contrast, the weld heating and cooling
cycles cannot be reproduced with the same degree of precision. This has to do with the fact
that the present analytical solution omits a consideration of the plastic straining occurring
during friction welding, which displaces the coordinates and alters the heat balance for the
system.

1.9 Moving Heat Sources and Pseudo-Steady State


In most fusion welding processes the heat source does not remain stationary. In the following
we shall assume that the source moves at a constant speed along a straight line, and that the net
power supply from the source is constant. Experience shows that such conditions lead to a
fused zone of constant width. This is easily verified by moving a tungsten arc across a sheet of
steel or aluminium, or by moving a soldering iron across a piece of lead or tin. Moreover,
zones of temperatures below the melting point also remain at constant width, as indicated by
the pattern of temper colours developed on welding ground or polished sheet.
It follows from the definition of pseudo-steady state that the temperature will not vary with
time when observed from a point located in the heat source. Under such conditions the temperature field around the source can be described as a temperature 'mountain' moving in the
direction of welding (e.g. see Fig. 15 in Ref. 11). For points along the weld centre-line, the
temperature at different positions away from the heat source (which for a constant welding
speed becomes a time axis) may be presented in a two-dimensional plot as indicated in Fig.
1.16. Specifically, this figure shows a schematic representation of the temperature in steel
welding from the base plate ahead of the arc to well into the solidified weld metal trailing the
arc. If we consider a fixed point on the weld centre-line, the temperature will increase very
rapidly during the initial period, reaching a maximum of about 2000-22000C for positions
immediately beneath the root of the arc.11 When the arc has passed, the temperature will start
to fall, and eventually (after long times) approach that of the base plate. In contrast, an observer moving along with the heat source will always see the same temperature landscape,
since this will not change with time according to the presuppositions.
It will be shown below that the assumption of pseudo-steady state largely simplifies the
mathematical treatment of heat flow during fusion welding, although it imposes certain restrictions on the options of the models.

1.10 Arc Welding


Arc welding is a collective term which includes the following processes*:
- Shielded metal arc (SMA) welding.
- Gas tungsten arc (GTA) welding.
- Gas metal arc (GMA) welding.
*The terminology used here is in accordance with the American Welding Society's recommendations. 12

- Flux cored arc (FCA) welding.


- Submerged arc (SA) welding.
The main purpose of this section is to review the classical models for the pseudo-steady
state temperature distribution around moving heat sources. The analytical solutions to the
differential heat flow equations under conditions applicable to arc welding were first presented
(a)
Cooling period

Temperature, 0C

Heating period

Predicted heating and cooling cycles


for the contact section (x=0)

Time, s

Peak temperature, 0C

(b)

Observed relationship

Predicted relationship

Distance from contact section, mm


Fig. 1.15. Comparison between measured and predicted temperatures in friction welding of Al-Cu-2Mg
alloys; (a) Temperature-time pattern, (b) Peak temperature distribution. Data from Tensi et al.10

by Rosenthal,1314 but the theory has later been extended and refined by a number of other
inve stigators .9'n*15"20
1.10.1 Arc efficiency factors
In arc welding heat losses by convection and radiation are taken into account by the efficiency
factor r\, defined as:
(1-40)
where qo is the net power received by the weldment (e.g. measured by calorimetry), / is the
welding current (amperage), and U is the arc voltage.
For submerged arc (SA) welding the efficiency factor (r\) has been reported in the range
from 90 to 98%, for SMA and GMA welding from 65 to 85%, and for GTA welding from 22
to 75%, depending on polarity and materials.11
A summary of ranges is given in Table 1.3.
7.70.2 Thick plate solutions
Model (after Rykalin9)
According to Fig. 1.17, the general thick plate model consists of an isotropic semi-infinite
body at initial temperature T0 limited in one direction by a plane that is impermeable to heat.
At time t = 0 a. point source of constant power qo starts on the surface at position O moving in
the positive x-direction at a constant speed U The rise of temperature T- T0 in point P at time
t is sought.
During a very short time interval from ^'to t'+ dt'the amount of heat released at the surface
is dQ = qodt'. According to equation (1-7) this will produce an infinitesimal rise of temperature in P at time t:

(1-41)

where

is the time available for conduction of heat over the distance

to point P.
For a convenient presentation of the pseudo-steady state solution, the position P should be
referred to that of the moving heat source. This is achieved by changing the coordinate system
from O to O'(see Fig. 1.17):

and

Hence, we may write:

Solidified
weld metal

Root of arc

Weld pool

X=Vt
Relative position along weld centre-line
Fig. 1.16. Schematic diagram showing weld centre-line temperature in different positions from the heat
source during steel welding at pseudo-steady state.

Table 1.3 Recommended arc efficiency factors for different welding processes. Data from Refs 11,21.
Arc efficiency factor j]
Welding Process

Range

Mean

SA welding (steel)

0.91-O.99

0.95

SMA welding (steel)

0.66-0.85

0.80

GMA welding (CO2-steel)

0.75-0.93

0.85

GMA welding (Ar-steel)

0.66-0.70

0.70

GTA welding (Ar-steel)

0.25-0.75

0.40

GTA welding (He-Al)

0.55-0.80

0.60

GTA welding (Ar-Al)

0.22-0.46

0.40

(1-42)

where
The total rise of temperature at P is obtained by substituting:

and

into equation (1-42), and integrating between the limits u = (R2l4af)m and u = o.This gives
(after some manipulation):
(1-43)
It is well-known that:

Hence, the general thick plate solution can be written as:

(1-44)

If u is sufficiently small (i.e. when welding has been performed over a sufficient period), we
obtain the pseudo-steady state temperature distribution:

(1-45)
This equation is often referred to as the Rosenthal thick plate solution,1314 in honour of
D. Rosenthal who first derived the relation by solving the differential heat flow equation directly for the appropriate boundary conditions.
1.10.2.1 Transient heating period
It follows from the above analysis that the pseudo-steady state temperature distribution is

Fig. 1.17. Moving point source on a semi-infinite slab.


attained after a transient heating period. The duration of this heating period is determined by
the integral in equation (1-44).
Taking the ratio between the real and the pseudo-steady state temperature equal to K1, we
have:
(1-46)
Equation (1-46) can be expressed in terms of the following parameters:
Dimensionless radius vector:
(1-47)
Dimensionless time:
(1-48)
Substituting these into equation (1-46) gives:
(1-49)

where

and

Equation (1-49) has been solved numerically for a limited range of cr3 and x. The results are
presented graphically in Fig. 1.18. A closer inspection of Fig. 1.18 reveals that the duration of
the transient heating period depends on the dimensionless radius vector a 3 . In practice this
means that the Rosenthal equation is not valid during the initial period of welding unless the
distance from the heat source to the observation point is very small. It should be noted, however, that a dimensionless distance o~3 may be 'short' for one combination of welding speed
and thermal diffusivity, while the same position may represent a 'long' distance for another
combination of V and a. Similarly, the dimensionless time T may be 'short' or 'long' at a
chosen number of seconds, depending on the ratio v/2a.
Example (1.5)

Consider stringer bead deposition on a thick plate of aluminium at a constant welding speed of
5 mm s"1. Calculate the duration of the heating period when the distance from the heat source
to the point of observation is 17 mm.
Solution

Ki = (T-T0)/(T-T0)p.s.

Taking a = 85 mm2 s"1, the dimensionless radius vector becomes:

T = v2t/2a
Fig. 1.18. Ratio between real and pseudo-steady state temperature in thick plate welding for
different combinations ofCT3and T.

It is seen from Fig. 1.18 that the pseudo-steady state temperature distribution is approached
when T ~ 3, which gives:

This corresponds to a total bead length of:

The above calculations show that the Rosenthal equation is not valid if the ratio between R
and L2 exceeds a certain critical value (typically 0.15 to 0.30 for aluminium welds and 0.4 to
0.6 for steel welds). This important point is often overlooked when discussing the relevance of
the thick plate solution in arc welding.
1.10.2.2 Pseudo-steady state temperature distribution
The Rosenthal equation gives, with the limitations inherent in the assumptions, full information on the thermal conditions for point sources on heavy slabs. Accordingly, in order to obtain
a general survey of the pseudo-steady state temperature distribution, it is convenient to present
the different solutions in a dimensionless form. The following parameters are defined for this
purpose:11
Dimensionless operating parameter:
(1-50)
Dimensionless jc-coordinate:
(1-51)
Dimensionless ^-coordinate:
(1-52)
Dimensionless z-coordinate:
(1-53)
By substituting these parameters into equation (1-45), we obtain:
(1-54)
where 6 and a 3 are the dimensionless temperature and radius vector, respectively (previously
defined in equations (1-9) and (1-47)).
Equation (1-54) has been solved numerically for chosen values of a 3 and . A graphical
presentation of the different solutions is shown in Fig. 1.19. These maps provide a good

e/n3

(a)

%
Fig, /./P.Dimensionless temperature maps for point sources on heavy slabs; (a) Vertical sections parallel to the ^-axis.
overall indication of the thermal conditions during thick plate welding, but are not suitable for
precise readings. Consequently, for quantitative analyses, the following set of equations can
be used:11
Isothermal zone widths
The maximum width of an isothermal enclosure is obtained by setting 3ln(0M3)/9a3 = 0:

From the definition of a 3 we have:

(b)

V
Fig. /.iP.Dimensionless temperature maps for point sources on heavy slabs (continued): (b) Isothermal
contours in the -\}/-plane for different ranges of 0/n3.
Partial differentiation of the Rosenthal equation gives:

and
(1-55)
Equation (1-55) can be used for calculations of isothermal zone widths V|/w and cross sectional areas A1. From Fig. 1.20 we have:
(1-56)
and
(1-57)
A graphical presentation of equations (1-55), (1-56), and (1-57) is shown in Fig. 1.21.
Length of isothermal enclosures
Referring to Fig. 1.20, the total length of an isothermal enclosure r is given by:
(1-58)
where ' and "are the distance from the heat source to the front and the rear of the enclosure,
respectively.

x,S
Heat source

y.v

z, C

Fig. 1.20.Three-dimensional graphical representation of Rosenthal thick plate solution (schematic).

Vm,O3m,Al

V9P

Fig. 1.21. Dimensionless distance a3m, half width \|/m and cross sectional area A1 vs n3 /Qp.
The coordinates ' and " are found by setting a 3 = ^ in equation (1-54). This gives:
(1-59)

and
(1-60)

Volume of isothermal enclosures


Since the assumption of a point heat source involves semi-circular isotherms in the \|/- plane,
the volume of an isothermal enclosure is obtained by integration over the total length from "
tor:
(1-61)
The former integral is readily evaluated by substituting:

which follows from a differentiation of equation (1-54). Hence, we may write:

(1-62)

Noting that <r3 = - "= n3 /0 at the lower limit of integration, we obtain:


(1-63)
The dimensionless volume T is related to the real volume of the enclosure V(mm3) through
the following equation:
(1-64)
Cooling conditions close to weld centre-line
For points located on the weld centre-line behind the heat source \|/ = = 0, and
Hence, equation (1-54) reduces to*:

d-65)
Equation (1-65) provides a basis for calculating the cooling time within a specific temperature interval (e.g. from O1 to B2):
(1-66)
For welding of low alloy steel the cooling time from 800 to 5000C is widely accepted as an
adequate index for the thermal conditions under which the austenite to ferrite transformation
takes place. From equation (1-66), we have:

(1-67)

(1-69)
Similarly, the cooling rate at a specific temperature is obtained by differentiating equation
(1-65) with respect to time:
(1-70)
By multiplying equation (1-70) with the appropriate conversion factor, we get:

(1-71)

Example (1.6)

Consider stringer bead deposition (GMAW) on a thick plate of low alloy steel under the following conditions:

Sketch the contours of the fusion boundary and the Ac3-isotherm (9100C) in the -\|/ (x-y)
plane at pseudo-steady state.
Solution

As shown in Fig. 1.22(a) it is sufficient to calculate the coordinates in four different (characteristic) positions to draw a sketch of the isothermal contours. If we neglect the latent heat of
melting, the d/n3 ratio at the melting point becomes, according to equations (1-9) and (1-50):

End-points
The end-points are readily obtained from equations (1-59) and (1-60):

and
Maximum width
The maximum width of an isothermal enclosure can generally be calculated from equations
(1-55) and (1-56), or read from Fig. 1.21. Taking O^ /n3 = QIn3 = 0.088, we obtain:

(a)

Pos. (3)
Pos. (2)
Pos. (4)

Pos.(1)
%

(b)

\
x(mm)

y(mm)

Fig. 1.22. Calculation of isothermal contours from Rosenthal thick plate solution; (a) Calculation procedure, (b) Sketch of fusion boundary and Acj-isotherm in position = 0 (Example 1.6).

and

Intersection point with y/ (y)-axis


In this case = = O, and cr3 = \|/. Hence, equation (1-54) reduces to:

which gives

Similarly, the contour of the Ac3-isotherm can be determined by inserting BZn3= 0.052 into
the same set of equations. Figure 1.22(b) shows a graphical representation of the computed
isothermal contours.
Example (1.7)

Consider GTA welding on a thick plate of low alloy steel under the following conditions:

Calculate the weld pool volume, the weld bead cross section, the width of the fully transformed HAZ, the cooling time from 800 to 5000C, and the cooling rate at the onset of the
austenite to ferrite transformation (e.g. at 6500C). The arc efficiency factor is taken equal to
0.5.
Solution

If we neglect the latent heat of melting, the GM3 ratio at the melting point becomes:

Weld pool volume


The coordinate ' may be calculated from equation (1-59):

We can now use equations (1-63) and (1-64) to calculate the weld pool volume. This gives:

and

Weld bead cross section


The weld bead cross section can be read from Fig. 1.21. Taking n3 /6 p = 1/0.445 = 2.196, we
obtain:

and

Width of fully transformed HAZ


When Tp = 9100C, the n3 /Qp ratio becomes:

From Fig. 1.21 we have:

which gives

Cooling time from 800 to 5000C


The cooling time, Af8/5, can be obtained from equation (1-68):

This value is also valid for positions outside the weld centre-line, since the cooling curves
at such low temperatures are reasonably parallel down to Tp ~ 9100C (see Fig. 1.19(a)).
Cooling rate at 6500C
The cooling rate at a specific temperature is given by equation (1-71). Taking the transformation start temperature equal to 6500C, we get:

1. 10.2.3 Simplified solution for a fast-moving high power source


Model (after Rykalin9)
It follows from Fig. 1.19(b) that the isotherms behind the heat source become increasingly
elongated as the arc power q0 and the welding speed v increase. In the limiting case, when
qo > oo? v oo and qo A) remains finite, the isotherms will degenerate into surfaces which are
parallel to the welding x direction as shown in Fig. 1.23(a). Conduction of heat will then occur
exclusively in directions normal to the jc-axis.
In a short time interval dt, the amount of heat released per unit length of the weld is equal to:

(1-72)
According to the assumptions this amount of heat will remain in a slice of thickness dx due
to the lack of a temperature gradient in the welding direction. Since symmetry demands that
the isotherms in the y-z plane are semi-circles*, the situation becomes identical to the temperature field around a linear instantaneous heat source in a thin plate, provided that the space
above the slab is replaced by solid material and the strength of the source is doubled (see Fig.
1.23(b)). The solution is then given by equation (1-6) if we replace QId by Iq0 Iv:
d-73)
where r* is the two-dimensional radius vector in the y-z plane.
Equation (1-73) represents the simplified solution for a fast moving high power source on a
semi-infinite slab, and is valid within a limited range of the more general Rosenthal equation
for three-dimensional heat flow (equation (1-45)). By introducing the dimensionless radius
vector Cr4 = Dr*12a, equation (1-73) transforms to:

(1-74)

A graphical presentation of equation (1-74) gives a family of curves which resembles the
thermal cycles shown in Fig. 1.19(a). Although the cooling conditions close to the weld centre-line are similar to those calculated from the Rosenthal equation, the predicted width/depth
of the isotherms will always be greater than that inferred from the general thick plate solution
as illustrated in Fig. 1.24 due to the assumption of 2-D heat flow. The parameter o~4m in Fig.
1.24 is obtained by differentiating equation (1-74) with respect to time:

*The isotherms must meet the plate surface at right angles in the absence of a temperature gradient across the boundary.

(a)

(b)

Symmetry
plane

Fig. 1.23. Fast moving high power source on a semi-infinite slab; (a) Sketch of model, (b) Analogy between a fast moving high power source and an instantaneous line source.

4m m

Asymptote

0 /n
P 3
Fig. 1.24. Theoretical width of isotherms under 2-D and 3-D heat flow conditions, respectively at pseudosteady state (thick plate welding).

which gives

Substituting this into equation (1-74) gives:

(1-75)
It is interesting to note that the dimensionless width a4m will approach \j/ m when the dp /n3
ratio becomes sufficiently small (i.e. less than about 0.1). Under such conditions the isotherms
will be strongly elongated in the welding direction (see Fig. 1.19(b)), which forces the heat to
flow primarily in directions normal to the x-axis.
A general graphical representation of the weld thermal programme can be obtained by
combining equations (1-74) and (1-75):

(1-76)
Equation (1-76) has been plotted in Fig. 1.25. This graph provides a basis for calculating
the retention time within specific temperature intervals under various welding conditions.

e/ep

W J
Fig. 1.25. Temperature-time pattern in thick plate welding at high arc power and high welding speed.
Example (1.8)

Consider SA welding on a thick plate of low alloy steel under the following conditions:

Calculate the retention time within the austenite regime (T > 9100C) for points located lmm
outside the fusion boundary.
Solution

If we use the melting point of the steel as a reference temperature, the parameter n3 becomes:

A comparison with Fig. 1.24 shows that the assumption of 2-D heat flow is justified when
dp < 1. Hence, the width of the fusion zone (8p = 1) can be calculated from equation (1-75):

which gives

The peak temperature lmm outside the fusion boundary is thus:

The total time spent in the thermal cycle from 0 = 0.59 (T= 9100C) to G^ = 0.78 and down
again to 0 = 0.59 can now be read from Fig. 1.25. Taking the ordinate 0/0^ equal to 0.76, we
obtain:

which gives

and

1.10.3 Thin plate solutions


Model (after Rykalin9)
As shown in Fig. 1.26, the general thin plate model considers a line source in a wide sheet of
thickness d and initial temperature T0. At time t = 0 the source starts to move at a constant
speed D in the positive x-direction. The rise of temperature T-T0 in point P at time t is sought.
According to equation (1-6) the elementary source dQ = qo df released at position Vt' will
cause a small rise of temperature dT in point P at time t:

(1-77)

where

is the time available for conduction of heat over the distance


to point P.

If we refer the position P to that of the heat source at time t, we shall expect a solution
independent of time at pseudo-steady state. This is achieved by changing the coordinate system from O to O' (see Fig. 1.26):

K0(U)1K1(U)

Fig. 1.26. Moving line source in a thin sheet.

U
Fig. 1.27. Graphical representation of the Bessel functions Ko(u) and Kx{u).

Hence, we may write:

d-78)
where

For integration of all contributions from

to

, we introduce:

from which

Substituting these parameters into equation (1-78) give:

d-79)
It is well-known that:

where K0(G5) is the modified Bessel function of the second kind and zero order. A graphical
representation of A^(W) is shown in Fig. 1.27.
Hence, the general thin plate solution can be written as:

(1-80)

When w is sufficiently large (i.e. when welding has been performed over a sufficient period), we obtain the pseudo-steady state temperature distribution:

(1-81)
Equation (1-81) is often referred to as the Rosenthal thin plate solution, in memory of its
originator D. Rosenthal.1314 It follows that this model is applicable to all types of welding
processes (including electron beam, plasma arc and laser welding), provided that a full throughthickness penetration is achieved in one pass.
1.10.3.1 Transient heating period
In thin plate welding the duration of the transient heating period is determined by the integral in equation (1-80). Taking the ratio between the real and pseudo-steady state temperature equal to K2, we obtain:

(1-82)

where w = x/2, and dw = dill.


Equation (1-82) has been solved numerically for a limited range of cr5 and x. The results are
presented graphically in Fig. 1.28.
Example (1.9)

Consider butt welding of a thin aluminium plate at a constant travel speed of 5mm s"1. Calculate the duration of the transient heating period when the distance from the heat source to the
point of observation is 17mm.
Solution

Taking a = 85 mm2 s"1, the dimensionless radius vector becomes:

It follows from Fig. 1.28 that the pseudo-steady state temperature distribution is approached
when T ~ 5, which gives:

and

This minimum bead length is nearly twice as large as that calculated for 3-D heat flow for
the same combination of welding speed, thermal diffusivity and radius vector (see Example

K2=(T-To)/(T-To)p.s.

t = v2t/2a
Fig. 1.28. Ratio between real and pseudo-steady state temperature in thin plate welding for different
combinations of a 5 and T.

(1.5)). Consequently, the duration of the transient heating period is significantly longer in thin
plate welding than in thick plate welding due the pertinent differences in the heat flow conditions.
1.10.3.2 Pseudo-steady state temperature distribution
A graphical presentation of the pseudo-steady state temperature distribution similar to that
shown in Fig. 1.19 for three-dimensional heat flow may be obtained by introducing the
dimensionless plate thickness 8 = vdlla, which is a measure of the relative speeds of the arc
and the heat flow in the material. By rearranging equation (1-81), we get:
(1-83)
Plots of this equation are shown in Fig. 1.29. It follows that the pseudo-steady state temperature distribution in thin plate welding depends on the parameter 68//?3. In practice, this
means that the shape of the isotherms is not influenced by the welding speed, since both 8 and
n3 are proportional to v.
Isothermal zone widths
The maximum width of an isothermal enclosure is obtained by setting 3(68 /n3) /die, = 0. Noting that (d/du)Ko(u) - -Kx(u), where K1 (u) is the modified Bessel function of the second kind
and first order (shown in Fig. 1.27), we get:

(1-84)

%
\

V
Fig. 1.29. Dimensionless temperature maps for line sources in thin plates for different ranges of 98/n3.

Heat source

Fig. 1.30. Graphical representation of Rosenthal thin plate solution (schematic)


This gives:
(1-85)
and
(1-86)

Equation (1-86) can be used for calculations of isothermal zone widths (v|/m) and cross
sectional areas (A2) in thin plate welding. Referring to Fig. 1.30, we have:

d-87)
and
(1-88)
Figure 1.31 shows a graphical presentation of equations (1-86), (1-87), and (1-88).
Length of isothermal enclosures
The distance from the heat source to the front (') and the rear (^') of an isothermal enclosure is obtained by substituting o~5 = into equation (1-83).
This gives:
(1-89)
and
(1-90)

^5m.Vm-A 2 /8

n3/6p5

s\-?.*t

Fig. l.Ji.Dimensionless distance a5m, half width \|/m and cross sectional area A2/S vs n3/Qph.

n 3 /98
Fig. 1.32. Dimensionless distance from heat source to front (') and rear (") of isothermal enclosure vs
3/68 (thin plate welding).

A graphical presentation of equations (1-89) and (1-90) is shown in Fig. 1.32. Included is
also a plot of the total length of the enclosure t vs the parameter n3 /08.
Cooling conditions close to weld centre-line
For points located on the weld centre-line behind the heat source a 5 = - = X. When x is larger
than about 1, (i.e. t > IaIv1), it is a fair approximation to set K0 (x) exp(-x)Vrc/2x (see Fig.
1.27). Hence, equation (1-83) reduces to:
(1-91)
Equation (1-91) provides a basis for calculating the cooling time within a specific temperature interval (e.g. from O1 to 02):

(1-92)

The dimensionless cooling time from 800 to 5000C is thus given by:

(1-93)
from which the real cooling time is obtained:
(1-94)
Taking as average values X = 0.025 W mm"1 0C"1, pc = 0.005 J mm"3 0C"1, and T0 = 200C
for welding of low alloy steels, we have:

d-95)
Similarly, the cooling rate at a specific temperature is obtained by differentiating equation
(1-91) with respect to time:
(1-96)

By multiplying equation (1-96) with the appropriate conversion factor, we get:

(1-97)

For welding of low alloy steels, the cooling rate becomes:

(1-98)

Example (LlO)

Consider GTA butt welding of a 2mm thick sheet of cold-rolled aluminium (Al-Mg alloy)
under the following conditions:

Sketch the contours of the fusion boundary and the Ar-isotherm in the -\|/ (x-y) plane at
pseudo-steady state. The recrystallisation temperature Ar of the base material is taken equal to
2750 C. Calculate also the cross sectional area of the fully recrystallised HAZ and the cooling
rate at 2750C for points located within this region.
Solution
Referring to Fig. 1.22(a) (Example (1.6)) it is sufficient to calculate the coordinates in four
different (characteristic) positions to sketch the contour of the fusion boundary. If we neglect
the latent heat of melting, the n3/68 ratio at the melting point becomes:

End-points
The end-points can be read from Fig. 1.32:

and

Maximum widths
The maximum width of an isothermal enclosure can generally be calculated from equations
(1-86) and (1-87) or read from Fig. 1.31. When n3/QpS = 0.84, we obtain:

and

Intersection point with y/(y)-axis


In this case = 0 and cr5 = \j/. Hence, equation (1-83) reduces to:

which gives

Similarly, the contour of the Ar-isotherm can be determined by inserting n3/db = 2.08 into
the same set of equations. Figure 1.33 shows a graphical representation of the calculated
isothermal contours.

x(mm)

y(mm)
Fig. 1.33. Calculated contours of fusion boundary and Ar-isotherm in GTA butt welding of a 2mm thick
aluminium plate (Example 1.10).

Cross sectional area of fully recrystallised HAZ


In general, cross sectional areas can be read from Fig. 1.31. Taking the n3/OpS ratio equal to
0.84 (Qp= 1) and 2.08 (8 p = 0.48), respectively, we have:

which gives

Cooling rate at 275 0C


The cooling rate at a specific temperature can be calculated from equation (1-97). In the present
case, we obtain:

1.10.3.3 Simplified solution for a fast moving high power source


Model (after Rykalin9)
It follows from Fig. 1.29 that the isotherms behind the heat source become increasingly elongated as the 08M3 ratio decreases. In the limiting case the isotherms will degenerate into
surfaces which are parallel to the welding x direction, as shown in Fig. 1.34.
In a short time interval dt the amount of heat released per unit length of the weld is equal to:

(1-99)
According to the assumptions this amount of heat will remain in a rod of constant cross
sectional area due to the lack of a temperature gradient in the welding direction. Under such
conditions the mode of heat flow becomes essentially one-dimensional, and the temperature
distribution is given by equation (1-5):
(1-100)
Equation (1-100) represents the simplified solution for a fast moving high power source* in
a thin sheet, and is valid within a limited range of the more general Rosenthal equation for twodimensional heat flow (equation (1-81)).
By substituting the appropriate dimensionless parameters into equation (1-100), we obtain:

* Since the shape of a given isotherm in the x-y plane is determined by the qjd ratio, the minimum welding speed
which is required to maintain 1-D heat flow increases with decreasing qjVd ratios. Hence, the term 'fast moving high
power source' is also appropriate in the case of the thin plate welding.

Fig. 1.34. Fast moving high power source in a thin plate.

(1-101)
The locus of the peak temperatures is readily evaluated from equation (1-101) by setting
3ln(e8M3)/3T = 0:

which gives

and
(1-102)
It is evident from the plot in Fig. 1.35 that the predicted width of the isotherms is always
greater than that inferred from the general thin plate solution (equation (1-83)) due to the
assumption of one-dimensional heat flow. However, such deviations become negligible at
very small Qpb/n3 ratios because of a small temperature gradient in the welding x direction
compared to the transverse y direction of the plate.
A general graphical representation of the weld thermal programme (similar to that shown in
Fig. 1.25 for a fast moving high power source on a heavy slab) can be obtained by combining
equations (1-101) and (1-102):

m(1-D)/m(a-D)

Asymptote

0 p 8/n 3

e/ep

Fig. 1.35. Theoretical width of isotherms under 1-D and 2-D heat flow conditions, respectively at pseudosteady state (thin plate welding).

2t/(vm)2
Fig. 1.36. Temperature-time pattern in thin plate welding at high arc power and high welding speed.

Next Page

(1-103)
Equation (1-103) has been plotted in Fig. 1.36.
Example (LU)
Consider butt welding of a 2mm thin plate of austenitic stainless steel with covered electrodes
(SMAW) under the following conditions:

Calculate the retention time within the critical temperature range for chromium carbide
precipitation (i.e. from 650 to 8500C) for points located at the 8500C isotherm.
Solution

If we use the melting point of the steel as a reference temperature, the parameter n3/5 becomes:

A comparison with Fig. 1.35 shows that the assumption of 1-D heat flow is justified when
Qp< 1. Hence, the total time spent in the thermal cycle from 6 = 0.43 (T = 6500C) to 0p = 0.56
(Tp = 8500C) and down again to 0 = 0.43 can be read from Fig. 1.36. Taking the ordinate 0/0p
equal to 0.76, we obtain:

which gives

and

1.10.4 Medium thick plate solution


In a real welding situation the assumption of three-dimensional or two-dimensional heat flow
inherent in the Rosenthal equations is not always met because of variable temperature gradients in the through thickness z direction of the plate.
Model (after Rosenthal14)

The general medium thick plate model considers a point heat source moving at constant speed
across a wide plate of finite thickness d. With the exception of certain special cases (e.g.
watercooling of the back side of the plate), it is a reasonable approximation to assume that the

Previous Page

(1-103)
Equation (1-103) has been plotted in Fig. 1.36.
Example (LU)
Consider butt welding of a 2mm thin plate of austenitic stainless steel with covered electrodes
(SMAW) under the following conditions:

Calculate the retention time within the critical temperature range for chromium carbide
precipitation (i.e. from 650 to 8500C) for points located at the 8500C isotherm.
Solution

If we use the melting point of the steel as a reference temperature, the parameter n3/5 becomes:

A comparison with Fig. 1.35 shows that the assumption of 1-D heat flow is justified when
Qp< 1. Hence, the total time spent in the thermal cycle from 6 = 0.43 (T = 6500C) to 0p = 0.56
(Tp = 8500C) and down again to 0 = 0.43 can be read from Fig. 1.36. Taking the ordinate 0/0p
equal to 0.76, we obtain:

which gives

and

1.10.4 Medium thick plate solution


In a real welding situation the assumption of three-dimensional or two-dimensional heat flow
inherent in the Rosenthal equations is not always met because of variable temperature gradients in the through thickness z direction of the plate.
Model (after Rosenthal14)

The general medium thick plate model considers a point heat source moving at constant speed
across a wide plate of finite thickness d. With the exception of certain special cases (e.g.
watercooling of the back side of the plate), it is a reasonable approximation to assume that the

plate surfaces are impermeable to heat. Thus, in order to maintain the net flux of heat through
both boundaries equal to zero, it is necessary to account for mirror reflections of the source
with respect to the planes of z = 0 and z = d. This can be done on the basis of the 'method of
images' as illustrated in Fig. 1.37. By including all contributions from the imaginary sources
...2q__2 , 2g_i , 2q\ , 2q2 ,...located symmetrically at distances 2id below and above the upper
surface of the plate, the pseudo-steady state temperature distribution is obtained in the form of
a convergent series*:

(1-104)

where
Note that equation (1-104) is simply the general Rosenthal thick plate solution (equation
(1-45)) summed for each source.

Fig. 1.37. Real and imaginary point sources on a medium thick plate.
*The number of imaginary heat sources necessary to achieve the required accuracy depends on the chosen values of
R0 and vd/2a.

By substituting the dimensionless parameters defined above into equation (1-104), we obtain:
(1-105)
where

It follows from equation (1-104) that the thermal conditions will be similar to those in a
thick plate close to the centre of the weld. Moreover, Rosenthal1314has shown on the basis of
a Fourier series expansion that equation (1-104) converges to the general thin plate solution
(equation (1-81) for points located sufficiently far away from the source. However, at intermediate distances from the heat source, the pseudo-steady state temperature distribution will deviate significantly from that observed in thick plate or thin plate welding because of variable
temperature gradients in the through-thickness direction of the plate. Within this 'transition
region', the thermal programme is only defined by the medium thick plate solution (equation
(1-104)).
1.10.4.1 Dimensionless maps for heat flow analyses
Based on the models described in the previous sections, it is possible to construct a series of
dimensionless maps which provide a general outline of the pseudo-steady state temperature
distribution during arc welding.20
Construction of the maps
The construction of the maps is done on the basis of the medium thick plate solution (equation
(1-105)). This model is generally applicable and allows for the plate thickness effect in a
quantitative manner. Since the other solutions are only valid within specific ranges of this
equation, they will have their own characteristic fields in the temperature-distance or the temperature-time space. The extension of the different fields can be determined from numerical
calculations of the temperature distribution by comparing each of these models with the medium thick plate solution, using a conformity of 95% as a criterion.
Similarly, when the 95% conformity is not met between the respective solutions, the fields
are marked 'transition region'. Since any combination of dimensionless temperature, operating parameter, and plate thickness locates a point in a field, it means that the dominating heat
flow mechanism can readily be read off from the maps.
Peak temperature distribution
The variation of peak temperature with distance in the \j/(^j-direction has been numerically
evaluated from equation (1-105) for different values of the dimensionless plate thickness (8 =
vdlla). The results are shown graphically in Fig. 1.38(a) and (b) for the two extreme cases of
= 0 (z = 0) and = 8 (z = d), respectively.
An inspection of the maps reveals that the temperature-time pattern in stringer bead
weldments can be classified into three main categories:

(a)

VV

Thin plate solution


(2-D heat flow) (1-D heat flow)

(b)
Thin plate solution
1-D heat flow

V"3

(2-D heat flow)

Fig. 1.38. Peak temperature distribution in transverse direction (\|/ = \\fm) of plate; (a) Upper plate surface
(^ = 0), (b) Lower plate surface (J = 8).

1. Close to the heat source, the thermal programme will be similar to that in a thick plate
(Fig. 1.38(a)), which means that the temperature distribution is determined by equation (154). For large values of the dimensionless plate thickness, the mode of heat flow may
become essentially two-dimensional. This corresponds to the limiting case of a fast moving high power source in a thick plate (equation (1-74)). Under such conditions the slope of
the Qp/n3-\ym curves in Fig. 1.38(a) attains a constant value of-2.
2. With increasing distance from the heat source, a transition from three-dimensional to
two-dimensional heat flow may occur, depending on the dimensionless plate thickness and
the operational conditions applied. Considering the upper surface of the plate (Fig. 1.38(a)),
the extension of the transition region is seen to decrease with increasing values of 8 as the
conditions for thick plate welding are approached. The opposite trend is observed for the
bottom plate surface (Fig. 1.38(b)), since a small dimensionless plate thickness generally
results in a more rapid equalisation of the temperature gradients in the t,(z) direction. When
the curves in Fig. 1.38(b) become parallel with the jc-axis, the temperature at the bottom of
the plate reaches its maximum value. Note that within the transition region, reliable predictions of the pseudo-steady temperature distribution can only be made from the medium
thick plate solution (equation (1-105)).
3. For points located sufficiently far away from the heat source, the temperature gradients in the through-thickness direction of the plate become negligible. This implies that the
temperature distribution at the upper and lower surface of the plate is similar, and can be
computed from the thin plate solution (equation (1-83)). When the conditions for onedimensional heat flow are approached (equation (1-101)), the slope of the dp/n3-\\fm curves
in Fig. 1.38(a) and (b) attains a constant value o f - 1 .
Cooling conditions close to weld centre-line
Figure 1.39 contains a plot of the cooling programme for points located on the weld centre-line
(\j/ = = 0), as calculated from equation (1-105). A closer inspection of Fig. 1.39 reveals that
the slope of the cooling curves increases gradually from -1 to -0.5 with increasing distance
from the heat source. This corresponds to a change from three-dimensional to one-dimensional heat flow.
From Fig. 1.39 it is possible to read-off the cooling time within specific temperature intervals for a wide range of operational conditions. These results are also valid for positions
outside the weld centre-line, since the cooling curves are virtually parallel in the transverse \|/
direction of the plate. A requirement is, however, that the peak temperature of the thermal
cycle is significantly higher than the actual temperature interval under consideration.
Retention times at elevated temperatures
The retention time, Axn is defined as the total time spent in a thermal cycle from a chosen
reference temperature 0 to the peak temperature dp and down again to 9. This parameter can
readily be computed from equation (1-105) by means of numerical methods. The results of
such calculations (carried out in position = 0) are shown graphically in Fig. 1.40 for 9 = 0.5
An inspection of Fig. 1.40 reveals a complex temperature-time pattern. In this case it is not
possible to determine the exact field boundaries between the respective solutions, since the

e/n3

e/n3

Thin plate solution


(2-D heat flow) (1-D heat flow)

T= ^

e/n 3

X
Fig. 1.39. Cooling programme for points located on the weld centre-line (\\f = =0).

AXf

Fig. 1.40. Total time spent in a thermal cycle from 9 through 9p to 9 for a chosen reference temperature of
9 = 0.59p.

mode of heat flow may vary within a single thermal cycle. Hence, the extension of the different fields is not indicated in the graph. The results in Fig. 1.40 provide a systematic basis for
calculating the retention time within specific temperature intervals under various welding conditions.
Isothermal contours
Because of the number of variables involved, it is not possible to present a two-dimensional
plot of the isotherms without first specifying the dimensionless plate thickness. Examples of
calculated isotherms in different planes are shown in Figs. 1.41 and 1.42 for 8 equal to 0.5 and
5, respectively. It is evident that an increase in the dimensionless plate thickness from 0.5 to 5
has a dramatic effect on the shape and position of the isothermal contours. However, in order
to explain these observations in an adequate manner, it is necessary to condense the results into
a two-dimensional diagram. As shown in Fig. 1.43, this can be done by plotting the calculated
field boundaries in Fig. 1.38(a) at maximum width of the isotherms vs the parameters Qp/n3

and vdlla.
It is seen from Fig. 1.43 that a large plate thickness generally will favour three-dimensional
heat flow. With decreasing values of Qp/n3, the conditions for a fast moving high power source
are approached before the transition from the thick plate to the thin plate solution occurs. In
such cases the isotherms at the bottom of the plate will be strongly elongated in the welding
direction and shifted to positions far behind the heat source. The opposite trend is observed at
small values of vdlla, since a rapid equalisation of the temperature gradients in the throughthickness direction of the plate will result in elliptical isotherms at both plate surfaces, located
in an approximately equal distance from the heat source. In either case the temperature at
which the cross-sectional isotherms approach a semi-circle or become parallel with the XXz)axis can be obtained from Fig. 1.43 by reading-off the intercept between the line for the
dimensionless plate thickness and the respective field boundaries.
Limitations of the maps
Since the maps have been constructed on the basis of the analytical heat flow equations, it is
obvious that they will apply only under conditions for which these equations are valid. The
simplifying assumptions inherent in the models can be summarised as follows:
(a)

The parent material is isotropic and homogeneous at all temperatures,


and no phase changes occur on heating.

(b)

The thermal conductivity, density, and specific heat are constant and
independent of temperature.

(c)

The plate is completely insulated from its surroundings, i.e. there are no
heat losses by convection or radiation from the boundaries.

(d)

The plate is infinite except in the directions specifically noted.

(e)

The electrode is not consumed during welding, and all heat is concentrated in a zero-volume point or line.

(a)

(b)

(C)

C
Fig. 1.41. Computed isothermal contours in different sections for 8 = 0.5;
(a) Front view (\j/ = \|/m),
(b) Side view (\|/ = 0),
(c) Top view ( = 0) and bottom view ( = 8).

(a)

(b)

(C)

\Fig. 1.42.Computed isothermal contours in different sections for 8 = 5;


(a) Front view (\|/ = \|/m);
(b) Side view (\|/ = 0);
(c) Top view ( = 0) and bottom view ( = 8).

ep/n3

Thick plate solution


(3-D heat flow)

Thin plate
solution

1-D heat flow


Thick plate
solution
(2-D heat flow)

5 = vd/2a
Fig. 1.43. Heat flow mechanism map showing calculated field boundaries in transverse direction (i|/
= i|/m) of plate vs Qp/n3 and 8 = vdlla.

(f)

Pseudo-steady state, i.e. the temperature does not vary with time when
observed from a point located in the heat source.

In general, the justification of these assumptions relies on a good correlation between theory
and experiments. However, since the analytical solutions ignore the important role of arc
energy distribution and directed metal currents in the weld pool, predictions of the weld thermal programme should be restricted to positions well outside the fusion zone where such
effects are of less importance (to be discussed below).
Example (1.12)

Consider stringer bead welding (GMAW) on a 12mm thick plate of aluminium (> 99% Al)
under the following conditions:

Based on Fig. 1.43, sketch the peak temperature contours in the transverse section of the
weld at pseudo-steady state.

Solution
If we neglect the latent heat of melting, the parameter n3 at the chosen reference temperature
(Tc = Tm) becomes:

Similarly, when v = 3mm s l and a = 85mm2 s"1 we obtain the following value for the dimensionless plate thickness:

Readings from Fig. 1.43 give:


ep

^, (0C)

Model System

Comments

0.50 1.0

340 660

Medium thick
plate solution

Heat flow in x and y directions,


partial heat flow in z direction

0.17 -> 0.50

130 -> 340

Thin plate solution


(2-D heat
flow)

Heat flow in x and y directions,


negligible heat flow in z direction

< 0.17

< 130

Thin plate solution


(1 -D heat
flow)

Heat flow in y direction, negligible


heat flow in x and z directions

A sketch of the HAZ peak temperature contours in the transverse section of the weld is
shown in Fig. 1.44.
Case Study (Ll)
Consider stringer bead GMA welding on 12.5mm thick plates of low alloy steel and aluminium
(i.e. a Al-Mg-Si alloy), respectively under the conditions E = 1.5 kJ mm"1 and r\ = 0.8. Details
of welding parameters for the four series involved are given in Table 1.4. Computed peak
temperature contours in the transverse section of the welds are given in Figs. 1.45 and 1.46.

Arrows indicate heat flow directions

Fusion
zone

Fig. 1.44. Schematic diagram showing specific peak temperature contours within the HAZ of an aluminium weld at pseudo-steady state (Example 1.12).

WELD A1
y(mm)

(b)

WELD A2

z(mm)

(a)

z(mm)

y(mm)

Fig. 1.45. Computed peak temperature contours in aluminium welding at pseudo-steady state (Case study
1.1); (a) Weld Al, (b) Weld A2. Black regions indicate fusion zone.

Aluminium welding
In general, the maximum width of the isotherms at the upper and lower surface of the plate can
be obtained from Fig. 1.38(a) and (b), although these maps are not suitable for precise readings. A comparison with the computed peak temperature contours in Fig. 1.45(a) and (b)
reveals a strong influence of the welding speed on the shape and position of the cross-sectional
isotherms at a constant gross heat input of 1.5 kJ mm"1. It is evident that the extension of the
fusion zone and the neighbouring isotherms becomes considerably larger when the welding
speed is increased from 2.5 to 5 mm s"1. This effect can be attributed to an associated shift
from elliptical to more elongated isotherms at the plate surfaces (e.g. see Fig. 1.43), which
reduces heat conduction in the welding direction. It follows from Fig. 1.43 that the upper left
corner of the map represents the typical operating region for aluminium welding.
Because of the pertinent differences in the heat flow conditions, the temperature-time pattern will also vary significantly between the respective series as indicated by the maps in Figs.
1.39 and 1.40. Hence, in the case of aluminium welding the usual procedure of reporting arc
power and travel speed in terms of an equivalent heat input per unit length of the bead is highly
questionable, since this parameter does not define the weld thermal programme. In general,
the correct course would be to specify both qo, v and d, and compare the weld thermal history
on the basis of the dimensionless parameters n3 and 8.
Steel welding
If welding is performed on a steel plate of similar thickness, the operating region will be
shifted to the lower right corner of Fig. 1.43. Under such conditions, the isotherms adjacent to
the fusion boundary will be strongly elongated in the x-direction even at very low welding
speeds (see Fig. 1.42). This implies that the thermal programme approaches a state where the
temperature distribution is uniquely defined by the net heat input r\E, corresponding to the

(a)

W E L D S1

z(mm)

y(mm)

(b)

WELD S2

z(mm)

y(mm)

Fig. 7.46.Computed peak temperature contours in steel welding at pseudo-steady state (Case study 1.1);
(a) Weld Sl, (b) Weld S2. Black regions indicate fusion zone.

Table 1.4 Operational conditions assumed in Case study (1.1).


qo

Series

(W)

(mras"1)

(mm)

(kJmrrr 1 )

Al-Mg-Si
alloy

Al
A2

6000
3000

5
2.5

12.5
12.5

Low alloy
steel

Sl
S2

9600
4800

8
4

12.5
12.5

Material

n3

1.5
1.5

0.36
0.09

0.50
0.25

1.5
1.5

32.6
8.2

10
5

limiting case of a fast moving high power source. As a result, the calculated shape and width
of the fusion boundary and neighbouring isotherms are seen to be virtually independent of
choice of qo and v as illustrated in Fig. 1.46(a) and (b).
1.10.4.2 Experimental verification of the medium thick plate solution
It is clear from the above discussion that the medium thick plate solution provides a systematic
basis for calculating the temperature distribution within the HAZ of stringer bead weldments
under various welding conditions. In the following, the accuracy of the model will be checked
against extensive experimental data, as obtained from in situ thermocouple measurements and
numerical analyses of a large number of bead-on-plate welds.
Weld thermal cycles
Examples of measured and predicted weld thermal cycles in aluminium welding are presented
in Fig. 1.47. It is evident that the medium thick plate solution predicts adequately the HAZ
temperature-time pattern under different heat flow conditions for fixed values of the peak
temperature. This, in turn, implies that the model is also capable of predicting the total time
spent in a thermal cycle within a specific temperature interval as shown in Fig. 1.48.
Weld cooling programme
At temperatures representative of the austenite to ferrite transformation in mild and low alloy
steel weldments, the conditions for a fast moving high power source are normally approached
before the transition from thick plate to thin plate welding occurs (see Fig. 1.39). In such
cases, it is possible to present the different solutions for Ax875 (at \|/ = = 0) in a single graph by
introducing the following groups of variables*:

Ordinate:

Abscissa:
Relevant literature data for the cooling time between 800 and 5000C are given in Fig. 1.49.
A closer inspection of the figure reveals a reasonable agreement between observed and predicted values in all cases. For welding of thick plates, the ordinate attains a constant value of
1. Similarly, in thin sheet welding, the slope of the curve becomes equal to 1.
In aluminium welding, the thermal conditions will be much more complex because of the
resulting higher base plate thermal diffusivity (see Fig. 1.39). Hence, it is not possible to
describe the weld cooling programme in terms of equations (1-74) and (1-101), which apply to
fast moving high power sources. The plot in Fig. 1.50 confirms, however, that the medium
thick plate solution is also capable of predicting the cooling time within specific temperature
intervals (e.g. from 300 to 1000C) in aluminium weldments.

These groups of variables can be obtained from equations (1-74) and (1-101).

Measured
Predicted

Temperature (0C)

GMAW: q 0 = 3872 W, v = 8.8 mm/s,


d =8 mm

Time (sec)

A tr(s), observed

Fig. 1.47. Comparison between measured and predicted weld thermal cycles in aluminium welding for
fixed values of Tp. Data from Myhr and Grong.20

P
t

A tr(s), predicted
Fig. /.4#. Comparison between measured and predicted retention times in aluminium welding for fixed
values of Tp. Data from Myhr and Grong.20

8/5
AT

n 3 [-i--4-

SMAW
SMAW
SAW
SAW
THICK PLATE
SOLUTION

THIN PLATE
SOLUTION

-2Hr + Ir 1
5 6SOO 6SOO
Fig. 1.49. Comparison between observed and predicted cooling times from 800 to 5000C in steel welding
(solid lines represent theoretical calculations). Data from Myhr and Grong.20

At

(sec), observed

GMAW (Al+2.5 wt% Mg)

T(0C)

t(sec)

At

(sec), predicted

Fig. 1.50. Comparison between observed and predicted cooling times from 300 to 1000C in aluminium
welding. Data from Myhr and Grong.20

Peak temperatures and isothermal contours


Figure 1.51 shows a comparison between measured and predicted HAZ peak temperatures in
aluminium welding. It is evident that the relative positions of the HAZ isotherms can be
calculated with a reasonable degree of accuracy from the medium thick plate solution, provided that the equation is precalibrated against a known isotherm (i.e. the fusion boundary).
Additional information on the HAZ peak temperature distribution in aluminium welding
can be obtained from the data of Koe and Lee21 reproduced in Fig. 1.52. These numerical
calculations* showed a good correlation with experimental measurements. A comparison with
the medium thick plate solution in Fig. 1.52 reveals a fair agreement between numerically and
analytically computed isothermal contours. It is interesting to note that even though the plate
thickness is small (i.e. 3.2mm), the mode of heat flow becomes essentially three-dimensional
close to the fusion boundary. This important point is often overlooked when discussing the
relevance of the analytical heat flow models in thin sheet welding.
LlOAJ Practical implications
The following important conclusions can be drawn from the results presented in Figs. 1.381.52:

Tp, 0C , observed

GMAW (Al +2.5 wt% Mg)

Tp, 0C , predicted
Fig. 1.51. Comparison between observed and predicted HAZ peak temperatures in aluminium welding.
Data from Myhr and Grong.20
Based on the finite difference method (FDM).

Z (mm)

y (mm)

Fig. 1.52. Comparison between numerically and analytically computed peak temperature contours in
GTA welding of a 3.2mm thin aluminium sheet. (Broken lines: numerical model; solid lines: analytical
model.) Data for welding parameters and material properties are given in Ref. 21.

1. Considering heat flow and temperature distribution in fusion welding, there exists no
defined plate thickness which can be regarded as 'thick' or 'thin'. Accordingly, in a real
welding situation, the mode of heat flow will vary continuously with increasing distance
from the heat source.
2. Close to the centre of the weld, the thermal conditions will be similar to those in a
heavy slab. This means that the temperature distribution is approximately described by the
Rosenthal thick plate solution.
3. At intermediate distances from the heat source, the temperature distribution will deviate significantly from that observed in thick plate welding because of variable temperature gradients in the through-thickness direction of the plate. Within this transition region,
reliable predictions of the pseudo-steady state temperature distribution can only be made
from the medium thick plate solution.
4. For points located sufficiently far away from the heat source, the temperature gradients in the through-thickness direction of the plate become negligible. Under such conditions, the weld thermal programme is approximately defined by the Rosenthal thin plate
solution.
5. In general, a full description of the weld thermal history requires that both the arc
power qo, the travel speed V, and the plate thickness d are explicitly specified. Hence, the
usual procedure of reporting welding variables in terms of an equivalent heat input per unit
length of the bead, (kJ mm"1), is highly questionable, since this parameter does not define
the weld thermal programme. An exception is welding of thick steel plates, where the
temperature distribution approaches that of a fast moving high power source because of a
low thermal diffusivity of the base metal.
6. A comparison between theory and experiments shows that the medium thick plate
solution predicts adequately both the peak temperature distribution and the temperaturetime pattern within the HAZ of stringer bead weldments for a wide range of operational
conditions (including aluminium and steel welding). A requirement is, however, that the
equation is calibrated against a known isotherm (e.g. the fusion boundary) due to the simplifying assumptions inherent in the model.

1.10.5 Distributed heat sources


In some cases it is also necessary to consider the important influence of filler metal additions,
arc energy distribution, and convectional heat flow in the weld pool on the resulting bead
morphology to obtain a good agreement between theory and experiments. Of particular interest in this respect is the formation of the so-called weld crater/weld finger, frequently observed
in GMA and SA stringer bead weldments (see Fig. 1.53). Although the nature of these phenomena is very complex, they can readily be accounted for by applying an empirical correction for the effective heat distribution in the weld pool.
1.10.5.1 General solution
Model (after Myhr and Grong22)
The heat distribution used to simulate the weld crater/weld finger formation is shown
schematically in Fig. 1.54(a). Here we consider two discreate distributions of elementary
point sources, which extend in the y- and z-direction of the plate, respectively. The contribution from each source to the temperature rise in an arbitrary point P located within the plate is
calculated on the basis of the "method of images", as shown in Fig. 1.54(b) and (c). For a heat
source displaced in the y-direction (Fig. 1.54(b)), the temperature field is given by equation (1 104):

(1-106)
where

y
Crater

HAZ"

Finger
Fusion line

Z
Fig. 1.53. Schematic diagram showing the weld crater/weld finger formation during stringer bead welding.

(a)

(b)

Z
Fig. 1.54. General heat flow model for welding on medium thick plates; (a) Physical representation of the
heat distribution by elementary point sources, (b) Method for calculating the temperature field around an
elementary point source displaced along the y-axis.

(C)

Z
Fig. 1.54. General heat flow model for welding on medium thick plates(continued): (c) Method for calculating the temperature field around an elementary (submerged) point source displaced along the z-axis.
Similarly, for a submerged point source located along the z-axis (Fig. 1.54(c))> we obtain:

(1-107)

where
and

Note that equation (1-107) correctly reduces to equation (1-106) when A approaches zero.
The total temperature rise in point P is then obtained by superposition of the temperature
fields from the different elementary heat sources, i.e.:
(1-108)
where

In practice, we can subdivide the heat distributions into a relatively small number of elementary point sources, and usually a total number of 8 to 10 sources is sufficient to obtain
good results (i.e. smooth curves). However, the relative strength of each heat source and their
distribution along the y- and z-axes must be determined individually by trial and error by
comparing the calculated shape of the fusion boundary with the real (measured) one.
Figure 1.55 shows the results from such calculations, carried out for a single pass (bead-ingroove) GMA steel weld. It is evident that the important effect of the weld crater/weld finger
formation on the HAZ peak temperature distribution is adequately accounted for by the present
model. A weakness of the model is, of course, that the shape and location of the fusion boundary must be determined experimentally before a prediction can be made.
1.10.5.2 Simplified solution
Similar to the situation described above, the point heat source will clearly not be a good
model when the heat is supplied over a large area. Welding with a weaving technique and surfacing with strip electrodes are prime examples of this kind.
Model (after Grong and Christensen19)

As a first simplification, the Rosenthal thick plate solution is considered for the limiting case
of a high arc power qo and a high welding speed D, maintaining the ratio qo Iv within a range
applicable to arc welding. Consider next a distributed heat source of net power density qo I2L
extending from -L to +L on either side of the weld centre-line in the y-direction*, as shown
schematically in Fig. 1.56. It follows from equation (1-73) that an infinitesimal source dqy
located between y and y + dy will cause a small rise of temperature in point P at time f, as:
(1-109)
where

and

Alternatively, we can use a Gaussian heat distribution, as shown in Appendix 1.4.

z (mm)

Shaded region indicates


fusion zone

y (mm)
Fig. 1.55. Calculated peak temperature contours in the transverse section of a GMA steel weldment (Operational conditions: / = 450A, U - 30V, v = 2.6mm s"1, d = 50mm).

2-D heat flow

Fig. 1.56. Distributed heat source of net power density qJ2L on a semi-infinite body (2-D heat flow).

(1-110)

where erf(u) is the Gaussian error function (previously defined in Appendix 1.3).
Peak temperature distribution
Because of the complex nature of equation (1-110), the variation of peak temperature with
distance in the y-z plane can only be obtained by numerical or graphical methods. Accordingly, it is convenient to present the different solutions in a dimensionless form. The following
parameters are defined for this purpose:
Dimensionless operating parameter:

(1-111)

Dimensionless time:
(1-112)
Dimensionless y-coordinate:
(1-113)
Dimensionless z-coordinate:
(1-114)
By substituting these parameters into equation (1-110), we obtain:

(1-115)

where 0 is the dimensionless temperature (previously defined in equation (1-9)).

The variation of peak temperature Qp with distance in the y-z plane has been numerically
evaluated from equation (1-115) for chosen values of P and 7 (i.e. P = 0, (3 = 3/4, P = I , and
7 = 0). The results are presented graphically in Figs. 1.57 and 1.58 for the through thickness
(z = zm) and the transverse (y = ym) directions, respectively. These figures provide a systematic
basis for calculating the shape of the weld pool and neighbouring isotherms under various
welding conditions.
In Fig. 1.59 the weld width to depth ratio has been computed and plotted for different
combinations of nw and 9p. It is evident that the predicted width of the isotherms generally is
much greater, and the depth correspondingly smaller than that inferred from the point source
model. Such deviations tend to become less pronounced with decreasing peak temperatures
(i.e. increasing distance from the heat source). At very large value of nw, the theoretical shape
of the isotherms approaches that of a semi-circle, which is characteristic of a point heat source.
Example (1.13)

Consider GMA welding with a weaving technique on a thick plate of low alloy steel under the
following conditions:

Sketch the contours of the fusion boundary and the Ac r isotherm (71O0C) in the y-z plane.
Compare the shape of these isotherms with that obtained from the point heat source model.
Solution
If we neglect the latent heat of melting, the operating parameter at the chosen reference temperature (Tc = Tm) becomes:

Fusion boundary
Here we have:

Readings from Figs. 1.57 and 1.58 give:

Ac j-temperature
In this case the peak temperature should be referred to 7200C, i.e.:

ep/nw

V -

ep/nw

Fig. /.57. Calculated peak temperature distribution in the through-thickness direction of the plate at different positions along the weld surface.

y m /L
Fig. 1.58. Calculated peak temperature distribution in the transverse direction of the plate at position y
(Z) = 0.

y m /z m

nw
Fig. 1.59. Effects of nw and dp on the weld width to depth ratio.

from which

Readings from Figs. 1.57 and 1.58 give:

Similarly, equation (1-75) provides a basis for calculating the width of the isotherms in the
limiting case where all heat is concentrated in a zero-volume point. By rearranging this equation, we obtain:

which gives
6.3 mm when Bp = 1(Tp = 15200C)

and
when
Figure 1.60 shows a graphical presentation of the calculated peak temperature contours.
Implications of model

It is evident from Fig. 1.60 that the predicted shape of the isotherms, as evaluated from equation (1-110), departs quite strongly from the semi-circular contours required by a point heat
source. Moreover, a closer inspection of the figure shows that inclusion of the heat distribution
also gives rise to systematic variations in the weld thermal programme along a specific
isotherm, as evidenced by the steeper temperature gradient in the v-direction compared with
the z-direction of the plate. This point is more clearly illustrated in Fig. 1.61, which compares
the HAZ temperature-time programme for the two extreme cases of z = 0 and y = 0, respectively. It is obvious from Fig. 1.61 that the retention time within the austenite regime is
considerably longer in the latter case, although the cooling time from 800 to 5000C, Af8/5, is
reasonably similar. These results clearly underline the important difference between a point
heat source and a distributed heat source as far as the weld thermal programme is concerned.
Model limitations

In the present model, we have used the simplified solution for a fast moving high power
source (equation 1-73)) as a starting point for predicting the temperature-time pattern. Since
the equations derived later are obtained by integrating equation (1-73), they will, of course,
apply only under conditions for which this solution is valid.
Moreover, a salient assumption in the model is that the heat distribution during weaving
can be represented by a linear heat source orientated perpendicular to the welding direction.
Although this is a rather crude approximation, experience shows that the assumed heat dis-

z, mm
Fig. /.60. Predicted shape of fusion boundary and Acpisotherm during GMA welding of steel with an
oscillating electrode (Example 1.13). Solid lines: Distributed heat source; Broken lines: Point heat source.

Temperature, 0C

Time, s
Fig. L61. Calculated HAZ thermal cycles in positions y = 0 and z = 0 (Example 1.13).
tribution is not critical unless the rate of weaving is kept close to the travel speed. However,
for most practical applications weaving at such low rates would be undesirable owing to an
unfavourable bead morphology.
Case Study (1.2)

Surfacing with strip electrodes makes a good case for application of equation (1-110). Specifically, we shall consider SA welding of low alloy steel with 60mm X 0.5mm stainless steel
electrodes. The operational conditions employed are listed in Table 1.5.
It is evident from the metallographic data presented in Fig. 1.62 that neither the bead penetration nor the HAZ depth (referred to the plate surface) can be predicted readily on the basis
of the present heat flow model when welding is carried out with a consumable electrode,
owing to the formation of a reinforcement. This situation arises from the simplifications made
in deriving equation (1-110). The problem, however, may be eliminated by calculating the
depth of the Ac3 and Ac1 regions relative to the fusion boundary, i.e. Azm = zm(Qp) - zm (Qp = 1),
or Aym = ym (8p) - ym (6p = 1), for specific positions along the weld fusion line, as shown by the
solid curves in Fig. 1.62 for Qp = 0.54 and 0.45, respectively. An inspection of the graphs
reveals satisfactory agreement between theory and experiments in all three cases, which implies that the model is quite adequate for predicting the HAZ thermal programme as far as strip
electrode welding is concerned. This result is to be expected, since the assumption of twodimensional heat flow is a realistic one under the prevailing circumstances.
Case Study (1.3)

As a second example we shall consider GTA welding (without filler wire additions) at various
heat inputs and amplitudes of weaving within the range from 1 to 2.5 kJ mm"1 and 0 to 15mm,
respectively. Data for welding parameters are given in Table 1.6.

Stainless steel!
y, mm

Weld S3

Fusion line

z, mm

Stainless steel
y, mm

Weld S4

z, mm
Stainless steel!
y, mm
Weld S5

z, mm
Fig. 1.62. Comparison between observed and predicted Ac3 and Ac1 contours during strip electrode welding (Case study 1.2). Data from Grong and Christensen.19
Table 1.5 Operational conditions used in strip electrode welding experiments (Case study 1.2).
Base metal/
filler metal
combination

Weld
No.

S3
Low alloy steel/
stainless steel

/
(A)

730

S4

73

S5

730

U
(V)

27
27

27

v
(mms"1)

2L
(mm)

nw
Cn = 0.7)

1.8

60

0.34

2 2

2.5

60
60

0.28
0.24

Table 1,6 Operational conditions used in GTA welding experiments (Case study 1.3).
I

2L
1

nw
"V=O-

12

WeIdNo.

(A)

(V)

(mms" )

(mm)

T] = 0.23

Bl

200

13.5

2.6

9.5

0.20

0.40

B2

200

14.0

1.1

15.0

0.20

0.40

B3

200

13.5

2.5

B4

200

12.5

1.0

Calibration procedure
In general, a comparison between theory and experiments requires that the arc efficiency factor can be established with a reasonable degree of accuracy. Unfortunately, the arc efficiency
factor for GTA welding has not yet been firmly settled, where values from 0.25 up to 0.75 have
been reported in the literature (see Table 1.3). Additional problems result from the fact that
only a certain fraction of the total amount of heat transferred from the arc to the base plate is
sufficiently intense to cause melting. This has led to the introduction of the melting efficiency
factor T]m, which normally is found to be 30-70% lower than the total arc efficiency of the
process, depending on the latent heat of melting, the applied amperage, voltage, shielding gas
composition, or electrode vertex angle.23 Consequently, since these parameters cannot readily
be obtained from the literature, the following reasonable values for r\m and r\ have been assumed to calculate nw in Table 1.6, based on a pre-evaluation of the experimental data: j \ m =
0.12 (fusion zone), TI = 0.23 (HAZ).
It should be noted that the above values also include a correction for three-dimensional heat
flow, since the assumption of a fast moving high power source during low heat input GTA
welding is not valid. Hence, both the arc efficiency factor and the melting efficiency factor
used in the present case study are seen to be lower than those commonly employed in the
literature.
Full weaving (welds Bl and B2)
The results from the metallographic examination of the two GTA welds deposited under full
weaving conditions are presented graphically in Fig. 1.63. Note that the shape of the fusion
boundary as well as the Ac3 and the Ac1 isotherms can be predicted adequately from the present
model for both combinations of E and L (an exception is the HAZ end points in position z = 0),
provided that proper adjustments of i\m and Tj are made. The good correlation obtained in Fig.
1.63 between the observed and the calculated peak temperature contours justifies the adaptation of the model to low heat input processes such as GTA welding, despite the fact that the
assumption of two-dimensional heat flow is not valid under the prevailing circumstances.
No weaving (welds B3 and B4)
For the limiting case of no weaving (Fig. 1.64), the concept of an equivalent amplitude of
weaving has been used in order to calculate the peak temperature contours from the model.
This parameter (designated Leq) takes into account the effects of convectional heat flow in the
weld pool on the resulting bead geometry, and is evaluated empirically from measurements of
the actual weld samples. At low heat inputs (Fig. 1.64(a)), the agreement between theory and

y, mm

Weld B1

z, mm
Fusion line

y, mm

Weld B2

z, mm

Fig. 1.63. Comparison between observed and predicted fusion line, Ac3 and Ac1 contours during GTA
welding under full weaving conditions (Case study 1.3). Data from Grong and Christensen.19
experiments is largely improved by inserting 2Leq = 7.5mm into equation (1-110), when comparison is made on the basis of the point source model. In contrast, at a heat input of 2.5 kJ
mm"1 (Fig. 1.64(b)), the measured shape of the HAZ isotherms is seen to approach that of a
semi-circle, and hence the deviation between the present model and the simplified solution for
a fast moving high power point source is less apparent.
Intermediate weaving
At intermediate amplitudes of weaving (2L = 5 and 7.5mm, respectively), convectional heat
flow in the weld pool will also tend to increase the bead width to depth ratio beyond the
theoretical value predicted from the present model, as shown in Fig. 1.65. The plot in Fig. 1.65

Weld B3
y, mm

z, mm

Fusion line

y, mm

Weld B4

z, mm
Fig. 1.64. Comparison between observed and predicted fusion line, Ac3 and Ac1 contours during GTA
welding with a stationary arc (Case study 1.3). Solid lines: Distributed heat source, Broken lines: Point
heat source. Data from Grong and Christensen.19
includes all data obtained in the GTA welding experiments with an oscillating arc, as reported
by Grong and Christensen.19 These results suggest that the applied amplitude of weaving must
be quite large before such effects become negligible. Consequently, adaptation of the model to
the weld series considered above would require an empirical calibration of the weaving amplitude similar to that performed in Fig. 1.64 for stringer bead weldments to ensure satisfactory
agreement between theory and experiments.
1.10.6 Thermal conditions during interrupted welding
Rapid variations of temperatures as a result of interruption of the welding operation can have
an adversely effect on the microstructure and consequently the mechanical properties of the
weldment.

Width

Width to depth ratio

Depth

Theoretical
curve

Fig. 1.65.Comparison between observed and predicted weld width to depth ratios during GTA welding
with an oscillating arc (Case study 1.3). Data from Grong and Christensen.19

T,e

t,t

Fig. 1.66. Idealised heat flow model for prediction of transient temperatures during interrupted welding.

Model (after Rykalin9)

The situation existing after arc extinction may be described as shown in Fig. 1.66. From time
t = t* there is no net heat supply to the weldment. This condition is satisfied if the real source
q0 is considered maintained by adding an imaginary source +qo and sink -qo of the same
strength at t*. The temperature at some later time t** in a given position R0 (measured from the
origin 0") is then equal to the difference of temperatures due to the positive heat sources qo
and +qo and the negative heat sink -qo. Each of these temperature contributions will be a
product of a pseudo-steady state temperature Tps, and a correction factor K1 or K2 (given by
equations (1-49) and (1-82), respectively). Hence, for 3-D heat flow, we have:
(1-116)
where
and
Similarly, for 2-D heat flow, we get:
(1-117)
where

(ro is the position of the weld with respect to the imaginary heat source at time f* in the x y
plane).
Example (1.14)

Consider repair welding of a heavy steel casting with covered electrodes under the following
conditions:

Suppose that a 50mm long bead is deposited on the top of the casting. Calculate the temperature in the centre of the weld 5 s after arc extinction.
Solution

The pseudo-steady state temperature for points located on the weld centre-line (i|/ = = 0) can
be obtained from equation (1-65). When t** - t* = 5 s, we get:

Referring to Fig. 1.67, the position of the weld with respect to the imaginary heat source at
time f* is 10mm, which gives:

Fig. 1.67. Sketch of weld bead in Example 1.14.

(a)
3-D heat flow

(b)
3-D heat flow

(C)
3-D heat flow

Fig. /.6#. Recommended correction factor/for some joint configurations; (a) Single V-groove, (b) Double V-groove, (c) T-joint.

Moreover, the dimensionless times T** and T** - x* are:

and
At these coordinates, the correction factor K1 is seen to be 1 and 0.62, respectively (Fig.
1.18). The temperature in the centre of the weld 5 s after arc extinction is thus:
which is equivalent to

LlOJ

Thermal conditions during root pass welding

During conventional bead-on-plate welding the angle of heat conduction is equal to 180 due
to symmetry effects (e.g. see Fig. 1.23). In order to apply the same heat flow equations during
root pass welding, it is necessary to introduce a correction factor,/, which takes into account
variations in the effective heat diffusion area due to differences in the joint geometry. Taking
/equal to 1 for ordinary bead-on-plate welding (b.o.p.), we can define the net heat input of a
groove weld as:9
(1-118)
Recommended values of the correction factor/for some joint configurations are given in
Fig. 1.68.
Example (Ll5)

Consider deposition of a root pass steel weld in a double-V-groove with covered electrodes
(SMAW) under the following conditions:

Calculate the cooling time from 800 to 5000C (Ar875), and the cooling rate (CR.) at 6500C
in the centre of the weld when the groove angle is 60.
Solution

The cooling time, Ar875, and the cooling rate, CR., can be obtained from equation (1-68) and
(1-71), respectively:
Cooling time, Atm

Cooling rate at 6500C

The above calculations show that the thermal conditions existing in root pass welding may
deviate significantly from those prevailing during ordinary stringer bead deposition due to
differences in the effective heat diffusion area. These results are in agreement with general
experience (see Fig. 1.69).
1.10.8 Semi-empirical methods for assessment of bead morphology
In fusion welding fluid flow phenomena will have a strong effect on the shape of the weld
pool. Since flow in the weld pool is generally driven by a combination of buoyancy, electromagnetic, and surface tension forces (e.g. see Fig. 3.10 in Chapter 3), prediction of bead morphology from first principles would require detailed consideration of the current and heat flux
distribution in the arc, the interaction of the arc with the weld pool free surface, convective
heat transfer due to fluid flow in the liquid pool, heat of fusion, convective and radiative losses
from the surface, as well as heat and mass loss due to evaporation.
Over the years, a number of successful studies have been directed towards numerical weld
pool modelling, based on the finite difference, the finite element, or the control volume approach.24"31 Although these studies provide valuable insight into the mechanisms of weld pool
development, the solutions are far too complex to give a good overall indication of the heatand fluid-flow pattern. The present treatment is therefore confined to a discussion of factors
affecting the nominal composition of single-bead fusion welds. This composition can be obtained from an analysis of the amount of deposit D and the fused part of the base material B,
from which we can calculate the mixing ratio BI(B + D) or DI(B + D). Methods have been
outlined in the preceding sections for handling such problems by means of point or line source
models. The following section gives a brief description of procedures which can be used for
predictions of the desired quantities in cases where the classic models break down, or where
the calculation will be too tedious.
1.10.8.1 Amounts of deposit and fused parent metal
The heat conduction theory does not allow for the presence of deposited metal. The rate of
deposition, dMwldt, is roughly proportional to the welding current /, and is often reported as a
coefficient of deposition, defined as:
(1-119)
Since the area of deposited metal D is frequently wanted, we may write:
(1-120)
where p is the density, and v is the welding speed.

Groove angle: <>


| = 60

Cooling time, AtQ/5(s)

Plate thickness:

Heat input, E (kJ/mm)


Fig. 1.69. Comparison between observed and predicted cooling times from 800 to 5000C in root pass
welding of steel plates (groove preparation as in Fig. 1.68(b)). Data from Akselsen and Sagmo.34
Recommended values of k'/p for some arc welding processes are given in Table 1.7. In
practice, the deposition coefficient k'/p will also vary with current density and electrode stickout
due to resistance heating of the electrode. Consequently, the numbers contained in Table 1.7
are estimated averages, and should therefore be used with care.
Example (1.16)

Consider stringer bead deposition (S AW) on a thick plate of low alloy steel under the following conditions:

Table 1.7 Average rates of volume deposition in arc welding. Data from Christensen.32
Welding Process

k'/p (mm3 A"1 s l)

SMAW

0.3-0.5

GMAW, steel

0.6-0.7

GMAW, aluminium

-0.9

SAW, steel

-0.7

Calculate the mixing ratio Bf(B + D) at pseudo-steady state.


Solution

The amount of fused parent material can be obtained from equation (1-75). If we include an
empirical correction for the latent heat of melting, the dimensionless radius vector a4m becomes:

This gives:

Similarly, the amount of deposited metal can be calculated from equation (1-120). Taking
ifc'/p equal to 0.7mm3 A"1 s"1 for SAW (Table 1.7), we get:

The mixing ratio is thus:

Example (L 17)

Consider stringer bead deposition with covered electrodes (SMAW) on a thick plate of low
alloy steel under the following conditions:

Calculate the mixing ratio BI(B + D) at pseudo-steady state.


Solution

In this particular case the conditions for a fast moving high power source are not met. Thus, in
order to eliminate the risk of systematic errors, the amount of fused parent metal should be
calculated from the general Rosenthal thick plate solution (equation (1-45)) or read from Fig.
1.21. When Tc = Tm (i.e. 8^ = 1), we obtain:

Reading from Fig. 1.21 gives:

and

Moreover, the amount of deposited metal can be calculated from equation (1-120). Taking
k'/p equal to 0.4mm3 A"1 s~l for SMAW (Table 1.7), we get:

and

The above calculations indicate a small difference in the mixing ratio between SA and
SMA welding, but the data are not conclusive. In practice, a value of BI(B + D) between 1/3
and 1/2 is frequently observed for SMAW, while the mixing ratio for SAW is typically 2/3 or
higher. The observed discrepancy between theory and experiments arises probably from difficulties in estimating the amount of fused parent metal from the point heat source model.
1.10.8.2 Bead penetration
It is a general experience in arc welding that the shape of the fusion boundary will depart quite
strongly from that of a semi-circle due to the existence of high-velocity fluid flow fields in the
weld pool.24"31 For combinations of operational parameters within the normal range of arc
welding, a fair prediction of bead penetration h can be made from the empirical equation
derived by Jackson et al.:33
(1-121)
A summary of Jackson's data is shown in Table 1.8. It is seen that the constant C in equation (1-121) has a value close to 0.024 for SAW and SMAW with E6015 type electrodes, and
about 0.050 for GMAW with CO2 -shielding gas. Penetration measurements of GMA/Ar + O2,
GMA/Ar, and GMA/He welds, on the other hand, show a strong dependence of polarity, and
shielding gas composition, to an extent which makes the equation useless for a general prediction. Such data have therefore not been included in Table 1.8.
Example (1.18)

Based on the Jackson equation (equation (1-121)), calculate the bead penetration for the two
specific welds considered in Examples (1.16) and (1.17). Use these results to evaluate the
applicability of the point heat source model under the prevailing circumstances.
Solution

From equation (1-121) and Table 1.8, we have:

Table 1.8 Recommended bead penetration coefficients for some arc welding processes. Data from
Jackson.33
Welding Process

Comments

SAW, steel

-0.024

Various types of fluxes


(Zz from 3 to 15 mm)

SMAW, steel
(E6015)

-0.024

Wide range of /, U, and v


(h from 0.7 to 5mm)

GMAW, steel
(CO2 - shielding)

-0.050

Electrode positive
(h from 6.5 to 8mm)

and

The corresponding values predicted from the point heat source model are:

and

Provided that the Jackson equation gives the correct numbers, it is obvious from the above
calculations that the point heat source model is not suitable for reliable predictions of the bead
penetration during arc welding. This observation is not surprising.
1.10.9 Local preheating
So far, we have assumed that the ambient temperature T0 remains constant during the welding
operation (i.e. is independent of time). The use of a constant value of T0 is a reasonable approximation if the work-piece as a whole is subjected to preheating. In many cases, however,
the dimensions of the weldment allow only preheating of a narrow zone close to the weld.
This, in turn, will have a significant influence on the predicted weld cooling programme, particularly in the low temperature regime where the classic models eventually break down when
T approaches T0.
Model (after Christensen n)

The idealised preheating model is shown in Fig. 1.70. Here it is assumed that the weld centreline temperature is equal to the sum of the contributions from the arc and from the field of
preheating. The former contribution is given by equation (1-45) for R = -x = Vt, provided that
the plate thickness is sufficiently large to maintain 3-D heat flow. Similarly, the temperature
field due to preheating can be calculated as shown in Section 1.7 for uniaxial heat conduction
from extended sources (thermit welding). By combining equations (1-45) and (1-22), we
obtain the following relation for the weld centre-line:
(1-122)

Temperature profile
att = 0

Weld
Preheated
zone

z
Fig. 1.70. Sketch of preheating model.
where T* is the local preheating temperature, and L* is the half width of the preheated zone.
Equation (1-122) can be written in a general form by introducing the following groups of parameters:
Dimensionless temperature:
(1-123)

Time constant:
(1-124)

Dimensionless time:
(1-125)

Dimensionless half width of preheated zone:


(1-126)
By inserting these parameters into equation (1-122), we obtain:
(1-127)

It is evident from the graphical representation of equation (1-127) in Fig. 1.71 that the
predicted weld cooling programme falls within the limits calculated for Q"-^ 0 (no preheating)
and Q /7 -> oo (global preheating). The controlling parameter is seen to be the dimensionless
half width of the preheated zone Q", which depends both on the actual width L*, the base plate
thermal properties a, X, and the net heat input qo Iv.
Example (1.19)

Consider stringer bead deposition with covered electrodes (SMAW) on a thick plate of low
alloy steel under the following conditions:

Calculate the cooling time from 800 to 50O0C (A%5), and the cooling time 1Oo measured
from the moment of arc passage to the temperature in the centre of the weld reaches 10O0C.
Solution
First we calculate the time constant to from equation (1-124):

e*-

from which we obtain

^6

Fig. 1.71. Graphical representation of equation (1-127).

Next Page
Cooling time, At8/5

The dimensionless temperatures conforming to 800 and 5000C are:

Reading from Fig. 1.71 gives:

from which

This cooling time is only slightly longer than that calculated from equation (1-68) for T0 =
200C (6.9s), showing that moderate preheating up to 1000C is not an effective method of
controlling Ar875.
Cooling time, t]00

When T=T0* = 1000C, the dimensionless temperature 9* = 1. Reading from Fig. 1.71 gives
T 6 - 10, from which:

The above value should be compared with that evaluated from the numerical data of Yurioka
et al.,35 replotted in Fig. 1.72 (see p.104). It follows from Fig. 1.72 that the weld cooling
programme in practice is also a function of the plate thickness d, an effect which cannot readily be accounted for in a simple analytical treatment of the heat diffusion process. For the
specific case considered above the parameter ^100 varies typically from 500 to 900s, depending
on the chosen value of d. This cooling time is significantly shorter than that calculated from
equation (1-127), indicating that the analytical model is only suitable for qualitative predictions.

References
1.
2.
3.
4.
5.
6.

H.S. Carslaw and J.C. Jaeger: Conduction of Heat in Solids; 1959, Oxford, Oxford University
Press.
British Iron and Steels Research Association: Physical Constants of some Commercial Steels
at Selected Temperatures; 1953, London, Butterworths.
R. Hultgren, R.L. Orr, RD. Anderson and K.K. Kelly: Selected Values of Thermodynamic
Properties of Metals and Alloys; 1963, New York, J. Wiley & Sons.
E. Griffiths (ed.): J. Iron and Steel Inst., 1946,154, 83-121.
J.E. Hatch (ed.): Aluminium Properties and Physical Metallurgy; 1984, Metals Park (Ohio),
American Society for Metals.
Metals Handbook, 9th edn., Vol. 2, 1979, Metals Park (Ohio), American Society for Metals.

Previous Page
Cooling time, At8/5

The dimensionless temperatures conforming to 800 and 5000C are:

Reading from Fig. 1.71 gives:

from which

This cooling time is only slightly longer than that calculated from equation (1-68) for T0 =
200C (6.9s), showing that moderate preheating up to 1000C is not an effective method of
controlling Ar875.
Cooling time, t]00

When T=T0* = 1000C, the dimensionless temperature 9* = 1. Reading from Fig. 1.71 gives
T 6 - 10, from which:

The above value should be compared with that evaluated from the numerical data of Yurioka
et al.,35 replotted in Fig. 1.72 (see p.104). It follows from Fig. 1.72 that the weld cooling
programme in practice is also a function of the plate thickness d, an effect which cannot readily be accounted for in a simple analytical treatment of the heat diffusion process. For the
specific case considered above the parameter ^100 varies typically from 500 to 900s, depending
on the chosen value of d. This cooling time is significantly shorter than that calculated from
equation (1-127), indicating that the analytical model is only suitable for qualitative predictions.

References
1.
2.
3.
4.
5.
6.

H.S. Carslaw and J.C. Jaeger: Conduction of Heat in Solids; 1959, Oxford, Oxford University
Press.
British Iron and Steels Research Association: Physical Constants of some Commercial Steels
at Selected Temperatures; 1953, London, Butterworths.
R. Hultgren, R.L. Orr, RD. Anderson and K.K. Kelly: Selected Values of Thermodynamic
Properties of Metals and Alloys; 1963, New York, J. Wiley & Sons.
E. Griffiths (ed.): J. Iron and Steel Inst., 1946,154, 83-121.
J.E. Hatch (ed.): Aluminium Properties and Physical Metallurgy; 1984, Metals Park (Ohio),
American Society for Metals.
Metals Handbook, 9th edn., Vol. 2, 1979, Metals Park (Ohio), American Society for Metals.

Cooling time, t100(s)

Heat input:
E=1.7 kJ/mm

Preheating temperature, T^ (0C)


Fig. 1.72. Cooling time to 1000C, tm, in steel welding for different combinations of T0*, L*, d and E.
Data from Yurioka et a/.35

7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.

K.P. Bentley, J.A. Greenwood, R McKnowlson and R.G. Bakes: Brit. Weld. J., 1963,10, 613619.
N.N. Rykalin, A.I. Pugin and V.A. Vasil'eva: Weld. Prod., 1959, 6, 42-52.
N.N. Rykalin: Berechnung der Warmevorgdnge beim Schweissen; 1953, Berlin, VEB Verlag
Technik.
H.M. Tensi, W. Welz and M. Schwalm: Aluminium, 1981, 58, 515-518.
N. Christensen, V. de L. Davis and K. Gjermundsen: Brit. Weld. J., 1965,12, 54-75.
Welding Handbook, 8th edn., Vol. 2, 1991, Miami (Florida), American Welding Society.
D. Rosenthal: Weld. / , 1 9 4 1 , 20, 220s-234s.
D. Rosenthal: Trans. ASME, 1946, 68, 849-866.
CM. Adams: Weld. J., 1958, 37, 210s-215s.
RS. Myers, O.A. Uyehara and G.L. Borman: Weld. Res. Bull., 1967, 123, 1-46.
T.W Eagar and N.S. Tsai: Weld. J., 1983, 62, 346s-355s.
M.F. Ashby and K.E. Easterling: Ada Metall, 1984, 32, 1935-1948.
0. Grong and N. Christensen: Mater. ScL Tech., 1986, 2, 967-973.
O.R. Myhr and 0 . Grong: Acta Metall. Mater., 1990, 38, 449-460.
S. Kou and Y. Le: Metall. Trans., 1983,14A, 2245-2253.

22.
23.
24.
25.
26.
27.
28.
29.
30.
32.
33.
34.
35.

O.R. Myhr and 0. Grong: Unpublished work, 1990, University of Trondheim, The Norwegian
Institute of Technology.
R.W. Niles and CE. Jackson: Weld. J., 1975, 54, 25s-32s.
G.M. Oreper, T.W. Eagar and J. Szekely: Weld J., 1983, 62, 307s-312s.
Y.H. Wang and S. Kou: Proc. Int. Conf. on Trends in Welding Research, Gatlinburg, TN, May,
1986, pp. 65-69, Publ. ASM International.
S.A. David and J.M Vitek: Int. Mater. Rev., 1989, 34, 213-245.
K.C. Mills and BJ. Keene: Int. Mater. Rev., 1990, 35, 185-216.
R.L. UIe, Y. Joshi and E.B. Sedy: Metall. Trans., 1990, 21B, 1033-1047.
T. Zacharia, S.A. David, J.M. Vitek and H.G. Kraus: Metall. Trans., 1991, 22B, 243-257.
A. Matsunawa: Proc. 3rd Int. Conf. on Trends in Welding Research, Gatlinburg, TN, 1992,
pp.3-16, Publ. ASM International.
N. Christensen: Welding Metallurgy Compendium, 1985, University of Trondheim, The Norwegian Institute of Technology.
CE. Jackson: Weld. J., 1960, 39, 226s-230s.
O.M. Akselsen and G. Sagmo: Technical Report STF34 A89147, 1989, Trondheim (Norway),
Sintef-Division of Metallurgy.
N. Yurioka, M. Okumura, S. Ohshita and S. Saito: HW Doc. XII-E-10-81, 1981.

Appendix 1.1
Nomenclature
General symbols
thermal diffusivity (mm2 s"1)

finite difference method


heat content per unit volume at Tc
(J mm"3)

cross section (mm2)


start temperature of ferrite to
austenite transformation (0C)
end temperature of ferrite to
austenite transformation (0C)
recrystallisation temperature (0C)

enthalpy increment referred to an


initial temperature T0 (J mm' 3 )
latent heat of melting (J mm"3)
amperage (A)
modified Bessel function of second
kind and zero order

cooling rate (0C s"1)


plate thickness (mm)

modified Bessel function of second


kind and first order

natural logarithm base number

integration parameter

Gaussian error function

integration parameter

complementary Gaussian error


function

start temperature of austenite to


martensite transformation (0C)

integration parameter

y-axis/transverse direction (mm)

net power (W)

z-axis/through thickness direction


(mm)

net heat input (J)


efficiency factor
two-dimensional radius vector
(mm)

dimensionless temperature

locus of peak temperature in T-r


space (mm)

dimensionless temperature conforming to 8000C

three-dimensional radius vector


(mm)

dimensionless temperature conforming to 5000C

locus of peak temperature in T-R


space (mm)

dimensionless temperature conforming to the melting point

isothermal zone width (mm)

dimensionless peak temperature

temperature (0C)

volume heat capacity


(J mm-3 0C-1)

reference temperature (0C)


ambient temperature (0C)

thermal conductivity
(W mm"1 0C-1)

melting point (0C)

dimensionless time

peak temperature (0C)

dimensionless cooling time

time (s)

dimensionless cooling time from


800 to 5000C

time variable (s)


dimensionless cooling time from
300 to 1000C

time variable (s)


cooling time (s)

Specific symbols
0

cooling time from 800 to 500 C (s)

Local Fusion in Arc Strikes

cooling time from 300 to 100 C (s)


integration parameter
voltage (V)
integration parameter
x-axis/welding direction (mm)

dimensionless operating parameter


arc ignition time (s)
isothermal zone width (mm)
dimensionless R-vector
locus of peak temperature in G-CT1
space

dimensionless isothermal zone


width

locus of peak temperature in 0'-2


space (star denotes a specific peak
temperature)

dimensionless time
dimensionless time
locus of peak temperature in 0-T1
space

dimensionless cooling time from


^ to 5000C

dimensionless cooling time


locus of peak temperature in (T-T3
space

Spot Welding
thickness of overlapping plates
(mm)
dimensionless operating parameter

Friction Welding
integral in equation (1-30)

heating time (s)

net power generation at weld interface (W)

dimensionless r-vector

duration of heating period (s)

locus of peak temperature in 0-cr2


space

contact section temperature at the


end of heating period (0C)

dimensionless time

dimensionless temperature

locus of peak temperature in 0-T2


space

dimensionless peak temperature

dimensionless cooling time

Thermit Welding

dimensionless jc-coordinate
locus of peak temperature in 6"Q.' space
dimensionless time

half width of groove (mm)


initial temperature of liquid metal
(0C)
distance from reference point to infinitesimal source (mm)
dimensionless temperature
dimensionless peak temperature
dimensionless .^-coordinate

Arc Welding
amount of fused parent metal
(mm2)
constant in Jackson equation
amount of deposited metal (mm2)
gross heat input per unit length of
weld (kJ mm"1)
correction factor for the net heat
input during root pass welding

FCAW

flux cored arc welding

imaginary heat source of net arc


power qo, qa, or qb (W)

GMAW gas metal arc welding


GTAW

maximum intensity of distributed


(Gaussian) heat source (W mm"1)

gas tungsten arc welding


bead penetration (mm)

power density of distributed


(Gaussian) heat source (W mm"1)

integer variables.... -1,0, L...


infinitesimal heat source (W)
constant in heat distribution function (mm"2)

two-dimensional radius vector in


y-z plane (mm)

coefficient of weld metal deposition (g A"1 s"1)

locus of peak temperature in T-r*


space (mm)

amplitude of weaving or half width


of strip electrode (mm)

distance from infinitesimal heat


source to point P in x-y or y-z plane
(mm)

half width of preheated zone (mm)

position of weld end-crater with


respect to imaginary heat source at
time t** in x-y plane (mm)

equivalent amplitude of weaving


(mm)
half width of linear source in
Gaussian heat distribution model
(mm)

distance from infinitesimal heat


source to point P in x-y-z space
(mm)

length of weld bead (mm)


k

mass of weld metal (g)

distances from real and imaginary


heat sources to point P in x-y-z
space (mm)

dimensionless operating parameter


position of weld with respect to
imaginary heat source at time f**
in x-y-z space (mm)

dimensionless operating parameter


in weaving model
reference point in stationary coordinate system
reference point in moving coordinate system

SAW

submerged arc welding

SMAW

shielded metal arc welding


time at moment of arc extinction
(s)

arbitrary reference point


time constant in preheating model
arbitrary point of observation
strength of elementary heat
sources (W)

(S)

cooling time to 1000C (s)

time constant in heat distribution


function (s)

displacement of elementary heat


source in z-direction (mm)

time referred to moment of arc ignition (s)

groove angle

retention time (s)

ratio between real and pseudo-steady


state temperature (thick plate
welding)

preheating temperature (0C)


welding speed (mm s"1)
volume of isothermal enclosure
(mm3)
welding direction in stationary coordinate system (mm)
^-coordinate at maximum width of
isotherm (mm)
transverse direction in stationary
coordinate system (mm)
distance from infinitesimal heat
source to point P in j-direction
(mm)
^-coordinate at maximum width of
isotherm (mm)

ratio between real and pseudosteady state temperature (thin plate


welding)
dimensionless temperature in
preheating model
dimensionless cross sectional area
of isothermal enclosure (thick plate
welding)
dimensionless cross sectional area
of isothermal enclosure (thin plate
welding)
density of weld metal (g mm 3 )
dimensionless distance from real
and imaginary heat sources to point
P
dimensionless /^-vector

isothermal zone width (mm)


dimensionless R0-vector
through-thickness direction in stationary coordinate system (mm)
z-coordinate at maximum width of
isotherms (mm)
isothermal zone width (mm)
dimensionless y-coordinate in
weaving model
dimensionless z-coordinate in
weaving model
dimensionless plate thickness

locus of peak temperature in 0-a3


space
dimensionless r*-vector
locus of peak temperature in 0-04
space
dimensionless r-vector
dimensionless r^-vector
locus of peak temperature in 8-CT5
space

dimensionless jc-axis

dimensionless zo -axis

dimensionless xo -axis

dimensionless z-coordinate at
maximum depth of isotherm

dimensionless jc-coordinate at
maximum width of isotherm

melting efficiency factor

dimensionless length of isothermal


enclosure

dimensionless volume of isothermal enclosure

dimensionless distance from heat


source to front of isothermal enclosure

dimensionless half width of


preheated zone

dimensionless distance from heat


source to rear of isothermal enclosure

dimensionless time in weaving


model
dimensionless time in preheating
model

dimensionless y-axis
dimensionless yo -axis
dimensionless ^-coordinate at
maximum width of isotherm
dimensionless isothermal zone
width
dimensionless z-axis

locus of peak temperature in 6-x


space
dimensionless retention time
dimensionless time at moment of
arc extinction
dimensionless time referred to moment of arc ignition

Appendix 1.2
Refined Heat Flow Model for Spot Welding
The refined model is based on the assumption that all heat is released instantaneously at time
t = 0 in a point located at the interface between the two overlapping plates, which implies that
equation (1-7) is valid. However, in order to maintain the net flux of heat through both plate
surfaces equal to zero, it is necessary to account for mirror reflections of the source with
respect to the planes z = dt/2 and z = - dt/2. This can be done on the basis of the method of
images, as illustrated in Fig. A 1.1. By including all contributions from the imaginary sources
"Q-2 >Q-i ,Q\,Qi > located symmetrically at distances idt below and above the centre-axis
of the joint, the temperature distribution is obtained in the form of a convergent series:

(Al-I)
where

Z
Fig. Al.l.Refined heat flow model for spot welding of plates.
and i is an integer variable (...-1, 0, 1...)A numerical solution of equation (Al.l) gives a peak temperature distribution similar to
that shown in Fig. 1.9.

Appendix 1.3
The Gaussian Error Function
The eiTor function erf(u) and the complementary error function erfc(u) are special cases of the
incomplete gamma function. Their definitions are:

and

The functions have the following limiting values and symmetries:

and

The following Fortran subroutine can be used for calculations of the error functions with a
fractional error less than 1.2 X 10~7:
FUNCTION ERFC(U)
Z=ABS(U)
T=l./(l.+0.5*Z)
ERFC=T*EXP(-Z*Z-1.26551223+T*(1.00002368+T*(.37409196+
*

T%09678418+T*(-.18628806+T%27886807+T*(-l.13520398+

T*(1.48851587+T*(-.82215223+T*.17O87277)))))))))

IF (U.LT.O.) ERFC=2.-ERFC
RETURN
END

Appendix 1.4
Gaussian Heat Distribution
Following the treatment of Rykalin,9 the situation may be described as shown in Fig. Al.2.
Here we consider a distributed heat source of net power density (in W mm"1):
(Al-2)
The total power of the source qo is obtained by integration of equation (Al-2). Substituting
and integrating from u = - to u = +<*>, gives:

from which

(Al-3)

It follows from equation (1-73) that an infinitesimal source dqy> located between j ' a n d
will cause a small rise of temperature dTy> in point P at time t, as:

(Al-4)
where

and

Integration of equation (Al-4) between the limits y'= -o and >>'= +00 gives:

(Al-5)

q(y)

2-D heat flow

z
Fig, Al.2. Distributed heat source of net power density q(y) on a semi-infinite body.

where

is a time constant)and n = Aat.

The latter integral can be evaluated by substituting:

from which

and integrating between the limits w = - and


This gives (after some manipulation):

(Al-6)
If we replace the Gaussian heat distribution by a linear source of the same strength, which
extends from -L0 to +L0 on either side of the weld centre-line in the y-direction (see Fig. A1.3),
we may write:

By rearranging this equation, we obtain:

(Al-7)
In practice, the parameter L0 has the same physical significance as the weaving amplitude
L in equation (1-110). Consequently, these solutions are equivalent in the sense that they
predict a similar temperature-time pattern.

q(y)

Fig. Al.3. Physical representation of a Gaussian heat distribution by a linear source of width 2LO.

2
Chemical Reactions in Arc Welding

2.1 Introduction
The weld metal composition is controlled by chemical reactions occurring in the weld pool at
elevated temperatures, and is therefore influenced by the choice of welding consumables (i.e.
combination of filler metal, flux, and/or shielding gas), the base metal chemistry, as well as the
operational conditions applied. In contrast to ladle refining of metals and alloys where the
reactions occur under approximately isothermal conditions, a characteristic feature of the arc
welding process is that the chemical interactions between the liquid metal and its surroundings
(arc atmosphere, slag) take place within seconds in a small volume where the metal temperature gradients are of the order of 1000C mm"1 with corresponding cooling rates up to
1000C s"1. The complex thermal cycle experienced by the liquid metal during transfer from
the electrode tip to the weld pool in GMA welding of steel is shown schematically in Fig. 2.1.
As a result of this strong non-isothermal behaviour, it is very difficult to elucidate the
reaction sequences during all stages of the process. Consequently, a complete understanding
of the major controlling factors is still missing, which implies that fundamentally based predictions of the final weld metal chemical composition are limited. Additional problems result
from the lack of adequate thermodynamic data for the complex slag-metal reaction systems
involved. However, within these restrictions, the development of weld metal compositions
can be treated with the basic principles of thermodynamics and kinetic theory considered in
the following sections.

2.2 Overall Reaction Model


The symbols and units used throughout this chapter are defined in Appendix 2.1.
In ladle refining of metals and alloys, the reaction kinetics are usually controlled by mass
transfer between the liquid metal and its surroundings (slag or ambient atmosphere). Examples of such kinetically controlled processes are separation of non-metallic inclusions from a
deoxidised steel melt or removal of hydrogen from liquid aluminium. In welding, the reaction
pattern is more difficult to assess because of the characteristic non-isothermal behaviour of the
process (see Fig. 2.1). Nevertheless, experience shows that it is possible to analyse mass
transfer in welding analogous to that in ladle refining by considering a simple two-stage reaction model, which assumes:1
(i)

A high temperature stage, where the reactions approach a state of local


pseudo-equilibrium.

(ii)

A cooling stage, where the concentrations established during the initial stage
tend to readjust by rejection of dissolved elements from the liquid.

Gas nozzle
Shielding gas
Filler wire

Contact tube

Arc plasma
temperature~10000C

Electrode tip droplet


(1600-20000C)
Falling droplet (24000C)
Hot part of weld
pool
(1900-22000C)

Cold part of
weld pool (< 19000C)

Weld pool
retention
time 2-1Os

Base plate

Fig. 2.1. Schematic diagram showing the main process stages in GMA welding. Characteristic average
temperature ranges at each stage are indicated by values in parenthesis.

As indicated in Fig. 2.2 the high temperature stage comprises both gas/metal and slag/metal
interactions occurring at the electrode tip, in the arc plasma, or in the hot part of the weld pool,
and is characterised by extensive absorption of elements into the liquid metal. During the
subsequent stage of cooling following the passage of the arc, a supersaturation rapidly increases because of the decrease in the element solubility with decreasing temperatures. The
system will respond to this supersaturation by rejection of dissolved elements from the liquid,
either through a gas/metal reaction (desorption) or by precipitation of new phases. In the latter
case the extent of mass transfer is determined by the separation rate of the reaction products in
the weld pool. It should be noted that the boundary between the two stages is not sharp, which
means that phase separation may proceed simultaneously with absorption in the hot part of the
weld pool.
In the following sections, the chemistry of arc welding will be discussed in the light of this
two-stage reaction model.

2.3 Dissociation of Gases in the Arc Column


As shown in Table 2.1, gases such as hydrogen, nitrogen, oxygen, and carbon dioxide will be
widely dissociated in the arc column because of the high temperatures involved (the arc plasma
temperature is typically of the order of 10 0000C or higher). From a thermodynamic standpoint, dissociation can be treated as gaseous chemical reactions, where the concentrations of
the reactants are equal to their respective partial pressures. Hence, for dissociation of diatomic
gases, we may write:
(2-1)
where X denotes any gaseous species.

'Cold' part of weld pool

Solid weld metal

Solid weld metal

Peak temperature

Grey j
zonei

Rejection of
dissolved elements

Peak concentration

Solid weld metal

Concentration

Absorption
of elements

Solid weld metal

Temperature

'Hot1 part of
weld pool

Equilibrium concentration at melting point

Time
Fig. 2.2. Idealised two-stage reaction model for arc welding (schematic).
Table 2.1 Temperature for 90% dissociation of some gases in the arc column. Data from Lancaster.2
Gas

Dissociation Temperature (K)

CO 2

3800

H2

4575

O2

5100

N2

8300

Next, consider a shielding gas which consists of two components, i.e. one inert component
(argon or helium) and one active component X2. When the fraction dissociated is close to
unity, the partial pressure of species X in the gas phase px is equal to:

(2-2)
where H1 and nx are the total number of moles of components / (inert gas) and X, respectively
in the shielding gas, andptot is the total pressure (in atm).
It follows from equation (2-1) that two moles of X form from each mole of X2 that dissociates. Hence, equation (2-2) can be rewritten as:

(2-3)

where nXl is the total number of moles of component X2 which originally was present in the
shielding gas.
If nXl and H1 are proportional to the volume concentrations of the respective gas components in the shielding gas, equation (2-3) becomes:

(2-4)

Taking vol% / = (100 - vol% X2) andp,ot = 1 atm, we obtain the following expression for
Px(2-5)

Similarly, if X2 is replaced by another gas component of the type YX2, we get:


(2-6)
and
(2-7)

It is evident from the graphical representations of equations (2-5) and (2-7) in Fig. 2.3 that
the partial pressure of the dissociated component X increases monotonically with increasing
concentrations of X2 and YX2 in the shielding gas. The observed non-linear variation of px
arises from the associated change in the total number of moles of constituent species in the gas
phase due to the dissociation reaction. Moreover, it is interesting to note that the partial pressure px is also dependent on the nature of the active gas component in the arc column (i.e. the
stoichiometry of the reaction). This means that the oxidation capacity of for instance CO2 is
only half that of O2 when comparison is made on the basis of equal concentrations in the
shielding gas (to be discussed later).

Px

Vol%)^ f VoRGYX2
Fig. 2.3. Graphical representation of equations (2-5) and (2-7).

2.4 Kinetics of Gas Absorption


In general, mass transfer between a gas phase and a melt involves:3
(i)

Transport of reactants from the bulk phase to the gas/metal interface.

(ii)

Chemical reaction at the interface.

(iii)

Transport of dissolved elements from the interface to the bulk of the metal.

2.4.1 Thin film model


In cases where the rate of element absorption is controlled by a transport mechanism in the gas
phase (step one), it is a reasonable approximation to assume that all resistance to mass transfer
is confined to a stagnant layer of thickness 8 (in mm) adjacent to the metal surface, as shown in
Fig. 2.4. Under such conditions, the overall mass transfer coefficient is given by:2

(2-8)
where Dx is the diffusion coefficient of the transferring species X (in mm2 s~*).
Although the validity of equation (2-8) may be questioned, the thin film model provides a
simple physical picture of the resistance to mass transfer during gas absorption.

Partial pressure

Distance
Fig. 2.4. Film model for mass transfer (schematic).

2.4.2 Rate of element absorption


Referring to Fig. 2.5, the rate of mass transfer between the two phases (in mol s"1) can be
written as:
(2-9)
where A is the contact area (in mm 2 ), R is the universal gas constant (in mm3 atm K"1 mol"1),
T is the absolute temperature (in K), px is the partial pressure of the dissociated species X in the
bulk phase (in atm), and px is the equilibrium partial pressure of the same species at the gas/
metal interface (in atm).
Based on equation (2-9) it is possible to calculate the transient concentration of element X
in the hot part of the weld pool. Let m denote the total mass of liquid weld metal entering/
leaving the reaction zone per unit time (in g s"1). If Mx represents the atomic weight of the
element (in g mol"1), we obtain the following relation w h e n / ? x px:

(2-10)

It follows from equation (2-10) that the transient concentration of element X in the hot part
of the weld pool is proportional to the partial pressure of the dissociated component X in the
plasma gas. Since this partial pressure is related to the initial content of the molecular species
X2 or YX2 in the shielding gas through equations (2-5) and (2-7), we may write:

Arc column

Bulk gas
phase

Stagnant gaseous
boundary layer
Gas/metal interface
Metal phase
Hot part of
weld pool
Fig. 2.5. Idealised kinetic model for gas absorption in arc welding (schematic).
(2-11)
and
(2-12)

where C1 and C2 are kinetic constants which are characteristic of the reaction systems under
consideration.

2.5 The Concept of Pseudo-Equilibrium


Although the above analysis presupposes that the element absorption is controlled by a
transport mechanism in the gas phase, the transient concentration of the active component X in
the hot part of the weld pool can alternatively be calculated from chemical thermodynamics by
considering the following reaction:
X(gas)

X (dissolved)

(2-13)

By introducing the equilibrium constant K{ for the reaction and setting the activity coefficient to unity, we get:
(2-14)
This equation should be compared with equation (2-10) which predicts a linear relationship

between wt% X and px. If the above analysis is correct, one would expect that the partial
pressure px at the gas/metal interface is directly proportional to the partial pressure of the
dissociated component in the bulk phase. Unfortunately, the proportionality constant is difficult to establish in practice.

2.6 Kinetics of Gas Desorption


During the subsequent stage of cooling following the passage of the arc, the concentrations
established at elevated temperatures will tend to readjust by rejection of dissolved elements
from the liquid. When it comes to gases such as hydrogen and nitrogen, this occurs through a
desorption mechanism, where the driving force for the reaction is provided by the decrease in
the element solubility with decreasing metal temperatures.
2.6.1 Rate of element desorption
Consider a melt which first is brought in equilibrium with a monoatomic gas of partial pressure px at a high temperature T1, and then is rapidly cooled to a lower temperature T2 and
immediately brought in contact with diatomic X2 of partial pressure pXl (see Fig. 2.6). Under
such conditions, the rate of element desorption (in mol s"1) is given by:

(2-15)

where k'd is the mass transfer coefficient (in mm s 1X and px is the equilibrium partial pressure of component X2 at the gas/metal interface (in atm).

Bulk gas
phase

Stagnant gaseous
boundary layer
Gas/metal interface
Metal phase
Cold part of
weld pool
Fig. 2.6. Idealised kinetic model for gas desorption in arc welding (schematic).

The partial pressure pX2 can be calculated from chemical thermodynamics by considering
the following reaction:
2X(dissolved) = X2 (gas)
(2-16)
from which
(2-17)
where K2 is the equilibrium constant, and [wt% X] is the concentration of element X in the
liquid metal (in weight percent). Note that the activity coefficient has been set to unity in the
derivation of equation (2-17).
The equilibrium constant K2 may be expressed in terms of the solubility of element X in the
liquid metal at 1 atm total pressure Sx. Hence, equation (2-17) transforms to:

(2-18)

By combining equations (2-15) and (2-18), we get:

(2-19)
Data for the solubility of hydrogen and nitrogen in some metals up to about 22000C are
given in Figs. 2.7 and 2.8, respectively. It is evident that the element solubility decreases
steadily with decreasing metal temperatures down to the melting point. This implies that the
desorption reaction is thermodynamically favoured by the thermal conditions existing in the
cold part of the weld pool.
2.6.2. Sievert's law
It follows from equation (2-19) that desorption becomes kinetically unfeasible when
Px2 ~ Px2' corresponding to:
(2-20)
Equation (2-20) is known as the Sievert's law. This relation provides a basis for calculating
the final weld metal composition in cases where the resistance to mass transfer is sufficiently
small to maintain full chemical equilibrium between the liquid metal and the ambient (bulk)
gas phase.

2.7 Overall Kinetic Model for Mass Transfer during Cooling in the Weld Pool
Because of the complexity of the rate phenomena involved, it would be a formidable task to
derive a complete kinetic model for mass transfer in arc welding from first principles. How-

(b)

Aluminium

ml H2/100 g fused metal

ml H2/100 g fused metal

(a)

Temperature, 0C

Solid Cu

Temperature, 0C

Iron

Temperature, 0C

ml H2/100g fused metal

(d)

(C)

ml H2/100g fused metal

Copper

Nickel

Temperature, 0C

Fig. 2.7. Solubility of hydrogen in some metals; (a) Aluminium, (b) Copper, (c) Iron, (d) Nickel. Data
compiled by Christensen.4
ever, for the idealised system considered in Fig. 2.9, it is possible to develop a simple mathematical relation which provides quantitative information about the extent of element transfer
occurring during cooling in the weld pool. Let [%X]eq denote the equilibrium concentration of
element X in the melt. If we assume that the net flux of element X passing through the phase
boundary A per unit time is proportional to the difference ([%X] - [%X]eqX the following
balance is obtained:3
(2-21)
where V is the volume of the melt (in mm 3 ), kd is the overall mass transfer coefficient (in
mm s"1), and A is the contact area between the two phases (in mm 2 ).

Temperature, 0C

log (wt% N)

Iron

104AT1 K
Fig. 2.8. Solubility of nitrogen in iron. Data from Turkdogan.5

Phase I i

Phase i

Distance

Net flux of X

Contact area (A)

Volume (V)

Concentration
Fig. 2.9. Idealised kinetic model for mass transfer in arc welding (schematic).
By rearranging equation (2-21) and integrating between the limife [%X]( (att = O) and [%X]
(at an arbitrary time t\ we get:

(2-22)

where to is a time constant (equal to VI kjA).

It is evident from the graphical representation of equation (2-22) in Fig. 2.10 that the rate of
mass transfer depends on the ratio Vl kji, i.e. the time required to reduce the concentration of
element X to a certain level is inversely proportional to the mass transfer coefficient kd. This
type of response is typical of a first order kinetic reaction.
Although the above model refers to mass transfer under isothermal conditions, it is also
applicable to welding if we assume that the weld cooling cycle can be replaced by an equivalent isothermal hold-up at a chosen reference temperature. Thus, by rearranging equation (222), we get:
(2-23)
It follows that the final concentration of element X in the weld metal depends both on the
cooling conditions and on the intrinsic resistance to mass transfer, combined in the ratio t/to.
When [%X]eq is sufficiently small, equation (2-23) predicts a direct proportionality between
[%X] and [%X\t (i.e. the initial concentration of element X in the weld pool). This will be the
case during deoxidation of steel weld metals where separation of oxide inclusions from the
weld pool is the rate controlling step. Moreover, when t/t0 1 (small resistance to mass
transfer), equation (2-23) reduces to:
(2-24)

(X-X^)Z(X1-Xeq)

Under such conditions the final weld metal composition can be calculated from simple
chemical thermodynamics.
Because of this flexibility, equation (2-23) is applicable to a wide range of metallurgical
problems at the same time as it provides a simple physical picture of the resistance to mass
transfer during cooling in the weld pool.

t,s
Fig. 2.10. Graphical representation of equation (2-22).

2.8 Absorption of Hydrogen


Some of the well-known harmful effects of hydrogen discussed in Chapters 3 and 7 (i.e. weld
porosity and HAZ cold cracking) are closely related to the local concentration of hydrogen
established in the weld pool at elevated temperatures due to chemical interactions between
the liquid metal and its surroundings.
2.8.1 Sources of hydrogen
Broadly speaking, the principal sources of hydrogen in welding consumables are:6
(i) Loosely bound moisture in the coating of shielded metal arc (SMA) electrodes and in
the flux used in submerged arc (SA) or flux-cored arc (FCA) welding. Occasionally, moisture
may also be introduced through the shielding gas in gas metal arc (GMA) and gas tungsten arc
(GTA) welding.
(ii) Firmly bound water in the electrode coating or the welding flux. This can be in the
form of hydrated oxides (e.g. rust on the surface of electrode wires and iron powder), hydrocarbons (in cellulose), or crystal water (bound in clay, astbestos, binder etc.).
(iii) Oil, dirt and grease, either on the surface of the work piece itself, or trapped in the
surface layers of welding wires and electrode cored wires.
It is evident from Fig. 2.11 that the weld metal hydrogen content may vary strongly from
one process to another. The lowest hydrogen levels are usually obtained with the use of lowmoisture basic electrodes or GMA welding with solid wires. Submerged arc welding and fluxcored arc welding, on the other hand, may give high or low concentrations of hydrogen in the
weld metal, depending on the flux quality and the operational conditions applied (note that the
former process is not included in Fig. 2.11). The highest hydrogen levels are normally associated with cellulosic, acid, and rutile type electrodes. This is due to the presence of large
amounts of asbestos, clay and other hydrogen-containing compounds in the electrode coating.
Table 2.2 (shown on page 132) gives a summary of measured arc atmosphere compositions
in GMA and SMA welding. Included are also typical ranges for the weld metal hydrogen
content.
2.8.2 Methods of hydrogen determination in steel welds
Hydrogen is unlike other elements in weld metal in that it diffuses rapidly at normal room
temperatures, and hence, some of it may be lost before an analysis can be made. This, coupled
with the fact that the concentrations to be measured are usually at the parts per million level,
means that special sampling and analysis procedures are needed. In order that research results
may be compared between different laboratories and can be used to develop hydrogen control
procedures, some international standardisation of these sampling and analysis methods is necessary.
Three methods are currently being used, as defined in the following standards:

Potential hydrogen level

FCAW

Very Low
Medium
low
Weld hydrogen level

High

Fig. 2.11. Ranking of different welding processes in terms of hydrogen level (schematic). The diagram is
based on the ideas of Coe.6
(i)
The Japanese method (JIS Z 313-1975), which has been adopted with important adjustments from the former ASTM designation A316-48T. This method involves collection of
released hydrogen from a single pass weld above glycerine for 48h at 45 0 C. The total volume
of hydrogen is reported in ml per 10Og deposit. Only 5 s of delay are allowed from extinction
of the arc to quenching.
(ii) The French method (N.F.A. 81-305-1975) where two beads are deposited onto core
wires placed in a copper mould. Hydrogen released from this bead is collected above mercury,
and the volume is reported in ml per 10Og fused metal (including the fused core wire metal).
(iii) The International Institute of Welding (HW) method (ISO 3690-1977), where a single
bead is deposited on previously degassed and weighed mild steel blocks clamped in a quickrelease copper fixture. The weldment is quenched and refrigerated according to a rigorously
specified time schedule. Hydrogen released from the specimens is collected above mercury
for 72 h at 25C, and the results are reported in ml per 10Og deposit, or in g per ton fused metal.
To avoid confusion, it is recommended to use the symbol HDM for the content reported in terms
of deposited metal (ml per 10Og deposit), and HFM for the content referred to fused metal (ml
per 100 g or g per ton fused metal). The relationship between HDM and HFM is shown in Fig.
2.12.
As would be expected, these three methods do not give identical results when applied to a
given electrode. Approximate correlations have been established between the HW criteria
HDM and HFM and the numbers obtained by the Japanese and the French methods (designated
HJIS and HFR, respectively). For covered electrodes tested at various hydrogen levels, we
have:7

Fig. 2.12. The relation between HDM and HFM (0.9 is the conversion factor from ml per 10Og to g per ton).
(2-25)
(2-26)
The conversion factor from HFR to HFM applies to a ratio of deposited to fused metal,
DI(B + D), equal to 0.6, which is a reasonable average for basic electrodes.
The use of HFM in preference of HDM is normally recommended, because it is a more rational criterion of concentration. Moreover, HDM values would be grossly unfair, if applied to
high penetration processes like submerged arc welding. In GTA welds made without filler
wire HDM cannot be used at all, since there is no deposit.
It should be noted that the present HW procedure gives the amount of 'diffusible hydrogen'.
For certain purposes the total hydrogen content may be wanted. It is obtained by adding the
content of 'residual hydrogen' determined on the same samples by vacuum or carrier gas
extraction at 6500C. A very small additional amount may be observed on vacuum fusion of the
sample, tentatively labelled 'fixed hydrogen'. There is no clear line of demarcation between
these categories of hydrogen. As will be discussed later, the extent of hydrogen trapping
depends both on the weld metal constitution and the thermal history of the metal. In singlebead basic electrode deposits the diffusible fraction is usually well above 90%.
2.8.3 Reaction model
Normally, measurements of hydrogen in weld metals are carried out on samples from solidified beads. Due to the rapid migration of hydrogen at elevated temperatures, such data do not
represent the conditions in the hot part of the weld pool. Quenched end crater samples would
be better in this respect, but they are not representative of normal welding. Further complications arise from the presence of hydrogen in different states (e.g. diffusible or residual hydrogen) and the lack of consistent sampling methods.
Nevertheless, experience has shown that pick-up of hydrogen in arc welding can be interpreted on the basis of the simple model outlined in Fig. 2.13. According to this model, two
zones are considered:
(i) An inner zone of very high temperatures which is characterised by absorption of atomic
hydrogen from the surrounding arc atmosphere.

Electrode
Hot part of weld pool
Absorption of atomic hydrogen
(controlled by pH in the arc column)

Cold part of weld pool


Desorption of hydrogen
(controlled by pH2 in ambient gas phase)

Hydrogen
trapped in
weld metal
Weld pool

Fig. 2.13. Idealised reaction model for hydrogen pick-up in arc welding.
(ii) An outer zone of lower temperatures where the resistance to hydrogen desorption is
sufficiently small to maintain full chemical equilibrium between the liquid weld metal and the
ambient (bulk) gas phase.
Under such conditions, the final weld metal hydrogen content should be proportional to the
square root of the initial partial pressure of diatomic hydrogen in the shielding gas, in agreement with Sievert's law (equation (2-20)).
2.8.4 Comparison between measured and predicted hydrogen contents
It is evident from the data in Table 2.2 that the reported ranges for hydrogen contents in steel
weld metals are quite wide, and therefore not suitable for a direct comparison of prediction
with measurement. For such purposes, the welding conditions and consumables must be more
precisely defined.
2.8.4.1 Gas-shielded welding
In GTA and GMA welds the hydrogen content is usually too low to make a direct comparison
between theory and experiments. An exception is welding under controlled laboratory conditions where the hydrogen content in the shielding gas can be varied within relatively wide
limits. The results from such experiments are summarised in Fig. 2.14, from which it is seen
that Sievert's law indeed is valid. A closer inspection of the data reveals that the weld metal
hydrogen content falls within the range calculated for chemical equilibrium at 1550 and 20000C,
depending on the applied welding current. This shows that the effective reaction temperature
is sensitive to variations in the operational conditions.
An interesting effect of oxygen on the weld metal hydrogen content has been reported by
Matsuda et al.9 Their data are reproduced in Fig. 2.15. It is evident that the hydrogen level is
significantly higher in the presence of oxygen. This is probably due to the formation of a thin
(protective) layer of slag on the top of the bead, which kinetically suppresses the desorption of
hydrogen during cooling.

Table 2.2 Measured arc atmosphere compositions in steel welding. Also included are typical ranges for
the weld metal hydrogen content. Data compiled by Christensen.4
Arc Atmosphere Composition
(vol%)
Method

Primary Source
of Hydrogen

Weld Metal
Hydrogen
Content (ppm)

CO2

CO

H 2 +H 2 O

Range

Average

98-80

2-20

<0.02

1-5

GMAW*
(CO2)

Moisture introduced
through the shielding gas

SMAW
(acid)

Firmly bound water in


the electrode coating

-4

-34

-62

10-30

25

SMAW
(rutile)

Firmly bound water in


the electrode coating

~4

-42

-54

10-30

25

SMAW
(basic)

Loosely bound water in


the electrode coating

-19

-77

-4

2-10

3-5

FCAW
(rutile)

Firmly bound water


influx

10-20

FCAW
(basic)

Loosely bound water


influx

2-5

SAW
(basic)

Loosely bound water


influx

2-10

*The arc atmosphere composition can vary within wide limits, depending on the operational conditions applied.

ml H2/100 g fused metal, HpM

GTAW (low-alloy steel)

ml H2 /100 g fused metal, HJ|S

GTAW (low-alloy steel)


Welding conditions: 300A-18V-2.5 mm/s

Weld metal oxygen content, wt%


Fig. 2.15. Hydrogen pick-up in GTA welding at different levels of oxygen in the weld metal. Data from
Matsuda et al.9
Example (2.1)

Consider GTA welding (Ar-shielding) on a thick plate of low-alloy steel under the following
conditions:
/ = 200A, U = 15V, v = 3 mm s"1, TI = 0.5, T0 = 20C
The shielding gas contains 0.1 vol% moisture (H2O) and is supplied at a rate of 15Nl mhr 1 .
Calculate the 'potential' hydrogen level, assuming that all hydrogen introduced through the
shielding gas is absorbed in the weld metal.
Solution

First we calculate the total mass of hydrogen per mm:

The resulting bead cross section and total mass of weld metal per mm can be estimated
from the Rosenthal equation by considering the dimensionless operating parameter at the melting
point (equation (1-50)):

Reading from Fig. 1.21 gives:

Taking the density of the steel equal to 7.85 X 10 3 g mm 3, we obtain:

The 'potential' hydrogen level is thus:

It is evident from the above calculations that the 'potential' hydrogen level is at least one
order of magnitude higher than the expected weld metal hydrogen content (1 to 3 ppm). This
shows that the hydrogen pick-up in GTA welding is not determined by the total amount of
hydrogen which is introduced through the shielding gas, but is mainly controlled by the resulting partial pressure of hydrogen in the ambient (bulk) gas phase.
2.8.4.2 Covered electrodes
In SMA welding the partial pressure of hydrogen is more difficult to assess due to the presence
of trapped moisture and hydrogen-containing compounds in the electrode coating. Such compounds will loose their identity at the stage of introduction into the arc atmosphere. Since very
little information is available on the species present in the arc column, we shall base our estimate on a simple thermodynamic approach, including only the molecular species H2 and H2O
which can be determined by analysis (see data in Table 2.2). It follows that the combined
partial pressure of H2 and H2O in the gas phase is given by:
(2-27)
The parameter pw can be estimated on the basis of combustion measurements of the electrode coating, assuming that no carbon is picked up or lost from the system in excess of the
amount calculated from an analysis of the base plate and the electrode wire. For a recorded
content of mw g H2O and mc g CO2 per 100 g of electrode coating, we obtain:
(2-28)
From a thermodynamic standpoint, replacement of pHl b y / ^ in the expression for Sievert's
law requires the use of a modified solubility of hydrogen, defined as:

(2-29)

where K3 is the equilibrium constant for the H 2 O-H reaction, and [%O] is the weld metal
oxygen content. In practice, the correction term ^ K3/(K3+[%0]) does not depart significantly from unity, which means that Sw ~ SH.

During welding with basic covered electrodes considerable amounts of CO2 may form as a
result of decomposition of calcium carbonate, according to the reaction:
(2-30)
Modern basic electrodes contain between 20 to 40 weight percent CaCO3, which is equivalent with a CO2 content of 9 to 18 percent. Taking as an average mc equal to 15 g CO2 per 100 g
electrode coating, we obtain:
(2-31)

In Fig. 2.16 the validity of equation (2-31) has been checked against relevant literature data
(compiled by Chew10). A closer inspection of the data reveals that the weld metal hydrogen
content falls within the range calculated for chemical equilibrium at 1520 to 2000C, taking Sw
equal to the solubility of hydrogen in pure iron at the indicated temperatures (i.e. 27 and 40 ml
H2 per 100 g fused metal, respectively). Although the observed scatter in the effective reaction
temperature is admittedly large, equation (2-31) points out a very interesting effect, namely
that the hydrogen content of SMA steel weld metals is controlled by the combined partial
pressure of H2 and H2O in the ambient gas phase. For this reason it is frequently recommended that calcium carbonate is added to the electrode coating, which on decomposition
produces considerable amounts of shielding gas in the form of CO2. Hydrogen shielding can
also be achieved by additions of volatile alkali-fluorides, which on heating will evaporate and
dilute the atmosphere with respect to hydrogen.

ml H2/100 g fused metal, HpM

SMAW (low-alloy steel)

Water content in electrode coating, wt%


Fig. 2.16. Hydrogen pick-up in SMA welding at different water contents in the electrode coating. Data
compiled by Chew.10

Example (2.2)

Consider SMA welding on mild steel with basic covered electrodes. The electrode coating
contains 35 wt% CaCO3 and 0.5 wt% H2O in the as-received condition. After drying at 3500C
for 1 h the water content is reduced to 0.2 wt% H2O. Estimate the weld metal hydrogen
content (in ppm) both before and after drying of the electrode. Assume in these calculations an
effective reaction temperature of 18000C.
Solution

First we calculate the CO2 content per 100 g of electrode coating. Taking the atomic weight of
CaCO3 and CO2 equal to 100.1 and 40.0, respectively, we obtain:

The combined partial pressure pw can now be estimated from equation (2-28). Before
drying we have:

After drying of the electrode, the partial pressure pw becomes:

From Fig. 2.7(c) it is evident that the solubility of hydrogen in liquid iron at 18000C is
about 37 ml H2 per 100 g fused metal. This corresponds to a modified solubility Sw (in ppm)
of:

Substituting this value into the expression for Sievert's law gives:
(before drying)
(after drying)
It follows from the above calculations that a low weld metal hydrogen level requires the use
of 'dry' basic electrodes. In practice, this can be achieved by protecting the electrodes against
moisture pick-up during storage (see Fig. 2.17). However, in certain cases it is necessary to
differentiate between strongly bound and loosely adsorbed moisture in the coating of basic
electrodes. This point is more clearly illustrated in Fig. 2.18, which shows the HDM content of
hydrogen in basic electrode deposits at various levels of coating moisture. It is seen that water
remaining from an insufficient baking treatment is more dangerous than moisture picked up by
exposure of a properly dried coating. This has to do with the fact that loosely adsorbed mois-

Water content in electrode coating, wt%

Exposure time, days

very low

low

medium

ml H2 /100 deposit, HDM

SMAW (low-alloy steel)

high

Fig. 2.17. Moisture content in basic electrode coating as a function of exposure time and relative humidity (R.H.) in ambient gas phase. Data from Evans.11

Water content, wt%


Fig. 2.18. Hydrogen pick-up in SMA welding at different levels (states) of adsorbed water in the electrode coating. Data from Evans and Bach.12

ture will tend to evaporate during the welding operation (before it enters the arc column)
because of resistance heating of the electrode, a process which is not feasible when the water is
bound in rust on the surface of the electrode wire or the iron powder.
2.8.4.3 Submerged arc welding
This method is usually classified as a pure slag-shielded process, because carbonates or other
gas-producing compounds are not present in large quantities. A closed arc cavity does exist,
however, as indicated by the falling volt-ampere curve characteristic of open arcs, and by
observations made by probes inserted through the flux cover.
It is reasonable to assume that the gas contained within this enclosure consists of metal
vapour, volatile constituents originating from the flux, and relatively small fractions of carbon
monoxide and water vapour. Acid fluxes of the calcium silicate type will probably generate
silicon monoxide, while agglomerated fluxes bonded with alkali silicate will produce volatile
alkali fluorides. In addition, carbon monoxide may be present as a result of oxidation of carbon, or decomposition of carbonates.
A small but important contribution to the cavity atmosphere is the trace of moisture remaining in the flux even after careful drying. No direct measurements of partial pressures are available, and the gas composition must therefore be inferred from observations of hydrogen absorption in the weld metal. Hydrogen pick-up during SA welding has been examined by Evans
and Bach.12 Their data are replotted in Fig. 2.19. The shape of the observed curve of hydrogen
vs residual water content would seem to indicate a relationship similar to that predicted by
Sievert's law. In fact, a very close fit can be obtained through empirical calibration of the
dilution term in equation (2-28). This, however, implies unreasonable amounts of CaCO3.
Carbon monoxide in addition to that delivered by carbonates could be formed by oxidation of
carbon. Again, an unreasonable amount of carbon loss would be required. Therefore, it must
be concluded that further research is needed for a proper interpretation of the factors controlling hydrogen pick-up in SA welding.

ml H2/100 deposit, HDM

Hydrogen content, HFM (ppm)

SAW (low-alloy steel)

Water content, wt%


Fig. 2.19. Hydrogen pick-up in SA welding at different water contents in the flux. Data from Evans and
Bach.12

Example (2.3)

Consider SA welding on a thick plate of low-alloy steel under the following conditions:

The flux contains 0.04 wt% H2O and is consumed at a rate of 0.6 g per g weld deposit.
Estimate both 'potential' and 'equilibrium' hydrogen levels when the total oxidation loss of
carbon in the weld pool is 0.03 wt%.
Solution

First we calculate the total amount of fused parent metal and weld deposit formed on welding.
From equations (1-75) and (1-120), we have:

and

When the dilution ratio DI(B + D) is known, it is possible to calculate the total flux consumption per gram fused weld metal:

The 'potential' hydrogen level is thus:

If we assume that all CO produced by reactions between dissolved carbon and oxygen is
infiltrated in the arc column, the following balance is obtained:
Total number of moles of CO per g fused weld metal:

Total number of moles of H2O per g fused weld metal:

This gives:

Since the effective reaction temperature of hydrogen absorption in SA welding is not known,
the maximum solubility of hydrogen at 1 atm total pressure is taken equal to 33 ppm, similar to
that in the previous example. By inserting this value in the expression for Sievert's law, we
obtain:

In practice, the 'potential' hydrogen level represents an upper limit for the hydrogen concentration which cannot be exceeded. Thus, the contradictory results obtained in the present
example clearly illustrate the difficulties involved in estimating the effective partial pressure
of hydrogen in SA welding.
2.8.4.4 Implications of Sievert's law
An important implication of Sievert's law is that the fraction of hydrogen picked up from the
arc atmosphere is very high at low hydrogen pressures:
(2-32)
As seen from equation (2-32), the first traces of hydrogen added to the atmosphere are
completely absorbed in the metal. At increasing partial pressures the fraction of hydrogen
picked up in the metal will gradually decrease, finally attaining a threshold of (SH/2) in the
case of pure H2. This shows that the concept of 'potential' hydrogen content frequently used to
characterise filler materials (see Fig. 2.11) is a dangerous one, since the rates of absorption are
so different in the high and low ranges of the hydrogen potential.
2.8.4.5 Hydrogen in multi-run weldments
So far, no standardised method is available for the determination of hydrogen in multi-layer
welds. Early measurements by Roux,13 using an arrangement similar to that subsequently
adopted in French standards, indicate a constant ratio of extracted hydrogen to the mass of
fused metal, regardless of the number of passes. If hydrogen is reported on the basis of deposited metal, this ratio may vary by a factor of 2.5 when comparing a deposit made in five passes
to a single bead.
Exploratory measurements of local hydrogen contents in large-size joints have been made
by Skjolberg,14 who butt welded a 40 mm plate with a self-shielding flux cored wire at an
interpass temperature of 2000C. Samples were cut from a refrigerated part of the weldment at
mid-thickness, including positions in the weld metal close to the fusion line and samples in the
HAZ. His results are summarised in Table 2.3.
Normal testing of the filler wire according to ISO 3690 gave fused metal hydrogen contents
of 3.3 ppm (diffusible) and 1.7 ppm (residual). A comparison with Table 2.3 shows that the
multi-run content of diffusible hydrogen is much lower than the corresponding ISO value,
probably as a result of a high interpass temperature which facilitates loss of hydrogen to the
surroundings through diffusion.

Table 2.3 Measured hydrogen contents in multi-run FCA steel weldment. Data from Skjolberg.14
HAZ
Distance from fusion line (mm)
Condition

Weld Metal

As-welded
PWHT*
(4h/150C)

Oto 5

5 to 10

0.6 ppm diffusible


0.9 ppm residual

0.25 ppm
diffusible

0.15 ppm
diffusible

0.35 ppm diffusible


2.25 ppm residual

0.15 ppm
diffusible

0.15 ppm
diffusible

10 to 15

0.10 ppm
diffusible

*Post weld heat treated.

2.8.4.6 Hydrogen in non-ferrous weldments


The solubility of hydrogen in metals and alloys of industrial importance increases with temperature, and passes through a maximum in the vicinity of the boiling point, where the opposing trends of increasing solubility and increasing dilution by metal vapour balance. Solubility
curves for hydrogen in aluminium, copper, and nickel up to about 22000C have previously
been presented in Fig. 2.7.
Since all these metals can dissolve considerable amounts of hydrogen, the risk of hydrogen
absorption during welding is imminent if moisture is present in the shielding gas. Results
obtained from arc melting experiments with Cu, Al, Ni in Ar-H 2 gas atmospheres indicate that
hydrogen is absorbed at a high temperature zone under the arc and is transported by fluid flow
to the outer, cooler regions of the pool.15 Rejection of the gas in the supersaturated outer
regions is slower than the absorption in the hot zone, so the gas content throughout the pool
approximates to that in the absorption zone. Typical estimates of the effective reaction temperature of hydrogen desorption (based on the Sievert's law) gave the following result:15
Copper:
16500C
Aluminium: 19000C
Nickel:
19000C
At present, it is not known whether these reaction temperatures also apply to conventional
GTA or GMA welding of the same materials or are mainly restricted to the operational conditions employed in the arc melting experiments.

2.9 Absorption of Nitrogen


It is generally recognised that interstitial nitrogen embrittles steel (e.g. see discussion in Chapter 7). In steel weld metals the associated loss of toughness due to free nitrogen has been
attributed to solid solution hardening and dislocation locking effects. In addition, excessive
nitrogen pick-up can cause porosity in steel weldments because of gas evolution during solidification.

2.9.7 Sources of nitrogen


Since the total nitrogen level in most welding consumables and shielding gases is quite low,
the main source of nitrogen contamination is air infiltrated in the arc column. For this reason,
the weld metal nitrogen content is very sensitive to variations in the operational conditions
(e.g. arc length, electrode stick-out, shielding gas flow rate etc.). The overall reaction of nitrogen absorption is similar to that of hydrogen:
(dissolved)

(2-33)

By introducing the equilibrium constant K4 for the reaction, we get:

(2-34)

where SN is the maximum solubility of nitrogen at 1 atm total pressure,/^ is the activity coefficient, and pNl is the resulting partial pressure of diatomic nitrogen in the gas phase.
The solubility of nitrogen in liquid iron is approximately given by:
(2-35)
where T is the temperature in K.
At 1600 and 20000C, this equation gives equilibrium concentrations of 446 and 465 ppm,
respectively. In alloyed steel containing large amounts of nitride-forming elements (e.g.
austenitic stainless steel), the activity coefficient of nitrogen fN is about 1/4 and hence, the
solubility will be about 4 times higher than that calculated from equation (2-35).
From a primitive model of pseudo-equilibrium between gaseous N2 and dissolved N a maximum solubility of about 465 ppm would be expected in welding under 1 atm total pressure.
Thus, the maximum pick-up of nitrogen in deposition of bare wire in air would be of the order
of 465A/OT8 ppm or 416 ppm. If a tentative estimate of air infiltration in the arc column is
made at 1 vol% N2, the expected pick-up of nitrogen would be 465 VoToT or about 47 ppm.
A comparison with the data in Table 2.4 shows that the measured weld metal nitrogen
contents are much higher than predicted from Sievert's law. This implies that the mechanism
of nitrogen desorption is different from that of hydrogen.
2.9.2 Gas-shielded welding
Information on the factors controlling nitrogen pick-up may be obtained from the work of
Kobayashi et a/.,16 who examined the GMA welding process in a systematic manner. Some of
their results are shown in Fig. 2.20.
Figure 2.20(a), for low-alloy steel, reveals that the square root relationship is a fair approximation only for welding in mixtures of N 2 and H 2 (curve No. 5). Mixtures of N2 + Ar (curve
No. 3), N 2 + CO2 (curve No. 2) and N2 + O 2 (curve No. 1) show increasing deviation from the
predicted behaviour. Pure N2 under reduced pressure gives a curve (No. 4) of an entirely different shape including a maximum at pN ~ 0.05.

Table 2.4 Summary of measured weld metal nitrogen contents. Data compiled by Christensen.4
Welding Method

Material

SMAW (basic electrodes)

Low-alloy steel

60-180

Stainless steel

550-650

Low-alloy steel

200-350

Stainless steel

600-750

SAW

Low-alloy steel

40-140

FCAW

Low-alloy steel

125-275

GMAW

Low-alloy steel

50-200

SMAW (rutile electrodes)

Nitrogen Content (ppm)

Similar features are seen from Fig. 2.20(b) for welding of stainless steel. Again, the deviation becomes more pronounced as the oxidation potential of the gas mixture is increased in the
sequence H 2 -Ar-CO 2 -O 2 . Moreover, a comparison with Fig. 2.20(a) reveals that the displacement of the nitrogen concentrations in the presence of chromium is larger than expected from
the calculated reduction of the nitrogen activity coefficient.
The trends observed in Fig. 2.20 have been confirmed by O'Brien and Jordan17 who studied
nitrogen pick-up during CO2-shielded welding of low-alloy steel. As can be seen from Fig.
2.21 (a) their curves are similar to those of Kobayashi et al.16 for short circuiting metal transfer,
while a mixed spray/globular transfer gives a sharp rise of nitrogen absorption up to pNi = 0.3
followed by a constant or slightly decreasing concentration (Fig. 2.21(b)). Both patterns are
clearly not in accordance with predictions based on Sievert's law (equation (2-34)).
An interpretation of the observed trends should be made with a view to absorption of hydrogen, where the concept of pseudo-equilibrium has proved useful for a semiquantitative
prediction. In both cases the molecular species H2 and N 2 are known to dissociate in the arc
column (see Table 2.1), and would therefore dissolve in the metal to an extent far beyond the
solubility controlled by pH or PN . The excess of dissolved hydrogen is probably released as
gas at weld pool temperatures. This will also be the case with nitrogen in the absence of oxygen, as shown previously in Fig. 2.20(a) and (b). However, under oxidising conditions the
desorption of gaseous nitrogen becomes suppressed by the presence of oxygen at the gas/metal
interface and hence, nitrogen is retained at a level which by far exceeds the solubility limit at
1 atm total pressure of N2. This has been confirmed experimentally by Uda and Ohno18 in their
classic work on surface active elements (i.e. oxygen, sulphur and selenium) in liquid steel. A
similar phenomenon was quoted in Section 2.8.4.1 from the work of Matsuda et al9 even in
the case of hydrogen, where increased entrapment of hydrogen was observed in the presence
of oxygen (see Fig. 2.15).
It appears thus that excessive absorption of nitrogen (and in some cases also hydrogen)
should be interpreted as a state of incomplete release of solute, as described previously in
Sections 2.6 and 2.7. As a consequence, Sievert's law cannot be used for an estimate of nitrogen pick-up in steel welding, unless the weld metal oxygen content is extremely low.
2.9.3 Covered electrodes
The nitrogen content of SMA weld deposits is known to be sensitive to variations in the arc

(a)

Nitrogen content, wt%

GMAW (low-alloy steel)

Vol% N 2 in shielding gas

(b)

Nitrogen content, wt%

GMAW (stainless steel)

Vol% N 2 in shielding gas


Fig. 2.20. Nitrogen pick-up in GMA welding at different concentrations of N2 in the shielding gas;
(a) Low-alloy steel, (b) Stainless steel. Data from Kobayashi et al.16

(a)

Nitrogen content, wt%

Low-alloy steel

Experiment

Vol% N 2 in shielding gas


(b)

Nitrogen content, wt%

Low-alloy steel

Experiment

Vol% N2in shielding gas

Fig. 2.21. Nitrogen pick-up in GMA welding at different concentrations of N2 in the


shielding gas; (a) Short circuting metal
transfer, (b) Mixed and free flight metal
transfer. Data from O'Brien and Jordan.17

length (voltage) because of the risk of air infiltration in the arc column. This point is more
clearly illustrated in Fig. 2.22, which shows that the resulting weld metal nitrogen level may
vary significantly from one weld to another, depending on the operational conditions applied.
Consequently, the use of long arcs in SMAW should be avoided in order to prevent excessive
pick-up of nitrogen from the surrounding atmosphere.
2.9.4 Submerged arc welding
In submerged arc welding the risk of air infiltration in the arc column is less imminent, since
welding is performed under the shield of a flux. Hence, in multipass welds the filler wire itself
will be the main source of nitrogen (see Fig. 2.23), while in single pass weldments the base
plate nitrogen content is more important because of the high dilution involved. The latter point
is illustrated by the following numerical example.
Example (2.4)

Consider SA (single pass) welding on a thick plate of low-alloy steel under the following
conditions:

Based on the 'rule of mixtures', calculate the weld metal nitrogen content. Assume in these
calculations that the nitrogen content of the base plate and the filler wire is 0.005 and 0.012
wt%, respectively.

Nitrogen content, ppm

SMAW (low-alloy steel)


4 and 5 mm basic covered electrodes

Welder No.
Fig. 2.22. Natural fluctuations in nitrogen pick-up during SMA welding due to variations in the arc
length. Data from Morigaki et al.19

Next Page

gain

loss

Weld metal nitrogen content, ppm

SAW (multipass steel weldments)

Nitrogen content in electrode wire, ppm


Fig. 2.23. Nitrogen pick-up in SA welding at different levels of nitrogen in the electrode wire. Data from
Bhadeshia et a/.20
Solution

First we calculate the total amount of fused parent metal and weld deposit formed on welding.
From equations (1-75) and (1-120), we have:

and

The 'rule of mixtures' gives us the nominal weld metal nitrogen content, which is defined
as:

The above calculations show that the nitrogen content of single pass SA steel welds is close
to that of the base plate because of the high dilution involved. This is in agreement with general experience.

Previous Page

2.10 Absorption of Oxygen


Partial oxidation almost invariably accompanies the welding of steel. It is well established that
considerable interaction takes place between the liquid weld metal and its surroundings (arc
atmosphere, slag) when welding is performed in the presence of oxygen. For slag-protected
processes, the flux is the main source of oxygen because of its content of easily-reduced oxides, such as iron oxide, manganese oxide, silica, or rutile. In gas metal arc (GMA) welding,
oxygen is often deliberately introduced through the shielding gas to improve the arc stability
and bead morphology, but at the expense of an increased oxygen content in the weld metal and
intensified losses of alloying elements. The oxidation reactions proceed very rapidly under the
prevailing conditions owing to the high metal temperatures and the large interfacial contact
area available for interactions.
A general survey of oxygen contents in fusion welds is shown in Table 2.5.
2.10.1 Gas metal arc welding
The GMA welding process offers a special advantage in the way that it allows the reactions to
be studied through variations of the shielding gas composition without the complicating presence of a flux. Gas-shielded welding is also interesting form a practical point of view, since the
process is readily mechanised and can be applied for welding of nearly all types of steel.
Moreover, many of the reactions in the weld pool will be similar to those occurring during
welding with covered electrodes or submerged arc fluxes.
The amount of oxygen transferred to the metal during the high-temperature stage will clearly
depend on the atmosphere. In GMA welding with active gases such as CO2, Ar + O2 or Ar + O2 +
CO2 the extent of element absorption is controlled by the oxygen potential of the shielding gas.
This type of gas/metal interaction will be discussed below.
The analytical weld metal oxygen content is also affected by the presence of deoxidisers,
because the transition from the high-temperature stage to the cooling stage is not sharp. As was
Table 2.5 Summary of measured weld metal oxygen contents. Data compiled by Christensen.4
Welding Method

Material

SMAW (basic electrodes)

Low-alloy steel

250^40

SMAW (rutile electrodes)

Low-alloy steel

500-1300

SMAW (acid electrodes)

Low-alloy steel

700-1400

GTAW (pure Ar or He)

Low-alloy steel

50-100

GMAW (Ar-O 2 mixtures)

Low-alloy steel

200-1300*

GMAW (CO2)

Low-alloy steel

500-800

FCAW (self-shielded cored wire)

Low-alloy steel

130-200

SAW (calcium silicate fluxes)

Low-alloy steel

450-2600

SAW (manganese silicate fluxes)

Low-alloy steel

400-750

SAW (basic fluxes)

Low-alloy steel

180-350

Depends on the oxygen potential of the shielding gas.

Oxygen Content (ppm)

discussed briefly in the introduction to Chapter 2, a separation of non-metallic inclusions may


occur towards the end of stage one and possibly at the beginning of stage two. A necessary
condition for removal of oxygen introduced into the metal is, of course, the formation of oxygen-rich phases and compounds (e.g. microslag or carbon monoxide). These matters will be
considered in a separate section on weld metal deoxidation reactions.
2.10.1.1 Sampling of metal concentrations at elevated temperatures
In order to understand the extent and direction of the oxidation reactions, it is not sufficient to
characterise the initial and final conditions (i.e. consumables/parent plate and the weld metal).
Accordingly, especially designed experiments are required for assessment of the reactions
taking place at the four main process stages:
(1) Electrode tip
(2) Arc column
(3) Hot part of the weld pool
(4) Cold part of the weld pool.
Proper sampling techniques are needed for basic studies of this kind, which allow sampling
of the falling droplets during their flight through the arc column. An additional requirement
is that the speed of quenching is sufficiently high to freeze-in the metal composition established at elevated temperatures. If not, spontaneous reactions and subsequent losses of dissolved elements due to CO gas formation and manganese silicate slag precipitation may take
place on cooling.
To overcome these problems, a special 'melt spinning' technique has been developed by
Grong and Christensen,1 utilising the same principles as those employed in production of amorphous alloy ribbons. By using a water-cooled, fast rotating copper wheel (spinner) as a cathode, rapid crystallisation of droplets released from the electrode tip can be obtained in the
absence of a weld pool*. The reactions which normally occur in the pool during regular multirun deposition may then be assessed by comparison of analytical data for chilled metal (i.e.
falling droplets) and normal multi-layer weldment, respectively.
2.10.1.2 Oxidation of carbon
Filler wires intended for GMAW are usually overalloyed with respect to carbon, silicon, and
manganese to compensate for heavy oxidation losses of these elements during the welding
operation. It is evident from the data presented in Fig. 2.24(a) and (b) that virtually identical
carbon contents are obtained for chilled and multilayer weld metal, except when welding is
performed in extremely oxidising atmospheres (beyond 20 vol% O2 in Ar). This implies that
oxidation of carbon does not normally take place in the weld pool, but is located to one or both
of the preceding stages, i.e. electrode tip or arc column. However, since the carbon oxidation at
the electrode tip is more than sufficient to account for the observed total loss1, it is likely that
little, if any, carbon is lost in the latter stage.
The fact that carbon is found to oxidise readily at the electrode tip but not in later stages,
indicates that the supply of oxygen and/or the conditions of nucleation are more favourable in
*The estimated cooling rate is of the order of 105 0C s"1 or higher.1
The measured electrode tip carbon content also includes oxidation after arc extinction. This accounts for the observed
loss of carbon beyond that recorded in chilled metal or multi-layer weld deposit.
f

(a)

Wt% C

Electrode tip
Chilled metal (falling droplet)
Multi-layer weld metal
Electrode wire

Vol% O 2 in Ar

(b)

wt%c

Electrode tip
Chilled metal (failing droplet)
Multi-layer weld metal
Electrode wire

V o l % C O 2 in Ar

Fig. 2.24. Measured carbon contents in electrode tip droplets, chilled metal and multi-layer weld deposit
vs the oxygen potential of the shielding gas; (a) Ar-C>2 gas mixtures, (b) Ar-CC>2 gas mixtures. Data from
Grong and Christensen.1

Carbon control

Silicon control

Pco>

atm

the former case. Normally, homogeneous nucleation of CO gas within the liquid metal is considered impossible, which means that the CO nucleation in practice must take place heterogeneously. However, the most probable site for CO evolution during droplet formation at the
electrode tip will be the gas/metal interface itself, which allows carbon to be oxidised simultaneously with silicon and manganese. It is reasonable to assume that most of the observed
carbon oxidation is located to the hot layers facing the arc, where the reaction is thermodynamically favoured. At other surface positions Si and Mn are expected to prevent carbon from
reacting due to a rather low metal temperature, stated to be only slightly above the melting
point at the time of detachment.21
It is evident from the data in Fig. 2.24(a) and (b) that the carbon losses increase with increasing O2 or CO2 contents in the shielding gas up to a certain critical level. Hence, supply of
oxygen to the tip droplet surface is the rate controlling step for oxidation of carbon at low
oxygen potentials. This conclusion is also consistent with calculations made by Corderoy et
al.,22 who found that transport of atomic oxygen through a stagnant gaseous boundary layer
close to the metal surface controls the oxidation rate of alloying elements at this stage of the
process.
The carbon oxidation will gradually decline with increasing oxygen concentrations in the
shielding gas, probably as a result of build-up of carbon monoxide in the surrounding atmosphere. When the critical CO gas pressure is reached, the carbon reaction is blocked, silicon
(and manganese) now exercising control of the oxygen level, as indicated in Fig. 2.25. For ArO 2 gas mixtures this critical pco pressure is attained at about 10 to 15 vol% O 2 in the shielding
gas, corresponding roughly to 0.05 wt% C oxidised in chilled metal. When welding is performed in Ar-CO 2 mixtures the reaction is blocked at a much earlier stage of carbon oxidation
(equal to about 0.02 wt% C lost in chilled metal), since dissociation of CO2 in this case will
produce an additional amount of CO to concentrate in the surrounding gas phase.

Temperature, 0C
Fig. 2.25. Break even equilibrium partial pressure of CO vs temperature for silicon control of oxygen
level at 0.8 wt% Si and silica saturation. Data from Elliott et al. 23

The mechanism indicated above is supported by the data presented in Fig. 2.26, which
show that a CO2-rich atmosphere even may act carburising if the initial carbon content of the
electrode wire is sufficiently low. Moreover, it is interesting to note that the carbon-oxygen
reaction is also influenced by the rate of droplet detachment. Since the highest carbon oxidation losses are normally associated with a coarse globular droplet transfer mode, this suggests
that the reaction time is more important than the effective contact area available for interaction
which depends on the droplet size.
2.10.1.3 Oxidation of silicon
It is evident from the data in Fig. 2.27 that loss of silicon mainly take place in the weld pool, as
indicated by the difference between the measured silicon content in chilled and multi-layer
weld metal. In Ar-O 2 mixtures the Si loss increases steadily with increasing oxygen potential
of the shielding gas. Thus, at 30 vol% O2 in Ar it amounts to 0.59 wt% Si (or 0.67 wt% O)
removed from the weld pool as a result of deoxidation reactions. A similar situation exists in
the case of CO2-shielded welding up to about 20 vol% CO2 in Ar. At higher CO2 contents, the
Si loss tends to drop off, finally attaining an upper limit of approximately 0.30 wt% Si corresponding to 0.34 wt% O removed from the weld pool.
In comparison, the amount of Si lost in the two preceding stages (i.e. electrode tip and arc
column) is much smaller, as shown by the data for the chilled metal Si content. Since no slag is
formed under the conditions of rapid cooling, silicon must escape in the form of a gaseous

Weld metal carbon content, wt%

GMAW (low-alloy steel)

(Gain)

(Loss)

Filler wire carbon content, wt%


Fig. 2.26. Correlation between filler wire and weld metal carbon contents in CO2-shielded welding.
Data from Ref.24.

(a)
Chilled metal (falling droplet)
Multi-layer weld metal

Wt% Si

Electrode wire

Vol% O 2 in Ar

(b)
Chilled metal (falling droplet)
Multi-layer weld metal

Wt% Si

Electrode wire

Vol% CO 2 in Ar

Fig. 2.27. Measured silicon contents in chilled and multi-layer weld metals vs the oxygen potential of the
shielding gas; (a) Ar-02 gas mixtures, (b) Ar-CC>2 gas mixtures. Data from Grong and Christensen.1

product. Evaporation losses can in this case be excluded due to a very low vapour pressure of
silicon at the prevailing temperatures. It is therefore reasonable to assume that SiO(g) forms as
a result of chemical reactions occurring at the electrode tip. The observed decrease in chilled
metal Si content with increasing O2 and CO2 contents in the shielding gas is probably caused
by the presence of CO at the gas/metal interface, which facilitates SiO formation according to
the reaction:
(slag)
(gas)
(gas)
(gas)
(2-36)

PSiO,10-3atm

Figure 2.28 shows a plot of the equilibrium partial pressure of SiO vs temperature at silica
saturation for three different CO levels. Note that stoichiometric amounts of CO2 have been
assumed to form in order to calculate psi0. It is evident from the thermodynamical data presented in Fig. 2.28 that the formation of SiO is strongly dependent on the CO partial pressure
at the slag/metal interface. Thus, from a thermodynamic standpoint the silicon loss at the electrode tip should be most pronounced during CO2-shielded welding due to the resulting higher
CO pressure. The data shown in Fig. 2.27(a) and (b) support this assumption. It is seen that the
silicon loss is increased by a factor of 3 to 5 in presence of CO2 when comparison is made on
the basis of equal oxygen potential of the shielding gas (i.e. equal loss of deoxidants in the
weld pool). Similar observations have also been made by Heile and Hill27 from determination
of silicon in collected GMA welding fumes. The recorded chilled metal Si loss in Fig. 2.27(a)
and (b) is in good agreement with the reported fume formation rates of silicon. Since all CO
consumed in reaction (2.36) is expected to be regenerated immediately by decomposition of
CO2 at the metal surface, the high CO partial pressure required for SiO formation at the electrode tip is maintained even in the case of extensive silicon losses. Consequently, the net reaction can be written as:

Temperature, 0C
Fig. 2.28. Equilibrium partial pressure of SiO vs temperature at different CO levels. Data from Refs. 25
and 26.

(slag)

(dissolved)

(gas)

(2-37)

When welding is performed in Ar-O 2 gas mixtures, pSiO andpco may be taken proportional
to the recorded loss of silicon and carbon in chilled metal (see data in Fig. 2.29). It is evident
from this plot that the silicon loss is directly proportional to the corresponding loss of carbon
up to a certain critical level. Thus, during the initial period of carbon oxidation at the electrode
tip the SiO formation is probably controlled by the resulting partial pressure of CO at the slag/
metal interface, according to reaction (2-36). When the carbon reaction is blocked, the chilled
metal silicon loss becomes independent of the CO partial pressure (see reaction (2-37)), since
all CO consumed in the SiO formation will immediately be recirculated within the system.
However, at the break even point for silicon control of the oxygen level at the electrode tip, the
CO pressure in the surrounding gas phase will be the same for both Ar-O 2 and Ar-CO 2 gas
mixtures. Hence, the recorded loss of silicon in chilled metal at 20 vol% CO2 in Ar is seen to be
similar to that in Ar + 10 vol% O2, as indicated by the heavy broken line in Fig. 2.29.
At the temperatures where liquid steel is normally deoxidised, silicon and manganese have
a strong affinity to oxygen. Their ability to form stable oxides decreases rapidly with increasing temperature, and above approximately 180O0C silicon and manganese do no longer act as
efficient deoxidation agents. Precipitation of manganese silicate slags is therefore favoured by
the lower metal temperatures prevailing at the electrode tip and in the cold part of the weld
pool. At higher temperatures, these oxides become unstable. Consequently, as a result of the
metal superheating occurring during droplet transfer through the arc column, the macroscopic
slag phase formed earlier at the electrode tip surface (as reported by Corderoy et ah12) will
redissolve in the metal. This gives rise to a relatively high chilled metal oxygen content (to be
discussed below).

Loss of silicon (%)

Ar+20 vol% CO2

Ar+10vol%O2

Loss of carbon (%)


Fig. 2.29. Correlation between loss of carbon and silicon in chilled metal at different O 2 levels in the
shielding gas. The corresponding loss of silicon at 20 vol% CO2 in Ar is indicated by the heavy broken
line in the graph. Data from Grong and Christensen.1

Example (2.5)

Consider CO2-shielded welding on a thin sheet of low-alloy steel with a 0.8mm dia. electrode
wire under the following conditions:

Based on the data presented in Fig. 2.27(b), calculate the fume formation rate (FFR) of
silicon (in mg per min) due to SiO formation at the electrode tip. The wire feed rate is 125mm s"1.
Solution

The total loss of silicon due to SiO formation may be taken equal to the observed difference
between the filler wire and the chilled metal silicon contents. For welding in pure CO2, we get:

The corresponding fume formation rate of silicon (in mg min"1) can readily be calculated
when the wire feed rate (WFR) is known. Taking the density of the steel equal to 7.85mg mm"3,
we obtain:

A comparison with the measured FFR of silicon in Table 2.6 (at / = 13OA) shows that the
calculated value is reasonable correct. Moreover, these data support our previous conclusion
that the SiO formation is favoured by a high CO2 content in the shielding gas due to the
dissociation reaction. In fact, more detailed studies of the reaction kinetics have confirmed that
the rate of SiO formation is proportional to the resulting partial pressure of CO at the gas/metal
interface,28 in agreement with equation (2-36).
2.10.1.4 Evaporation of manganese
It is seen from the data in Fig. 2.30(a) and (b) that the amount of manganese lost in chilled
metal is virtually independent of the oxygen potential of the shielding gas, as indicated by the
constant difference of about 0.35 wt% between the filler wire and the chilled metal Mn contents in both graphs. This implies that significant amounts of manganese are lost during
droplet transfer through the arc column as a result of evaporation.
At the prevailing temperatures, the vapour pressure of iron is also high due to the almost
unity activity of Fe. If the average arc metal temperature is taken equal to about 24000C,29 the
data in Fig. 2.31 indicate that the vapour pressures of iron and manganese (at 1.27 wt% Mn)
are nearly identical and close to 0.05 atm. In the hot surface layers of liquid metal facing the
arc the temperature will be even higher, which means that iron vapour will dominate.
Measurements of collected GMA welding fume reported by Heile and Hill27 (see data in
Table 2.6) show a substantial higher loss of iron than that derived from simple thermodynamical calculations taking the rate of element loss proportional to the vapour pressure. From their
results a reasonable value of the average mass ratio Fe to Mn in dust is about 5. Consequently,
as a preliminary estimate the loss of iron may be taken 5 times the amount of manganese lost

Table 2.6 Measured fume formation rates in GMA welding of ferrous materials. Data from Heile and
Hill.27

Shielding Gas

Argon

Ar+ 2 % O 2

Ar+ 5% O 2

Ar + 25% CO2

Pure CO2

Current

Voltage

(A)

(V)

Fume Formation Rate


(mg min"1)
Mn

Si

Fe

250

29

22

300
350

31
35

1
5

0
4

12
51

150
200
300
400

28
28
29
34

15
12
5
18

8
10
4
16

134
75
35
86

100
200
300

28
28
28

29
16
10

33
22
23

273
129
76

100

23

11

20

75

150
300

27
35

13
29

35
62

105
191

130
150
200
250
300

27
30
30
30
30

13
14
19
31
36

59
63
73
112
125

86
120
126
216
214

in chilled metal, corresponding to a total loss of 1.7 wt% (Fe + Mn) or 500mg (Fe + Mn) per
min. The reported fume formation rate of Fe and Mn under nearly similar conditions is only
one half of that stated above, which indicates that the calculated flux of metal vapour may be
somewhat overestimated. However, if these two values are considered to represent borderline
cases, the volume of metal vapour is in the range from 300 to 500 times that of the droplets
and will therefore be more than sufficient to protect the metal from arc atmosphere oxidation
at this stage of the process. This conclusion is consistent with statements made by Distin et al?1
who claim that iron vapour acts as an effective oxygen getter already at about 19000C.
Very large amounts of manganese are also lost in the weld pool stage, as shown by the
difference between the measured Mn contents in chilled and multi-layer weld metals. This
situation appears to be quite similar to that of silicon. For Ar-O 2 mixtures the manganese loss
increases steadily with increasing oxygen contents in the shielding gas, and reaches a value of
about 1.08 wt% Mn (or 0.31 wt% O) removed from the weld pool at 30 vol% O2 in Ar. In the
case OfAr-CO2 shielded welding the Mn oxidation loss starts to drop off at a CO2 content of
about 20 vol%, finally attaining an upper limit of about 0.50 wt% Mn corresponding to 0.15
wt% O removed from the weld pool.
A more detailed discussion of oxidation reactions in GMA welding is given in Section
2.10.1.5.
Example (2.6)

Consider GMA welding low-alloy of steel under conditions similar to those in Example 2.5.

(a)

Wt% Mn

Chilled metal
(falling droplet)

Multi-layer weld metal


Electrode wire

Vol% O 2 in Ar

Wt% Mn

(b)

Chilled metal (falling droplet)


Multi-layer weld metal
Electrode wire

Vol%CO 2 inAr
Fig. 2.30. Measured manganese contents in chilled and multi-layer weld metals vs the oxygen potential
of the shielding gas; (a) Ar-02 gas mixtures, (b) Ar-CO2 gas mixtures. Data from Grong and Christensen.1
Based on the data presented in Fig. 2.30(b), calculate the fume formation rate (FFR) of manganese due to evaporation losses occurring during droplet transfer through the arc column. Estimate also the effective mass transfer coefficient for manganese evaporation under the prevailing circumstances by utilising the vapour pressure data in Fig. 2.31. The surface temperature
of the falling droplets is assumed constant and equal to 26000C.

PMr/PFe

LogpMn,atm

Temperature, 0C
Fig. 2.31. Equilibrium manganese vapour pressure and corresponding vapour pressure ratio pMn to PFe vs
temperature at 1.27 wt% Mn in iron. Data from Kubaschewski and Alcock. 30
Solution

The total loss of manganese due to evaporation may be taken equal to the observed difference
between the filler wire and the chilled metal manganese contents. For welding in pure CO2, we
have:

The corresponding fume formation rate of manganese (in mg min"1) can readily be calculated when the wire feed rate (WFR) is known. Taking the density of the steel equal to 7.85mg
mm~3, we obtain:

A comparison with the data in Table 2.6 reveals that the reported fume formation rate of
manganese (at / = 13OA) is much lower than computed in the present example. This discrepancy can probably be attributed to differences in the filler wire manganese content. Assuming
that the evaporation loss of manganese is controlled by a transport mechanism in the gas phase,
we can estimate the effective mass transfer coefficient from equation (2-15) by inserting a
reasonable average value for the manganese vapour pressure at the gas/metal interface (pMn in
the bulk phase is taken equal to zero). Reading from Fig. 2.31 (at T= 26000C) gives:

If the diameter of the falling droplets is taken equal to the diameter of the filler wire, it is
possible to calculate the total loss of manganese associated with one droplet (in mol):

Since the average flight time of large, globular droplets through the arc column is of the
order of one second, the corresponding flux of manganese vapour per unit time is close to 9.97
X 10~8 mol s"1. Thus, by rearranging equation (2-15), we obtain the following value for the
effective mass transfer coefficient:

Although the above value is rather uncertain, the calculated mass transfer coefficient is of
the expected order of magnitude. This supports our previous conclusion that the evaporation
kinetics are controlled by a transport mechanism in the gas phase.
2.10.1.5 Transient concentrations of oxygen
It is evident from the data summarised in Fig. 2.32 that the oxygen content in both chilled and
multi-layer weld metal increases with increasing oxygen potential of the shielding gas. The
chilled metal analysis is representative of the oxygen absorption occurring at the electrode tip
due to the lack of gas/metal interaction in the arc column. Measurements of manganese oxidation losses during droplet formation have been performed by Corderoy et al.22 From the curve
presented for 20ms tip melting cycles, a reasonable estimate for the manganese oxidation loss
in Ar-O 2 shielding gas mixtures may be:
(2-38)
Taking the Si to Mn mass ratio in precipitated slag equal to 0.66,22 the corresponding loss of
silicon is equal to:
(2-39)
Based on equations (2-38) and (2-39) it is possible to calculate the oxygen absorption at the
electrode tip which is associated with the MnO and the SiO2 slag formation. If corrections also
are made for the amount of oxygen simultaneously removed as iron oxide*, we obtain:
(2-40)

*An approximate correction can be made from an analysis of the Fe to Mn mass ratio in precipitated slag, which under
the prevailing circumstances is close to 0.16.22

(a)

Wt% O

Chilled metal (falling droplet)


Multi-layer weld metal
Electrode wire

Vol%O2inAr
(b)

Wt% O

Chilled metal (falling droplet)


Multi-layer weld metal
Electrode wire

Vol% CO2 in Ar
Fig. 2.32. Measured oxygen contents in chilled and multi-layer weld metals vs the oxygen potential of
the shielding gas; (a) Ar-C>2 gas mixtures, (b) Ar-CO2 gas mixtures. Data from Grong and Christensen.1
At 30 vol% O2 in Ar the absorption of oxygen is roughly 0.27 wt% O, which is reasonably
close to that recorded in the chilled metal analysis (about 0.23 wt% O). The rate controlling
step for metal oxidation at this stage of the process is believed to be transport of atomic oxygen
through a stagnant gaseous boundary layer of Ar and/or CO adjacent to the metal surface.22
On the other hand, the measured chilled metal oxygen contents are much too low to account
for the heavy oxidation losses of deoxidants observed in multi-layer welds. This indicates that

considerable amounts of oxygen are introduced in the hot part of the weld pool immediately
beneath the root of the arc. Although the gas/metal interfacial contact area available for reaction is much smaller than that of the electrode tip (or falling) droplets, the strong turbulence
existing in the hot part of the pool will provide an effective circulation of the liquid metal
through the reaction zone.32 Moreover, absorption of oxygen at this stage of the process will be
favoured by the increased time available for gas/metal interaction. If the diameter of the reaction zone is taken equal to about 3mm, the time available for oxygen absorption in the weld
pool is of the order of one second for a typical welding speed of 3mm s"1. In comparison, the
corresponding reaction time during droplet formation at the electrode tip is only 20 to 50 ms.
Calculations of the total oxygen absorption can be done on the basis of the measured difference between silicon and manganese contents in chilled and multi-layer weld metals, designated A[%S7] and A[%Mn], respectively*. Moreover, corrections should be made for the amount
of oxygen simultaneously removed as iron oxide from the weld pool. For deoxidation with
silicon and manganese the mass ratio wt% Fe to wt% Mn in precipitated slag is equal to about
O.I.1 This leads to the following balance:
(2-41)
At 30 vol% O2 in Ar the total oxygen absorption amounts to:

It is seen from the graphical representation of equation (2-41) in Fig. 2.33(a) and (b) that
most of the oxygen pick-up takes place in the weld pool. However, at present it is not clear
whether the calculated values represent a real transient concentration in the hot part of the
weld pool or a number in concentration units representing precipitation of manganese silicate
slags. According to Fischer and Schumacher33 the solubility of oxygen in liquid iron is 0.94
wt% at 19000C and 1.48 wt% at 20000C This temperature range is probably relevant with respect to the hot part of the weld pool, although the surface metal temperature will be even
higher.34 However, in steel weld metals the oxygen concentrations should be well below this
solubility limit in the presence of silicon and manganese. Taking the average weld pool silicon
content equal to 0.50 wt%, the corresponding oxygen content in equilibrium with a silicasaturated slag is roughly 0.10 and 0.20 wt% at 1900 and 20000C, respectively.23 On the other
hand, under the prevailing circumstances there will be no solid/liquid interface available for
nucleation of oxide particles, and hence homogeneous nucleation is the only possibility.
According to Sigworth and Elliott35 this requires a relatively high degree of supersaturation,
which means that the oxygen concentration most likely will exceed the Si-Mn-O equilibrium
value before slag precipitation occurs.
The data in Fig. 2.33(a) and (b) indicate that the weld pool oxygen absorption is controlled
by a complex transport mechanism in the gas phase. However, since mass transfer in gas-jet/
liquid systems is not fully understood,3 we shall only consider the limiting case where the
resistance to mass transfer is confined to a stagnant gaseous boundary layer adjacent to the
metal surface. Under such conditions equation (2-10) predicts that the transient oxygen concentration is determined by the partial pressure of atomic oxygen in the plasma gas, as shown
*Oxygen consumed in CO and SiO formation is not included.

(a)

Wt% O

Calculated total oxygen


absorption
Analytical weld metal
oxygen content

Vol% O2 in Ar

(b)

Wt% O

Calculated total oxygen


absorption
Analytical weld metal
oxygen content

Vol% CO2 in Ar
Fig. 2.33. Calculated total oxygen absorption in GMA welding at different oxygen potentials of the
shielding gas; (a) Ar-O 2 gas mixtures, (b) Ar-CO 2 gas mixtures. The analytical weld metal oxygen
content is indicated by the broken lines in the graphs. Data from Grong and Christensen.1

by the plots in Fig. 2.34. If we as a borderline case assume that all oxygen present in the plasma
gas is immediately absorbed in the liquid metal, the slope of the curve in Fig. 2.34 for Ar-O 2
gas mixtures (equal to about 2.67 wt% dissolved oxygen at one atmosphere total pressure of
atomic oxygen) is representative of the total amount of gas passing through the arc column.
Taking the mass of liquid metal leaving/entering the reaction zone equal to 3Og min"1 under
the prevailing circumstances,1 the following gas flow rate is obtained (in Nl min"1).

This value corresponds to about 6 per cent of the total shielding gas flow rate.
In contrast to the situation described in Fig. 2.34 for Ar-O 2 gas mixtures, a large deviation
from the expected relationship is observed for welding in CO2-rich atmospheres. At present,
the reason for this shift in the reaction kinetics is not known. However, it is evident from the
data presented in Fig. 2.34 that the effective mass transfer coefficient decreases by a factor of
about 2.5 when the CO2 content in the shielding gas increases from 10 to 100 vol%. This
implies that the oxidation capacity of pure CO2 is comparable with that of Ar + 13 vol% O2,
although the partial pressure of atomic oxygen in the plasma gas is equivalent with an oxygen
content of about 33 vol%.
Example (2.7)

[%O]t0t,wt%

Consider GMA welding of low-alloy steel in Ar + 20 vol% CO2 and pure CO2, respectively.
Based on the results in Figs. 2.27(b) and 2.30(b), calculate the total weight of top bead slag (in
gram per 100 gram weld deposit) which forms as a result of deoxidation reactions. Assume in
these calculations that the iron to manganese mass ratio in precipitated slag is equal to 0.1.

Pure CO2

P0, atm
Fig. 2.34. Calculated total oxygen absorption in GMA welding at different partial pressures of atomic
oxygen in the plasma gas. Data from Grong and Christensen.1

Solution

The amount of silicon and manganese lost as a result of deoxidation reactions is equal to the
observed difference in the chilled and multi-layer weld metal Si and Mn contents. Assuming
that these elements are removed from the weld pool as SiO2 and MnO, respectively, the following balance is obtained:

Thus, for welding in Ar + 20 vol% CO2, the total weight of slag amounts to:

Similarly, for CO2-shielded welding, we get:

A comparison with the experimental data in Fig. 2.35 shows that there is a fair agreement
between the amount of slag recorded by weighing and that calculated from a simple mass
balance of silicon, manganese and iron.

g slag/10Og deposit

Calculated weight of slag


Measured weight of slag

Vol% CO2 in Ar
Fig. 2.35. Comparison between measured and calculated weight of top bead slag in CO2-shielded welding. Data from Grong and Christensen.1

2.10.1.6 Classification of shielding gases


The data in Fig. 2.34 provide a basis for evaluating the oxidation capacity of various shielding
gases. For welding in Ar + O2 and Ar+CO2 gas mixtures up to 10 vol% CO2 in argon, the total
oxygen absorption is approximately given by the following equation:

(2-42)
Similarly, for welding with CO2-rich shielding gases (i.e. between 10 and 100 vol% CO2),
we obtain:
(2-43)
Equal oxidation capacity means that the total weld metal oxygen absorption is the same for
both shielding gas mixtures. Hence, we may write:

(2-44)
when the CO2 content in the shielding gas is less than 10 vol%, and

(2-45)

when welding is performed with CO2-rich shielding gases (more than 10 vol% CO2 in Ar).
Based on equations (2-44) and (2-45) it is possible to compare the oxidation capacity of
various shielding gases (see Table 2.7). Included in Table 2.7 is also a slightly modified version of the International Institute of Welding (HW) classification system,36 which is based on
an evaluation of retained (analytical) oxygen in the weld deposit. It is evident from these data
that both systems are applicable and mutual consistent, although the former one utilises a more
rational criterion for the shielding gas oxidation capacity.
2.10.1.7 Overall oxygen balance
In GMA welding with solid wires, the CO content in the exhaust gas provides a direct measure
of the extent of gas/metal interaction. This CO content should be compatible with that calculated from an overall oxygen balance for the reaction system.37
Example (2.8)

Consider GMA welding of low-alloy steel under the following conditions:

Table 2.7 Proposed shielding gas classification scheme for GMA welding of low-alloy steel according to
equations (2-44) and (2-45). Included is also a modified version of the corresponding IIW's classification
system.36
Shielding Gas Composition

[%O] tot

Vol%CO2

Vol%O 2

Vol%Ar

(wt%)

IIW's Terminology*

0-4
0-2

0-2
0-1

balance
balance
balance

<0.10

Lightly oxidising
(<0.02)

4-10
2-5

2-5
1-2V2

balance
balance
balance

0.1-0.25

Oxidising
(0.025-0.035)

10-25
25-100

5-8
-

balance
balance

0.25-0.40

Oxidising
(0.035-0.045)

8-13
-

balance
balance

0.40-0.60

Strongly oxidising
(0.045-0.07)

> 13

balance

>0.60

Extremely oxidising
(> 0.07)

The analytical weld metal oxygen content (in wt%) is given by the values in brackets.
The shielding gas is pure CO2 and is supplied at a constant rate of 15Nl min *. Based on the
composition data in Table 2.8 calculate the resulting CO content in the welding exhaust gas.
Solution
First we calculate the nominal weld metal chemical composition by neglecting oxidation loss
of alloying elements due to chemical reactions:

The dilution ratios BI(B + D) and DI(B + D) can be estimated from the classic heat flow
theory presented in Chapter 1. From equations (1-75) and (1-120), we have:

Table 2.8 Chemical composition of filler wire, base plate and weld metal used in Example 2.8.
C
(wt%)

O
(wt%)

Si
(wt%)

Mn
(wt%)

Filler wire

0.10

0.01

0.93

1.52

Baseplate

0.14

0.007

0.40

1.30

Weld metal

0.09

0.065

0.35

0.81

Element

and

This gives:

The extent of gas/metal interaction can then be evaluated from the observed concentration
displacements:

Calculated values for the concentration displacements of carbon, oxygen, silicon and manganese utilising the composition data in Table 2.8 are given below.
Element
\-%X\nom.

[A%X]

C
0

1 2

-0.03

O
0

0 0 8

0.057

'

Si

Mn

6 3

L 3 9

-0.28

-0.58

The total CO evolution (in mol min l) can now be computed from an overall oxygen balance for the reaction system. In these calculations we shall assume that Si and Mn lost as
SiO(g) and Mn(vap.) immediately react with CO2 to form SiO2 and MnO, respectively*. Taking the density of steel equal to 7.85 X 10~3 g mm"3, the total mass of weld metal produced per
unit time amounts to:

Overall oxygen balance (i.e. consumption of CO2)


Oxidation of carbon:

Oxidation of silicon:

CO 2 consumed in oxidation of iron vapour is disregarded.

Oxidation of manganese:

Increase in oxygen content:

Total CO evolution (sum):(13.4 + 53.5 + 28.3 + 9.6) X 10~3 mol CO min - 1 = 104.8 X 10~3 mol
CO min' 1
Based on this information it is possible to calculate the resulting CO content in the welding
exhaust gas:

A comparison with the data in Table 2.2 shows that a CO content of about 15 vol% is
reasonably close to that determined by analysis.
2.10.1.8 Effects of welding parameters
So far, gas/metal interactions in GMA welding has mainly been discussed in terms of the
oxygen potential of the shielding gas. In the following, some consideration will be given to the
effects of welding parameters on the weld metal chemistry.
Amperage

When the welding current is raised, the time available for interaction decreases due to the more
rapid detachment of the electrode tip droplets. At the same time the interfacial contact area
increases as the average droplet size becomes smaller. From measurements of fume formation
rates in GMA welding,27 it has been shown that these two counteracting effects will almost
cancel, i.e. the total amount of emitted dust (in mg per g deposit) is found to be constant and
nearly independent of the applied amperage. On the other hand, the total fume formation rate
is probably not a reliable index for the burn-off of Si and Mn, since the evolution of iron
vapour during droplet transfer will tend to conceal the corresponding loss of alloying elements.
The effect of amperage (or more correct the droplet detachment frequency) on the burn-off
of carbon, silicon, and manganese in CO2-shielded welding has been investigated by Smith et
al.38 They found that the recovery of alloying elements in the weld deposit increased with
increasing welding current (i.e. droplet detachment frequency). In view of the previous discussion, it is reasonable to assume that the higher weld metal carbon and silicon contents reported
by Smith et al3S are a result of a reduced CO and SiO gas evolution at the electrode tip due to
the shorter time available for chemical interaction. In the case of manganese reduced evaporation losses because of a more rapid transfer of the droplets through the arc column offers a

reasonable explanation to the increased element recovery. This shows that the weld metal
chemistry is sensitive to variations in the welding current.
Arc voltage

Since the arc voltage neither affects the melting rate nor the droplet size to any great extent,39
variations in the arc voltage should only have a minor effect on the weld metal chemistry. This
conclusion is apparently in conflict with observations made by Lindborg,40 who found that the
oxidation reactions in GMA welding were strongly voltage dependent and at the same time
independent of the welding current, the droplet detachment frequency, and the mode of metal
transfer (spray or short-circuiting). Consequently, further investigations are required to explain these discrepancies.
Welding speed

It can be inferred from the data of Grong and Christensen1 that the analytical weld metal
carbon and oxygen contents are virtually independent of the welding speed v within the normal range of GMAW (i.e. from 0.4 to 6 mm s"1). However, the intensified losses of silicon and
manganese observed at low welding speeds indicate that more oxygen is absorbed in the weld
pool under such conditions. This point is more clearly illustrated in Fig. 2.36 which shows a
plot of [%O]tot vs v for a series of multi-pass GMA welds deposited under the shield of Ar +
10 vol% O2. It is evident that the total oxygen absorption increases nearly by a factor of two
when the welding speed decreases from 6 to 0.4mm s"1. This shows that the welding speed has
a marked effect on the transient oxygen pick-up in the hot part of the weld pool during GMA
welding, since it controls the time available for element absorption.
2.10.2 Submerged arc welding
In flux-shielded processes the reaction pattern is much more difficult to assess because of the

[%O]tot,wt%

Calculated total oxygen absorption

Travel speed, mm/s


Fig. 2.36. Calculated total oxygen absorption in GMA welding at different travel speeds. Data from
Grong and Christensen.1

complicating presence of the slag. For this reason most investigators have chosen to analyse
empirically slag/metal reactions in SA welding. Nevertheless, some authors have been able to
interpret their results on more theoretical grounds in spite of the complex reaction systems
involved.41"*5 Unfortunately, these thermodynamic approaches give, at best, only a qualitative
description of the compositional changes occurring during the welding operation.
Recently, a kinetic model has been developed by Mitra and Eagar46 to account for variations in the element recovery in both single-pass and multi-pass SA steel weldments. From
their work it is evident that the transfer of alloying elements between the slag phase and the
weld metal cannot be adequately described by means of a primitive model of pseudo-equilibrium without including a more detailed analysis of the reaction kinetics. This shows that the
conditions existing in SA welding are quite similar to those prevailing during GMA welding,
although the experimental and theoretical challenges are much greater in the former case due
to the complicating presence of a macroscopic slag phase.
2.10.2.1 Flux basicity index
During SA welding of steel, oxygen may be transferred from the slag to the weld metal due to
decomposition of easily reduced oxides at elevated temperatures according to the overall reaction:
MxOy = xM (dissolved) + y O (dissolved)

(2-46)

where MxOy denotes any oxide component in the slag phase (e.g. SiO2, MnO or FeO).
The basicity index (B.I.), originally adopted from steel ladle refining practice, is most frequently employed for assessment of oxygen pick-up in SA welding, since it gives an approximate measure of the flux oxidation capacity. A number of different expressions exists in the
literature, but for the purpose of convenience the basicity index defined by Eagar47 has been
adopted here:
basic oxides"
non- basic oxides

(2-47)
where the concentration of each flux component is given in weight percent.
It is evident from Fig. 2.37, which shows a typical correlation between the weld metal
oxygen content and B.L, that the oxygen level of welds produced under acid fluxes (i.e. low
B.I.) is strongly dependent on the basicity index. In contrast, the oxygen concentrations of
welds deposited under basic fluxes are seen to be essentially independent of B.I., as indicated
by the horizontal part of the curve in Fig. 2.37. It should be noted that this analysis gives no
information about the extent of slag/metal interaction, since it is based on data for retained
oxygen in the weld deposit. Consequently, because of the empirical nature and limited applicability of the basicity index, its role in the choice of welding fluxes for SA welding is a keenly
debated question.

Oxygen content, wt%

Basic fluxes

Acid fluxes

Flux basicity index


Fig. 2.37. Correlation between retained oxygen and flux basicity in SA welding. Data from Eagar.47

2.10.2.2 Transient oxygen concentrations


In SA welding of C-Mn steels, the transient flux of oxygen passing through the weld pool can
be estimated from the observed concentration displacements of silicon and manganese, which
may be taken equal to the difference between absorbed and rejected Si and Mn, respectively:
(2-48)
and
(2-49)
If we assume that rejection of Si and Mn in the weld pool occurs as a result of MnSiO3
microslag precipitation and subsequent phase separation, [%Mn]rej^ is bound to [%Si]rej. through
the following stoichiometric relationship:
(2-50)
Taking the ratio between absorbed Mn and Si in the weld metal equal to k, a combination of
equations (2-48), (2-49) and (2-50) gives:

(2-51)
The value of k is difficult to evaluate in practice, but in view of the reported mass transfer
coefficients for Mn and Si a reasonable estimate would be about 0.5 in the case of manganese
silicate fluxes.46 Under such conditions the total oxygen absorption, [%O]abs is given by:

(2-52)
where [A%O] is the observed concentration displacement of oxygen in the weld metal, and
[%O]rej. is the amount of oxygen rejected from the weld pool as a result of deoxidation reactions.
Based on equation (2-52) it is possible to estimate the total oxygen absorption during SA
welding of C-Mn steels from an analysis of measured concentration displacements of oxygen,
silicon, and manganese in the weld metal. The results of such calculations are shown graphically in Fig. 2.38, using data from Indacochea et a/.44 It is evident from this plot that the total
oxygen absorption during SA welding is much larger than that inferred from an analysis of
retained oxygen in the weld deposit. The situation is thus quite similar to that observed experimentally in GMA welding (see Fig. 2.33).
It should be noted that the calculated values for [%O]abs. in Fig. 2.38 may be encumbered
by systematic errors due to the number of simplifying assumptions inherent in equation (2-52).
However, this does not affect our main conclusion regarding the significance of the oxygen
absorption, since more refined calculations give a pattern similar to that observed above (see
Fig. 2.39).
2.10.3 Covered electrodes
Chemical reactions during SMA welding have been studied by several investigators in the

Oxygen content, wt%

Experimental MnO-FeO-SiO2 fluxes

Retained oxygen

[A% Si]0.5 - [A%Mn]


Fig. 2.38. Calculated total oxygen absorption in SA welding with experimental MnO-FeO-SiO2 fluxes.
Data from Indacochea et al.AA

Oxygen content, wt%

Flux Type CS: Bead on Plate


: Two Wires
Filled Symbols: Flux FB

Absorbed
oxygen
Retained
oxygen

Silicon content, wt%


Fig. 2.39. Calculated total oxygen absorption in SA welding with commercial calcium silicate (CS) and
fluoride-basic (FB) fluxes. Data from Christensen and Grong.45
past.47"51 Most of these investigators have interpreted their results as a high-temperature equilibrium between the slag and the weld metal, but a verified quantitative understanding of the
transfer of elements during welding is lacking. This situation arises mainly from the lack of
adequate thermodynamic data for the complex slag/metal systems involved.
2.10.3.1 Reaction model
The reaction model presented here is restricted to welding with basic covered electrodes.
During SMA welding gases are generated by decomposition of compounds present in the
electrode coating. In the case of basic covered electrodes, the decomposition of limestone
results in an atmosphere consisting predominantly of carbon monoxide and carbon dioxide,
containing only small amounts OfH2 and H2O (see data in Table 2.2). The characteristic high
concentrations of CO and CO2 in the arc atmosphere would be expected to lead to extensive
absorption of carbon and oxygen in the weld metal. Under the prevailing circumstances, it is
reasonable to assume that these reactions approach a state of local pseudo-equilibrium during
droplet transfer through the arc column.
During the subsequent stage of cooling in the weld pool, a supersaturation with respect to
the various deoxidation reactions is initially increasing, which is released when the conditions
for nucleation of the respective reaction products are reached. Since carbon is a much stronger
deoxidant than silicon and manganese at temperatures above about 17000C,23 it is reasonable
to assume that carbon will be in control of the oxygen level during the initial stage of cooling*,
in accordance with the reaction:
C (dissolved) + O (dissolved) = CO(gas)

(2-53)

* Although gases such as CO and CO2 are widely dissociated and ionised in the arc column, from a thermodynamic
standpoint, there is no objection to the choice of molecular species as components for the system, provided that
equilibrium is maintained down to temperatures where such species are stable.

Oxygen content, wt%

Carbon boil in the weld pool has been detected experimentally during welding with covered electrodes,49 which implies that heterogeneous nucleation of CO is kinetically feasible
under the prevailing circumstances. Possible nucleation sites for CO are gas bubbles present in
the macroscopic slag layer covering the metal, created by the vigorous stirring action of the arc
plasma jet. It should be noted that this behaviour is in sharp contrast to experience with GMA
welding, where little or no oxidation of carbon takes place in the weld pool, as shown previously in Section 2.10.1.2. It is tentatively suggested that the apparent difference between SMA
and GMA welding regarding the possibilities for CO nucleation in the weld pool arises from
the lack of a macroscopic slag layer in the latter case.
Unlike carbon, the deoxidation capacity of silicon (and manganese) increases rapidly with
decreasing metal temperatures (se Fig. 2.40), which means that carbon oxidation becomes
gradually suppressed during cooling in the weld pool. Upon reaching the critical temperature
indicated in Fig. 2.40, the carbon reaction is blocked, silicon and manganese now control the
oxygen level. An unknown but significant fraction of the manganese silicate inclusions precipitated in the hot part of the weld pool beneath the root of the arc are brought by convection
currents to the interface between the macroslag and the metal, where they are readily absorbed.
The remaining fraction formed in the cold and unstirred part of the weld pool is trapped in the
metal solidification front in the form of finely dispersed oxide particles. This results in a high
and rather unpredictable weld metal oxygen content.
The above reaction model has been tested experimentally against data obtained from a
series of hyperbaric welding experiments carried out in a remotely controlled pressure chamber with basic covered electrodes containing various levels of ferrosilicon in the electrode
coating (see Table 2.9). Welding under hyperbaric conditions offer the special advantage of
assessing the reactions through variations in the ambient pressure without changing the composition of the electrode coating or the core wire. Consequently, if the proposed reaction model

0.1 wt%Cat10bar.

0.1 wt% C at 1 bar

Temperature,C
Fig. 2.40. The break even equilibrium temperature for silicon control of oxygen level at 0.1 wt% C and
0.3 wt% Si. Data from Elliott et al.23

Table 2.9 Contents of ferrosilicon and iron powder in the electrode coating of experimental
consumables used in the hyperbaric welding experiments.
Electrode

FeSi (76 wt% Si)

Iron Powder

R*

4.5 wt%

31wt%

5.5 wt%

30 wt%

6.5 wt%

29 wt%

7.5 wt%

28 wt%

Carbon content, wt%

^Reference electrode (E8018-C1 type electrode).

Electrode R
A
B
C

Low FeSi levels

Total pressure, bar

Fig. 2.41. Carbon absorption in hyperbaric SMA welding. Data from Grong et al.51
is at least qualitatively correct would expect a correlation between the weld metal carbon
content and the concentrations of oxygen, silicon and manganese, both under atmospheric and
hyperbaric welding conditions. The main effect of pressure on weld metal chemistry is thus to
suppress the carbon-oxygen reaction in the weld pool at the expense of intensified oxidation
losses of silicon and manganese, as indicated by the thermodynamic data in Fig. 2.40.
2.103.2 Absorption of carbon and oxygen
It is evident from the data presented Fig. 2.41 that the weld metal carbon content increases
monotonically with pressure from 1 to 31 bar for all four electrodes involved. This indicates
that the carbon oxidation in the weld pool is systematically suppressed under hyperbaric welding conditions. Moreover, Fig. 2.41 reveals a small but important effect of electrode deoxidation
capacity on the weld metal carbon content. Since ferrosilicon itself is an insignificant source of
carbon, the observed increase in the carbon concentrations with increasing additions of
ferrosilicon to the electrode coating is an indication that carbon oxidation in the weld pool is
blocked at an earlier stage of the process at high silicon levels, according to the reaction:

Si (dissolved) + 2CO(gas) = 2 C (dissolved) + SiO2 (slag)

(2-54)

This interpretation is further supported by the results from the oxygen determination contained in Fig. 2.42. Although there is considerable scatter in the data in this figure, it is evident
that the recorded enhancement of the weld metal carbon content at high ferrosilicon levels in
the electrode coating is accompanied by a corresponding reduction in the oxygen concentrations.
2.10.3.3 Losses of silicon and manganese
Suppression of carbon oxidation in the weld pool at elevated pressures gives rise to intensified
oxidation losses of silicon and manganese, as shown in Figs. 2.43 and 2.44. Moreover, it is
apparent that increased additions of ferrosilicon to the electrode coating result in a corresponding increase in both the silicon and the manganese concentrations. This finding suggests that
the final weld metal content of the deoxidants is controlled by the reaction:
Si (dissolved) + 2MnO (slag) = 2 Mn (dissolved) + SiO2 (slag)

(2-55)

Assuming the activity ratio (tf Mn0 ) / (aSio2 )in precipitated slag to be constant and independent of pressure, equation (2-55) may be rewritten as:
(2-56)

Oxygen content, ppm

In Fig. 2.45 the weld metal manganese content has been plotted versus the square root of
the silicon content by inserting data from Figs. 2.43 and 2.44. As it appears from Fig. 2.45, the

Electrode R

A
B
C

Total pressure, bar


Fig. 2.42. Oxygen absorption in hyperbaric SMA welding. Data from Grong et al.51

Silicon content, wt%

Electrode R
A
B
C

Total pressure, bar

Manganese content, wt%

Fig. 2.43. Silicon oxidation in hyperbaric SMA welding. Data from Grong et al.51

Electrode R
A
B
C

High FeSi levels

Low FeSi levels

Total pressure, bar


Fig. 2.44. Manganese oxidation in hyperbaric SMA welding. Data from Grong et al.51
experimental data cluster around a straight line passing through the origin, which confirms
that the silicon and manganese concentrations are balanced by a reaction according to equation
(2-55).

Manganese content, wt%

Electrode R
A
B
C

1/2
[Silicon content, wt%]
Fig. 2.45. Correlation between weld metal manganese and silicon contents. Data from Grong et a/.51

2.103.4 The product [%C] [%O]


From steelmaking practice, the product [%C] [%O] is generally accepted as an adequate index
of the interaction between carbon and oxygen during the refining stage. This product is related
to the equilibrium content of dissolved carbon and oxygen in contact with carbon monoxide of
a partial pressure pco:
(2-57)
Here K5 is the equilibrium constant for reaction (2-53) (equal to about 2.0 X 10"3 at 16000C
and 2.6 X 10~3 at 20000C),23 Nco is the mole fraction of CO in the reaction product (equal to
the partial pressure of CO at 1 bar), andptotis the total ambient pressure.
During the initial stage of cooling in the weld pool, the oxygen content in an assumed
equilibrium with carbon would be expected to be higher than the analytical values. This situation applies in particular to welds made under hyperbaric conditions, where significant quantities of oxygen clearly are removed from the weld pool in the form of oxide inclusions after
the completion of the carbon oxidation.
The concentration of dissolved oxygen at the break even temperature for silicon control of
the oxygen level can be estimated from the measured concentration displacements of oxygen,
silicon and manganese in the weld deposit with increasing pressures, relative to 1 bar (designated A[%(9], A[%Si] and A[%Mrc], respectively). If the total amount of oxygen which reacts
with silicon and manganese at 1 bar, as a first approximation, is taken equal to the analytical
weld metal oxygen content, the following balance is obtained:

(2-58)

Next Page

Here [%0]eq. is the oxygen concentration in an assumed equilibrium with carbon at a given
pressure, and [%O]anaL is the analytical weld metal oxygen content. For this correlation, minor
vaporisation losses of manganese as well as possible reactions between oxygen and liquid iron
have been neglected.
In Fig. 2.46 the product m = [%C] [%O] is plotted vs the total ambient pressure. Calculations of m have been done both on the basis of [%O]anaL and [%O]eq. It can be seen from Fig.
2.46 that the former set of data (i.e. open symbols in the graph) cannot be represented by a
straight line passing through the origin, which should apply to a true equilibrium reaction.
However, when proper corrections are made for the amount of oxygen removed from the weld
pool after the completion of the carbon oxidation, such a correlation may be obtained as shown
by the solid line in Fig. 2.46. No clear effect of the electrode deoxidation capacity (i.e. ferrosilicon
content) on the product m = [%C] [%O] can be observed within the precision of measurements. This result is to be expected if the weld metal carbon content is controlled by a local
equilibrium with oxygen established at elevated temperatures in the weld pool. Also, inspection of the slope of the curve (i.e. heavy solid line in Fig. 2.46) indicates that the product
K5Nco is about 1.14 X 10~3 under the prevailing circumstances. If a reasonable average value
for the equilibrium constant K5 of 2.3 X 10~3 is assumed within the specific temperature range
of the reaction, we get:
Nco~0.5
(2-59)
The above calculations suggest that the controlling partial pressure of CO in the reaction
product is significantly lower than the ambient pressure under hyperbaric welding conditions.
This probably arises from an extensive infiltration of helium in the nucleating bubbles at the
slag/metal interface which, thermodynamically, will enhance the deoxidation capacity of carbon according to Le Chatelier's Principle. The conditions existing in hyperbaric SMA welding
thus appear to be similar to OBM/Q-BOP steelmaking, where simultaneous injection of oxygen and inert gas from the bottom of the convenor during the decarburisation stage results in a
steel carbon content which is typically below the value calculated for equilibrium between
oxygen and carbon at 1 atm partial pressure of CO.52

2.11 Weld Pool Deoxidation Reactions


During cooling, the metal concentrations established at high temperatures due to dissolution of
oxygen tend to readjust by precipitation of new phases. Accordingly, a supersaturation with
respect to the various deoxidation reactants initially increases and thus provides the driving
force for nucleation of oxides. Subsequently, the deoxidation reactions will proceed rapidly
through growth of nuclei above a critical size. Equilibrium conditions will finally establish the
limits for the degree of deoxidation that can be achieved.
In spite of the fact that large amounts of oxygen are removed from the weld pool during the
deoxidation stage, the analytical weld metal oxygen content exceeds by far the value predicted
from chemical thermodynamics, assuming that equilibrium conditions are maintained down to
the solidification temperature (see data in Table 2.5). This situation cannot be ascribed to a
large deviation from chemical equilibrium, but is mainly a result of an incomplete phase separation. Consequently, due consideration must be given to the kinetics. The three basic consecutive steps in steel deoxidation are shown in Fig. 2.47.

Previous Page

Here [%0]eq. is the oxygen concentration in an assumed equilibrium with carbon at a given
pressure, and [%O]anaL is the analytical weld metal oxygen content. For this correlation, minor
vaporisation losses of manganese as well as possible reactions between oxygen and liquid iron
have been neglected.
In Fig. 2.46 the product m = [%C] [%O] is plotted vs the total ambient pressure. Calculations of m have been done both on the basis of [%O]anaL and [%O]eq. It can be seen from Fig.
2.46 that the former set of data (i.e. open symbols in the graph) cannot be represented by a
straight line passing through the origin, which should apply to a true equilibrium reaction.
However, when proper corrections are made for the amount of oxygen removed from the weld
pool after the completion of the carbon oxidation, such a correlation may be obtained as shown
by the solid line in Fig. 2.46. No clear effect of the electrode deoxidation capacity (i.e. ferrosilicon
content) on the product m = [%C] [%O] can be observed within the precision of measurements. This result is to be expected if the weld metal carbon content is controlled by a local
equilibrium with oxygen established at elevated temperatures in the weld pool. Also, inspection of the slope of the curve (i.e. heavy solid line in Fig. 2.46) indicates that the product
K5Nco is about 1.14 X 10~3 under the prevailing circumstances. If a reasonable average value
for the equilibrium constant K5 of 2.3 X 10~3 is assumed within the specific temperature range
of the reaction, we get:
Nco~0.5
(2-59)
The above calculations suggest that the controlling partial pressure of CO in the reaction
product is significantly lower than the ambient pressure under hyperbaric welding conditions.
This probably arises from an extensive infiltration of helium in the nucleating bubbles at the
slag/metal interface which, thermodynamically, will enhance the deoxidation capacity of carbon according to Le Chatelier's Principle. The conditions existing in hyperbaric SMA welding
thus appear to be similar to OBM/Q-BOP steelmaking, where simultaneous injection of oxygen and inert gas from the bottom of the convenor during the decarburisation stage results in a
steel carbon content which is typically below the value calculated for equilibrium between
oxygen and carbon at 1 atm partial pressure of CO.52

2.11 Weld Pool Deoxidation Reactions


During cooling, the metal concentrations established at high temperatures due to dissolution of
oxygen tend to readjust by precipitation of new phases. Accordingly, a supersaturation with
respect to the various deoxidation reactants initially increases and thus provides the driving
force for nucleation of oxides. Subsequently, the deoxidation reactions will proceed rapidly
through growth of nuclei above a critical size. Equilibrium conditions will finally establish the
limits for the degree of deoxidation that can be achieved.
In spite of the fact that large amounts of oxygen are removed from the weld pool during the
deoxidation stage, the analytical weld metal oxygen content exceeds by far the value predicted
from chemical thermodynamics, assuming that equilibrium conditions are maintained down to
the solidification temperature (see data in Table 2.5). This situation cannot be ascribed to a
large deviation from chemical equilibrium, but is mainly a result of an incomplete phase separation. Consequently, due consideration must be given to the kinetics. The three basic consecutive steps in steel deoxidation are shown in Fig. 2.47.

[%C][%O]x102

Electrode R
A
B
C

Total pressure, bar

Separation Stage
Buoyancy
Surface tension
Stirring

Growth Stage
Diffusion of reactants in the
melt to the oxide nuclei
Particle coalescence

Nucleation Stage
Homogeneous nucleation
Heterogeneous nucleation

Average particle size

Fig. 2.46. The product [%C][%0] in hyperbaric SMA welding. Solid symbols: calculations based on
[%O]eq.. Open symbols: calculations based on [%O]anai.- Data from Grong et al. 51

Reaction time

Fig. 2.47. The three major consecutive steps in steel deoxidation (schematic).

Although rate phenomena in ladle refining of liquid steel are extensively investigated and
reported in the literature,53"55 only recently attempts have been made to include such effects in
an analysis of deoxidation reactions in arc welding. 1 ' 56 ' 57
2.11.1 Nucleation of oxide inclusions
During ladle-refining of liquid steel, it is well established that homogeneous nucleation of
oxide inclusions may occur in certain regions of the melt where the supersaturation is sufficiently high.55'58 Over the composition range normally applicable to deoxidation of steel, the
number of nuclei formed at the time of addition of deoxidisers is approximately 105 mirr 3 . 55
However, in steel weld metals, the number of oxide nuclei formed during the initial stage of
deoxidation must be considerably higher to account for the observed inclusion number density
of about 107 mm~3 to 108 mm"3.56'57 This implies that the supersaturation established in the
weld pool on cooling as a result of rapid temperature fluctuations (~103 0C s"1) exceeds by far
that obtained by additions of deoxidisers to a liquid steel melt under approximately isothermal
conditions.
There are several theories available for treating nucleation phenomena, but for the purpose
of convenience, a simplified version of the model of Turpin and Elliott58 has been adopted
here. Consider a steel melt which is brought to a state of supersaturation by first equilibrating
it with pure MxOy at a high temperature T\ and then rapidly cooling it to a lower temperature
T2, as shown schematically in Fig. 2.48. It follows from the classical theory of homogeneous
nucleation that the required temperature difference T1-T2 necessary to achieve spontaneous
precipitation of MxOy is approximately given by the following relationship:

(2-60)
where AH is the standard enthalpy of reaction, G is the oxide-steel interfacial energy (assumed to be constant and independent of temperature), and Vm is the molar volume of the
nucleus.
The derivation of equation (2-60) is shown in Appendix 2.2.
Example (2.9)

Assume that precipitation of manganese silicates in the weld pool occurs according to the
following reaction:
Si (dissolved) + Mn (dissolved) + 3 O (dissolved) = MnSiO3 (slag)

(2-61)

where
AG(J) = 858620 + 3457.
Based on equation (2-60), calculate the critical temperature interval of subcooling for homogeneous nucleation OfMnSiO3. Typical physical data for liquid steel and manganese silicate slags are given in Table 2.10.

Time

Fig. 2.48. Idealised model for homogeneous nucleation of oxide inclusions in steel weld metals (schematic).
Table 2.10 Physical data for liquid steel and manganese silicate slags at 16000C. Data from Refs. 3 and
53.

Property

Density
(kgnr 3 )

Viscosity
(kgm^s- 1 )

Steel

6900

4.3 X 10~3

Silicate slag

2300

Interfacial Energyt
(Jm" 2 )

0.8

In contact with liquid steel.

Solution

First we estimate the molar volume of the nucleus:

By inserting the appropriate values for Vm, AH, and a in equation (2-60), we obtain:

from which

If we assume that the supersaturation is released at T2 = 16000C (1873K), the initial temperature of the liquid T1 becomes:

The critical temperature interval of subcooling is thus:

Critical temperature interval, AT (0C)

Similar calculations can also be carried out for other types of oxide inclusions, e.g. FeO(I),
SiO2(s), and Al2O3(S). The results of such computations are presented graphically in Fig. 2.49,
using data from Refs 55 and 58. It is evident from these plots that the critical temperature
interval of subcooling depends on the interfacial energy, G. Although data for oxide-steel
interfacial energies are scarce, the following average values are frequently used in the literature, 5558 i.e. a(FeO-Fe) = 0.3 J m"2; G(SiO2-Fe) = 0.9 J m~2; and G(Al2O3-Fe) - 1.5 J mr2. If
these values are accepted, the results in Fig. 2.49 indicate that the critical temperature interval
of subcooling for homogeneous nucleation of FeO (1), SiO2(s), and Al2O3(S) is of the order of
200 to 30O0C. Considering the fact that the liquid weld metal spans a temperature range of
about 2200 to 15000C,34 it is not surprising to find that nucleation of oxide inclusions occurs
readily in the weld pool during cooling. It should be noted that the quoted data for G are
representative of ladle-refined steel deoxidised at 16000C. At higher metal temperatures, in the
presence of large amounts of dissolved oxygen, the oxide-steel interfacial energies would be
expected to be significantly lower.59 Hence, it is reasonable to assume that the actual temperature interval of subcooling required for spontaneous oxide precipitation in a weld pool is well
below 2000C.

Interfacial energy,o(J/m2)
Fig. 2.49. Critical temperature interval of subcooling for homogeneous nucleation of oxide inclusions in
steel weld metals at 16000C as a function of the interfacial energy a.

2.11.2 Growth and separation of oxide inclusions


In practice, there are three major growth processes in steel deoxidation:55

(i) Collision
(ii) Diffusion
(iii) Ostwald ripening.
From deoxidation and ladle refining of liquid steel it is well established that the flotation
rate of the oxides generally depends on their growth rates, since large inclusions separate
much more rapidly than small ones, in agreement with Stokes law.55 Growth of the oxides can
proceed either through diffusion of reactants in the melt to the oxide nuclei or by collision and
coalescence of ascending inclusions, and is therefore influenced by factors such as the number
density of the nuclei, interfacial tensions, and the extent of melt stirring.53"55 The last factor is
of particular importance in welding, because the stirring action will increase the possibilities
for collision and coalescence of inclusions and, depending on the direction of flow, can give
rise to circulation of inclusion-laden metal to the surface. As a result, the separation of small
oxide particles, i.e. microslag, is strongly favoured by the turbulent conditions existing in the
hot part of the weld pool immediately beneath the root of the arc.
2.11.2.1 Buoyancy (Stokes flotation)
Assessment of the role of buoyancy (Stokes flotation) in the separation of deoxidation products from the weld pool can be done on the basis of the experimental data of Grong et ai56
reproduced in Fig. 2.50. It is evident that a change in the welding position (i.e. from flat to
overhead position) has no significant effect on the weld metal oxygen content. This shows that
the buoyancy effect does not play an important role in the separation process of oxide inclusions during welding.

Oxygen content, wt%

Upward welding
Downward welding

Horizontal

Welding orientation, degrees


Fig. 2.50. Effect of welding position on retained oxygen in GMA steel weld deposits. Data from Grong et
a/. 56

The above conclusion is also in agreement with predictions based on the Stokes law, which
gives the terminal velocity of the ascending particles (u) relative to the liquid:55
(2-62)
where g is the gravity constant, dv is the particle diameter, Ap is the difference in densities
between the liquid steel and the inclusions, and |x is the steel viscosity.
Taking the Stokes parameter, gAp/18|ji, equal to 0.6 jjinr1 s"1 for manganese silicate slags in
steel (Table 2.10), equation (2-62) becomes:
(2-63)
Generally, the majority of non-metallic inclusions in steel weld metals are of a diameter
below 2 jxm.56'57 According to equation (2-63), such particles have a relative velocity less than
2.4 |xm s"1. This implies that the buoyancy effect alone is far too insignificant to promote
flotation of the inclusions out of the weld pool before solidification, when it is recognised that
the average fluid flow velocity in the weld pool is four to five orders of magnitude higher (to
be discussed later).
Stokes law is based on the assumption that the inclusions are completely wetted by the
liquid steel, i.e. there is no slip at the oxide/metal interface.55 Normally, interfacial tension
effects promote slip at the particle/metal interface, which, in turn, enhances the flotation rate
of the ascending oxides (often referred to as the Plockinger effect).60 The concept of a wetting
angle has been used in this context. But, the slip phenomenon is probably perceived better in
terms of a secondary flow in the interfacial region between the liquid metal and the solid
precipitate, which is produced by gradients in interfacial tension. In the rare case of no wetting
(0 = 180), it has been shown that the average terminal velocity of the inclusions is approximately 50% higher than that given by equation (2-62).55 For silica slags, the wetting angle (in
the presence of air) is close to 115,54 indicating that the correction in the particle terminal
velocity due to the slip at the oxide/metal interface in less than about 30%. Consequently,
interfacial tension effects between the slag and the steel do not significantly affect the flotation
rate of the particles, and, therefore, can be ignored.
2.11.2.2 Fluid flow pattern
It is evident from the above discussion that the separation of deoxidation products during arc
welding is controlled by the fluid flow fields set up within the weld pool. The typical flow
pattern in a submerged arc weld pool is shown in Fig. 2.51. It can be seen from the figure that
a depression is formed at the forward edge of the pool, which forces the melted metal to flow
underneath and on either side of the depression, following the arrows in Fig. 2.51. At the rear
of the weld pool the flow direction is reversed, and the metal streams back along the pool
surface. A similar flow pattern has also been observed in GMA weld pools.61 As a result, the
inclusions precipitated in the turbulent part of the weld pool are rapidly brought to the upper
surface behind the arc due to the high-velocity flow field created within the liquid metal, and
disengaged by the surface tension effects in the pool.
The drag force exerted on the particles because of the liquid flow-field velocity can be
estimated from published data for weld pool flow velocities following electromagnetic fluid
mechanics theory. Normally, for weld pools the Peclet number for heat transfer lies within the

range from 10 to 5000, which indicates that the heat flow is predominately convectional.62 The
limits quoted above for the Peclet number correspond to a range in the average weld pool fluid
velocity from about 0.025 to 0.4 m s"1 in the case of SA welding.6364 Unfortunately, specific
information about the fluid flow velocity gradients in the weld pool are lacking, which prevents a more complete analysis of variations in the flow pattern with increasing distance from
the root of the arc.
Generally, the drag force, Fd, acting on a spherical particle in relative motion to a fluid can
be expressed as:3
(2-64)
where Cd is the drag coefficient, p is the fluid density, u is the bulk velocity of the fluid
relative to the particle, and dv is the particle diameter. The drag coefficient is, in turn, a function of the particle Reynolds number, A^,:3
(2-65)
where |x is the fluid viscosity. Typical physical-property data for liquid steel are given in Table
2.10. By inserting the values for p and JJL from Table 2.10, then using the relative velocity
calculated in equation (2-63) (2.4 |xm s"1) as one extreme and the fluid flow velocity (0.4 m s"1) as
the other (i.e. assuming a stationary particle), one can demonstrate that the particle Reynolds
numbers obtained never exceed 1.3 for most weld metal inclusions. The limit calculated above
for the Reynolds number is within the so-called creeping flow region where Stokes law is
indeed valid. Under such conditions, the ratio between the particle drag force Fd and the corresponding gravity force Fg is given by the expression:3
(2-66)
Note that equation (2-66) is the basis for obtaining Stokes law (equation (2-62)).
For steady motion of the particle, Fd and Fg are, of course, equal. However, the significance
of Fd can also be interpreted in a transient sense, by considering the limiting case where the
particle is stationary and its instantaneous motion is governed by the forces acting on it. In this
case, the ratio of Fd (max) acting on the particle relative to the net gravity force (given by
Electrode

Pulsating cavity

Base plate

Fig. 2.51. Typical flow pattern in SA weld pools. For clarity, the arc, slag and flux have been omitted.
The sketch is based on the ideas of Lancaster.32

equation (2-66)) can determine the dominant force responsible for the particle's trajectory. In
Fig. 2.52 the above ratio has been calculated and plotted against the particle diameter, dv, for
values of u' equal to 0.025 and 0.4 m s"1, which represent the typical velocities reported for
SA steel weld pools. 63 ' 64 It can be seen from the figure that the drag force is always several
orders of magnitude greater than the gravity force for particles within the typical size range of
weld metal inclusions (i.e. less than 2 |xm in diameter). This result is to be expected, since the
relative velocity, based on Stokes law and calculated in equation (2-63), is negligible compared with the liquid velocities in the weld pool (2.4 \xm s"1 vs 0.02 to 0.4 m s"1). This calculation thus supports our previous conclusion that the separation of the oxide inclusions is controlled solely by the fluid flow behaviour in the weld pool. The fact that the phase separation
proceeds under strongly turbulent conditions is also evidenced by the large number of iron
droplets being mechanically dispersed in the top bead slag of GMA steel welds, as shown by
the optical micrographs in Fig. 2.53.
In the case of GMA welding, the non-metallic inclusions that are brought to the upper
surface behind the arc coalesce rapidly to form large slag clusters that float on the top of the
bead. Generally, re-entrapment of the slag does not occur owing to the decrease in the total
surface free energy of the system, which is caused by the emergence of the inclusions from the
weld metal.54 Consequently, the slag will remain floating on the top of the bead even when
welding is performed in the overhead position, as shown previously in Fig. 2.50.

*ci(max/Fg

2.11.2.3 Separation model


Based mainly on experience with the GMA welding process, a simple model for the assessment of the sequence of deoxidation reactions in arc welding has been proposed,1 and is shown
schematically in Fig. 2.54. The model is based on the assumption that equilibrium between the
reacting elements and precipitated slag is maintained down to low metal temperatures, and
divides the weld pool into the following two main reaction zones:

Inclusion diameter, |i.m


Fig. 2.52. The ratio between maximum particle drag force, Fornax) and the corresponding gravity force,
F8, vs the inclusion diameter at two different flow velocities in the weld pool. Calculations are based on
equation (2-66).

(a)

(b)

Fig. 2.53. Optical micrographs showing characteristic "comet's tails" of trapped iron droplets (light
areas) in collected top bead slags of GMA steel welds; (a) Ar + 5vol%O2, (b) Ar + 5vol%CO2.

Electrode
(a)

Cold part of weld pooN

Arc

Hot part of weld pool


Base plate

S e c t i o n A-A

Cold part of weld pool: Hot part of weld pool:


Deoxidation/phase
Deoxidation/incomplete
separation
phase separation

Temperature

(b)

Final weld metal,


oxygen concentration

Oxygen content
Fig. 2.54. Schematic diagrams showing the sequence of reactions occurring during weld metal deoxidation;
(a) Longitudinal section of weld pool, (b) Cross section of weld pool along A-A. The diagrams are based
on the ideas of Grong and Christensen.1

(i) The hot part of the weld pool, characterised by simultaneous oxidation and deoxidation
of the metal, where the separation of microslag takes place continuously as a result of highly
turbulent flow conditions.
(ii) The cold part of the weld pool, where precipitated slag will largely remain in the metal
as finely dispersed particles as a result of inadequate melt stirring.
Under such conditions equation (2-23) predicts a direct correlation between absorbed and
retained (analytical) oxygen in the weld metal, i.e.:

(2-67)
in agreement with experimental observations (see plots in Figs. 2.33, 2.38, and 2.39). Typically, the proportionality constant C4 varies between 0.1 to 0.2, which corresponds to a range
in the t/to ratio from 2.3 to 1.6. This shows that the boundary between 'hot' and 'cold' parts of
the weld pool is not well defined, but depends on the welding system under consideration as
well as on the operational conditions applied.
2.11.3 Predictions of retained oxygen in the weld metal
Although the weld metal oxygen content is controlled by a transport mechanism in the weld
pool, the concept of pseudo-equilibrium can still be used for an assessment of slag/metal reactions in arc welding.
2.11.3.1 Thermodynamic model
In the case of silicon deoxidation of steel weld metals, we may write:
Si (dissolved) + 2 O (dissolved) = SiO2 (slag)

(2-68)

On introduction of the equilibrium constant for the reaction, we obtain:


(2-69)
To allow for the decrease in the silica activity with increasing manganese-to-silicon ratios,
it is essential to establish a correlation that links the activity of silica to the concentrations of
the deoxidation elements in the weld metal. A semi-empirical corr61ation of this kind has been
presented by Walsh and Ramachandran,65 derived from a re-analysis of activity data for silica
in the Fe-Mn-Si-O system previously published by Hilty and Crafts.66 Within the temperature range from 1550 to 16500C, they showed that the silica activity in the deoxidation product
can be approximately expressed as:
(2-70)
where K1 represents the manganese-to-silicon ratio at which the activity of silica becomes

unity for a given temperature. A check of this equation against more recent data for silica
activities in MnO-SiO2 slags reported by Turkdogan67 supports the findings of Walsh and
Ramachandran65 that the activity of silica is approximately given by equation (2-70) for a wide
range in the steel manganese-to-silicon ratio (i.e. from 0.1 to 50). By combining equations (269) and (2-70), it is possible to obtain an expression for the equilibrium oxygen content, solely
in terms of the silicon and manganese concentrations:
(2-71)
where K^ is a temperature-dependent parameter equal to (K1I K6)05.
The temperature dependence of the Si-O reaction (equation (2-68)) is well established and
is approximately given by the relationship:23
(2-72)
when pure SiO2 is used as the standard state for the silica activity.
For a rough estimate of the temperature dependence of equation (2-70), the results of
Turkdogan55 can be used. It should be noted that Walsh and Ramachandran65 did calculate the
temperature dependence of K1 within the range from 1550 to 16500C. However, because equation (2-70) is empirical, the function cannot be extrapolated beyond these temperature limits.
The data quoted in Ref. 55 are derived directly from the Si-Mn reaction (equation (2-55)) and
activity data for MnO at silica saturation. On introduction of the equilibrium constant for
equation (2-55), we obtain:
(2-73)
By using data from Ref.55, the initial [%Mn]2/[%Si] ratio for precipitation of silica saturated slags (equal to K9(aMnO)2) at 1500,1550,1600 and 16500C has been recorded and replotted
against temperature, as shown in Fig. 2.55. The figure shows that the critical [%Mn]2/[%Si]
ratio for precipitation of silica saturated slags is temperature dependent and decreases from
about 5 at 15000C to below 1.5 at 16500C. On the basis of a crude extrapolation of the data to
higher metal temperatures, it can, however, be argued that the ratio would approach a constant
value of approximately 0.75 at temperatures beyond about 17500C (indicated by the broken
horizontal line in Fig. 2.55). The above observation reflects the fact that the Si-Mn-O reaction
equilibrium (equation (2-55)) is not very sensitive to a change in temperature (i.e. the enthalpy
for the reaction is small). Over the composition range normally applicable to Si-Mn deoxidation
of steel weld metals the observed threshold for the critical [%Mn]2/[%Si] ratio for precipitation of silica-saturated slags in Fig. 2.55 corresponds to a manganese-to-silicon ratio closely
equal to unity. Consequently, at temperatures higher than about 17500C, K1 can, as a first
approximation, be taken constant and independent of temperature (i.e. K1- Y). The temperature dependence of equation (2-71) is thus simply (1/^ 6 ) 05 or:
(2-74)

for temperatures higher than about 17500C (2023K).

[%Mn]2/[%Si]

Temperature, 0C
Fig. 2.55. The critical [%Mn]2/ [%Si] ratio for precipitation of silica-saturated slags as a function of
temperature. Data from Turkdogan.55

2.11.3.2 Implications of model


Figure 2.56 shows plots of retained (analytical) oxygen in GMA and SMA steel weld metals vs
the theoretical deoxidation parameter ([%Si][%Mn])~25, using relevant literature data. 5668
This parameter allows for the inherent decrease in the silica activity with increasing weld
metal Mn to Si ratios. A closer inspection of the slopes of the curves reveals that the effective
reaction temperature falls within the range calculated for chemical equilibrium between silicon, manganese, and oxygen at 1800 to 19000C. Although the boundary between 'hot' and
'cold' parts of the weld pool for possible inclusion removal in practice is not sharp, the results
in Fig. 2.56 literally suggest that all oxides precipitated above 1800 to 19000C are simultaneously removed from the weld pool under the prevailing turbulent flow conditions. At lower
temperatures, the degree of melt stirring is too low to promote separation of the deoxidation
products out of the pool before solidification and hence, they are trapped in the metal solidification front in the form of finely dispersed inclusions. Moreover, the observed difference in
the effective reaction temperature between GMA and SMA welding supports our previous
conclusion that any change in welding parameters, flux or shielding gas composition, which
alters the fluid flow pattern in the weld pool, will often have a stronger influence on the weld
metal oxygen content than variations in the deoxidation practice. This implies that control of
the weld metal oxygen level through additions of deoxidants, in practice, is difficult to achieve.

2.12 Non-Metallic Inclusions in Steel Weld Metals


Inclusions commonly found in steel weldments will either be exogenous or indigenous, dependent on their origin. The first type arises from entrapment of welding slags and surface
scale, while indigenous inclusions are formed within the system as a result of deoxidation
reactions (oxides) or precipitation reactions (nitrides, sulphides). The latter group is almost
always seen to be heterogeneous in nature both with respect to chemistry (multiphase parti-

Oxygen content, wt%

GMA Welding
SMA Welding

([%SI] [%Mn]f"25
Fig. 2.56. Examples of pseudo-equilibrium in GMA and SMA welding of C-Mn steels. The solid lines in
the graph represent thermodynamical calculations at indicated temperatures. Data from Refs. 56 and 68.
cles), shape (angular or spherical particles), and crystallographic properties as a result of the
complex alloying systems involved69 (see Fig. 2.57). An exception may be C-Mn steel welds,
where the oxide inclusions will be predominately glassy, spherical, manganese silicates.1
A survey of important weld metal inclusion characteristics is given in Table 2.11.
2.12.1 Volume fraction of inclusions
It is evident from Table 2.11 that the volume fraction of non-metallic inclusions in steel weld
metals normally falls within the range from 2 X 10~3 to 8 X 10~3, depending on the type of
weld under consideration. Based on simple stoichiometric calculations it is possible to convert
the analytical weld metal oxygen and sulphur contents to an equivalent inclusion volume fraction when the chemical composition of the reaction products is known. This is shown below.
Table 2.11 Summary of weld metal inclusion characteristics. The data are compiled from miscellaneous sources.
Size Distribution*

Chemical Composition

Type of Weld

C-Mn steel
weld metals
Low-alloy
steel weld
metals

Constituent
elements

Reported
phases

10-50

Si, Mn, O, S
(traces of Al,
Ti, and Cu)

SiO2,
MnOSiO2,
MnS, (CuxS)

10-40

Al, Ti, Si,


Mn, O, S, N
(Cu)

MnOAl2O3,
7-Al2O35TiN,
MnOSiO2, SiO2,
a-MnS, (3-MnS,
(Cu^S)

Vv X 10"3

dv

NvX 107

Sv

3-8

0.3-0.6

1-10

2-6

0.3-0.7

0.5-5

*Vy: volume fraction; dv: arithmetic mean (3-D) particle diameter (jxm); Nv: number of particles per unit volume
(mm"3); Sv\ total particle surface area per unit volume (mm2 per mm3).

Fig. 2.57. Digital STEM brightfield image and Si, Al, S, Mn and Ti X-ray images of a multiphase weld
metal inclusion. After Kluken and Grong.57

Solution

First we calculate the total weight of retained MnOSiO2 per 100 g weld metal:

This corresponds to an equivalent volume fraction of:

Similarly, we can calculate the weight and volume fraction of MnS in the weld metal:

and

The total volume fraction of MnOSiO2 and MnS is thus:

In practice, the stoichiometric conversion factors for oxygen and sulphur are virtually constant for a wide spectrum of inclusions70 and hence, they can be regarded as independent of
composition. Taking the solubility of sulphur in solid steel equal to 0.003 wt%, the following
relationship is obtained for steel weld metals:57'70
(2-75)
The validity of equation (2-75) has been confirmed experimentally by comparison with
microscopic assessment methods.5771
In steel weld metals the majority of the inclusions will be in the submicroscopic range
owing to the limited time available for growth of the oxides. From the histogram in Fig. 2.58 it
is seen that particles with diameters between 0.3 to 0.8 |xm contribute to nearly 50 percent of
the total inclusion volume fraction. This trend is not significantly changed by additions of
strong deoxidisers, such as aluminium and titanium, or by a moderate increase/decrease in the
heat input.57
2.12.2 Size distribution of inclusions
As shown in Fig. 2.59 the majority of the three-dimensional (3-D) inclusion diameters fall
within the range of 0.05 to 1.5 |xm, with a characteristic peak in the particle frequency at about
0.4 to 0.5 |jim. These data obey the log-normal law, since a plot of the frequencies against the
logarithms of the diameters approximately gives a symmetrical curve. Considering specific
inclusion size classes, deoxidation with aluminium generally results in a higher fraction of
coarse particles (> 1 |xm) due to incipient clustering of Al2O3.57' 69 However, the observed
particle clustering has no significant influence on the arithmetic mean 3-D inclusion diameter,
as shown by the data in Fig. 2.59.

Relative volume fraction (%)

Particle diameter (jim)


Fig. 2.58. Percental contribution of different size classes to the total volume fraction of non-metallic
inclusions in a low-alloy steel weld metal. Data from Kluken and Grong.57

2.12.2.1 Effect of heat input


In contrast to the situation described above, the 3-D inclusion size distribution is strongly affected an increase in the heat input (see Fig. 2.60). At 1 kJ mm"1, the measured 3-D inclusion
diameters fall within the range from 0.05 to 1 |xm, with a well-defined peak in the particle frequency at about 0.3 |xm. When the heat input is increased to 8 kJ mm"1, the content of coarse
inclusions will dominate (>0.5 |xm), which results in a broader distribution curve and a shift
in the peak frequency towards larger particle diameters. A comparison with Fig. 2.61 reveals
that the arithmetic mean 3-D inclusion diameter is approximately a cube-root function of the
heat input. This result is to be expected if Ostwald ripening is the dominating coarsening
mechanism in the cold part of the weld pool (to be discussed below).
2.12.2.2 Coarsening mechanism
As already mentioned in Section 2.11.2 there are three major growth processes in steel deoxidation, i.e. (i) collision, (ii) diffusion, and (iii) Ostwald ripening. In the cold part of the weld
pool, particle growth by collision can be excluded in the absence of adequate melt stirring because of a low collision probability of inclusions while ascending in the molten steel within the
regime of Stokes law.72 In addition, the diffusion-controlled part of the deoxidation reaction
(which involves diffusion of reactants in the melt to the oxide nuclei) would be expected to
be essentially complete within a fraction of a second when the number of nuclei is greater
than 107 mm"3.55 This implies that the observed increase in the inclusion diameter with increasing heat inputs (Fig. 2.61) can be attributed solely to Ostwald ripening effects.
Before discussing details of the inclusion growth kinetics, it is essential to clarify the temperature level in the 'cold' part of the weld pool. As shown by the results in Fig. 2.62, the liquid
metal temperature in the trailing edge of the weld pool is fairly constant and slightly above
the melting point of the steel. Accordingly, inclusion growth in welding (at a fixed volume fraction) can be treated as an isothermal process, where the time dependence of the mean particle diameter dv is approximately given by the Wagner equation:

(a)

Frequency (%)

Low Al (0.018 wt%)


Low Ti (0.005 wt%)

Particle diameter (fxm)

(b)
High Al (0.053 wt%)
Frequency (%)

High Ti (0.053 wt%)

Particle diameter, p,m


Fig. 2.59. Three-dimensional (3-D) size distribution of non-metallic inclusions in two different lowalloy steel weld metals; (a) Low weld metal aluminium and titanium levels, (b) High weld metal aluminium and titanium levels. Data from Kluken and Grong.57

(2-76)

Here do is the initial particle diameter, a is the oxide-steel interfacial energy, Dm is the
element diffusivity, Cm is the element bulk concentration, V'm is the molar volume of the oxide
per mole of the diffusate, and t is the retention time.

Frequency (%)

Particle diameter, p,m

Particle diameter, u,m

Fig. 2.60. Effect of heat input on the 3-D inclusion size distribution in low-alloy steel weld metals. Data
from Kluken and Grong.57

Heat input, kJ/mm

Fig. 2.61. Variation of arithmetic mean 3-D inclusion diameter with heat input during SA welding. Data
from Kluken and Grong.57
For welding of thick plates, the time available for growth of particles in the 'cold' part of
the weld pool can be estimated from the Rosenthal equation, i.e. equation (1-45) in Section
1.10.2 (Chapter 1). If the characteristic length of the cooling zone is taken equal to the weld
ripple lag (defined in Fig. 2.63), the retention time t is approximately given by the following
relationship:

Temperature, C

"Tpeak = 1538C (about 8 mm from edge of weld pool)

Time, seconds
Fig. 2.62. Measured temperature level in the trailing edge of the weld pool during GMA welding. Data
from Kluken and Grong.57

Top view of weld crater (z = 0)

Max. width

Retention time
Weld ripple lag
Welding speed
Heat source

Fusion boundary

Fig. 2.63. Definition of weld ripple lag xm, and retention time t.

(2-77)
where J\ is the arc efficiency (equal to about 0.95 for SAW and 0.80 for GMAW /SMAW), and
E is the gross heat input (kJ mm"1). Note that equation (2-77) assumes constant values for the
steel thermal diffusivity and volume heat capacity (5 mm2 s"1 and 0.0063 J mm"3 0C"1, respectively), and no preheating (T0* = 200C).
In Fig. 2.64 the Ostwald ripening theory has been tested against relevant literature data,
which may be considered representative of the 3-D particle size distribution. Although there is
some scatter in the data, the observed inclusion growth rates fall within the range calculated
for oxygen diffusion-controlled coarsening of SiO2 and Al2O3 at 15500C, using the Wagner
equation. In these calculations, a reasonable average value for the bulk diffusivity of oxygen
has been assumed (i.e. 10~2 mm2 s"1).55 If the effective inclusion growth rate constant for lowalloy steel weld metals is taken equal to the slope of the curve in Fig. 2.64, the following
relationship is achieved:

GMAW
SAW

dv,jLim

SMAW

1/3
i i sJ/3
Fig. 2.64. Relation between arithmetic mean (3-D) inclusion diameter dv and retention time t for different
arc welding processes. Data compiled by Kluken and Grong.57
CoId1 part of
weld pool

Deoxidation
Phase separation
Deoxidation
Incomplete phase;
separation
Homogeneous
nucleation

Growth of
inclusions

Oxidation

Solid weld metal

Temperature

Gas/metalslag/metal
reactions

Solid weld metal

'Hot' part of
weld pool

Ostwald ripening

Time
Fig. 2.65. Proposed deoxidation model for steel weld metals (schematic). The diagram is based on the
ideas of Kluken and Grong.57

(2-78)
By substituting equation (2-77) into equation (2-78), we obtain a direct correlation between
the arithmetic mean 3-D inclusion diameter dv and the net heat input T\E:
(2-79)

Equation (2-79) predicts that dv is a simple cube root function of E, in agreement with the
experimental data in Fig. 2.61.
It should be noted that the measured shape of the particle distributions (see Figs. 2.59 and
2.60) deviates somewhat from that required by the Wagner equation, which assumes a quasistationary distribution curve, and that the maximum stable particle diameter is about 1.5 times
the mean diameter of the system.73 Although the origin of this discrepancy remains to be
resolved, this suggests that particle clustering is also a significant process in steel weld metals
as it is in other metallurgical systems. In fact, such effects would be expected to be most
pronounced at high aluminium levels because of a large interfacial energy between Al2O3 and
liquid steel, in agreement with experimental observations.57'69
2.12.2.3 Proposed deoxidation model
Referring to Fig. 2.65, the sequence of reactions occurring during weld metal deoxidation can
be summarised as follows. In general, nucleation of oxide inclusions occurs homogeneously
as a result of the supersaturation established during cooling in the weld pool. The diffusioncontrolled deoxidation reactions (i.e. diffusion of reactants to the oxide nuclei) will be essentially complete when the liquid metal temperature attains a constant level of about 15500C at
some distance behind the root of the arc. Growth of the particles may then proceed under
approximately isothermal conditions at a rate controlled by the Wagner equation until the
temperature reaches the melting point of the steel. Since retention times in welding generally
depend on the heat input, it follows that choice of operating parameters will finally determine
the degree of particle coarsening to be achieved.
Example (2.11)

Consider SA welding with a basic flux on a thick plate of low-alloy steel under the following
conditions:

Previous experience has shown that this steel/flux combination gives a weld metal oxygen
and sulphur content of 0.035 and 0.008 weight percent, respectively. Based on the stereometric
relationships given below, calculate the total number of particles per unit volume Nv, the total
number of particles per unit area Na, the total particle surface area per unit volume Sv, and the
mean particle centre to centre volume spacing Xv in the weld deposit:74'75

(2-80)
(2-81)

(2-82)

(2-83)

Solution

First we calculate the total volume fraction of oxide and sulphide inclusions from equation
(2-75):
Vv = 10-2 [5.0(0.035) + 5.4(0.008 -0.003)] = 2.0 X 10~3
The arithmetic mean (3-D) inclusion diameter can then be evaluated from equation (2-79):

This gives:
particles per mm3

particles per mm 2
mm 2 per mm3

particles per mm3

particles per mm 2
mm 2 per mm3

2.123 Constituent elements and phases in inclusions


It is evident from Table 2.11 that non-metallic inclusions commonly found in low-alloy steel
weld metals may contain a considerable number of constituent elements and phases.
2.12.3.1 Aluminium, silicon and manganese contents
Figure 2.66 shows examples of measured X-ray intensity histograms for silicon, manganese,
and aluminium in inclusions extracted from a low-alloy steel weld metal. These results have
been converted into arbitrary elemental weight concentrations by normalising the collected Xray data for all preselected elements (i.e. Al, Si, Mn, Ti, Cu, and S) to 100%. The characteristic
normal distribution curves recorded for silicon, manganese, and aluminium show that the content of each oxide phase may vary significantly from one particle to another. This observation
is not surprising, considering the complex chemical nature of the weld metal inclusions (see
Fig. 2.57).
2.12.3.2 Copper and sulphur contents
In addition, the inclusions may contain significant levels of both copper and sulphur in addition to aluminium, silicon, manganese and titanium. Sulphide shells around extracted inclusions have frequently been observed in SA and SMA steel welds, often in combination with
copper. This has been taken as an indication of copper sulphide formation.
However, based on the wavelength dispersive X-ray (WDX) data reported by Kluken and
Grong,57 it can be argued that copper sulphide is a rather unlikely reaction product in steel
weld metals as it is in ordinary ladle-refined steel.76 From their data it is evident that the total

Silicon

Manganese

Aluminium

Frequency (%)

26 wt%

20 wt%

28 wt%

Arbitrary elemental weight concentrations, wt%


Fig. 2.66. Measured X-ray intensity histograms for silicon, manganese and aluminium in inclusions
extracted from a low-alloy steel weld metal. Arrows indicate average composition. Data from Kluken
and Grong.57
number of counts recorded for copper in discrete particles is not significantly higher than the
corresponding matrix value, which shows that the copper content of the inclusions is low.
Since these measurements were carried out on mechanically polished specimens and not on
carbon extraction replicas, as done in the EDX analysis, the indications are that the higher
inclusion copper level observed in the latter case mainly results from surface copper contamination inherent from the extraction process. In contrast, the WDX analysis of sulphur revealed
evidence of sulphur enrichment in most of the particles. Considering the fact that these particles also contained significant amounts of manganese, it is reasonable to assume that most of
the sulphur is present in the form of MnS (possibly with some copper in solid solution).76
2.12.3.3 Titanium and nitrogen contents
From the literature reviewed, it is apparent that conflicting views are held about the role of
titanium in weld metal deoxidation. From a thermodynamic standpoint, Ti2O3 is the stable
reaction product in titanium deoxidation,76 but this phase has not yet been detected experimentally in steel weldments (only in continuous cast steel).77 However several authors have
reported the presence of crystalline patches containing titanium toward the edges of
inclusions.78"80 This constituent has a cubic crystal structure with a lattice parameter close to
0.42 nm, conforming to either 7-TiO, TiN, or TiC (note that the 7-T1O phase is the hightemperature modification of the titanium monoxide).81
In general, formation of titanium monoxide requires strongly reducing conditions, which

implies that the 7-TiO constituent is not an equilibrium reaction product in steel deoxidation.55'76
Hence, it is reasonable to assume that the titanium-rich crystalline patches observed toward
the edges of weld metal inclusions are titanium nitride. This conclusion has later been confirmed experimentally by Kluken and Grong.57
2.12.3.4 Constituent phases
Based on the above discussion, it is possible to rationalise the formation of primary and secondary reaction products (i.e. oxides, sulphides and nitrides) during cooling in the weld pool,
as shown in Fig. 2.67. In general, the inclusions will consist of an oxide core which is formed
during the primary deoxidation stage. The chemical composition of the deoxidation product
can vary within wide limits, depending on the activities of Al, Ti, Si, Mn, and O in the weld
metal. The surface of the oxides will partly be covered by MnS and TiN (see also STEM
micrograph in Fig. 2.57). Precipitation of these phases occurs after the completion of the weld
metal deoxidation, probably during solidification, where the reactions are favoured by solute
enrichment in the interdendritic liquid. Additional precipitation of TiN may occur in the solid
state as a result of diffusion of titanium and nitrogen to the surface of the inclusions.
2.12.4 Prediction of inclusion composition
Since the diffusion-controlled deoxidation reactions are completed within a fraction of a second when the number of nuclei is of the order of 107 mm"3 or higher,55 the average chemical
composition of the inclusions should be compatible with that calculated for chemical equilibrium at temperatures close to the melting point (e.g. 15500C).
2.12.4.1 C-Mn steel weld metals
Over the composition range normally applicable to silicon-manganese deoxidation of steel
weld metals (i.e. between 0.4 to 0.7 wt% Si and 0.8 to 1.5 wt% Mn) the equilibrium reaction
product at 15500C should be silica-saturated slags with a mole fraction of SiO2 close to 0.55. 5565
Since the two other slag components are MnO and FeO, we may write:
(2-84)
The activity coefficients for MnO and FeO in the ternary system SiO 2 -MnO-FeO can be
computed from the equations presented by Sommerville et al.:S2
(2-85)
and
(2-86)
For the specific case of silica-saturated slags, we obtain:
(2-87)
and
(2-88)

TiN (secondary reaction product)


Oxide Core (primary reaction product]
MnS (secondary reaction product)
Fig. 2.67. Schematic diagram showing the presence of primary and secondary phases in weld metal
inclusions.
Under such conditions the activity ratio aMnO/aFeO in the slag is given by:

(2-89)

The activity ratio aMnO/aFeO can also be estimated from the equilibrium constant for the FeMn reaction at 15500C,23 i.e.:
(2-90)

The corresponding mole fractions of MnO and FeO in the slag phase are then obtained by
combining equations (2-89) and (2-90):
(2-91)
and
(2-92)

Equations (2-91) and (2-92) provide a basis for calculating the chemical composition of the
inclusions under different deoxidation conditions. A requirement is, however, that the weld
metal Si to Mn ratio is sufficiently high to promote precipitation of silica-saturated slags at
15500C.

and

This gives the following chemical composition of the inclusions (in wt%):

The above calculations should be compared with the composition data presented in Fig.
2.68. It is evident from this plot that the agreement between predictions and experiments is
reasonably good both at high and low weld metal manganese levels. This justifies the
simplifications made in deriving equations (2-91) and (2-92).
2.12.4.2 Low-alloy steel weld metals
In principle, a procedure similar to that described above could also be used to establish a
theoretical basis for predicting the chemical composition of the inclusions in the case lowalloy steel weld metals. Unfortunately, adequate activity data for the Fe-Al-Ti-Si-Mn-O system are not available in the literature.
An alternative approach would be to calculate the average inclusion composition from simple mass balances, assuming that all oxygen combines stoichiometrically with the various
deoxidation elements to form stable oxides in the order Al2O3, Ti2O3, SiO2, and MnO, according to their oxygen affinity in liquid steel (see Fig. 2.69). If reasonable average values for the
inclusion and steel densities are used (i.e. 4.2 and 7.8 g cm"3, respectively), the following set of
equations* can be derived from a balance of O, Al, Ti, S, Si, and Mn and data for acid soluble
aluminium and titanium in the weld metal:57
Aluminium

The average aluminium content of the inclusions, [%Al]inch can be estimated from the measured difference between total and acid soluble aluminium in the weld metal, (k%Al)we[d. This
difference is, in turn, equal to the total mass of aluminium in the inclusions:
(2-93)
where mind and Vv (cal) are the total mass and volume fraction of non-metallic inclusions in
the weld deposit, respectively.
*The numerical constants in the constitutive equations given below could alternatively be expressed in terms of
atomic weights etc. to bring out more clearly their physical significance (e.g. see the treatment of Bhadeshia and
Svensson83).

Mole fraction

Ar - O2 gas mixtures
Ar - CO2 gas mixtures

Manganese content, wt%


Fig. 2.68. Comparison between measured and predicted microslag composition in GMA welding of
C-Mn steels. Solid lines represent theoretical calculations based on equations (2-91) and (2-92). Data
from Grong and Christensen.1

Oxygen in solution, wt%

Iron oxide in solution at 16000C


Mn

Si
Ti

Al

Deoxidizer in solution, wt%


Fig. 2.69.Deoxidation equilibria in liquid steel at 1600C.Data from Turkdogan.55

By rearranging equation (2-93), we get:

(2-94)

However, since data for acid soluble aluminium (and titanium), in practice, may contain
large inherent errors, the following restriction applies:
(2-95)
where [%O]anaL is the analytical weld metal oxygen content.
Titanium

Similarly, the average titanium content of the inclusions, [%Ti]incl, can be estimated from the
measured difference between total and acid soluble titanium, (A%Ti)weld. However, since TiN
dissolves readily in strong acid, it is necessary to include an empirical correction for the amount
of titanium nitride which simultaneously forms at the surface of the inclusions during solidification. This can be done on the basis of published data for the solubility product of TiN in
liquid steel.84 If we assume that the nitrogen content of the inclusions is proportional to the
calculated difference between total and dissolved nitrogen at the melting point (15200C), the
following relationship is obtained:

(2-96)

where [%N]anai is the analytical weld metal nitrogen content, and [%Ti]soL is acid soluble
titanium. Note that the correction term for TiN, in practice, neither can be negative nor exceed
[%Ti]soh

In this case, the maximum amount of titanium which can be bound as Ti2O3 is determined
by the overall oxygen balance:
(2-97)
Sulphur

If the solubility of sulphur in solid steel is taken equal to 0.003 wt%,70 the average sulphur
content of the inclusions, [%S]ind is given by:

(2-98)
where [%S]anai is the analytical weld metal sulphur content.
Silicon and manganese

From the experimental data of Saggese et al.S5 reproduced in Fig. 2.70, it is evident that the
mass ratio between SiO2 and MnO in the oxide phase may be considered constant and virtually
independent of composition (i.e. equal to about 0.94). This implies that the average silicon
content of the inclusions, [%Si]ind, can be calculated from a balance of oxygen:

SiO2

Al2O3

MnO
wt% AI2O3

Fig. 2.70. Measured inclusion compositions in low-alloy steel weld metals. Data from Saggese et alP

(2-99)
Considering manganese, proper adjustments should also be made for the amount of MnS
formed at the surface of the inclusions during solidification. Hence, the average manganese
content of the inclusions, [%Mn]inci, is given by the sum of the oxygen and the sulphur contributions:

(2-100)

Experimental verification of model

In Fig. 2.71, the accuracy of the model has been tested against the experimental SA/GMA
inclusion data reported by Kluken and Grong,57 taking the sum (%A1 + %Ti + %S + %Si +
%Mn) equal to 100%. A closer inspection of the graphs reveals a reasonable agreement between calculated and measured average inclusion compositions in all cases, which confirms
that the model is sound. Included in Fig. 2.71 is also a collection of data for SMA low-alloy
steel weld metals (3 kJ mm"1 basic electrodes). Since these results follow the same pattern,
it implies that the model is generally applicable and, therefore, can be adopted to all relevant
arc welding processes.

Measured composition, wt%

Aluminium
Sulphur

Marked symbols: SMAW

Measured composition, wt%

Calculated composition, wt%

Titanium
Manganese
Silicon

Marked symbols: SMAW

Calculated composition, wt%


Fig. 2.71. Comparison between measured and predicted inclusion compositions; (a) Aluminium and
sulphur, (b) Titanium, manganese and silicon. Data from Kluken and Grong57'86
Implications of model

It can be inferred from equations (2-94) and (2-95) that the chemical composition of the inclusion oxide core is directly related to the aluminium to oxygen ratio in the weld metal.
Referring to Fig. 2.72, the fraction of MnOAl2O3 and 7-Al2O3 in the inclusions is seen to increase steadily with increasing (A%Al)wdJ[%O]anal ratios until the stoichiometric composition for precipitation of aluminium oxide is reached at 1.13. At higher ratios, the deoxidation product will be pure Al2O3, since aluminium is present in an over-stoichiometric amount
with respect to oxygen. When titanium is added to the weld metal, titanium oxide (in the form
of Ti2O3) may also enter the reaction product. At the same time both TiN and a-MnS form
epitaxially on the surface of the inclusions during solidification. Consequently, in Al-Ti-Si-Mn
deoxidised steel weld metals the total number of constituent phases within the inclusions may
approach six. The kinetics of inclusion formation are further discussed in Ref. 87.

SiO2

MnO

Fig. 2.72. Coexisting phases in inclusions at different weld metal aluminium-to-oxygen ratios. Shaded
region indicates the approximate composition range for the oxide phase. The diagram is constructed on
the basis of the model of Kluken and Grong57 and relevant literature data.

Example (2.13)

Consider SA welding of low-alloy steel with two different basic fluxes. Data for the weld
metal chemical compositions are given in Table 2.12. Based on Fig. 2.72, estimate the total
number of constituent phases in the inclusions in each case.
Solution

It follows from Fig. 2.72 that the chemical composition of the deoxidation product is determined by the aluminium to oxygen ratio in the weld metal. For weld A, we have:

Since this ratio is higher than the stoichiometric factor of 1.13, all oxygen is probably tied up
as aluminium oxide. In addition, weld A contains small amounts of titanium and sulphur,
which may give rise to precipitation of TiN and MnS at the surface of the inclusions during
solidification. Hence, the three major constituent phases in the weld metal inclusions are
7-Al2O3, TiN, and ot-MnS.

Table 2.12 Chemical composition of SA steel weld metals considered in Example 2.13.
Weld
No.

C
(wt%)

O
(wt%)

Si
(wt%)

Mn
(wt%)

S
(wt%)

N
(wt%)

Al*
(wt%)

Ti*
(wt%)

0.09

0.021

0.45

1.52

0.01

0.006

0.028
(0.003)

0.010
(0.009)

0.09

0.045

0.45

1.52

0.01

0.006

0.028
(0.003)

0.028
(0.010)

*Data for acid soluble Al and Ti are given by the values in brackets.

In the case of weld B the situation is much more complex due to a higher content of oxygen
and titanium. From Table 2.12, we have:

and

These calculations show that Al and Ti are not present in sufficient amounts to tie-up all oxygen. Under such conditions Fig. 2.72 predicts that the total number of constituent phases in the
inclusions is six, i.e.: SiO2, MnOAl2O3, 7-Al2O3, Ti2O3, a-MnS and TiN.
References
1
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.

0 . Grong and N. Christensen: Scand. J. Metall, 1983,12, 155-165.


J.F. Lancaster: Metallurgy of Welding, 3rd Edn, 1980, London, George Allen & Unwin Ltd.
J. Szekely and NJ. Themelis: Rate Phenomena in Process Metallurgy, 1971, New York, John
Wiley & Sons, Inc.
N. Christensen: Welding Metallurgy Compendium, 1985, University of Trondheim, The Norwegian Institute of Technology, Trondheim, Norway.
E.T. Turkdogan: Physical Chemistry of Oxygen Steelmaking, Thermochemistry and Thermodynamics, 1970, United States Steel Corporation.
F.R. Coe: Welding Steels without Hydrogen Cracking, 1973, Abington (Cambridge), The Welding
Institute.
Doc. IIS/IIW-532-77: Weld. World, 1977,15, 69-72.
B. Chew and R. A. Willgoss: Proc. Int. Conf Weld Pool Chemistry and Metallurgy, London,
April 1980, Paper 25,155-165. Publ. The Welding Institute (England).
F. Matsuda, H. Nakagawa, K. Shinozaki and H. Kihara: Trans. JWRI, 1978, 7, 135-137.
B. Chew: Weld. J., 1973, 52, 386s-391s.
GM. Evans: Schweissmitteilungen (Oerlikon), 1977, No. 79, 4-8.
GM. Evans and H. Bach: HW Doc. IIA-363-74, 1974.

13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.

R. Roux: Rev. Met. 1954,192-209.


E.M. Skjolberg: M.Sc. thesis, 1980, University of Trondheim, The Norwegian Institute of Technology, Trondheim, Norway.
D.G. Howden and D.R. Milner: Brit. Weld. J., 1963,10, 304-316.
T. Kobayashi, T. Kuwana and T. Kikuchi: HW Doc. XII-265-65, 1965.
J.E. O'Brien and M.F. Jordan: Met. Constr. Brit. Weld. J., 1971,3, 299-303.
M. Uda and S. Ohno: Trans. Nat. Res. Inst. for Metals, 1973,15, 20-28.
O. Morigaki, T. Tanigaki, M. Kuwabara and K. Fujibayashi: HW Doc. 11-746-75, 1975.
H.K.D.H. Bhadeshia, L.E. Svensson and B. Gretoft: J. Mater. ScL Lett, 1988, 7, 610-612.
E. Halm0y: Proc. Int. Conf. on Arc Physics and Weld Pool Behaviour, London, May, 1979,4957. Publ. The Welding Institute (England).
D.J.H. Corderoy, B. Wills and G.R. Wallwork: Proc. Int. Conf. on Weld Pool Chemistry and
Metallurgy, London, April 1980, 147-153. Publ. The Welding Institute (England).
J.F. Elliott, M.Gleiser and V. Ramakrishna: Thermochemistry for Steelmaking (Vol. II), 1963,
London, Addison-Wesley Publ. Company (Pergamon Press).
Doc. IIS/IIW-343-70: Weld. World, 1970, 8, 28-35.
I. Barin and O. Knacke: The Thermochemical Properties of Inorganic Substances, 1973, Berlin, Springer Verlag.
Janaf, Thermochemical Tables - 2nd Edn (NSRDS); 1971, New York, National Bureau of
Standards.
R.F. Heile and D.C. Hill: Weld. J., 1975, 54, 201s-210s.
B. Ozturk and RJ. Fruehan: Metall Trans. B, 1985,16B, 801-806.
G. Jelmorini, G.W. Tichelaar and GJ.P.M. Van den Heuvel: HW Doc. 212-411-77, 1977.
O. Kubaschewski and C B . Alcock: Metallurgical Thermochemistry, 5th Edn, 1979, Oxford,
Pergamon Press.
PA. Distin, S.G. Whiteway and CR. Masson: Can. Metall. Quart., 1971,10, 13-18.
J.F. Lancaster: The Physics of Welding, 1984, Oxford, HW-Pergamon Press, 204-267.
WA. Fischer and J.F. Schumacher: Arch. Eisenhiittenwesen, 1978, 49, 431-435.
N. Christensen, V. de L. Davis and K. Gjermundsen: Brit. Weld. J., 1965,12, 54-75.
G.K. Sigworth and J.F. Elliott: Metall. Trans., 1973, 4,105-113.
D.N. Shackleton and A.A. Smith: HW Doc. XII-832-84, 1984.
0. Grong, N.H. Rora and N. Christensen: Scand. J. Metall., 1984,13,154-156.
A.A. Smith et al: Weld. World., 1970, 8, 28-35.
A.A. Smith: CO2-Shielded Consumable Electrode Arc Welding, 2nd Edn, 1965, Brit. Weld.
Res. Assoc.
U. Lindborg: Met. Constr. and Brit. Weld. J., 1972, 4, 52-55.
T.H. North, H.B. Bell, A. Nowicki and I. Craig: Weld. J., 1978, 57, 63s-71s.
CS. Chai and T. W. Eagar: Metall. Trans., 1981,12B, 539-547.
U. Mitra and T.W. Eagar: Metall. Trans., 1984,15A, 217-227.
J.E. Indacochea, M. Blander, N. Christensen and D.L. Olson: Metall. Trans., 1985, 16B, 237245.
N. Christensen and 0 . Grong: Scand. J. Metall, 1986,15, 30-40.
U. Mitra and T.W. Eagar: Metall Trans., 1991, 22B, 65-100.
T.W. Eagar: Weld. J., 1978, 57, 76s-80s.
N. Christensen and J. Chipman: Weld. Res. Coun. Bull. Series, 1953, No. 15, New York, Welding Research Council (USA).
A. Apold: Carbon Oxidation in the Weld Pool, 1962, Oxford, Pergamon Press.
D.A. Wolstenholme: Proc. Int. Conf on Trends in Steel and Consumables for Welding, London, 1978,123-134. Publ. The Welding Institute (England).
0 . Grong, D.L. Olson and N. Christensen: Metal Constr., 1985,17, 810R-814R.

52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
78.
79.
80.
81.
82.
83.
84.
85.

86.
87.

P. Nilles, P. Dauby and J. Claes: Proc. Int. Conf. Basic Oxygen Steelmaking A New Technology Emerges, London, 1978, 60-72, The Metals Society (England).
U. Lindborg and K. Torsell: Trans. TMS-AIME, 1968, 242, 94-102.
N.F. Grevillius: Jernkont. Ann., 1969,153, 547-572.
E.T. Turkdogan: Proc. Int. Conf. on Chem. Metall. of Iron and Steel, Sheffield, July 1971,153170, Publ. The Iron and Steel Institute (England).
0 . Grong. T.A. Siewert, G.P. Martins and D.L. Olson: Metall. Trans. A, 1986,17A, 1797-1807.
A.O. Kluken and 0 . Grong: Metall. Trans. A, 1989, 2OA, 1335-1349.
M.L. Turpin and J.F. Elliott: J. Iron Steel Inst., 1966, 204, 217-225.
E.T. Turkdogan: Physicochemical Properties of Molten Slags and Glasses, 1983, London, The
Metals Society.
E. Plockinger and M. Wahlster: Stahl und Eisen, 1960, 80, 659-669.
BJ. Bradstreet: HW Doc. 212-138-68, 1968.
J.F. Lancaster: Phys. Technol, 1984,15, 73-79
F. Eickhorn and A. Engel: HW Doc. 212-201-70, 1970.
N. Mori and Y. Horii: HW Doc. 212-188-70, 1970.
R.A. Walsh and S. Ramachandran: Trans. TMS-AIME, 1963, 227, 560-562.
D.C. Hilty and W. Crafts: Trans.-AIME, 1950,188, 425-436.
E.T. Turkdogan: Trans. TMS-AIME, 1965, 233, 2100-2112.
G.M. Evans: HW Doc. IIA-630-84, 1984.
0 . Grong and D.K Matlock: Int. Met. Rev., 1986, 31, 27-48.
A.G. Franklin: J. Iron Steel Inst., 1969, 207, 181-186.
A.O. Kluken, 0. Grong and J. Hjelen: Mat. ScL Technol., 1988, 4, 649-654.
L.M. Hocking: Quart. J. Royal Meterol. Soc, 1959, 85, 44-50.
C. Wagner: Z Electrochemie, 1961, 65, 581-591.
E.E. Underwood: Quantitative Stereology, 1970, London, Addison-Wesley Publ. Co.
M.F. Ashby and R. Ebeling: Trans. TMS-AIME, 1966, 236, 1396-1404.
R. Kiessling: Non-Metallic Inclusions in Steel, 1978, London, The Metals Society (TMS).
H. Homma, S. Ohkita, S. Matsuda and K. Yamamoto: Weld. J., 1987, 66, 301s-309s.
G. Thewlis: HW Doc. IIA-736-88, 1988.
J.M. Dowling, J.M. Corbett and H.W. Kerr: Metall. Trans. A, 1986,17A, 1611-1623.
G.M. Evans: Metal Constr., 1986,18, 631R-636R.
J.L. Murray and H.A. Wriedt: Bull. Alloy Phase Diagr., 1987, 8, 148-165.
LD. Sommerville, I. Ivanchev and H.B. Bell: Proc. Int. Conf Chem. Metall of Iron and Steel,
Sheffield, July 1971, 23-25, Publ. The Iron and Steel Inst. (1973).
H.K.D.H. Bhadeshia and L.E. Svensson: Mathematical Modelling of Weld Phenomena (Eds
H. Cerjak and K.E. Easterling), 1993, London, The Institute of Materials, 109-180.
S. Matsuda and N. Okumura: Trans. ISIJ, 1978,18, 198-205.
M.E. Saggese, A.R. Bhatti, D.N. Hawkins and J.A. Whiteman: Proc. Int. Conf on the Effect of
Residual, Impurity and Micro-Alloy ing Elements on Weldability and Weld Properties, London,
Nov. 1983, Paper 15, Publ. The Welding Institute (England).
A.O. Kluken and 0 . Grong: Report No. STF34 F87125, 1987, Sintef, Trondheim, Norway.
S.S. Babu, S.A. David, J.M. Vitek, K. Mundra and T. DebRoy: Mater. ScL Technol., 1995,11,186199.

Appendix 2.1
Nomenclature
contact area (mm2)

activity coefficient of
element X

difference between
total and acid soluble Al in
weld metal (wt%)

gravity constant
(9.81ms- 2 )

thermal diffusivity
(mm2 s 1 )

standard free energy of


reaction (J mol"1 or
kJ mol-1)

activity of arbitrary slag


component

energy barrier for homogeneous nucleation


(kJ mol-1)

cross section of fused


parent metal (mm2)

driving force for precipitation of oxide inclusions


(J nr 3 )

flux basicity index


kinetic constants
GMAW

gas metal arc welding

GTAW

gas tungsten arc welding

drag coefficient
molar concentration of
element X in the liquid
(mol m~3)
cross section of deposited
metal (mm2)
element diffusivity in liquid
phase (m2 s"1 or mm2 s"1)
diffusivity of element X in
gas phase (mm2 s"1)
gross heat input (kJ mm"1)
FCAW

standard enthalpy of
reaction (J mol-1 or kJ
mol-1)
hydrogen content related
to deposited metal (ml per
10Og deposit)
hydrogen content related
to fused metal (ml per 10Og
or g per ton)

flux cored arc welding

hydrogen content related


to French practice (g per
ton)

drag force acting on


spherical particle in relative
motion to a fluid (N)

heat content per unit


volume at the melting point
(J mm"3)

gravity force acting on a


spherical particle in relative
motion to a fluid (N)

hydrogen content related


to Japanese practice (ml per
10Og deposit)

fume formation rate


(mg miir 1 )

amperage (A)

nucleation rate of oxide


inclusions in the weld pool
(nuclei nr 3 s"1)
constant in equation (A2-1)
(nuclei rrrV 1 )
ratio between absorbed Si
and Mn in the weld metal
coefficient of weld metal
deposition (g A-1S"1)
mass transfer coefficient for
gas absorption (mm s"1)
mass transfer coefficient for
gas desorption (mm s"1)
overall mass transfer
coefficient (mm s"1)
equilibrium constants

total mass of liquid metal


leaving/entering the
reaction zone per unit time
(g s-1)
arbitrary flux or slag
component
concentration displacement
of manganese referred to
different standard states
(wt%)
total manganese absorption
in the weld metal due to
slag/metal interactions
(wt%)
rejected manganese during
cooling in the weld pool
(wt%)
oxidation loss of manganese at electrode tip (wt%)

the product [%C][%O]


atomic weight of element X
mass of CO2 per 100 g of
electrode coating (g or
wt%)

dimensionless operating
parameter

total mass of inclusions per


100 g weld deposit (g)

total number of moles of


component /

mass of H2O per 100 g of


electrode coating (g or
wt%)

absorption/desorption rate
of element X in the weld
pool (mol s"1)

total weight of retained


MnOSiO2 (g per 100 g weld
deposit)

total number of moles of


component X

total weight of precipitated


microslag (in g per 100 g
weld deposit)
total weight of precipitated
MnS (g per 100 g weld
deposit)

total number of moles of


component X2
number of particles per
unit area (mm"2)
Avogadro constant (6.022
X 1023 mor 1 )

mole fraction of arbitrary


slag component

combined partial pressure


of H2 and H2O (atm)

analytical weld metal


nitrogen content (wt%)

post weld heat treatment


parts per million (g per ton)

mole fraction of CO in gas


phase
Reynolds number
number of particles per unit
volume (No. per mm"3)
calculated oxygen absorption in electrode tip or
falling droplets (wt%)
analytical weld metal
oxygen content (wt%)

net arc energy (W)


universal gas constant
(J K-1 mol-1)
relative humidity (%)
analytical weld metal
sulphur content (wt%)
standard entropy of
reaction (J K -1 mol"1)

equilibrium weld metal


oxygen content (wt%)

concentration displacement
of silicon referred to
different standard states
(wt%)

rejected oxygen during


cooling in the weld pool
(wt%)

total silicon absorption in


the weld metal due to slag/
metal interactions (wt%)

total oxygen absorption in


hot part of weld pool (wt%)

rejected silicon during


cooling in the weld pool
(wt%)

total pressure (atm or bar)


partial pressure of component X in bulk gas phase
(atm)
equilibrium partial pressure
of component X at gas/
metal interface (atm)
partial pressure of component X2 in bulk gas phase
(atm)
equilibrium partial pressure
of component X2 at gas/
metal interface (atm)

oxidation loss of silicon at


electrode tip (wt%)
solubility of element X at 1
atm total pressure (ml per
10Og, ppm or wt%)
total particle surface area
per unit volume (mm2 per
mm3)
modified solubility of
hydrogen at 1 atm total
pressure (ml per 100 g, ppm
or wt%)

SAW

submerged arc welding

SMAW

shielded metal arc welding

STEM

scanning transmission
electron microscope

volume fraction of
inclusions
concentration of component / (vol%)
concentration of component X2 (vol%)

time (s)
time constant (s)
acid soluble titanium in
weld metal (wt%)
difference between total
and acid soluble Ti in weld
metal (wt%)
temperature (0C or K)

concentration of
component YX2 (vol%)
WDX

wavelength dispersive Xray analysis

WFR

wire feed rate (mm s -1 or


m min"1)
arbitrary element or
gaseous species

ambient temperature (0C or


K)

weld ripple lag (mm)

preheating temperature (0C


or K)

concentration of element X
in the weld metal (wt%)

reference temperatures (0C


or K)

concentration displacement
of element X referred to
nominal weld metal
composition (wt%)

rising velocity of ascending


particles relative to the
liquid (|jLms-1)
bulk velocity of the fluid
relative to the particles
(ms- 1 )

equilibrium concentration
of element X in the weld
pool (wt%)

voltage (V)

initial concentration of
element X in the weld pool
(wt%)

welding (travel) speed


(mm s"1)

average content of element


X in inclusions (wt%)

volume of melt (mm3)

nominal concentration of
element X in weld metal
(wt%)

molar volume of nucleus


(m3 mol-1)
molar volume of oxide per
mole of the diffusate
(m3 mol-1)

concentration of element X
in base plate (wt%)

concentration of element X
in filler wire (wt%)
arbitrary element or
gaseous species
density (kg irr 3 or
g mm"3)

thickness of stagnant
gaseous boundary layer
(mm)
wetting angle
mean particle centre to
centre volume spacing
(jxm)

difference in density
between liquid steel and
inclusion (kg irr 3 or
g mm"3)

dimensionless y-coordinate
at maximum width of
isotherm

slag/metal interfacial
energy (J nr 2 )

activity coefficient for


MnO in slag phase

arc efficiency factor

activity coefficient for


SiO2 in slag phase

viscosity (kg nr 1 s"1)

Appendix 2.2
Derivation of equation (2-60)
The nucleation rate / as a function of temperature can be expressed as:
(A2-1)
where J0 is a constant (with the unit nuclei per m3 and s) and AG* is the energy barrier for
nucleation.
By rearranging equation (A2-1) and inserting reasonable values for J and J0 for the specific
case of homogeneous nucleation of oxide inclusions in liquid steel,55 we obtain:

(A2-2)

From the classic theory of homogeneous nucleation AG* is given by:

(A2-3)
where NA is the Avogadro constant, o is the interfacial energy between the nucleus and the
liquid (in J nr 2 ) and AGV is the driving force for the precipitation reaction (in J nr 3 ).

The parameter AGV can be expressed as:

(A2-4)
where AH and AS are the standard entalpy and entropy of the precipitation reaction, respectively and Vm is the molar volume of the nucleus (in m3 mol"1).
It is evident from Fig. 2.48 that AGV = 0 when T=Tu which gives \S = A//7 T\. Hence,
equation (A2-4) may be rewritten as:
(A2-5)

By combining equations (A2-2), (A2-3) and (A2-5), we obtain the following relationship
between T\ and T^-

(A2- 6)

3
Solidification Behaviour of Fusion Welds

3.1 Introduction
Inherent to the welding process is the formation of a pool of molten metal directly below the
heat source. The shape of this molten pool is influenced by the flow of both heat and metal,
with melting occurring ahead of the heat source and solidification behind it. The heat input
determines the volume of molten metal and, hence, dilution and weld metal composition, as
well as the thermal conditions under which solidification takes place. Also important to solidification is the crystal growth rate, which is geometrically related to weld travel speed and weld
pool shape. Hence, weld pool shape, weld metal composition, cooling rate, and growth rate
are all factors interrelated to heat input which will affect the solidification microstructure.
Some important points regarding interpretation of weld metal microstructure in terms of these
four factors will be discussed below.
Since the properties and integrity of the weld metal depend on the solidification microstructure, a verified quantitative understanding of the weld pool solidification behaviour is essential. At present, our knowledge of the chemical and physical reactions occurring during solidification of fusion welds is limited. This situation arises mainly from a complex sequence of
reactions caused by the interplay between a number of variables which cannot readily be accounted for in a mathematical simulation of the process. Nevertheless, the present treatment
will show that it is possible to rationalise the development of the weld metal solidification
microstructure with models based on well established concepts from casting and homogenising
treatment of metals and alloys.

3.2 Structural Zones in Castings and Welds


The symbols and units used throughout this chapter are defined in Appendix 3.1.
During ingot casting, three different structural zones can generally be observed, as shown
schematically in Fig. 3.1. The chill zone is produced by heterogeneous nucleation in the region adjacent to the mould wall as a result of the pertinent thermal undercooling. These grains
rapidly become dendritic, and dendrites having their <100> direction (preferred easy growth
direction for cubic crystals) parallel to the maximum temperature gradient in the melt will
soon outgrow those grains that do not have this favourable orientation. Competitive growth
occurring during the initial stage of the solidification process leads to an alignment of the
crystals in the heat flow direction and eventually to the formation of a columnar zone. 12 Finally, an equiaxed zone may develop in the centre of the casting, mainly as a result of growth
of detached dendrite arms within the remaining, slightly undercooled liquid.
A similar situation also exists in welding, as indicated in Fig. 3.2 However, in this case the
chill zone is absent, since the partly melted base metal grains at the fusion boundary act as seed
crystals for the growing columnar grains.3 In addition, the growth direction of the columnar

Shrinkage pipe

Chill zone
Columnar zone
Equiaxed zone
Mould

Fig. 3.1. Transverse section of an ingot showing the chill zone, the columnar zone and the equiaxed zone
(schematic).
grains will change continuously from the fusion line towards the centre of the weld due to a
corresponding shift in the direction of the maximum temperature gradient in the weld pool.
This change in orientation may result in a curvature of the columnar grains (Fig. 3.2(a)). Alternatively, new grains can nucleate and grow in a columnar manner, producing a so-called 'stray'
structure as shown schematically in Fig. 3.2(b). Finally, if the conditions for nucleation of new
grains are favourable, an equiaxed zone will form near the weld centreline similar to that
observed in ingots or castings (see Fig. 3.2(b)).
Although the process of weld pool solidification is frequently compared with that of an
ingot in 'miniature', a number of basic differences, already mentioned, exist which strongly
influence the microstructure and properties of the weld metal. Of particular importance is also
the disparity in cooling rate between a fusion weld and an ingot (see Fig. 3.3). For conventional processes such as shielded metal arc (SMA), gas metal arc (GMA), submerged arc (SA)
or gas tungsten arc (GTA) welding the cooling rate may vary from 10 to 103 0 C s"1, while for
modern high energy beam processes such as electron beam (EB) and laser welding the cooling
rate is typically of the order of 103 to 106 0 C s"1.4 Consequently, to appreciate fully the implications of these differences in general solidification behaviour between a weld pool and an
ingot, it is necessary to consider in detail the sequence of events taking place in the solidifying
weld metal beginning with the initiation of crystal growth at the fusion boundary.

3.3 Epitaxial Solidification


It is well established that initial solidification during welding takes place epitaxially, where the
partly melted base metal grains at the fusion boundary act as seed crystals for the columnar
grains. This process is illustrated schematically in Fig. 3.4.

(a)
Welding direction

(b)
Columnar zone
Welding direction

Equiaxed zone
Columnar zone

Rapid solidification technology

SMAW, SAW,
GMAW, GTAW

Electron beam welding


Laser welding

Cooling rate, C/s

Fig. 3.2. Examples of structural zones in fusion welds (schematic); (a) Curved columnar grains,
(b) Stray grain structure.

Process
Fig. 3.3. Disparity in cooling conditions between casting, welding and rapid solidification.

Fusion
boundary
HAZ

Weld metal

Fig. 3.4. Schematic illustration showing epitaxial growth of columnar grains from partly melted base
metal grains at fusion boundary.
Liquid (L)

Substrate (S)

Fig. 3.5. Schematic representation of heterogeneous nucleation.

3.3.1 Energy barrier to nucleation


During epitaxial solidification, a solid embryo (nucleus) of the weld metal forms at the meltedback surface of the base metal grain. Assuming that the interfacial energy between the embryo
and the liquid is isotropic, it can be shown, for a given volume of the embryo, that the interfacial energy of the whole system is minimised if the embryo has the shape of a spherical cap.
Under such conditions, the following relationship exists between the interfacial energies (see
Fig. 3.5):
(3-1)
where (3 is the wetting angle.
The change in free energy, AGhet, accompanying the formation of a solid nucleus with this
configuration is given by:5
(3-2)

where VE is the volume of the solid embryo, AGV is the free energy change associated with the
embryo formation, AEL and AES are the areas of the embryo-liquid and embryo-substrate interfaces, respectively, and/(P) is the so-called shape factor, defined as:

(3-3)

The critical radius of the stable nucleus, r / , is found by differentiating equation (3-2) with
respect to rs and equating to zero:
(3-4)

By substituting equation (3-4) into equation (3-2), we obtain the following expression for
the energy barrier to heterogeneous nucleation (AG^ r ):

(3-5)

where AHm is the latent heat of melting, Tm is the melting point, and AJT is the undercooling.
It is easy to verify that the first term in equation (3-5) is equal to the energy barrier to
homogeneous nucleation, AG^om. Hence, we may write:
(3-6)
Equation (3-6) shows that AG^ is a simple function of the wetting angle O). Since the

chemical composition and the crystal structure of the two solid phases are usually very similar,
we have:6

Under such conditions equation (3-1) predicts that the wetting angle 3 ~ 0 (cos(3 ~ 1),
which implies that there is a negligible energy barrier to solidification of the weld metal (}*het
~ 0), i.e. no undercooling of the melt is needed, and solidification occurs uniformly over the
whole grain of the base metal. This is in sharp contrast to conventional casting of metals and
alloys where some undercooling of the melt is always required to overcome the inherent energy barrier to solidification (see Fig. 3.6).
3.3.2 Implications of epitaxial solidification
Since the initial size of the weld metal columnar grains is inherited directly from the grain
growth zone adjacent to the fusion boundary, the solidification microstructure depends on the
grain coarsening behaviour of the base material. This is particularly a problem in high energy
processes such as submerged arc and gas metal arc welding, where grain growth of the base
metal can be considerable. In such cases the size of the columnar grains at the fusion boundary
will be correspondingly coarse, as indicated by the data in Fig. 3.7.
Moreover, during multipass welding the columnar grains can renucleate at the boundary
between for instance the first and the second weld pass and subsequently grow across the
entire fusion zone, as illustrated in Fig. 3.8. This type of behaviour is usually observed in
weldments which do not undergo transformations in the solid state (e.g. aluminium, certain
titanium alloys, stainless steel etc.). In practice, the problem can be eliminated by additions of
inoculants via the filler wire, which facilitates a refinement of the columnar grain structure
through heterogeneous nucleation of new (equiaxed) grains ahead of the advancing interface
(to be discussed later).

AG

Welding

Casting
Homogeneous
nucleation

Fig. 3.6. The free energy change associated with heterogeneous nucleation during casting and weld
metal solidification, respectively (schematic). The corresponding free energy change associated with
homogeneous nucleation is indicated by the broken curve in the graph.

Weld metal
prior austenite grain size (jim)

Fusion
line

HAZ

Weld
metal

GMAW (low-alloy steel)

HAZ prior austenite grain size (jim)


Fig. 3.7. Correlation between HAZ prior austenite grain size at the fusion boundary and the corresponding weld metal prior austenite grain size. Data from Grong et al?

2. pass

1. pass
HAZ

Base metal

Fig. 3.8. Optical micrograph showing renucleation of columnar grains during multipass GMA welding
of a P-titanium alloy.

3.4 Weld Pool Shape and Columnar Grain Structures


Growth of the columnar grains always proceeds closely to the direction of the maximum thermal gradient in the weld pool, i.e. normal to the fusion boundary. This implies that the columnar grain morphology depends on the weld pool geometry.
3.4.1 Weld pool geometry
The weld pool geometry is a function of the welding speed and the balance between the heat
input and the cooling conditions, as influenced by the base plate thermal properties. At pseudosteady state, these conditions establish a dynamic equilibrium between heat supply and heat
extraction so that the shape of the weld pool remains constant for any given speed. Following
the treatment in Chapter 1, the weld pool geometry depends on the dimensionless operating
parameter n3, defined as:
(3-7)
where qo is the net arc power, v is the welding speed, a is the thermal diffusivity of the base
plate, and Hm-Ho is the heat content per unit volume at the melting point.
As shown in Fig. 3.9(a), a tear-shaped weld pool is favoured by a high n3 value, which is
characteristic of fast moving high power sources. In contrast, at a low arc power and a low
welding speed the shape of the weld pool becomes more elliptical because of a shift in the
mode of heat flow (see Fig. 3.9(b)). Note, however, that the thermal properties of the base
metal is also of importance in this respect, since the n3 parameter is a function of both a and
Hm-Ho. Consequently, a tear-shaped weld pool is usually observed in weldments of a low
thermal diffusivity (e.g. austenitic stainless steel), whereas an elliptical or spherical weld pool
is more likely to form during aluminium welding owing to the resulting higher thermal diffusivity of the base metal.
In addition to the factors mentioned above, the geometry of the weld pool is also affected by
convectional heat transfer due to the presence of buoyancy, electromagnetic or suface tension
gradient forces. Recently, attempts have been made to include such effects in heat flow models for welding.8"11 Referring to Fig. 3.10(a) the buoyancy force will promote the formation of
a shallow, wide weld pool because of transport of 'hot' metal to the surface and 'cold' metal to
the bottom of the pool. In the presence of the electromagnetic force the flow pattern is reversed, since the latter force will tend to push the liquid metal in the central part of the pool
downward to the root of the weld. This makes the weld pool deeper and more narrow, as
shown in Fig. 3.10(b).
Moreover, it is generally accepted that surface tension gradients can promote circulation of
liquid metal within the weld pool from the region of low surface tension to the region of higher
surface tension.9 In the absence of surface active elements such as oxygen and sulphur, the
surface tension decreases with increasing temperature as illustrated in Fig. 3.10(c), which forces
the metal to flow outwards towards the fusion boundary. This results in the formation of a
relatively wide and shallow weld pool. However, if oxygen or sulphur is present in sufficient
quantities a positive temperature coefficient of the surface tension may develop, which facilitates an inward fluid flow pattern and an increased weld penetration (see Fig. 3.10(d)). The
important influence of surface active elements on the resulting bead morphology is well docu-

HAZ isotherms
Fusion boundary

Weld
pool

(a)
HAZ isotherms
Fusion boundary

Weld
pool

(b)
Fig. 3.9. Theoretical shape of fusion boundary and neighbouring isotherms under different operational
conditions; (a) High n3 values, (b) Low ^-values.
merited for ordinary GTA austenitic stainless steel welds. 1 2 1 3 The indications are that such
effects become even more important under hyperbaric welding conditions. 14

3.4.2 Columnar grain morphology


It is evident from the above discussion that a change in the weld pool geometry, caused by
variations in the operational conditions, may strongly alter the weld metal solidification microstructure. In fact, more than nine different grain morphologies have been observed during
fusion welding.15 The two most important are shown in Fig. 3.11. Referring to Fig. 3.11(a) a
spherical or elliptical weld pool will reveal curved and tapered columnar grains owing to a
shift in the direction of the maximum thermal gradient in the liquid from the fusion boundary
towards the weld centre-line. In contrast, a tear-shaped weld pool yields straight and broad

Electrode

Weld pool

Arc

(a)

Electrode

Weld pool

Arc

(b)

Fig. 3.10. Schematic diagrams illustrating the major fluid flow mechanisms operating in a weld pool;
(a) Buoyancy force (b) Electromagnetic force.
columnar grains as shown in Fig. 3.11(b), since the direction of the maximum temperature
gradient in the melt does not change significantly during the solidification process. The latter condition is known to promote formation of centre-line cracking because of mechanical
entrapment of inclusions and enrichment of eutectic liquid at the trailing edge of the weld
pool.
3.4.3 Growth rate of columnar grains
The growth rate of the columnar grains is geometrically related to the weld travel speed and
the weld pool shape.
3.4.3.1 Nominal crystal growth rate
Since the shape of the weld pool remains constant during steady state welding, the growth rate
of the columnar grains must vary with position along the fusion boundary. This point is more
clearly illustrated in Fig. 3.12 which shows a sketch of a single columnar grain growing parallel with the steepest temperature gradient in the weld pool. Taking the angle between the

Surface tension
Temperature
Electrode

Weld pool

Arc

Surface tension

(C)

Temperature

Electrode

Arc
Weld pool

(d)

Fig. 3.10. Schematic diagrams illustrating the major fluid flow mechanisms operating in a weld pool
(continued); (c) Surface tension gradient force (negative gradient); (d) Surface tension gradient force
(positive gradient).

Heat source

Heat source

(a)

(b)

Fig. 3.11. Schematic comparison of columnar grain structures obtained under different welding conditions; (a) Elliptical weld pool (low n3 values), (b) Tear-shaped weld pool (high n3 values). Open arrows
indicate the direction of the maximum temperature gradient in the weld pool.

Fusion boundary

Heat source
Crystal
Fig. 3.12. Definition of the nominal crystal growth rate RN.
growth direction and the welding direction equal to a, the steady state growth rate, R N , becomes:
(3-8)
where v is the welding speed.
Considering spherical or elliptical weld pools, the nominal crystal growth rate is lowest at
the edge of the weld pool (a>90, cosa-^0) and highest at the weld centre-line where R N
approaches v (a->0, c o s a ^ l ) . In contrast, columnar grains trailing behind a tear-shaped weld
pool will grow at an approximately constant rate which is significantly lower than the actual
welding speed (a 0), since the direction of the maximum temperature gradient in the weld
pool does not change during the solidification process. This is also in agreement with practical
experience (see Fig. 3.13).

(a)
Nominal growth rate (RN), mm/s

Niobium (1 mm plate thickness)

Relative position from edge of weld pool (%)

(b)

Equiaxed zone

Nominal growth rate (RN), mm/s

Stainless steel (1 mm plate thickness)

Relative position from edge of weld pool (%)


Fig. 3.13. Measured crystal growth rates in thin sheet electron beam welding; (a) Niobium, (b) Stainless
steel. Data from Senda et al.16

Example (3.1)

Consider electron beam (EB) welding of a lmm thin sheet of austenitic stainless steel under
the following conditions:

Estimate on the basis of the Rosenthal thin plate solution (equation 1-83) the steady state
growth rate of the columnar grains trailing the weld pool.
Solution

The contour of the fusion boundary can be calculated from the Rosenthal thin plate solution
according to the procedure shown in Example (1.10). If we include a correction for the latent
heat of melting, the QbZn3 ratio at the melting point becomes:

Substitution of the above value into equation (1-83) gives the fusion boundary contour
shown in Fig. 3.14. It is evident from Fig. 3.14 that the weld pool is very elongated under the
prevailing circumstances due to a constrained heat flow in the ^-direction. This implies that the
angle a will not change significantly during the solidification process. Taking a as an average,
equal to about 70, the steady-state crystal growth rate R N becomes:

This value is in reasonable agreement with the measured crystal growth rates in Fig. 3.13(b).
3.4.3.2 Local crystal growth rate
Equation (3-8) does not take into account the inherent anisotropy of crystal growth. For faceted materials the dendrite growth directions are always those that are 'capped' by relatively
slow-growing (usually low-index) crystallographic planes.1 Figure 3.15 shows examples of
faceted cubic crystals delimited by {100} and {111} planes, respectively. If the {111} planes
are the slowest growing ones, the {100} planes will grow out, leaving the {111} facets and a
new crystal growing in the <100> directions as shown schematically in Fig. 3.15(b).
Although most metals and alloys do not form faceted dendrites, the anisotropy of crystal
growth is still maintained during solidification.2 In fact, experience has shown that the major
dendrite growth direction is normally the axis of a pyramid whose sides are the most closely
packed planes with which a pyramid can be formed.1 These directions are thus <100> for
body- and face-centred cubic structures, < 1010 > for hexagonal close-packed structures, and
<110> for body-centred tetragonal structures.
Because of the existence of preferred growth directions, the local growth rate of the crystals
RL will always be higher than the nominal growth rate R N defined in equation (3-8). Consider
now a cubic crystal which grows along the steepest temperature gradient in the weld pool, as
shown schematically in Fig. 3.16. If denotes the angle between the interface normal and the
<100> direction, the following relationship exists between RN and RL:

-y(mm)
Columnar zone

Heat source

Equiaxed zone
+x(mm)

Columnar zone

+yjmm)

Fusion boundary

Fig. 3.14. Predicted shape of fusion boundary during electron beam welding of austenitic stainless steel
(Example (3.1)).

(a)

(b)

Fig. 3.25.Examples of faceted cubic crystals; (a) Crystal delimited by {100} planes, (b) Crystal delimited by {111} planes.

Columnar grain

Welding direction (x)

Tip temperature, 0C

Fig. 3.16. Definition of the local crystal growth rate RL.

Liquidus temperature

Tip velocity, mm/sFig. 3.17. Calculated dendrite tip temperature vs dendrite growth velocity for an Fe-15Ni-15Cr alloy.
The undercooling of the dendrite tip is given by the difference between the liquidus temperature and the
solid curve in the graph. Data from Rappaz et alP

(3-9)
which gives:

(3-10)
Equation (3-10) shows that the local growth rate increases with increasing misalignment of

the crystal with respect to the direction of the maximum temperature gradient in the weld pool.
Since such crystals cannot advance without a corresponding increase in the undercooling ahead
of the solid/liquid interface (see Fig. 3.17), they will soon be outgrowed by other grains which
have a more favourable orientation. Fusion welds of the fee and bcc type will therefore develop a sharp <100> solidification texture in the columnar grain region, similar to that documented for ingots and castings. The weld metal columnar grains may nevertheless be separated by 'high-angle' boundaries, as shown in Fig. 3.1&, due to a possible rotation of the grains
in the plane perpendicular to their <100> length axes.
Example (3.2)

Consider electron beam welding of a 2mm thick single crystal disk of Fe-15Ni-15Cr under
the following conditions:

The orientation of the disk with respect to the beam travel direction is shown in Fig. 3.19.
Calculate on the basis of the minimum velocity (undercooling) criterion the growth rate of
the dendrites trailing the weld pool under steady state welding conditions (assume 2-D
heat flow). Make also schematic drawings of the solidification microstructure in different
sections of the weld. Relevant thermal properties for the Fe-15Ni-15Cr single crystal are
given below:

Solution

Since the base metal is a single crystal, separate columnar grains will not develop. Nevertheless, under 2-D heat flow conditions growth of the dendrites can occur both in the [100] and
the [010] (alternatively the [010]) direction. Referring to Fig. 3.20 the growth rate of the
[100] and the [010] deridrites is given by:

and

Fig. 3.18. Spatial misorientation between two columnar grains growing in the <100> direction (schematic).

Heat source

Weld

Fig. 3.19. Orientation of the single crystal Fe-15Ni-15Cr disk with respect to beam travel direction (Example (3.2)).

Welding direction

Fig. 3.20. Schematic diagram showing the pertinent orientation relations between the fusion boundary
interface normal and the dendrite growth directions (Example (3.2)).
From this it is seen that the velocity of the [100] dendrites is always equal to that of the heat
source v. In contrast, the growth rate of the [010] dendrites depends both on v and a, and will
therefore vary with position along the fusion boundary. It follows from minimum velocity
criterion that the [100] dendrites will be selected when the interface normal angle a is less than
45, while the [010] dendrites will develop at larger angles. This is shown graphically in Fig.
3.21.
At pseudo-steady state the fusion boundary can be calculated from the Rosenthal thin plate
solution (equation (1-83)) according to the procedure shown in Example (1.10). If we include

hk ,

/V

dendrites

a, degrees
Fig. 3.21.Normalised minimum dendrite tip velocity vs interface normal angle a (Example 3.2)).

a correction for the latent heat of melting, the QbIn3 ratio at the melting point becomes:

Substitution of this value into equation (1-83) gives the fusion boundary contour shown in
Fig. 3.22(a). Included in Fig. 3.22 are also schematic drawings of the predicted solidification
microstructure in different sections of the weld.
The results in Fig. 3.22 should be compared with the reconstructed 3-D image of the solidification microstructure in Fig. 3.23, taken form Rappaz et al.17 Due to partial heat flow in the
z-direction, [001] dendrite trunks will also develop. Nevertheless, these data confirm the general validity of equations (3-8) and (3-10) relating crystal growth rate to welding speed and
weld pool shape.
3.4.4 Reorientation of columnar grains
In principle, there are two different ways a columnar grain can adjust its orientation during
solidification in order to accommodate a shift in the direction of the maximum temperature
gradient in the weld pool, i.e.:
(i)
(ii)

Through bowing
Through renucleation.

-y (mm)

(a)

dendrites

Heat source

dendrites

+x (mm)

dendrites
+y (mm)
Fusion boundary
(b)

Fusion zone (4.4 mm)

Base plate

dendrites

dendrites

dendrites

Fig. 3.22. Schematic representation of the weld metal solidification micro structure (Example 3.2));
(a) Top view of fusion zone, (b) Transverse section of fusion zone.

3.4.4.1 Bowing of crystals


A continuous change in the crystal orientation due to bowing will result in curved columnar
grains, as shown previously in Fig. 3.2(a). This type of grain morphology has been observed
in for instance electron beam welded aluminium and iridium alloys.34 Normally, the adjustment of the crystal orientation is promoted by multiple branching of dendrites present within
the grains. Alternatively, the reorientation can be accommodated by the presence of surface
defects at the solid/liquid interface, e.g. screw dislocations, twin boundaries, rotation boundaries, etc. The latter process presumes, however, a faceted growth morphology, and is therefore
of minor interest in the present context.
Example (3.3)

Consider a curved columnar grain of iridium which grows from the fusion boundary towards
the weld centre-line along a circle segment of length L, as shown schematically in Fig. 3.24.
Based on the assumption that the bowing is accommodated solely by branching of [010] dendrites
in the [100] direction, calculate the maximum local growth rate of the crystal during solidification.

y
X

2
Fig. 3.23. Reconstructed 3-D image of solidification microstructure in an electron beam welded Fe-15Ni15Cr single crystal. The letters (a), (b) and (c) refer to [100], [010] and [001] type of dendrites, respectively. After Rappaz et al.17
Weld centre-line

Fusion line

Fig. 3.24. Sketch of curved columnar grain in Example (3.3).


Solution
In principle, the solution to this problem is identical to that presented in Example (3.2). Referring to Fig. 3.24 the growth rate of [100] and the [010] dendrite stems is given by:

and

It follows from Fig. 3.21 that growth will occur preferentially in the [010] direction as long
as the interface normal angle a is larger than 45, while the [100] direction is selected at
smaller angles. This means that the local growth rate of the dendrites, in practice, never will
exceed the welding speed v.
3.4.4.2 Renucleation of crystals
In ingots and castings, three different mechanisms for nucleation of new grains ahead of the
advancing interface are operative:12
(i)
(ii)
(iii)

Heterogeneous nucleation
Dendrite fragmentation
Grain detachment.

The former mechanism is of particular importance in welding, since the weld metal often
contains a high number of second phase particles which form in the liquid state. These particles can either be primary products of the weld metal deoxidation or stem from reactions
between specific alloying elements which are deliberately introduced into the weld pool through
the filler wire. The latter process is also known as inoculation.
Nucleation potency of second phase particles
In general, the effectiveness of individual particles to act as heterogeneous nucleation sites can
be evaluated from a balance of interfacial energies, analogous to that described in Section
3.3.1 for epitaxial nucleation. It follows from the definition of the wetting angle (3 in Fig. 3.5
that the energy barrier to heterogeneous nucleation is a function of both the substrate/liquid
interfacial energy ySL, the substrate/embryo interfacial energy yES, and the embryo/liquid interfacial energy yEL. Complete wetting is achieved when:
(3-11)
Under such conditions, the nucleus will readily grow from the liquid on the substrate. Unfortunately, data for interfacial energies are scarce and unreliable, which makes predictions
based on equation (3-11) rather fortuitous.18
In pure metals, experience has shown that the solid/liquid interfacial energies are roughly
proportional to the melting point, as shown by the data in Fig. 3.25. On this basis, it can be
expected that the higher melting point phases will reveal the highest ySL values, and thus be
nucleants for lower melting phases. A similar situation also exists in the case of non-metallic
inclusions in liquid steel, where the high-melting point phases are seen to exhibit the highest
solid/liquid interfacial energies (see Fig. 3.26).
In contrast, very little information is available on the substrate/embryo interfacial energy
yES. For fully incoherent interfaces, yES would be expected to be of the order of 0.5 to 1 J m~2.5
However, this value will be greatly reduced if there is epitaxy between the inclusions and the
nucleus, which results in a low lattice disregistry between the two phases. In general, assessment of the degree of atomic misfit between the nucleus n and the substrate s can be done on

Melting point, K

Interfacial energy, J / m 2

lnterfacial energy, J/m2

Fig. 3.25. Values of solid/liquid interfacial energy ySL of various metals as function of their melting points.
Data from Mondolfo.18

Melting point, 0 C
Fig. 3.26. Values of interfacial energy 7 5L for different types of non-metallic inclusions in liquid steel at
16000C as function of their melting points. Data compiled from miscellaneous sources.

the basis of the Bramfitt's planar lattice disregistry model :19

(3-12)

a low-index
a low-index
a low-index
a low-index

where

plane of the substrate;


direction in (hkl)s
plane in the nucleated solid;
direction in (hkl)n;

the interatomic spacing along [wvw]n;


the interatomic spacing along [WVH>]5; and
the angle between the [wvw]^ and the [wvw]w.

Undercooling, 0C

In practice, the undercooling Ar (which is a measure of the energy barrier to heterogeneous


nucleation) increases monotonically with increasing values of the planar lattice disregistry, as
shown by the data in Fig. 3.27. This means that the most potent catalyst particles are those
which also provide a good epitaxial fit between the substrate and the embryo. Examples of
such catalyst particles are TiAl3 in aluminium 18 and TiN in steel.19 Nucleation of delta ferrite
at titanium nitride will be considered below.

>
Fig. 3.27. Relationship between planar lattice disregistry and undercooling for different nucleants in
steel. Data compiled from miscellaneous sources.

Example (3.4)

In low-alloy steel weld metals, titanium nitride can form in the melt due to interactions between dissolved titanium and nitrogen. Assume that the TiN particles are faceted and delimited by {100} planes. Calculate on the basis of equation (3-12) the minimum planar lattice
disregistry between TiN and the nucleating delta-ferrite phase under the prevailing circumstances. Indicate also the plausible orientation relationship between the two phases. The
lattice parameters of delta ferrite and TiN at 15200C may be taken equal to 0.293 and 0.43 lnm,
respectively.
Solution

Titanium nitride has the NaCl crystal structure, while delta ferrite is body-centred cubic, as
shown in Fig. 3.28(a) and (b). It is evident from Fig. 3.29(a) that a straight cube-to-cube orientation relationship between TiN and 8-Fe will not result in a small lattice disregistry.
However, the situation is largely improved if the two phases are rotated 45 with respect to
each other (see Fig. 3.29(b)), conforming to the following orientation relationship:

The resulting crystallographic relationship at the interface is shown schematically in Fig.


3.29(c). Since the lattice arrangements are similar in this case, equation (3-12) reduces to:

A comparison with the data in Fig. 3.27 shows that the calculated lattice disregistry conforms to an undercooling of about 1 to 2C. This value is sufficiently small to facilitate heterogeneous nucleation of new grains ahead of the advancing interface during solidification.
Considering other inclusions with more complex crystal structures, the chances of obtaining a small planar lattice disregistry between the substrate and the delta ferrite nucleus are

Fe- atoms

N-atoms

Ti- atoms
(a)

(b)

Fig. 3.28. Crystal structures of phases considered in Example (3.4); (a) Titanium nitride, (b) Delta ferrite.

TiN

(a)

TiN

(C)
(b)

Ti atoms

N atoms

8-Fe atoms

Fig. 3.29. Possible crystallographic relationships between titanium nitride and delta ferrite (Example (3.4));
(a) Straight cube-to-cube orientation, (b) Twisted cube-to-cube orientation, (c) Details of lattice arrangement along coherent TiN/d-Fe interface.

rather poor (see Fig. 3.27). Nevertheless, such particles can act as favourable sites for heterogeneous nucleation if 7 ^ is sufficiently large compared with yEL and 7 ^ . This is illustrated by
the following example:
Example (3.5)

In low-alloy steel weld metals 7-Al2O3 inclusions can form during the primary deoxidation
stage as discussed in Section 2.12.4.2 (Chapter 2). Based on the classic theory of heterogeneous nucleation, evaluate the nucleation potency of such inclusions with respect delta ferrite.

Solution

It is readily seen from Fig. 3.27 that the planar lattice disregistry between delta ferrite and
Al2O3 is very large, which indicates of a fully incoherent interface (i.e. yES 0.75 J m" 2 ).
Moreover, readings from Figs. 3.25 and 3.26 give the following average values for the delta ferrite/liquid and the inclusion/liquid interfacial energies:

and

According to equation (3-11) complete wetting is achieved when ySL > yES + yEL. This
requirement is clearly met under the prevailing circumstances.
Similar calculations can also be performed for other types of non-metallic inclusions in
steel weld metals. The results are presented graphically in Fig. 3.30. It is evident that the
nucleation potency of the inclusions increases in the order SiO2-MnO, Al 2 O 3 -Ti 2 O 3 -SiO 2 MnO, Al2O3, reflecting a corresponding increase in the inclusion/liquid interfacial energy ySL.
The resulting change in the weld metal solidification microstructure is shown in Fig. 3.31,
from which it is seen that both the average width and length of the columnar grains decrease
with increasing Al2O3-contents in the inclusions. This observation is not surprising, considering the characteristic high solid/liquid interfacial energy between aluminium oxide and steel
(see Fig. 3.26). The important effect of deoxidation practice on the weld metal solidification
microstructure is well documented in the literature.320"22
Rate of heterogeneous nucleation
It can be inferred from the classic theory of heterogeneous nucleation that the nucleation rate

Complete wetting

No wetting

Embryo'

G*he/AGhom

p (degrees)

Inclusion

(Y sf Y ES )/7 EL
Fig. 3.30. Nucleation potency of different weld metal oxide inclusions with respect to delta ferrite.

AI2O3 content (wt%)

(a)

Average width of grains, ^i m

95% confidence limit

Pure AI2O3

(A%AI)weld/[%O]anaL
AI2O3 content (wt%)

(b)

Average length of grains, ^m

95% confidence limit


Pure AI2O3

<A%AIW[%0W
Fig. 3.31.Effect of deoxidation practice on the columnar grain structure in low-alloy steel weld metals;
(a) Average width of columnar grains, (b) Average length of columnar grains. Data from Kluken et al.22

Nhet(i.e. number of nuclei which form per unit time and unit volume of the melt) is interrelated to the energy barrier AG^ through the following equation:5

(3-13)
whereZ1 is a frequency factor, Nv is the density of nucleation sites per unit volume of the melt,

AGD is the activation energy for diffusion of atoms across the interface, and k is the Boltzmann
constant.
Since AGD is often negligible compared with AG^et in liquids, equation (3-13) reduces to:

(3-14)
Equation (3-14) shows that the nucleation rate Nhet depends both on Nv and &Ghe{. Hence,
under full wetting conditions (AGhet ~ 0), the number of nuclei which form per second and mm3
ahead of the advancing interface is directly proportional to the instantaneous concentration of
catalyst particles in the melt. Examples of such particles are TiAl3 in aluminium and TiN/Al2O3
in steel. The important effect of controlled titanium additions and subsequent TiAl3 precipitation on the columnar grain structure in 1100 aluminium welds is illustrated in Fig. 3.32.

Critical cell/dendrite alignment angle


It follows from the minimum growth rate criterion and the definition of the local crystal growth
rate in equation (3-10) that reorientation of the columnar grains will occur when the cell/
dendrite alignment angle $ reaches a critical value <\>*. The value of 4>* will, in turn, depend on
the nucleation potency of the catalyst particles and can be estimated for different types of
welds. If growth of the columnar grains is assumed to occur along a circle segment of length
L (see Fig. 3.33), the critical cell/dendrite alignment angle is given by:
(3-15)

Average width of grains, (x m

where co is the total grain rotation angle, and / is the average length of the columnar grains (in
mm).

Titanium content, wt%


Fig. 3.32. Effect of titanium on the columnar grain structure in 1100 aluminium welds. The value Y is
the fractional distance from fusion line to top surface of weld metal. Data from Yunjia et alP

Weld metal

Fig. 3.33. Characteristic growth pattern of columnar grains in bead-on-plate welds (schematic).
By introducing reasonable average values for co and / in the case of SA welding of lowalloy steel,22 we obtain:
(3-16)
Calculated values for 4>* in steel weld metals are presented in Fig. 3.34, using data from
Kluken et al.22 An expected, the critical cell/dendrite alignment angle in fully aluminium
deoxidised steel welds is seen to be very small (of the order of 2), reflecting the fact that
nucleation of delta ferrite occurs readily at Al2O3 inclusions. The value of 4>* increases gradually with decreasing Al2O3 contents in the inclusions and reaches a maximum of about 4 for
Si-Mn deoxidised steel weld metals. This situation can be attributed to less favourable nucleating opportunities for delta ferrite at silica-containing inclusions, which reduces the possibilities of obtaining a change in the crystal orientation during solidification through a nucleation
and growth process.
Dendrite fragmentation
In principle, nucleation of new grains ahead of the advancing interface can also occur from
random solid dendrite fragments contained in the weld pool. Although the source of these
solid fragments has yet to be investigated, it is reasonable to assume that they are generated by
some process of interface fragmentation due to thermal fluctuations in the melt or mechanical
disturbances at the solid/liquid interface.3 At present, it cannot be stated with certainty whether
grain refinement by dendrite fragmentation is a significant process in fusion welding.26
Grain detachment
Since the partially melted base metal grains at the fusion boundary are loosely held together by
liquid films between them, there is also a possibility that some of these grains may detach
themselves from the base metal and be trapped in the solidification front.26 Like dendrite
fragments, such partially melted grains can act as seed crystals for the formation of new grains
in the weld metal during solidification if they are able to survive sufficiently long in the melt.

R L / v cos a

Next Page

Calculated from
equation (3-10)

Critical cell/dendrite alignment angle ($*)


Fig. 3.34. Critical cell/dendrite alignment angle ()>* for reorientation of delta ferrite columnar grains during solidification of steel weld metals. Data from Kluken et al.22

3.5 Solidification Microstructures


So far, we have discussed growth of columnar grains without considering in detail the weld
metal solidification microstructure. In general, each individual grain will exhibit a substructure consisting of a parallel array of dendrites or cells. This substructure can readily be revealed by etching, also in cases where it is masked by subsequent solid state transformation
reactions (as in ferrous alloys).2224
3.5.1 Substructure characteristics
A cellular substructure within a single grain consists of an array of parallel (hexagonal) cells
which are separated from each other by 'low-angle' grain boundaries, as shown schematically
in Fig. 3.35. In the presence of solute, these boundaries respond to etching even in the absence
of segregation. When the cellular to dendritic transition occurs, the cells become more distorted and will finally take the form of irregular cubes, as indicated by the optical micrograph
in Fig. 3.36. This is actually a dendritic type of substructure, where the formation of secondary
and tertiary dendrite arms is suppressed because of a relatively small temperature gradient in
the transverse direction compared with the longitudinal (growth) direction. Fully branced
dendrites may, however, develop in the centre of the weld if the thermal conditions are favourable. Branching will then occur in specific crystallographic directions, e.g. along the three
<100> easy growth directions for bcc and fee crystals, as illustrated in Fig. 3.37.
Besides the difference in morphology, the distinction between cells and dendrites lies primarily in their sensitivity to crystalline alignment. Cells do not necessarily have the <100>
axis orientation, while dendrites do.2 Hence, cells can grow with their axes parallel to the heat
flow direction, regardless of the crystal orientation. This important point is often overlooked
when discussing competitive grain growth in fusion welding.

R L / v cos a

Previous Page

Calculated from
equation (3-10)

Critical cell/dendrite alignment angle ($*)


Fig. 3.34. Critical cell/dendrite alignment angle ()>* for reorientation of delta ferrite columnar grains during solidification of steel weld metals. Data from Kluken et al.22

3.5 Solidification Microstructures


So far, we have discussed growth of columnar grains without considering in detail the weld
metal solidification microstructure. In general, each individual grain will exhibit a substructure consisting of a parallel array of dendrites or cells. This substructure can readily be revealed by etching, also in cases where it is masked by subsequent solid state transformation
reactions (as in ferrous alloys).2224
3.5.1 Substructure characteristics
A cellular substructure within a single grain consists of an array of parallel (hexagonal) cells
which are separated from each other by 'low-angle' grain boundaries, as shown schematically
in Fig. 3.35. In the presence of solute, these boundaries respond to etching even in the absence
of segregation. When the cellular to dendritic transition occurs, the cells become more distorted and will finally take the form of irregular cubes, as indicated by the optical micrograph
in Fig. 3.36. This is actually a dendritic type of substructure, where the formation of secondary
and tertiary dendrite arms is suppressed because of a relatively small temperature gradient in
the transverse direction compared with the longitudinal (growth) direction. Fully branced
dendrites may, however, develop in the centre of the weld if the thermal conditions are favourable. Branching will then occur in specific crystallographic directions, e.g. along the three
<100> easy growth directions for bcc and fee crystals, as illustrated in Fig. 3.37.
Besides the difference in morphology, the distinction between cells and dendrites lies primarily in their sensitivity to crystalline alignment. Cells do not necessarily have the <100>
axis orientation, while dendrites do.2 Hence, cells can grow with their axes parallel to the heat
flow direction, regardless of the crystal orientation. This important point is often overlooked
when discussing competitive grain growth in fusion welding.

Cell wall'

Fig. 3.35. Schematic representation of the cellular substructure.

Fig. 3.36. Optical micrograph showing the characteristic cellular-dendritic substructure in a low-alloy
steel weld. The metallographic section is normal to the columnar grain growth direction. After Kluken
etal.22

Fig. 3.37. Schematic representation of the development of


primary, secondary and tertiary dendrite arms along <100>
directions in cubic crystals.

A characteristic feature of cellular and cellular-dendritic growth is also that the boundary
between two adjacent columnar grains will closely follow the contours of the original cell
boundaries. An illustration of this point is contained in Fig. 3.38. Consequently, since all cell
walls are preferential sites for segregation during solidification (see ion micrograph in Fig.
3.39), the presence of solute at the columnar grain boundaries can strongly alter the kinetics of
subsequent solid state transformation reactions. The indications are that for instance phosphorus segregations at prior austenite grain boundaries will promote the formation of grain boundary ferrite in low-alloy steel weld metals during the 7 to a transformation because of the
associated increase in the A<?3-temperature.22'25 We will return to this question in Chapter 6.

(a)

(b)

Fig. 3.39. Ion (SIMS) micrograph showing evidence of phosphorus segregations at primary solidification (cell) boundaries in a low-alloy steel weld. After Kluken et al22

3.5.2 Stability of the solidification front


The stability of the solid/liquid interface is critical in determining the microstructural characteristics of the weld metal.
3.5.2.1 Interface stability criterion
The simplified treatment given here is based on the constitutional undercooling criterion which
ignores the important effect of capillarity on the interface stability.
Consider a simple binary system, the phase diagram of which is given in Fig. 3.40(a).
During solidification, a solute-rich layer will form in front of the growing interface, as shown
schematically in Fig. 3.40(b). At steady state, the mass of solute transferred from the liquid to
the interface, R1(C1Jp is equal to the mass of solute which accumulates in the solid, R1(C8)^
minus the amout which diffuses back into the liquid, D1(O1C1Za1X)1. Hence, we may write:

(3-17)
Provided that equilibrium exists at the solid/liquid interface, i.e. (Cs)t = IcJC1J1, equation
(3-17) can be rewritten as:
(3-18)
where DL is the diffusivity of the solute in the liquid, and ko is the equilibrium partition coefficient (Jco < 1).

Temperature

(a)

Concentration

Liquid composition

(b)

Distance (x)

Temperature

(C)

Liquidus temperature (TL)

Constitutionally undercooled region

Distance (x)
Fig. 3.40. Constitutional undercooling in alloy solidification; (a) Schematic representation of binary phase
diagram, (b) Build-up of solute-enriched layer in front of solid/liquid interface, (c) Undercooled region
ahead of solid/liquid interface.

With the aid of the phase diagram in Fig. 3.40(a) it is easy to verify that the equilibrium
liquidus temperature increases with distance from the interface because of the lower solute
content. Referring to Fig. 3.40(c) the latent heat of melting AHm will diffuse away from the
interface (thereby stabilising possible interface protuberances) if the actual temperature gradient in the liquid (dTA/dx)t is less than the equilibrium liquidus temperature gradient (3T1ZdX)1.
The latter gradient can be expressed as:
(3-19)

where (dTL/dCL)i denotes the slope of the liquidus curve in the phase diagram (designated mL).
By combining equations (3-18) and (3-19), we obtain the following criterion for the interface stability:
(3-20)

Taking (C1). = C0Ik0 and Rj = RL this equation can alternatively be written as:

(3-21)

3.5.2.2 Factors affecting the interface stability


It follows from equation (3-21) that the stability of the weld metal solidification front is controlled by the extent of constitutional undercooling ahead of the advancing interface, and is
therefore influenced by factors such as the total amount of alloying or impurity elements present,
C0, the local crystal growth rate, RL, and the thermal gradient in the weld pool, GL. The
combined effects of alloy content, growth rate, and thermal gradient are presented schematically
in Fig. 3.41. It can be seen from the figure that a planar solidification front is favoured by a
high GL/RL ratio, in agreement with predictions based on equation (3-21). At lower GLIRL
ratios, the morphology of the interface changes to cellular, cellular-dendritic or dendritic, depending on the degree of constitutional undercooling ahead of the advancing interface.
Normally, the GL/RL ratio close to the fusion line is large enough to facilitate cellular solidification. The temperature gradient in the weld pool decreases, however, with distance from
the fusion boundary and hence, a cellular-dendritic type of substructure is often observed in
the central areas of the columnar grain region.22 This substructure may change to dendritic
close to the weld centre-line if the degree of constitutional undercooling in front of the advancing interface is sufficiently large. The various types of growth products that may develop
during normal solidification of fusion welds are shown schematically in Fig. 3.42.
Example (3.6)

In aluminium welding, binary Al-Si and Al-Mg alloys are frequently used as filler metals.
Consider stringer bead deposition with an Al-4.5wt%Si filler wire at a constant welding speed
of 1.5mm s"1. Based on equation (3-21) calculate the critical (minimum) temperature gradient

Alloying level (C )

Planar

GL/RL
Fig. 3.41. Schematic representation of the combined effect of crystal growth rate RL and melt thermal
gradient GL on the weld metal solidification microstructure.

Fusion zone
v

Larger R L,
Small RL, Small GL
Large G1

Weld centre-line

Equiaxed-dendritic
Cellular-dendritic
Cellular

HAZ grains

Fig. 3.42. Schematic diagram illustrating structural variations in the weld metal solidification microstructure across the fusion zone.
in the weld pool which gives a planar solidification front. Relevant physical data for the Al-Si
system are given below:2

Solution

When mL, ko and D1 are known the critical temperature gradient (G1)cr can readily be calculated from equation (3-21). Taking RL ~ v, we obtain:

The above calculations show that the temperature gradient in the weld pool must be extremely large in order to promote a planar solidification front. The calculated value for (GL)cr
corresponds to a cooling rate of about 92 4000C s"1 (CR. = G1R1). Besides low-heat input
electron beam and laser welding, such high cooling rates are rarely observed in fusion welding
(see Fig. 3.3). On this basis it is not surprising to find that the weld metal solidification microstructure is normally of the cellular or the dendritic type.
For a rough evaluation of the weld metal solidification microstructure the diagram in Fig.
3.43 can be used. This diagram summarises the various microstructures which can be obtained
(using a typical alloy with a melting range of 500C) when the imposed temperature gradient GL
and crystal growth rate RL are varied. Moving from the lower left to the upper right along the
lines at 45 leads to a refinement of the structure without changing the morphology (constant
GL/RL ratio). Crossing these lines by passing from the upper left to the lower right leads to
changes in the morphology from planar to cellular or dendritic growth, while the scale of the
microstructure remains essentially the same. The gray bands define the regions over which
one structure changes into another. An example of the application of this diagram is given
below.
Example (3.7)

Consider GTA butt welding of a 2mm thin sheet of aluminium (Al-Mg alloy) under the following conditions:

Based on the diagram in Fig. 3.43, estimate the weld metal solidification microstructure in
different positions from the fusion boundary. In these calculations we shall assume that the
actual thermal gradient in the weld pool is equal to the average thermal gradient within the
temperature interval from 680 to 5500C, as evaluated from the Rosenthal thin plate solution
(equation (1-83)). Relevant thermal data for the Al-Mg alloy are given in Table 1.1 (Chapter
1).
Solution

The contours of the fusion boundary and the 680 and the 5500C isotherms can be calculated
according to the procedure shown in Example (1.10). If we neglect the latent heat of melting,
the corresponding 68//?3 ratios at these temperatures become:

GL,C/mm

RL, mm/s
Fig. 3.43. Variation of weld metal solidification microstructure with GL and RL. The diagram is based on
the ideas of Kurz and Fisher.2

Substitution of the above values into equation (1-83) gives the isothermal contours shown
in Fig. 3.44. It follows from Fig. 3.44 that the thermal gradient decreases from about 25C
mnr 1 at the fusion boundary to approximately 100C mm"1 close to the weld centre-line. This
occurs parallel with an increase in the nominal crystal growth rate from 0.8 to 4mm s"1.
According to Fig. 3.43 the calculated values of GL and RL conform to a cellular-dendritic
solidification microstructure within the central regions of the weld and an equiaxed-dendritic
microstructure close to the weld centre-line. Both types of substructures are commonly observed in aluminium weldments.2728
Weld pool (680 0C)
Fusion boundary (650 0C)
HAZ (550 0C)

+x (mm)

+Y (mm)
Fig. 3.44. Predicted shape of fusion boundary and neighbouring isotherms during thin plate aluminium
welding (Example (3.6)).

3.5.3 Dendrite morphology


Within the fully dendritic region of the weld, the dendrite morphology remains largely unchanged over a wide range of cooling rates. Nevertheless, it will become finer as the heat is
extracted at greater rates.

Tip radius,\irr\

3.5.3.1 Dendrite tip radius


A combination of stability analysis with diffusion considerations and the specification of a
dendrite tip shape and a radius selection criterion can lead to an unique solution for the morphology and solute distribution around a growing dendrite.2930 Based on such analyses it is
possible to calculate the dendrite tip radius rd as function of the growth rate RL and the thermal
gradient GL in the weld pool for a given alloy system.17
Referring to Fig. 3.45, the dendrite growth is seen to occur between two stability limits. At
low speeds, the rd vs R1 curve reaches a threshold where the tip radius increases rapidly and
finally approaches an infinite value corresponding to a planar solidification front. At high
speeds, an absolute stability limit is also found. This corresponds to the limiting case where
the transport of solute across the interface becomes too sluggish to keep pace with the advancing solidification front. However, within the central areas of the curve, the dendrite tip radius
is seen to decrease steadily with increasing values of RD typically as l/(RL)m. Note that the
thermal gradient in the weld pool GL will have a strong influence on the low growth rate
stability limit. In contrast, the absolute stability limit is essentially independent of GL, as
indicated in Fig. 3.45.

Growth rate, mm/s


Fig. 3.45. Predicted relationship between dendrite tip radius rd, crystal growth rate RL and melt thermal
gradient GL for a Fe-15Cr-15Ni alloy. Data from David and Vitek.4

3.5.3.2 Primary dendrite arm spacing


The primary dendrite arm spacing (defined in Fig. 3.46) is an important characteristic of the
solidification microstructure and has a marked effect on the weld metal mechanical properties.
If it is assumed that the dendrite envelope, representing the mean cross-section of the trunk,
can be described by an ellipse (see Fig. 3.47), the radius of curvature of the ellipse is given by:2
(3-22)
In general, the half width b of the dendrite stem is proportional to the primary arm spacing
X1. In cases where the arrangement of the dendrite trunks can be represented by a simple
close-packed hexagonal array, the value of b is exactly equal to X1 IV3 -2 Similarly, the total
length of the dendrite stem g can be calculated by considering the difference between the tip
temperature Ttip and the root temperature Troot, as shown in Fig. 3.48. Taking the thermal
gradient in the mushy zone equal to GL, we obtain:

(3-23)

This gives the following relationship between the primary dendrite arm spacing X1, the tip
radius rd and the thermal gradient GL:
(3-24)

rd

b
Fig. 3.46. Definition of primary X1 and secondary X2 dendrite
arm spacings.

tig. 3.47.Geometrical relationship


between the radius of curvature rd
and the width to length ratio of an
elliptical dendrite.

Ttip

Troot
9
Heat flow

Cell/dendrite

Fig. 3.48. Definition of the dendrite tip Ttip and root Troot temperatures.
If we also take into account that the dendrite tip radius rd is inversely proportional to the
square root of the crystal growth rate RL within the central range of RL (see Fig. 3.45), equation
(3-24)reduces to:
(3-25)

where c{ is a kinetic constant which is characteristic of the alloy system under consideration.
It is evident from the above analysis that the primary dendrite arm spacing cannot readily
be characterised by one single parameter (e.g. the cooling rate), since its dependence on GL
and R1 have different exponents.

Example (3.8)

Based on equation (3-25), show that the following relationships exist between the primary
dendrite arm spacing X1, the net arc power qo, and the weld travel speed v during thick plate
and thin plate welding, respectively:
Thick plate welding:
(3-26)
Thin plate welding:
(3-27)
where c2 and c3 are proportionality constants.

Solution

Under 3-D heat flow conditions, the pseudo-steady state temperature distribution is given by
the Rosenthal thick plate solution (equation (1 -45)). For points located on the weld centre-line
behind the heat source y = z = 0, and R* = -x. Hence, equation (1-45) reduces to:

The thermal gradient GL in the mushy zone close to the weld centre-line can be obtained
by differentiating of the above equation with respect to the x-coordinate:

(3-28)

By inserting the appropriate expressions for GLandRL into equation (3-25) (noting that RL=v
at the weld centre-line), we arrive at an expression for X1 which is identical with the one
presented in equation (3-26).
Similarly, under 2-D heat flow conditions, the pseudo-steady state temperature distribution
is given by the Rosenthal thin plate solution (equation 1-81). For points located on the weld
centre-line behind the heat source y = 0 and r = -x = vt. If IxI is sufficiently large, it is a fair
approximation to set Ko(u) ~ exp(-u)^n!2u
. Hence, equation (1-81) reduces to:

The thermal gradient GL is then given as:

(3-29)
From this we see that the primary dendrite arm spacmg X1 during thin plate welding is
interrelated to qo and v through a relationship of the type shown in equation (3-27).
In Fig. 3.49 the validity of equation (3-26) has been checked against the experimental data
of Jordan and Coleman,27 who measured the primary dendrite arm spacing in different GMA
Al-Mg-Mn welds. It is evident from this plot that their data-points can approximately be
represented by straight lines passing through the origin, as required by the theory. Moreover,
a closer inspection of the figure shows that the dendrite arm spacing close to the centre-line
varies systematically from the bottom to the top of the weld. This observation is not surprising, considering the fact that the rate of solidification increases progressively from the toe to
the surface of the plate.
Example (3.9)

Consider GTA butt welding of a 2mm thin aluminium sheet (Al-Mg alloy) under conditions
similar to those employed in Example (3.7). Based on equation (3-25) calculate the relative
change in the primary dendrite arm spacing X1 from the fusion boundary to the weld centreline during solidification.

Primary dendrite arm spacing, p,m

GMAW (Al-Mg-Mn alloys)


Just above toe of weld
Level with surface of
plate

Parallel to plate
surface
Just above toe.
of weld

[(qo)1/2/(v)1/4],(WS1/2/mm1/2)1/2
Fig. 3.49. Experimental verification of equation (3-26). Data from Jordan and Coleman.27

Solution

As shown in Fig. 3.44, the thermal gradient, GL, decreases from about 25C mm"1 at the
fusion boundary to approximately 10C mm"1 close to the weld centre-line. This occurs
parallel with an increase in the nominal crystal growth rate from 0.8 to 4mm s"1. Taking the
primary dendrite arm spacing at the fusion boundary equal to X1*, the ratio X1 /X1* in different
positions of the weld becomes:
Central region of columnar zone:

Weld centre-line:

In contrast to that predicted above, the smallest dendrite arm spacing is normally observed at
the weld centre-line.26 This has to do with the fact that the constant C1 in equation (3-25), in
practice, decreases with increasing distance from the fusion boundary due to solute segregation, which gradually reduces the coarsening rate of the dendrites.
3.5.3.3 Secondary dendrite arm spacing
It follows from the above analysis that the primary dendrite arm spacing, once it has been

established, will remain constant during subsequent cooling of the weld. This is not true of the
secondary arms (see definition in Fig. 3.46), which undergo a continuous ripening process.
Hence, their size and morphology will change with time as the thicker branches grow larger at
the expense of the smaller ones.2
The driving force for the ripening process is the disparity in chemical potential (interfacial
energy) between branches with different curvature. Since this process is analogous to the
Ostwald ripening of precipitates (see equation (2-76)), the spacing of the branches, X2, will be
a simple cube root function of the local solidification time fo:1'2
(3-30)
where M is a mobility term defined as:2
(3-31)

The parameter Q in equation (3-31) refers to the Gibbs-Thomson coefficient and is given as:
(3-32)
where a is the solid/liquid interfacial energy, and AS, is the entropy of fusion.
In practice, the value of M can easily vary by an order of magnitude. Nevertheless, its effect
on the secondary dendrite arm spacing is rather weak, since X2 is proportional to the cube root
of M. Hence, a plot of the secondary dendrite arm spacing vs the local solidification time in a
logarithmic diagram will normally reveal a straight-line relationship between X2 and to with a
slope close to 0.33, as shown in Fig. 3.50.

X2^m

GTAW (Stainless steel)

Slopes 0.33

to'5
Fig. 3.50. Relation between secondary dendrite arm spacing X2 and local solidification time to in type
AISI 310 stainless steel welds. Data from Kou and Lee.31

Example (3.10)

Based on equation (3-30), derive a relationship between the secondary dendrite arm spacing
X2, the net arc power qo, and the weld travel speed v during thick plate welding. Calculate then
the secondary dendrite arm spacing in the centre of a thick GTA Al-Si weld deposited under
the following conditions:

Relevant physical data for the Al-Si system are given below:2

Solution

The local solidification time to is the time for the dendrite array to pass an arbitrary point in the
weld and is therefore a measure of the solidification rate. Referring to Fig. 3.48, the local
solidification time is defined as:
(3-33)

By inserting the appropriate expressions for GL (equation (3-28)) and RL in position y = z = 0,


we obtain the following relationship between to, qo and v during thick plate welding:

(3-34)

A combination of equations (3-30) and (3-34) gives:

(3-35)

In the present example, the numerical values of M and X2 are:

and

Although reliable experimental data are not available for a direct comparison, the calculated value for X2 is considered reasonably correct. In Al-Cu castings a secondary dendrite
arm spacing of about 4(im corresponds to a cooling rate of the order of 103 C s"1, as shown in
Fig. 3.51. In the present example, the cooling rate at the solid/liquid interface is close to:

This value is compatible with a secondary dendrite arm spacing of 4|am.


Example (3.11)

The operating parameters given in Example (3.10) are also applicable to single pass butt welding of thin aluminium plates. Based on equation (3-30), estimate the secondary dendrite arm
spacing in the centre of a 2mm thick Al-Si butt weld deposited under such conditions.
Solution

In thin plate welding the local solidification time to is obtained by combining equations (3-29)
and (3-33). Noting that RL = v at the weld centre-line, we get:

(3-36)

X 2 ,nm

AI-4.5 wt% Cu

Cooling rate, C/s


Fig. 3.57.Relation between secondary dendrite arm spacing X2 and cooling rate CR. in Al-Cu castings.
Data compiled by Munitz.32

The secondary dendrite arm spacing is thus given as:

(3-37)

Taking pc equal to 0.0027 J mm

C ] in the case of Al-Si alloys, we obtain:

It is evident from the above calculation that the secondary dendrite arm spacing is significantly coarser in thin plate welding compared with thick plate welding. This observation is not
surprising, considering the pertinent difference in the heat transfer mode and thus the cooling rate between these two types of weldments. In the former case, we have:

A comparison with the data in Fig. 3.51 shows that a cooling rate of about 78C s"1 is
compatible with a secondary dendrite arm spacing of 13 jim.

3.6 Equiaxed Dendritic Growth


As already stated in the introduction of the chapter, an equiaxed zone is often observed close to
the weld centre-line. This zone can in certain cases be very dominating and completely override the columnar grain zone, as shown in Fig. 3.52.
As long as growth occurs in a columnar manner the crystals are in contact with solid metal.
The heat will therefore be conducted through the crystals in a direction which is parallel and
opposite to their growth direction, as illustrated in Fig. 3.53(a). In equiaxed dendritic growth
the situation is different. Here the melt will be slightly undercooled because of extensive segregation of solute to the weld centre-line where the solidification fronts growing from each
side impinge. This makes the crystals hotter than the liquid and gives rise to a radial heat flow
away from the crystals in the same direction as that of growth (see Fig. 3.53(b)).
3.6.1 Columnar to equiaxed transition
In fusion welds, the GL/RL ratio decreases continuously from the fusion boundary to the weld
centre-line during the crystallisation process. For a given alloy system containing a fixed
number of heterogenous nucleation sites it is reasonable to assume that the columnar to equiaxed

CL

FL

FL

CL

Fig. 3.52.Example of grain refinement in an Al-Mg-Si plasma arc weld due to TiAl3-precipitation (FL:
fusion line, CL: centre-line). Courtesy of M.I. Onsoien, SINTEF, Trondheim, Norway.
transition occurs when the GJR1 ratio drops below a certain critical value fcn. Hence, we may
write:
(3-38)
During thick plate welding (3-D heat flow) the temperature gradient at the weld centre-line
is given by equation (3-28). Taking RL = v, a combination of equations (3-28) and (3-38) gives:

(3-39)

where c4 is a kinetic constant which is characteristic of the alloy system under consideration.
Similarly, for welding of medium thick plates (mixed heat flow), we get:
(3-40)
where c5 is a new kinetic constant, and n is an exponent which varies between zero and unity
depending on the mode of heat flow (i.e. thin, medium thick or thick plate welding).

(a)

(b)

Heat flow

Fig. 3.53. Schematic diagrams showing different dendrite growth morphologies in castings; (a) Columnar dendritic growth, (b) Equiaxed dendritic growth. The diagrams are based on the ideas of Kurz and
Fisher.2

Equation (3-40) predicts that the columnar to equiaxed transition in fusion welds occurs at
critical combinations of the net arc power qo and the welding speed v. Thus, a decrease in qo
must always be compensated by a corresponding increase in v in order to maintain equiaxed
dendritic growth at the weld centre-line. This is also in agreement with general experience
(see Fig. 3.54).
Example (3.12)

Based on the experimental data in Fig. 3.54, calculate the critical G1IR1 ratio which provides
equiaxed dendritic growth during thick plate welding of Al-Mg alloys (5083 series). Relevant
thermal data for the Al-Mg system are given in Table 1.1 (Chapter 1).

Net arc power, kW

Plate thickness: 10-12 mm

Welding speed, mm/s


Fig. 3.54. Effect of net arc power qo and welding speed v on the columnar to equiaxed transition in
different aluminium welds. Data from Matsuda et al.33

Solution

The critical G1IR1 ratio is obtained by rearranging equation (3-39):

By inserting data from Fig. 3.54 and Table 1.1, we get:

By multiplying fcr with the welding speed we see that the calculated value corresponds to
a critical temperature gradient of about 928C mm"1. This means that the thermal conditions
existing in welding will favour growth of equiaxed dendrites close to the weld centre-line. A
requirement is, of course, that the melt contains a sufficient number of seed crystals to facilitate heterogeneous nucleation of new grains ahead of the advancing solid/liquid interface.

Next Page

3.6.2 Nucleation mechanisms


Several nucleation mechanisms have been proposed to explain the columnar to equiaxed transition during weld pool solidification, including:26
(i)
(ii)
(iii)

Heterogeneous nucleation
Dendrite fragmentation
Grain detachment.

The former mechanism is particularly relevant to welding, since the weld metal often contains a high number of second phase particles which form in the liquid state. As already
mentioned in Section 3.4.4.2, these particles can either be primary products of the weld metal
deoxidation or stem from reactions between specific alloying elements which are deliberately
introduced into the weld pool through the filler wire. The important effect of deoxidation
practice (inclusions) on the columnar to equiaxed transition in ferritic stainless steel GTA
welds is shown in Figs. 3.55 and 3.56.

3.7 Solute Redistribution


During solidification of fusion welds, alloying and impurity elements tend to segregate extensively to the centre parts of the intercellular or interdendritic spaces under the conditions of
rapid cooling. This, in turn, alters the kinetics of the subsequent solid state transformation
reactions.
3.7.1 Micro segregation
In general, microsegregation in ingots and castings is caused by the interplay between a number
of variables which cannot readily be accounted for in a mathematical simulation of the process 12 Nevertheless, several important deductions can be made from the well known nonequilibrium lever rule or Scheil equation which, in spite of its simple nature, gives a reasonable description of the segregation pattern during weld metal solidification.4'34
The Scheil equation applies to directional solidification of long bars and is based on the
following assumptions:
(i)
(ii)
(iii)
(iv)
(v)

Uniform liquid composition (i.e. complete mixing in the liquid state).


A flat solid/liquid interface.
Local equilibrium at the solid/liquid interface (ko is constant).
Negligible solid-state diffusion.
Equal solid and liquid densities.

Under such conditions it is fairly simple to derive an expression for the solute concentration
in the metal as a function of the fraction solidified. Referring to Fig. 3.57, a mass balance
gives:
(3-41)
Since dz z and Cs = ko C1, equation (3-41) can be rewritten as:

Previous Page

3.6.2 Nucleation mechanisms


Several nucleation mechanisms have been proposed to explain the columnar to equiaxed transition during weld pool solidification, including:26
(i)
(ii)
(iii)

Heterogeneous nucleation
Dendrite fragmentation
Grain detachment.

The former mechanism is particularly relevant to welding, since the weld metal often contains a high number of second phase particles which form in the liquid state. As already
mentioned in Section 3.4.4.2, these particles can either be primary products of the weld metal
deoxidation or stem from reactions between specific alloying elements which are deliberately
introduced into the weld pool through the filler wire. The important effect of deoxidation
practice (inclusions) on the columnar to equiaxed transition in ferritic stainless steel GTA
welds is shown in Figs. 3.55 and 3.56.

3.7 Solute Redistribution


During solidification of fusion welds, alloying and impurity elements tend to segregate extensively to the centre parts of the intercellular or interdendritic spaces under the conditions of
rapid cooling. This, in turn, alters the kinetics of the subsequent solid state transformation
reactions.
3.7.1 Micro segregation
In general, microsegregation in ingots and castings is caused by the interplay between a number
of variables which cannot readily be accounted for in a mathematical simulation of the process 12 Nevertheless, several important deductions can be made from the well known nonequilibrium lever rule or Scheil equation which, in spite of its simple nature, gives a reasonable description of the segregation pattern during weld metal solidification.4'34
The Scheil equation applies to directional solidification of long bars and is based on the
following assumptions:
(i)
(ii)
(iii)
(iv)
(v)

Uniform liquid composition (i.e. complete mixing in the liquid state).


A flat solid/liquid interface.
Local equilibrium at the solid/liquid interface (ko is constant).
Negligible solid-state diffusion.
Equal solid and liquid densities.

Under such conditions it is fairly simple to derive an expression for the solute concentration
in the metal as a function of the fraction solidified. Referring to Fig. 3.57, a mass balance
gives:
(3-41)
Since dz z and Cs = ko C1, equation (3-41) can be rewritten as:

Fraction equiaxed grains (%)


(a)

Average grain size, Jim

Aluminium content, wt%

(b)
Aluminium content, wt%
Fig. 3.55. Effect of deoxidation practice (aluminium additions) on the columnar to equiaxed transition in
ferritic stainless steel GTA welds; (a) Average fraction of equiaxed grains observed at the surface of the
welds vs aluminium content, (b) Surface equiaxed grain size vs aluminium content. Data from Villafuerte
etal21
(3-42)
Integration of this equation gives:

(3-43)

Fraction equiaxecl grains (%)

(a)

Titanium content, wt%

Average grain size, JI m

(b)

Titanium content, wt%


Fig. 3.56. Effect of deoxidation practice (titanium additions) on the equiaxed transition in ferritic stainless steel GTA welds; (a) Average fraction of equiaxed grains observed at the surface of the weld vs
titanium content, (b) Surface equiaxed grain size vs titanium content. Data from Villafuerte et al.21
from which
(3-44)
and
(3-45)
where fs is the fraction solidified (equal to z/L*).

Liquid

Concentration, C

Solid

Distance, z
Fig. 3.57. Solute redistribution during non-equilibrium freezing according to the Scheil equation.
Equation (3-44) is valid up to CL = Ceut where the remaining melt solidifies in the form of
intercellular or interdendritic eutectics. The eutectic fraction feut is, in turn, given as:

(3-46)

Figure 3.58 shows how the Scheil equation can be used for an evaluation of the
microsegregation pattern in binary alloy systems by considering a small volume element of
length L* which solidifies perpendicular to the cell/dendrite growth direction.
Since the Scheil equation does not allow for solid state diffusion during solidification, a
slightly refined version of this equation also exists in the literature:1

(3-47)

where a* is a dimensionless diffusion parameter, defined as:

(3-48)

Solid
Liquid

Fig. 3.58.Idealised model for microsegregation in ingots and castings (schematic).


Equation (3-47) has been used by Brody and Flemings34 to evaluate the effect of solid
diffusion on the amount of eutectic in different cast structures. The extent of this diffusion
depends on the dimensionless product a* ko and it becomes significant only for values of
a* Jc0 greater than about O.I.1 Note that in cellular or cellular-dendritic growth the primary
dendrite arm spacing X1 provides a measure of the diffusion length. Taking L* = X ,/2, equation
(3-48) can be rewritten as:
(3-49)
In contrast, during equiaxed dendritic growth the secondary dendrite arm spacing X2 is a
more appropriate dimension for the solidification microstructure, since the back diffusion process
here occurs mainly between secondary arms and not between primary trunks.2 In such cases
we may write:
(3-50)

Example (3.13)

Consider GTA welding of an Al-2wt%Cu alloy under the following solidification conditions:
Welding speed
Cooling rate
Primary dendrite arm spacing

Estimate on the basis of the Scheil equation the degree of microsegregation occurring during weld metal solidification. Relevant physical data for the Al-Cu system are given below:

Solution

The local solidification time can be calculated from equation (3-33). Taking RL = v, we get:

Since the solidification conditions in this case facilitate the formation of a cellular-dendritic
type of substructure close to the weld centre-line (see Fig. 3.43), the characteristic diffusion
length L* is determined by the primary dendrite arm spacing X1. The product a* ko is then
given as:

Because the numerical value of a* ko is very small, the contribution from diffusion in the
solid state can be neglected. Hence, the extent of microsegregation occurring during solidification can be evaluated from equation (3-45). Taking C0 = 2wt% and ko = 0.17, we obtain:

The results are presented graphically in Fig. 3.59.


As expected, the copper concentration is seen to increase monotonically from the core to
the periphery (surface) of the dendrite stem. When the eutectic composition is reached, the
remaining fraction solidifies as Al(5.6wt%Cu) + CuAl2. According to equation (3-46) the
eutectic fraction is equal to:

These results should be compared with the experimental data of Brooks and Baskes35
replotted in Fig. 3.60. It is evident that the measured copper concentration profile for the Al2wt%Cu GTA weld is similar to that inferred from the Scheil equation, although the observed
dendrite core concentration lies significantly above the predicted one. Consequently, the Scheil
equation gives a reasonable description of the segregation pattern during weld metal solidification, in spite of the simplifying assumptions inherent in the model.

Cu concentration, wt%

z.^m

Normalized distance (z/L*)


Fig. 3.59.Predicted Cu concentration profile based on the Scheil equation (Example (3.13)).

Cu concentration, wt%

GTAW (Al-2wt% Cu)

Position, ]im
Fig. 3.60. Electron microprobe analysis of Cu across primary solidification (cell) boundaries in an Al2wt% Cu GTA weld. Data from Brooks and Baskes.35

3.7.2 Macro segregation


Macrosegregation in the form of solute banding is also frequently observed after weld metal
solidification.34 This type of segregation arises from a periodic enrichment and/or depletion
of solute elements, caused by a non-steady advancement of the solidification front (see Fig.

Fig. 3.6/. Example of solute banding (dark stripes) in a phosphorus-containing copper weld. After Garland and Davis.36
3.61). Although the origin of the phenomenon is not yet fully understood, it is reasonable to
assume that the pertinent fluctuations in the solidification rate occurs as a result of frequent
variations in the heat flux during welding. Direct experimental evidence for such a correlation
can be obtained from the data of Garland and Davis.36
Moreover, there is a pronounced tendency for alloying and impurity elements to segregate
to the weld centre-line where the columnar grains growing from each side impinge. This, in
turn, may produce hot tearing as a result of the formation of low-melting eutectics between the
dendrite arms. In general, the risk of hot tearing decreases with increasing width to depth ratio
of the weld because of a more favourable crystal growth mode, as illustrated in Fig. 3.62.
3.7.3 Gas porosity
Troublesome impurities in fusion welds are those which precipitate, alone or in combination
with other elements, to form various gaseous reaction products.37 These reaction products may
be simple diatomic gases such as H2 and N2 or more complex gaseous compounds like CO, or
H2O. An illustration of gas porosity due to hydrogen evolution is contained in Fig. 3.63.
Following the discussion in Chapter 2, absorption of gases occurs readily in the hot part of
the weld pool during welding because of interactions with the surrounding arc atmosphere.
During the subsequent stage of cooling a supersaturation rapidly increases due to the associated decrease in the element solubility with decreasing temperatures (see Figs. 2.7 and 2.8).
The system will respond to this supersaturation by rejection of dissolved elements from the
liquid, either through a gas/metal reaction (desorption) or by precipitation of new phases. The
latter incident may result in porosity if the gas bubbles become trapped in the weld metal
solidification front.
3.7.3.1 Nucleation of gas bubbles
In liquid metals a bubble will be stable if the gas pressure inside the bubble is sufficiently high
to balance the external forces. These external forces are the liquid/vapour interfacial energy

(b) Incorrect width to depth ratio

(a) Correct width to depth ratio

Width

Depth

Depth

Width

Fig. 3.62. Effect of weld width to depth ratio on the tendency to centre-line cracking; (a) Correct width to
depth ratio, (b) Incorrect width to depth ratio.

Fig. 3.63. Example of gas porosity in a GTA 7106 aluminium weld. After D'annessa.39

Transport of dissolved
gaseous species

Liquid
Gas bubble

Solid
Fig. 3.64. Growth of a gas bubble due to diffusion in the liquid phase (schematic).

a*, the metallostatic pressure head/?m, and the ambient pressure pa. Thus, for a stable bubble,
we may write:1
(3-51)
where pg is the total gas pressure inside the bubble, and rg is the radius of the gas bubble.
For shallow welds, the contribution from the metallostatic pressure head pm can be ignored. Hence, equation (3-51) reduces to:
(3-52)
Since a* is typically of the order of 1 J m 2 (9.87 atm |im) for most gas-metal systems, we
may write:
(3-53)
It is evident from equation (3-53) that the interfacial energy term is negligible at large
values of r . However, if the radius of curvature becomes sufficiently small, extremely large
pressures are required to maintain a stable bubble. Thus, there is a bubble nucleation problem, which is formally similar to that of nucleation of a solid from a liquid (discussed in
Section 3.3.1). In fact, it can be shown on the basis of classic nucleation theory that the driving
force normally associated with rejection of dissolved gases in liquid metals is by far too small
to allow for homogeneous nucleation of gas bubbles in the weld pool during cooling. This, in
turn, implies that solid particles (e.g. inclusions) entrained in the liquid metal will be the most
probable sites for gas bubble formation in fusion welds.
3.7.3.2 Growth and detachment of gas bubbles
As shown in Fig. 3.64, growth of gas bubbles in liquids is a diffusion-controlled process where
the time dependence of the mean bubble radius is given by:38
(3-54)
Here Q.* is the growth constant, defined as:

(3-55)
where CL is the molar concentration of solute in the supersaturated liquid, C6 is the equilibrium
molar concentration of the solute at the gas/liquid interface, and p^ is the gas density (in the
same units as CL and C6).
Equations (3-54) and (3-55) may be used to estimate the growth rate of a bubble while it is
still attached to the solid/liquid interface. The bubble becomes detached when the buoyancy
force, which is pushing it upwards, exceeds the surface tension force, which tends to keep it
attached to the solid surface. The bubble radius at which detachment occurs is given by the socalled Fritz equation:38

(3-56)

where gc is the gravity constant, and P is the wetting angle (in degrees).
Based on equations (3-54) and (3-56) it is possible to evaluate the conditions for growth
and detachment of gas bubbles during weld metal solidification. This is a subject of considerable importance in welding, since the pore formation will inevitably affect the mechanical
integrity of the weldment.
Example (3.14)

Consider GTA butt welding of a 3mm thin Al-Mg sheet under the following conditions:

Suppose that gas bubbles form at the solid/liquid interface during solidification due to rejection of dissolved hydrogen from a supersaturated liquid. Based on equations (3-54) and
(3-56) estimate the maximum theoretical radius of the gas bubbles and the critical radius at
which the bubbles detach themselves from the solid/liquid interface during welding. Relevant
physical data for the Al-Mg system are given below:

Solution

Since particles located at the solid/liquid interface are the most probable sites for hydrogen
gas evolution, the local solidification time t0 provides a conservative estimate of the growth
time t in equation (3-54). From equation (3-36), we have:

In order to calculate the growth constant from equation (3-55), it is necessary to convert the
concentration driving force to molar units:

The molar density of the gas p is obtained from the ideal gas law:

The value of the growth constant is thus:

The maximum theoretical radius of the gas bubbles can now be evaluated from equation
(3-54) by inserting the appropriate values for Q*, DH, and to:

Similarly, the critical radius at which the hydrogen bubbles become detached may be estimated from equation (3-56). Since PM > > PH 2 , we obtain:

By inserting this value into equation (3-54) it is also possible to estimate the average bubble
detachment frequency under the prevailing circumstances:

Since r (crit.) r (max), the maximum pore radius will probably be closer to 0.7mm than
30mm in a real welding situation. This is also in agreement with practical experience (see Fig.
3.65).

Cumulative probability (%)

GTAW (Aluminium)

Diameter of pores, mm
Fig. 3.65. Measured distribution of pore diameters in some GTA aluminium welds deposited with different hydrogen-containing shielding gases. Data from Tomii et al.40

3.7.3.3 Separation of gas bubbles


It follows that the gas bubbles will start to migrate towards the surface of the weld immediately
after they become detached from the solid/liquid interface. Small, spherical bubbles (characterised by a bubble Reynold number less than 2) will rise at a terminal velocity determined by
Stokes law:38
(3-57)
where dg is the diameter of the gas bubbles, and jLl is the viscosity of the liquid.
Larger gas bubbles (characterised by a bubble Reynold number between 2 and 400) will
also rise in a rectilinear manner, but their terminal velocity may be as much as 50% greater
than that predicted from Stokes law.38 Depending on their flotation rate, such ascending gas
bubbles will either escape to the weld surface or be trapped in the weld metal solidification
front in the form of macroscopic gas porosity (see Fig. 3.63).
Example (3.15)

Based on Stokes law (equation (3-57)) calculate the rising velocity of a 0.2mm large hydrogen
bubble ascending in liquid aluminium. Relevant physical data for liquid aluminium are given
below:

Solution

Since p z p , we may write:

It is evident from the above calculations that the flotation rate of such gas bubbles is quite
high and of the same order of magnitude as the weld pool fluid flow velocity (discussed in
Section 2.11.2). Hence, the buoyancy force would be expected to play a significant role in the
separation process of detached gas bubbles in the weld pool. On this basis it is not surprising to
find that a change in the welding position (e.g. from flat to overhead) results in a dramatic
increase in the volume of porosity during GTAW of aluminium alloys (see Fig. 3.66).
Although a great deal has been reported on the causes and effects of porosity in weld metals
(see Ref. 37 for an excellent discussion), little is known about the mechanism of pore formation relative to solidification mechanics, nucleation, growth and transport of gas bubbles in the
weld pool. Consequently, a more fundamental approach to the porosity problem in fusion
welding (along the lines indicated above) is necessary in order to obtain a verified, quantitative
understanding of the phenomenon.

GTAW

Volume of pores, ml/10Og

Overhead position

Flat position

Welding orientation, degrees


Fig. 3.66. Porosity in GTA welds deposited on 2mm sheets of aluminium at various orientations from flat
to overhead. Data compiled by Devletian and Woods.37

3.7.4 Removal of microsegregations during cooling


As shown in Section 3.7.1, the characteristic growth pattern of cellular and dendritic solidification, in combination with the rapid cooling rates normally associated with fusion welding,
lead to extensive segregation of alloying and impurity elements to the intercellular or
interdendritic spaces. Segregation produced by this means is remarkably persistent, and can in
certain cases only be eliminated by prolonged high-temperature heat treatment. A simplified
analysis of homogenisation of microsegregations in fusion welds is given below.
3.7.4.1 Diffusion model
It is a reasonable approximation to regard microsegregations in cast structures as periodic,
where the concentration at any point and time along an arbitrary line, C(x,t), is given by:41

where C and Cmax are as indicated in Fig. 3.67, lavg is the average distance between adjacent maxima and minima, and Ds is the diffusivity of the solute in the solid.
Equation (3-58) states that the concentration remains constant and equal to Cavg at positions
JC = 0, JC = lavg, x = 2/ etc., while the peak of the sine wave is attained at distances x = lavg /2,
x = 5lavg /2 etc. during the decay. If only the peak concentration is considered, the sine term
becomes equal to unity and equation (3-58) reduces to:
(3-59)
from which the homogenisation time during isothermal heat treatment thom can be obtained:

(3-60)

Since the diffusion length I , in practice, is equal to the half dendrite arm spacing (X1 /2 or
X2/2), equation (3-60) predicts that the homogenisation time is proportional to the square OfX1
or X2. The latter parameters are, in turn, determined by the thermal conditions existing within
the mushy zone during weld metal solidification, and are therefore sensitive to variations in
welding variables such as the net arc power qo and the travel speed v.
3.7.4.2 Application to continuous cooling
During cooling of the weld in the solid state, some equalisation of microsegregations will
occur through diffusion. The extent of this diffusion can be reported in terms of an equivalent
isothermal homogenisation time at a chosen reference temperature Tr. If Qs denotes the activation energy for diffusion of the solute in the solid, the equivalent isothermal homogenisation
time tr at Tr is given by:

(3-61)

Concentration

Dendrite arm spacing

x
Fig. 3.67. Relaxation of sinusoidal distribution of solute during isothermal annealing.

The integral on the right-hand side of equation (3-61) represents the kinetic strength of the
weld thermal cycle with respect to homogenisation (solid state diffusion), and can be determined by means of numerical methods when the cooling programme is known. The extent of
solute diffusion may then be evaluated from equation (3-60) by inserting representative
values for t and / :
T

UY'g.

(3-62)

Example (3.16)

[n low-alloy steel, both carbon, phosphorus, and manganese are known to segregate to the
interdendritic spaces during solidification. Consider SA welding on a thick plate of steel
under the following conditions:

Based on equations (3-61) and (3-62) calculate the extent of homogenisation occurring
within the solid weld metal during cooling in the austenite regime (i.e. from 1520 to 7000C).
Relevant data for the diffusivity of carbon, phosphorus, and manganese in austenite are given
below:

Solution

During thick plate welding, the cooling programme can be calculated from equation (1-45).
For points located on the weld centre-line behind the heat source y = z = 0, and R* = -x = vt.
Hence, equation (1-45) reduces to:

By inserting representative values for T0, qo, v and \ , we obtain:

From this it is seen that the actual transformation temperatures (i.e. 1520 and 7000C) are
reached after 20.9 and 46.2s, respectively. If 13500C is used as a reference temperature, the
kinetic strength of the cooling cycle with respect to solute diffusion can be expressed as:

Numerical integration of this equation over the weld cooling cycle gives:

At 13500C (1623 K) the diffusivities of C, P and Mn in austenite are:

The extent of homogenisation is thus:

It is evident from the above calculations that interdendritic segregations of carbon are
readily removed during cooling of the weld from the solidification temperature due to a high
diffusivity of C in austenite. In contrast, segregations of phosphorus and manganese are much
more persistent, since the diffusion cannot keep pace with the falling temperature. In the latter case the initial concentration gradients will largely be maintained down to temperatures
where the austenite to ferrite transformation occurs (see ion micrograph of phosphorus segregations in Fig. 3.39). This, in turn, can promote the formation of different types of transformation products (ranging from ferrite to martensite depending on the nature of the segregants) along the primary solidification boundaries due to local variations in the steel
hardenability. An example of martensite banding in an AISI 4340 SA steel weld is shown in
Fig. 3.68.
It should be noted that the conditions for equalisation of interdendritic segregations (e.g.
phosphorus) are not dramatically altered by a change in the heat input as long as the tr I l\vg
ratio in equation (3-62) remains fairly constant. For instance, if the net arc power qQ in the
above example is reduced by a factor of four (conforming to welding with covered electrodes),
the corresponding change in / and J1350 (P) will be:

and
from which

This value is reasonably close to that obtained during high-heat input SA welding
(0.036 s jirrr2), as shown in the above example.

Fig. 3.68. Banding of martensite (M) along primary solidification boundaries in an AISI 4340 SA steel
weld. After Burck.42

3.8 Peritectic Solidification


Referring to the schematic phase diagram in Fig. 3.69, crystallisation within peritectic systems
starts with primary precipitation of the 8p-phase from the liquid. At the peritectic temperature
T a new solid phase 7 forms, according to the reaction:
5p+ liquid -> yp

<3"63)

Because of the nature of this reaction, there is a strong tendency for the secondary phase 7
to grow along the 8^/liquid interface and, thus, to isolate the primary phase from contact with
the liquid.43 Depending on the growth mode, the decomposition of the primary 8^-phase is
said to occur either through a peritectic reaction or by a peritectic transformation,44 as
shown schematically in Fig. 3.70 and 3.71, respectively.
3.8.1 Primary precipitation of the y -phase
Primary precipitation of the 7^-phase from the liquid is also possible if the temperature is
gradually decreasing during solidification.43 As shown by Kerr et al.44^6 this type of precipitation is favoured by the presence of solid particles within the liquid metal which can act as
seed crystals for the 7^-phase in a selective manner. Particularly at high cooling rates, the formation of the primary 8^-phase can be completely suppressed, thereby allowing the secondary
7 -phase to nucleate and grow directly from the supercooled liquid.
3.8.2 Transformation behaviour of low-alloy steel weld metals
In general, low-alloy steels undergo multiple phase transformations during solidification and
subsequent cooling. Depending on the cooling rate, carbon or the substitutional alloy content,
the primary solidification product will either be delta ferrite 8 Fe , austenite 7 Fe or a mixture of
both. 47 ^ 9
3.8.2.1 Primary precipitation of delta ferrite
When a low-alloy steel is slowly cooled below the crystallisation temperature, delta ferrite
will be the first phase to form.50 In fusion welds, the delta ferrite grains will reveal an anisotropic
columnar morphology, with their major axes aligned in the direction of the steepest temperature gradient in the weld pool. On further cooling below the peritectic temperature, austenite
nucleates epitaxially at the primary delta ferrite grain boundaries, as shown schematically in
Fig. 3.72, since these sites provide the lowest energy barrier against heterogeneous nucleation
(requires a Kurdjumow-Sachs or a Nishiyama-Wasserman type of orientation relationship
between 8Fe and 7Fe).51 Subsequent growth of the austenite into the delta ferrite may then
proceed by a peritectic transformation at a rate which is controlled by diffusion of carbon in
the austenite. This reaction pattern is analogous to that documented for growth of 7 F e Widmanstatten sideplates in duplex stainless steel.5253
Since the austenite is bound by an orientation relationship with the delta ferrite,51 the
austenite cannot grow across the primary delta ferrite solidification boundaries. Hence, after
the peritectic transformation the columnar 7Fe grains will adopt the original 8Fe morphology,
as shown in Fig. 3.72. Direct experimental evidence for such a correspondence can be obtained from the optical micrographs in Fig. 3.38 and the texture data presented in Section 6.3.5
(Chapter 6).

Temperature

Liquid (L)

Fig. 3.69.Peritectic phase diagram (schematic).

Growth direction
Liquid

Diffusion of
solute atoms

Fig. 3.70.Peritectic reaction by which the secondary 7p-phase grows along the surface of the primary
8p-phase (schematic).

Liquid

Fig. 3.7L Peritectic transformation involving long-range diffusion of solute atoms through the secondary
7^-phase (schematic).

Delta ferrite
Liquid

Solute segregations

Austenite

Austenite grain
boundary
Fig. 3.72. Primary delta ferrite solidification with subsequent growth of austenite along the boundaries of
the primary 8Fe-phase (schematic).

3.8.2.2 Primary precipitation of austenite


Under certain welding conditions the weld metal can solidify directly as austenite without any
primary precipitation of delta ferrite (see Fig. 3.73). This solidification mode has been observed during welding with covered electrodes on high-carbon steels.54 Because of competitive growth occurring during the initial stage of the solidification process, the columnar austenite
grains will have one of their <100> axes aligned in the direction of the maximum temperature
gradient in the weld pool. Consequently, the weld metal will develop a solidification texture
which is similar to that observed during primary precipitation of delta ferrite.51
3.8.2.3 Primary precipitation of both delta ferrite and austenite
In certain low-alloy steel weld metals the austenite grain boundaries will systematically cross
the original delta ferrite solidification boundaries, as indicated in Fig. 3.74 because of a shift in
the mechanism of the peritectic transformation.3'22'2454"56 This point is more clearly illustrated in Fig. 3.75, which shows ion (SIMS) micrographs of phosphorus and boron segregations

Austenite
Columnar
grain
Liquid

Fig. 3.73. Primary austenite solidification (schematic).

Fig. 3.74. Optical micrograph showing austenite grain boundaries crossing primary delta ferrite solidification boundaries in a low-alloy SA steel weld (indicated by arrows). After Kluken et al.22
at prior solidification (cell) and austenite grain boundaries*, respectively in a SA steel weld
metal deposited under the shield of a basic flux. It is obvious from Fig. 3.75 that there is no
matching between the two types of boundaries in this particular case.
It follows from the analysis of Kluken et al?2 that the observed shift in the mechanism of
the peritectic transformation can probably be attributed to heterogeneous nucleation of
austenite at inclusions (e.g. Al2O3), which is energetically more favourable than nucleation at
8Fe/8Fe grain boundaries. Under such conditions, the austenite is not bound by an orientation
relationship with the delta ferrite and is thus free to grow across the original delta ferrite
columnar grain boundaries, as shown schematically in Fig. 3.76. The austenite grains will
therefore adopt a morphology which is different from the columnar one.245556

References
1.
2.
3.
4.
5.
6.
7.
8.
9.

M.C. Flemings: Solidification Processing, 1974, New York, McGraw-Hill Book Company.
W. Kurz and DJ. Fisher: Fundamentals of Solidification, 3rd Edn, 1989, Aedermannsdorf
(Switzerland), Trans. Tech. Publications.
GJ. Davis and J.G. Garland: Int. MetalL Rev., 1975, 20, 83-106.
S.A. David and J.M. Vitek: Int. Mater. Rev., 1989, 34, 213-245.
D. A. Porter and K.E. Easterling: Phase Transformations in Metals and Alloys, 1981, Wokingham
(England), Van Nostrand Reinhold Co. Ltd.
K.E. Easterling: Introduction to the Physical Metallurgy of Welding, 1983, London, Butterworths
& Co (Publisher) Ltd.
0. Grong, TA. Siewert and G.R. Edwards: Weld. J., 1986, 65, 279s-288s.
G.M. Oreper, TW. Eagar and J. Szekely: Weld. J., 1983, 62, 307s-312s.
Y.H. Wang and S. Kou: Proc. Int. Conf. on Advances in Welding Science, Gatlinburg, TN,
May, 1986, 65-69, Publ. ASM International.

*Because of a similar atomic weight of BO2~ and AlO~, both types of ions will be sampled in the SIMS analysis.
Consequently, the white spots observed within the interior of the grains are probably traces of aluminium oxide
inclusions.

(a)

(b)

Fig. 3.75. Ion (SIMS) micrographs showing austenite grain boundaries crossing primary delta ferrite
solidification boundaries in a low-alloy SA steel weld; (a) Phosphorus segregations at primary solidification (cell) boundaries, (b) Boron segregations at prior austenite grain boundaries (same area as in Fig.
3.75(a)). After Kluken et alP

Delta ferrite.

Solute segregations

Liquid

Inclusions (e.g. AI2O3)

Austenite

Austenite grain
boundary

Fig. 3.76. Primary delta ferrite solidification with subsequent nucleation of austenite at inclusions (schematic).
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.

K.C. Mills and BJ. Keene: Int. Mater. Rev., 1990, 35, 185-216.
T. Zacharia, S.A. David, J.M. Vitek and H.G. Kraus: Metall. Trans., 1991, 22B, 243-257.
CR. Heiple and J.R. Roper: Weld. J., 1982, 61, 97s-102s.
CR. Heiple and P. Burgardt: Weld. J., 1985, 64, 159s-162s.
T. Habrekke, H.O. Knagenhjem and J.O. Berge: Proc. 9th. Int. Conf. on Offshore Mech. and
Arctic Engin., Houston, TX, Feb., 1990, 511-515, Publ. ASME.
H. Biloni: Physical Metallurgy, 3rd Edn (Eds R.W. Chan and P. Haasen), 1983, Amsterdam,
North-Holland Physics Publ., 478-579.
T. Senda, F. Matsuda and M. Kato: Techn. Reports of the Osaka University (Japan), 1970,20,
527-558.
M. Rappaz, S.A. David, J.M. Vitek and L.A. Boatner: Metall. Trans. A, 1989,2OA, 1125-1138.
L.F. Mondolfo: Mat. ScL and TechnoL, 1989, 5, 118-122.
B.L. Bramfitt: Metall. Trans., 1970, 1, 1987-1995.
G.N. Heintze and R. McPherson: Weld. /., 1986, 65, 71s-81s.
J.C. Villafuerte, E. Pardo and H.W. Kerr: Metall. Trans. A, 1990, 21A, 2009-2019.
A.O. Kluken, 0. Grong and G. R0rvik: Metall. Trans. A, 1990, 21 A, 2047-2058.
H. Yunjia, R.H. Frost, D.L. Olson and G.R. Edwards: Weld. J., 1989, 68, 280s-289s.
CE. Cross, 0. Grong, S. Liu and J.F. Capes: Applied Metallography, (Ed. G.F. Vander Voort),
1986, New York, Van Nostrand Reinhold Inc., 197-210.
A.O. Kluken and 0 . Grong: Proc. Int. Conf. on Recent Trends in Welding ScL and TechnoL
(TWR 89), Gatlinburg, TN, May, 1989, 781-786, Publ. ASM International.
S. Kou: Welding Metallurgy, 1987, New York, John Wiley & Sons.
M.F. Jordan and M.C. Coleman: Brit. Weld. J., 1968,15, 553-558.
CE. Cross and G.R. Edwards: Treatise on Mat. ScL TechnoL, 1989, 31, 171-187.
W. Kurz, B. Giovanola and R. Trivedi: Acta Metall., 1986, 34, 823-830.

30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.

R. Trivedi and W. Kurz: Acta MetalL, 1986, 34, 1663-1670.


S. Kou and Y. Lee: MetalL Trans., 1982,13A, 1141-1152.
A. Munitz: MetalL Trans., 1985,16B, 149-161.
F. Matsuda, K. Nakata, K. Tsukamoto and K. Arai: Trans. JWRI, 1983,12, 81-87.
H.D. Brody and M.C. Flemings: Trans. AIME, 1966, 236, 615-623.
J.A. Brooks and M.I. Baskes: Proc. Int. Conf. on Trends in Welding Research (TWR 86),
Gatlinburg, TN, May, 1986, 93-99, Publ. ASM International.
J.G. Garland and GJ. Davis: Brit. Weld. J., 1970,17, 171-175.
J.H. Devletian and W.E. Woods: Weld. Res. Council Bull., 1983, 290, 1-18.
J. Szekely and NJ. Themelis: Rate Phenomena in Process Metallurgy, 1971, New York, John
Wiley & Sons, Inc.
A.T. D'annessa: Weld. J., 1967, 46, 491s-499s.
Y. Tomii, A. Sakaguchi and M. Mizuno: Proc. 4th Int. Conf. on Aluminium Weldments, Tokyo
(Japan), 1988, 46-60, Publ. Japan Light Metal Welding and Construction Association.
J.D. Verhoeven: Fundamentals of Physical Metallurgy, 1975, New York, John Wiley & Sons,
Inc.
P. Burck: M.Sc. thesis, 1984, Colorado School of Mines, Golden, Colorado (USA).
M. Hillert: Solidification and Casting of Metals, 1979, London, The Metals Society, 81-87.
H. W. Kerr, J. Cisse and G.F. Boiling: Acta MetalL, 1974, 22, 677-686.
J. Cisse, G.F. Boiling and H.W. Kerr: /. Cryst. Growth, 1972,13/14, 777-781.
J. Cisse, H.W. Kerr and G.F. Boiling: MetalL Trans., 1974, 5, 633-641.
H. Fredriksson: Met. ScL, 1976,10, 77-86.
H. Fredriksson and J. Stjerndahl: MetalL Trans., 1977, 8A, 1107-1115.
H. Fredriksson and J. Stjerndahl: Met. ScL, 1982,10, 575-585.
N.S. Pottore, CL Garcia and AJ. DeArdo: MetalL Trans., 1991, 22A, 1871-1880.
A.O. Kluken, 0. Grong and J. Hjelen: MetalL Trans., 1991, 22A, 657-663.
N. Suutala, T. Takalo and T. Moisio: MetalL Trans., 1979,10A, 1183-1190.
S.A. David: Weld. J., 1981, 60, 63s-71s.
A.A.B. Sugden and H.K.D.H. Bhadeshia: MetalL Trans., 1987,19A, 669-674.
J.G. Garland and RR. Kirkwood: Proc. Int. Symp. on Welding of Line Pipe Steels, St. Louis
(USA), 1977, 176-227, Publ. Welding Research Council (New York).
R.C. Cochrane: Weld, in the World, 1983, 21, 16-29.

Appendix 3.1
Nomenclature
thermal diffusivity
(mm2 s-1)

difference between total and


acid soluble Al in weld metal
(wt%)

lattice parameter (nm)


b
equilibrium 7/a solvus temperature (C)

(|Lim)

bcc
area of embryo-liquid interface (m2)
area of embryo-substrate interface (m2)

half width of dendrite stem


body-centred cubic structure
various kinetic constants and
temperature-dependent parameters

average solute concentration


in solid phase (wt%)

eutectic fraction
fraction solidified

initial alloy concentration


(wt%)

shape factor
face-centred cubic structure

equilibrium concentration of
solute at gas/liquid interphase
(wt% or mol cirr3)

length of dendrite stem (jam)

eutectic concentration (wt%)

gravity constant (m s"2)

concentration of alloying element in liquid phase (wt%


or mol cm"3)

temperature gradient in weld


pool (C mm"1 or K irr 1 )

maximum solute concentration in solid phase (wt%)


concentration of alloying element in solid phase (wt%)
solute concentration at position x and time t (wt%)
cooling rate (0C s~l or K s"1)
plate thickness (mm)
the interatomic spacing along
[MVW] n

GMAW

gas metal arc welding

GTAW

gas tungsten arc welding


activation energy for diffusion of atoms across the interface (J)
energy barrier to heterogeneous nucleation (J)
energy barrier to homogeneous nucleation (J)
free energy change associated with the embryo formation (J irr 3 )

the interatomic spacing along

[uvw\

heat content per unit volume


at the melting point (J mm"3)

diffusion coefficient in liquid


(mirr s [)

latent heat of melting (J irr 3 )

diffusion coefficient in solid


(mm2 s-1)

a low-index plane in the nucleated solid

diffusion coefficient of arbitrary element X


(mm2 s~!)

a low-index plane of the


substrate
amperage (A)

EBW

electron beam welding


frequency factor (s"1)

Boltzmann constant (1.381


X 10" 23 JK- 1 )

detachment frequency of gas


bubbles (s"1)

equilibrium partition coefficient

critical GJR1 ratio which facilitates equiaxed dendritic


growth(Csmm" 2 )

modified Bessel function of


second kind and zero order

length of columnar grain


(mm)

radius of spherical nucleus


(m)

average distance between adjacent maxima and minima


concentrations (|im)

critical radius of stable nucleus (m)

total length of circle segment


(mm)

universal gas constant (8.314


J K-1 mol-1)

total length of volume element (mm or Jim)

three-dimensional radius
vector (mm)

slope of the liquidus curve


(C per wt%)

dendrite growth rate in hkl


direction (mm s"1)

mobility term (mm3 s"1)

steady-state growth rate


of solid/liquid interface
(mm s"1)

exponent

local crystal growth rate (mm

dimensionless operating parameter

nominal crystal growth rate


(mm s'1)

rate of heterogeneous nucleation (nuclei per s and mm3)


density of nucleation sites per
unit volume of melt (nuclei
per mm3)
analytical weld metal oxygen
content (wt%)
ambient pressure (atm)
total gas pressure in bubbles
(atm)

SAW

submerged arc welding

SIMS

secondary
spectrometry

SMAW

shielded metal arc welding

ion

mass

entropy of fusion per unit


volume (J nr 3 Kr1)
time (s)
integration limits (s)

metallostatic pressure head


(atm)

homogenisation time (s)

net arc power (W)

local solidification time (s)

activation energy for diffusion of solute in solid (kJ


mol-1)

equivalent isothermal homogenisation time (s)

two-dimensional radius vector (mm)


dendrite tip radius (Jim)
radius of gas bubble (mm)

temperature (C or K)
ambient temperature (C or
K)
actual temperature in the liquid (C or K)

liquidus temperature according to phase diagram (C or


K)

primary peritectic phase

melting point (C or K)

interface normal angle (degrees)

peritectic temperature (C or
K)
reference temperature (C or
K)
dendrite root temperature (C
or K)
dendrite tip temperature (C
or K)

efficiency factor

dimensionless diffusion parameter


molar density of gas (mol
cm"3)
density of liquid (kg n r 3 or
g cm"3)
volume heat capacity
(J mm~3 0C"1)

undercooling (C or K)
temperature difference between tip and root of dendrite
stem (C or K)

cell/dendrite alignment angle


(degrees)
critical cell/dendrite alignment angle (degrees)

a low-index direction in
(UcQn

Gibbs-Thomson coefficient
(mm C or mK)

a low-index direction in

W),

dimensionless growth constant

voltage (V)
dimensionless temperature
rising velocity of gas bubbles
(mm s"1)

dimensionless temperature at
the melting point

volume of solid embryo (m3)


wetting angle (degrees)
welding speed (mm s"1)
Jt-axis/welding direction
(mm)

viscosity of liquid metal (kg


nr 1 s-1)
angle between [MVW]^ and

y-axis/transverse direction
(mm)

[vw]n (degrees)
austenite

z-axis/through thickness direction (mm)

secondary peritectic phase

dimensionless plate thickness

total grain rotation angle (degrees)

delta ferrite

solid/liquid interfacial energy


(J mr2)

lattice disregistry between a


nucleus n and a substrate s

liquid/vapour interfacial energy (J nr 2 )

embryo/liquid interfacial energy (J rrr2)

primary dendrite arm spacing


(Mm)

embryo/substrate interfacial
energy (J nr 2 )

primary dendrite arm spacing


close to fusion boundary
(Mm)

substrate/liquid interfacial
energy (J rrr2)
thermal conductivity (W
mm-10C"1)

secondary dendrite arm spacing (Mm)

4
Precipitate Stability in Welds

4.1 Introduction
Precipitate stability is an important aspect of welding metallurgy. Normally, modern structural
steels and aluminium alloys derive their balanced package of high strength, ductility and toughness via optimised thermomechanical processing to produce a fine-grained, precipitation
strengthened matrix. This delicate balance of microalloy precipitation and microstructure,
however, is significantly disturbed by the heat of welding processes, which, in turn, affects the
mechanical integrity of the weldment.
When a commercial alloy is subjected to welding or heat treatment several competitive
processes are operative which may contribute to a change in the volume fraction and size
distribution of the base metal precipitates. The two most important are:1
(i)
(ii)

Particle coarsening (Ostwald ripening)


Particle dissolution (reversion)

Referring to Fig. 4.1, particle coarsening occurs typically at temperatures well below the
equilibrium solvus Te of the precipitates, while particle dissolution is the dominating mechanism at higher temperatures. On the other hand, there exists no clear line of demarcation
between these two processes, which means that particle coarsening can take place simultaneously with reversion in certain regions of the weld where the peak temperature of the thermal
cycle falls within the 'gray zone' in Fig. 4.1. Nevertheless, it is important to regard them as
separate processes, since the reaction kinetics are so different (coarsening is driven by the
surface energy alone, whereas dissolution, which involves a change in the total volume fraction, is driven by the free energy change of transformation).

4.2 The Solubility Product


The symbols and units used throughout this chapter are defined in Appendix 4.1.
The solubility product is a basic thermodynamic quantity which determines the stability of
the particles under equilibrium conditions. Because of its simple nature, the solubility product
is widely used for an evaluation of the response of grain size-controlled and dispersion-hardened materials to welding and thermal processing.23
4.2.1 Thermodynamic background
In general, the solubility product can be derived from an analysis of the Gibbs free energy AG
of the following dissolution reaction:

'Grey zone1

Particle coarsening |

Increasing heating rate

Temperature

Particle dissolution

%B
Fig. 4.1. Schematic diagram showing the characteristic temperature ranges where specific physical
reactions occur during reheating of grain size-controlled and dispersion-hardened materials.

(4-1)
At equilibrium, we have:

(4-2)
where AH and AS are the standard enthalpy and entropy of reaction, respectively. The other
symbols have their usual meaning (see Appendix 4.1).
When pure An Bm is used as a standard state, the activity of the precipitate {aAn Bm) is equal
to unity. In addition, for dilute solutions it is a fair approximation to set aA~[%A] and aB~[%B],
where the matrix concentrations of elements A and B are either in wt% or at%*. Hence, the
solubility product can be written as:
(4-3)

*For the solute, the standard state is usually a hypothetical 1 % solution. This implies that the activity coefficient is
equal to unity as long as Henry's law is obeyed.

Table 4.1 gives a summary of equilibrium solubility products for a wide range of precipitates in low-alloy steels and aluminium alloys.
In addition to the compounds listed in Table 4.1, different types of mixed precipitates may
form within systems which contain more than two alloying elements.3"6 However, since the
presence of such multiphase particles largely increases the complexity of the analysis, only
pure binary intermetallics will be considered below.
4.2.2 Equilibrium dissolution temperature
Based on equation (4-3) it is possible to calculate the equilibrium dissolution temperature Td of
the precipitates. By rearranging this equation, we get:

(4-4)
where [%A]o and [%B]o refer to the analytical content of elements A and B in the base metal,
respectively.
Equation (4-4) shows that the equilibrium dissolution temperature increases with increasing concentrations of solute in the matrix. This is in agreement with the Le Chatelier's principle.
Table 4.1 Equilibrium solubility products for different types of precipitates in low-alloy steels and
aluminium alloys. Data compiled from miscellaneous sources.
Material/ phase

log [%A]n [%B]m

Type of
Precipitate

C* = AS/R'

D* = Mi0IR'

TiN

0.32

8000

TiC

5.33

10475

NbN

4.04

10230

Low-alloy steel

NbC

2.26

6770

(austenite)t

VN

3.02

7840

VC

6.72

9500

AIN

1.79

7184

Mo2C

5.0

7375

Al-Mg-Si^

Mg2Si

5.85

5010

Al-Cu-Mg$

CuMg

6.64

4005

MgZn

5.33

2985

Zn2Mg

7.72

4255

Al-Zn-Mgij:

All concentrations in wt%


All concentrations in at.%

Example (4.1)

Consider a low-alloy steel with the following chemical composition:

Calculate on the basis of the reported solubility products in Table 4.1 the equilibrium dissolution temperature of each of the following three nitride precipitates, i.e. NbN, AlN, and TiN.

Solution

The equilibrium dissolution temperature of the precipitates can be computed from equation (44) by inserting the correct values for C* and D* from Table 4.1:

It is evident from these calculations that precipitates of the NbN and the AlN type will
dissolve readily at temperatures above 1050 to 11000C, while TiN is thermodynamically stable up to about 14500C. In practice, however, a certain degree of superheating is always
required to overcome the inherent kinetic barrier against dissolution, particularly if the heating
rate is high. Consequently, in a real welding situation the actual dissolution temperature of the
precipitates may be considerably higher than that inferred from simple thermodynamic calculations based on the solubility product (to be discussed later).
4.2.3 Stable and metastable solvus boundaries
Due to the lack of adequate phase diagrams for the complex alloy systems involved, thermodynamic calculations based on the solubility product represent in many cases the only practical means of estimating the solid solubility of alloying elements in commercial low-alloy steels
and aluminium alloys.
4.2.3.1 Equilibrium precipitates
In the case of large, incoherent precipitates (where the Gibbs-Thomson effect can be neglected),
the concentration of element A in equilibrium with pure An Bm at different temperatures can be
inferred directly from equation (4-3). If we replace / ^ by R (i.e. switch from common to
natural logarithms), this equation yields:
(4-5)

Equation (4-5) describes the solvus surface within the solvent-rich corner of the phase diagram. However, when a pure binary compound dissolves the concentration of elements A and
B in solid solution is fixed by the stoichiometry of the reaction. The following relationship
exists between [%B] and [%A]:

(4-6)
or

where MA and MB are the atomic weight of elements A and B, respectively.


Figure 4.2 shows a graphical representation of equations (4-5) and (4-6), and the corresponding change in the matrix concentrations during dissolution of pure AnBm for a given set
of starting conditions.
Alternatively, we can express T as function of the product [%A]n [%B]m by utilising equations (4-3) and (4-6). The combination of these equations provides a mathematical description
of the 'solvus boundary' of an equilibrium precipitate in a multi-component alloy system. It is
evident from the graphical representation in Fig. 4.3 that the solid solubility will always in-

[%B]

[%A]0

[%B]o

Excess B
[%A]
Fig. 4.2. Concentration displacements during dissolution of binary intermetallics (equilibrium conditions).

Increased additions
_of element B

Nominal alloy composition

Reduced solid solubility


of element A

Temperature

I Increased dissolution temperature

Concentration of element A
Fig. 4.3. Factors affecting the solid solubility of a binary intermetallic compound in a multi-component
alloy system (schematic).

crease with increasing temperature when AH is positive. This type of behaviour is characteristic of intermetallics in metals and alloys, since the dissolution process in such systems is
endothermic.7 As a result, increased additions of a second alloying element B will also reduce
the solubility of the first alloying element A by shifting the 'solvus boundary' towards higher
temperatures when an intermetallic compound between A and B is formed.
With the aid of Fig. 4.3 it is easy to verify that the equilibrium volume fraction of the
precipitates/^ at a fixed temperature is given by:

(4-7)

where fmax is the maximum possible volume fraction precipitated at absolute zero.
Equation (4-7) provides a basis for estimating the equilibrium volume fraction of binary
intermetallics in complex alloy systems at different temperatures in cases where the concentration of element B is sufficiently high to tie-up all A in the form of precipitates. Similarly, if A
is present in an overstoichiometric amount with respect to B, we may write:

(4-8)

Example (4.2)

In Al-Mg-Si alloys the equilibrium Mg2Si phase may form during prolonged high temperature annealing. Consider a pure ternary alloy which contains 0.75 wt% (0.83 at.%) Mg and 1.0
wt% (0.96 at.%) Si. Estimate on the basis of the solubility product the equilibrium volume
fraction of Mg2Si at 4000C. Make also a sketch of the Mg2Si solvus in a vertical section
through the ternary Al-Mg-Si phase diagram. Relevant physical data for the Al-Mg-Si system are given below:

Solution

The maximum possible volume fraction of Mg2Si precipitated at absolute zero (fmax) can be
estimated from a simple mass balance by considering the stoichiometry of the reaction:

Moreover, the solubility product [at.% Mg]2 [at.% Si] at 4000C (673K) can be obtained
from equation (4-3) by utilising data from Table 4.1:

from which

If we also take into account the stoichiometry of the reaction, the solubility product can be
expressed solely in terms of the Mg-concentration. Substituting

into the above equation gives [at.% Mg] ~ 0.20. The equilibrium volume fraction OfMg2Si at
4000C is thus:

Similarly, the equilibrium Mg2Si solvus can be calculated from the solubility product by
substituting

into equation (4-3). By inserting data from Table 4.1 and rearranging this equation, we get:

It is seen from the graphical representation of the above equation in Fig. 4.4 that the Mg2Si
compound is thermodynamically stable up to about 3000C. At higher temperatures the phase
will start to dissolve until the process is completed at 5600C.
It is obvious from these calculations that the microstructure of overaged Al-Mg-Si alloys
should be very persistent to the heat of welding processes. In practice, only a narrow solutionised
zone forms adjacent to the fusion boundary. However, within this zone significant strength
recovery may occur after welding due to reprecipitation of hardening phases from the
supersaturated solid solution. Consequently, in such weldments the ultimate HAZ strength
level is usually higher than that of the base metal, as illustrated in Fig. 4.5.
4.2.3.2 Metastable precipitates
In practice, the solid solubility is also affected by the size of the particles. If, for instance, a

Nominal alloy composition

Temperature, 0C

Dissolution temperature: 5600C

at% Mg
at%Mg Si
Fig. 4.4. Solubility of Mg2Si in aluminium (Example (4.2)).

Strength level

After artificial ageing


After natural ageing
Immediately after welding

HAZ

Unaffected base metal


Distance from fusion line

Fig. 4.5. Response of overaged Al-Mg-Si alloys to welding and subsequent heat treatment (schematic).
spherical precipitate is acted on by an external pressure of say 1 atm, the same precipitate is
also subjected to an extra pressure AP due to the curvature of the particle/matrix interface, just
as a soap bubble exerts an extra pressure on its content (see Fig. 4.6(a)). The pressure AP is
given as:8
(4-9)
where 7 is the particle/matrix interfacial energy, and r is the radius of the precipitate.
Because of this extra pressure, the Gibbs energy of a small precipitate will be higher than
that of a large one, which, in turn, increases its solubility (see Fig. 4.6(b)). The important
influence of particle curvature on the solid solubility has been extensively investigated and
reported in the literature.18 Usually, the phenomenon is referred to as the capillary or the
Gibbs-Thompson effect.
In the following we shall assume that the thermodynamic and crystallographic properties
of the metastable precipitates are similar to those of the equilibrium phase and that the reduced thermal stability is only associated with capillary effects. For single phase precipitates
in binary alloy systems, it is fairly simple to show that the concentration of solute across
a curved interface, [%A]r, is interrelated to the equilibrium concentration of solute across a
planar interface, [%A], through the following equation:8

(4-10)
where Vn is the molar volume of the precipitate (in m3 mol"1), and Q is the contribution of the
interface curvature to the reaction enthalpy (equal to IyVJr).

(a)

Atmospheric
pressure

Matrix

(b)

Gibbs energy

Small precipitate

Matrix

Large precipitate

[%A]

[%A] r
Concentration

Fig. 4.6. Effect of interfacial energy on the solubility of small particles; (a) Schematic representation of
spherical particles embedded in a metal matrix, (b) Integral molar Gibbs energy of matrix and precipitates at a constant temperature.
Assuming that this relationship also holds in the case of binary intermetallics, a combination of equations (4-5) and (4-10) gives:

(4-11)

where
(4-12)

or

Alternatively, we can express T as a function of the product [%A]rn [%B]rm. This gives the
following expression for the solvus temperature of metastable precipitates T'eq:
(4-13)
It is evident from the graphical representation of equation (4-13) in Fig. 4.7 that the solid
solubility at a given temperature is significantly increased at small particle radii. Taking as an
example 7 = 0.5 J n r 2 , Vm = 10~5 m3 moH, R = 8.314 J Kr1 moH, T = 500 K, we obtain from
equation (4-10):

or

where r is the particle radius in nm.


From this it is seen that quite large solubility differences can arise for particles in the range
from r = 1 - 50nm.

Temperature

Large (equilibrium) precipitates

Small (metastable) precipitates

Concentration
Fig. 4.7. Graphical representation of equation (4-13) (schematic).

Example (4.3)

In Al-Mg-Si alloys metastable (hardening) p"(Mg2Si)-precipitates m a y form during artificial


ageing in the temperature range from 160-2000C. Consider a T6 heat treated ternary alloy
which contains 0.75 wt% (0.83 at.%) Mg and 1.0 wt% (0.96 at.%) Si. Based on equation (413) make a sketch of the metastable P "(Mg2Si) solvus in a vertical section through the ternary
Al-Mg-Si phase diagram. In these calculations we shall assume that the thermodynamic
properties of the metastable (3"(Mg2Si) phase are similar to those of the equilibrium (3 (Mg2Si)
phase, i.e. the reduced thermal stability is only related to the Gibbs-Thomson effect. Relevant
physical data for the Al-Mg-Si system are given below:

Solution

First we estimate the molar volume of the precipitate:

The contribution of the particle curvature to the reaction enthalpy is thus:

The metastable [3"(Mg2Si) solvus can now be calculated from the solubility product by
substituting

into equation (4-13). By inserting data from Table 4.1 and rearranging this equation, we get:

It is evident from the graphical representation of the above equation in Fig. 4.8 that the
particle curvature has a dramatic effect on the solid solubility. A comparison with Fig. 4.4
shows that the dissolution temperature drops from about 5600C in the case of the equilibrium
Mg2Si phase to approximately 225C for the metastable |3"(Mg2Si)-phase. On this basis it is
not surprising to find that artificially aged (T6 heat treated) Al-Mg-Si alloys suffer from severe softening in the HAZ after welding, as shown schematically in Fig. 4.9. Moreover, it is

Nominal alloy
composition A

Temperature, 0C

Dissolution temperature: 225 0C

Metastable solvus
boundary

at% Mg
at% Mg 2 Si

Strength level

Fig. 4.8. Solubility of (3"(Mg2Si) in aluminium (Example (4.3)).

After artificial ageing


After natural ageing
Immediately after welding

HAZ
Distance from fusion line
Fig. 4.9. Response of artificially aged Al-Mg-Si alloys to welding and subsequent heat treatment (schematic).

evident that the characteristic low dissolution temperature of the precipitates also gives rise to
the formation of a heat affected zone which is significantly wider than that observed during
welding of overaged Al-Mg-Si alloys.9 This shows that the response of age-hardenable aluminium alloys to welding and thermal processing depends strongly on the initial base metal
temper condition.
With the aid of equation (4-11) it is also possible to calculate an average (apparent) metastable
solvus boundary enthalpy for hardening |3"(Mg2Si)-precipitates in Al-Mg-Si alloys. A closer
evaluation of the exponent gives:

This value is in close agreement with the reported solvus boundary enthalpy for (3"(Mg2Si)precipitates in 6082-T6 aluminium alloys. 910

4.3 Particle Coarsening


When dispersed particles have some solubility in the matrix in which they are contained, there
is a tendency for the smaller particles to dissolve and for the material in them to precipitate on
larger particles. The driving force is provided by the consequent reduction in the total interfacial energy and ultimately, only a single large particle would exist within the system.
4.3.1 Coarsening kinetics
The classical theory for particle coarsening was developed independently by Lifshitz and
Slyovoz11 and by Wagner.12 The kinetics are generally controlled by volume diffusion through
the matrix. At steady state, the time dependence of the mean particle radius r is found to
be:11'12
(4-14)
where ro is the initial particle radius, 7 is the particle-matrix interfacial energy, Dm is the element diffusivity, Cm is the concentration of solute in the matrix, Vm is the molar volume of the
precipitate per mole of the diffusate, and t is the retention time.
Although the classic Lifshitz-Wagner theory suffers from a number of simplifying assumptions, experimental observations usually reveal a cubic growth law of the form given by equation (4-14).13
4.3.2 Application to continuous heating and cooling
Ion, Easterling and Ashby14 have shown how equation (4-14) can be applied to continuous
heating and cooling. In their analysis equation (4-14) was used in a more general form:
(4-15)

where c{ is a kinetic constant, and Qs is the activation energy for the coarsening process (for
binary intermetallics Qs may be taken equal to the activation energy for diffusion of the less
mobile constituent atom of the precipitates in the matrix).
4.3.2.1 Kinetic strength of thermal cycle
It follows that the extent of particle coarsening occurring during a weld thermal cycle can be
calculated by integration of equation (4-15) between the limits t = t{ and t = t2:

(4-16)
The integral on the right-hand side of equation (4-16) represents the kinetic strength of the
thermal cycle with respect to particle coarsening, and can be determined by means of numerical methods when the weld thermal (T-t) programme is known. The resulting radius of the
precipitates may then be evaluated from equation (4-16) by inserting representative values
for the constants ro and C1 (e.g. obtained from quantitative particle measurements).
4.3.2.2 Model limitations
A salient assumption in the classic Lifshitz-Wagner theory is that the particles coarsen at
almost constant volume fraction, i.e. no solute is lost to the surrounding matrix during the
coarsening process. Consequently, equation (4-16) should only be applied in cases where the
peak temperature of the thermal cycle is well below the equilibrium solvus of the precipitates.
Example (4.4)

Consider stringer bead deposition (GMAW) on a thick plate of a Ti-microalloyed steel under
the following conditions:

Assume that the base metal contains a fine dispersion of TiN precipitates in the as-received
condition. Calculate on the basis of equation (4-16) and the Rosenthal thick plate solution
(equation (1-45)) the extent of particle coarsening occurring within the fully transformed heat
affected zone during welding. Relevant physical data for titanium-microalloyed steels are
given below:

(activation energy for diffusion of Ti in austenite)


Solution

In the present example the problem is to calculate the size of the TiN precipitates in different

positions from the fusion boundary. This requires detailed information about the weld thermal
programme, as shown in Fig. 4.10(a). By substituting the appropriate values for qo, X, a and v
into the Rosenthal thick plate solution, the governing heat flow equation becomes:

where /?* refers to the three-dimensional radius vector in the moving coordinate system (designated R in equation (1-45)), while x is the welding direction (equal to vt at pseudo-steady
state).
Since titanium nitride is thermodynamically stable up to the melting point of the steel,
equation (4-16) can be used to calculate the extent of particle coarsening occurring within the
transformed parts of the HAZ. In the present example, we may write:

where the times tx and t2 are defined in Fig. 4.10(a).


The kinetic strength of the weld thermal cycle with respect to particle coarsening can now
be evaluated from these two equations by utilising the numerical integration procedure shown
in Fig. 4.10(b). The results from such computations are presented graphically in Fig. 4.11.
It is evident from this figure that significant coarsening of the precipitates occurs within the
HAZ during welding, particularly in regions close to the fusion boundary where the peak
temperature of the thermal cycle is high. A comparison with the experimental data of Ion et
al.l4 (reproduced in Fig. 4.12) shows that the theory gives a fairly good prediction of particle
size as a function of the peak temperature, provided that the kinetic constant C1 in equation (416) can be estimated with a reasonable degree of accuracy. In practice, however, the numerical
value of C1 will vary significantly with the chemical composition and thermal history of the
base metal (see Fig. 4.13). This means that empirical calibration of equation (4-16) to experimental data is always required to avoid systematic deviations between theory and experiments.

4.4 Particle Dissolution


During welding, the thermal pulse experienced by the heat affected zone adjacent to the fusion
boundary can result in complete dissolution of the base metal precipitates. Since this may give
rise to subsequent strength loss and grain growth, it is important to understand how variations
in welding parameters and operational conditions affect the dissolution rate. In the following,
the kinetics of particle dissolution will be discussed from a more fundamental point of view.
4.4.1 Analytical solutions
Over the years, several analytical models have been developed which describe the kinetics of
particle dissolution in metals and alloys at elevated temperatures.16 None of these solutions are
exact, since they represent different approximations to the diffusion field around the dissolv-

(a)
Weld metal

Temperature

HAZ

Y~ regime

Time

(i/r)exp(-Qs/RT)

(b)

Time

Fig. 4.10. Kinetic strength of weld thermal cycle with respect to particle coarsening (Example (4.4)); (a)
HAZ temperature-time programme (schematic), (b) Numerical integration procedure (schematic).

Weld metal

Partly transformed HAZ

Particle radius, nm

Fully transformed HAZ

Peak temperature, 0C
Fig. 4.11. Coarsening of TiN during steel welding (Example (4.4)).

Frequency, %

Rosenthal thick plate heat cycle:

Particle radius, nm

Fig. 4.12. Measured size distribution of TiN before (broken lines) and after (full lines) weld thermal
simulation. Operational conditions as in Example (4.4). Data from Ion et al.14
ing precipitates. Nevertheless, it will be shown below that at least some of them are sufficiently accurate to capture the essential physics of the problem and to give valuable quantitative information on the extent of particle dissolution occurring during the weld thermal cycle.

Particle radius, nm

Annealing temperature: 1350 0C

Steel A
Steel B
Steel C

Annealing time, s
Fig. 4.13. Effects of annealing time and steel chemical composition on the mean particle size of TiN.
Data from Matsuda and Okumura.15

4.4. L1 The invariant size approximation


The model described here is due to Whelan.17 Consider a spherical particle embedded in an
infinite matrix, as shown schematically in Fig. 4.14. The corresponding matrix concentration
profile is plotted in the lower part of the figure. In this case the concentration of the constituent
element A is higher close to the particle/matrix interface than in the bulk. Hence, there is a
tendency for the element to diffuse away from the particle and into the surrounding matrix (i.e.
the particle dissolves). Based on the assumption that the particle/matrix interface is stationary
(i.e. the diffusion field has no memory of the past position of the interface), Whelan17 arrived
at the following expression for the dissolution rate of a spherical precipitate at a constant
temperature:
(4-17)
where r is the particle radius, a is the dimensionless supersaturation (defined in Fig. 4.14), and
Dm is the element bulk diffusivity.
The term Hr on the right-hand side of equation (4-17) stems from the steady-state part of
the diffusion field, while the (1 A/7) term arises from the transient part. Because of the complex form of equation (4-17) it cannot be integrated analytically and hence, numerical
methods must be applied. However, if the transient part of equation (4-17) is neglected (conforming to the solution after long times), it is possible to obtain a simple expression for the
particle radius as a function of time:
(4-18)
where ro is the initial particle radius.
Equation (4-18) is identical with the so-called invariant-field solution developed independ-

Concentration

Distance
Fig. 4J4. Schematic representation of the concentration profile around a dissolving spherical particle in
an infinite matrix.
ently by Aaron et al.16 and is valid after a certain period of time, provided that there is no
impingement of diffusion fields from neighbouring precipitates. As shown in Fig. 4.15, this
simplified solution gives a reasonable description of the dissolution kinetics of small spherical
precipitates in steel during reheating above the AC1 -temperature.
Following the treatment of Agren,18 the time required for complete dissolution of a spherical precipitate td can be obtained from equation (4-18) by setting r = 0:

(4-19)
Moreover, the volume fraction of the precipitates/as a function of time is given by:

(4-20)
where fo is the initial particle volume fraction.
The former equation shows that the dissolution time td depends strongly on the initial particle size rQ.
Example (4.5)

The following example illustrates the direct application of equations (4-18) and (4-19). Consider a niobium-microalloyed steel which contains a fine dispersion of NbC precipitates. Provided that impingement of diffusion fields from neighbouring particles can be neglected, calculate the total time required for complete dissolution of a 100 nm large NbC precipitate at

Particle radius, jam

Numerical solution (Agren)

Simplified analytical
solution (Whelan)

Time, s
Fig. 4.15. Dissolution kinetics of spherical cementite particles in austenite at 8500C. Data from Agren.18
135OC. Data for the steel chemical composition and the diffusivity of Nb in austenite at
13500C are given below:
Steel chemical composition:

Diffusivity of Nb in austenite at 13500C:

Atomic weight of Nb:


Atomic weight of C :
Solution

In the present example it is reasonable to assume that the dissolution rate of the precipitate is
controlled by diffusion of Nb in austenite. For a single NbC precipitate embedded in a Nbdepleted matrix, the dimensionless supersaturation becomes:

The equilibrium concentration of niobium at the particle/matrix interface can be estimated

from the solubility product by utilising data from Table 4.1. If we assume that the carbon
concentration at the interface is constant and equal to the nominal value of 0.12 wt% (i.e. the
stoichiometry of the dissolution reaction is neglected), equation (4-5) reduces to:

Moreover, the concentration of Nb in the precipitate is equal to:

This gives:

The dissolution time td can now be calculated from equation (4-19) by inserting the appropriate values for ro, aNt)J and Dm\

A comparison with Fig. 4.16 shows that the predicted value is off by a factor of about 4
compared with that obtained from more sophisticated numerical calculations. This degree of
accuracy is acceptable and justifies the use of equation (4-18) for prediction of the dissolution
rate of spherical precipitates under different thermal conditions provided that the model is
calibrated against experimental data points.
4.4.1.2 Application to continuous heating and cooling
Application of the model to continuous heating and cooling requires numerical integration of
equation (4-18) over the weld thermal cycle:

(4-21)

Under such conditions the volume fraction of the precipitates is given by:

(4-22)

Equations (4-21) and (4-22) provide a basis for predicting the extent of particle dissolution
occurring within the HAZ during welding in the absence of impingement of diffusion fields
from neighbouring precipitates.

Dissolution time, s

Particle diameter, nm
Fig. 4.16. The dissolution time of NbC in austenite at 13500C as function of initial particle diameter lro
for different Nb and C levels (numerical solution). Data from Suzuki et al.6
Example (4.6)

Consider stringer bead deposition (SAW) on a thick plate of a Nb-microalloyed steel (0.10
wt% C - 0.03 wt% Nb) under the following conditions:

Assume that the base metal contains a fine dispersion of NbC precipitates in the as-received
condition. Calculate on the basis of equation (4-22) and the Rosenthal thick plate solution
(equation (1-45)) the extent of particle dissolution occurring within the fully transformed HAZ
during welding. Relevant physical data for Nb-microalloyed steels are given below:

Solution

In the present example the problem is to calculate the variation in the//fo ratio across the fully
transformed HAZ. By substituting the appropriate values for qo, X, a, and v into the Rosenthal
thick plate solution, the governing heat flow equation becomes:

Since it is reasonable to assume that the dissolution rate of the precipitates is controlled by
diffusion of Nb in austenite, the dimensionless supersaturation reduces to:

As shown in example (4.5), the equilibrium concentration of niobium at the particle/matrix


interface (in wt%) can be estimated from the solubility product by utilising data from Table
4.1. If we assume that the carbon concentration at the interface is constant and equal to the
nominal value of 0.10 wt%, equation (4-5) becomes:

Moreover, the concentration of Nb in the precipitate is equal to:

This gives:

By substituting the appropriate expressions for aNb and DNb into equation (4-22), we obtain:

Here the lower and upper integration limits refer to the total time spent in the thermal cycle
from Ac3 to T and down again to Ac3.
The extent of particle dissolution occurring within the HAZ during welding can now be
calculated in an iterative manner by numerical integration of the above equation over the weld
thermal cycle. The results from such computations are presented graphically in Fig. 4.17.
It is evident from these data that NbC starts to dissolve when the peak temperature of the
thermal cycle T exceeds the equilibrium dissolution temperature Td of the precipitate. The
process is completed when T approaches 13300C, conforming to a temperature interval of
1900C. This shows that considerable superheating is required in order to overcome the inherent kinetic barrier against particle dissolution under the prevailing circumstances.
4.4.2 Numerical solution
In the previous treatment, no consideration is given to impingement of diffusion fields from
neighbouring precipitates or the position of the particle/matrix interface during the dissolution
process. In certain cases, however, such phenomena will have a marked effect on the dissolution kinetics.18"22 A good example is Al-Mg-Si alloys where the hardening P"(Mg2Si)-phase
forms a very fine distribution of needle-shaped precipitates along <100> directions in the
aluminium matrix. These precipitates are closely spaced and will therefore interact strongly
with each other during dissolution (coupled reversion).

f/fo

No dissolution-

Complete dissolution
Peak temperature, 0C
Fig. 4.17. Dissolution of NbC during steel welding (Example (4.6)).

4.4.2.1 Two-dimensional diffusion model


For rod or needle-shaped precipitates in a finite, depleted matrix, the rate of dissolution can be
calculated by numerical methods from a simplified two-dimensional diffusion model. Assuming that the precipitates are mainly aligned in one crystallographic direction, it is reasonable to
approximate their distribution by that of a face-centered cubic (close-packed) space lattice, as
shown in Fig. 4.18(a). If planes are placed midway between the nearest-neighbour particles,
they enclose each particle in a separate cell. Since symmetry demands that the net flux of
solute through the cell boundaries is zero, the dissolution zone is approximately defined by an
inscribed cylinder whose volume is equivalent to that of the hexagonal cell. The modelling
principles outlined in Fig. 4.18(a) and (b) have previously been used by a number of other
investigators to describe particle dissolution during isothermal heat treatment.18"22 Consequently, readers who are unfamiliar with the concept should consult the original papers for
further details.
It follows from Fig. 4.18(b) that the rate of reversion can be reported as:
(4-23)

where ro is the initial cylinder (particle) radius.


(a)

Concentration

(b)

Distance
Fig. 4.18. Numerical model for dissolution of rod-shaped particles in a finite, depleted matrix; (a) Dissolution cell geometry, (b) Particle/matrix concentration profile (moving boundary).

For a specific alloy, the ratio between ro and L (the mean interparticle spacing) can be
calculated from a simple mass balance, assuming that all solute is tied-up in precipitates. Taking this ratio equal to 0.06 for rod-shaped precipitates in diluted alloys,9 the kinetics of particle
dissolution during isothermal heat treatment have been examined for a wide range of operational conditions. These results are presented in a general form in Fig. 4.19 by the use of the
following groups of dimensionless parameters:
Dimensionless time

(4-24)

Dimensionless supersaturation

(4-25)

(a is defined previously in Fig. 4.14).


The data in Fig. 4.19 suggest that the reaction kinetics during the initial stage of the process
are approximately described by the relation:
(4-26)

log <1 - f/fo)

where c2 is a kinetic constant, and n{ is a time exponent (assumed constant and equal to 0.5
under the prevailing circumstances).
The rate of particle dissolution will gradually decline with increasing values of T as a result
of impingement of diffusion fields from neighbouring precipitates which reduces a. In practice, this is seen as a continuous decrease in the slope of the flfo-x curves in Fig. 4.19 (nx < 0.5).
In such cases equation (4-26) will only be valid within small increments of X.

log T
Fig. 4.19. Dissolution kinetics of rod-shaped particles in a finite, depleted matrix. Data from Myhr and
Grong.9

4.4.2.2 Generic model


Myhr and Grong9 have shown how this model can be applied to specific alloy systems.
From equation (4-26) we have:

(4-27)
where n{ < 0.5.
This equation can further be simplified if we assume that

and

(4-28)
For isothermal heat treatment at a chosen reference temperature (Tn), the rate of particle
dissolution is determined by the retention time tr{. Let f* denote the maximum hold time
required for complete dissolution of the precipitates. It follows that equation (4-28) can be
written in a general form by normalising tn with respect to t*x. The parameter t*n is obtained
by setting flfo = 0:
(4-29)
where C3 is a new kinetic constant.
If heat treatment is carried out at a different temperature (T * Tn) , the maximum hold time
t* is simply given by:

(4-30)
By inserting the approximate expressions for C1 and Dm into equation (4-30) (see previous
examples), and rearranging equation (4-28), we obtain:

(4-31)

and
(4-32)
where Q'apP. is the apparent (metastable) solvus boundary enthalpy (defined in Section 4.2.3.2),
and Qs is the activation energy for diffusion of the less mobile constitutive atom of the precipitates.

Equations (4-31) and (4-32) exploit some good modelling techniques. For example, the use
of a dimensionless time eliminates an unknown kinetic constant which premultiplies t and ^1*
in the derivation of equation (4-32). Moreover, by raising the dimensionless time to a power
H1 means that the premultiplying constant, here unity, is independent of the value of nv and is
itself also dimensionless. Finally, the form of equation (4-31) eliminates further unknown
kinetic constants, and may readily be calibrated using an experimental time r* at a reference
temperature.
Figure 4.20 shows the variation inflfo with time (on log axes), from a range of isothermal
experiments carried out on 6082-T6 aluminium alloys, using hardness (or electrical conductivity) measurements to evaluate/7/\ The curve (equation (4-32)) extrapolates back to a slope
of 0.5 (the exponent n{) for the case of the early stages of dissolution before impingement of
the diffusion fields. The exponent nx is seen to fall to lower values when the proportion dissolved is higher, in agreement with the theoretical curves in Fig. 4.19.

log(i-f/f 0 )

4.4.2.3 Application to continuous heating and cooling


Myhr and Grong9 have also shown how this model can be applied to situations where the
temperature varies with time (as in welding).
In order to obtain a general description of particle dissolution under non-isothermal conditions, it is convenient to introduce the related concepts of an isokinetic reaction and the kinetic
strength of a thermal cycle.23 A reaction is said to be isokinetic if the increments of transformation in infinitesimal isothermal time steps are additive. Christian24 defines this mathematically by stating that a reaction is isokinetic if the evolution equation for some state variable X
may be written in the form:

log (t/t*)
Fig. 4.20. 'Master-curve' for dissolution of hardening p"(Mg2Si)-precipitates in 6082-T6 aluminium
alloys. Data from Myhr and Grong.9

(4-33)
where G(X) and H{T) are arbitrary functions of X and T, respectively.
For a given thermal history T(t), this essentially means that the differential equation contains separable variables of X and T. The same criterion may also be applied to the models
derived above. In the case of coupled reversion, we may write:
(4-34)
and
(4-35)
Since dfldlx and t\ are unique functions of/and T, respectively, the additivity condition is
satisfied. Consequently, when the temperature varies with time, we replace the term 11 t*x in
equation (4-34) by dt I t*x and integrate over the thermal cycle, giving:
(4-36)
This integral is called the kinetic strength of the thermal cycle with respect to reversion.
The resulting volume fraction of the precipitates following a heating cycle is then found by
evaluating the integral Z1 numerically (e.g. by utilising input data from Table 4.2) and replacing 111* with Z1 in equation (4-32), yielding a value for flfQ from the master curve of Fig. 4.20.
Case Study (4.1)
By utilising equation (4-36) and the general heat flow model for welding on medium thick
plates (i.e. equation (1-104)), it is possible to calculate the variation in the f/fo ratio (i.e. the
solute distribution) across the HAZ of single pass 6082-T6 aluminium weldments for a wide
range of operational conditions (see Table 4.3). The results from such computations are presented graphically in Fig. 4.21 -4.23.
When stringer bead deposition is carried out on a plate of medium thickness, the solute
distribution in the transverse y direction is expected to vary with distance from the plate surface due to a continuous change in the heat flow conditions. A closer inspection of Figs. 4.21
and 4.22 shows that this is correctly accounted for in the present model. In contrast, a full
penetration butt weld will always reveal a similar solute distribution in the transverse direction
of the weld, as shown in Fig. 4.23. This situation arises from the lack of a temperature gradient
in the through-thickness z direction of the plate.
Table 4.2 Basic input data in dissolution model for hardening p" (Mg2Si)- precipitates in 6082-T6
aluminium alloys. Data from Myhr and Grong.9
Parameter

Q'app

Qs

nx
(starting value)

t\
(375C)

Value

3OkJm0I-1

13OkJm 0 I- 1

0.5

600 s

Wo

Peak temperature, 0C

(a)

Comolete dissolution

Aym ,mm

f/fo

Peak temperature, 0C

(b)

Complete dissolution

ym,mm
Fig. 4.21. Dissolution of p"(Mg2Si)-precipitates during aluminium welding (Weld 1); (a) Upper plate
surface, (b) Lower plate surface. Operational conditions as in Table 4.3.
Table 4.3 Operational conditions used in aluminium welding experiments (Case study 4.1).
Weld

Material

No.*

Plate thickness

Net arc power

Welding speed

(mm)

(kW)

(mms"1)

AA6082

15

9.1

4.2

AA6082

15

5.7

9.1

3
AA6082
13
14.0
5.8
* Calibration of heat flow model is done by including an empirical correction for heat consumed in melting of the
parent material (thermal data for the AA6082 alloy are given in Table 1.1, Chapter 1).

Complete dissolution

Peak temperature, 0C

Mo

(a)

Aym,mm

Complete dissolution

Peak temperature, 0C

f/fo

(b)

ym,mm
Fig. 4.22. Dissolution of (3"(Mg2Si)-PrCCiPiIaIeS during aluminium welding (Weld 2); (a) Upper plate
surface, (b) Centre of plate. Operational conditions as in Table 4.3.

4.4.2.4 Process diagrams for single pass 6082-T6 butt welds


For single pass butt welding of plates, the heat flow model (equation 1-104)) can largely be
simplified if the net arc power is kept sufficiently high compared with the plate thickness (i.e.
q /d>0.5 kW mm"1). Under such conditions the mode of heat flow becomes essentially onedimensional, and the temperature distribution is determined by the ratio qo/vd, kJ ram"2 (see
Sections 1.10.3.3 and 1.10.4.1 in Chapter 1).

Peak temperature, 0C

Wo

Complete dissolution

Ay ,mm
m
Fig. 4.23. Dissolution of p"(Mg2Si)-precipitates during aluminium welding (Weld 3). Operational conditions as in Table 4.3.

f/fo

Figure 4.24 shows plots of the variation in the flfo ratio across the HAZ of 6082-T6 aluminium weldments for different values of qo Ivd. It follows that a narrow width of the dissolution zone requires the use of a low energy per mm2 of the weld. In practice, this can be
achieved by the use of an efficient welding process (e.g. electron beam or laser welding) which
facilitates deposition off a full penetration butt weld without employing a groove preparation
(i.e. eliminates the need for filler metals).

Scale:

Distance from fusion line


Fig. 4.24. Process diagram showing the solute distribution within the HAZ of single-pass 6082-T6 aluminium butt welds for different values of qo Ivd.

References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.

24.

R.D. Doherty: Physical Metallurgy, 3rd Edn (Eds R.W. Chan and R Haasen), 1983, Amsterdam, North-Holland Physics Publ., 934-1030.
K.E. Easterling: Introduction to the Physical Metallurgy of Welding, 1983, London, Butterworths
& Co., Ltd.
H. Adrian and RB. Pickering: Mater. ScL TechnoL, 1991, 7, 176-182.
B. Loberg, A. Nordgren, J. Strid and K.E. Easterling: MetalL Trans., 1984,15A, 33-41.
J. Strid and K.E. Easterling: Acta MetalL, 1985, 33, 2057-2074.
S. Suzuki, G.C. Weatherly and D.C. Houghton: Acta. MetalL, 1987, 35, 341-352.
J.L. Petty-Galis and R.D. Goolsby: J. Mater. ScL, 1989, 24, 1439-1446.
D. A. Porter and K.E. Easterling: Phase Transformations in Metals and Alloys, 1981, Wokingham
(England), Van Nostrand Reinhold Co. Ltd.
O.R. Myhr and 0. Grong: Acta MetalL Mater., 1991, 39, 2693-2702; ibid., 2703-2708.
H.R. Shercliff and M.F. Ashby: Acta MetalL Mater., 1990, 38, 1789-1802; ibid., 1803-1812.
J.M. Lifshitz and V.V. Slyozov: /. Phys. Chem. Solids, 1961,19, 35-50.
C. Wagner: Z. Electrochem., 1961, 65, 581-591.
L.C. Brown: Acta MetalL Mater., 1992, 40, 1293-1303.
J.C. Ion, K.E. Easterling and M.F. Ashby: Acta MetalL, 1984, 32, 1949-1962.
S. Matsuda and N. Okumura: Trans. ISIJ, 1978,18, 198-205.
H.B. Aaron, D. Fainstein and G.R. Kotler: J. Appl. Phys., 1970, 41, 4404-4410.
MJ. Whelan: Metal ScL J., 1969, 3, 95-97.
J. Agren: Scand. J. MetalL, 1990,19, 2-8.
R.A. Tanzilli and R.W. Heckel: Trans. Met. Soc. AIME, 1968, 242, 2313-2321.
H.B. Aaron and R. Kotler: MetalL Trans., 1971, 2, 393-408.
R. Asthana and S.K. Pabi: Mat. ScL Eng., 1990, A128, 253-258.
U.H. Tundal and N. Ryum: MetalL Trans., 1992, 23A, 433-444; ibid., 445-449.
H.R. Shercliff, 0. Grong, O.R. Myhr and M.F. Ashby: Proc. 3rd Int. Conf. on Aluminium
Alloys Their Physical and Mechanical Properties, Trondheim, Norway, June 1992, Vol. Ill,
pp. 357-369, The University of Trondheim, The Norwegian Inst. of Technol.
J.W. Christian: Phase Transformations in Metals and Alloys, 1975, Oxford, Pergamon Press.

Appendix 4.1
Nomenclature
thermal diffusivity (mm2 s l)

end temperature of ferrite to


austenite transformation (0C)

activity of element A in alloy


activity of precipitate An Bm
in alloy
activity of element B in alloy
arbitrary alloying element
start temperature of ferrite to
austenite transformation (0C)

equilibrium concentration of
element A in matrix (wt% or
at.%)
analytical content of element
A in alloy (wt% or at.%)
equilibrium concentration of
element A across a curved
particle/matrix interface
(wt% or at.%)

standard free energy of reaction (J mol"1 or kJ mol"1)

arbitrary alloying element


equilibrium concentration of
element B in matrix (wt% or
at.%)
analytical content of element
B in alloy (wt% or at.%)
equilibrium concentration of
element B across a curved
particle/matrix interface
(wt% or at.%)
various kinetic constants and
temperature-dependent parameters
normalised (dimensionless)
entropy of reaction
concentration of solute at
particle/matrix interface
(wt% or at.%)
nominal alloy composition
(wt% or at.%)
concentration of solute in
matrix (mol irr 3 , wt% or
at.%)
concentration of solute inside
the precipitates (wt% or
at.%)
plate thickness (mm)
normalised enthalpy of reaction (K)
element diffusivity (mm2 s"1
or m2 s"1)
particle volume fraction
initial particle volume fraction

GMAW

gas metal arc welding

H(T)

arbitrary function of T
standard enthalpy of reaction (J mol"1 or kJ mol"1)
amperage (A)
kinetic strength of thermal
cycle with respect to reversion
equilibrium constant
mean interparticle spacing in
dissolution model (m, Jim or
nm)
integer
atomic weight of element A
(g mol"1)
atomic weight of binary precipitate (g mol"1)
atomic weight of element B
(g mol-1)
integer
time exponent in dissolution
model
pressure caused by curvature
effects (J nr 3 )
net arc power (W)
apparent (metastable) solvus
boundary enthalpy (kJ mor 1 )
activation energy for diffusion (kJ moH)

equilibrium particle volume


fraction

particle radius (m, Jim or nm)

maximum particle volume


fraction at absolute zero

initial particle radius (m, (im


or nm)

arbitrary function of X

universal gas constant


(8.314JK-1InOl-1)

artificially aged condition


voltage (V)

universal gas constant multiplied by InIO (19.14J Kr1


mol"1)

welding speed (mm s"1)

three-dimensional radius
vector in Rosenthal equation
(mm)

state variable

standard entropy of reaction


(J K mol"1)
SAW

submerged arc welding

welding direction (mm)

transverse direction (mm)


width of HAZ referred to fusion boundary (mm)
through-thickness direction
(mm)

time (s)
time necessary for complete
dissolution of precipitate(s)
integration limits (s)
time necessary for complete
reversion at T(s)

molar volume of precipitate


(m3 mol 1 )
molar volume of precipitate
per mole of the diffusate (m3
mol-1)

retention time at T1. (s)

density of precipitate An Bm
(g cm"3 or kg rrr3)

time necessary for complete


reversion at Trl(s)

thermal conductivity (W
mm-1 0C-1)

temperature (0C or K)

dimensionless supersaturation in dissolution model

ambient temperature (0C or


K)
equilibrium dissolution temperature of precipitate (0C or
K)

density (g cm"3 or kg nr 3 )
arc efficiency factor
particle/matrix interfacial
energy (J m~2)

melting point (0C or K)


peak temperature of thermal
cycle (0C or K)
chosen reference temperature
in dissolution model (0C or
K)
equilibrium solvus temperature (0C or K)
solvus temperature of
metastable precipitates (0C or
K)

dimensionless time in dissolution model


contribution of particle curvature to reaction enthalpy (J
mol"1 or kJ mol"1)
equilibrium phase in Al-MgSi alloys
hardening precipitate in AlMg-Si alloys

5
Grain Growth in Welds

5.1 Introduction
Grain growth is an important aspect of welding metallurgy. Normal grain growth in metals
and alloys is a thermally activated process driven by the reduction in the grain boundary energy. Physically, it occurs by growth of the larger grains at the expense of the smaller ones
which tend to shrink.
Under isothermal heat treatment conditions, normal grain growth is well described by the
following empirical equation:1
(5-1)
where D is the mean grain size (diameter), D0 is the initial grain size, n is the time exponent,
/ is the isothermal annealing time, Qapp. is the apparent activation energy for grain growth, and
C1 is a kinetic constant. The other symbols have their usual meaning.
For most metals and alloys the time exponent n in equation (5-1) varies typically in the
range from 0.1 to 0.4, as shown in Fig. 5.1. Only in the case of ultrapure metals annealed at
very high temperatures the time exponent may approach a constant value of 0.5. This corresponds to the limiting case where the grain boundary migration rate is directly proportional to
the driving pressure 7 /D (7 denotes the grain boundary interfacial energy per unit area, while
D is the grain size).
It is well recognised that alloying and impurity elements both in the dissolved state and in
the form of inclusions or second phase particles will retard grain growth.2"* Consequently, a
comprehensive theoretical treatment of grain growth in welds must include a consideration of
such effects. The present analysis will therefore start with a closer examination of factors
affecting the grain boundary mobility in metals and alloys under conditions applicable to welding.

5.2 Factors Affecting the Grain Boundary Mobility


The symbols and units used throughout this chapter are defined in Appendix 5.1.
5.2.7 Characterisation of grain structures
A critical aspect of modelling grain growth is the quantitative description of grain structures,
which is essential in making a comparison between theoretical predictions and experimental

(U) JU8U0CJX9 91U|1

Z.R.AI

Brass

Homologous temperature (T/Tm)


Fig. 5.1. Temperature-dependence of the time exponent in isothermal grain growth (Z.R.: zone refined,
H.R: high purity). Data compiled by Hu and Rath.1
measurements. Different parameters are used to describe the size and the shape of individual
grains. In three dimensions, individual grain volumes cannot be determined directly from
measurements made in single cross sections through the structure. Therefore, certain geometric assumptions must be employed to obtain these quantities.
Since most grain size measurements seek to correlate the interaction of grain boundaries
with specific properties (e.g. the transformation behaviour), an estimate of the grain boundary
area per unit volume Sv is often required. This parameter can be calculated without assumptions concerning grain shape and size distribution from measurements of the mean linear grain
intercept D*.5
(5-2)
If the mean grain diameter D is required from 5V, this may be obtained by assuming a
spherical grain shape. Noting that each boundary is shared by two adjacent grains, we obtain:

(5-3)
from which
(5-4)
It follows that the mean linear grain intercept D* is always smaller than the actual grain size
D.
A common observation in metals and alloys is that the size distributions of grain aggregates

at different annealing times become equivalent when the measured grain size parameter, D, is
normalised (scaled) by the time-dependent average of this metric, D (see Fig. 5.2). This
means that grain structures can be completely characterised, in a statistical sense, by simple
probability functions of the standard deviation of the distribution together with the time dependence of the average size scale D.
5.2.2 Driving pressure for grain growth
The thermodynamic driving pressure for motion of a spherically curved element of grain boundary is given by:7
(5-5)
where y is the grain boundary interfacial energy, and p* is the radius of curvature.
It is conventional practice to replace p* in equation (5-5) with some measure of the average
grain size, such as the mean linear grain intercept D*, or with the diameter of the equivalent
spherical volume of some geometrically modelled average grain size. Experimental measurements performed on high purity aluminium indicate that p* ~ 3.23 D*.7 This observation is
consistent with the model of Hellman and Hillert,8 which predicts that the curvature of the
most critical element of grain boundary that must be stabilised is p* = 3 D . Under such conditions, the driving pressure becomes:
(5-6)

F (arbitrary units)

Zone refined iron

Rayleigh distribution

Log-normal
distribution

D/D
Fig. 5.2. Comparison of measured grain size data in iron with the Rayleigh and log-normal distribution
functions F. The similarity of the size distributions at different annealing times illustrates the self-similar
scaling behaviour of normal grain growth. After Pande.6

In practice, the numerical constant in equation (5-6) can vary by, at least, a factor of three,
depending on the assumptions of the models. Consequently, in the general case the average
driving pressure is given by:
(5-7)
where c2 is a constant which is characteristic of the system under consideration.
5.2.3 Drag from impurity elements in solid solution
In the dissolved state impurity and alloying elements will retard grain growth through elastic
attraction of the atoms towards the open structure of the grain boundary. The boundary must
then either drag the impurity atoms along (so that its speed is limited by the diffusion rate of
these atoms) or break away if the impurity concentration is sufficiently small or the driving
pressure or temperature is high enough.
An analysis of such effects can be done on the basis of the classic impurity drag theories of
Chan2 and Lucke and Stiiwe,3 which deal with the following two extreme cases:
(i) A low velocity limit, where the rate of grain boundary migration is controlled by diffusion of impurity atoms perpendicular to the boundary.
(ii) A high velocity limit, where the grain boundary migration process is mainly governed
by the diffusion of solvent atoms across the boundary (i.e. controlled by the rate of boundary
self diffusion).
The low velocity limit is associated with either a low driving pressure PG or a high impurity
level C0, and is characterised by a linear type of relationship between the grain boundary
migration rate Va and PG:2
(5-8)
Here e denotes the intrinsic drag coefficient, while 1F is a parameter depending on the
diffusivity and the interaction energy between the grain boundary and the impurity atoms.
For the other extreme (i.e. the high velocity limit), the grain boundary migration rate Vb is
described by the relationship:2
(5-9)
where 4Vp2 denotes another complex function of the impurity diffusivity and the interaction
energy between the grain boundary and the impurity atoms.
In the case of spherical grains, the classic grain growth equation predicts that the grain
boundary migration rate V is a power function of the driving pressure (y /D). This is readily
seen by differentiating equation (5-1) and inserting 7 (the grain boundary interfacial energy)
into the resulting expression:

(5-10)
where c3 is a new kinetic constant.
Note that at very high driving pressures or low impurity levels, equation (5-9) approaches
the limiting case where the migration rate Vb becomes directly proportional to PG, corresponding to a time exponent n = 0.5 in the grain growth equation. At this point the grain boundary
will break away from its surrounding impurity atmosphere and migrate at a rate close to the
rate of boundary self diffusion. In most cases, however, the relationship between V and PG
derived from equation (5-10) will be different from the theoretical one due to the empirical
nature of the grain growth equation.
For high driving pressures and intermediate impurity concentrations, the classic impurity
drag theories predict a discontinuous transition from the high to the low velocity limit. It has,
however, been argued by Vandermeer9 that the observed transition may be considered as continuous. Thus, in a log V vs log PG plot it would appear as a steep curve connecting the lines
for the high and low velocity extremes together, as illustrated schematically in Fig. 5.3.
5.2.4 Drag from a random particle distribution
The retardation of grain growth by second phase particles was first theoretically investigated
by Zener.10 There seems to be general agreement that the maximum pinnig force Fp exerted by
a single particle of radius r on a grain boundary is given by: 410
(5-11)

log V

High velocity limit

Transition region

Low velocity limit

log PG
Fig. 5.3. Schematic variation of grain boundary migration rate V with driving pressure PQ according to
the classic impurity drag theories of Cahn2 and Liicke and Stiiwe.3 The diagram is based on the ideas of
Vandermeer.9

If the number of interacting particles per unit area of the grain boundary is taken equal to na,
the resulting retardation pressure becomes:
(5-12)
Assuming that only one half of the particles which touches the grain boundary will interact
with a maximum force, na is related to Nv (the number of particles per unit volume) through the
following equation:410
(5-13)
Given that the particles are spherical and of uniform size, Nv can be expressed as:
(5-14)
where/is the particle volume fraction.
A combination of equations (5-12), (5-13), and (5-14) leads to the well-known expression
for the so-called Zener drag (or Zener retardation pressure):410
(5-15)
In practice, the numerical constant in equation (5-15) can vary by, at least, a factor of five,
depending on the assumptions of the models.11 Consequently, in the general case the Zener
retardation pressure is given by:
(5-16)
where C4 is a constant which is characteristic of the system under consideration.
5.2.5 Combined effect of impurities and particles
As already stated in the introduction of this chapter, the time exponent n is a measure of the
resistance to grain boundary motion in the presence of impurity and alloying elements in solid
solution. Based on equation (5-1) Hu and Rath 112 have shown that the grain boundary migration rate V is related to the effective driving pressure APG and the time exponent n through the
following equation:
(5-17)
where M is the grain boundary mobility.
In alloys containing grain boundary pinning precipitates, the effective driving pressure APG
is defined as the numerical difference between PG and Pz. By inserting the correct expressions
for PG and Pz into equation (5-17), we obtain:

(5-18)
It follows from equation (5-18) that the grain boundary migration rate V becomes inversely
proportional to the average grain size D when n = 0.5 a n d / = 0. This corresponds to the
limiting case where the grain boundary will break away from its surrounding impurity atmosphere and migrate at a rate which is controlled by the diffusion of solvent atoms across the
boundary. In most cases, however, the observed relationship between V and APG will be
different from the theoretical one due to drag from second phase particles (f> 0) or impurity
elements in solid solution (n < 0.5).

5.3 Analytical Modelling of Normal Grain Growth


By substituting V = V2 (dD/dt) and M = M0 exp (- QappIRT) into equation (5-18), it is
possible to obtain a simple differential equation which describes the variation in the average
grain size D with time t and temperature T in the presence of impurities and grain boundary
pinning precipitates:
(5-19)
Equation (5-19) can be written in a more general form by setting

and
(5-20)

From this it is seen that the parameters M0 and k are true physical constants which are
related to the grain boundary mobility and the pinning efficiency of the precipitates, respectively.
5.3.1 Limiting grain size
Equation (5-20) shows that the grain structure is stabilised when (d D ldi) - 0. The stable
(limiting) grain size is given by:
(5-21)
The parameter k (which in the following is referred to as the Zener coefficient) is defined as
the ratio between the numerical constants in equations (5-7) and (5-16), respectively. In the
original Zener's model k = 4/3, while other investigators have arrived at different results.811"14
As shown in Fig. 5.4, the limiting grain size may vary by over one order of magnitude, depending upon the assumptions of the models. This makes it difficult to apply equations (5-20) and
(5-21) for quantitative grain size analyses without further background information on the Zener
coefficient.

Diim.nm

Gladman

r/f, um
Fig. 5.4. Relation between limiting grain size Dum., particle radius r, and volume fraction/predicted by
different models.
Example (5. J)

Consider multipass GMA welding on a thick steel plate under the following conditions:

Based on the models of Zener,10 Hellman and Hillert,8 and Gladman13 estimate the limiting
austenite grain size Dlim in the transformed parts of the weld HAZ when the oxygen and
sulphur contents of the as-deposited weld metal are 0.04 and 0.01 wt%, respectively.
Solution

As shown in Chapter 2 of this textbook the volume fraction of oxide and sulphide inclusions
can be calculated from equation (2-75):

Similarly, the average radius of the grain boundary pinning inclusions can be obtained from
equation (2-79):

This gives the following values for the limiting austenite grain size:

Zener:

Hellman and Hillert:

Gladman:
and
As expected, the limiting austenite grain size is seen to vary by more than one order of
magnitude, depending on the assumptions of the models. In practice, the Zener coefficient in
low-alloy steel weld metals falls within the range from 0.32 to 0.93, as shown in Fig. 5.5. The
average value of A: is close to 0.52, which is the same as that inferred from the Gladman model
(upper limit). When it comes to intermetallic compounds such as titanium nitride, the Zener
coefficient varies typically between 0.75 and 0.25 during grain growth in the austenite regime. 1617 This suggests that k ~ 0.50 is a reasonable estimate of the grain boundary pinning
efficiency of oxides and nitrides in steel.
5.3.2 Grain boundary mobility
Direct application of equation (5-20) requires also reliable information on the time exponent n
and the grain boundary mobility M. When n = 0.5 and/= 0, the classic impurity drag theories
predict that the activation energy Qapp, should be close to the value for boundary self diffusion
in the matrix material.2'3 This borderline case is approximately attained in steel welding, as
shown in Fig. 5.6(a) and (b), since the driving pressure for austenite grain growth immediately
following the dissolution of the pinning precipitates is usually so large that the grain boundary
migration rate approaches the higher velocity limit defined in equation (5-9).18 On this basis it
is not surprising to find that Qapp falls within the range reported for lattice self diffusion (284
kJ mol"1) and boundary self diffusion (170 kJ mol~l) in pure 7-iron19 during welding.18 In
most cases, however, the activation energy will be different from the theoretical one due to
complex interactions between impurity atoms and grain boundaries (characterised by a time
exponent n < 0.5). Under such conditions, the value of Qapp has no physical meaning.1
5.3.3 Grain growth mechanisms
Equation (5-20) provides a basis for evaluating the grain growth inhibiting effect of impurity
elements and second phase particles under different thermal conditions. This also includes
situations where the grain boundary pinning precipitates either coarsen or dissolve during the
heat treatment process.
5.3.3.1 Generic grain growth model
Equation (5-20) can readily be integrated to give the average grain size D as a function of
time. In the general case we may write:
(5-22)

(a)
Austenite grain size, Jim

SA steel weld metal

Annealing temperature, 0 C

(b)

D||m. Hm

GMA and SA steel


weld metals

r/f, urn

Fig. 5.5. Evaluation of the Zener coefficient in steel weld metals containing stable oxide and sulphide
inclusions; (a) Determination of Dum. from isothermal grain growth data (holding time: 30 min),
(b) Variation in Dum. with the inclusion rlf ratio. Data from Skaland and Grong.15
where D{im is the limiting grain size (defined in equation (5-21)).
The integral I1 on the right-hand side of equation (5-22) represents the kinetic strength of the
thermal cycle with respect to grain growth and can be determined by numerical methods when
the temperature-time programme is known. In practice, however, it is not necessary to solve this
integral to evaluate the grain growth mechanisms. Consequently, the left-hand side of equation

(a)
Steel A
Log (DxZD1)

Slope: n = 0.4

Log [number of cycles]

(b)

LogD7

Steel A
Steel B

Fig. 5.6. Evaluation of the time exponent n and the activation energy Q for austenite grain growth in
steel under thermal conditions applicable to welding; (a) Time exponent n, (b) Activation energy Qapp.
Data from Akselsen et ah18

(5-22) can be solved explicitly for different values of DUm, n, and Z1. The results may then be
presented in the form of novel diagrams which show the competition between the various
processes that lead to grain growth during heat treatment of metals and alloys.
A more thorough documentation of the predictive power of the model and its applicability
to welding is given in Section 5.4.
5.3.3.2 Grain growth in the absence ofpinning precipitates
In the absence of grain boundary pinning precipitates, we have:/= 0, Dlim > ~ , and (1/ DlitrL) = 0.
Under such conditions, equation (5-22) reduces to:

(5-23)

After integration this equation yields:


(5-24)
Referring to Fig. 5.7, the average grain size D becomes a simple cube root function of Z1
when n = 0.5 and D 0 = 0. In other situations (n < 0.5), the grains will coarsen at a slower rate
due to drag from alloying and impurity elements in solid solution. This is seen as a general
reduction in the slope of the D-Ix curves in Fig. 5.7.
The important austenite grain growth inhibiting effect of phosphorus and free nitrogen in
steel following particle dissolution is shown in Fig. 5.8.
5.3.3.3 Grain growth in the presence of stable precipitates

If grain growth occurs in the presence of stable precipitates, the limiting grain size {Dlim) in
equation (5-22) becomes constant and independent of the thermal cycle. In the specific case
when n = 0.5 the integral on the left-hand side of equation (5-22) has the following analytical
solution:

D, ^m

(5-25)

I1W"
Fig. 5.7. Predicted variation in average grain size D with /, and n f o r / = 0 and D0 = 0 ('free' grain
growth).

(a)

D Y ,fim

Steel A

Number of cycles
(b)

D y ,um

Steel B

Number of cycles Fig. 5.8. Illustration of the austenite grain growth inhibiting effect of phosphorus and free nitrogen in
low-alloy steel during reheating above the Ac^ temperature (multi-cycle weld thermal simulation);
(a) Steel A (50ppm P, 20ppm N), (b) Steel B (180ppm P, 80ppm N). Data from Akselsen et a/.18

from which the average grain size D is readily obtained. In other cases, numerical methods
must be employed to evaluate D.
It is evident from the graphical representation of equation (5-25) in Fig. 5.9 that the grain
growth inhibiting effect of the precipitates is very small during the initial stage of the process
when D D lim. Under such conditions the grains will coarsen at a rate which is comparable
with that observed for free grain growth (n = 0.5,/= 0). The grain coarsening process becomes
gradually retarded as the average grain size increases because of the associated reduction in
the effective driving pressure APG until it comes to a complete stop when AP0 = 0 (i.e. D =
D Hm)-

D, jim

I 1 ^m 2

D, (im

Fig. 5.9. Predicted variation in average grain size D with Z1 and Dnm. for n = 0.5 and D0 = 0 (stable
precipitates).

I 1 , [nm]1/n

Fig. 5.10. Predicted variation in average grain size D with Z1 and n for Dum. = 250|Jin and D0 = 0
(stable precipitates). Dotted curves correspond to grain growth in the absence of pinning precipitates.

If grain growth at the same time occurs under the action of a constant drag from impurity
elements in solid solution, the situation becomes more complex. As shown in Fig. 5.10, a
decrease in the time exponent from say 0.5 to 0.2 gives rise to a marked reduction in the slope
of the D-I 1 curves, similar to that observed in Fig. 5.7 for particle-free systems (/= 0). However, the predicted grain coarsening rate is lower than that evaluated from equation (5-24) due
to the extra drag exerted by the grain boundary pinning precipitates. This leads ultimately to a
stabilisation of the microstructure when D = DUm.
5.3.3.4 Grain growth in the presence of growing precipitates
Very little information is available in the literature on the matrix grain growth behaviour of
metals and alloys in the presence of growing second phase particles. So far, virtually all
modelling work has been carried out on two phase a-(3 titanium alloys.14 Unfortunately, none
of these models can be extended to more complex alloy systems such as steels or aluminium
alloys.
When grain growth occurs in the presence of growing second phase particles, Dum. will no
longer be constant due to the associated increase in the particle rlf ratio with time. As shown in
Chapter 4 of this textbook, the Lifshitz-Wagner theory2021 provides a basis for modelling
particle growth during welding and heat treatment of metals and alloys in cases where the peak
temperature of the thermal cycle is kept well below the equilibrium solvus of the precipitates.
Under such conditions, the particles will coarsen at almost constant volume fraction (f=fo), in
accordance with equation (4-16):
(5-26)
where Qs is the activation energy for the coarsening process, C5 is a kinetic constant, and I2 is
the kinetic strength of the thermal cycle with respect to particle coarsening. The other symbols
have their usual meaning.
If the base metal contains particles of an initial radius ro and volume fraction/^, the limiting
grain size at I2 = 0 (D lim) can be defined as:

(5-27)
from which
(5-28)
Similarly, when I2 > 0, we may write:
(5-29)
By combining equations (5-26), (5-28), and (5-29), we arrive at the following relationship
between (D Um ) and I2:
(5-30)

It is seen from equation (5-30) that the limiting grain size in the presence of growing
particles depends on the product (k/fo)3I1. In practice, the grain boundary pinning effect of the
precipitates is determined by the relative rates of particle coarsening and grain growth in the
material, i.e. whether the grain boundary mobility is sufficiently high to keep pace with the
increase in DUm during heat treatment. Generally, the pinning conditions are defined by the
(k/fo)3 I1IIx ratio, which after substitution and rearranging yields:

(5-31)

In cases where the parameters c5 ,Qs, M0*, and Qapp are known, the average grain size D
can readily be evaluated from equations (5-22), (5-30), and (5-31) by utilising an appropriate
integration procedure. However, since Qs normally differs from Qapp^ the (klfo)3I1I Ix ratio
will depend on the thermal path during continuous heating and cooling. Consequently, solution of these coupled equations generally requires stepwise integration in temperature-time
space via a fourth heat flow equation. This problem will be dealt with in Section 5.4.
The situation becomes much simpler if heat treatment is carried out isothermally. Under
such conditions the product (k/fo)311 will only differ from Ix by a proportionality constant m,
which is characteristic of the system under consideration. Accordingly, equation (5-30) can be
rewritten as:
(5-32)
From this we see that the two coupled equations (5-22) and (5-32) can be solved explicitly
for different values of Dnm., n, m, and Z1. Hence, it is possible to present the results in the form
of novel 'mechanism maps' which show the competition between particle coarsening and
grain growth during isothermal heat treatment for a wide range of operational conditions.
Examples of such diagrams are given in Figs. 5.11 and 5.12.
It is evident from these figures that the grain coarsening behaviour during isothermal heat
treatment is very sensitive to variations in the proportionality constant m. For large values of
m, the matrix grains will coarsen at a rate which is comparable with that observed in Fig. 5.7
for particle-free systems (f = 0). This corresponds to a situation where the grain boundary
pinning precipitates will completely outgrow the matrix grains. It is interesting to note that
particle outgrowing is more likely to occur if the time exponent n is small, as shown in Fig.
5.12, because of the associated reduction in the grain boundary mobility in the presence of
impurity elements in solid solution. In other systems, where the proportionality constant m is
closer to unity, the reduced coarsening rate of the precipitates gives rise to a higher Zener
retardation pressure and ultimately to a stagnation in the matrix grain growth. In the limiting
case, when m = 0, the grain growth behaviour becomes idential to that observed in Figs. 5.9
and 5.10 for stable precipitates.

D.jim

Time exponent n = 0.5

I 1 ,^m 2
Fig. 5.11. Predicted variation in average grain size D with Ix and m for Dum. = 50jLim, n = 0.5, and
D0 =0 (growing precipitates).

D,fxm

Time exponent n = 0.3

I1^m1'"
Fig. 5.12. Predicted variation in average grain size D with I1 and m for Dum. - 50|im, n = 0.3, and
D0 = 0 (growing precipitates).

Example (5.2)

Consider a titanium-microalloyed steel with the following chemical composition:


Ti(total): 0.016 wt%, Ti(soluble): 0.009 wt%, N: 0.006 wt%
Assume that the base metal contains an uniform dispersion of TiN precipitates in the asreceived condition, conforming to a limiting austenite grain size T>um. of 50 |iim. Provided that
boundary drag from impurity elements in solid solution can be neglected (i.e. n ~ 0.5), estimate on the basis of Fig. 5.11 the average austenite grain size D1 in the material after 25 s of
isothermal annealing at 13000C. Relevant physical data for titanium-microalloyed steels are
given below:
(activation energy for diffusion of Ti in austenite)

Solution

The initial volume fraction of TiN in the material can be estimated from simple stoichiometric
calculations by considering the difference between total and soluble titanium. Taking the
atomic weight of Ti and N equal to 47.9 and 14.0 g mol"1, respectively, we obtain:

From this we see that the initial radius of the TiN precipitates in the base metal is close to:

Since heat treatment is carried out under isothermal conditions, the parameters m and Z1 can
be obtained directly from equations (5-31) and (5-22) without performing a numerical integration:

Similarly, in the case of Z1 we get:

The average austenite grain size can now be read from Fig. 5.11 by linear interpolation
between the curves for m = 10 and 100 Jim. This gives:

Although experimental data are not available for a direct comparison, the predicted grain
size is of the expected order of magnitude. From this it is obvious that considerable austenite
grain growth may occur in titanium-microalloyed steels because of particle coarsening, in
spite of the fact that TiN, from a thermodynamic standpoint, is stable up to the melting point of
the steel. The process can, to some extent, be counteracted by the use of a finer dispersion of
TiN precipitates in the material. For example, if the initial particle radius is reduced by a factor
of five (conforming to a change in I W from 50 to 10 Jim), the austenite grain size of the
annealed material decreases from 75 to 65 jLim, as shown in Fig. 5.13. Nevertheless, since
particle coarsening is a physical phenomenon occurring during high temperature heat treatment of metals and alloys, austenite grain growth cannot be avoided. This explains why, for
instance, conventional titanium-microalloyed steels are not suitable for high heat input welding due to their tendency to form brittle zones of Widmanstatten ferrite and upper bainite in the
coarse grained HAZ region adjacent to the fusion boundary.22

D,jim

Time exponent n = 0.5

Stable particles
I1-Hm2
Fig. 5.13. Predicted variation in average grain size D with Z1 and m for DHm. = 10 um, n = 0.5, and
Do = 0 (growing precipitates).

5.3.3.5 Grain growth in the presence of dissolving precipitates


Little information is available in the literature on the matrix grain growth behaviour of metals
and alloys in the presence of dissolving precipitates. As shown in Chapter 4, the model of
Whelan23 provides a basis for calculating the dissolution rate of single precipitates embedded
in an infinite matrix. If the transient part of the diffusion field is neglected, the variation in the
particle radius r with time t at a constant temperature is given by equation (4-18):
(5-33)
where a is the dimensionless supersaturation (defined in Fig. 4.14), and Dm is the element
diffusivity.
Application of the model to continuous heating and cooling requires numerical integration
of equation (5-33) over the weld thermal cycle:
(5-34)

where I3 is the kinetic strength of the thermal cycle with respect to particle dissolution.
From this relation the following expression for the particle volume fraction can be derived
(see equation (4-22), Chapter 4):

(5-35)
where fo is the initial particle volume fraction.
By substituting Dlim - k(rlf) and Dnm. = Kr0If0) into equations (5-34) and (5-35), it is
possible to obtain a simple mathematical relation which describes the variation in the limiting
grain size with I3 during particle dissolution. After some manipulation, we obtain:

(5-36)

It is seen from equation (5-36) that the limiting grain size increases from Dum. at I3 = 0 to
infinite when I3 = (fo Ik)2 (Dum. )2 Since the magnitude of the Zener drag, in practice, depends
on the relative rates of grain growth and particle dissolution in the material, the pinning conditions are defined by the (klfo)2131 Ix ratio:

(5-37)

Equation (5-37) shows that the (k/fo)2I3111 ratio is contingent upon the thermal path during
continuous heating and cooling. Consequently, application of the model to welding generally
requires numerical integration of the coupled equations (5-22), (5-36), and (5-37) over the
weld thermal cycle.
However, the integration procedure is largely simplified if heat treatment is carried out
isothermally. In such cases the product (k/fo)213 will only differ from Ix by a proportionality
constant m*, which is characteristic of the system under consideration. By substituting m*Ix
into equation (5-36), we obtain:

(5-38)
From this we see thaUhe two coupled equations (5-22) and (5-38) can be solved explicitly
for different values of Dun., n,m*, and I1. Hence, it is possible to present the results in the
form of novel 'mechanism maps' which show the competition between particle dissolution
and grain growth during isothermal heat treatment for a wide range of operational conditions.
Examples of such diagrams are given in Figs. 5.14 and 5.15.
As expected, the stability of the second phase particles is sensitive to variations in the
proportionality constant m*. Normally, the precipitates will exert a drag on the grain boundaries as long as they are present in the metal matrix. However, when the dissolution process is
completed, the matrix grains are free to grow without any interference from precipitates. This

Time exponent n = 0.5

D,jim

Complete particle
dissolution

Stable particles

I 1 ,nm 2
Fig. 5.14. Predicted variation in average grain size D with Z1 and ra* for Dum. = 50 um, n = 0.5, and
D0 =0 (dissolving precipitates).

D,jim

Time exponent n = 0.3


Complete particle dissolution

I 1 ^m 1 ' 0
Fig. 5.15. Predicted variation in average grain size D with I1 and m* for Dum. - 50 Jim, n - 0.3, and
Do =0 (dissolving precipitates).
means that the grains, after prolonged high temperature annealing, will coarsen at a rate which
is comparable with that observed in Fig. 5.7 for particle-free systems. In the limiting case,
when m* = 0, the grain growth behaviour becomes identical to that shown in Figs. 5.9 and 5.10
for stable precipitates.
Example (5.3)

Consider a niobium-microalloyed steel with the following composition:


Nb(total): 0.025 wt%, Nb(soluble): 0.010 wt%, C: 0.10 wt%
Assume that the base metal contains a fine dispersion of NbC precipitates in the as-received
condition, conforming to a limiting austenite grain size Dnm. of 50 jim. Provided that the
boundary drag from impurity elements in solid solution can be neglected (i.e. n ~ 0.5), estimate on the basis of Fig. 5.14 the average austenite grain size D 7 in the material after 25 s of
isothermal annealing at 13000C. Relevant physical data for niobium-microalloyed steels are
given below:

Solution

The initial volume fraction of NbC in the material can be estimated from simple stoichiometric
calculations by considering the difference between total and soluble niobium. Taking the
atomic weight of Nb and C equal to 92.9 and 12.0 g mol"1, respectively, we obtain:

From this we see that the radius of the NbC precipitates in the base metal is close to:

As shown in Example 4.6 (Chapter 4), the dimensionless supersaturation of niobium a ^


adjacent to the particle/matrix interface during dissolution can be written as:

By substituting this value into the expression for the proportionality constant m*, we obtain:

Moreover, at 1300C the value OfZ1 becomes:

The average austenite grain size can now be read from Fig. 5.14 by interpolation between
the curves for m* = 1 and/= 0 (free grain growth). This gives:

Since the calculated value of D1 is reasonably close to that observed for a particle-free
system, it means that the presence of a fine dispersion of NbC in the base metal has no significant effect on the resulting austenite grain size under the prevailing circumstances. Other

Next Page

types of niobium microalloyed steels may reveal a different grain coarsening behaviour, depending on the chemical composition, size distribution, and initial volume fraction of the base
metal precipitates. However, the pattern remains essentially the same, i.e. the growth inhibition is always succeeded by grain coarsening as long as the precipitates are thermally unstable.

5.4 Grain Growth Diagrams for Steel Welding


In welding the temperature will change continuously with time, which makes predictions of
the HAZ grain coarsening behaviour rather complicated. The method adopted from Ashby et
^ 24,25 j s b asec i o n m e jd e a o f integrating the elementary kinetic models over the weld thermal
cycle where the unknown kinetic constants are determined by fitting the integrals at certain
fixed points to data from real or simulated welds. Although the introduction of the Zener drag
in the grain growth equation largely increases the complexity of the problem, the methodology
and calibration procedure remain essentially the same. This means that the results from such
complex computations can be presented in the form of simple grain growth diagrams which
show contours of constant grain size in temperature-time space.
5.4.1 Construction of diagrams
A grain growth model for welding consists of two components, i.e. a heat flow model, and a
structural (kinetic) model.
5.4.1.1 Heat flow models
As a first simplification, the general Rosenthal equations26 are considered for the limiting case
of a high net power qo and a high welding speed v, maintaining the ratio qo/v within a range
applicable to arc welding. It has been shown in Chapter 1 that under such conditions, where no
exchange of heat occurs in the .^-direction, the following equations apply:
Thick plate welding (2-D heat flow)
(5-39)
Thin plate welding (1-D heat flow)

Previous Page

types of niobium microalloyed steels may reveal a different grain coarsening behaviour, depending on the chemical composition, size distribution, and initial volume fraction of the base
metal precipitates. However, the pattern remains essentially the same, i.e. the growth inhibition is always succeeded by grain coarsening as long as the precipitates are thermally unstable.

5.4 Grain Growth Diagrams for Steel Welding


In welding the temperature will change continuously with time, which makes predictions of
the HAZ grain coarsening behaviour rather complicated. The method adopted from Ashby et
^ 24,25 j s b asec i o n m e jd e a o f integrating the elementary kinetic models over the weld thermal
cycle where the unknown kinetic constants are determined by fitting the integrals at certain
fixed points to data from real or simulated welds. Although the introduction of the Zener drag
in the grain growth equation largely increases the complexity of the problem, the methodology
and calibration procedure remain essentially the same. This means that the results from such
complex computations can be presented in the form of simple grain growth diagrams which
show contours of constant grain size in temperature-time space.
5.4.1 Construction of diagrams
A grain growth model for welding consists of two components, i.e. a heat flow model, and a
structural (kinetic) model.
5.4.1.1 Heat flow models
As a first simplification, the general Rosenthal equations26 are considered for the limiting case
of a high net power qo and a high welding speed v, maintaining the ratio qo/v within a range
applicable to arc welding. It has been shown in Chapter 1 that under such conditions, where no
exchange of heat occurs in the .^-direction, the following equations apply:
Thick plate welding (2-D heat flow)
(5-39)
Thin plate welding (1-D heat flow)

5.4.1.2 Grain growth model


The use of equation (5-22) for prediction of the HAZ grain structure requires quantitative
information about the different kinetic constants entering the model. In the following we shall
assume that data for the time exponent n, the activation energies Qam and Qs, the temperature
dependence of the dimensionless supersaturation a and the element diffusivity Dm as well as
the initial and limiting grain sizes D0 and Dum. are available in the literature. From knowledge of these parameters it is possible to calibrate the model against data from real or simulated welds by adjusting the remaining kinetic constants so that a good agreement is obtained
between theory and experiments.
5.4.1.3 Calibration procedure
The calibration procedure involves the following basic steps:
Evaluation of experimental data
Supposejhat the mean grain size at two different locations in the HAZ is known (designated
D1 and Z)2, respectively). At each of these locations the peak temperature T and the cooling
time within a specific temperature range (e.g. from 800 to 5000C), Af8/5, are given. The thermal
cycles for the points can then be computed from either equation (5-39) or (5-40). Using these
temperature profiles and an appropriate value for the activation energy Q , the I1IM* ratio
is calculated from the integral:

(5-41)
where the limits tx and t2 refer to the total time spent in the thermal cycle from the chosen
reference temperature Tc to the peak temperature Tp and down again to Tc.
Tuning of coarsening model
When the parameters n, Qapp., Qs, D0, and Dlim are known, it is possible to evaluate the
remaining (unknown) kinetic constants from equations (5-22) and (5-30) by an iterative procedure. The following parameter is defined for this purpose:
(5-42)
The next step is to calculate the integral
(5-43)

for the same pairs of values of Tp and Ar875 as above. The differential grain growth equation
can now be solved by selecting an appropriate starting value for M0* and evaluating the corresponding Q-value which conforms to a mean grain size of D\ and D2, respectively. The
computations are repeated by adjusting M0* until a contour in Mo*-Q space is built up for each
grain size. The accepted values of M* and Q are then found by considering the intersection
point between the two curves, as shown schematically in Fig. 5.16(a).

(a)

MS.

Accepted,
"" value
Accepted^
" ~vaTue~

a
(b)

MJ

Accepted.
~ value" "
Acce|3ted_
value" "

Q*
Fig. 5.16. Method for calibrating unknown kinetic constants to experimental grain growth data;
(a) M*o-Q. (coarsening model), (b) M*o-Q* (dissolution model).

Tuning of dissolution model


In this case the unknown kinetic constants are combined in a single calibration parameter 2*,
defined as:
(5-44)
where a and Dm include all constants entering the expressions for the dimensionless
supersaturation a and the element diffusivity Dm, respectively.

Under such conditions equation (5-37) becomes:


(5-45)
where AH* is the standard enthalpy of the dissolution reaction per mole of the diffusate (defined by the solubility product in equation (4-5), Chapter 4), and Qd is the activation energy for
diffusion of the less mobile constituent atom of the precipitates in the matrix.
By calculating the integral in equation (5-45) for the same pairs of values of Tp and Af8/5 as
above and selecting an appropriate starting value for Mo* in equation (5-41), it is possible to
build up a contour in Mo*-T space for each grain size that satisfies the differential grain
growth equation. The accepted values of M* and Q* are then found by considering the intersection point between the curves representing D1 and D2 in Fig. 5.16(b).
5.4.1.4 Axes and features of diagrams
The microstructural information calculated using the mathematical models described in the
previous sections may be plotted on various kinds of welding diagrams. The graphical representation chosen here has been adopted from Ashby et al?^25 The process diagrams have
axes of weld input energy and peak temperature, and display contours of constant grain size
along with information about the thermal stability of the grain boundary pinning precipitates.
The axes can, in turn, be converted into an equivalent cooling time, Atm, and isothermal zone
width Ar* or Aym through equations (5-39) and (5-40):
Thick plate welding (2-D heat flow)

(5-46)

(5-47)

where Tp < Tn, (Tn, is the melting point).


Thin plate welding (1-D heat flow)

(5-48)

(5-49)
where Tp < Tn,.

Welding diagrams of this kind are very useful, since they summarise the effect of the important process variables in a systematic and illustrative manner at the same time as they provide
a good overall indication of the grain growth behaviour of materials during welding. In addition, the diagrams can be used for quantitative predictions of the austenite grain size across the
HAZ of steel welds for a wide range of operational conditions. This will be illustrated below
in various numerical examples.
5.4.2 Case studies
The following section describes grain growth diagrams for different types of steels. The parameters used to construct the maps are either tabulated or included in the text. Some of these
are taken from the literature, while others are arrived at by fitting the theory of the previous
sections to data from real or simulated welds according to the procedure shown in Fig. 5.16.
To obtain a consistent basis, all grain sizes reported here conform to the mean linear intercept.28 Conversion to three-dimensional grain sizes may then be done through equation (5-4)
or by the use of other appropriate conversion factors (e.g. 1.776 as recommended by Ashby et
a/.24'25).
5.4.2.1 Titanium-microalloyed steels
Titanium-microalloyed steels are widely used in welded structures. From a thermodynamic
standpoint, additions of small amounts of titanium to steel would be expected to impede austenite
grain coarsening during welding by virtue of its ability to form stable nitrides even at high
temperatures (see discussion in Chapter 4). However, certain restrictions must be adhered
during casting and subsequent thermomechanical processing of the steel so that the number
density of TiN particles is sufficiently high to retard grain growth.22 This currently limits the
use of titanium for austenite grain size control to continuously cast and controlled rolled steels.
Typical HAZ grain growth diagrams for Ti-microalloyed steels can be constructed on the
basis the experimental data reported by Ion et al.25 Tables 5.1 and 5.2 contain information
about the steel chemical composition and the parameters used in the computations, while Fig.
5.17 gives an overall indication of the accuracy of the predictions after calibration of the model
against two experimental data points.
Table 5.1 Chemical composition of Ti-microalloyed steel used by Ion et al.25 (in wt%).
C

Si

Mn

Al

Ti

0.12

0.23

1.53

0.005

0.006

0.02

0.011

0.009

Table 5.2 Data used to construct welding maps for Ti-microalloyed steel (compiled from miscellaneous sources).

"
0.5

I ~PP I Q? I Hm.

I K I ^* I

(kJmol-1)

(kJmol-1)

(|xm)

(jun)

(IJLm2S-1)

(1JLm3Ks-1)

224

240

20

7.7 X 109

9.9 X 1014

!Activation energy for diffusion of Ti in austenite.

Ti-microalioyed steel
E
=i
s
(a
I
Oi
I
i
Calibration point

Predicted grain size,jim


Fig. 5.17. Comparison between measured and predicted HAZ austenite grain sizes after calibration of
model to data reported by Ion et al.25 for Ti-microalloyed steel (simulated thick plate welds).
The response of the base material to welding under 2-D and 1-D heat flow conditions is
shown in Fig. 5.18(a) and (b), respectively. As expected, the presence of TiN particles is seen
to retard austenite grain growth within the heat affected zone during welding. However, since
particle coarsening is a physical process occurring at temperatures well below the equilibrium
solvus of the precipitates, the problem cannot be eliminated. This means that a coarse grained
region will always form adjacent to the fusion boundary, even at very low heat inputs, as
indicated by the nomograms in Fig. 5.18(a) and (b).
Example (5.4)

Consider SA welding on a thick plate of a titanium-microalloyed steel under the following


conditions:

Evaluate on the basis of the nomograms in Fig. 5.18(a) the variation in the austenite grain
size across the fully transformed HAZ after welding. Estimate also the total width of the HAZ
(referred to the fusion boundary) under the prevailing circumstances.
Solution

First we calculate the net heat input per unit length of the weld:

Readings from Fig. 5.18(a) give the HAZ austenite grain size profile shown in Fig. 5.19.

Thick plate welding (2-D heat flow)


Prior austenite grain size (^m)
Relative size of pinning precipitates (r/ro)

Cooling time, At B/5 , s

Net heat input (qo/v), kJ/mm

(a)

Peak temperature, 0C
Thin plate welding (1-D heat flow)
Prior austenite grain size dim)
Relative size of pinning precipitates (r/ro)

Net heat input (qo/vd), kJ/mm2

(b)

Peak temperature, 0C
FZg. 5.18. HAZ grain growth diagrams for titanium-microalloyed steel; (a) Thick plate welding (2-D heat
flow), (b) Thin plate welding (1-D heat flow). No preheating (T0 = 200C).

Austenite grain size, ^m

SAW (Ti-microalloyed steel)

Peak temperature, 0C
Fig. 5.19. Predicted variation in austenite grain size across the fully transformed HAZ of a Ti-microalloyed
steel weld (Example 5.4).
The total width of the fully transformed HAZ can now be estimated from equation (5-47)
by using data from Table 1.1 (Chapter 1). Taking the Ac3 -temperature equal to 9100C, we
obtain:

Based on the quoted value of 2 in Table 5.2, it is also possible to estimate the initial volume
fraction of TiN in the base metal. Taking C5 6.67 X 104 |im3 K s"1 and k 0.5 for titaniummicroalloyed steels, we obtain from equation (5-42):

A comparison with the experimental data of Ringer et aill shows that a volume fraction of
2 X 10"4 is reasonably close to that measured by microscopic assessment methods.

5.4.2.2 Niobium-microalloyed steels


This class of steel is used for a variety of applications, including ship building, pressure vessels, oil platforms and bridges. The steels contain small amounts of microalloying elements
such as niobium, vanadium, and aluminium which during thermomechanical processing combine with carbon and nitrogen to form fine dispersions of grain boundary pinning precipitates.
They are readily weldable, but suffer from severe HAZ grain coarsening due to dissolution of
the carbo-nitrides at elevated temperatures.1822

Typical HAZ grain growth diagrams for Nb-microalloyed steels (with chemical composition as in Table 5.3) can be constructed on the basis of the experimental data of Ion et al. 25 The
parameters used to compute the diagrams are listed in Table 5.4. Figure 5.20 shows a correlation between predicted and observed grain sizes after calibration of the model against two
experimental data points.
It is evident from the nomograms in Fig. 5.21 (a) and (b) that considerable austenite grain
growth occurs during welding of niobium-microalloyed steels because of dissolution of the
grain boundary pinning precipitates. On the average, the HAZ austenite grain size adjacent to
the fusion boundary is four to six times larger than that observed for titanium-microalloyed
steels. This gives rise to a high HAZ hardenability, which facilitates formation of low temperature transformation products such as martensite and bainite during welding ?2
Example (5.5)

In Table 5.4 the quoted value for the effective grain boundary pinning constant Q* is 5.0 X
1018 |im 2 s"1. Based on equations (5-27) and (5-44), estimate the initial volume fraction/, and
radius ro of NbC in the parent material under the prevailing circumstances.
Table 5.3 Chemical composition of Nb-microalloyed steel used by Ion et al.25 (in wt%).
C

Si

Mn

Al

Nb

0.12

0.16

0.91

0.002

0.005

0.04

0.021

0.011

Measured grain size, urn

Nb-microalloyed steel
SAW
GMAW
Laser welding
Weld simulation
Calibration point

Predicted grain size, jam


Fig. 5.20. Comparison between measured and predicted HAZ austenite grain sizes after calibration of
model to data reported by Ion et al.25 for Nb-microalloyed steel (real and simulated thick plate welds).

Thick plate welding (2-D heat flow)


Prior austenite grain size (urn)
Relative volume fraction of pinning precipitates (f/fo)

Cooling time, M8^, s

Net heat input (qo/v), kJ/mm

(a)

Peak temperature, 0C
Thin plate welding (1-D heat flow)
Prior austenite grain size (urn)
Relative volume fraction of pinning precipitates (f/fo)

Net heat Input (qo/vd), kJ/mm2

(b)

Peak temperature, 0C
Fig. 5.21. HAZ grain growth diagrams for niobium-microalloyed steel; (a) Thick plate welding (2-D heat
flow), (b) Thin plate welding (1-D heat flow). No preheating (T0 = 200C).

Table 5.4 Data used to construct welding maps for Nb-microalloyed steel (compiled from miscellaneous sources).

n
0.5

I Qapp. I ^*f I 0} I Dlim.


(kJmol-1)

(kJmol-1)

(kJ mol-1)

((Jim)

224

130

343

11

I ^

(|xm)
6

I K* I

"^

(1Xm2S-1)

(1Xm2S-1)

37 X 109

5.0 X 1018

fEstimated from the solubility product of NbC in austenite.


^Activation energy for diffusion of Nb in austenite.

Reasonable average values for a 0 , Dm, k, and Dnm. are given below:
20.5 (estimated from the solubility product of NbC in austenite, Example
4.6)
5.9 X 1010 Jim2 s"1 (diffusion constant for Nb in austenite)
0.5 (Zener coefficient)
11 |Lim (from Table 5.4)
Solution

The initial volume fraction of NbC can be estimated from equation (5-44). After rearranging
this equation, we obtain:

The corresponding radius of the precipitates can now be obtained from equation (5-27):

Although experimental data are not available for a direct comparison, the calculated values
for/o and ro are reasonable and of the expected order of magnitude (see Example 5.3).
5.4.2.3 C-Mn steel weld metals
In C-Mn steel weld metals the volume fraction of non-metallic inclusions is considerably
higher than for normal cast steel products because of the limited time available for growth and
separation of the deoxidation products (see discussion in Chapter T). For the same reason, the
weld metal inclusions are also significantly smaller in dimension and more finely dispersed.
Since oxides and sulphides will neither coarsen nor dissolve during the weld thermal cycle,
they will ultimately lead to a stabilisation of the austenite grain structure within the reheated
regions of multipass steel weld metals. This is illustrated schematically in Fig. 5.22.
Grain growth diagrams for C-Mn steel weld metals (with chemical compostion as in Table
5.5) can be constructed on the basis of the theory described in the previous sections by selecting reasonable average values for the kinetic constants in equation (5-22). The combination of
parameters listed in Table 5.6 gives a fair agreement between predictions and measurements,
as shown in Fig. 5.23.
The response of the material to reheating above the Ac3-temperature under 2-D and 1-D
heat flow conditions is illustrated in Fig. 5.24(a) and (b). As expected, the presence of finely

Weld metal

Base plate

AustenJte grain size

Fully transformed HAZ

Peak temperature
Fig. 5.22. Schematic diagram illustrating the effect of non-metallic inclusions on the weld metal grain
coarsening behaviour.
Table 5.5 Chemical composition of C-Mn steel weld metal used by Kluken et al.29 (in wt%).
C

Si

Mn

Nb

Al

Ti

0.09

0.034

0.48

1.86

0.01

0.01

0.005

0.004

0.02

0.018

0.005

Table 5.6 Data used to construct welding maps for C-Mn steel weld metal (compiled from
miscellaneous sources).

"

I
0.5

~pp. I K
1

I bHm. I K?

(kJmol" )

(|xm)

(|xm)

(JJLm2S-1)

224

10

95

20X109

dispersed inclusions within the weld metal gives rise to a strong austenite grain boundary
pinning effect, similar to that documented for TiN in microalloyed steels. However, since no
particle coarsening occurs in the present case, it means that a small austenite grain size
(< 95jnm) is preserved at all relevant peak temperatures, irrespectively of the applied heat
input. Accordingly, the weld metal grain growth behaviour is seen to be quite different from
that of the base metal, even when the nominal chemical composition has not been significantly
changed by the welding process.

Austenite grain size, jim

C-Mn steel weld metal


Peak temperature: 1350 C

Cooling time, A t . . s
8/5

Fig. 5.23. Comparison between measured and predicted austenite grain sizes after calibration of model
to data reported by Kluken et al.29 for C-Mn steel weld metal (simulated thick plate heat cycles).
Example (5.6)

Consider deposition of a cap layer (GMAW) on the top of a thick multipass Nb-microalloyed
steel weld under the following conditions:

Estimate on the basis of the nomograms in Figs. 5.21(a) and 5.24(a) the variation in the
prior austenite grain size across the fully transformed HAZ at different locations along the
periphery of the weld after arc extinction. The situation is illustrated in Fig. 5.25.
Solution

First we calculate the net heat input per unit length of the weld:

Readings from Figs. 5.21(a) and 5.24(a) give the HAZ austenite grain size profiles shown
in Fig. 5.26. From this we see that the prior austenite grain size adjacent to the fusion boundary of a cap layer will vary significantly with position along the periphery of the weld, depending on the type of material sampled (i.e. base plate or weld metal).
5.4.2.4 Cr-Mo low-alloy steels
Chromium-molybdenum low-alloy steels are widely used in the petroleum industry and in
high-power-generating equipment because of their good corrosion and oxidation resistance

Thick plate welding (2-D heat flow)

Cooling time, At 8/5 , s

Net heat input (qo/v), kJ/mm

(a)

Peak temperature, 0C
Thin plate welding (1 -D heat flow)

Net heat input (qo/vd), kJ/mm2

(b)

Peak temperature, 0C
Fig. 5.24. Grain growth diagrams for reheated C-Mn steel weld metal; (a) Thick plate welding (2-D heat
flow), (b) Thin plate welding (1-D heat flow). No preheating (T0 = 200C).

HAZ

Base plate

Dissolution of NbC

Austenite grain size, ^m

Fig. 5.25. Sketch of multipass steel weld in Example 5.6.

Peak temperature, 0C
Fig. 5.26. Predicted variation in austenite grain size across the fully transformed HAZ at different locations along the periphery of the weld (Example 5.6).

and their high creep strength. They are readily weldable, although reheat cracking and cold
cracking in the HAZ and fusion zone may be a problem.30 In practice, these difficulties can be
overcome by the choice of an appropriate welding procedure (e.g. preheating in combination
with a low heat input), which reduces austenite grain growth and maximises the proportion of
the HAZ refined by subsequent weld passes.3031
The grain growth data of Miranda and Fortes31 provide a basis of calibrating the kinetic
constants in equations (5-22) and (5-45), as shown in Fig. 5.27. Information about steel chemical
composition and parameters used to construct the maps are contained in Table 5.7 and 5.8,
respectively.
It is seen from the nomograms in Fig. 5.28(a) and (b) that considerable austenite grain
growth occurs in Cr-Mo low-alloy steel weldments due to dissolution of molybdenum carbide
(Mo2C) at elevated temperatures. However, the maximum HAZ austenite grain size is signifiTable 5.7 Chemical composition of Cr-Mo steel used by Miranda and Fortes31 (in wt%).
C

Si

Mn

Cr

Mo

Ni

Cu

Al

0.10

0.24

0.46

0.01

0.01

2.19

0.94

0.23

0.1

0.01

0.01

Predicted grain size, jam

Calibration points

Symbol * w s >

Measured grain size, |im


Fig. 5.27. Comparison between measured and predicted HAZ austenite grain sizes after calibration of
model to data reported by Miranda and Fortes31 for Cr-Mo low-alloy steel (SAW-thick plates).
Table 5.8 Data used to construct welding maps for Cr-Mo low-alloy steel (from Refs. 24 and 31).

n
0.32

I QapP. I A//*+ I QP I blm

T ~ F I MO* I QT~

(kJmol-1)

(kJmol-1)

(kJmor"1)

(^m)

(fim)

(IJLm1711S-1)

(IXm2S-1)

180

71

300

20

1.3 X 109

7.8 X 1017

f Estimated from the solubility product of Mo2C in austenite.


f f Activation energy for diffusion of Mo in austenite.

cantly smaller than that observed during welding of Nb-microalloyed steels, in spite of the fact
that Mo2C is less stable than NbC. This situation can probably be attributed to drag from
alloying and impurity elements in solid solution, which retard grain growth through elastic
attraction of the atoms towards the open structure of the grain boundary (indicated by a time
exponent n = 0.32 in Table 5.8).
5.4.2.5 Type 316 austenitic stainless steels
The 316 series of stainless steels have good general corrosion resistance, but are sensitive to
stress corrosion cracking in the presence of chlorides. They are used extensively in the chemical industry and in power plants, particularly in advanced nuclear technology. During welding
dissolution of the base metal chromium carbides (e.g. C^C 6 ) will occur in parts of the HAZ
where the peak temperature of the thermal cycle has been above about 8500C. This gives rise
to considerable austenite grain growth.24
HAZ grain growth diagrams for this type of steel (with chemical composition as in Table
5.9) can be constructed on the basis of the experimental data of Ashby and Easterling,24 using
the parameters listed in Table 5.10. Figure 5.29 shows a correlation between predicted and
observed grain sizes after calibration of the model against two experimental data points.

Thick plate welding (2-D heat flow)


Prior austenite grain size (urn)
Relative volume fraction of pinning precipitates (Vi0)

Cooling time, At 8751 s

Net heat input (qo/v), kJ/mm

(a)

Peak temperature, 0C
Thin plate welding (1-D heat flow)
Prior austenite grain size (urn)
Relative volume fraction of pinning precipitates (f/y

Net heat input (qo/vd), kJ/mm2

(b)

Peak temperature, 0C
Fig. 5.28. HAZ grain growth diagrams for Cr-Mo low-alloy steel; (a) Thick plate welding (2-D heat
flow), (b) Thin plate welding (1-D heat flow). No preheating (T0 = 200C).

Table 5.9 Chemical composition of type 316 austenitic stainless steel used by Ashby and Easterling24
(in wt%).
C

Mn

Si

Cr

Ni

Mo

0.05

2.0

1.0

19

11

2.3

0.045

0.030

Table 5.10 Data used to construct welding maps for type 316 austenitic stainless steel (compiled from
miscellaneous sources).

I Qapp.

I Atf*+ I Q/ I Dlm.
1

I K

o*

(kJmol- )

(kJmol- )

(kJ mol" )

(|jim)

(fxm)

(1JLm2S-1)

224

60

240

18

18

3.4 X IQ9

0.5

I ^*

(1Xm2S-1)

2.1 X IQ14

Measured grain size, u.m

!Estimated from data quoted by Kou32 for Cr23C6 in 304 stainless steel.
^Activation energy for diffusion of Cr in austenite.

Calibration point

Predicted grain size, ^m

Fig. 5.29. Comparison between measured and predicted HAZ austenite grain sizes after calibration of
model to data reported by Ashby and Easterling24 for type 316 stainless steel (simulated thick plate
welds).
It follows from the nomograms in Fig. 5.30(a) and (b) that the presence of Cr23C6 has little
influence on the HAZ grain coarsening behaviour because it dissolves at a fairly low temperature. This results in a rather coarse austenite grain structure adjacent to the fusion boundary,
which in certain cases may exceed 100 jim.

Cooling time, At8/5, s

Net heat input (qo/v), kJ/mm

Thick plate welding (2-D heat flow)


Austenite grain size ^x m)
Relative volume fraction of pinning precipitates (f/fo)

Peak temperature, 0C

Net heat input (qo/vd), kJ/mm2

Thin plate welding (1 -D heat flow)


Austenite grain size (JLIITI)
Relative volume fraction of pinning precipitates (f/fo)

Peak temperature, 0C
Fig. 5.30. HAZ grain growth diagrams for type 316 austenitic stainless steel; (a) Thick plate welding
(2-D heat flow), (b) Thin plate welding (1-D heat flow). No preheating (T0 = 200C).

Example (5.7)

Consider GTA butt welding of a 2mm thin sheet of type 316 austenitic stainless steel under the
following conditions:

Provided that the conditions for one dimensional (1-D) heat flow are met, estimate on the
basis of the nomograms in Fig. 5.30(b) the variation in the austenite grain size across the HAZ
after welding. Estimate also the total width of the HAZ (referred to the fusion boundary)
under the prevailing circumstances.
Solution

First we calculate the net heat input per mm2 of the weld:

Austenite grain size, jim

Readings from Fig. 5.30(b) give the HAZ austenite grain size profile shown in Fig. 5.31.
The total width of the grain growth zone can be estimated from equation (5-49) by using
data from Table 1.1 (Chapter 1). Taking the peak temperature for incipient dissolution of
Cr23C6 equal to 9000C, we obtain:

Peak temperature, 0C
Fig. 5.31. Predicted variation in austenite grain size across the HAZ of a 316 austenitic stainless steel
weld (Example 5.7).

5.5 Computer Simulation of Grain Growth


The multivariable characteristics of grain growth impose several restrictions on the use of
analytical modelling techniques for a mathematical description of the grain evolution during
welding. However, many of these restrictions can be relaxed by employing novel computational techniques in mainframe facilities which enable the modelling of exact grain shapes in
topological connected microstructures.
Historically, the application of the computer simulation technique to poly cry stalline microstructures has evolved along two different paths. The first approach treats the grain boundaries
as continuous interfaces, which are governed by determistic equations of motion.33"37 The
other approach is a Monte Carlo-based technique that takes into account explicitly the interactions among individual grains by discretising the microstructure to construct an image of the
grain aggregates in the computer.38"43 Both approaches have been applied to the modelling of
grain growth phenomena in welding 374445 with the objective of incorporating important sideeffects which cannot readily be accounted for in a simple analytical treatment of the process.
5.5.1 Grain growth in the presence of a temperature gradient
In comparing HAZ grain growth in thermally simulated and real welds it is sometimes found
that the maximum grain size is larger in simulated specimens than in a weld HAZ when comparison is made on the basis of a similar temperature-time programme.16 A possible explanation to this observation is the phenomenon of 'thermal pinning' which can be attributed to the
presence of steep temperature gradients in a weld HAZ. This effect has been disregarded in the
previous analytical analysis, but can readily be incorporated in a numerical model.
Following the treatment of Fortes and Soares37 the effect of 'thermal pinning' can be simulated by considering 1-D grain growth in a thin polycrystal with a Gaussian grain distribution.
As an illustration of principles, a constant mobility gradient was assumed, the extreme mobilities
being in the ratio of 5:1 and 10:1 in different simulations. Simulation with a uniform grain
mobility (UNIF) was then carried out for comparison. The polycrystal was subsequently divided into three regions of the same size, i.e. cold (C), intermediate (IM), and hot (H) regions,
respectively, and the grain size distributions were obtained for each of these regions and for the
entire polycrystal (G: global region). The simulations were carried out until the total number
of grains were reduced from initially 5000 to 1500 grains. The results of the computations are
summarised in Fig. 5.32.
It is evident from Fig. 5.32(a) that grain growth is slower in a mobility gradient than under
isothermal heat treatment conditions (UNIF). This effect becomes more pronounced as the
mobility gradient increases. Moreover, a closer inspection of Figs. 5.32(b) and (c) reveals that
the grain growth rate of the entire crystal (global region: G) is slower than the corresponding
coarsening rate of the intermediate region (IM), which again indicates that grain growth is
slower in a mobility gradient. Although the simulation results in Fig. 5.32 are not directly
applicable to welding (in welding the temperature, and hence the mobility, at a fixed point in
the crystal will not be constant but vary with time), these findings provide a strong motivation
for incorporating the 'thermal pinning' effect in future refinements of the HAZ grain growth
models.

D x 10 (arbitrary units)

(a)

t1/2 (arbitrary units)

D x 10 (arbitrary units)

(b)

t1/2 (arbitrary units)

D x 10 (arbitrary units)

(C)

t1/2 (arbitrary units)


Fig. 5.32. Effect of 'thermal pinning' on grain growth; (a) Plot of D vs t for different simulation conditions, (b) Coarsening kinetics of different thermal regions for a mobility gradient of 5:1, (c) Same as in
(b) for a mobility gradient of 10:1. Data from Fortes and Soares.37

5.5.2 Free surface effects


In a weld HAZ the fusion line represents a physical barrier against grain growth which cannot
be exceeded. In principle, it can be regarded as a free surface, which means that the grain
boundaries must meet the fusion line at right angles.
The simulation results of Saetre and Ryum36 provide a basis for evaluating to what extent
grain growth under isothermal conditions is affected by the presence of a free surface. Figure
5.33 shows the evolution of a 2-D Voronoi structure with time. In Fig. 5.33(a) all grain boundaries
are straight lines, but the triple line junctions are not in equilibrium. However, adjustments of
the triple line junctions into equilibrium positions lead to local curvatures of the grain boundaries near the junctions, and the grain growth process is thus initiated. Initially, the Voronoi
structure contained 485 grains, but this number is gradually reduced during the coarsening
process (Figs. 5.33(b), (c) and (d)). A qualitative inspection of the diagrams reveals, on the
other hand, no clear difference in the grain size in the radial position, which indicates that the
constrain provided by the free surface is only of minor importance in the present context. It
should be noted that this does not exclude the possibility that the HAZ grain size is influenced
by the presence of a fusion boundary, since other effects such as surface grooving and solute
segregation (not studied here) can impose additional restrictions on the system by contributing
to physical pinning of the grain boundaries. Consequently, further modelling work is required
to explore these possibilities.

References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.

H. Hu and B.B. Rath: Metall Trans., 1970,1, 3181-3184.


J.W. Cahn: Acta Metall, 1962,10, 789-798.
K. Lucke and H. Stuwe: in Recovery and Recrystallization of Metals, 1963, New York,
Interscience.
E. Nes, N. Ryum and O. Hunderi: Acta Metall, 1985, 33, 11-22.
E.E. Underwood: Quantitative Stereology, 1970, London, Addison-Wesley Publ. Co.
C S . Pande: Acta Metall, 1987, 35, 2671-2678.
B.R. Patterson and Y. Liu: Metall Trans., 1992, 23A, 2481-2482.
P. Hellman and M. Hillert: Scand. J. Metall, 1975, 4, 211-219.
R.A. Vandermeer: Acta Metall, 1967,15, 447-458.
C. Zener (quoted by C S . Smith): Trans. AIME, 1948,175, 15-51.
CH. Worner and P.M. Hazzledine: JOM, 1992, 44 (No. 9), 16-20.
B.B. Rath and H. Hu: Trans. TMS-AIME, 1969, 245, 1243-1252, ibid., 1577-1585.
T. Gladman: Proc. Roy. Soc, 1966, 294A, 298-309.
G. Grewal and S. Ankem: Acta Metall Mater, 1990, 38, 1607-1617.
T. Skaland and 0. Grong: University of Trondheim, The Norwegian Institute of Technology,
Trondheim Norway (unpublished work).
K.E. Easterling: Introduction to the Physical Metallurgy of Welding, 2nd Edn, 1992, Oxford,
Butterworth-Heinemann Ltd.
S.P. Ringer, W.B. Li and K.E. Easterling: Acta Metall, 1989, 37, 831-841.
O.M. Akselsen, 0. Grong, N. Ryum and N. Christensen: Acta Metall, 1986, 34, 1807-1815.
A.M. Brown and M.F. Ashby: Acta Metall, 1980, 28, 1085-1101.
J.M. Lifshitz and V.V. Slyozov: J. Physics Chem. Solids, 1961,19, 35-50.
C. Wagner: Z. Electrochem., 1961, 65, 581-591.
0. Grong and O.M. Akselsen: Metal Construction, 1986,18, 557-562.

(a)

(b)

(C)

(d)

Fig. 5.33. The evolution of 2-D grain structures during normal grain growth; (a) Initial Voronoi structure,
485 grains, (b) 456 grains, (c) 349 grains, (d) 251 grains. Data from Saetre and Ryum.36

23.
24.
25.
26.
27.
28.
29.

MJ. Whelan: Metal ScL J., 1969, 3, 95-97.


M.F. Ashby and K.E. Easterling: Acta MetalL, 1982, 30, 1969-1978.
J.C. Ion, K.E. Easterling and M.F. Ashby: Ada MetalL, 1984, 32,1949-1962.
D. Rosenthal: Trans. ASME, 1946, 68, 849-866.
O.R. Myhr and 0. Grong: Acta MetalL Mater., 1990, 38, 449-460.
ASTM Standard, El 12-84 (1984).
A.O. Kluken, 0. Grong and H. Hemmer: Technical Report STF34 F87093, 1987, Trondheim
(Norway), Sintef-Division of Metallurgy.

30. RJ. Alberry and W.K.C. Jones: Metals Technology, 1977, 4, 45-51; ibid., 360-364; ibid., 557566.
31. R.M. Miranda and M.A. Fortes: Mater. ScL Eng., 1989, A108, 1-8.
32. S. Kou: Welding Metallurgy, 1987, Toronto, John Wiley & Sons.
33. V.E. Fradkov and D.G. Udler: in Simulation and Theory of Evolving Microstructures (Eds M.R
Anderson and A.D. Rollett), 1990, TMS, Warrendale, Pa, pp. 15-29.
34. HJ. Frost, CV. Thompson and D.T. Walton: ibid., pp.31-39.
35. K. Kawasaki, T.Nagai and S. Ohta: ibid., pp.65-77.
36. T.O. Saetre and N. Ryum: in Modelling of Coarsening and Grain Growth (Eds C S . Pande and
S.R Marsh), 1993, The Minerals, Metals & Materials Society, pp.281-294.
37. M.A. Fortes and A. Soares: ibid., pp. 257-270.
38. A.D. Rollett, DJ. Srolovitz and M.R Anderson: Acta Metall, 1989, 37, 1227-1240.
39. G.S. Grest, M.R Anderson, DJ. Srolovitz and A.D. Rollett: Scripta Metall Mater., 1990, 24,
661-665.
40. DJ. Srolovitz, G.S. Grest and M.R Anderson: Acta Metall, 1986, 34, 1833-1845; ibid, 1988,
36,2115-2128.
41. T.O. Sartre, O. Hunderi and E. Nes: Acta Metall, 1986, 34, 981- 987.
42. K. Marthinsen, O. Lohne and E. Nes: Acta Metall, 1989, 37, 135-145.
43. A.D. Rollett, MJ. Lutony and DJ. Srolovitz: Acta Metall. Mater., 1992, 40, 43-55.
44. A. Kern, W. Reif and U. Schriever: Proc. Int. Conf. on Grain Coarsening, 18-21 June, 1991,
Rome, Mater. ScL Forum, 1992, vols 94-96, 709-714, Trans. Tech Publications.
45. B. Radhakrishnan and T. Zacharia: Proc. Int. Conf. on Modeling and Control of Joining Processes, 8-10 Dec, 1993, Orlando, FL, American Welding Society.

Appendix 5.1
Nomenclature
thermal diffusivity (mm2 s"1)

limiting grain size at h - 0 (|im or m)

end temperature of ferrite to austenite


transformation (0C or K)

element diffusivity (jum2 s"1, mm2 s"1


or m2 s"1)

nominal alloy composition (wt%, at.%


or mol rrr3)

kinetic constant in expression for Dm


(|Ltm2 s"1, mm2 s"1 or m2 s"1)

various kinetic constants and temperature-dependent parameters

particle volume fraction

plate thickness (mm)

initial particle volume fraction

grain size (jLirn or m)

maximum pinning force exerted by a


single particle ( J m 1 )

mean grain size (\im or m)


GMAW gas metal arc welding
mean linear grain intercept (JJJTI or m)
GTAW

gas tungsten arc welding

initial grain size (jim or m)


limiting grain size (jim or m)

standard enthalpy of dissolution reaction per mole of the diffusate (J mol"1


or kJ mol"1)

amperage (A)

activation energy for particle coarsening (J mor 1 or kJ mor 1 )

kinetic strength of thermal cycle with


respect to grain growth (jam1/n)

particle radius (nm, jam or m)

kinetic strength of thermal cycle with


respect to particle coarsening (fxm3)

initial particle radius at t - 0 (nm, jam


or m)

kinetic strength of thermal cycle with


respect to particle dissolution (jim2)

two-dimensional radius vector in y-z


plane (mm)

Zener coefficient

universial gas constant (8.314 J Kr1


mor 1 )

constant related to the relative rates of


grain growth and particle coarsening in
the material (jum3"1/n)

grain boundary area per unit volume


(jam2 per jum3 or m2 per m3)

constant related to the relative rates of


grain growth and particle dissolution in
the material (jam2"I/n)

time (s or min)

grain boundary mobility (variable units)

integration limits (s)

kinetic constant in expression for M


(variable units)

cooling time from 800 to 5000C (s)

submerged arc welding

temperature (0C or K)

modified kinetic constant in expression


for M(J^m1711S-1)

reference temperature (0C or K)

time exponent in grain growth equation

ambient temperature (0C or K)

number of interacting particles per unit


area of grain boundary (nr 2 )

melting point (0C or K)

number of particles per unit volume


(nr 3 )
driving pressure for grain growth
(J m-3)
effective driving pressure for grain
growth (J nr 3 )
retardation pressure due to second
phase particles (J irr 3 )
net arc power (W)
apparent activation energy for grain
growth (J mor 1 or kJ mor 1 )

peak temperature (0C or K)


voltage (V)
welding speed (mm s"1)
grain boundary migration rate (jam s~{
or m s"1)
grain boundary migration rate conforming to the low velocity limit (|Lim s 1 or
ms- 1 )
grain boundary migration rate conforming to the high velocity limit (jam s~l or
ms" 1 )
x-axis/welding direction (mm)

activation energy for element diffusion


in dissolution model (J moT 1 or
kJ mor 1 )

y-axis/transverse direction (mm)

z-axis/through thickness direction (mm)


dimensionless supersaturation in dissolution model

grain boundary interfacial energy


(J m-2)
calibration constant in coarsening
model (|Lim3 K s"1)

kinetic constant in expression for a


parameter depending on the diffusivity
and the interaction energy between the
grain boundary and the impurity atoms

calibration constant in dissolution


model (|Lim2 s"1)
density (g cm"3 or kg rrr3)

another complex function of the impurity diffusivity and the interaction energy between the grain boundary and
the impurity atoms

radius of curvature of a spherical grain


(jum or m)

intrinsic drag coefficient

thermal conductivity (W mm"1 0C"1)

arc efficiency

volume heat capacity (J mm"3 0C"1)

6
Solid State Transformations in Welds

6.1 Introduction
The majority of phase transformations occurring in the solid state take place by thermally
activated atomic movements. In welding we are particularly interested in transformations that
are induced by a change in temperature of an alloy with a fixed bulk composition. Such
transformations include precipitation reactions, eutectoid transformations, and massive transformations both in the weld metal and in the heat affected zone.
Since welding metallurgy is concerned with a number of different alloy systems (including
low and high alloy steels, aluminium alloys, titanium alloys etc.), it is not possible to cover all
aspects of transformation behaviour. Consequently, the aim of the present chapter is to provide the background material necessary for a verified quantitative understanding of phase
transformations in weldments in terms of models based on thermodynamics, kinetics, and
simple diffusion theory. These models will then be applied to specific alloy systems to illuminate the basic physical principles that underline the experimental observations and to predict
behaviour under conditions which have not yet been studied.

6.2 Transformation Kinetics


In order to understand the extent and direction of a transformation reaction, it is essential to
know how far the reaction can go and how fast it will proceed. To answer the first question we
need to consider the thermodynamics, whereas kinetic theory provides information about the
reaction rate.
6.2.1 Driving force for transformation reactions
The symbols and units used throughout this chapter are defined in Appendix 6.1.
In practice, solid state transformations require a certain degree of undercooling, which is
essential to accommodate the surface and strain energies of the new phase.1 Generally, this
minimum molar free energy of transformation, AG, can be written as a balance between the
following four contributions:
(6-1)
Here AGy (the volume free energy change associated with the transformation) and AGD
(free energy donated to the system when the nucleation takes place heterogeneously) are negative, since they assist the transformation, while AGS (increase in surface energy between the
two phases) and AGE (increase in strain energy resulting from lattice distortion) are both positive because they represent a barrier against nucleation. It follows that the transformation

Molar free energy

Stable (B)

Stable (a)

Temperature
Fig. 6.1. Schematic representation of the molar free energies of two solid a and P phases as a function of
temperature (allotrophic transformation no compositional change).

Temperature

reaction can proceed when the driving force AG becomes greater than the right-hand side of
equation (6-1).
For an allotropic transformation, in which there is no compositional change, AG will be a
simple function of temperature, as illustrated in Fig.6.1. For alloys the situation is slightly
more complex, since there is an additional variable, i.e. the composition. In such cases the
temperature at which the a-phase becomes thermodynamically unstable (Teq) corresponds to a
fixed point on the a-(3 solvus boundary in the equilibrium phase diagram, as shown
schematically in Fig. 6.2. Since phase diagrams are available for many of the important industrial alloy systems, it means that the driving force for a transformation reaction can readily be
obtained from such diagrams in the form of a characteristic undercooling (AT).

%B
Fig. 6.2. Schematic representation of the a-(3 solvus boundary in a simple binary phase diagram.

6.2.2 Heterogeneous nucleation in solids


In general, solid state transformations in metals and alloys occur heterogeneously by nucleation at high energy sites such as grain corners, grain boundaries, inclusions, dislocations and
vacancy clusters. The potency of a nucleation site, in turn, depends on the energy barrier
against nucleation (AG*^) which is a function of the 'wetting' conditions at the substrate/
nucleus interface.1
It can be seen from Fig. 6.3 that nucleation at for instance inclusions or dislocations is
always energetically more favourable than homogeneous nucleation (AG*het < tsG*hom ) but
less favourable than nucleation at grain boundaries or free surfaces. As a result, the transformation behaviour is strongly influenced by the type and density of lattice defects and second
phase particles present within the parent material.
6.2.2.1 Rate of heterogeneous nucleation
Whereas every atom is a potential nucleation site during homogeneous nucleation, only those
associated with lattice defects or second phase particles can take part in heterogeneous nucleation. In the latter case the rate of nucleation (Nhet) is given by:1'2

(6-2)
where v is a vibration frequency factor, Nv is the total number of heterogeneous nucleation
sites per unit volume, AG^, is the energy barrier against nucleation, and Qd is the activation
energy for atomic migration across the nucleus/matrix interface.

Grain boundary

Vacancy clusters

Dislocations/stacking faults

Inclusions

Grain
boundaries

Inclusion

Grain
corners

Free
surfaces

AG* /AG*
het.
horn.

Free
surface

Nucleation site
Fig. 6.3 Schematic diagram showing the most potent sites for heterogeneous nucleation in metals and
alloys.

It follows from the graphical representation of equation (6-2) in Fig. 6.4 that the nucleation
rate Nhet is highest at an intermediate temperature due to the competitive influence of
undercooling (driving force) and diffusivity on the reaction kinetics. This change in Nhet with
temperature gives rise to corresponding fluctuations in the transformation rate, as shown
schematically in Fig. 6.5. Note that the peak in transformation rate is due to two functions,
growth and nucleation (which peak at different T) whereas peak in Nhet is due to nucleation
only.
6.2.2.2 Determination

of AGhet and Qd

During the early stages of a precipitation reaction, the reaction rate may be controlled by the
nucleation rate Nhet. Under such conditions, the time taken to precipitate a certain fraction of
the new phase t* is inversely proportional to Nhet\

(6-3)
where C1 and C2 are kinetic constants.
By taking the natural logarithm on both sides of equation (6-3), we obtain:

(6-4)
If the complete C-curve is known for a specific transformation reaction, it is possible to
evaluate AG*het and Qd from equation (6-4) according to the procedure described by Ryum.3
In general, a plot of In t* vs HT will yield a distorted C-curve with well-defined asymptotes, as
shown in Fig. 6.6. At high undercoolings, when &G*het is negligible, the slope of the curve
becomes constant and equal to QdIR. The mathematical expression for this asymptote is:
T

Low undercooling
High diffusivity

High undercooling
Low diffusivity
%B

Nhet.

Fig. 6.4. Schematic diagram showing the competitive influence of undercooling (driving force) and
diffusivity on the heterogeneous nucleation rate.

Fraction transformed

logt

logt
Fig. 6.5. Fraction transformed as a function of time referred to the C-curve (schematic).
(6-5)
At the chosen reference temperature Tr the time difference between the real C-curve and
the extension of the lower asymptote amounts to (see Fig. 6.6):
(6-6)
from which
(6-7)
It follows that equations (6-5) and (6-7) provide a systematic basis for obtaining quantitative information about Qd and AGhet from experimental microstructure data through a simple
graphical analysis of the shape and position of the C-curve in temperature-time space.

1/r

lnt
Fig. 6.6. Determination of AG*het and Qd from the C-curve (schematic).

6.2.2.3 Mathematical description of the C-curve


In order to obtain a full mathematical description of the C-curve, we need to know the variation in the energy barrier AG*het with undercooling AT. For heterogeneous nucleation of precipitates above the metastable solvus, the strain energy term entering the expression for AG*het
can usually be ignored. In such cases the energy barrier is simply given as:1
(6-8)

where TV4 is the Avogadro constant, ^ n is the interfacial energy per unit area between the nucleus and the matrix, AGV is the driving force for the precipitation reaction (i.e. the volume free
energy change associated with the transformation), and 5(0) is the so-called shape factor which
takes into account the wetting conditions at the nucleus/substrate interface.
For a particular alloy, AGV is for small Ar proportional to the degree of undercooling:l
(6-9)
where C3 is a kinetic constant. This equation follows from the definition of AGv in diluted alloy
systems and the mathematical expression for the solvus boundary in the binary phase diagram.
By substituting equation (6-9) into equation (6-8), we get:
(6-10)
It follows that A0 is a characteristic material constant which is related to the potency of the
heterogeneous nucleation sites in the material. The value of A0 is, in turn, given by equations
(6-7) and (6-10):
(6-11)

In cases where A0 is known, it is possible to obtain a more general expression for t* by


substituting equation (6-10) into equation (6-3):

(6-12)

Equation (6-12) can further be modified to allow for compositional and structural variations in the parent material by using the calibration procedure outlined in Fig. 6.7. Let tr
denote the time taken to precipitate a certain fraction of (3 at a chosen reference temperature
T= Tr in an alloy containing Nv nucleation sites per unit volume. If we take the corresponding
solvus temperature of the (3-phase equal to T*q , the expression for t* becomes:

(6-13)

A combination of equations (6-12) and (6-13) then yields:

(6-14)

Equation (6-14) provides a basis for predicting the displacement of the C-curve in temperature-time space due to compositional or structural variations in the parent material. In genT

C-cun/e(Nv=N^)

logt
Fig. 6.7. Method for eliminating unknown kinetic constant in expression for t*.

eral, an increase in Nv will shift the nose of the C-curve to the left in the diagram (i.e. towards
shorter times), as shown schematically in Fig. 6.8, because of the resulting increase in the
nucleation rate. Moreover, in solute-depleted alloys the critical undercooling for nucleation
will be reached at lower absolute temperatures where the diffusion is slower. This results in a
marked drop in Nhet, which displaces the C-curve towards lower temperatures and longer times
in the IT-diagram, as indicated in Fig. 6.9.
Example (6.1)

Isothermal transformation (IT) or continuous cooling transformation (CCT) diagrams are available for many of the important alloy systems.4 In the case of aluminium, so-called temperature-property diagrams exist for different types of wrought alloys.45 Suppose that the C-curve
in Fig. 6.10 conforms to incipient precipitation of [3'(Mg2Si) particles at manganese-containing dispersoids in 6351 extrusions. Use this information to estimate the values of A0 and Qd in
equation (6-3) when the solvus temperature of (3'(Mg2Si) is 5200C.
Solution

The parameters A0 and Qd can be evaluated from the C-curve according to the procedure
shown in Fig. 6.6. Referring to Fig. 6.11, the value of AG^ at the chosen reference temperature Tr = 35OC (623K) is equal to:

When AGhet is known, the parameter A0 can be obtained from equation (6-11):
T

iogt
Fig. 6.8. Effect of Nv on the shape and position of C-curve in temperature-time space (schematic).

logt
\e,
Fig. 6.9. Effect of solute content on the shape and position of C-curve in temperature-time space (schematic).
%B

Similarly, Qd can be read from Fig. 6.11 by considering the slope of the lower asymptote:

This value is in good agreement with the reported activation energy for diffusion of magnesium in aluminium.6

Temperature, 0C

AA 6351 - T6

Time, s
Fig. 6.10. C-curve for 99.5% maximum yield strength of an AA6351-T6 extrusion. After Staley.5

103/T, K"1

Solvus temperature: 520 0C

lnt
Fig. 6.11. Determination of kG*heU and Qd from the C-curve in Fig. 6.10 (Example 6.1).

6.2.3 Growth of precipitates


If the embryo is larger than some critical size, it will grow by a transport mechanism which
involves diffusion of solute atoms from the bulk phase to the matrix/nucleus interface.
6.2.3.1 Interface-controlled growth
When transfer of atoms across the a/(3-interface becomes the rate-controlling step, the reaction is said to be interface-controlled. This growth mode is therefore observed during the
initial stage of a precipitation reaction before a large, solute-depleted zone has formed around
the particles. In the case of incoherent precipitates, the variation in the particle radius r with
time is given by:7
(6-15)

where M1 is a mobility term, C0 is the concentration of solute in matrix, Ca is the concentration


of solute at the particle/matrix interface, and Cp is the concentration of solute inside the precipitate.
In general, the mobility of incoherent interfaces is high, since the solute atoms can easily
'jump' across the interface and find a new position in the particle lattice, as shown schematically
in Fig. 6.12(a). In contrast, a coherent interface is essentially inmobile because transfer in this
case involves trapping of atoms in an intermediate lattice position, as indicated in Fig. 6.12(b).
As a result, semi-coherent precipitates are forced to grow by lateral movement of ledges along
a low energy interface in a direction where the matrix is incoherent with respect to the particle
lattice (see Fig. 6.13). In such cases the thickening rate of the precipitates U*aj^ is given by:3'7

Incoherent Interface

Coherent interface

(a)
(b)

a/p

Fig. 6J2. Schematic illustration of atom transfer across different kinds of interfaces; (a) Incoherent
interface, (b) Coherent interface.

Lateral movement of incoherenf


interface

Fig. 6.13. Thickening of plate-like precipitates by the ledge mechanism (schematic).

(6-16)

where M1* is a new mobility term, and / is the interledge spacing.


6.2.3.2 Diffusion-controlled growth
For growth of incoherent precipitates above the metastable solvus, the rate-controlling step
will be diffusion of solute in the matrix. If precipitation of the P-phase occurs from a

supersaturated a, the reaction proceeds by diffusion of solute to the growing p-particle, as


shown schematically in Fig. 6.14. On the other hand, when the (3-phase is formed by rejection
of solute from the a-phase, the transformation occurs by diffusion of atoms away from the Pparticle, as indicated in Fig. 6.15.
Aron et al.s have presented general solutions for diffusion-controlled growth of both flat
plates and spheres under such conditions. In the former case the half thickness AZ of the plate
is given by:
(6-17)
The parameter E1 in equation (6-17) is frequently referred to as the one-dimensional parabolic thickening constant, and is defined as:

Temperature

Liquid

Concentration

%B

Diffusion of solute

Distance
Fig. 6.14. Schematic representation of concentration profile ahead of advancing interface during precipitation of (B from a supersaturated a-phase.

Temperature
Concentration

%B

Diffusion of solute

Distance
Fig. 6.15. Schematic representation of concentration profile ahead of advancing interface during growth
of solute-depleted P into a metastable a-phase.

(6-18)

where Dm is the diffusivity of the solute in the matrix, and erfc(u) is the complementary error
function (defined previously in Appendix 1.3, Chapter 1).
Similarly, for growth of spherical precipitates, the variation in the radius r with time can be
written as:8
(6-19)
where e 2 *s t n e corresponding parabolic thickening constant for a spherical geometry, defined
as:

(6-20)

The parabolic relations in equations (6-17) and (6-19) imply that the growth rate slows
down as the (3-phase grows. This is due to the fact that the total amount of solute partitioned
during growth decreases with time when the diffusion distance increases. Moreover, the form
of equations (6-18) and (6-20) suggests that the maximum in the growth rate is achieved at an
intermediate temperature because of the competitive influence of undercooling (driving force)
and diffusivity on the reaction kinetics. Consequently, a plot of E1 or 2 vs temperature will
reveal a pattern similar to that shown in Fig. 6.4 for the nucleation rate, although the thickening constants generally are less temperature-sensitive.
In addition to the models presented above for plates and spheres, approximate solutions
also exist in the literature for thickening of needle-shaped precipitates, based on the Trivedi
theory for diffusion-controlled growth of parabolic cylinders.9 However, because of space limitations, these solutions will not be considered here.
6.2.4 Overall transformation kinetics
The progress of an isothermal phase transformation may be conveniently represented by an ITdiagram of the type shown in Fig. 6.5. Among the factors that determine the shape and position of the C-curve are the nucleation rate, the growth rate, the density and the distribution of
the nucleation sites as well as the physical impingement of adjacent transformed volumes.
Due to the lack of adequate kinetic models for diffusion-controlled precipitation, we shall
assume that the overall microstructural evolution with time can be described by an Avramitype of equation:10
(6-21)
where X is the fraction transformed, n is a time exponent, and k is a kinetic constant which
depends on the nucleation and growth rates.
The exponential growth law summarised in the Avrami equation is valid for linear growth
under most circumstances, and approximately valid for the early stages of diffusion-controlled
growth.10 Table 6.1 gives information about the value of the time exponent for different experimental conditions.
In general, the value of n will not be constant, but change due to transient effects until the
steady-state nucleation rate is reached and n attains its maximum value. Subsequently, the
nucleation rate starts to decrease as the sites become filled with nuclei and eventually approach zero when complete saturation occurs. This is because the heterogeneous nucleation
sites are not randomly distributed in the volume, but are concentrated near other nucleation
sites leading to an overall reduction in n. From then on, the transformation rate is solely
controlled by the growth rate.
6.2.4.1 Constant nucleation and growth rates
For a specific transformation reaction, the value of k in equation (6-21) can be estimated from

Table 6.1 Values of the time exponent n in the Avrami equation. After Christian.10
Polymorphic changes, discontinuous precipitation, eutectoid reactions, interface
controlled growth, etc.
Increasing nucleation rate
Constant nucleation rate
Decreasing nucleation rate
Zero nucleation rate (saturation of point sites)
Grain edge nucleation after saturation
Grain boundary nucleation after saturation
Diffusion controlled growth
All shapes growing from small dimensions, increasing nucleation rate
All shapes growing from small dimensions, constant nucleation rate
All shapes growing from small dimensions, decreasing nucleation rate
All shapes growing from small dimensions, zero nucleation rate
Growth of particles of appreciable initial volume
Needles and plates of finite long dimensions, small in comparison with their separation
Thickening of long cylinders (needles) (e.g. after complete end impingement)
Thickening of very large plates (e.g. after complete edge impingement)
Precipitation on dislocations (very early stages)

kinetic theory, using the classic models of nucleation and growth described in the previous
sections. In practice, however, this is a rather cumbersome method, particularly if the base
metal is of a heterogeneous chemical nature. Alternatively, we can calibrate the Avrami equation against experimental microstructure data, e.g. obtained from generic IT-diagrams. A convenient basis for such a calibration is to write equation (6-21) in a more general form:
(6-22)
where k* is a new kinetic constant (equal to kr1/n). In the latter equation the parameter k* can be
regarded as a time constant, which is characteristic of the system under consideration. Note
that this form of the Avrami equation is mathematically more appropriate, as the dimensions of
the k* constant are not influenced by the value of the time exponent n.
During the early stages of a transformation reaction, the reaction rate is controlled by the
nucleation rate. Let f denote the time taken to precipitate a certain fraction of P (X = Xc) at
an arbitrary temperature T (previously defined in equation (6-14)). It follows from equation
(6-22) that the value of k* in this case is given as:
(6-23)
A combination of equations (6-22) and (6-23) then gives:

(6-24)

from which
(6-25)
Equation (6-25) represents an alternative mathematical description of the Avrami equation,
and is valid as long as the nucleation and growth rates do not change during the transformation. It has therefore the following limiting values and characteristics: X=O when t = 0, X =
Xc when t = t*, and X>1 when r> <*>.
6.2.4.2 Site saturation
If the nucleation rate is considered to be zero by assuming early site saturation, the subsequent
phase transformation simply involves the reconstructive thickening of the p-layer. In the onedimensional case, the process can be modelled in terms of the normal migration of a planar
a/p interface, as shown schematically in Fig. 6.16. Let Aa/^ denote the interfacial area between a and (3 per unit volume and Ua/^ the growth rate of the incoherent a/p-interface. From
Fig. 6.16 we see that the volume fraction of the transformed (3-phase is given as:
(6-26)
By using the standard Johnson-Mehl correction for physical impingement of adjacent transformation volumes, we may write in the general case:
(6-27)
which after integration yields:
(6-28)
This specific form of the Avrami equation is valid under conditions of early site saturation
where the a/p-interface is completely covered by P nuclei at the onset of the transformation.
6.2.5 Non-isothermal transformations
So far, we have assumed that the phase transformations occur isothermally. This is, of course,
a rather unrealistic assumption in the case of welding where the temperature varies continuously with time. From the large volume of literature dealing with solid state transformations in

Fig. 6.16, Schematic illustration of the planar geometry assumed in the site saturation model.

metals and alloys, it appears that the bulk of the research has been concentrated on modelling
of microstructural changes under predominantly isothermal conditions.1"411 In contrast, only
a limited number of investigations has been directed towards non-isothermal transformations. 51012 " 18 However, these studies have clearly demonstrated the advantage of using analytical modelling techniques to describe the microstructural evolution during continuous cooling, instead of relying solely on empirical CCT-diagrams.
6.2.5.1 The principles of additivity
From the literature reviewed it appears that there is considerable confusion regarding the application of isothermal transformation theory for prediction of non-isothermal transformation
behaviour. These difficulties are mainly due to the independent variations of the nucleation
and growth rate with temperature. In fact, it can be shown on theoretical grounds that the
problem is only tractable when the instantaneous transformation rate is a unique function of
the fraction transformed and the temperature.10 This leads to the additivity criterion described
below.
The principles of additivity are based on the theory advanced by Scheil.12 He proposed that
the start of a transformation under non-isothermal conditions could be predicted by calculating
the consumption of fractional incubation time at each isothermal temperature, with the transformation starting when the sum is equal to unity. The Scheil theory has later been extended to
phase transformations to predict continuous cooling transformation kinetics from isothermal
microstructure data.10'17'18
Let t* again denote the time taken to precipitate a certain fraction of P (X - Xc) at an arbitrary temperature T. If the reaction is additive, the total time to reach Xc under continuous
cooling conditions is obtained by adding the fractions of time to reach this stage isothermally
until the sum is equal to unity. Noting that t* varies with temperature, we may write in the
general case:
(6-29)

A schematic illustration of the Scheil theory is contained in Fig. 6.17.


T

Cooling curve

Subdivision of time into


infinitesimal steps of isothermal heat treatments.

logt
Fig. 6.17. Schematic illustration of the Scheil theory.

6.2.5.2 Isokinetic reactions


The concept of an isokinetic reaction has previously been introduced in Section 4.4.2.3 (Chapter 4). A reaction is said to be isokinetic if the increments of transformation in infinitesimal
isothermal time steps are additive, according to equation (6-29). Christian10 defines this mathematically by stating that a reaction is isokinetic if the evolution equation for some state variable X may be written in the form:
(6-30)
where G(X) and H(T) are arbitrary functions of X and T, respectively.
For a given thermal history, T(t), this essentially means that the differential equation contains separable variables of X and T.
6.2.5.3 Additivity in relation to the Avrami equation
The concept of an isokinetic reaction can readily be applied to the Avrami equation. Differentiation of equation (6-22) with respect to time leads to the following expression for the rate of
transformation:
(6-31)

In a typical diffusion-controlled nucleation and growth process, the fraction transformed X


will not be independent of temperature, since the equilibrium volume fraction of the precipitates decreases with temperature (e.g. see equation (4-7) in Chapter 4). However, for dilute
alloys it is a fair approximation to neglect this variation as the solvus boundary becomes increasingly steeper and in the limiting case approaches that of a straight (vertical) line. Thus, if
n is constant and k* depends only on the transformation temperature, the reaction will be isokinetic in the general sense defined by Christian.10
Because of the independent variations of the nucleation and growth rate with temperature,
the transformation rate will not be a simple function of temperature. However, by considering
the form of the constitutive equations, it is obvious that the change in the nucleation rate with
temperature is far more significant than the corresponding fluctuations in the growth rate.
This point is more clearly illustrated in Fig. 6.18, which shows the temperature-dependency of
the nucleation and growth rates of grain boundary ferrite in a C-Mn steel. It is evident from
these data that the change in the parabolic thickening constant , is negligible compared with
the fluctuations in the nucleation rate. Consequently, in transformations that involve continuous cooling it is sufficient to allow for the variation of Nhet with temperature, provided that
site saturation has not been reached. Thus, if n is constant we can apply the Scheil theory
directly and rewrite equation (6-25) in an integral form:
(6-32)
In equation (6-32) Z1 represents the kinetic strength of the thermal cycle with respect to Pprecipitation. This parameter is generally defined by the integral:

Thickening constant Ce1),|ims"

-2 -1
Nucleation rate (N*het), cm s

Temperature, 0C
Fig. 6.18. Predicted variation in N*het and E1 with temperature during the austenite to ferrite transformation in a C-Mn steel (0.15 wt% C, 0.40 wt% Mn). Data from Umemoto et al.19

(6-33)

where dt is the time increment at T, and f is the corresponding hold time required to reach Xc
at the same temperature (given by equation (6-14)). The derivation of equation (6-32) is shown
in Appendix 6.2
The principles of additivity are also applicable under conditions of early site saturation. If
only U^p varies with temperature, it is possible to rewrite equation (6-28) in an integral form:

(6-34)

This equation can readily be integrated by numerical methods when the temperature-time
programme is known.
6.2.5.4 Non-additive reactions
If the additivity condition is not satisfied, it means that the fraction transformed is dependent
on the thermal path, and the differential equation has no general solution. This, in turn, implies
that the C-curve concept breaks down and cannot be applied to non-isothermal transforma-

Next Page

tions. Solution of the differential equation then requires stepwise integration in temperature-time space, using an appropriate numerical integration procedure. As already pointed
out, this will generally be the case for diffusion-controlled precipitation reactions, since the
evolution parameter X is a true function of temperature. Under such conditions, experimentally based continuous cooling transformation (CCT) diagrams must be employed.

6.3 High Strength Low-Alloy Steels


High-strength low-alloy steels are typically produced with a minimum yield strength in the
range 300-500 MPa, depending on the plate thickness.2021 During welding microstructural
changes take place both within the heat affected zone (HAZ) and the fusion region, which, in
turn, affect the mechanical integrity of the weldment.2122 In the HAZ, for instance, nitrides
and carbides coarsen and dissolve, and grain growth occurs to an extent that depends on the
distance from the fusion boundary and the exposure time characteristic of the welding process.
This can have a profound effect on the subsequent structure and properties of the weld by
displacing the CCT curve to longer times, thereby producing more Widmanstatten ferrite, or
increasing the possibility of bainite and martensitic transformation products on cooling. The
formation of such microstructures may reduce the toughness of the weld and increase the risk
of hydrogen cracking.2122
6.3.1 Classification of microstructures
It is appropriate to start this section with a detailed classification of the various microstructural
constituents commonly found in low-alloy steel weldments.
During the austenite to ferrite transformation, a large variety of microstructures can develop, depending on the cooling rate and the steel chemical composition. Normally, the microstructure formed within each single austenite grain after transformation will be a complex
mixture of two or more of the following constituents, arranged in approximately decreasing
order of transformation temperature:
(i)
(ii)
(iii)
(iv)
(v)
(vi)
(vii)

grain boundary (or allotriomorphic) ferrite (GF)


polygonal (or equiaxed) ferrite (PF)
Widmanstatten ferrite (WF)
acicular ferrite (AF)
upper bainite (UB)
lower bainite (LB)
martensite (M).

The microstructural constituents listed above are indicated in Fig. 6.19, which shows
photomicrographs of typical regions within low-alloy steel weldments.
6.3.2 Currently used nomenclature
Quantification of microstructures in steel welds is most commonly done by means of optical
microscopy. Several systems have been introduced throughout the years for the classification
of the various constituents, with each system reflecting different investigator's views and

Previous Page

tions. Solution of the differential equation then requires stepwise integration in temperature-time space, using an appropriate numerical integration procedure. As already pointed
out, this will generally be the case for diffusion-controlled precipitation reactions, since the
evolution parameter X is a true function of temperature. Under such conditions, experimentally based continuous cooling transformation (CCT) diagrams must be employed.

6.3 High Strength Low-Alloy Steels


High-strength low-alloy steels are typically produced with a minimum yield strength in the
range 300-500 MPa, depending on the plate thickness.2021 During welding microstructural
changes take place both within the heat affected zone (HAZ) and the fusion region, which, in
turn, affect the mechanical integrity of the weldment.2122 In the HAZ, for instance, nitrides
and carbides coarsen and dissolve, and grain growth occurs to an extent that depends on the
distance from the fusion boundary and the exposure time characteristic of the welding process.
This can have a profound effect on the subsequent structure and properties of the weld by
displacing the CCT curve to longer times, thereby producing more Widmanstatten ferrite, or
increasing the possibility of bainite and martensitic transformation products on cooling. The
formation of such microstructures may reduce the toughness of the weld and increase the risk
of hydrogen cracking.2122
6.3.1 Classification of microstructures
It is appropriate to start this section with a detailed classification of the various microstructural
constituents commonly found in low-alloy steel weldments.
During the austenite to ferrite transformation, a large variety of microstructures can develop, depending on the cooling rate and the steel chemical composition. Normally, the microstructure formed within each single austenite grain after transformation will be a complex
mixture of two or more of the following constituents, arranged in approximately decreasing
order of transformation temperature:
(i)
(ii)
(iii)
(iv)
(v)
(vi)
(vii)

grain boundary (or allotriomorphic) ferrite (GF)


polygonal (or equiaxed) ferrite (PF)
Widmanstatten ferrite (WF)
acicular ferrite (AF)
upper bainite (UB)
lower bainite (LB)
martensite (M).

The microstructural constituents listed above are indicated in Fig. 6.19, which shows
photomicrographs of typical regions within low-alloy steel weldments.
6.3.2 Currently used nomenclature
Quantification of microstructures in steel welds is most commonly done by means of optical
microscopy. Several systems have been introduced throughout the years for the classification
of the various constituents, with each system reflecting different investigator's views and

(a)

(b)

(C)

(d)

Fig. 6.19. Optical micrographs showing various microstructural constituents commonly found in lowalloy steel weldments; (a) Coarse grained HAZ (low heat input welding), (b) Coarse grained HAZ (high
heat input welding), (c) As-deposited weld metal (low heat input welding), (d) Reheated weld metal (low
heat input welding). Letters in micrographs are defined in the text.

discretions. This controversy in terminology has been a source of confusion, and the work by
Sub-Commission IXJ of the International Institute of Welding (HW)23 for developing guidelines for quantification of microstructures is, therefore, an important step towards a standardised system of nomenclature.
The IIW recommendations are based on the scheme originally proposed by Abson and
Dolby.23 The IIW system involves a simplified classification procedure compared with the
outline used in Fig. 6.19, since the distinction between acicular ferrite and the various sideplate
structures is based on features such as aspect ratio, relative lath size, and number of parallel

laths. This has led to the introduction of the FS-constituent (ferrite with aligned second phase),
which, in principle, comprises both Widmanstatten ferrite and upper bainite.
In contrast to the HW approach to classifying microstructural elements based on their appearance in the optical microscope, other investigators rank the various constituents solely in
terms of their transformation behaviour, according to the scheme originally proposed by Dube
et ai26 From a scientific point of view, this classification system is more correct, since it does
not violate common terminology based on thermodynamics and kinetics of transformation
reactions. However, with the omission of the FS-constituent grouping utilised by the HW, the
Dube system is more inconvenient to use in practice because the different transformation products often cannot readily be identified on the basis of their transformation characteristics. Consequently, both classification systems appear to have their weaknesses, which, in turn, limit
their applicability.
6.3.3 Grain boundary ferrite
Grain boundary (or allotriomorphic) ferrite is the first phase to form on cooling below the Ae3temperature. It nucleates preferentially at austenite grain corners and boundaries, since these
sites generally provide the lowest energy barrier against nucleation (see Fig. 6.3). The fundamental aspects of grain boundary ferrite have been reviewed in detail by Bhadeshia,26 where
many of the original references can also be found.
6.3.3.1 Crystallography of grain boundary ferrite
The grain boundary ferrite allotriomorphs nucleate having a Kurdjumow-Sachs type orientation relationship with one of the austenite grains:20

This orientation relationship, which lies within the so-called Bain orientation region,27 is
adopted in order to minimise the increase in the strain energy resulting from lattice distortion
AGEby formation of a low-energy interface between the ferrite nucleus and the parent austenite
phase.1 Subsequent growth of the ferrite may then occur into the adjacent austenite grain with
which the ferrite has a random orientation relationship,28 since a disordered (incoherent) interface generally has a higher mobility than an ordered (coherent/semi-coherent) interface at low
undercoolings.
6.3.3.2 Nucleation of grain boundary ferrite
As shown in Fig. 6.20, allotriomorphic ferrite can nucleate both at grain corners, grain edges
and grain faces, the former ones being the most potent sites for ferrite nucleation.1 If we
assume that ferrite nucleation occurs preferentially at austenite grain faces, the A^_/ Nv term
in equation (6-14) may be replaced by the inverse grain size ratio Dy /Dy, where Dy and Dy
refer to the austenite grain size in the actual and the reference material, respectively. This
leads to the following expression for t*:
(6-35)

Edge

Corner

Fig. 6,20. Sketch of an austenite grain showing different sites for ferrite nucleation.

Based on equation (6-35) it is possible to predict the displacement of the C-curve in temperature-time space due to structural or compositional variations in the parent material. As an
illustration, we shall assume that the parameters listed in Table 6.2 are representative of nucleation of grain boundary ferrite in a low-alloy steel with an initial austenite grain size of 10
jim. In addition, we need information about the Ae3-temperature in the equilibrium phase
diagram. This temperature can readily be obtained from thermodynamic calculations, even for
multicomponent systems.29 Alternatively, we can use the empirical relationship quoted by
Leslie:30

(6-36)

where all compositions are given in weight %.


By lowering the A<?3-temperature, the austenite stabilising elements C, Mn, Ni and Cu reduce the undercooling and hence, the driving force for transformation of austenite, at any
lower temperature. On the same basis, the ferrite stabilising elements will assist the transformation because they increase the undercooling. In particular, the very large ferrite stabilising
effects of P, Al and Ti are obvious from the above relationship.
Effect of austenite grain size
A closer inspection of equation (6-35) reveals that the position of the C-curve shifts by a factor
of (Dy I Dy) when the austenite grain size changes from D1 to Dy. This is due to the fact
that the austenite grain boundary area per unit volume is inversely proportional to the grain
size. Consequently, grain growth which occurs during welding will have a profound effect on
the subsequent transformation behaviour of the weld. An illustration of this point is given
below.
Table 6.2 Input data used to construct C-curve for allotriomorphic ferrite in reference steel.

Parameter I T
Value

(K)

(s)

T*eq I
(K)

T0
(Jmol"1)

823

0.6

1108

700

f Activation energy for diffusion of carbon in austenite.

Q]

I Dy

(kJmol-1)

(jim)

135

10

Example (6.2)

Consider SA welding on a thick plate of a Nb-microalloyed steel under the following conditions:

Evaluate on the basis of the grain growth diagram in Fig. 5.21 (a) (Chapter 5) and equations
(6-32), (6-33), and (6-35) the conditions for ferrite formation at two different positions within
the HAZ corresponding to a peak temperature Tpof 13500C and 10000C, respectively. Assume in these calculations that the ferrite may form within the temperature range from 800 to
6000C, and that the equilibrium volume fraction of ferrite (f^qFe) in the fully transformed
steel is 0.9.
Solution

First we calculate the net heat input per unit length of the weld:

Readings from Fig. 5.21 (a) give the following HAZ grain sizes:

By substituting data from Table 6.2 into equation (6-35), we obtain:


Grain refined region ( Dy - 1 OfJm):

Grain coarsened region ( Dy = 150 jlm):

As expected, the theoretical C-curves in Fig. 6.21 reveal a strong effect of the austenite
grain size on the HAZ transformation kinetics. By considering the superimposed weld cooling curve, it is possible to estimate the volume fraction of ferrite/""7^ which forms in each case
from equations (6-32) and (6-33). Taking the time exponent n in the Avrami equation equal
to 5/2 for nucleation of ferrite at austenite grain boundaries10 and 1 Xc = 0.98, we obtain:
Grain refined region (Dy = 10 jlm):

Temperature, 0C

Time, s
Fig. 6.21. Effect of austenite grain size on the HAZ transformation kinetics (Example 6.2). The superimposed cooling curve corresponds to a cooling time, A%5, of 21s.
and

Grain coarsened region ( Dy = 150 jim):

and

From this we see that polygonal ferrite dominates the microstructure within the grain refined region, whereas ferrite hardly forms within the grain coarsened HAZ under the prevailing circumstances. Although experimental data are not available for a direct comparison, the
predicted effect of the austenite grain size on the HAZ transformation kinetics is reasonable
and consistent with general experience (e.g. see experimental CCT-diagrams in Fig. 6.22).
Effect of austenite and ferrite stabilising elements
In practice, the transformation behaviour of steel weldments is affected both by the prior
austenite grain size and by alloying additions. In welding metallurgy, the combined effects are
often discussed by considering their influence on hardenability. Broadly speaking, additions of
hardenability elements may serve two purposes:
(i)

To ensure the desired strength level by solid solution or precipitation strengthening;

(ii)

To control the microstructure through modification of the nucleation and growth rates of
proeutectoid ferrite.

Temperature, 0C

Time, s
Fig. 6.22. CCT-diagrams for a low-carbon Cu-Ni containing steel. Superimposed on the CCT-diagrams
are two cooling curves corresponding to Af875 of 10 and 100 s, respectively. Austenitising conditions;
Heavy solid lines: 9000C for 5 min, Heavy broken lines: 13000C for 5 s. Data from Cross et al?x

In the latter case the effect is related to a shift in the Ae3-temperature of the steel, which
alters the undercooling and hence, the driving force for the austenite to ferrite transformation.
This point is illustrated by the following example.
Example (6.3)

Consider SA welding on a thick plate of a Nb-microalloyed steel under conditions similar to


those employed in Example (6.2). Based on equations (6-32), (6-33), and (6-35) estimate the
volume fraction of grain boundary ferrite in the grain coarsened HAZ (Tp = 13500C) after
welding when the A^-temperature of the steel is 863C (1136K).
Solution

By substituting data from Table 6.2 into equation (6-35), we arrive at the following expression
for r *50:

It is evident from the graphical representation of the above equation in Fig. 6.23 that an increase in the v4e3-temperature (e.g. from 1108 to 1136K) displaces the C-curve towards higher
temperatures and shorter times in the IT-diagram. This, in turn, gives rise to improved conditions for ferrite nucleation. Taking the time exponent n in the Avrami equation equal to 5/2
and (1 Xc) = 0.98 as in the previous example, we obtain after integration:

Temperature, 0C

Time, s
Fig. 6.23.Effect of steel chemical composition (Ae3-temperature) on the HAZ transformation kinetics
(Example 6.3). The superimposed cooling curve corresponds to a cooling time, Ar875, of 21s.

and

Although the calculated value offa'Fe is rather uncertain, the trends predicted in the present
example are reasonable and consistent with general experience.

Effect of boron alloying


The role of boron in steel transformation kinetics has been a subject of research for many
years. A number of hardenability mechanisms have been proposed to explain the behaviour of
boron in steels. Of these, only four have survived to the present.32 All assume that boron
influences hardenability by increasing the energy barrier against ferrite nucleation at austenite
grain boundaries* and that it does not influence the thermodynamic properties of the austenite
and ferrite phases, i.e. reduces the A<?5-temperature of the steel.32 Moreover, it is now generally accepted that impurity-boron interactions are responsible for most of the observed erratic
behaviour of boron steels with respect to hardenability.32"34 In particular, if the steel contains
small amounts of dissolved oxygen and nitrogen, these elements may combine with boron to
reduce the free, diffusible boron content.32 The hardenability effect of boron is therefore sensitive to variations in the applied steel deoxidation and alloying practice.32"37
*Based on classic nucleation theory it can be argued that a decrease in the y-y grain boundary energy due to boron
segregations will suppress the formation of ferrite at these sites. Further enhancement of the energy barrier against
ferrite nucleation through an increase in the total strain energy of the embryo is possible if the free grain boundary
volume becomes filled with either boron atoms or borocarbide precipitates (e.g. Fe23(B9C)6).

During welding, quantitative information about the extent of boron segregation which occurs to the austenite grain boundaries under various thermal programmes can be obtained on
the basis of a well established theoretical model for quench-induced segregation of boron in
steel.3839 At peak temperatures above 1000 to 11000C, borocarbides and -nitrides present in
the base plate will rapidly dissolve in the matrix,4041 leading to a significant increase in the
amount of free diffusible boron. Generally, solute atoms in a crystal lattice will have an associated strain energy,38 which implies that it is energetically feasible to pair the solute boron
atom with a vacancy. Since the formation of vacancies is a thermally activated process, it follows that the fraction of boron occupying such sites, [B]v, increases exponentially with temperature:39
(6-37)

Here m contains various geometric and entropy terms, Zy is the vacancy formation energy,
Eb is the vacancy-boron binding energy, and [B] is the bulk concentration of free boron. By
substituting reasonable average values for ra, Ep and Eb into equation (6-37), we arrive at the
following expression for [B]V:39A2
(6-38)

During cooling, the vacancy concentration initially established at elevated temperatures


tends to readjust by elimination of excess point defects at grain boundaries through diffusion.
This, in turn, results in an associated flow of solute boron to the austenite grain boundaries (the
extent of which is controlled by the peak temperature and the bulk boron concentration), provided that the diffusivity of the boron-vacancy complexes is higher than that of the vacancies
and the boron atoms at all relevant temperatures.
Consider now the limiting case where the boron-vacancy complex diffusion to the grain
boundaries occurs sufficiently rapid to keep pace with the falling temperature; i.e. the equilibrium concentration, [B]v, is maintained from the peak temperature Tp down to the start temperature of the austenite to ferrite transformation Ae3. In view of the high diffusivity of boronvacancy complexes in austenite, this is not an unrealistic assumption when the cooling rate is
of the order of 500C s"1 (representative of a weld HAZ).39 Under such conditions, the amount
of boron which segregates to the austenite grain boundaries within the temperature interval
from Tp to Ae3 on cooling, [B]gb, can approximately be written as:42

when Tp Ae3.
In addition to the diffusible boron content, the grain size is also an important variable in
steel hardenability. A quantitative estimate of the combined effect of boron segregations and

austenite grain size on the HAZ transformation kinetics can be obtained by assuming that the
ferrite nucleates primarily on grain faces. In such cases the total number of heterogeneous
nucleation sites per unit volume Nvis given as:
(6-40)
where na is the number of nucleation sites per unit grain boundary area, and Sv is the grain
boundary surface area per unit volume (equal to 2/Dy).
If we, as a second approximation, assume that na is inversely proportional to the amount of
boron which diffuses to the grain boundaries on cooling, the N* / Nv ratio in equation (6-14)
can be written as:
(6-41)
Equation (6-41) predicts that the position of the ferrite C-curve in temperature-time space
depends on the l[B]gb Z)7) I l[B]*gb Z>y J ratio, as shown schematically in Fig. 6.24. Consequently,
this ratio can be regarded as a measure of the HAZ hardenability during welding of boroncontaining steels.
In Fig. 6.25 the microstructure data of Akselsen et al.42 have been replotted vs the
hardenability parameter [B]gbDy, taking the product [B]*gbDy in the reference steel equal to
unity for a direct comparison between theory and experiments. It is evident from the graph
that the HAZ martensite content of the two boron-containing steels can be represented by one
single curve under the prevailing circumstances. This result is to be expected if the displacement of the ferrite C-curve in temperature-time space is determined by a relationship of the
type shown in equation (6-41).

Temperature

High values of [B]

Low values of [Bl D


gb Y

Cooling curve

log time
Fig. 6.24. Effect of boron alloying on the shape and position of ferrite C-curve in temperature-time space
according to the site blocking mechanism (schematic).

Martensite, vol%

Steel A (11 ppm B)


Steel B (26 ppm B)

Hardenability parameter ([B] D )


Fig. 6.25. Relation between the volume fraction of martensite and the hardenability parameter
[B]gb Dy for simulated weld HAZs ([B]gb in ppm, Dy in urn, [B]gbDy = 1). Data from Akselsene#/.42
Example (6.4)

Consider GMA welding on a thick plate of a boron-containing steel under the following conditions:
/=300A, U = 30V, v = 4mm s-1, TI = 0.8, T0 = 200C
Suppose that the free (diffusible) boron content of the steel at elevated temperatures is 40
ppm. Estimate on the basis of the theory outlined in the previous sections the conditions for
ferrite/martensite formation in the grain refined HAZ (Tp = 11000C) when the austenite grain
size is 15|Lim. In these calculations we shall assume that the [B]*gb Dy product in the reference
steel (with thermodynamic properties as in Table 6.2) is close to unity.
Solution

First we estimate the amount of boron which segregates to the austenite grain boundaries
during cooling from equation (6-39). When Tp = 11000C (1373K), we get:

This gives the following value of the Af* / Nv ratio (equation 6-41):

The resulting displacement of the ferrite C-curve can now be calculated from equation (614), using input data from Table 6.2:

As expected, the theoretical C-curves in Fig. 6.26 reveal a strong effect of boron alloying
on the HAZ transformation kinetics. By considering the superimposed weld cooling curve, it
is possible to estimate the volume fraction of ferrite f^~Fe which forms from equations (6-32)
and (6-33). Taking the time exponent n in the Avrami equation equal to 5/2 for nucleation of
ferrite at austenite grain boundaries and (1 - Xc) = 0.98 as in the previous examples, we obtain after integration within the temperature range from 800 to 600 C:

and

If the same calculations are performed for the reference steel in Table 6.2 (characterised by

[*]gb = iBfgb )> we get:

and
From this we see that boron, even in small quantities, can have a dramatic effect on the
HAZ transformation kinetics by promoting the formation of bainite and martensite at the expense of grain boundary ferrite. This is in good agreement with general experience (see experimental CCT-diagrams in Fig. 6.27).
Effect of solidification-induced segregation
As shown in Chapter 3 of this textbook, the characteristic solidification pattern of mild and
low-alloy steel weld metals leads to extensive segregation of alloying and impurity elements
to grain boundaries and interdendritic spaces. Of particular interest in this respect is phosphorus segregations at columnar austenite grain boundaries. Referring to Figs. 3.38 and 3.39
(Chapter 3), the phosphorus-rich zone is seen to extend about 7jam on either side of the grain
boundaries. Since phosphorus is among the strongest ferrite stabilising elements in steel (see
equation (6-36)), the existence of such solidification-induced segregations would be expected

Temperature, 0C
Time, s
Fig. 6.26. Predicted displacement of ferrite C-curve in temperature-time space due to segregation of
boron to prior austenite grain boundaries (Example 6.4). The superimposed cooling curve corresponds
to a cooling time, Ats/5, of 9.2s.
to enhance the nucleation rate of grain boundary ferrite at these sites due to the associated
increase in the Ae3-temperature. The above phenomenon should not be confused with equilibrium segregation of phosphorus to austenite grain boundaries during heat treatment of steel,
which stems form attraction of the atoms towards the open structure of the boundary. In the
latter case phosphorus may act as a hardenability element by occupying favourable sites for
ferrite nucleation at the austenite grain boundaries analogous to that documented for boron in
steel.43'44
Example (6.5)

Suppose that the local phosphorus content adjacent to the columnar austenite grain boundaries
in a low-alloy steel weld metal is 500 ppm, whereas the bulk concentration of phosphorus is
100 ppm. Estimate on the basis of the theory developed in the previous sections the resulting
displacement of the ferrite C-curve in temperature-time space when the columnar austenite
grain size is 80|jun. In these calculations we shall assume that the transformation characteristic
of the bulk metal is similar to that of the reference steel in Table 6.2.
Solution

First we estimate the actual Ae3-temperature within the phosphorus-rich region adjacent to the
columnar austenite grain boundaries from equation (6-36). Taking the A<?3-temperature of the
bulk phase equal to 835C (1108K), the local phase boundary temperature becomes:

By substituting data from Table 6.2 into equation (6-35), we arrive at the following expression for ^ 0 :

(a)

Temperature, 0C

Steel A (11 ppmB)

Cooling time, At0,., s


(b)

Temperature, 0C

Steel B (26 ppm B)

Cooling time, A t 8/5 , s


Fig. 6.27. CCT-diagrams for boron-containing steels; (a) Steel A (llppm B), (b) Steel B (26ppm B).
Data from Akselsen et al.42

It is evident from the graphical representation of the above equation in Fig. 6.28 that the
observed increase in the yl^-temperature from 1108 to 1136K displaces the ferrite C-curve
towards higher temperatures and shorter times in the IT-diagram. The resulting effect on the
weld metal transformation kinetics is obvious, since an increase in the nucleation rate of ferrite will favour early site saturation at the austenite grain boundaries. On this basis it is not
surprising to find that allotriomorphic ferrite in low-alloy steel weld metals tends to form continuous veins of blocky ferrite along the columnar austenite grain boundaries, as shown in Fig.
6.19(c). Outside the fusion zone the conditions for early site saturation are less favourable,
since modern steelmaking practice implies that solidification-induced segregations are removed by prolonged high-temperature annealing prior to the welding operation. Hence, ferrite veining of the type shown in Fig. 6.19(c) is not commonly observed within the reheated
regions of the base plate, unless the heat input is extremely large (see Fig. 6.19(a) and (b)).

Temperature, 0C

Time, s
Fig. 6.28. Effect of solidification-induced phosphorus segregations on the austenite to ferrite transformation in low-alloy steel weld metals (Example 6.5).

Effect of cooling rate on ferrite grain size


In steel metallurgy it is well accepted that accelerated cooling refines the ferrite grain size, and
thus improves both strength and toughness of the parent material. 1920 If the austenite is allowed to recrystallise before it transforms, the final ferrite grain size Da in the base plate will
be an unique function of the prior austenite grain size D1 and the cooling rate through the
critical transformation temperature range for ferrite formation (CR.).
The fundamental aspects of ferrite grain refinement by accelerated cooling have been considered by Umemoto et al.19 By allowing for the variation in the nucleation and growth rates
of allotriomorphic ferrite with temperature (see Fig. 6.18), they arrived at the following theoretical expression for Da when austenite grain faces are the dominant nucleation site of ferrite:
(6-42)
where D1 is in |im and (CR.) is in 0C s"1.
In practice, Umemoto et al.19 observed a discrepancy between theory and experiments due
to competitive nucleation of ferrite at austenite grain edges and corners (characterised by a
theoretical grain size dependence of D12/3 and D1, respectively). Consequently, the real ferrite grain size in the steel varied with austenite grain size and cooling rate as:
(6-43)
where C4 is a kinetic constant which is characteristic of the alloy system under consideration.
An indication of the applicability of equation (6-43) to welding can be obtained from the
micro structure data of Evans,45 reproduced in Fig. 6.29. In this plot, the reported heat inputs
have been converted into an equivalent cooling rate at 7000C via equation (1-71) (Chapter 1).
It is evident from the figure that the observed variation in the ferrite grain size with cooling
rate is consistent with calculations based on equation (6-43). Hence, the model of Umemoto et

Ferrite grain size,jim

Equation (6-43)

Weld A (0.04 wt% C, 0.58 wt% Mn)


Weld C (0.05 wt% C, 1.33 wt% Mn)

Cooling rate at 700 0C1 C/s


Fig. 6.29. Variation of ferrite grain size with cooling rate in reheated C-Mn steel weld metals. Data from
Evans.45
al19 can readily be employed for prediction of the ferrite grain size within the grain refined
region of both single pass and multipass steel welds, provided that the austenite grain size and
the cooling rate can be estimated with a reasonable degree of accuracy.
Example (6.6)

Consider multipass welding with covered electrodes (SMAW) on a thick plate of low-alloy
steel under the following conditions:

Previous experience has shown that polygonal ferrite forms within the low-temperature
reheated regions of the weld metal, typically 1.0 to 1.7 mm beneath the surface (fusion boundary) of subsequent weld passes. Estimate on the basis of the grain growth diagram in Fig.
5.24(a) (Chapter 5) and equation (6-43) the maximum variation in the ferrite grain size across
the weld HAZ under the prevailing circumstances. Thermal data for low-alloy steels are given
in Table 1.1 (Chapter 1).
Solution

First we need to convert the depths at which polygonal ferrite appears beneath the fusion
boundary to an equivalent (characteristic) peak temperature range. If we neglect the contribution from heat flow in the welding direction, this conversion can be done on the basis of
equation (5-47) in Chapter 5:

Taking Ar*m equal to 1 and 1.7mm, respectively, we get:

and

The prior austenite grain size Dy at these two locations can now be read from Fig. 5.24(a)
and Table 5.6, respectively:

Since the cooling rate (CR.) at a given temperature is essentially the same across the weld
HAZ, the maximum variation in the ferrite grain size can be evaluated directly from equation
(6-43) without further background information:

From this we see that the variation in the ferrite grain size is significantly smaller than the
corresponding change in the prior austenite grain size. This result is in good agreement with
general experience.
6.3.3.3 Growth of grain boundary ferrite
If the austenite grain boundaries become rapidly decorated with a continuous layer of ferrite
(so that the subsequent transformation involves the reconstructive thickening of these layers),
the evolution of allotriomorphic ferrite is determined solely by its growth kinetics.37 As shown
in Example 6.5, this is a realistic assumption in the case of as-deposited steel weld metals,
where the presence of solidification-induced phosphorus segregations at the columnar austenite
grain boundaries favours an early site saturation.
Several investigators have modelled the evolution of allotriomorphic ferrite in low-alloy
steel weld metals along the lines indicated above.46"48 The most thorough analysis is probably
that of Bhadeshia et al.46 who were able to account for the combined effect of temperature and
steel chemical composition on the growth kinetics. However, in order to illustrate the competition between the different variables that contribute to the formation of grain boundary ferrite
in as-deposited steel weld metals, the simplified treatment of Liu and Olson47 and Fleck et al.4S
has been adopted here.
As a starting point the Avrami equation17 is considered for the limiting case of zero nucleation rate by assuming early site saturation. If subsequent growth of the grain boundary ferrite
allotriomorphs occurs in both directions perpendicular to the austenite grain boundaries at a
rate which is controlled by diffusion of carbon in austenite, the time dependence of the ferrite
thickness, AZa, is given by equation (6-17):
(6-44)

where E1 is the one-dimensional parabolic thickening constant (defined in equation (6-18)).


The corresponding growth rate of allotriomorphic ferrite Ua is then obtained by differentiating equation (6-44) with respect to time:
(6-45)
From a stereological standpoint, the grain boundary surface area per unit volume Sv cannot
be calculated without further assumptions regarding the shape of the columnar austenite grains.
Bhadeshia et al.31*46*49 solved this problem by representing the grain morphology by a uniform,
space-filling array of hexagonal prisms. However, with the precision aimed at here, it is sufficient to assume that Sv is equal to the surface area per unit volume of an inscribed cylinder
whose volume is equivalent to that of the hexagonal prisms (see Fig. 6.30). Noting that each
grain boundary is shared by two adjacent grains, the expression for Sv (in the absence of end
effects) becomes:47'48
(6-46)

where dy and I1 are the diameter and length of the inscribed cylinder, respectively.
By substituting the above expressions for Ua and Sv into equation (6-34), we obtain:

(6-47)

In practice, the one-dimensional parabolic thickening constant in equation (6-47) varies


both with temperature and steel chemical composition, as shown by the data of Bhadeshia et
aL46 reproduced in Fig. 6.31. However, if we instead use a reasonable average value for E1
within the characteristic transformation temperature range for allotriomorphic ferrite (i.e.
E1 = E1), the integral I2 in equation (6-47) has the following analytical solution:
x Growth direction of columnar grains

Fig. 6.30. Hexagonal prism model for the columnar austenite grain morphology in low-alloy steel weld
metals. The inscribed cylinder in the figure has approximately the same surface to volume ratio as the
hexagonal prisms.

Parabolic thickening constant (E1), ^m/s1/2

Temperature, 0C
Fig. 6.31. Effect of temperature and steel chemical composition on the one-dimensional parabolic thickening constant for allotriomorphic ferrite. (1): 0.03 wt% C, (2): 0.06 wt% C, (3): 0.08 wt% C and (4):
0.10 wt% C. Data from Bhadeshia et al.46

(6-48)
from which
(6-49)

Equation (6-49) predicts that the volume fraction of grain boundary ferrite in the as-deposited weld metal depends on the combined action of the following three main variables:
(i) The one-dimensional parabolic thickening constant ei which is determined by the weld
metal chemical composition (i.e. the content of austenite and ferrite stabilising elements).
(ii) The columnar austenite grain size dy which is controlled by the weld metal solidification
microstructure (i.e. the weld metal chemistry, the weld pool geometry, and the thermal
conditions under which solidification occurs).
(iii) The retention time within the critical transformation temperature range for allotriomorphic
ferrite, as determined by the applied heat input and the mode of heat flow (i.e. thick plate,
medium thick plate, or thin plate welding, respectively).

It can be seen from the microstructure data of Grong et al.50 reproduced in Fig. 6.32 that the
influence of these variables are adequately accounted for in the present model. However, the
calculated volume fractions of grain boundary ferrite are consistently lower than the measured
ones. This discrepancy can probably be attributed to the use of constant values for the onedimensional parabolic thickening constant. Consequently, if proper corrections are made for
the inherent variation in E1 with temperature and steel chemical composition, the agreement
between theory and experiments is significantly improved, as shown by the data of Bhadeshia
et al.46 reproduced in Fig. 6.33.
Example (6.7)

Consider GMA welding on a thick plate of a low-alloy steel under the following conditions:

Suppose that the grain boundary allotriomorphs form within the temperature range from
750 to 6000C at a constant rate ei of 3 jam s~1/2. Estimate on the basis of the theory developed
in the previous section the volume fraction of allotriomorphic ferrite in the weld deposit when
the columnar austenite grain size is 80|Lim. In these calculations we shall assume that the
equilibrium volume fraction offerritefq~Fe in the fully transformed steel is 0.9. Thermal data
for low-alloy steels are given in Table 1.1 (Chapter 1).
Solution

The situation is described in Fig. 6.34. In this case the problem is to estimate the retention time
/Str within the critical temperature range for ferrite formation from the Rosenthal thick plate
solution (Chapter 1). From equation (1-66), we have:

Volume fraction of GF

Regression line

C-Mn steel weld metals


Si content: 0.50-1.03 wt%
Mn content: 0.33-2.72 wt%

Fig. 6.32. Experimental verification of equation (6-49). Data from Grong et al.5Q

Measured volume fraction of GF

C-Mn steel weld metals

Line of unit slope

Calculated volume
fraction of GF

Temperature

Fig. 6.33. Comparison between measured and predicted volume fractions of grain boundary ferrite in
C-Mn steel weld metals. Data from Bhadeshia et al.46

log time
Fig. 6.34. Conditions for allotriomorphic ferrite formation in low-alloy steel weld metals (Example 6.6).

By inserting this value into equation (6-49), we obtain:

Although experimental data are not available for a direct comparison, the calculated value
of fa'Fe is reasonable and of the expected order of magnitude.
6.3.4 Widmanstdtten ferrite
With increasing degree of undercooling the redistribution of carbon becomes insufficient to
maintain a planar growth mode, and hence, further growth of the ferrite can only take place by
lateral movement of ledges along a low-energy interface*.120 This, in turn, implies a
Kurdjumow-Sachs-type orientation relationship between the austenite and the ferrite, i.e.
{Ill }1_Fe parallel with {110}a_F6, and <110> 7 .^ parallel with <11 l>a_Fe, which is a characteristic feature of the Widmanstatten ferrite structure. The ferrite sideplates, once nucleated,
grow very rapidly under the prevailing conditions because of an efficient redistribution of
carbon to the sides of the advancing interface.1 As a result, parallel arrays of ferrite laths of
high aspect ratios (typically 10:1 to 20:1) are often found in the areas adjacent to the austenite
grain boundaries, as shown by the optical micrographs in Fig. 6.19(b) and (c).
Following the treatment of Bhadeshia et al.,46 the lengthening rate of Widmanstatten ferrite
can be estimated using the Trivedi theory9 for diffusion-controlled growth of parabolic cylinders with correction for the assumed displacive character of the Widmanstatten ferrite transformation. As shown in Fig. 6.35, the pertinent growth rates are normally so large that the
formation of Widmanstatten ferrite is essentially complete within a fraction of a second. This
implies that the transformation, for all practical purposes, can be treated as being isothermal.37
When the growth rate is known, the volume fraction of Widmanstatten ferrite in the asdeposited weld metal may be estimated by assuming that the ferrite sideplates nucleate at a
constant rate at the yFelaFe boundaries and subsequently grow into the interior of the columnar austenite grains until they physically impinge with intragranularly nucleated acicular ferrite or allotriomorphic ferrite.46 It is seen from the microstructure data of Bhadeshia et al.46
reproduced in Fig. 6.36 that the calculated volume fractions of Widmanstatten ferrite are in
reasonable agreement with experiments, although the discrepancy in certain cases is admittedly large.
*A different view is suggested by Bhadeshia et al.31'46'51'52 who claim that growth of Widmanstatten ferrite occurs in
a displacive manner analogous to that documented for martensite in steel, with the exception that carbon must diffuse
during growth.

Next Page

Growth rate of WF, u,m/s

Temperature, 0C
Fig. 6.35. Predicted growth rates of Widmanstatten ferrite in C-Mn steel weld metals; (a) Growth rate
calculations for weld compositions listed in Fig. 6.31, (b) Growth rate calculations after modifying compositions to allow for carbon enrichment due to grain boundary ferrite formation. Data from Bhadeshia

etai46

6.3.5 Acicular ferrite in steel weld deposits


Simultaneously with or immediately after the formation of Widmanstatten ferrite at the austenite
grain boundaries, acicular ferrite may start to nucleate intragranularly at non-metallic inclusions. This phase is commonly observed in low-alloy steel weld metals, where the fine dispersion of oxide inclusions provides favourable sites for heterogeneous nucleation.3653 There
seems to be general agreement that microstructures primarily consisting of acicular ferrite
provide optimum weld metal mechanical properties, both from a strength and toughness point
of view, by virtue of its small lath size and high dislocation density. 365354 Consequently, the
formation of this particular microconstituent is of significant commercial importance and has
therefore attracted substantial research interest over the years. 363746 " 58 In spite of all this effort, the acicular ferrite transformation in low-alloy steel weld metals is still a subject of considerable controversy.
6.3.5.1 Crystallography of acicular ferrite
It is well established that acicular ferrite nucleates in the transformation temperature range
between Widmanstatten ferrite and lower bainite.55"59 Based on conventional diffraction pattern analyses in the transmission electron microscope (TEM), Bhadeshia 6tf a/.54'59have shown
that the acicular ferrite plates exhibit an orientation relationship with the austenite grain in
which they grow. The observed orientation relationship lies within the Bain orientation region27 and can approximately be described as:

Previous Page

Growth rate of WF, u,m/s

Temperature, 0C
Fig. 6.35. Predicted growth rates of Widmanstatten ferrite in C-Mn steel weld metals; (a) Growth rate
calculations for weld compositions listed in Fig. 6.31, (b) Growth rate calculations after modifying compositions to allow for carbon enrichment due to grain boundary ferrite formation. Data from Bhadeshia

etai46

6.3.5 Acicular ferrite in steel weld deposits


Simultaneously with or immediately after the formation of Widmanstatten ferrite at the austenite
grain boundaries, acicular ferrite may start to nucleate intragranularly at non-metallic inclusions. This phase is commonly observed in low-alloy steel weld metals, where the fine dispersion of oxide inclusions provides favourable sites for heterogeneous nucleation.3653 There
seems to be general agreement that microstructures primarily consisting of acicular ferrite
provide optimum weld metal mechanical properties, both from a strength and toughness point
of view, by virtue of its small lath size and high dislocation density. 365354 Consequently, the
formation of this particular microconstituent is of significant commercial importance and has
therefore attracted substantial research interest over the years. 363746 " 58 In spite of all this effort, the acicular ferrite transformation in low-alloy steel weld metals is still a subject of considerable controversy.
6.3.5.1 Crystallography of acicular ferrite
It is well established that acicular ferrite nucleates in the transformation temperature range
between Widmanstatten ferrite and lower bainite.55"59 Based on conventional diffraction pattern analyses in the transmission electron microscope (TEM), Bhadeshia 6tf a/.54'59have shown
that the acicular ferrite plates exhibit an orientation relationship with the austenite grain in
which they grow. The observed orientation relationship lies within the Bain orientation region27 and can approximately be described as:

Measured volume fraction of WF

Calculated volume
fraction of WF
Fig. 6.36. Comparison between measured and predicted volume fractions of Widmanstatten ferrite in CMn steel weld metals. Data from Bhadeshia et al.46

This corresponds to the well-known Kurdjumow-Sachs (K-S) orientation relationship* and


suggests that growth of the AF plates occurs either by a ledge mechanism or by a pure shear
transformation similar to that reported for upper bainite in steel. 1^0-60"62
6.3.5.2 Texture components of acicular ferrite
Kluken et al.63 have studied the development of transformation textures in as-deposited steel
weld metals containing acicular ferrite, using the electron backscattering pattern (EBSP) technique. Referring to the (200) stereographic projection in Fig. 6.37, the measured acicular
ferrite orientations form a symmetrical and consistent pattern and disclose evidence of a coupled solidification/solid state transformation texture in the weld metal. A closer examination
of the data reveals the existence of three major texture components within the acicular ferrite
microstructure, i.e. one <100> component and two complementary <111> components.63 The
former component includes acicular ferrite plates which have a <100> direction approximately
*The orientation relationship can alternatively be described by the Nishiyama-Wasserman (N-W) correspondence
which also lies within the Bain orientation region. However, since the K-S and the N-W orientation relationships
only differ from each other by a 5.26 rotation of the close packed planes, they can be regarded as equivalent.

(a)

(b)

Fig. 6.37. The development of transformation textures in as-deposited steel weld metals containing acicular
ferrite; (a) (200) pole figure showing the crystallographic orientations of acicular ferrite referred to the
original cell/dendrite growth direction, (b) Backscattered electron channeling contrast image of delta
ferrite/austenite columnar grain (the metallographic section is normal to the cell/dendrite growth direction). After Kluken et al.63
parallel with the cell/dendrite growth direction (Fig. 6.38(a)). The other plates have a <111>
direction aligned in the same crystal growth direction (Fig. 6.38(b)). These data can be represented by two sub-components which are displaced with respect to each other by a 60 rotation
about a common <111> axis.
It is evident from the measurements of Kluken et al.63 that the acicular ferrite plates in asdeposited steel weld metals exhibit an orientation relationship with both the austenite and the
prior delta ferrite columnar grains in which they grow. This 'memory' effect arises from the
characteristic solidification pattern and transformation behaviour of low-alloy steel welds (e.g.
see discussion in Section 3.8.2, Chapter 3). As shown schematically in Fig. 6.39, the columnar
grain region will exhibit a sharp <100> solidification texture which has its origin in the phenomenon of preferred crystal growth. At the onset of the peritectic reaction, the austenite
adopts a K-S type of orientation relationship with the delta ferrite in order to minimise the
energy barrier against nucleation.64 The austenite subsequently grows around the periphery of
the primary phase until impingement occurs on neighbouring columnar grain boundaries. During
the yFe to aFe transformation, this memory effect gives rise to the formation of acicular ferrite
plates which have a <100> direction approximately parallel with the original cell/dendrite
growth direction. The presence of the two other texture components within the weld metal is
thus a result of complementary crystal rotations taking place within the same orientation region.
The proposed sequence of reactions is in excellent agreement with the texture analysis of
Hu65 who made theoretical calculations of the resulting orientations of iron after Fe to yFe and
yFe to aFe transformations in succession according to the scheme outlined in Fig. 6.39.
6.3.5.3 Nature of acicular ferrite
Currently, the mechanisms of acicular ferrite formation in low-alloy steel weld metals are not

Cell/dendrlte growth direction

(a)

(b)

Fig. 6.38. Schematic diagrams showing the three main texture components in acicular ferrite according
to the Kurdjumow-Sachs orientation relationship; (a) The <100> texture component, (b) The two complementary <111> texture components. After Kluken et a/.63
fully understood. However, detailed TEM studies performed by Bhadeshia et aL37>59>62>66 have
clearly demonstrated that acicular ferrite is a form of intragranularly nucleated bainite. In
practice, this means that the microconstituent may be present either as 'upper' or 'lower'
acicular ferrite in the weld deposit (depending on the carbon concentration), as shown
schematically in Fig. 6.40.
In general, the ferrite component of upper bainite is composed of groups of thin parallel
laths (subplates) with a well-defined crystallographic habit.2061 Although the growth mechanism of upper bainite is still a subject of considerable controversy, it has been postulated that
the subplates advance into the austenite with their own tip configurations. One model is shown
in Fig. 6.41, where each subplate forms as a ledge upon the adjacent subplates through a
nucleation and growth process. These ferrite laths possess the same variant of the K-S orientation relationship, which means that they are separated by low-angle grain boundaries. A
typical austenite grain will contain numerous sheaves of bainitic ferrite exhibiting different
variants of the K-S orientation relationship. This implies that the boundary between adjacent
plates of acicular ferrite should alternately be of the low-angle and high-angle type, a feature
which also has been observed experimentally (see data in Fig. 6.42). Hence, both the morphology and the crystallography of acicular ferrite bear a close resemblance to upper bainite.

Columnar grain
Fig. 6.39. Schematic diagram showing the sequence of reactions occurring during cooling of a low-alloy
steel weld through the critical transformation temperature ranges. After Kluken et al.63

6.3.5.4 Nucleation and growth of acicular ferrite


Since the thickening rate of plate-like precipitates is very high, the acicular ferrite laths, once
nucleated, will grow into the austenite phase until they physically impinge with neighbouring
plates. At present, there exist no models which allow the acicular ferrite content to be calculated from first principles. The reason for this is quite obvious, since the final weld metal
microstructure generally depends on complex interactions between a number of different variables, including:36
(i)
(ii)
(iii)
(iv)
(v)

The total alloy content.


The concentration, chemical composition, and size
distribution of non-metallic inclusions,
The solidification microstructure.
The prior austenite grain size,
The weld thermal cycle.

Carbon supersaturated plate


Carbon diffusion into
austenite and carbide
precipitation in ferrite

Carbon diffusion
into austenite

Carbide precipitation
from austenite

'Upper' acicular
ferrite

'Lower' acicular
ferrite

Fig. 6.40. Schematic illustration of the transition from 'upper' to 'lower' acicular ferrite in low-alloy
steel weld metals. The diagram is based on the ideas of Bhadeshia and Svensson.37
Successive
nucleation and
growth of
parallel plates

Misorientation, degrees

Fig. 6.41. Proposed model for nucleation and growth of upper bainite in steel (schematic). After
Verhoeven.61

Plate number
Fig. 6.42. Measured spatial misorientation between adjacent plates of acicular ferrite in a low-alloy steel
weld. Data from Kluken et a/.63

A more realistic approach would be to estimate the volume fraction of acicular ferrite
via the equation:37
(6-50)
where f%FFe and f%FFe are the corresponding volume fractions of grain boundary ferrite and
Widmanstatten ferrite, respectively (note that in equation (6-50) the formation of microphases
has been disregarded).
The method outlined above has shown to work well for numerous welds (e.g. see Fig. 6.43),
but fails when the primary microstructure consists of a mixture of acicular ferrite and martensite,
as is the case in high strength steel weld deposits6768 In spite of this shortcoming, equation (650) expresses in an explicit manner the real essence of the problem, namely that the evolution
of the acicular ferrite microstructure depends on the interplay between several competing nucleation and growth processes which occur consecutively during cooling from the Ae3-temperature. This important point is often overlooked when discussing the conditions for acicular
ferrite formation in low-alloy steel weld metals.
Size effects in heterogeneous nucleation
The important influence of second-phase particles on the austenite to ferrite transformation
has been examined theoretically by Ricks et al69 using classical nucleation theory. By assuming inert, incoherent, and non-deformable inclusions and constant values for the volume free
energy change and the surface free energy of both the yFe /yFe and yFe /aFe boundaries, the
normalised energy barrier to nucleation has been calculated and plotted against the particle
radius as shown in Fig. 6.44. It can be seen from the figure that the nucleation of ferrite at
inclusions is always energetically more favourable than homogeneous nucleation, but less
favourable than nucleation at austenite grain boundaries, irrespectively of the inclusion size.
The most potent nucleation sites are particles of a radius greater than about 0.2-0.5 jum, which

Volume fraction

Allotriomorphic
ferrite

Widmanstatten
ferrite
Acicular
ferrite

Carbon content, wt%


Fig. 6,43. Predicted variation in microstructure as a function of carbon concentration in C-Mn steel
weld deposits (SMAW-IkJ mm~ l). Data from Bhadeshia and Svensson.37

AG* /AG*
het.
horn.

Inclusion
, Austenite
grain boundary

Inclusion radius, jim


Fig. 6.44. Effect of particle radius on the normalised energy barrier against ferrite nucleation at inclusions. The corresponding energy barrier against nucleation of ferrite at austenite grain boundaries is
indicated by the horizontal broken line. Data from Ricks et al.69

is within the typical size range of most weld metal inclusions (see Figs. 2.57-2.61 in Chapter
2). This finding is in excellent agreement with the results of Barbaro et al.70 reproduced in Fig.
6.45, showing that a certain minimum inclusion size (say 0.2-0.3 jam) is required for acicular
ferrite nucleation in steel weld deposits.
It should be noted, however, that Ricks et al69 omitted a consideration of the effects of
plastic strain produced as a result of differences in thermal contraction between the austenite
and the particles as well as the possibility for the ferrite to adopt reasonable orientation relationships with both the austenite and the catalyst particles. Based on nucleation theory it can
be argued that these factors will influence the transformation process.1 This, in turn, may
explain why certain types of inclusions appear to be more favourable nucleation sites for acicular
ferrite than other (to be discussed below).
Catalyst effects in heterogeneous nucleation
As mentioned above, there is considerable circumstantial evidence available in the scientific
literature that intragranular nucleation of acicular ferrite is preferentially associated with specific types of non-metallic inclusions (i.e. 7-Al2O3, MnOAl2O3, TiN). 365571 " 78 Different mechanisms have been proposed over the years to explain these phenomena, including:3655'73'76'77
(i)

Nucleation resulting from a small lattice disregistry between the inclusions and the
ferrite.

(ii)

Nucleation in the vicinity of inclusions caused by local compositional inhomogeneity


in the steel matrix.

(iii) Nucleation in the vicinity of inclusions resulting from favourable strain or dislocation
arrays due to differences in the thermal contraction between the particles and the
matrix.

Probability of nucleation

Inclusion radius, JLX m


Fig. 6.45. Effect of inclusion size on the probability of acicular ferrite nucleation in steel weld deposits.
Data from Barbara et al.10

Because of the complexity of the weld metal inclusions, and the experimental difficulties
involved in performing controlled in situ measurements, it cannot be stated with certainty
which of these three mechanisms that are operative during the acicular ferrite transformation.
However, based on simple theoretical calculations it can be argued that the contribution from
the elastic strain fields around the particles due to differential contraction effects probably is
too insignificant to influence the free energy of transformation and that the resulting punchingstress at the particle/matrix interface is well below the critical value required to generate new
dislocations in the austenite.47'81'82 Moreover, detailed STEM/EDS microanalyses have failed
to reveal detectable variations in the matrix composition in the vicinity of the inclusions.83
Hence, nucleation resulting from a small lattice disregistry between the inclusions and the
ferrite appears to be the most likely explanation to the observed effects of deoxidation practice
on the weld metal transformation behaviour.
From a theoretical standpoint, the development of a faceted ferrite nucleus which exhibits a
rational orientation relationship with both the austenite and the inclusions would require that
the substrate and the austenite have similar crystal structures and identical lattice orientations.
The catalyst particles must therefore be cubic and bear an orientation relationship with the
austenite which lies within the Bain region.27 However, considering the fact that the weld
metal inclusions form in the liquid state prior to the solidification process, the latter requirement cannot generally be met.37 Nevertheless, even if the orientation of the inclusions were
perfectly random, it is apparent that orientation relations within the Bain region would be
observed purely by chance. In view of the high symmetry of the cubic system, the probability
of this happening must be calculated. An approximative estimate is given below for single
phase cubic inclusions, based on the method described by Ryder et al?4
Figure 6.46 contains a standard stereographic projection of the austenite crystal, showing
the <100> rFe -poles (squares) and the <110>rFe-poles (dots). The Bain orientation region is
represented by small circles of radius 11 centered on the <100> r / v - and the <110> rFe -poles.84

Fig. 6.46. Stereographic projection of Bain regions represented by 11 circles, round <100>7_Fe-poles
(squares) and <110>7_Fe-poles (dots) of the austenite lattice. After Ryder et aiS4
The austenite/inclusion(i) orientation relationship is within the orientation region derived
from the Bain correspondence if one <100>rpole lies within a <100>7_Fe-region and the other
two <100>rpoles lie within <110>rF<?-regions. Assuming a random inclusion orientation, the
probability P 1 that a given <100>rpole lies within a given <100> r ^-region is (o/47T, where o>
is the solid angle enclosed by one <100> r ^-region (equal to 27t(l-cos 11) = 0.115). Since
there are three <100>/ directions and six <100>7_Fe-regions, the probability that at least one
<100>rpole lies within a <100>7_F,-region is equal to:
(6-51)
Imagine now that the inclusion lattice is rotated through 360 about the <100>raxis which
lies within a <100> y . Fe -region. The other two <100> r poles will then lie within
<110>7.Fe-regions for at the most 4 X 22 = 88 of this rotation, since the diameter of the
<100>7_Fe-regions is 22. Hence, the probability P2 that, if one <100> r pole lies within a
<100>7.Fe-region, the other two will lie within <110>7.Fe-regions is given in the upper limit
by:
(6-52)
The total probability that a given orientation relationship lies within the Bain region purely
by chance is thus:
(6-53)
Therefore, assuming random orientation, about 4% of the weld metal inclusions would lie
within the Bain region purely by chance if they were single phase cubic crystals. In practice,
however, inclusions commonly found in low-alloy steel weld metals are of a very heterogeneous chemical and crystalline nature. As shown in Fig. 2.72 in Chapter 2, a typical inclusion
may contain up to six different constituent phases, including the three cubic phases 7-Al2O3,
MnOAl2O3, and TiN. This implies that at least 12% of the inclusions may contain a cubic
phase which lies within the Bain orientation region.

Measurements of orientation relationships between specific inclusion constituent phases


(i.e. 7-Al2O3, MnOAl2O3, and TiN) and contiguous acicular ferrite plates performed by Grong
et a/.85 support the above interpretation (see Table 6.3). Referring to the standard stereographic
projections in Fig. 6.47, a very high proportion of the ferrite/inclusion orientations falls within
the Bain region. In addition, two other variants (i.e. No. 5 and 6) are indicated for TiN which
do not meet this requirement. They are therefore regarded as spurious (in the sense that the
observed orientation relationships do not stem from a catalyst nucleation event) and should be
ignored. Hence, it may be concluded that the observed orientation relationships between acicular
ferrite and specific inclusion constituent phases are not fully reproducible in the true meaning
of the word, since only those combinations which satisfy the inherent crystallography of the
acicular ferrite microstructure are acceptable.
An interesting observation from the data in Table 6.3 is that nucleation of acicular ferrite on
inclusions is always associated with low-index planes of the {100} or the {110} type, which
indicates a faceted growth morphology of the inclusions. Faceted growth may occur as a result
of anisotropy in the growth rates between high-index and low-index crystallographic planes,
and can in the extreme case lead to a morphology of the type shown in Fig. 6.48. Consequently, formation of faceted inclusions in the liquid steel during deoxidation appears to be an
intrinsic feature of low-alloy steel weld metals.
Simple verification on the basis of classic nucleation theory shows that the associated reduction of the energy barrier to nucleation, AGlet , is the primary cause for the ferrite nucleus
to develop epitaxial orientation relationships with the substrate and the austenite.8687 Referring to Fig. 6.49, a qualitative ranking of the different inclusion constituent phases with respect
to nucleation potency of acicular ferrite can be made from the data presented in Table 6.3. It is
evident that both 7-Al2O3, MnOAl2O3, and TiN reveal a good lattice matching with the ferrite
phase in one crystallographic direction. In addition, nucleation of acicular ferrite at TiN offers
Table 6.3 Observed orientation relationships between acicular ferrite and different inclusion constituent phases in a SA low-alloy steel weld. Data from Grong et al.85
Substrate

Orientation Relationship

(s)

No.

Plane Combinations

Ratio^

(100) s ~//(011) a _ Fe
[011]s ~ // [533] a . Fe

{011}a_Feand{400}s

1.02

MnOAl2O3

(200) s ~//(110) a F e

{110}a_Feand {400}s

0.99

(spinel)

(011)s // (010)a Fe
[Oil],//[QOlLp 6

{200}a Fe and {440}s

0.99

(TlO) 8 // (100)a Fe
(112)s//(011)a_Fe
[lll] s //[011] a . F e

{200}a Fe and {220}s


{011} a . Fe and{112} s

0.97
1.17

(101) s //(103) a _ Fe
(320)s//(112)a_Fe
[232] s //[351] a _ Fe

{310} a . Fe and{330} s
{112}a.Fe and {320}s

0.91
0.99

{133} aFe and{550} s


{200}a.Feand {221 }s

1.10
1.01

7-Al2O3
(distorted spinel)

TiN
(NaCl)

Variant

Interplanar Spacing

(101) s //(133) a _ Fc
[221]s//[200]a_Fe
Defined as d(hkl)a.Fe /d(hkl)s

Fig. 6.47. Standard (100) stereographic projections of the orientation relationships listed in Table 6.3; (a)
Variant (1) and (3), (b) Variant (4), (5) and (6). The Bain orientation region is indicated by the 11 circles
in the graphs (see Fig. 6.46 for details).
the advantage of partial lattice coherence in a second (independent) direction, which further
contributes to a reduction of AG^e/ through a minimisation of the interfacial energy between
the two phases. This makes TiN an extremely efficient nucleant for acicular ferrite.
Microstructure data available for submerged arc (SA) steel weld deposits clearly support
the above findings that nucleation of acicular ferrite occurs preferentially at inclusions which
contain aluminium or titanium. As shown in Fig. 6.50, a high volume fraction of acicular
ferrite is always achieved when sufficient amounts of titanium are added either through the
filler wire or the flux, irrespectively of the aluminium and oxygen concentrations. This is in
sharp contrast to welds produced with welding consumables containing low levels of titanium,
where the acicular ferrite content drops rapidly with decreasing [A%Al]weld/[%O]anal ratios

Fig. 6.48, Example of a faceted crystal delimited by {100} and {110} planes (schematic).

TiN

MnOAI2O3

T-Al2O3

AG

;et.

Austenite grain boundary

Nucleation site
Fig. 6.49. Qualitative ranking of different inclusion constituent phases with respect to nucleation potency of acicular ferrite.
due to the presence of lower fractions of 7-Al2O3 and MnOAl2O3 in the inclusions (see Fig.
2.72 in Chapter 2). Similar observations have also been made by other investigators.3672"
76,78,79

It should be noted that the weld metal transformation behaviour in practice depends on
complex interactions between a number of important variables, including alloying and
deoxidation practice, the solidification microstructure, the prior austenite grain size, and the
weld thermal cycle.36'5358 This means that the presence of 7-Al2O3, MnOAl2O3 or TiN at the
surface of the inclusions is perhaps a necessary but not a sufficient criterion for formation of
acicular ferrite in steel weld metals.
Example (6.8)

Consider a partly Ti-Al deoxidised steel weld metal which contains a total number of 4 X 107
inclusions per mm 3 . Based on Fig. 2.72 in Chapter 2 and the theory developed above, estimate
an upper limit for the volume of a typical plate of acicular ferrite when the weld metal [A%Al]weld/
\yO\anaL r a t i o i s - 80 -

Vol% AF

[A%A|lweld/[%o]anal
Fig. 6.50. Effect of deoxidation practice (inclusion chemistry) on the acicular ferrite transformation in
low-alloy steel weld metals. Data compiled from miscellaneous sources.
Solution

From Fig. 2.72 it is seen that the total number of constituent phases in the inclusions is six,
including the three cubic phases 7-Al2O3, MnOAl2O3, and TiN. If we assume a random orientation and only one nucleation event per inclusion, the following upper limit for the acicular
ferrite plate volume is obtained:

The above volume corresponds to an acicular ferrite plate which has the shape of a square
lath of side lOjim and thickness 2|im. Although this estimate is in reasonable agreement with
experimental observations,37 the prediction is conservative in the sense that it assumes only
one nucleation event per inclusion. In practice, an oxide inclusion which is orientated within
the Bain region has the capability of nucleating several acicular ferrite plates, as shown by the
SEM micrograph in Fig. 6.51. In addition, the acicular ferrite plates may nucleate
autocatalytically at aFe /yFe boundaries, a process which also is referred to as sympathetic
nucleation in the literature.36'37'7083 At present, it is not clear to what extent autocatalytic
nucleation plays a role in the development of the acicular ferrite microstructure.
Hardenability effects
Since acicular ferrite is one of the last phases to form after the growth of allotriomorphic and
Widmanstatten ferrite, it is bound to be influenced by the prior transformation products, as
indicated by equation (6-50). The strong dependence of the acicular ferrite content on the
austenite grain size must therefore be understood on this basis. 3747 ' 48 ' 7088 " 90 It is evident from
the data of Barbaro et al.70 reproduced in Fig. 6.52 that a coarse austenite grain size favours
intragranular nucleation of acicular ferrite at the expense of formation of allotriomorphic and

Fig. 6.51. SEM micrograph of a carbon extraction replica showing evidence of multiple nucleation of
acicular ferrite at a weld metal inclusion.
Widmanstatten ferrite. This effect is most pronounced during slow cooling, since the combination of a small austenite grain size and a slow cooling rate implies that much of the yFephase already has transformed to allotriomorphic ferrite before the temperature for intragranular
nucleation of acicular ferrite is reached.
Example (6.9)

Consider a low-alloy steel weld metal which contains a total number of 5 X 107 inclusions per
mm3 with an average radius of 0.25 Jim. Use this information to evaluate the conditions for
acicular ferrite formation within the as-deposited weld metal and the low-temperature reheated
region of the weld when the austenite grain size is 100 and 10 |iim, respectively. In these
calculations we shall assume that the equilibrium volume fraction of ferrite ff~Fe in the fully
transformed steel is 0.9.
Solution

First we need to estimate the total surface area per unit volume available for ferrite nucleation at austenite grain boundaries, Sv (GB), and non-metallic inclusions, Sv (/), respectively.
As-deposited weld metal

Reheated weld metal

(as before)

Volume fraction of AF

Austenjte grain size,jim


Fig. 6.52. Effect of austenite grain size on the acicular ferrite transformation in low-alloy steel weld
deposits. Data from Barbara et al.70
From the above calculations it is apparent that the conditions for intragranular nucleation of
acicular ferrite at inclusions are particularly favourable within the as-deposited weld metal,
since SV(I) > SV(GB), whereas nucleation and growth of ferrite at austenite grain boundaries
will dominate within the reheated region of the weld (SV(GB) SV(I)). This conclusion is also
consistent with predictions based on the Avrami equation (equation (6-49)). If we, as an illustration, assume that the volume fraction of grain boundary ferrite in the as-deposited weld
metal is 0.3, the corresponding fraction of GF in the reheated weld metal becomes:

On this basis it is not surprising to find that the microstructure within the grain refined
region of low-alloy steel welds is usually polygonal ferrite, while the as-deposited weld metal
also contains high proportions of acicular ferrite (see Fig. 6.19(c) and (d), respectively).
It is important to realise, however, that the presence of allotriomorphic ferrite at the austenite
grain boundaries has the benefical effect of suppressing the formation of bainitic sheaves at the
austenite grain boundaries, which, in turn, allows the acicular ferrite to develop on intragranular
nucleation sites.9091 Consequently, due to the number of competing nucleation and growth
processes involved, the volume fraction of acicular ferrite is often seen to pass through a maximum when the weld metal hardenability is successively increased by additions of alloying
elements, as shown in Fig. 6.53.

Microstructural component, %

GF and PF
WF

Fertile with aligned


second phase (FS)

Chromium content, wt%


Fig. 6.53. Effect of chromium additions on the C-Mn primary weld metal microstructure. After Babu
and Bhadeshia.91

6.3.6 Acicular ferrite in wrought steels


In recent years a new class of low-carbon microalloyed steels has emerged to meet the need for
improved weldability, particularly with respect to the HAZ toughness. 70'92 These steels are
not aluminium-killed (they usually contain less than 30 ppm Al), but are instead deoxidised
with titanium (120-140 ppm) to produce a relatively coarse distribution of submicroscopic Ti
oxide inclusions within the base metal (presumably Ti 2 O 3 ). During high heat input welding (>
5 kJ mm"1), the oxides will not retard austenite grain growth but instead act as favourable
nucleation sites for acicular ferrite within the interior of the austenite grains, as shown
schematically in Fig. 6.54(a). However, since the inclusion number density is quite low (<106
particles per mm3), the microstructure is significantly coarser than that normally observed in
steel weld deposites.92 Nevertheless, their presence contributes to suppress the formation of
Widmanstatten ferrite and upper bainite within the grain coarsened HAZ, which is a common
problem with the traditional Al-Ti microalloyed steels22 (see Fig. 6.54(b)).

63.7 Bainite
Bainitic microstructures (besides acicular ferrite) are frequently observed during welding, particularly in the HAZ of low-carbon microalloyed steels,22'93"95 but also within the fusion region of the weld if the nucleation conditions are favourable.37'55'5991 Two main forms can be
identified, i.e. upper and lower bainite, as indicated in Fig. 6.19(a).
6.3.7.1 Upper bainite
In general, the morphology of upper bainite bears a close resemblance to Widmanstatten ferrite (both are typical sideplate structures), while its crystallography is more like that of lowcarbon martensite (the K-S orientation relationship is usually less precise than for WF).20 In
spite of this similarity, the formation of upper bainite cannot be fully understood in terms of

HAZ

WF/UB

HAZ
Fig. 6.54. Schematic illustration of the HAZ transformation behaviour during high heat input welding;
(a) Ti-oxide containing steel, (b) Conventional Al-Ti microalloyed steel. The diagrams are based on the
ideas of Homma et al.92

the classic theory of martensite nucleation assuming a pure invariant plane strain deformation.1'20'61
HAZ transformation behaviour
As shown in Fig. 6.19(a), upper bainite is frequently observed within the grain coarsened HAZ
of low-carbon microalloyed steels. The bainite sheaves nucleate preferentially at austenite
grain boundaries and subsequently grow into the interior of the grains until they physically
impinge with other transformation products such as martensite or Widmanstatten ferrite. Because of these competitive nucleation and growth processes, the highest volume fractions of
upper bainite are normally attained at an intermediate cooling rate, as indicated by the microstructure data in Fig. 6.55.
Weld metal transformation behaviour
Upper bainite may also form within the columnar grain region of high heat input SA steel
welds, as shown by the optical micrograph in Fig. 6.56. The indications are that the observed

Microstructural component, %

iUBj

Ferrite with
aligned second dhase (FS)

WF

Mill
Cooling time, At 8/5 , s
Fig. 6.55. Effect of cooling time, Af8/5, on the grain coarsened HAZ transformation behaviour (simulated
thick plate heat cycles, Tp 13500C). Data from Akselsen et al96

Fig. 6.56. Optical micrograph showing formation of upper bainite within the columnar grain region of a
SA steel weld.

shift in the weld metal transformation behaviour is related to a change in the deoxidation
practice which alters the kinetics of the subsequent solid state transformation reactions through
a modification of the solidification microstructure.64 Solidification induced phosphorus
segregations are of particular interest in this respect, since previous examinations have shown
that phosphorus can strongly enhance the formation of grain boundary ferrite by raising the
local Ae3 temperature of the steel97 (see also Example 6.5).
Following the discussion in Section 6.3.5.4, the evolution of allotriomorphic ferrite at the
austenite grain boundaries has the beneficial effect of suppressing the formation of bainitic
sheaves at these sites, which, in turn, allows the acicular ferrite to develop on intragranular

Microstructural component, %

Carbides

Lower bainite

Martensite
Fig. 6.58. TEM micrograph showing the formation of lower bainite within the HAZ of a low-carbon
microalloyed steel.

6.3.8 Martensite
At very high undercoolings, the austenite decomposes to martensite by means of an invariant
plane strain deformation, which implies that there is no change in the steel chemical composition. The reaction product will either be lath or plate (twinned) martensite, depending on the
alloy level. Lath martensite is commonly found in plain carbon and low-alloy steels up to
about 0.5wt% C, and is formed by a slip mechanism, as shown schematically in Fig. 6.59(a).
When the carbon content exceeds this threshold, the martensite transformation occurs rather
by formation of deformation twins (Fig. 6.59(b)). The crystal structure of plate martensite is
bet (body-centred tetragonal), while lath martensite reveals a bcc (body-centred cubic) structure,61 which becomes increasingly distorted with increasing steel carbon contents.20 Both
transformation products exhibit the characteristic Kurdjumow-Sachs orientation relationship
with the austenite, but this relationship tends to be less precise at high carbon levels.1'20
6.3.8.1 Lath martensite
Lath martensite is frequently observed within the grain coarsened HAZ during low heat input
welding of microalloyed steels. As shown in Fig. 6.60, the microstructure will be fully
martensitic for values of Ar875 up to about 5-10s, depending on the steel hardenability. The
hardness of the martensite formed is usually below 400 VPN, which is significantly lower than
the corresponding peak value achieved after water-quenching.98 This means that considerable
autotempering occurs during cooling from the M^-temperature, which in the case of low-carbon microalloyed steels can be as high as 5000C.
6.3.8.2 Plate (twinned) martensite
In the intercritical HAZ (i.e. partly transformed region), carbon-rich austenite, formed by decomposition of pearlite or isolated carbides, may transform to high carbon (twinned) martensite
on cooling, as shown schematically in Fig. 6.61. However, since the Mf temperature in this
case will be far below room temperature, significant amounts of retained austenite may be
present in the areas adjacent to the martensite islands. This transformation product is therefore

Slip

Twin

Fig. 6.59. Mechanisms of martensite formation in steel (schematic); (a) Slip along parallel planes, (b)
Generation of deformation twins. The diagram is based on the ideas of Verhoeven.61

Steel A

Martensite content, vol%

Steel B
Steel C

Cooling time, At0._ ,s


8/5
Fig. 6.60. Effect of cooling time, Af8/5, on the volume fraction of lath martensite within the grain coarsened HAZ of microalloyed steels (simulated thick plate heat cycles). Data from Akselsen et /.98

frequently referred to as the martensite-austenite (M-A) constituent in the scientific literature.99"103


Kinetics of austenite formation
In general, the formation of austenite during intercritical annealing of low-carbon microalloyed
steels can be separated into three main stages:99
(i)

Rapid growth of austenite into pearlite until the dissolution process is completed.

(ii)

Slower growth of austenite into ferrite at a rate which is either controlled by carbon
diffusion in the austenite or by diffusion of substitutional elements such as manganese in the ferrite, depending on the applied annealing temperature.

wt% C

Fig. 6.61. Schematic illustration of the formation of plate (twinned) martensite within the intercritical
HAZ of low-carbon microalloyed steels; Heating leg of thermal cycle: Pearlite -> -yFe (>0.5 wt% C),
Cooling leg of thermal cycle: yFe(> 0.5 wt% C) -> a'Fe (twinned martensite).

(iii) Very slow final equilibration of solute concentration gradients within the austenite or
the ferrite through diffusion.
At temperatures above say 7700C, the rate of pearlite decomposition is sufficiently high for
complete dissolution, even during low-heat input welding. Hence, in hot rolled and normalised steels all pearlite will transform to austenite when the peak temperature of the thermal
cycle exceeds this threshold. However, if the starting microstructure is a mixture of martensite
and upper bainite (as often will be the case in multi-pass welding), the austenite will nucleate
both at prior austenite grain boundaries and along the interfaces between laths of bainite or
martensite, as shown by the optical micrographs in Fig. 6.62. A similar pattern has also been
observed during intercritical annealing of dual-phase steels.104'105 Considering the kinetics,
carbides precipitated within autotempered low-carbon martensite or between the laths of bainite
would be expected to dissolve at a rate comparable with that of pearlite.105 Hence, it is reasonable to assume that no retained carbides will be present intragranularly after reheating to say
Tp 7700C when the starting microstructure is a mixture of martensite and bainite.
Further growth of the austenite into the ferrite requires, however, that the peak temperature
of the thermal cycle is raised significantly above 7700C in order to reach the kinetic
(paraequilibrium) stage where the reaction is controlled by diffusion of carbon in austenite."
Consequently, the possibilities of obtaining growth of the austenite colonies within the low
peak temperature regions of the intercritical HAZ are strongly limited, which implies that the
carbon content of the austenite islands in these regions should be close to the saturation level
of carbon in austenite at the temperature of dissolution.
Conditions for M-A formation
On subsequent cooling, the carbon-rich austenite can decompose to a variety of microstructures (ranging from twinned martensite to pearlite) or remain untransformed, depending on the
base plate hardenability, the weld cooling programme, and the initial size of the austenite
islands.102103'106 When the starting microstructure is a mixture of ferrite and pearlite (as frequently observed in normalised steels), twinned martensite forms readily along the prior base

(a)

(b)

Fig. 6.62. Optical micrographs showing favourable sites for austenite formation during two-pass weld
thermal simulation. (First cycle: Tp 13500C, Atm 12 s, Second cycle: Tp 775C, Atm 12 s);
(a) Intergranular, (b) Intragranular. After Akselsen et aim

metal pearlite bands or within isolated pearlite colonies at high cooling rates, as shown by the
TEM micrographs in Fig. 6.63(a) and (b). With decreasing cooling rates, however, the transformation product shifts from twinned martensite to predominatly pearlite (see Fig. 6.63(c)
and (d)) in the absence of strong hardenability elements such as molybdenium and boron which
can stabilise the M-A constituent.102'107"109
In controlled rolled and accelerated cooled steels, where the carbides are mainly present in
the form of submicroscopic colonies or isolated particles located at ferrite/ferrite grain boundaries
(see Fig. 6.64(a)), the situation becomes slightly more complex. Under such conditions, the
majority of the austenite islands formed within the low peak temperature region of the
intercritical HAZ will be of a size below 1 jam. It has been verified experimentally that such

(a)

(C)

(b)

(d)

Fig. 6.63. Examples of transformation products formed after intercritical thermal cycling of a ferritic/
pearlitic starting microstructure to a peak temperature of 775C; (a) M-A islands (black areas) surrounded by ferrite (At6/4 ~ 5 s), (b) Close-up of twinned martensite within a M-A island (At6/4 ~ 5 s),
(c) Isolated pearlite colony formed at an intermediate cooling rate (Ar674 ~ 12 s), (d) Pearlite colonies
formed during slow cooling (At614 ~ 35 s). After Akselsen et al.103

small particles do not readily transform to martensite, but will largely remain in the steel in the
form of retained austenite.100'103'110'111 This can be explained by the lack of nucleation opportunities for martensite or by the volume restraint provided by the surrounding ferrite matrix.1 x x
Consequently, twinned martensite in controlled rolled and accelerated cooled steel is seen to
form within a few, relatively large austenite islands which stem from decomposition of single
carbide colonies, as illustrated by the TEM micrograph in Fig. 6.64(b).
During multipass welding, the initial carbide distribution in the base plate will be of less
importance because of the transformations imposed by the heat of previous weld passes. Hence,
after full reaustenitising of the steel the M-A constituent may form both intergranularly and
intragranularly, depending on the starting microstructure, as shown previously in Fig. 6.62(a)
and (b). An example of intergranularly nucleated twinned martensite is contained in Fig. 6.65.
Volume fraction of M-A constituent
At present, an adequate kinetic model for partial reaustenitising of steel during continuous
heating and cooling is lacking. However, in view of the previous discussion it is obvious that
the volume fraction of the M-A constituent within the intercritical HAZ of single pass steel
weldments must be closely related to the initial base plate pearlite content. This is also in
agreement with experimental observations (see Fig. 6.66).

Next Page

(a)

(b)

Fig. 6.64. Conditions for martensite formation within the intercritical HAZ of a controlled rolled and
accelerated cooled steel containing copper and nickel; (a) TEM micrograph of the initial base plate carbide distribution, (b) TEM micrograph showing evidence of twinned martensite within a l(im large
austenite colony formed during intercritical thermal cycling to T 775C (Ar674 35 s). After Akselsen
etal^

6.4 Austenitic Stainless Steels


Simple austenitic stainless steels generally contain between 18 and 30 wt% chromium, 8 to 20
wt% nickel, and between 0.03 to 0.1 wt% carbon.20 Since the solubility of carbon decreases
rapidly with temperature (see Fig. 6.67), reheating of the steels within the temperature range
from 500 to 9000C will lead to the rejection of carbon from solid solution, usually by chromium carbide precipitation (Cr23C6). These carbides nucleate preferentially at the austenite
grain boundaries, which, in turn, results in depletion of the regions adjacent to the grain boundaries with respect to chromium, as shown schematically in Fig. 6.68. The presence of such
chromium-depleted regions within the steel makes it sensitive to intergranular corrosion in
service.20'112

Previous Page

(a)

(b)

Fig. 6.64. Conditions for martensite formation within the intercritical HAZ of a controlled rolled and
accelerated cooled steel containing copper and nickel; (a) TEM micrograph of the initial base plate carbide distribution, (b) TEM micrograph showing evidence of twinned martensite within a l(im large
austenite colony formed during intercritical thermal cycling to T 775C (Ar674 35 s). After Akselsen
etal^

6.4 Austenitic Stainless Steels


Simple austenitic stainless steels generally contain between 18 and 30 wt% chromium, 8 to 20
wt% nickel, and between 0.03 to 0.1 wt% carbon.20 Since the solubility of carbon decreases
rapidly with temperature (see Fig. 6.67), reheating of the steels within the temperature range
from 500 to 9000C will lead to the rejection of carbon from solid solution, usually by chromium carbide precipitation (Cr23C6). These carbides nucleate preferentially at the austenite
grain boundaries, which, in turn, results in depletion of the regions adjacent to the grain boundaries with respect to chromium, as shown schematically in Fig. 6.68. The presence of such
chromium-depleted regions within the steel makes it sensitive to intergranular corrosion in
service.20'112

Content of M-A constituent, vol%-

Fig. 6.65. TEM micrograph showing evidence of intergranularly nucleated twinned martensite in double-cycled specimens (First cycle: Tp ~ 13500C, A;6/4 -12s, Second cycle: Tp - 775C, A;6/4 - 12s) After
Akselsen et al.103

Base plate pearlite content, v o l %

Fig. 6.66. Effect of base plate pearlite content on the volume fraction of M-A constituent in thermally
cycled specimens (simulation conditions: Tp -775C, Atm 5s). Data from Akselsen et al.m

Temperature, 0C

Carbon content, wt%


Fig. 6.67. Solvus temperatures for different types of carbides in 304 austenitic stainless steel as a function of carbon content. Data from Kou.112
Cr depleted zone

Grain boundary

Cr carbide precipitate
Fig. 6.68. Schematic illustration OfCr23C6 precipitation at grain boundaries in austenitic stainless steels.
The diagram is based on the ideas of Kou.112

6.4.1 Kinetics of chromium carbide formation


Precipitation of Cr23C6 in austenitic stainless steels is a typical nucleation and growth process.
The temperature-time transformation curve is therefore C-shaped, and can be modelled according to the general principles outlined in Section 6.2.2.3. As an illustration, we shall assume that the parameters listed in Table 6.4 are representative of nucleation of Cr23C6 in a 304
type austenitic stainless steel containing 0.07 wt% C. If we at the same time allow for the
variation in austenite grain size with distance from the fusion boundary, the shape and position
of the C-curve in temperature-time space at different locations within the HAZ is given by
equation (6-35):

(6-54)

A graphical representation of equation (6-54) in Fig. 6.69 shows that the nose of the
C-curve for the reference steel (characterised by Teq = T*eq and D1 = D 7 ) is strongly shifted to
the left in the IT-diagram, thereby providing favourable conditions for Cr23C6 precipitation in
the heat affected zone during welding. The problem becomes less imminent if the base plate
carbon content is reduced from say 0.07 to 0.04 wt%. In that case the associated reduction in
the solvus temperature Teq from 920 to 8000C will displace the C-curve towards longer times
in the diagram which by far exceed the duration of a typical weld thermal cycle, as shown in
Fig. 6.69.
6.4.2 Area of weld decay
Experience shows that corrosion attact within the HAZ of single pass austenitic stainless steel
welds usually starts in regions where the peak temperature of the thermal cycle has been close
to the nose of the C-curve in Fig. 6.69. This observation cannot be explained by just considering the HAZ temperature-time pattern, since the retention time within the critical temperature
range for chromium carbide precipitation is virtually the same for both high and low peak
temperature thermal cycles, as indicated in Fig. 6.70. The explanation is, of course, that the
C-curve becomes increasingly shifted to the right in the IT-diagram as the peak temperature
increases due to austenite grain growth occurring during welding. This point is more clearly
illustrated below.
Example (6.10)

Consider single pass butt welding of 2mm sheets of 304 austenitic stainless steels with covered electrodes under the following conditions:

Evaluate on the basis of equations (6-32), (6-33), and (6-54) in combination with the grain
growth diagram in Fig. 5.30(b) (Chapter 5) the conditions for chromium carbide formation
within the HAZ during welding when the base plate carbon content is 0.07 and 0.04 wt% C,
respectively. Thermal data for austenitic stainless steels are given in Table 1.1 (Chapter 1).

Temperature, 0C

Time, s
Fig. 6.69. Effect of carbon content on the isothermal precipitation of C ^ C 6 in 304 austenitic stainless
steels. Broken curve: 0.07 wt% C, Solid curve 0.04 wt% C. The diagrams are constructed on the basis
of equation (6-54).
T

Grain growth Area of weld


zone
decay
Fig. 6.70. Schematic illustration of thermal cycles and sensitisation region within the HAZ of an
unstabilised austenitic stainless steel.
Table 6.4 Input data used to construct C-curve for Cr23C6 precipitates in reference steel.
Parameter

Value

tr

T*q

A0

Q/

>*

(K)

(s)

(K)

(Jmol- 1 )

(kJmol-1)

(jxm)

923

60

1193

150

200

18

[ Activation energy for grain boundary diffusion of Cr in austenite.

Solution

If we, as a first approximation, neglect the contribution from heat flow in the welding direction, the temperature-time pattern is given by equation (1-100) in Chapter 1. Taking n = 4 and
(1 Xc) = 0.98 (assuming interface-controlled growth), the extent of chromium carbide precipitation occurring within the HAZ of the weld can be calculated from equations (6-32) and
(6-33) by numerical integration over the actual thermal cycles. The results of such computations are presented graphically in Fig. 6.71.
As expected, precipitation of chromium carbides occurs readily within the low peak temperature region of the weld when the carbon content is 0.07 wt%. For points located within
the grain growth zone, the resulting displacement of the C-curve towards longer times in the
CCT-diagram will gradually retard the reaction, which implies that the fraction transformed
starts to decline when the peak temperature of the thermal cycle exceeds about 11500C. In
contrast, the low-carbon steel reveals no sign of chromium carbide precipitation within the
HAZ. This result is in good agreement with general experience, and explains why the carbon
content of modern austenite stainless steels has been gradually lowered to values below 0.03
wt % in step with the progress in the steel manufacturing technology.20112

6.5 Al-Mg-Si Alloys

Austenite grain size, pm

Grain growth zone

Unaffected base metal

Fraction transformed

Age-hardenable Al-Mg-Si alloys are widely used as structural components in welded assemblies. They offer tensile strength values higher than 350 MPa in the artificially aged T6 condition owing to the presence of very fine, needle shaped (3"(Mg2Si) precipitates along <100>
directions in the aluminium matrix.113 Although Al-Mg-Si alloys are readily weldable, they
suffer from severe softening in the heat affected zone (HAZ) because of reversion (dissolution) of the 3"(Mg2Si) precipitates during the weld thermal cycle. 6112 This type of mechanical
impairment represents a major problem in engineering design.114

Peak temperature, 0C
Fig. 6.71. Conditions for Cr23C6 precipitation within the HAZ of a single pass austenitic stainless steel
butt weld (Example 6.10).

6.5.1 Quench-sensitivity in relation to welding


High strength alloys such as AA 6082 contains manganese in addition to magnesium and
silicon. Manganese is added to control recrystallisation and grain growth in the material during hot forming. The disadvantage is that it increases the quench sensitivity of the alloy.115
The reason for this is that the Mn-bearing dispersoids (which form during homogenisation)
provide favourable nucleation sites for the non-hardening (3'(Mg2Si) phase, as shown by the
TEM micrograph in Fig. 6.72. The resulting reduction in the solute content leads to a reduced
HAZ strength both in the naturally aged TA and peak aged T6 conditions. 56116
6.5.1.1 Conditions for (3 '(Mg2Si) precipitation during cooling
Myhr and Grong6 have shown how the quench-sensitivity concept can be applied to welding
of Al-Mg-Si alloys. Their model has later been refined and extended to heat treatment of AA
6082 alloys by Shercliff et al.ni and Bratland et a/.118119
As an illustration, we shall assume that the parameters listed in Table 6.5 are representative
of nucleation of (3'(Mg2Si) particles in an AA 6082 alloy homogenised at 5800C. If we further
assume that the P'(Mg2 Si) particles only form at Mn-bearing dispersoids, the (N* /Nv) term
in equation (6-14) may be replaced by the corresponding (5* / Sv) ratio, where S* and Sv refer
to the total surface area per unit volume of dispersoids in the reference and the actual aluminium alloy, respectively. This leads to the following expression for t*:

Fig. 6.72. TEM micrograph showing nucleation of non-hardening (3'(Mg2Si) precipitates at Mn-bearing
dispersoids in an AA 6082 alloy.
Table 6.5 Input data used to construct C-curve for p' (Mg2Si) precipitates in reference aluminium alloy.
Parameter

Value

Tr

t\

T*q.

A0

(K)

(s)

(K)

(Jmol"1)

573

20

766

350

f" Activation energy for diffusion of Mg in aluminium.

Qj

S^

(kJmol"1) (mm2 per mm3)


130

100

(6-55)
A graphical representation of equation (6-55) in Fig. 6.73 shows that the nose of the Ccurve for the reference alloy (characterised by Teq = r and Sv = S^ is strongly shifted to
the left in the IT-diagram, thereby providing favourable conditions for (3'(Mg2Si) formation
dunng welding. In general, an increase in T or Sv will enhance the quench-sensitivity of the
material because of the resulting increase in the nucleation rate. This will be the case if the
alloy contams large amounts of excess silicon in solid solution or is homogenised at a temperature lower than 5800C.118119
Example (6.11)

Consider plasma-MIG butt welding of a 10 mm thick Al-Mg-Si plate under the following
conditions:
qo = 10 kW, v = 10 mm s~\ T0 = 200C
Evaluate on the basis of equations (6-32), (6-33), and (6-55) the conditions for P'(Mg,Si)
precipitation within the high peak temperature region of the HAZ during welding (T7 > T )
In these calculations we shall assume that the transformation behaviour of the base'metafis
similar to that of the reference alloy in Table 6.5. Relevant thermal data for Al-Mg-Si alloys
are given in Table 1.1 (Chapter 1).
Solution

Temperature,C

Referring to Fig. 1.43 in Chapter 1, the mode of heat flow becomes essentially one-dimensional if the net arc power is kept sufficiently high compared with the plate thickness (i.e.

Time, s
Fig. 6.73. C-curve for precipitation of 3'(Mg2Si) in the reference AA 6082 alloy. The diagram is constructed on the basis of equation (6-55).

qold>0.05 kW mm"1). Since this requirement is met in the present case, the HAZ temperaturetime pattern is given by equation (1-100).Taking n = 0.75 and (1 - Xc) = 0.84 for precipitation
of p'(Mg2Si) in AA6082 aluminium alloys,118 we obtain after integration of equations (6-32)
and (6-33) over the weld cooling cycle:

and

The above calculations show that precipitation of (3'(Mg2Si) particles at dispersoids is, indeed, a significant process under the prevailing circumstances. Since the mode of heat flow
during single pass butt welding of aluminium plates is essentially one-dimensional, it is possible to construct general transformation diagrams which give the fraction transformed as a
function of the applied heat input. An example of such a diagram is contained in Fig. 6.74.

Fraction transformed

6.5.1.2 Strength recovery during natural ageing


Myhr and Grong6 have shown how equation (6-15) can be applied for modelling of the HAZ
room temperature ageing characteristics. Figure 6.75 shows a typical natural ageing curve for
AA 6082 alloys after full solution heat treatment. Due to enhanced precipitation of GP-zones,
the hardness will increase from about 42 VPN to a maximum of 80 VPN after a period of 5 to
7 days. However, this ultimate hardness is significantly reduced if solute is lost during the
weld cooling cycle because of precipitation of non-hardening (3'(Mg2Si) particles at
dispersoids.6117 The relationship between the solute content, C0, and the maximum fraction of
hardening precipitates which form during natural ageing, AXp, can be obtained from a simple

logt
Net heat input ( q /vd), kJ/mm2
Fig. 6.74. Conditions for 0'(Mg2Si) precipitation within the HAZ of single pass AA 6082 butt welds
(Tp>Teq).

Hardness (VPN)

Water-quenched
specimens

5-7 days

Log time
Fig. 6.75. Typical ageing curve for an AA 6082 aluminium alloy at room temperature (schematic).
2-D kinetic (cell) model, assuming that the reaction is interface-controlled. Let r denote the
radius of the growing precipitates (defined in Fig. 6.76(a)). Since we are only interested in the
terminal value of r at a fixed temperature, the time t in the expression for r can be regarded as
constant. Hence, we may write (when C0 Ca and Cp C0):
(6-56)
where C4 is a kinetic constant.
If the distribution of the precipitates is approximated by that of a 2-D face-centered cubic
space lattice (see Fig. 6.76(b)), the parameter, AXp, is simply given as:

(6-57)

where rm denotes the maximum particle radius which forms within the system if all alloying
elements are present in solid solution at the onset of the ageing reaction (C0 = C*).
Because of the stoichiometry of the precipitation reaction, C0 and C* in the expression for
AXp may be taken proportional to the magnesium concentration in solid solution. Hence, we
may write:
(6-58)

Equation (6-58) predicts that AXp is a simple power function of CMg.


The magnesium concentration in solid solution can be obtained from electrical resistivity
measurements, and is given by the following equation:6118
(6-59)
where pm is the measured resistivity, pss is the resistivity in the as-quenched condition, while p0

is the corresponding resistivity in the fully annealed condition (i.e. when all Mg and Si are tied
up in precipitates).
A comparison between equation (6-58) and the electrical resistivity data in Fig. 6.77 confirms the relevance of this power-law-relationship, although the deviation in certain cases in
admittedly large.
Example (6.12)

Consider plasma-MIG butt welding of a 10 mm thick Al-Mg-Si plate under conditions similar
to those employed in Example 6.11. Estimate on the basis of equation (6-58) the relative
(b)

Concentration

(a)

Distance

AXp

Fig. 6.76. Simplified 2-D kinetic (cell) model for precipitation of hardening particles in Al-Mg-Si alloys
during natural ageing; (a) Particle/matrix concentration profile, (b) Cell geometry.

Mg /C*Mg

Fig. 6.77. Experimental verification of equation (6-58).

fraction of hardening precipitates which forms within the fully reverted region of the HAZ
after prolonged room temperature ageing.
Solution

Referring to Example (6.11), the relative fraction of (3'(Mg2Si) precipitates which forms at
dispersoids during the weld cooling cycle amounts to:

from which

This gives:

Since the resulting precipitation strength increment is directly proportional to AXp,ul loss
of solute in the form of (3'(Mg2Si) particles during the weld cooling cycle will inevitably lead
to a reduced HAZ strength in the naturally aged (T4) condition.6 We shall return to this question in Chapter 7 (Section 7.4.3).
6.5.2 Sub grain evolution during continuous drive friction welding
Continuous drive friction welding is a solid-state joining process that produces coalescence by
the heat developed between two surfaces by mechanically induced rubbing motion. When the
appropriate rotation velocity is reached, the two workpieces are brought together and an axial
force is applied. The two surfaces are held under pressure, and due to the heat developed, a
plasticised layer forms at the interface. After a predetermined time the rotation stops and the
pressure is increased to facilitate forging or local upsetting of the heated metal. Filler metal,
flux, or shielding gas is not required with this process.
The structural changes within the fully plasticised region of friction welded Al-Mg-Si
alloys arise from the combined action of work hardening and softening due to dynamic recovery.
When the steady state conditions for deformation in friction welding are reached, the rate of
dislocation generation in the plasticised material balances the dislocation annihilation rate,
which means that large strains can be imposed without any changes in theflowstress or subgrain
size. Referring to Fig. 6.78, the subgrain structure within the fully plasticised region consists
of sheaves of virtually equiaxed grains (2-3 |jim in size) that are separated from each other by
low-angle grain boundaries. This type of microstructure is characteristic of hot worked aluminium alloys.120
In general, the Zener-Hollomon (Zh) parameter provides a basis for evaluating the evolution of the subgrain structure during hot working. As shown by McQueen and Jonas,121 the
following relationship exists between the subgrain diameter ds and the Z^-parameter in
aluminium alloys:
(6-60)

Partly deformed
region

Fully plasticized
region

Partly deformed
region

(a)

(b)

Fig. 6.78. Micrographs showing the subgrain structure within the fully plasticised and partly deformed
region of a friction welded Al-Mg-Si alloy; (a) Overview, (b) Close-up of the subgrain structure (EBSP
image).

where
(6-61)

The peak temperature T and the strain rate i distributions within the fully plasticised region of a friction weld can be computed on the basis of the generic models developed by
Midling and Grong.122 Examples of such calculations are shown in Fig. 6.79(a). Plots of the resulting Zener-Hollomon parameter and subgrain diameter at different locations within the
HAZ are contained in Fig. 6.79(b).

Partly deformed region

Strain rate, s"1

Radial position:

Peak temperature, 0C

Fully plasticized region

Radial position:

Fully plasticized region

Subqrain diameter, urn

Zener-Hollomon parameter, s

Axial distance, mm

Partly deformed region

Axial distance, mm
Fig. 6.79. Modelling of the subgrain evolution in the fully plasticised region of a friction welded AA
6082 aluminium alloy; (a) Predicted peak temperature Tp and strain rate (e) distributions, (b) Plots of
the Zener-Hollomon Zh parameter and resulting subgrain diameter ds at different locations within
the HAZ. Data from Midling and Grong.122

It is evident from the graphs that the value of the Zener-Hollomon parameter is of the order
of 1014-1012 s"1 within the fully plasticised region of the HAZ during continuous drive friction
welding of Al-Mg-Si alloys. The predicted range in Zh corresponds to a subgrain size of 2-3
jam, in agreement with experimental observations (see EBSP image in Fig. 6.78(b)). Outside
the fully plasticised region the Zener-Hollomon parameter drops from about 1012 to 1010 s"1
due to a sudden change the strain rate from 102 to 10 s"1 as the contribution from the material
flow field in the rotational direction ceases. This value is outside the validity range of equation
(6-60), and leads to the formation of coarse subgrains at the boundary between the fully plasticised and the partly deformed region, as shown by the TEM micrographs in Fig. 6.80.

Fig. 6.80. TEM micrographs showing evidence of coarse subgrains at the boundary between the fully
plasticised and partly deformed region of a friction welded AA 6082 aluminium alloy.

References
1.
2.
3.
4.

D. A. Porter and K.E. Easterling: Phase Transformations in Metals and Alloys, 1981, Wokingham
(England), Van Nostrand Reinhold.
M.E. Fine: Phase Transformations in Condensed Systems, 1964, New York, The Macmillian Company.
N. Ryum: Compendium in Physical Metallurgy, 1973, Trondheim (Norway), The University of
Trondheim, The Norwegian Institute of Technology.
G.F. Vander Voort (ed.): Atlas of Time-Temperature Diagrams for Nonferrous Alloys, 1991, Materials Park (Ohio), ASM International.

5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.

J.T. Staley: Mater. ScL TechnoL, 1987, 3, 923-935.


O.R. Myhr and 0. Grong: Acta Metall Mater., 1991, 39, 2693-2702; ibid., 2703-2708.
R.D. Doherty: In Physical metallurgy, 3rd Edn (Eds R.W. Cahn and P. Haasen), 1983, Amsterdam, 934-1030, North-Holland Physics Publishing.
H.B. Aaron, D. Fainstein and G.R. Kotler: J. Appl. Phys., 1970, 41, 4404-4410.
R. Trivedi: Metall. Trans., 1970, 1, 921-927.
J. W. Christian: The Theory of Phase Transformations in Metals and Alloys -Part I, 1975, Oxford
(England), Pergamon Press.
Atlas of isothermal transformation and cooling transformation diagrams, 1977, Materials Park
(Ohio), ASM International.
E. Scheil: Arch. Eisenhuttenwes., 1934/35, 8, 565-571.
J.S. Kirkaldy: Metall. Trans., 1973, 4, 2327-2333.
M.B. Kuban, R. Jayaraman, E.B. Hawbolt and J.K. Brimacombe: Metall Trans., 1986, 17A,
1493-1503.
LA. Wierszykkowski: Metall Trans., 1991, 22A, 993-999.
R.G. Kamat, E.B. Hawbolt, L.C. Brown and J.K. Brimacombe: Metall. Trans., 1992, 23A, 24692480.
M. Avrami: J. Chem. Phys., 1939, 7, 1103-1112; ibid, 1940, 8, 212-224; ibid, 1941, 9, 177-184.
J.W. Chan: Acta Metall, 1956, 4, 572-575.
M. Umemoto, Z.H. Guo and I. Tamura: Mat. ScL Technol, 1987, 3, 249-255.
R.W.K. Honeycombe: Steels Microstructure and Properties, 1980, London, Edward Arnold
(Publishers) Ltd.
H. Suzuki: Weld. World, 1982, 20, 121-148.
0. Grong and O.M. Akselsen: Metal Construction, 1986, 18, 557-562.
R.C. Cochrane: HW Doc. IXJ-102-85 (1985).
DJ. Abson and R.E. Dolby: HW Doc. IXJ-29-80 (1980).
CA. Dube, H.I. Aaronson and R.F. Mehl: Rev. Met, 1958, 55, 201-210.
H.K.D.H. Bhadeshia: Progress in Mater. ScL, 1985, 29, 321-386.
E.C. Bain: Trans. Am. Inst. Min. Eng., 1924, 70, 25-46.
S.S. Babu and H.K.D.H. Bhadeshia: Mater. ScL Eng., 1991, A142, 209-219.
H.K.D.H. Bhadeshia and D.V. Edmonds: Acta Metall, 1980, 28, 1265-1273.
W.C. Leslie: The Physical Metallurgy of Steels, 1981, London (England), McGraw-Hill.
CE. Cross, 0 . Grong, S. Liu and J.F. Capes: In Applied Metallography, (Ed. G.F. Vander Voort),
1986, New York, Van Nostrand Reinhold, 197-210.
J.E. Morral and T.B. Cameron: Proc. Int. Symp. on Boron in Steels, Milwaukee (Wisconsin),
September, 1979, 19-32, Publ. The Metall. Soc. AIME (1980).
S. Pakrasi, E. Just, J. Betzold and F. Hollrigl-Rosta: ibid, pp. 147-164.
L. F. Porter: ibid., pp. 199-211.
J.H. Devletian: Weld. J., 1976, 55, 5s-12s.
0. Grong and D.K. Matlock: Int. Met. Rev., 1986, 31, 27-48.
H.K.D.H. Bhadeshia and L.E. Svensson: In Mathematical Modelling of Weld Phenomena, (Eds
H. Cerjack and K.E. Easterling), 1993, London, The Institute of Materials, 109-180.
E.D. Hondros and M.P. Seah: In Physical metallurgy, (Eds R.W. Cahn and P. Haasen), 1983,
Amsterdam, North Holland Physics Publishing B.V., 855-931.
T.M. Williams, A.M. Stoneham and D.R. Harris: Met. ScL J., 1976,10, 14-19.
R.W. Fountain and J. Chipman: Trans. AIME, 1962, 224, 599-606.
Ph. Maitrepierre, D. Thivellier and R. Tricot: Metall Trans., 1975, 5A, 287-301.
O.M. Akselsen, 0 . Grong and PE. Kvaale: Metall Trans., 1986,17A, 1529-1536.
CJ. McMahon: Metall. Trans., 1980, HA, 531-535.
FA. Jackobs and G. Krauss: J. Heat Treating, 1981, 2, 139-146.

45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.

60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.

74.
75.
76.
77.
78.
79.
80.
81.
82.
83.
84.
85.
86.

G.M. Evans: HW Doc. II-A-469-79 (1979).


H.K.D.H. Bhadeshia, L.E. Svensson and B. Gretoft: Acta Metall., 1985, 33, 1271-1283.
S. Liu and D.L. Olson: Weld. J., 1986, 65, 139s-15Os.
N.A. Fleck, 0. Grong, G.R. Edwards and D.K. Matlock: Weld. J., 1986, 65, 113s-121s.
H.K.D.H. Bhadeshia, L.E. Svensson and B. Gretoft: J. Mat. ScL, 1986, 21, 3947-3951.
0. Grong, T.A. Siewert and G.R. Edwards: Weld. J., 1986, 65, 279s-288s.
H.K.D.H. Bhadeshia: Mater. ScL Forum, 1990, 56-58, 263-274.
A. AIi and H.K.D.H. Bhadeshia: Mater. ScL Technol., 1990, 6, 781-784.
DJ. Abson and RJ. Pargeter: Int. Met. Rev., 1986, 31, 141-194.
J.R. Yang and H.K.D.H. Bhadeshia: Weld. /., 1990, 69, 305-307.
RL. Harrison and R.A. Ferrar: Int. Mater. Rev., 1989, 34, 35-51.
R.C. Cochrane: Weld, in the World, 1983, 21, 16-29.
J.M. Dowling, J.M. Corbett and H.W. Kerr: Metall. Trans., 1986,17A, 1611-1623.
DJ. Abson: Weld, in the World, 1989, 27, 11-28.
M. Strangwood and H.K.D.H. Bhadeshia: Proc. Int. Conf. on Advances in Welding Science and
Technology, Gatlinburg, TN, May 1986, pp. 209-213. Publ. ASM INTERNATIONAL, Metals
Park, OH, 1986.
R.F. Hehemann, K.R. Kinsman and H.I. Aaronson: Metall. Trans., 1972, 3, 1077-1094.
J.D. Verhoeven: Fundamentals of Physical Metallurgy, 1975, New York, John Wiley & Sons.
H.K.D.H. Bhadeshia and J.W. Christian: Metall. Trans., 1990, 21A, 767-797.
A.O. Kluken, 0. Grong and J. Hjelen: Metall. Trans., 1991, 22A, 657-663.
A.O. Kluken, 0 . Grong and G. R0rvik: Metall. Trans., 1990, 21A, 2047-2058.
H. Hu: Trans. AIME, 1965, 233, 1071-1075.
A.A.B. Sugden and H.K.D.H. Bhadeshia: Metall. Trans., 1989, 2OA, 1811-1818.
P. Deb, K.D. Challenger and A.E. Therrien: Metall. Trans., 1987,18A, 987-999.
H.K.D.H. Bhadeshia and L.E. Svensson: J. Mater. ScL, 1989, 24, 3180-3188.
R.A. Ricks, G.S. Barritte and PR. Howell: Proc. Int. Conf. on Solid-Solid Phase Transformations,
Pittsburg (Pennsylvania), August 1981, pp. 463-468. Publ. The Metall. Soc. AIME (1982).
FJ. Barbaro, P. Krauklis and K.E. Easterling: Mater. ScL Technol., 1989, 5, 1057-1068.
S. Liu: In Ferrous Alloy Weldments, (Eds D.L. Olson and T.H. North), 1992, Zurich (Switzerland), Trans. Tech. Publications, pp. 1-18.
0. Grong and A.O. Kluken: ibid., pp. 21 -46.
L. Devillers, D. Kaplan, B. Marandet, A. Ribes and PV. Ribound: Proc. Int. Conf on Effects of
Residual, Impurity and Alloying Elements on Weldability and Weld Properties, London, November 1983, Paper 1, Publ. The Welding Institute (England).
C. Bonnet and F.P. Charpentier: ibid., Paper 8.
M.E. Saggese, D.N. Hawkins and J.A. Whiteman: ibid., Paper 15.
A.R. Mills, G. Thewlis and J.A. Whiteman: Mater. ScL Technol., 1987, 3, 1051-1062.
R.A. Farrar and PL. Harrison: J. Mat. ScL, 1987, 22, 3812-3820.
G.M. Evans: Weld. J., 1992, 71, 447s-454s.
M. Es-Souni, PA. Beaven and G.M. Evans: Mater. ScL Eng., 1990, A130, 173-184.
S. Liu and D.L. Olson: /. Mater. Eng., 1987, 9, 237-251.
S. Liu: PhD thesis, 1984, Dept. of Metall. Eng., Colorado School of Mines, Golden, Colorado
(USA).
0 . Grong and A.O. Kluken: Unpublished work.
R.A. Ricks, PR. Howell and G.S. Barritte: J. Mat. Sci, 1982,17, 732-740.
PL. Ryder, W. Pitsch and R.F. Mehl; Acta Metall., 1967,15, 1431-1440.
0 . Grong, A.O. Kluken, H.K. Nylund, A.L. Dons and J. Hjelen: Metall. Mater. Trans., 1995,26A,
525-534.
WC. Johnson, C. L. White, RE. Marth, RK. Ruf, S.M. Tuominen, K.D. Wade, K.C. Russell and
H.I. Aaronson: Metall Trans., 1975,6A, 911-919.

87.
88.
89.
90.
91.
92.
93.
94.
95.
96.
97.
98.
99.
100.
101.
102.

103.
104.
105.
106.
107.
108.
109.
110.
111.
112.
113.
114.
115.
116.
117.

118.

119.
120.

121.
122.

RE. Marth, H.I. Aaronson, G.W. Lorimer, T.L. Bartel and K.C. Russell: Metall. Trans., 1976, 7A,
1519-1528.
CM. Dallam and D.L. Olson: Weld. J., 1989, 68, 198s-2O5s.
G. Thewlis: HW Doc. IXJ-165-90 (1990).
R.A. Farrar, Z. Zhang, S.R. Bannister and G.S. Barrite: J. Mater. ScL, 1993, 28, 1385-1390.
S.S. Babu and H.K.D.H. Bhadeshia: Mater. ScL Technol, 1990, 6, 1005-1020.
H. Homma, S. Ohkita, S. Matsuda and K. Yamamoto: Weld. J., 1987, 66, 301s-309s.
R.E. Dolby: Weld. J., 1979, 58, 225s-238s.
B.C. Kim, S. Lee, NJ. Kim and D.Y. Lee: Metall Trans., 1991, 22A, 139-149.
R.H. Phillip: Weld. J., 1983, 62, 12s-18s.
O.M. Akselsen, J.K. Solberg, G. R0rvik and AJ. Paauw: Technical Report STF34 F87013, 1987,
Trondheim (Norway), Sintef-Division of Metallurgy.
A.O. Kluken and 0. Grong: Proc. Int. Conf. on Recent Trends in Welding Science and Technology,
Gatlinburg, TN, May, 1989, pp. 781-786. Publ. ASM International, Materials Park, OH, 1990.
O.M. Akselsen, G. R0rvik, M.I. Ons0ien and 0. Grong: Weld. J., 1989, 68, 356s-362s.
G.R. Speich, V.A. Demarest and R.L. Miller: Metall. Trans., 1981,12A, 1419-1428.
C.A.N. Lanzillotto and FB. Pickering: Metal ScL, 1982, 16, 371-382.
J.H. Chen, Y. Kikuta, T. Araki, M. Yoneda and Y. Matsuda: Acta Metall, 1984, 32, 1779-1788.
M. Ramberg, O.M. Akselsen and 0 . Grong: Proc. Int. Conf. on Advances in Welding Science and
Technology, Gatlinburg, TN, May, 1986, pp. 679-684. Publ. ASM International, Materials Park,
OH, 1986.
O.M. Akselsen, 0. Grong and J.K. Solberg: Mater. ScL Technol, 1987, 3, 649-655.
X, -L. Cai, AJ. Garratt-Reed and W.S. Owen: Metall. Trans., 1985, 16A, 543-557.
JJ. YI, LS. Kim and H.S. Choi: Metall. Trans., 1985, 16A, 1237-1245.
O.M. Akselsen, J.K. Solberg and 0 . Grong: Stand. J. Metall, 1988,17, 194-200.
A.R. Marder: Metall. Trans., 1981, 12A, 1569-1579.
X.P Shen and R. Priestner: Metall Trans., 1990, 21 A, 2547-2553.
NJ. Kim and YG. Kim: Mater. ScL Eng., 1990, A129, 35-44.
K.R. Kinsman, G. Das and R.F Hehemann: Acta Metall, 1977, 25, 359-365.
N.K. Balliger andT. Gladman: Met. ScL, 1981, 15, 95-108.
S. Kou: Welding Metallurgy, 1987, Toronto (Canada), John Wiley & Sons.
J.E. Hatch (ed.): Aluminium Properties and Physical Metallurgy, 1984, Metals Park (Ohio),
American Society for Metals.
FM. Mazzolani: Aluminium Alloy Structures, 1985, Boston (USA), Pitman Publish. Inc.
O. Lohne and A.L. Dons: Scand. J. Metall, 1983, 12, 34-36.
R.P Martukanitz: Proc. Int. Conf. on Advances in Welding Science and Technol, Gatlinburg
(Tenn.), May, 1986, pp. 193-201. Publ. ASM International, Materials Park (OH).
H.R. Shercliff, 0. Grong, O.R. Myhr and M.F Ashby: Proc. 3rd. Int. Conf. on Aluminium Alloys
Their Physical and Mechanical Properties, Trondheim (Norway), June, 1992, Vol. Ill, pp. 357369. Publ. The Norwegian Institute of Technology, Trondheim, Norway.
D.H. Bratland, 0. Grong, O.R. Myhr and H.R. Shercliff: Proc. Int. Conf on Computer-Assisted
Materials Design and Process Simulation, Tokyo (Japan), September, 1993, pp. 135-141. Publ.
The Iron and Steel Institute of Japan.
D.H. Bratland, 0 . Grong, H.R. Shercliff, O.R. Myhr and S.Tjotta:,4cta Metall Mater., Overview
No. 124,1997,45,1-22.
CM. Sellars: Proc. 3rd Int. Conf. on Aluminium AlloysTheir Physical and Mechanical
Properties, Trondheim (Norway), June, Vol. Ill, pp. 89-106. Publ. The Norwegian Institute of
Technology, Trondheim, Norway.
HJ. McQueen and JJ. Jonas: Treatise on Mater. ScL Technol, 1975,6, 393-493.
O.T. Midling and 0 . Grong: Acta Metall Mater., 1994,42,1595-1609: ibid., 1611-1622.

Appendix 6.1
Nomenclature
start temperature of ferrite to
austenite transformation (0C or
K)

concentration of solute at particle/matrix interface (at.% or


wt%)

end temperature of ferrite to


austenite transformation (0C or
K)

concentration of solute inside the


precipitates (at.% or wt%)

equilibrium 7/a solvus temperature (0C or K)


interfacial area per unit volume
(jLim2 per |um3 or m2 per m3)
material constant related to the
potency of the heterogeneous
nucleation sites (J mol"1)
difference between total and acid
soluble aluminium in weld metal
(wt%)

concentration of solute in matrix


(at.% or wt%)
maximum concentration of solute in matrix (at.% or wt%)
continuous cooling transformation
cooling rate (0C s"1 or K s 1 )
plate thickness (mm)
subgrain diameter (jum or m)

acicular ferrite
integration parameter

average diameter of columnar


austenite grains (\xm or m)

bulk concentration of free boron


in steel (ppm)

ferrite grain size (\xm or m)

amount of boron which segregates to austenite grain boundaries on cooling (ppm)


amount of boron which segregates to austenite grain boundaries in reference steel on cooling (ppm)

austenite grain size (jum or m)


austenite grain size in reference
material (jum or m)
element diffusivity (JLim2 s"1 or
!!TV1)
complementary error function

fraction of boron atoms associated with vacancies (ppm)


various kinetic constants and
temperature-dependent parameters

vacancy-boron binding energy (J


mol"1 or kJ mol"1)
vacancy formation energy (J
mor 1 or kJ mol"1)

electron backscattering pattern

arbitrary function of (T)

electron dispersive system

amperage (A)

volume fraction of transformation product

kinetic strength of thermal cycle


with respect to p-precipitation

equilibrium volume fraction of


transformation product

modified Z1 integral

ferrite with aligned second phase

integral in equation (6-47)

molar free energy of a-phase (J


mol"1 or J m"3)

kinetic constant in Avrami equation (variable units)

molar free energy of p-phase (J


mol"1 or J nr 3 )

time constant in Avrami equation


(s)

arbitrary function of X

Kurdjumow-Sachs orientation
relationship

molar free energy of transformation (J mol"1 or J m~3)

isothermal transformation

free energy donated to the system when nucleation occurs heterogeneously (J mol"1 or J nr 3 )
increase in strain energy
resulting from lattice distortion
(J mol 1 or J m 3 )
increase in surface energy due to
nucleus formation (J mol"1 or J
ITT3)

interledge spacing (nm)


length of columnar austenite
grains (|Lim or m)
lower bainite
proportionality constant in boron
diffusion model
martensite

volume free energy change associated with transformation (J


mol"1 or J m 3 )

mobility parameters related to


the jump-frequency of atoms
across a particle/matrix interface
(variable units)

energy barrier against heterogeneous nucleation (J or J mol"1)

end temperature of martensite


transformation (0C or K)

energy barrier against homogeneous nucleation (J or J mol"1)

start temperature of martensite


transformation (0C or K)

grain boundary ferrite


gas metal arc welding

martensite-austenite constituent
metal inert gas

time exponent in Avrami equation

two-dimensional radius vector in


y-z plane (mm)

number of nucleation sites per


unit grain boundary area (irr 2 or
finr 2 )

isothermal zone width referred


to the fusion boundary (mm)

Avogadro constant (6.022 X


1023 moH)
rate of heterogeneous nucleation
(m-3 s-1)

universal gas constant (8.314 J


mol-1 K"1)
grain boundary or particle surface area per unit volume (m2
per m3 or mm2 per mm 3 )

rate of grain boundary nucleation (rrr2 s"1 or CITT2 s"1)

shape factor related to the


wetting conditions at nucleus/
substrate interface

total number of heterogeneous


nucleation sites per unit volume
(rrr3)

submerged arc welding

total number of heterogeneous


nucleation sites per unit volume
in reference material (m~3)

shielded metal arc welding


scanning transmission electron
microscope
time (s)

Nishiyama-Wasserman orientation relationship


analytical weld metal oxygen
content (wt%)
probability

time taken to precipitate a certain


fraction of P at an arbitrary temperature T (s)
time taken to precipitate a certain fraction of (3 at a chosen reference temperature Tr (s)

polygonal ferrite

integration limits (s)

net arc power (W or kW)

cooling time from 600 to 4000C


(S)

activation energy for diffusion (J


mol"1 or kJ mol 1 )

cooling time from 800 to 5000C


(S)

radius of precipitates (nm, Jim or


m)

retention time (s)


temperature (0C or K)

maximum particle radius following prolonged room temperature


ageing (nm, fim or m)

undercooling or temperature interval (0C or K)

equilibrium solvus temperature


(0C or K)

integration parameter
half thickness of plate (jam, m)

ambient temperature (0C or K)


Zener-Hollomon parameter (s"1)
0

peak temperature ( C or K)
frequency factor (s"1)
chosen reference temperature
(0C or K)

interfacial energy between nucleus and matrix (J irr 2 )

naturally aged condition


wetting angle (degrees)
peak-aged condition
transmission electron microscope

non-hardening (metastable) precipitates in Al-Mg-Si alloys


hardening (metastable) precipitates in Al-Mg-Si alloys

voltage (V)
migration rate of incoherent
interface (|im s"1)

strain rate (s~!)

migration rate of semi-coherent


interface (jam s"1)

parabolic thickening constant for


plate-like precipitates (jam s~1/2
or m s-1/2)

upper bainite
1

welding speed (mm s" )

average value of E1 within a


specific temperature range
(|im s~1/2 or m s"1/2)

volume (|iim3, mm3, or m3)


Vickers pyramid number

parabolic thickening constant for


spherical precipitates (|im s~1/2
or m s"1/2)

Widmanstatten ferrite
arc efficiency
fraction transformed (corrected
for physical impingement of adjacent volumes)
fraction transformed (not corrected for physical impingement
of adjacent volumes)
fraction transformed defined by
the C-curve
maximum fraction of hardening
precipitates which form during
natural ageing

solid angle enclosed by one


<100>7_Fe-pole in a standard
stereographic projection of the
Bain region
measured electrical resistivity
(nQm)
electrical resistivity in solutedepleted matrix (nQm)
electrical resistivity in asquenched condition (nQm)

Appendix 6.2
Additivity in relation to the Avrami Equation
In order to prove that equation (6-32) is the isokinetic version of the Avrami equation, we first
need to rewrite it in a differential form. From equation (6-25), we have:
(A6-1)

Substituting
(A6-2)

(A6-3)

into equation (A6-1) then gives:

(A6-4)

(A6-5)
Provided that equation (A6-5) contains separable variables of X and t* (T), it can be integrated as follows:
(A6-6)

Taking b = 1 - XIn, y = 1 - X and dy = - dX, we get:

Rearranging equation (A6-7), gives:


(A6-8)
where
(A6-9)

Equation (A6-8) is identical with equation (6-32) in the text.

7
Properties of Weldments

7.1 Introduction
Weldments are prime examples of components where the properties obtained depend upon the
characteristics of the microstmcture. Since failure of welds often can have dramatic consequences, a wealth of information is available in the literature on structure-property relationships. However, in order to fit some of the apparently conflicting results into a more consistent
picture, a theoretical approach is adopted here rather than a review of the literature. This
procedure also involves the use of phenomenological models for the quantitative description
of structure-property relationships in cases where a full physical treatment is not possible.

7.2 Low-Alloy Steel Weldments


The symbols and units used throughout the chapter are defined in Appendix 7.1.
The major impetus for developments in high-strength low-alloy (HSLA) steels has been
provided by the need for: (i) higher strength, (ii) improved toughness, ductility, and formability,
and (iii) increased weldability. In order to meet these contradictory requirements, the steel
carbon contents have been progressively lowered to below 0.10 wt% C. The desired strength
is largely achieved through a refinement of the ferrite grain size, produced by the additions of
microalloying elements such as aluminium, vanadium, niobium, and titanium in combination
with various forms for thermomechanical processing.1 This procedure has made it possible to
improve the resistance of steels to hydrogen-assisted cold cracking, stress corrosion cracking,
and brittle fracture initiation in the weld heat-affected zone (HAZ) region, without sacrificing
base metal strength, ductility, or low-temperature toughness.2 Controlled rolled HSLA steels
are currently produced with a minimum yield strength in the range from 350-550 MPa. Above
this strength level, quenched and tempered steels are commonly employed.
72.1 Weld metal mechanical properties
The recent progress in steel plate manufacturing technology has, in turn, called for new developments in welding consumables to produce weld metal deposits with mechanical properties
essentially equivalent to the base metal.3 From the large volume of literature dealing with
HSLA steel filler metals, it appears that the bulk of weld metal research over the past decade
has been concentrated on the achievement of a maximum toughness and ductility for a given
strength level by control of the weld metal microstructure.34 There seems to be general agreement that microstructures primarily consisting of acicular ferrite provide optimum weld metal
mechanical properties, both from a strength and toughness point of view, by virtue of its high

dislocation density and small lath size. The formation of large proportions of upper bainite,
Widmanstatten ferrite, or grain boundary ferrite, on the other hand, are considered detrimental
to toughness, since these structures provide preferential crack propagation paths, especially
when continuous films of carbides are present between the ferrite laths or plates. Attempts to
control the weld metal acicular ferrite content have led to the introduction of welding
consumables containing complex deoxidisers (Si, Mn, Al, Ti) and balanced additions of various alloying elements (Nb, V, Cu, Ni, Cr, Mo, B).
The final weld metal microstructure will depend on complex interactions between several
important variables such as:3"5
(i)
(ii)

The total alloy content.


The concentration, chemical composition, and size distribution of non-metallic
inclusions.
(iii) The solidification microstructure.
(iv) The prior austenite grain size,
(v) The weld thermal cycle.
Although the microstructural changes taking place within the weld metal on cooling through
the critical transformation temperature range in principle are the same as those occurring during rolling and heat treatment of steel, the conditions existing in welding are significantly
different from those employed in steel production because of the characteristic strong nonisothermal behaviour of the arc welding process. For example, in steel weld deposits the
volume fraction of non-metallic inclusions is considerably higher than that in normal cast steel
products because of the limited time available for growth and separation of the particles. Oxygen is of particular interest in this respect, since a high number of oxide inclusions is known to
influence strongly the austenite to ferrite transformation both by restricting the growth of the
austenite grains as well as by providing favourable nucleation sites for various types of
microstructural constituents (e.g. acicular ferrite). Moreover, during solidification of the weld
metal, alloying and impurity elements tend to segregate extensively to the centre parts of the
interdendritic or intercellular spaces under the conditions of rapid cooling.67 The existence of
extensive segregations further alters the kinetics of the subsequent solid state transformation
reactions. Accordingly, the weld metal transformation behaviour is seen to be quite different
from that of the base metal, even when the nominal chemical composition has not been significantly changed by the welding process.3"5 This, in turn, will affect the mechanical integrity of
the weldment.
7.2.1.1 Weld metal strength level
In low-alloy steel weld metals there are at least four different strengthening mechanisms which
may contribute to the final strength achieved. These are:
(i)
(ii)
(iii)
(iv)

Solid solution strengthening,


Dislocation strengthening,
Grain boundary strengthening,
Precipitation strengthening.

The relative contribution from each is determined by the steel chemical composition and

the weld thermal history. Because of the number of variables involved, a full physical treatment of the problem is not possible. Consequently, the simplified treatment of Gladman and
Pickering8 has been adopted here.
Figure 7.1 shows the individual strength contributions in low-carbon bainite, which is the
dominating microconstituent in as-deposited steel weld metals (includes both upper and lower
bainite as well as acicular ferrite). Firstly, there are the solid solution strengthening increments
from alloying and impurity elements such as manganese, silicon and uncombined nitrogen,
which in the present example correspond to a matrix strength of about 165 MPa. Secondly, the
grain size contribution to the yield stress is shown as a very substantial component, the magnitude of which is determined by the bainite lath size. Finally, a typical increment for dispersion
strengthening is indicated. This contribution is negligible at large lath sizes typical of upper
bainite, but becomes significant at small grain sizes because of a finer intralath carbide dispersion.8 Hence, in steel weld deposits containing high proportions of acicular ferrite or lower
bainite carbides will make a direct contribution to strength, even at relatively low carbon levels.
The results in Fig. 7.1 are of significant practical importance, since they show the inherent
limitations of the system with regard to the maximum strength that can be achieved through
control of the microstructure. As shown in Section 6.3.5.4 (Chapter 6), the typical lath size
(width) of acicular ferrite in low-alloy steel weld metals is about 2 jam. According to Fig. 7.1,
this corresponds to a maximum yield strength of approximately 650 MPa, which is in good
agreement with the observed threshold strength of acicular ferrite containing steel weld deposits.9 If higher strength levels are required, it is necessary to decrease the grain (lath) size
through a refinement of the microstructure, i.e. by replacing acicular ferrite with either lower

Yield strength, MPa

p 1/4
-f/o
Number of carbides per mm Nv (mm" )

Dispersion

Grain size

Matrix strength

-1/2
Bainitic ferrite grain size, mm

Fig. 7.1. Contributions to strength in low-carbon bainite. Data from Gladman and Pickering.8

bainite or martensite. Development along these lines has led to the introduction of a new
generation of high strength steel weld metals with a yield strength in the range from 650 to 900
MPa.10'11
7.2.1.2 Weld metal resistance to ductile fracture
It is well established that the weld metal resistance to ductile fracture is strongly influenced by
the volume fraction, shape, and size distribution of non-metallic inclusions.12"15 Although a
verified quantitative understanding of the fracture process is still lacking, there seems to be
general agreement that it involves the following three basic steps:16
(i)

Nucleation of internal cavities during plastic flow, preferentially at non-metallic


inclusions.
(ii) Growth of these cavities with continued deformation,
(iii) Final coalescence of the cavities to produce complete rupture.
Details of these three stages may vary widely in different materials and with the state of
stress existing during deformation. Similarly, the fractographic appearance of the final fracture surface is also influenced by the same factors.
Effect of inclusion volume fraction
The primary variables affecting the true strain at fracture 8/ are the inclusion diameter dv, and
the inclusion volume fraction Vv. The relation between Ef and Vv has been determined experimentally for a wide variety of materials, and can most simply be expressed as:17
(7-1)
where c\ is an empirical constant.
The tensile test data of Widgery12 reproduced in Fig. 7.2 reveal a strong dependence of /
on Vv, but the relationship appears to be linear rather than non-linear, as predicted by equation
(7-1). Due to a similar fracture mechanism, a correlation also exists between the Charpy Vnotch (CVN) upper shelf energy and the true fracture strain in tensile testing, as shown in Fig.
7.3. For this reason, the weld metal impact properties are normally seen to decrease with
increasing oxygen concentrations when testing is performed in the upper shelf region. From
Fig. 7.4 we see that the CVN upper shelf energy is a linear function of the weld metal oxygen
content. This observation is not surprising, considering the fact that the inclusion volume
fraction is directly proportional to the oxygen level (see equation (2-75) in Chapter 2).
Effect of inclusion size distribution
Void nucleation may occur both by cracking of the inclusions and by interface decohesion. In
the former case, the critical stress for particle cracking ap is given by:16

(7-2)

where 7^ is the surface energy of the particle, Ep is the Young's modulus of the particle, A is the
stress concentration factor at the particle, and dv is the particle diameter.

True fracture strain

GMAW
(E=1.6kJ/mm)

Inclusion volume fraction

CVN upper shelf energy, J

Fig. 7.2. Variation of true fracture strain /with inclusion volume fraction Vv. Data from Widgery.12

SAW and FCAW

True fracture strain


Fig. 7.3. Correlation of CVN upper shelf energy with true fracture strain in tensile testing. Data from
Akselsen and Grong.20
Equation (7-2) predicts that large inclusions will tend to form voids first as the stress required for initiation is proportional to (l/dv )1/2. This result is also in agreement with experimental observations. As shown in Fig. 7.5, the size distribution of inclusions located in the
centre of voids at the fracture surface is significantly coarser than the corresponding particle
size distribution in the material. In particular, large, angular shaped aluminium oxide (AI2O3)

CVN upper shelf energy, J

SAW

Oxygen content, wt%


Fig. 7.4. Correlation of CVN upper shelf energy with analytical weld metal oxygen content. Data from
Devillers et aL 13

inclusions appear to be preferential nucleation sites for microvoids in low-alloy steel weld
metals (see Fig. 7.5(b)). Although the combined effect of particle size and local stress concentration on the ductile fracture behaviour cannot readily be accounted for in a mathematical
simulation of the process, the CVN data in Fig. 7.6 suggest that the content of large inclusions
(e.g. of a diameter greater than about 1.5 Jim) should be minimised in order to maintain a high
resistance against dimpled rupture. In practice, this requires careful control of the weld metal
aluminium-oxygen balance and the heat input applied during welding (see Section 2.12 in
Chapter 2).
Effect of strength level
The toughness of a material reflects its ability to absorb energy in the plastic range. One way of
looking at toughness is to assume that it scales with the total area Uj under the stress-strain
curve. Several mathematical expressions for this area have been suggested. For ductile materials we may write:19
(7-3)
where Rm is the ultimate tensile strength (UTS).
If Uj is regarded as a material constant, one would expect that Rm and y are reciprocal

Total inclusion
population

SAW

Frequency, %

Frequency, %

Inclusions associated
with dimples

Inclusion diameter, jum

Frequency, %

Total inclusion
population

(b)

SAW

Inclusions associated
with dimples

Frequency, %

(a)

Inclusion diameter, jum

Fig. 7.5. Histograms showing the size distribution of non-metallic inclusions in the weld metal and in the
centre of microvoids at the fracture surface, respectively; (a) Low aluminium level (Al-containing manganese silicate inclusions), (b) High aluminium level (AI2O3 inclusions). Data from Andersen.18

High Ti levels
CVN upper shelf energy, J

Medium Ti levels
Low Ti levels

SAW

Nv (d v >1.5 um)-105
Fig. 7.6. Correlation of CVN upper shelf energy with number of particles per mm3 greater than 1.5 urn,
Nv(dv > 1.5 um). Data from Grong and Kluken.15

True fracture strain

SAW and FCAW

Ultimate tensile strength, MPa


Fig. 7.7. Correlation of true fracture strain with ultimate tensile strength (low-alloy steel weld metals).
Data from Akselsen and Grong.20
quantities, i.e. an increase in Rm is always associated with a corresponding decrease in Ef,
according to the equation:
(7-4)
where c^ is a constant which is characteristic of the alloy system under consideration.

It is evident from the tensile test data in Fig. 7.7 that the fracture strain is a true function of
Rm, although the relationship appears to be linear rather than non-linear, as predicted by equation (7-4). These results are of considerable practical importance, since they imply that the
upper shelf energy absorption, and hence, the shape of the CVN transition curve is strongly
affected by the weld metal strength level. Accordingly, control of the weld metal microstructure becomes particularly urgent at high strength levels to avoid problems with the cleavage
fracture resistance (to be discussed below).
7.2.1.3 Weld metal resistance to cleavage fracture
Cleavage fracture is characterised by very little plastic deformation prior to the crack propagation, and occurs in a crystallographic fashion along planes of low indicies, i.e. of high atomic
density.1 Body-centred cubic (bcc) iron cleaves typically along {100} planes, which implies
that the cracks must be deflected at high angle grain (or packet) boundaries, as shown
schematically in Fig. 7.8. Consequently, in steel weld metals the ferrite grain size and the
bainite packet width are the main microstructural features controlling the resistance to cleavage crack propagation.
Since the microstructure which forms within each single austenite grain will not be uniform
but a complex mixture of two or more constituent phases, it is difficult, in practice, to define
a meaningful grain size or packet width. For this reason, most investigators have attempted to
correlate toughness with the presence of specific microconstituents in the weld metal.3"5 For
example, an increase in the volume fraction of acicular ferrite will result in a corresponding
increase in toughness (i.e. decrease in the CVN transition temperature), as shown in Fig. 7.9.

(a)

(b)
Fig. 7.8. Schematic diagrams showing cleavage crack deflection at interfaces; (a) High angle ferriteferrite grain boundaries, (b) High angle packet boundaries (bainitic microstructures).

Transition temperature, 0C

SAW (E = 5.2-6.2 kJ/mm)

Al: 0.018-0.062 wt%


Ti: 0.005 - 0.065 wt%
O: 0.018-0.058 wt%

Acicular ferrite content, vol%


Fig. 7.9. Correlation between the weld metal 35J CVN transition temperature and the acicular ferrite
content. Data from Grong and Kluken.15
This observation is not surprising, considering the extremely fine lath size of the acicular
ferrite microstructure (typically less than 5jim).
Moreover, results obtained from fractographic examinations of SMA and FCA steel weld
metals have demonstrated that large non-metallic inclusions (> l|im) can strongly influence
the cleavage fracture resistance, either by acting as cleavage cracks themselves of by providing internal sites of stress concentration which facilitate carbide-initiated cleavage in the adjacent matrix.21'22 In the former case, the critical stress required for crack propagation in the
matrix, Cf(M), is given by the Griffith's equation:19

(7-5)

where En is the Young's modulus of the matrix, ye^ is the effective surface energy (equal to
the sum of the ideal surface energy and the plastic work), and c is the half crack length.
Since c is proportional to the particle diameter dv, equation (7-5) predicts that welds containing large inclusions should be more prone to cleavage cracking than others. This result is
also in agreement with general observations. For example, in self-shielded FCA steel weld
metals it has been demonstrated that cleavage crack initiation is usually associated with large
aluminium-containing inclusions which form in the molten pool before solidification (see Fig.
7.10). Consequently, control of the inclusion size distribution is essential in order to ensure an
adequate low-temperature toughness.
7.2.1.4 The weld metal ductile to brittle transition
In addition to the parameters mentioned above, there are several other factors, some interrelated, which play an important part in the initiation of cleavage fracture. These are:1
(i)

The temperature dependence of the yield stress.

(a)

(b)

(C)

Fig. 7.10. Initiation of cleavage fracture in a self-shielded FCA steel weld from an aluminium-containing
inclusion; (a) Initiation site short distance ahead of the notch, (b) Detail of initiation site showing cracked
inclusion, (c) Detail of cracked inclusion (remnants of particle are left in the hole).
(ii)
(iii)
(iv)

Dislocation locking effects caused by interstitials or alloying elements in solid


solution (e.g. nitrogen and silicon),
Nucleation of cracks at twins,
Nucleation of cracks at carbides.

In general, this picture is too complicated to establish a physical framework within which
the various theoretical models for the ductile to brittle transition in steel can be embedded. We
are therefore forced to base our judgement and understanding of how key parameters affect the
position and shape of the CVN transition curve solely on scattered phenomenological observations and empirical models (e.g. see the reviews of Grong and Matlock3 or Abson and Pargeter4).
An example of how far the latter approach has been developed is given below.
Akselsen and Grong20 have established a series of empirical equations which relates toughness to the weld metal acicular ferrite content and the ultimate tensile strength (UTS). Figures
7.11 and 7.12 show how each of these parameters influences the CVN transition curve. It is
evident from the diagrams that control of the weld metal acicular ferrite content becomes
particularly important at high strength levels to avoid problems with the fracture toughness. In
cases where undermatch is aimed at (i.e. a weld metal to base plate strength ratio less than
unity), the weld metal tensile strength is typically of the order of 450 to 550 MPa. Within this
range a volume fraction of acicular ferrite beyond 25 vol% will generally be sufficient to meet
current toughness requirements (35 J at -40 0 C). If overmatch is desired, the volume fraction
of acicular ferrite becomes more critical, partly because of a higher weld metal strength level
and partly because of more stringent toughness requirements (e.g. 45 J rather than 35 J at
-40 0 C). Process diagrams of the type shown in Figs. 7.11 and 7.12 can therefore serve as a
basis for proper selection of consumables for welded steel structures.
It should be noted that Akselsen and Grong20 in their analysis omitted a consideration of the
important influence of free (uncombined) nitrogen and non-metallic inclusions on the CVN
transition curve. Based on the experimental data in Fig. 7.13 it can be argued that such
compositional variations can be equally detrimental to toughness as a decrease in the acicular
ferrite content. Consequently, further refinements of the models are required if a verified
quantitative understanding of the ductile to brittle transition in low-alloy steel weld metals is
to be obtained.
Example (7.1)

Consider multipass FCA steel welding with two different electrode wires, one with titanium
additions and one without. Table 7.1 contains a summary of weld metal chemical compositions. Provided that the microstructure and the inclusion size distribution are similar in both
cases, use this information to evaluate the low-temperature toughness of the welds, as revealed
by CVN testing.
Solution

Since the nitrogen content of both welds is quite high (0.011 wt%), the risk of a toughness
deterioration due to strain ageing is imminent, particularly at low Ti levels. Taking the atomic
weight of titanium and nitrogen equal to 47.9 and 14.0 g mol"1, respectively, the stoichiometric
amount of titanium that is necessary to tie-up all nitrogen as TiN can be calculated as follows:

WeIdA
In weld A most of the nitrogen is free (uncombined) due to an unbalance in the titanium content. This means that the risk of a toughness deterioration due to strain ageing is high.

(a)

Absorbed energy, J

Tensile strength: 600 MPa


Vol% acicular ferrite

35 Joules

Test temperature, 0C

(b)

Absorbed energy, J

Tensile strength: 800 MPa

Vol% acicular ferrite

35Joules.

Test temperature, 0C
Fig. 7.11. Predicted effect of weld metal acicular ferrite content on the CVN transition curve at two
different tensile strength levels; (a) Rm = 600 MPa, (b) Rm = 800 MPa. Data from Akselsen and Grong.20

WeIdB
Weld B contains 0.030 wt% Ti, which is not far from the stoichiometric amount of titanium
necessary to tie-up all nitrogen. Although some titanium also is bound as Ti2O3, it is reasonable to assume that the free nitrogen content in this case is sufficiently low to eliminate problems with strain ageing. Consequently, weld B would be expected to exhibit the highest toughness (i.e. the lowest CVN transition temperature) of the two, as indicated in Fig. 7.14.

(a)
25 vol% acicular ferrite
Absorbed energy, J

UTS

35 Joules

Test temperature, 0C

(b)

Absorbed energy, J

75 v o l % acicular ferrite

UTS

-..35J.Q.ute$-

Test temperature, 0 C

Fig. 7.12. Predicted effect of weld metal ultimate tensile strength (UTS) on the CVN transition curve at
two different volume fractions of acicular ferrite; (a) 25 vol% AF, (b) 75 vol% AF. Data from Akselsen
and Grong.20
Table 7.1 Chemical composition of FCA steel weld metals considered in Example (7.1).
Element
Weld

wt% C

wt% Si

wt% Mn

wt% Al

wt% Ti

wt% S

wt% N

wt% O

0.10

0.40

1.50

0.005

0.006

0.008

0.011

0.031

0.10

0.40

1.50

0.005

0.030

0.008

0.011

0.031

(a)
SMAW (basic electrodes)
Absorbed energy, J

95% confidence interval

Testing temperature: -400C

Nitrogen content, ppm


(b)

Absorbed energy, J

Low content of coarse inclusions

High content of coarse


inclusions ( > 1}im )

Self-shielded FCA steel


weld metals
Test temperature, 0C
Fig. 7.13. Effect of impurities on weld metal CVN toughness; (a) Nitrogen content, (b) Inclusion level.
Data from ESAB AB (Sweden) and Grong et al. 22

7.2.1.5 Effects of reheating on weld metal toughness


In principle, improvement of weld properties can be achieved through a post-weld heat treatment (PWHT). This may have the benefits of:3
(i)

Enhancing the fatigue strength through a general reduction of welding residual


stresses.

Absorbed energy

WeIdB

WeIdA

35 J

Test temperature, 0C
Fig. 7.14. Schematic drawings of the CVN transition curves for welds A and B (Example (7.1)).

(a)

(b)

Fig. 7.15. Typical low-temperature fracture modes of Ti-B containing steel weld metals; (a) Quasicleavage (as-welded condition), (b) Intergranular fracture (after PWHT).

(ii)

Increasing the toughness by recovery (i.e. removal of strain-aged damage) and


martensite tempering.

For these reasons local PWHTs are commonly required for all structural parts above a
specified plate thickness (e.g. 50 mm according to current North Sea offshore specifications).
Post-weld heat treatment is usually carried out in the temperature range from 550 to 6500C.
In practice, however, the toughness achieved will depend on the weld metal chemical composition, and in some cases deterioration rather than improvement of the impact properties is
observed after PWHT. In such cases the reduction in toughness can be ascribed to:3'4
(i)

Precipitation hardening reactions. Present experience indicates that elements such


as vanadium, niobium, and presumably titanium can produce a marked deterioration in toughness because of precipitation of carbonitrides in the ferrite, provided
that these elements are present in the weld metal in sufficiently high concentrations.

(ii)

Segregation of impurity elements (e.g. P, Sn, Sb and As) to prior austenite grain
boundaries, which, in turn, can give rise to intergranular fracture. The indications
are that this type of embrittlement is strongly enhanced by the presence of second
phase particles at the grain boundaries.

Experience shows that Ti-B containing steel weld metals often fail by intergranular fracture in the columnar grain region after PWHT,23 as evidenced by the SEM fractographs in Fig.
7.15. The observed shift in the fracture mode is associated with a significant drop in toughness
(Fig. 7.16) and arises from the combined action of solidification-induced phosphorus
segregations and borocarbide precipitation along the prior columnar austenite grain bounda-

ACVN, J

SAW

Open symbols: 5 - 8 ppm B


Filled symbols: 20 - 25 ppm B

Base line

Titanium content, wt%


Fig. 7.16. Observed displacement in the CVN toughness after PWHT (ACVN) as a function of the weld
metal boron and titanium contents. Negative values indicate loss of toughness. Data from Kluken and
Grong.23

Fig. 7.17. TEM micrograph showing precipitation of borocarbides, Fe23(B,C)6, along prior austenite
grain boundaries in a Ti-B containing steel weld metal after PWHT (600 0 C-Ih).

ries (Fig. 7.17). Since borocarbides are brittle and partly incoherent with the matrix, they can
be regarded as microcracks (of length dp) ready to propagate. In such cases there is virtually
no plastic deformation occurring before crack propagation, which implies that the intergranular
fracture stress is given by the Griffith's equation:24
(7-6)

where 7 ^ is again the effective surface energy (equal to the sum of the ideal surface energy
and the plastic work), and dp is the particle diameter.
Although the value of yeg_ would be expected to be low in the presence of solidificationinduced phosporus segregations,24 this alone is not sufficient to initiate intergranular fracture
in the weld metal. However, during PWHT the borocarbides will start to grow from an initially small value to a maximum size of about 0.1 to 0.2jim (Fig. 7.17), following the classic
growth law for grain boundary precipitates dpatl/4?5 This implies that the intergranular fracture stress, Oj(I), will gradually decrease with increasing annealing times, as indicated in Fig.
7.18. When the matrix fracture strength, Cj(M), is reached, the fracture mode shifts from
predominantly quasi-cleavage in the as-welded condition (Fig. 7.15(a)) to intergranular rupture after PWHT (Fig. 7.15(b)). This is observed as a marked reduction in the CVN toughness,
as shown previously in Fig. 7.16.
7.2.2 HAZ mechanical properties
The last twenty years have seen a revolution in the metallurgical design of steel. Whereas old
steels relied on the use of carbon for strength, the trend today is to rely more on grain refinement in combination with microalloy precipitation to meet the current demand for an improved weldability. This includes both the sensitivity to weld cracking and the HAZ mechanical properties required by service conditions and test temperatures. The latter aspect is of
particular interest in the present context and will be discussed later.

Stress

Intergranular
fracture mode

Quasi-cleavage
fracture mode
Particle diameter

[Annealing time]174
Fig. 7.18. Schematic illustration of the mechanisms of temper embrittlement in Ti-B containing steel
weld metals (Gf(M): matrix fracture strength, (*/(/): intergranular fracture strength).

p0.2> R nv M P a
R

HV 5 ,kp/mm 2

Martensite content, vol%

7.2.2.1 HAZ hardness and strength level


The HAZ hardness and strength level is of significant practical importance, since it influences
both the cracking resistance and the toughness. Although the peak strength is mainly controlled by the martensite content (see Fig. 7.19), the relationship is generally too complicated to
allow reliable predictions to be made from first principles. This implies that our understanding
of the HAZ strength evolution, at best, is semi-empirical.

Cooling time, At 8 / 5 , s
Fig. 7.19. Structure-property relationships in the grain coarsened HAZ of low-carbon microalloyed steels
(vol% M: martensite content, Rp : 0.2% proof stress, Rm: ultimate tensile strength, HV5: Vickers hardness, A%5.* cooling time from 800 to 5000C). Data from Akselsen et al.26

A number of different empirical models exist in the literature for prediction of HAZ peak
hardness and strength.26"31 However, the aptness of some of these models is surprisingly
good, which justifies construction of iso-hardness and iso-strength diagrams for specific grades
of steels.32 Examples of such diagrams are given in Fig. 7.20. It is