Abstract
A method based on nucleation theory but using fitted constants was used to model the time-temperature-transformation
(TTT) diagrams of steels. The transformation-start C-curves on the diagrams were represented as functions of the free
energy of nucleation and of an activation energy. The subsequent increase in volume fraction with time is treated using the
Johnson-Mehl-Avrami equation. These results, and published methods for determination of the variety of transforma-
tion-start temperatures, are demonstrated to estimate properly the published TTT diagrams for a large variety of
commercial steels containing C, Mn, Si, Ni, Mo and Cr as solute additions.
temperature; (3) the martensite-start M s temperature; The martensite-start temperature M s was predicted
(4) a set of C-curves describing various degrees of according to ref. 26. Martensite is considered to form
reconstructive transformation (allotriomorphic ferrite, when the driving force for diffusionless transformation
pearlite); (5) a set of C-curves describing various achieves a critical value which is a function of the
degrees of displacive transformation (Widmanst~itten carbon concentration of the parent austenite [26]. The
ferrite, bainite). driving force for diffusionless transformation was
calculated as in ref. 22.
2.1. Transformation temperatures
The Ae 3 temperature was calculated in each case 2.2. Transformation start and finish C-curves
according to the method of Kirkaidy and coworkers The equation that has been used previously to repre-
[18-20]. The bainite and Widmanst~itten ferrite-start sent the incubation period for a specified proportion of
temperatures can be estimated according to thermo- transformation is [11]
dynamic methodology discussed elsewhere [11,
21-23], but a simpler and equally reliable procedure is
adopted here.
(5)
It was emphasized earlier that the TTT diagram can
be described in terms of two C-curves, the lower curve where T is the absolute temperature, R is the universal
representing the Widmanst~itten ferrite and bainite gas constant, Gmax is the maximum free energy change
reactions. Both bainite and Widmanst~itten ferrite available for nucleation [11 ], Q is an activation energy,
nucleate by the same mechanism; the nucleus develops and C, p and z = 20 are empirical constants obtained
into Widmanst~itten ferrite if at the transformation by fitting to experimental data.
temperature the driving force available cannot sustain The above function has been demonstrated to
diffusionless transformation. By contrast, bainite forms reproduce the shape of C-curves rather well, and was
from the same nucleus if transformation can occur consequently employed here to describe the transfor-
without diffusion. Thus, the highest temperature 1"h of mation-start and finish curves. By varying p and plot-
the lower C-curve corresponds either to the Widman- ting ln(rGmaxP/T2°) against 1/RT it is possible to
st~itten ferrite-start or the bainite-start temperature. choose an optimum value of p for each transformation
These ideas are due to relatively recent studies of the curve by maximizing the regression coefficient R~.
mechanisms of transformation, whereas the experi- Once p has been determined, Q and C follow from
mentally determined TTT diagrams do not address respectively the slope and intercept of the plot.
such issues and always identify 1"h with B s. To be con- It is worth noting that in most cases, the bainite
sistent with this, we also refer to Th as the bainite-start finish C-curve of the experimental TTT diagrams not
temperature whilst recognizing that it can in principle only shifts to longer times relative to the corresponding
also describe the Widmanst~itten ferrite-start tempera- curve for the start of the reaction, but is also depressed
ture. The temperature was calculated using the empiri- to lower temperatures (eqn. (1)). This is taken into
cal equation of Steven and Haynes [24]. account by plotting ln[rGmJ/(T+120) 2°] against
The bainite transformation does not reach comple- l/R( T+ 120) in order to determine the values of p, Q
tion if the austenite enriches with carbon. However, in and C for the bainite finish C-curve.
many steels, carbide precipitation from austenite
eliminates the enrichment and allows the austenite to 2.3. Fractional transformation curves
transform completely. In those circumstances it is To estimate the increase in the volume fraction of
appropriate to discuss a "bainite finish" temperature, bainite as a function of time, a Johnson-Mehl-Avrami
which according to Leslie [25] is given approximately equation was adopted, as developed by Umemoto et al.
by [8]:
B~(°C) = B s - 120 (1)
Representations of intermediate states of transforma- x =l-exp~ KLt"
d']I (6)
tion between 0% and 100% were derived by fitting to
the experimental TTT diagrams as follows:
where x is the transformed volume fraction, K, is a rate
Bx(°C) = B s - 300x x<0.1 (2)
constant which is a function of temperature and d is the
Bx(°C)=Bs-30-75(x-O.1) 0.1 < x < 0 . 9 (3) austenite grain size; n and m are empirical constants.
Since all the steels analysed in the present work have
Bx(°C) = B s - 90 - 300(x - 0.9) 0.9 < x< 1 (4)
rather similar austenite grain sizes (within the range
where x refers to the fraction of transformation. ASTM No. 7-8), the grain size effect can be neglected,
J.-L. Lee, tt. K. D. tt. Bhadeshia / Prediction of TTT diagrams 225
/
1.01
eqn. (5), the values of K~ and n can be solved for each
transformation product as a function of temperature.
Fractional transformation curves for arbitrary values of 0.9
"U
x can therefore be determined using eqn. (7).
8
¢3
0.8
e'-.
o EN18
3. Determination of the fitting parameters .r'~
• EN16
0.7
o ENI4
The experimental data listed in ref. 10 were selected
for the determination of p, Q and C. The chemical 0.6 i I I t
The p parameter
Before determining the best fit for the parameters Q T A B L E 2. Optimum values of fitting constants, the concentra-
and C, it is necessary to determine the value of p tions are in weight per cent
in eqn. (5). Figure 1 illustrates how the correlation
FS p=4, C = - 155-9Ni-30Mo,
coefficient R~ varies as a function of p for steels Enl 4, Q = 3 3 1 3 4 0 + 83000Ni + 2 5 5 0 0 0 M o + 1 2 9 0 0 0 C r
En 16 and En 18. A value of p = 4 was selected for the FF p = 8 , C = - 1 5 5 - 9 N i - 30Mo,
start curve of ferrite formation since R~ ceases to be Q = 5 9 6 7 5 0 + 78000Ni + 265(100Mo + 115000Cr
sensitive to p beyond a value of 4. The results of similar - 110000C
BS p=5, C = -133,
determinations for the other curves are presented in
Q = 2 1 9 8 8 0 + 10000C + 6000Ni - 3 0 0 0 M o
Table 2. + 5000Cr
BF p=5, C= -133,
Q = 2 3 4 8 6 0 + 2 7 0 0 0 C + 40(l(JNi + 11000Cr
Q and C values) seemed to vary with steel chemistry of the published TTT curve data. The results are
(Fig. 3). Thus, it was found necessary to express these presented in Table 2.
parameters as a function of the chemical composition
of the steel in order to obtain acceptable representation
3.3. Other restrictions
900 . . . . I " " I .... I " " " t ..... I ...... I ....
The TTT diagrams of ref. 9 show that for most of
E N I 4 the steels, the bainite C-curve has a plateau at its
highest temperature, and this was incorporated by con-
800 Ae~ -
structing a horizontal line at the B s temperature and
joining it to a C-curve calculation for temperatures
700 _ , ' / ...... i'!"-' ' - -':"-".............. below B~. A similar construction was used for other
t.)
"7. 10 8
0
E E Fs
,-j 9' 7-
u') u-i
ov- 8" v- 6
7" ideallty
0 0 ~ ,/~J tdeahty
"0
5
•oQ 6" 0
P >
• 5" 4
J~ J~
0 0
4 3
4 5 6 7 8 9 10 3
(a) Predicted Q / 10 5 j moll (b) Predicted Q / 10 5 j mol-1
0 2.2 0 2.4
2.2 2.3 2.4 2.4 21s 216 2.r
(c) Predicted Q / 10 5 j mol-1 (d) Predicted O / 10 5 d mol-1
Fig. 4. Relation between the observed and predicted Q (J mol- ') value for (a) FS, (b) FF, (c) BS and (d) BF curves.
J.-L. Lee, tt. K. D. tt. Bhadeshia / Prediction of 77Tdiagrams 227
700 700
iiii
L) C.)
o o
(i,"I ~" .
rn 600 ,.~ 600
B_ B,
~ 500 500
n.
400 400
M,
M,
300 300
ENI4 ENI6
....... I ....... I ....... I ....... J ....... J ...... J ..... ..... I ........ I ........ I ........ I ........ I ....... ~ ......
1.0E+00 1.0E401 i . 0 E + 0 2 1.0E+03 1.0E+04 i . 0 E + 0 5 1.0E+06 1.0E+00 1.0E+01 1.0E+02 1.0E+03 1.0E+04 1.0E+05 1.0E+06
700 700
0 {A
600 600
;D
[-.,
s00 n.
500
04
[..,
4O0 400
MI
300 300
ENI8 ENIIO
Fig. 5. Comparison of the experimental and predicted q T r diagrams for BS steels: (a) En 14, (b) En 16, (c) En 18 and (d) En I 10.
b e l o w the B s t e m p e r a t u r e , especially for the high alloy exaggerated for these two alloys. The difficulty lies in
steels illustrated in Figs. 5(b)-(d). T h e estimates are the fact that although steel EN19 and steel US4140
better than the earlier m o d e l of B h a d e s h i a [11 ]. This have almost identical chemical compositions, their
better a c c u r a c y also applies to the lower alloy steels reported TTT curves are quite different (Fig. 7). This is
w h e r e the bay region is (correctly) not e x a g g e r a t e d as a contradiction which needs to be resolved with further
was the case in ref. 11. experiments.
T h e r e is also a r e a s o n a b l e fit with A m e r i c a n specifi- The prediction of fraction transformation curves
cation steels (Fig. 6), although not for alloys U S 4 1 4 0 using eqn. (7) is illustrated in Fig. 8; the agreement with
and U S 5 1 4 0 . T h e calculated bay region is highly experimental data is found to be reasonable.
228 J.-L. Lee, H. K. D. H. Bhadeshia / Predictionof TTT diagrams
9~0 .... l " " l - [ " I I " I 900 " " I 'I " i " " i " I " l
.......... E X P E R I M E N T A L .......... EXPERIMENTAL
800 _ _ PREDICTED
Ae 800 - PREDICTED
A%
700 700
¢.J
o
ta 600 600
500
400 400
M.
300 300 M, ,,
........ I ....... L ...... I , , .._.] ........ I ........ I ....... J ........ I ....... J ....... J .... I ...... I
l 0E+0o 1.OE+Ol 1.0E+e 1.0E+03 ,.0E+04 1.0E+0S .OE+e l .~-001 L.oE,,oo t.o --,ol l. E+02 1. +03 1.0E+04 1.0E+0S 1.0E+0 ,
(a) TIME / S (b) TIME / S
900 .... I " I " i " I " I ..... i ' ] 900 I " I I I ..... q . . . . 1
800
..........
_ _
EXPERIMENTAL
PREDICTED t 000
.......... EXPERIMENTAL
PREDICTED
A%
700
(--
.................
<.-- ~-
~ 1 .
°'*
o
m b00 b00
/"//" ...
[.-.,
500 ~- ~,"'..........
"'"
J ....................... *''J --i ~ 500
L_ . . . . . I ...... I ....... I ....... I ....... J . . . . . . . ...... J ....... I ....... I ....... J ....... I ...... J .....
1.~01 i.OE~O I.OD0I 1.0E+02 1.0E+03 i.0E+04 1.0E+05 I.OE+06 1.~2~81 1.0E+08 I.OE+81 1.0E+02 I.eE+03 1.0E+04 i.0E+85 1.0E+06
(C) TIME / S (d) TIME / S
Fig, 6. C o m p a r i s o n of the experimental and predicted TTT diagrams for US steels: (a) 4140, (b) 4150, (c) 4340 and (d) 5140.
900 ..... I ..... I .... I "1 " "1 .... I ..... 900 --- i i i i l ~ I
600 B,__=
iv,
s00
04 ( .-- .......... -- .........
40O
Mo
3OO
~.2°~E-°0~~.0e;;0'0i.~e;~-'e5 i.i~;i;'6
i.0E+00 t.OE-,-Ol 1.0E+02 t.0E+03 1.0E+94 1.0E+O5i.0E~-O6 (a] TIME / S
TIME / S
900 "'" I " I I " I " I - " 1 ""
Fig. 7. C o m p a r i s o n o f e x p e r i m e n t a l T T T d i a g r a m s f o r s t e e l s
. . . . . . . . . . EXPERIMENTAL
E n 1 9 a n d 4141). T h e s e t w o s t e c l s h a v c a l m o s t i d e n t i c a l c o m p o s i -
tion. 800 _ _ PREDICTED
Ae;
700 O. I _ _ ~
L~
o
o7o ,,
r.a b00
2~1 - t;,/2
3Og
5.3. Some limitations
EN24
The model tends to overestimate the transformation
...... d ....... J ........ I ........ I ....... J ........ I
time at temperatures just below the A e 3 temperature. l.l~-OO'I l.Oe+o0 I.OE÷Ol 1.0E+02 1.0E+83 1.0[+84 1.0[+85 i.0E+06
This is because the driving force t e r m A-Gmax is calcu- (b) TIME / S
lated on the basis of paraequilibrium, and becomes
zero at some temperature less than the Ae 3 tempera- Fig. 8. Comparison of the experimental and predicted TTT
diagrams including fractional transformation curves at 0.1, 0.5
ture. and 0.9 transformedfractions:(a) En 19 and (b) En24.
The coefficients utilized in the calculations were
derived by fitting to experimental data, so that the
model may not be suitable for extrapolation outside of
that dataset. Thus, calculations should be limited to the
following concentration ranges (in weight per cent): C it should nevertheless be useful in procedures for the
0.15-0.6; Si (I.l 5-0.35; Mn 0.5-2.0; Ni 0.0-2.0; Mo calculation of microstructure in steels.
0.0-0.8; Cr 0.0-1.7.
Acknowledgments
6. Conclusions
The authors are grateful to Professor C. J.
A procedure has been developed which enables the Humphreys for the provision of laboratory facilities at
estimation of the time-temperature-transformation the University of Cambridge. The work of J.L.L. was
diagram of steels, as a function of the alloy chemistry. It funded by the China Steel Corporation and the
is recognized that the model is empirical in nature, but National Science Council of Taiwan (ROC).
230 J.-L. Lee, H. K. D. H. Bhadeshia / Prediction of TTT diagrams
900 " " I " 1 " " J . . . . I " I " " I . . . . . 3 H. K. D. H. Bhadeshia, L.-E. Svensson and B. Gretoft, Acta
.......... 0.3C,0.SMn Metall., 33(1985) 1271-1283.
4 L.-E. Svensson, B. Gretoft and H. K. D. H. Bhadeshia,
800 0.3C,I.SMn ............. .A% Scand. J. Metall., 15 (1986) 97-103.
5 D. E Watt, L. Coon, M. Bibby, J. Goldak and C. Henwood,
Acta Metall., 36 (1988) 3029-3035.
700 6 E L. Harrison and R. A. Farrar, Int. Mater. Rev., 34 (1989)
o 35-51.
7 J.S. Kirkaldy, Metall. Trans., 4(1973) 2327-2333.
~a b00 8 M. Umemoto, N. Nishioka and I. Tamura, Trans. Iron Steel
[.., Inst. Jpn., 22 (1982) 629-636.
9 K.C. Russell, Acta Metall., 17(1969) 1123-1131.
~500 / _