Coating
Alex Williams
i
Acknowledgments
Prof. H., Benkreira for his support and guidance throughout this project. Dr., R.P., Alpin
for the work carried out during his thesis that has enabled this work to be possible.
ii
Contents
Abstract i
Acknowledgments ii
List of Figures iv
List of Tables v
1 Introduction 1
1.1 Coating of the Modern World . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Coating Flows: Principles . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Fluid Mechanics of Coating Flows . . . . . . . . . . . . . . . . . . . . 7
1.4 Aims and Objectives of this Study . . . . . . . . . . . . . . . . . . . 10
2 Literature Review 11
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Single Layer Down a Slide . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 Dual Layer Down a Slide . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4 Triple and Multilayer Down a Slide . . . . . . . . . . . . . . . . . . . 16
4 Data Correlation 34
4.1 Single Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2 Dual Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.3 Triple Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Appendix A Listings 43
iii
LIST OF FIGURES
Appendix B Correlations 47
List of Figures
3.1 Measured height versus predicted height for the single layer 9.5 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Measured height versus predicted height for the single layer 60 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3 Measured height versus predicted height for the single layer 200 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.4 Height ratio versus Reynolds number for all points of the 200 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.5 Measured height versus predicted height for the dual layer 9.5 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.6 Measured height versus predicted height for the dual layer 60 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.7 Measured height versus predicted height for the dual layer 200 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.8 Measured height versus predicted height for the dual layer all fluids. 27
3.9 Measured height versus predicted height for the dual layer 60/9.5 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.10 Measured height versus predicted height for the dual layer 200/60 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
iv
3.11 Measured height versus predicted height for the dual layer 9.5/60 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.12 Measured height versus predicted height for the dual layer 9.5/200 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.13 Measured height versus predicted height for the dual layer 60/200 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.14 Measured height versus predicted height for the triple layer 9.5 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.15 Measured height versus predicted height for the triple layer 60 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.16 Measured height versus predicted height for the triple layer 9.5/60/200 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.17 Measured height versus predicted height for the triple layer 9.5/200/60 mPa s
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.1 Relation of viscosity coefficient to the viscosity ratio with the linear
fit model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2 Relation of viscosity coefficient to the viscosity ratio with the non-
linear fit model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
B.1 correlation with 95% confidence interval for 9.5 mPa s dual layer fluid. 47
B.2 correlation with 95% confidence interval for 60 mPa s dual layer fluid. 48
B.3 correlation with 95% confidence interval for 200 mPa s dual layer fluid. 48
B.4 correlation with 95% confidence interval for 9.5/60 mPa s dual layer
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
B.5 correlation with 95% confidence interval for 9.5/200 mPa s dual layer
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
B.6 correlation with 95% confidence interval for 60/9.5 mPa s dual layer
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
B.7 correlation with 95% confidence interval for 60/200 mPa s dual layer
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
B.8 correlation with 95% confidence interval for 200/60 mPa s dual layer
fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
B.9 correlation with 95% confidence interval for 9.5 mPa s triple layer fluid. 51
B.10 correlation with 95% confidence interval for 60 mPa s triple layer fluid. 52
B.11 correlation with 95% confidence interval for 10 mPa s/60 mPa s/200 mPa s
triple layer fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
B.12 correlation with 95% confidence interval for 10 mPa s/200 mPa s/60 mPa s
triple layer fluid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
List of Tables
v
LIST OF TABLES
4.1 coefficients for the correlation in the case of same fluid for dual layer. 35
4.2 coefficients for the correlation in the case of different fluid for dual
layer with linear and non-linear fit. . . . . . . . . . . . . . . . . . . . 35
4.3 coefficients for the correlation in the case of same fluid for triple layer. 37
4.4 coefficients for the correlation in the case of different fluid for triple
layer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
vi
Listings
vii
Chapter 1
Introduction
Coating has been quite eloquently defined as any process that replaces gas at a solid
surface by a layer of liquid (Christodoulou and Scriven 1989a). As encapsulating as
this definition is, one nuance that can muddy its interpretation is the presence of
technologies capable of producing similar results. These usually entail the formation
of much thinner films, typically in the Angstrom range, and is commonly termed
thin film coating (Kistler and Schweizer 1997). For example a coating of ZnFe,
zinc phosphate and some organic layer formed by electrodeposition is not the type
of multilayer coating discussed here (Panagopoulos et al. 2011).
Figure 1.1: Schematic of a typical coating process. This instance shows a slide coater
transferring the layer onto a roll-to-roll substrate.
Coating in this report refers to liquid film coating, wherein liquid layers (in
the order of 100 microns) are deposited onto a substrate in a continuous process as
shown (Fig. 1.1). This chapter provides an introduction to many of the concepts
surrounding this field and relevant background information. Section one recapitu-
lates the historical applications of coating to date. Section two details the manner
in which coating flows can be classified, and provides typical examples of coaters in
each category. The third section gives a qualitative overview of the fluid mechanics
involved in coating flows; central to this is the idea of film thickness and inherent
1
CHAPTER 1. INTRODUCTION
difficulties during coating that give rise to an assortment of film defects. Finally,
section four dictates the aims and objectives of this study, specifying how its results
can be translated into an industrial environment.
2
CHAPTER 1. INTRODUCTION
printed silver electrode micro-patterns used in flexible printed circuit boards. The
conclusion of their work suggested that roll-to-roll rotogravure was apt at mass pro-
duction of large-area flexible printed circuit boards. The production of organic pho-
tovoltaics (OPVs) by coating processes is another area of popular interest. Wengeler
et al. 2014 have given in-depth research on the fluid dynamic properties of common
materials used to create polymer-based solar cells by roll-to-roll coating processes.
On the other hand, Kubota et al. 2015 looked at increasing the power conversion
efficiency of OPVs by multilayered antireflective coating and moth eye texturing.
It is obvious that coating processes are essential in the application of a coat
itself, from something simple as paint to flame retardants. Coating processes have
been used as integral steps in the fabrication of special membranes. Lu et al. 2017
dip-coated commercial stainless steel mesh membranes into a polymeric solution con-
taining divinyl sulfone (DVS) and PDSH (a superhydrophilic terpolymer). The DVS
and PDSH was then crosslinked to form a membrane that exhibited antimicrobial,
antifouling and hydrophilic-oeleophobic properties.
Pharmaceutical thin films that can be applied on the skin, mouth and even eye
are another use. Their clinical popularity is owed to the ease of administration,
especially where choking is a hazard (infants and the elderly), and their rapid onset
(Larraneta et al. 2016). The ability to increase productivity on a continuous roll-
to-roll basis is also commercially attractive (Karki et al. 2016).
Innovative uses of coating technology have been seen in the work of Tammisola
et al. 2010, where the capacity for spin coating to produce the protective layer of a
Blu-Ray disk was modeled. The inherent challenge to this is forming a layer that
is thick enough to simultaneously provide adequate protection but not obscure the
laser reader. The model was validated, but it is not known whether this has been
translated to anything practical. Zhu et al. 2015 investigated the use of coating tech-
nology as a means of surface finishing 3D printed parts. Surface finishing improves
the surface quality by overcoming the stepped nature of a 3D part due to layer-by-
layer construction. Typical methods involve sanding or chemicals which often spoil
the surface detail. Coating has the benefit of sealing gaps whilst simultaneously
providing beneficial properties such as corrosion resistance.
3
CHAPTER 1. INTRODUCTION
line plays a fundamental role in the integrity of the formed film. As the substrate
speed increases, there comes a point wherein the wetting line becomes disrupted
and momentarily detaches from the substrate. Usually this incurs the entrainment
of air, which results in the formation of non-uniform films containing air bubbles.
This mode of failure is known as dynamic wetting failure of which a plethora of film
defects can be associated with (Blake and Ruschak 1997). For a quality coating of
constant thickness, the line must remain straight and advance steadily.
Consider the case of Fig. 1.3 the flow originates from some distribution cavity
with Poiseuille flow. The liquid emerges from the slot, developing a velocity profile
a la Couette flow down the inclined slide. This region is known as the film forming
zone. Upon reaching the edge of applicator (in this case, the slide), the fluid enters
free fall. This is the sheet forming zone. Where the fluid makes contact underneath
the lip of the slide is known as the static wetting line. It plays a somewhat less con-
sequential role in coating processes, but still must be controlled as it may contribute
to leakage and contamination of the substrate (Blake and Ruschak 1997).
The fluid in free-fall is referred to as the liquid curtain. The viscous drag of
air on the fluid is considered negligible, therefore no velocity gradient can be estab-
lished. This results in a flat velocity profile known as plug flow at all points during
the descent. Effectively, the flow is inviscid. Where the curtain makes contact with
the substrate and forms the dynamic wetting line is referred to the upstream menis-
cus. The downstream meniscus forms the free surface. The film developed on the
substrate is referred to as the coated film forming zone (Blake and Ruschak 1997).
Fig. 1.3 is representative of slide curtain coating. It is called such because the film
is formed on a slide and transferred onto the substrate via the curtain. Of course,
the flow need not develop on a slide; it could originate from a die. In fact, when the
slide in Fig. 1.3 is lowered, such that the gap between the slide and substrate is small,
the point of no contact is suddenly referred to as a bead, not a curtain. Films may
be transferred by rollers, forced through dies, or simply dipped into a pool of liquid.
There are many geometries and mechanisms in which a film on a substrate can be
realized (Fig. 1.4. All of which may or may not change the physics of situation.
Clearly some form of categorization is needed. Two groups that commonly arise
are self-metered and premetered flows. The former takes into account the liquid
properties, web speed and geometry to determine the film thickness. In the latter
case a precise liquid delivery system is used; the film thickness here is controlled
by the flow rate (Weinstein and Ruschak 1985). A more descriptive classification
is provided by Benkreira et al. 1994, suggesting four categories: free coating flow,
metered coating flow, transfer coating flow, and print (or gravure) coating flow.
Free coating and metered coating flows may be considered a subcategory of self-
metered coating flows, whereas transfer coating and print/gravure coating flows
4
CHAPTER 1. INTRODUCTION
Figure 1.3: Example of a coating flow, detailing the film forming zone, sheet forming
zone and coated film forming zone.
5
CHAPTER 1. INTRODUCTION
in fact alternating patches of dry and wet. Unlike ribbing, they are formed due to
disturbances to the upstream meniscus (Tjiptowidjojo 2016).
6
CHAPTER 1. INTRODUCTION
7
CHAPTER 1. INTRODUCTION
speed of withdrawal. A numerical solution has been obtained by Javidi and Hrymak
2015 for the coating film thickness, showing good agreement with experimental data.
The coating thickness was found to be dependent on the coating bath dimensions.
Slot coating is a type of transfer coating flow (Benkreira et al. 1994). Film
thickness is determined by the coating flow, and it is independent of the coating
liquid properties. This makes it apt for precision coating. The substrate emerges
from the die slot and fills a small gap between the die lips and substrate. The liquid
bridge formed here is referred to as the coating bead. This bead is particularly
susceptible to vibrations associated with the movement of the substrate (Lee and
Nam 2015). The flow in the bead is related to the quality of the formed film
(Horiuchi, Suszynski, and Carvalho 2015). Sometimes this can result in the change
of the distance between the die lip and substrate. This usually occurs with substrate
8
CHAPTER 1. INTRODUCTION
thickness variations and is called gap disturbance. The web speed has been found
to be the most influencing parameter. Low viscosity Newtonian fluids have shown
to be effective at mitigating the effect of gap disturbance (Lee and Nam 2015).
Air entrainment of multilayer slot coating has been found to be dependent on
the ratio of the bottom to top thickness ratio, with the bottom layer being thicker
than 1/3 of the coating gap (Horiuchi, Suszynski, and Carvalho 2015). A minimum
thickness is also necessary for single layers to prevent air entrainment (Schmitt et al.
2015).
The static wetting line is very important in slot coating, with the position of the
upstream meniscus being indicative of the onset of ribbing and other film defects
(Lee et al. 2011).
A gravure coating flow is depicted in Fig. 1.9. The cavities in the cog-like wheel
are referred to as cells. These are filled with liquid from the bath and transferred
onto the substrate as it rotates. This is partially transferred to the substrate as the
liquid bridge formed between the two moving surfaces breaks apart. The thickness
of this film is determined by the capillary number (Miura and Yamamura 2015).
9
CHAPTER 1. INTRODUCTION
2. any relevant work in the study of multilayer film thickness and slide coating
will be reviewed.
3. relevant data of Alpin 2016 will be extracted into a new Excel environment,
in which the data will be analyzed.
4. MATLAB will be used to devise correlations relating the ratio of the measured
to predicted height based on the observed influencing parameters.
10
Chapter 2
Literature Review
In this section, the background information relevant to slide coating is first presented.
Each section is then dedicated to studies pertaining to single, dual and triple layer
flow systems.
2.1 Introduction
The remainder of the discussion on coating flow fluid mechanics is restricted to slide
coating for relevancy. The slide coating process is represented by Fig. 2.1. Slide
coating is a type of transfer coating flow (Benkreira et al. 1994). The applicator
is comprised of an inclined die face with a series of spaced slots. Fluid emerges
from these slots and flows laminarly down the slide onto a moving substrate. In
multilayered coating, subsequent layers flow down over the first without mixing
(Christodoulou and Scriven 1989b). The length of the die tends to increase with
the number of layers in the die. This tends to decrease stability since the increase
in length provides a longer time for the growth of free surface waves (Hens and
Abbenyen 1997). A vacuum is fitted underneath to stabilize the bead, enabling
operation outside of normal coating speeds (Christodoulou and Scriven 1989b). The
vacuum must withdraw air at a uniform rate as fluctuations in pressure interfere with
the coating bead (Hens and Abbenyen 1997).
The geometry of the slot exits influence the coating flow. Some configurations
are shown in Fig. 2.2. When the slide is all on one plane (a), the formation of
vortexes can occur, meaning the flow is neither two-dimensional or uniform. In
general, these geometries are designed with the intention of preventing back flow
(b) and re-circulation or vortexes (c) (Hens and Abbenyen 1997).
Wave formation is particularly relevant in the slide region due to its implication
on the quality of the final coated film. Eqn. 2.2 states that for any set of parameters
one value of film thickness is prescribed. In reality, the formation of free surface
waves occur. It was shown by (Yih 1965) that above a critical Reynolds number,
11
CHAPTER 2. LITERATURE REVIEW
5
Rec = cot (2.1)
6
12
CHAPTER 2. LITERATURE REVIEW
course the film thickness. The solution is obtained from applying the Navier-Stokes
equations to the above situation with a rectangular coordinate system. It is assumed
that the flow is unidirectional in the path of x, this simplifies the equation to:
2 vx
0= + g sin (2.3)
y 2
The partial notation of the Navier-Stokes in Eqn. 2.3 can be dropped since the
velocity in the x direction changes only in response to the height. To further simplify
the problem, the parameters density, gravitation acceleration, angle and viscosity
13
CHAPTER 2. LITERATURE REVIEW
Figure 2.3: Diagram for the analysis of a film flowing down an inclined plane.
are lumped together as A since they are all constant. The second order differential
equation can then be written as being equal to:
d2 vx g sin
= 2
= A (2.4)
dy
To determine the velocity profile, the indefinite integral of Eqn. 2.4 with respect to
y is taken. This introduces the first constant of integration C1 .
dvx
= Ay + C1 (2.5)
dy
Integrating Eqn. 2.6 for the velocity, introduces the second constant, C2 .
y2
vx = A + C1 + C2 (2.6)
2
In order to solve for the velocity, Eqn. 2.7 must be evaluated at two boundary
conditions. Both the no-slip condition and free surface flow assumption are invoked.
The former states that the velocity is zero at the liquid-solid interface (B.C. 1), and
the latter that the change in velocity is zero at the free surface (B.C. 2). That is to
14
CHAPTER 2. LITERATURE REVIEW
say:
B.C. 1: at y = 0, vx = 0
dvx
B.C. 2: at y = , =0
dy
Thus,
gsin
= vx = [2y y 2 ] (2.7)
2
Integrating the product of Eqn. 2.7 and the differential area W dy results in the
volumetric flowrate which can then be rearranged to return Nusselts proposed model
of film thickness. 1/3
3Q
=
g sin
Since then, several other models for the film thickness have been proposed. A
comparison of these down a vertical plate was done by citeportalski:1963. Fluids
of greater kinematic viscosity gave way to thicker films with regular wave patterns.
Viscosity was suspected to play a role in damping the oscillations. Plots of the
measured thickness versus the flow rate produced a family of curves. The exponent
of this was found to be indicative of the flow state.
The work of Henry et al. 2014 is quite similar to that of Alpin 2016 (which is
the basis of this study). They looked at layers ranging from one to three down a
slide coater using flow visualisation technology. From this they could extrapolate
the heights of fluid layers. For the single layer they found that the the measured film
thickness showed good agreement with the Nusselt approximation (Eqn. 2.2). Their
study involved a water-glycerol mixture for the coating fluid. The concentrations
in which the water is mixed with glycerol results in the fluids having appreciably
close properties to compare. For instance, (Henry et al. 2014) used fluids of 8.4, 72,
and 219 mPa s, whereas (Alpin 2016) used 9.5 mPa s, 60 mPa s, and 200 mPa s.
In a study by (Takamura, Fischer, and Morrow 2012) it was shown that viscosity
was the most sensitive property in response to a change in composition. Therefore,
provided that the viscosity is reasonably close they can be assumed to be practically
the same composition.
The free surface profile of a single layer down a slide has been studied by Gaskell
et al. 2014 using finite element solutions to the Navier-Stokes equations.
Flow visualization of a single layer down a slide has also been studied by Njifenju
et al. 2013, using water as the coating fluid. Their data showed good agreement with
the Nusselt approximation. Deviation from theory was pinned down to the formation
of waves.
15
CHAPTER 2. LITERATURE REVIEW
16
Chapter 3
This chapter begins by explaining the methodology for this investigation. The fol-
lowing subsection then presents the observations of the data.
3.1 Methodology
The raw data was obtained from the work of Alpin 2016. This data looks at one,
two and three layers flowing down the slide. The coating fluid used was a mixture
of water and glycerol such that the flow properties were altered (density, viscosity,
et cetera). In the multilayer case, the fluid position was varied. For each layer the
height value associated with the corresponding density, viscosity, angle of inclination
and flow rate were given. The height was measured just outside the slot exit. The
various combinations for each flow system are given in Table 3.1 and Table 3.2. The
combinations for the tiple layer are explained in the relevant section.
The first step was to plot the measured height against the theoretical height.
For the single layer, the Nusselt approximation was used.
1/3
3Q
h=
g sin
The film thickness of a multilayered scenario was approximated by the 1-layer ap-
proximation previously mentioned.
1/3
31 QT
hT =
1 g sin
These equations accept the flow rate on a per unit width basis. This is due to the
fact that upon derivation of these equations, the film height is assumed to be equal
17
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
Die angle Flow rate layer 1 Flow rate layer 2 Viscosity layer 1 Viscosity layer 2
10 0.10 0.50 10 10
15 0.50 0.15 60 60
23 0.50 0.50 200 200
30 0.10 1.00 10 60
35 1.00 0.15 60 10
10 200
60 200
200 60
18
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
across the width of the slide. The flow rates were divided by the slot width (0.15 m)
to convert them into a compatible form. For the multilayer flow it is stressed again
that the total flow rate is used, in addition to the fluid properties of the bottom
layer. Ideally, the plot of measured versus predict height should align as y = x if
the approximation is accurate. This is later shown not to be the case for every flow
situation.
In many cases the approximations provided above are not adequate to describe
the real film thickness and must be correlated for better accuracy. The approach for
this correlation is to relate the measured film thickness to a set of parameters that
are known beforehand, or can be calculated that influence the final film thickness.
These parameters can be inferred from the approximations (angle, density, viscosity,
flow rate). In order to keep dimensional consistency, the parameters will be nondi-
mensionalized, with the dimensionless height being a function of the formulated
dimensionless groups. The dimensionless height, the height ratio (H), was then
defined as the measured height (hm ) divided by the predicted height (hp ). Thus,
if H > 1 then the measured height is greater than the predicted. Alternatively, if
H < 1, then the predicted height is greater than the measured.
A more heuristic approach is then taken for developing the dimensionless groups;
for them to be meaningful they must ideally incorporate every parameter in the
problem. For the single layer, both the Reynolds number (Re) and capillary (Ca)
contain many of the investigated numbers.
Q
Re = (3.1)
U
Ca = (3.2)
Where U is the characteristic velocity:
Q
U= (3.3)
hp
The height ratio was then plotted against these groups to discern any relationship.
From which, a correlation was derived base on these numbers for the single layer
case.
The multilayered situation followed a similar protocol to establish the correla-
tions. Except that the Reynolds number used the total flow rate of the layers, just
like the 1-layer approximation. In addition, it was corrected for the angle used. This
is partly due to the observed deviation from the approximation with changing angle.
Thus, the corrected Reynolds number is:
19
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
QT
Recorrected = tan() (3.4)
From the observations, the role of the viscosity ratio, flow rate and angle in the
final height ratio was inferred. The parameters flow rate and angle were contained
in the corrected Reynolds number, thus the general format is given as:
2
Hd = Cd Recorrected (3.5)
1
For the dual layer case, where Cd is the constant coefficient of the dual layer, and
is the viscosity ratio constant exponent and is the constant corrected Reynolds
number exponent. The subscripts 1 and 2 refer to the fluid layer. This notation can
be extended to the triple layer case, giving:
1 2
2 3
Ht = Ct Recorrected (3.6)
1 1
MATLAB Student R2017a was used with Curve Fitting Toolbox for the rendition
of each correlation. The correlations are discussed in more detail in the next chapter.
20
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
Figure 3.1: Measured height versus predicted height for the single layer 9.5 mPa s
fluid.
Previously it has been mentioned that the amplitude of surface waves increases
with Reynolds number. One possible explanation is that the sinusoidal oscillation
creates too much variance in the average height and so the recorded values are above
the predicted. Although this would suggest the presence of points both below and
above the line which is not the case here. Data points of the steeper inclination,
but similar Reynolds number, tend to hug the line more closely, whereas the lower
angles represented by light blue and orange tend to be more scattered. This is likely
due to the previous definition of the Reynolds number having no consideration for
the incline. In summary, the measured height is below expectation for smaller
heights, the middle range reflects theory excellently, and a deviation towards the
actual height being greater that theory is noted towards the end.
The 60 mPa s fluid showed the greatest adherence to the Nusselt approximation
out of the three fluids. There is a notable, but slight deviation at the lower predicted
heights; this is more telling of the steeper inclinations, whereas the shallow inclines
are more obedient.
Something unusual occurred in the 200 mPa s fluid. A clearly linear trend is
exhibited when plotting the measured height versus the predicted height, however
the measured height appears offset by a fixed constant. Remarkably, the gradient
is unchanged, suggesting the fit is more akin to hm = hp + K, where K is the
intercept. However, upon inspection of the last few points for each angle, a shift
21
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
Figure 3.2: Measured height versus predicted height for the single layer 60 mPa s
fluid.
Figure 3.3: Measured height versus predicted height for the single layer 200 mPa s
fluid.
22
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
towards the line y = x is seen. This convergence increases with the inclination.
Since the more viscous fluid implies a lower Reynolds number at the same flow rate,
the 200 mPa s may gradually fit the Nusselt approximation at greater Reynolds
numbers. To substantiate this claim, a plot of all the height ratios versus the
Reynolds number was created. In Fig. 3.4 it can generally be seen that as the
Reynolds number increases, the height ratio (hm /hp ) approaches unity, implying
that the Nusselt approximation is a good fit. However, it should be cautioned that
more data points at higher Reynolds numbers would be needed to ensure that this
becomes asymptotic at one, rather than continuing to decrease.
Figure 3.4: Height ratio versus Reynolds number for all points of the 200 mPa s
fluid.
23
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
addition, the 1-layer approximation is used to predict the film thickness. The spirit
of this is that the total film thickness is imagined to be the just one film (with the
bottom layer fluid properties), but with the cumulative flow rates of each individual
layer. The scenario with identical fluids is presented first.
In Fig. 3.5-Fig. 3.7, the measured height versus predicted height is plotted in-
dividually. Fig. 3.8 paints the bigger picture, from which an overall trend can be
established. The 9.5 mPa s dual layer shows good agreement with theory with the
smaller film thicknesses. There is an even distribution about the line y = x which
might represent the error in measuring the oscillations at the free surface. What is
undeniable, however, is the broadening of the variance from the line y = x as the film
thicknesses get larger. This is more exaggerated in the shallower angles, whereas
the steeper inclines seem to hug the line more, and the approximation becomes
more valid. The 60 mPa s fluid is shown in Fig. 3.6. Most points lie above the line,
implying the measured height is greater than theory. A better fit is observed in
the 200 mPa s case, however the points have greater variance about the line y = x
compared to the 9.5 mPa s situation. Fig. 3.8 takes a collective view of all these data
points. There appears to be some form of heteroskedasticity, in that the variation of
points grows as the predicted height increases. The role of some of the influencing
variables can be determined here. The circles, squares and triangles represent data
points in order of increasing viscosity. The dispersion of points increases with in-
creasing viscosity. It can also be said that the lower angles also increase dispersion.
One final point to is the repeated pattern regarding the variance of data. The reason
this pattern is repeated is likely because of the differences in flow rates of the two
layers. Generally, when the flow rate of the second layer has a greater flow rate than
the first layer, then the total film thickness is exceeds the predicted value. When
the first layer has a substantially greater flow rate than the second, then the total
measured film thickness is less than the predicted.
The situation immediately becomes much more complicated with the introduc-
tion of viscosity differences in addition to flow rate differences. For maximum clarity
each graph is presented separately. There are five in total, however it is better to
imagine them as two groups: a) where the bottom layer contains the most viscous
fluid and b) where the top layer contains the most viscous fluid. The first group
behaves more ideally and reflects the 1-layer approximation well, and can be seen in
Fig. 3.9 and Fig. 3.10. The second group does not, with the measured height being
consistently in excess of the predicted height. These are shown in Fig. 3.11, Fig. 3.12
and Fig. 3.13. Presented in these cases are varying amounts of difference in viscosity.
There are two mirror cases, that is, where the same fluids are used, but the order is
reversed. This exists for the 60 mPa s-200 mPa s and the 9.5 mPa s-60 mPa s case.
For group a this is Fig. 3.9 and Fig. 3.10; for group b this is Fig. 3.11 and Fig. 3.13.
24
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
Figure 3.5: Measured height versus predicted height for the dual layer 9.5 mPa s
fluid.
Figure 3.6: Measured height versus predicted height for the dual layer 60 mPa s
fluid.
25
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
Figure 3.7: Measured height versus predicted height for the dual layer 200 mPa s
fluid.
Considering group b for the moment, as the difference in viscosity between the top
and bottom layer increases, so does the deviation from the 1-layer approximation.
In group a, as the difference between the bottom and top layer increases, there is a
very slight depression of the points, implying the actual height is less than predicted.
Note in the mirror cases however, that for equal changes in viscosity differences the
response is not equal. In other words, if the least viscous is at the bottom then any
change in viscosity stratification will be more consequential than if the most viscous
fluid was on the bottom. This is of course with respect to the 1-layer approximation.
Retaining the grouping system mentioned previously, and looking at repeated
patterns in the data for each angle can provide some insight into the influence of
flow rate stratification. Consider group b first, where the lower viscous liquid is the
bottom layer. When the flow rate of the top layer is greater than or equal to the
bottom, then the measured height is the most over-predicting. This over-prediction
is proportional to the top layer flow rate, so the as the top layer flow rate increases,
so does the over prediction. Generally, as the angle increases, the amount by which
the measured height is greater decreases. This rule of thumb holds very tenuously
when the differences in viscosity are larger (9.5 mPa s-200 mPa s).
26
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
Figure 3.8: Measured height versus predicted height for the dual layer all fluids.
For the different layers there are only two combinations studied. The first is the
9.5 mPa s, 60 mPa s and 200 mPa s corresponding to layers one, two and three.
The second is 9.5 mPa s, 200 mPa s and 60 mPa s. These are shown in Fig. 3.16,
Fig. 3.17 respectively. In the first case, each layer above the previous has a greater
viscosity. From the dual layer section it was observed that when this happened,
increased scatter and deviation from the 1-layer approximation was seen. In the
second case there is more deviation, but much less scatter. This may be due to the
27
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
Figure 3.9: Measured height versus predicted height for the dual layer 60/9.5 mPa s
fluid.
28
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
Figure 3.10: Measured height versus predicted height for the dual layer 200/60 mPa s
fluid.
Figure 3.11: Measured height versus predicted height for the dual layer 9.5/60 mPa s
fluid.
29
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
Figure 3.12: Measured height versus predicted height for the dual layer
9.5/200 mPa s fluid.
30
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
Figure 3.13: Measured height versus predicted height for the dual layer 60/200 mPa s
fluid.
Figure 3.14: Measured height versus predicted height for the triple layer 9.5 mPa s
fluid.
31
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
Figure 3.15: Measured height versus predicted height for the triple layer 60 mPa s
fluid.
Figure 3.16: Measured height versus predicted height for the triple layer
9.5/60/200 mPa s fluid.
32
CHAPTER 3. EXPERIMENTAL DATA AND OBSERVATION
Figure 3.17: Measured height versus predicted height for the triple layer
9.5/200/60 mPa s fluid.
33
Chapter 4
Data Correlation
A complete discussion of the correlations are presented here. This includes previous
attempts as to show the development. The single layer relates the height ratio as
a function of the Reynolds or capillary number due to the relative simplicity of the
interacting parameters. This is also because of how well the Nusselt approximation
has already shown to be. The dual and triple layer height ratios take the general
form of Eqn. 3.6 Eqn. 3.5 as seen in the previous chapter. The m-file associated
with each layer can be found in Appendix A.
Hd = Cd Recorrected (4.1)
34
CHAPTER 4. DATA CORRELATION
Table 4.1: coefficients for the correlation in the case of same fluid for dual layer.
Table 4.2: coefficients for the correlation in the case of different fluid for dual layer
with linear and non-linear fit.
The values of Cd and for the same fluid dual layer case are given in Table 4.1.
The reduction in the coefficient Cd coincides with the reduction in variance. The
problem in this case is not the viscosity term, however, and it is unlikely that this
proposed model can help correlate the height ratio. It would appear that the flow
rate stratification plays a more important role which should be looked at in future.
The model is satisfactory in that most points lie within the confidence interval.
When the viscosity was different in each layer, the Cd term was found to be weak.
It was eliminated, so that only the exponent of the viscosity ratio and Reynolds
number could influence the situation. This lead to Eqn. 4.2:
2
Hd = Recorrected (4.2)
1
The problem with this rendition is that now the Reynold exponent was small, sug-
gesting the influence was trivial. However, upon moving this term the intercept is
depressed and the model does not fit well at all. In an ad hoc fashion, the model
was converted to a linear form to see if this would increase the magnitude of the
exponent.
2
Hd = + Recorrected (4.3)
1
This solution was not particularly effective as it merely reversed the signs and the
magnitude was relatively unchanged. These can be found in Table. 4.2. A plot of the
viscosity exponent versus viscosity ratio showed linear relation (Fig. 4.1, Fig. 4.2). In
the linear model, this relationship was very strong, suggesting that = 12 . However,
this appears too coincidental to be true which may call for a re-investigation.
35
CHAPTER 4. DATA CORRELATION
Figure 4.1: Relation of viscosity coefficient to the viscosity ratio with the linear fit
model.
Figure 4.2: Relation of viscosity coefficient to the viscosity ratio with the non-linear
fit model.
Table 4.3: coefficients for the correlation in the case of same fluid for triple layer.
Table 4.4: coefficients for the correlation in the case of different fluid for triple layer.
37
Chapter 5
A literature review was carried out to investigate the influence of parameters in slide
coating on the film thickness. Experimental data was then used to plot the measured
height versus predicted height. It was determined that the viscosity stratification
had an apparent effect on the accuracy of the approximation models used. Thus,
a correlation relating the height ratio to dimensionless groups was posed. The
dimensionless groups for the single layer were the Reynolds and capillary number.
For the multilayer cases, these were inferred from the plots of measured versus
predicted height. It was decided that the viscosity played an important role, thus
the viscosity ratio and correct Reynolds number was used. The single layer fluid
actually showed excellent fit with the Nusselt approximation. Generally, the two
layer flow height was predicted well by the 1-layer approximation provided that
either the same fluid was used, or the most viscous one was at the bottom. When
this was not the case, a correlation was more appropriate since the approximations
were inaccurate.
The proposed correlation for the dual layer relates the height ratio to the viscosity
ratio and corrected Reynolds number. It was found that the exponent of the viscosity
ratio was proportional to viscosity ratio. This appears too coincidental to be correct
and should be reviewed. More work can be done with the data, including the
comparison of the data with the derivation of a new model to approximate the film
thickness. The use of neural-networks to develop correlations may also be suitable
for this data type due to the wide amount of variables. This might guide a better
solution for the film thickness problem given the strength of the coefficient of each
parameter.
38
Bibliography
39
BIBLIOGRAPHY
Hens, J. and W. V. Abbenyen (1997). Slide coating. In: Liquid Film Coating.
Ed. by S. F. Kistler and P. M. Schweizer. Glasgow: Springer Science+Business
Media Dordrecht. Chap. 11, pp. 427459.
Horiuchi, R, W. J. Suszynski, and M. S. Carvalho (2015). Simultaneous multilayer
coating of water-based and alcohol-based solutions. In: Journal of Coatings
Technology and Research 12.5, pp. 819826. doi: 10.1007/s11998-015-9689-
9.
Javidi, M. and A. N. Hrymak (2015). Numerical simulation of the dip-coating
process with wall effects on the coating film thickness. In: Journal of Coatings
Technology and Research 12.5, pp. 843853. doi: 10.1007/s11998-015-9699-7.
Jiang, W. Y. et al. (2005). Low-Reynolds-number instabilities in three-layer flow
down an inclined wall. In: Journal of Fluid Mechanics 539.2005, pp. 387416.
doi: 10.1017/S0022112005005781.
Karki, S. et al. (2016). Thin films as an emergying platform for drug delivery. In:
Asian Journal of Pharmaceutical Sciences 11.5, pp. 559574. doi: 10.1016/j.
ajps.2016.05.004.
Kistler, S. F. and P. M. Schweizer (1997). Coating science and technology: an
overview. In: Liquid Film Coating. Ed. by S. F. Kistler and P. M. Schweizer.
Glasgow: Springer Science+Business Media Dordrecht. Chap. 1, pp. 13.
Kubota, S. et al. (2015). Hybrid antireflection structure with moth eye and multi-
layer coating for organic photovoltaics. In: Journal of Coatings Technology and
Research 12.1, pp. 3747. doi: 10.1007/s11998-014-9614-7.
Larraneta, E. et al. (2016). Microneedle arrays as transdermal and intradermal drug
delivery systems: Materials science, manufacture and commercial development.
In: Materials Science & Engineering: R: Reports 104.June, pp. 132. doi: 10.
1016/j.mser.2016.03.001.
Lee, S. and J. Nam (2015). Response of slot coating flow to gap disturbances: effects
of fluid properties, operating conditions, and die configurations. In: Journal of
Coatings Technology and Research 12.5, pp. 949958. doi: 10.1007/s11998-
015-9684-1.
Lee, S. H. et al. (2011). Operability coating windows and frequency response in slot
coating flows from a viscocapillary model. In: Chemical Engineering Science
21.1, pp. 49534959. doi: 10.1016/j.ces.2011.04.044.
Lu, C., S.-Y. Jiang, and R.-Q. Duan (2016). Wave Characteristics of Falling Film
on Inclination Plate at Moderate Reynolds Number. In: Science and Technology
of Nuclear Installations 2016, pp. 17. doi: 10.1155/2016/6586097.
Lu, M. et al. (2017). Combined spear and shield: superhydrophilic antimicrobial
and antifouling mesh membrane for efficient oil-water separation through facile
40
BIBLIOGRAPHY
41
BIBLIOGRAPHY
42
Appendix A
Listings
20 figure ()
21 p l o t (X, y , ko ) % o b s e r v e d data
22 h o l d on
23 p l o t ( xrange , ypred , k , LineWidth , 2 )
24 p l o t ( xrange , [ l o w e r ; upper ] , r , LineWidth , 1 . 5 )
43
APPENDIX A. LISTINGS
32 figure ()
33 p l o t (X, y , ko ) % o b s e r v e d data
34 h o l d on
35 p l o t ( xrange , ypred , k , LineWidth , 2 )
36 p l o t ( xrange , [ l o w e r ; upper ] , r , LineWidth , 1 . 5 )
44
APPENDIX A. LISTINGS
11 %modelfun = @( b ,X) ( b ( 1 ) X. b ( 2 ) ) ;
12 %b e t a 0 = [ 1 , 1 ] ;
13
25 figure ()
26 p l o t (X, y , ko ) % o b s e r v e d data
27 h o l d on
28 p l o t ( xrange , ypred , k , LineWidth , 2 )
29 p l o t ( xrange , [ l o w e r ; upper ] , r , LineWidth , 1 . 5 )
45
APPENDIX A. LISTINGS
27 f i g u r e ( ) %p l o t
28 p l o t (X, y , ko ) % o b s e r v e d data
29 h o l d on
30 p l o t ( xrange , ypred , k , LineWidth , 2 )
31 p l o t ( xrange , [ l o w e r ; upper ] , r , LineWidth , 1 . 5 )
46
Appendix B
Correlations
Figure B.1: correlation with 95% confidence interval for 9.5 mPa s dual layer fluid.
47
APPENDIX B. CORRELATIONS
Figure B.2: correlation with 95% confidence interval for 60 mPa s dual layer fluid.
Figure B.3: correlation with 95% confidence interval for 200 mPa s dual layer fluid.
48
APPENDIX B. CORRELATIONS
Figure B.4: correlation with 95% confidence interval for 9.5/60 mPa s dual layer
fluid.
Figure B.5: correlation with 95% confidence interval for 9.5/200 mPa s dual layer
fluid.
49
APPENDIX B. CORRELATIONS
Figure B.6: correlation with 95% confidence interval for 60/9.5 mPa s dual layer
fluid.
Figure B.7: correlation with 95% confidence interval for 60/200 mPa s dual layer
fluid.
50
APPENDIX B. CORRELATIONS
Figure B.8: correlation with 95% confidence interval for 200/60 mPa s dual layer
fluid.
Figure B.9: correlation with 95% confidence interval for 9.5 mPa s triple layer fluid.
51
APPENDIX B. CORRELATIONS
Figure B.10: correlation with 95% confidence interval for 60 mPa s triple layer fluid.
52
APPENDIX B. CORRELATIONS
53