You are on page 1of 19

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/225730093

Fully developed flow of a viscoelastic film down a vertical cylindrical or planar


wall

Article  in  Rheologica Acta · December 2009


DOI: 10.1007/s00397-009-0386-4

CITATIONS READS

7 85

3 authors, including:

John Tsamopoulos
University of Patras
122 PUBLICATIONS   2,790 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

On the flow and deformation of Newtonian, viscoplastic and viscoelastic fluids based on the Finite element method along with robust and cutting edge 2D & 3D grid
generation techniques View project

On the flow and deformation of Newtonian, viscoplastic and viscoelastic fluids based on the Finite element method along with robust and cutting edge 2D & 3D grid
generation techniques View project

All content following this page was uploaded by John Tsamopoulos on 15 July 2014.

The user has requested enhancement of the downloaded file.


Rheol Acta (2009) 48:1031–1048
DOI 10.1007/s00397-009-0386-4

ORIGINAL CONTRIBUTION

Fully developed flow of a viscoelastic film down


a vertical cylindrical or planar wall
M. Pavlidis · Y. Dimakopoulos · J. Tsamopoulos

Received: 27 March 2009 / Accepted: 2 September 2009 / Published online: 17 September 2009
© Springer-Verlag 2009

Abstract The one-dimensional, gravity-driven film Stokes number decreases as R̃/ H̃ decreases down to
flow of a linear (l) or exponential (e) Phan-Thien and zero for a film on the outer cylindrical surface, while
Tanner (PTT) liquid, flowing either on the outer or on it asymptotes to very large values when R̃/ H̃ decreases
the inner surface of a vertical cylinder or over a planar down to unity for a film on the inner surface. When ξ =
wall, is analyzed. Numerical solution of the governing 0, an upper limit in De exists above which a solution
equations is generally possible. Analytical solutions cannot be computed. This critical value increases with
are derived only for: (1) l-PTT model in cylindrical ε and decreases with ξ .
and planar geometries in the absence of solvent, β ≡
η̃s / η̃s + η̃ p = 0, where  η p and 
ηs are the zero-shear Keywords Viscoelastic film flow ·
polymer and solvent viscosities, respectively, and the Gravity-driven flow · PTT fluid model
affinity parameter set at ξ = 0; (2) l-PTT or e-PTT
model in a planar geometry when β = 0 and ξ = 0;
(3) e-PTT model in planar geometry when β = 0 and Introduction
ξ = 0. The effect of fluid properties, cylinder radius,
R̃, and flow rate on the velocity profile, the stress The Phan-Thien and Tanner (PTT) constitutive
components, and the film thickness, H̃, is determined. equation is based on network theory and can be used
On the other hand, the relevant dimensionless num- primarily for polymer melts, but also for polymer
bers, which are the Deborah, De = λ̃Ũ/ H̃, and Stokes, solutions when a Newtonian “solvent” is included by
2   
St = ρ̃ g̃H̃ / η̃ p + η̃s Ũ, numbers, depend on H̃ and setting β ≡ η̃s / η̃s + η̃ p = 0 (Phan-Thien and Tanner
the average film velocity, U.  This makes necessary a 1977; Phan-Thien 1978). It is one of the most widely ac-
trial and error procedure to obtain H̃ a posteriori. We cepted viscoelastic rheological models, due to its ability
find that increasing De, ξ , or the extensibility parameter to represent realistically the experimentally measured
ε increases shear thinning resulting in a smaller St. The elongational properties and the shear-thinning behav-
ior of polymers (Quinzani et al. 1995) and its simplicity,
because it requires only two additional parameters
besides fluid elasticity (Saramito and Piau 1994). These
M. Pavlidis · Y. Dimakopoulos · J. Tsamopoulos additional parameters control the elongational and
Laboratory of Computational Fluid Dynamics, shear-thinning behavior of the model, ε, or the slip of
Department of Chemical Engineering, University of Patras,
the polymeric chains with respect to the medium, ξ .
Patras 26500, Greece
Therefore, investigation of its predictions for velocities
Present Address: and stresses in simple flows is of great importance
Y. Dimakopoulos (B) but also necessary for comparisons with experiments.
Department of Biomedical Engineering,
Eindhoven University of Technology, P.O. Box 513,
Particularly valuable are analytical solutions to such
Eindhoven, 5600 MB, The Netherlands problems, because they provide a convenient way to
e-mail: dimako@chemeng.upatras.gr examine the ability of a fluid model to represent a
1032 Rheol Acta (2009) 48:1031–1048

specific fluid behavior. However, these are not always either the Upper Convected Maxwell or the Oldroyd-B
available because of the non-linearity of this model, models, to which the PTT model reduces for particular
especially in its exponential form and the need to solve values of its parameters. Also, Ruyer-Quil et al. (2008)
the momentum and constitutive model equations as a have studied film flow of a Newtonian fluid down a
coupled set, in contrast to simpler models such as the fiber. They examined conditions for wave formation
UCM. On the other hand, simplified one-dimensional and growth leading eventually to liquid beads around
problems give physical insight and can be used to check the fiber. Analytical or numerical solutions for film flow
results of numerical computations as limiting test cases around a cylindrical surface, even under fully devel-
of complex geometries. Moreover, such solutions of oped conditions, using the PTT fluid model are still
simple flows are necessary as inlet (Karapetsas and lacking, but are presented herein.
Tsamopoulos 2008) or outlet (Dimakopoulos and Analytical solutions for the PTT model in duct flows
Tsamopoulos 2004) boundary conditions in numerical (i.e., a 2D channel or a pipe) have been derived by
simulations of more complex flows using the PTT Oliveira and Pinho (1999) using either the linear or the
model. Examples of such flows are the planar or exponential stress coefficient function in the absence
axisymmetric sudden contraction or expansion flows of a Newtonian solvent β = 0. Alves et al. (2001) have
(Baaijens 1993; Baloch et al. 1996), extrusion from added the effect of the affinity parameter, ξ = 0. Cruz
rectangular, circular, or annular dies (Sunwoo et al. et al. (2005) have explored the existence of analyt-
2001), flows around solid bodies (Baaijens et al. 1997), ical solutions, when the PTT fluid model includes a
etc. Using the readily available solutions for Newtonian Newtonian solvent, β = 0, while retaining the no-slip
fluids as inlet or outlet conditions for viscoelastic fluids condition between the molecular network and the
inevitably leads to inaccurate results. continuum medium ξ = 0. The two latter works in-
Here, we consider the fully developed flow of a volve only the linear variation of the PTT constitutive
viscoelastic film, either on the outer or on the inner equation.
surface of a long cylinder. The corresponding planar As we will demonstrate below, there are certain
geometry is obtained when the ratio of the cylin- similarities between one-dimensional film flow and duct
der radius to the film thickness becomes very large. flows. For example, in both geometries, (1) the single
Film flow of a Newtonian fluid in the planar geom- velocity component varies only in the shear direction;
etry is considered a textbook problem (Bird et al. (2) normal polymeric stresses arise in the streamwise
2002), although much more sophisticated and realistic direction and, only when ξ = 0, in the shear direction;
models and experiments exist, e.g., Vlachogiannis and and (3) the boundary conditions are no slip on the
Bontozoglou (2001). Generally, in polymeric flows, the wall and zero shear, either on the symmetry line or
shear viscosity is very large, resulting in rather low on the undeformable free surface (the ambient air ap-
velocities, and the surface tension is small, making the plies no force). On the other hand, there are certain
capillary number very large and surface tension effects differences between the flows in these two geometries.
negligible, even for the curved free-surface flow on For example: (1) the duct width is known, whereas the
a cylindrical surface. Moreover, we do not consider film thickness is unknown (i.e., it is a free boundary
very thin films, which, in addition, require examination problem), and must be calculated after the velocity
of short-range intermolecular forces. Hence, the vis- profile is determined from the imposed flow rate; (2)
coelastic film thickness is not affected by the complex duct flows are pressure-driven, whereas film flow is
interplay of surface tension, inertia, and viscous and gravity-driven; and (3) an additional parameter arises
intermolecular forces, as discussed in Quere (1999), but in film flow around a cylinder, the ratio of the cylinder
it is affected by the imposed flow rate, the velocity radius to the film thickness. This latter geometric effect
of the solid substrate, and gravity, depending on the adds to the linear dependence of the total shear stress
application. Examples of such flows arise in the so- on the distance from the wall, a term that is inversely
called premetered coating flows (Ruschak 1985), such proportional to this distance, which results in an addi-
as slide and curtain coating (Kistler and Schweitzer tional logarithmic variation of the velocity profile, even
1997), in spin coating (Borkar et al. 1994; Tsamopoulos for a Newtonian fluid. Therefore, the present geometry
et al. 1996), etc. Very recently, Saprykin et al. (2007) cannot be reduced to the simpler flow inside a pipe.
investigated the steady and transient gravity-driven Consequently, both the dimensionless pressure drop
flow of a viscoelastic film flowing in the presence of and the dimensionless gravity force (Stokes number)
inertia and topography. They examined the effect of involve the average film velocity, whereas the pressure
elasticity on the ridge height that is formed near a step drop is imposed, the Stokes number is calculated, and
change in the topography of the solid substrate using when combined with the dimensional flow rate yields
Rheol Acta (2009) 48:1031–1048 1033

the dimensional film thickness. Moreover, we will show a long cylinder, whereas in Fig. 1b the fluid coats the
that the cylinder radius even for Newtonian fluids af- inner surface of a hollow cylinder. For a planar film,
fects the Stokes number. it was generally more convenient to start directly with
In this paper, we present numerical or analytical, Cartesian coordinates and extract the corresponding
when possible, solutions for the fully developed gravity- equations than to deduce them from the cylindrical
driven film flow of a PTT fluid model over either the ones. Even in this geometry, the coordinate normal to
outer or the inner surface of a cylindrical wall. More- the solid wall is indicated by r.
over, we analyze the effect of fluid elasticity and of the It is more convenient to work with non-dimensional
other two parameters of the PTT fluid model on the quantities obtained after scaling distances with the film
velocity profiles and the Stokes number. In each partic- thickness H̃, velocities with the cross-sectional average
ular case, we present velocity profiles and stress distrib- velocity Ũ, although both depend on the viscoelastic
utions across the film thickness and finally we calculate fluid model used and, hence, they are a priori un-
the Stokes number. When the cylinder radius becomes known.
 Moreover, we scale pressure or stresses with
much larger than the film thickness ( R̃/ H̃ → ∞), the η̃s + η̃ p Ũ/ H̃, where η̃ p and η̃s are the zero-shear
curvature tends to zero and the film flow over a flat polymer and solvent viscosities, while the dimensional
surface results and Cartesian coordinates are used. mean velocity Ũ is defined for both geometries as usual

 H̃
R̃±  R̃±
 H̃
Governing equations Ũ = ũzr̃dr̃ r̃dr̃. (1)
R̃ R̃
We consider a film of an incompressible PTT fluid
falling either on the outer or on the inner surface Since the average dimensional velocity is used as a
of a long, vertical cylinder of radius R̃. The flow is characteristic velocity, its dimensionless counterpart is
gravity-driven and the ambient air applies no force on unity and Eq. 1 reduces to
the film. We assume a fully developed flow resulting
in a film of constant thickness H̃ and that capillary R±1
1
and intermolecular forces are negligible compared to uzrdr = ±R + , (2)
2
viscous and elastic forces. Including the former forces R
in a fully developed flow would only alter the constant
pressure in the film without affecting the velocities or where the dimensionless cylinder radius is defined as
the stresses. We employ cylindrical coordinates with the R ≡ R̃/ H̃ and the upper/lower sign corresponds to film
z-axis along the axis of the cylinder and pointing in the flowing on the outer/inner cylindrical surface, a conven-
direction of gravity. Figure 1 illustrates the coordinate tion that will be kept throughout. Hence, the dimen-
system and the flow geometries examined herein. In sionless group that arises in the momentum  balance
Fig. 1a, the fluid ring is located on the outer surface of is the Stokes number, St = ρ̃ g̃ H̃ 2 / η̃ p + η̃s Ũ, whereas

Fig. 1 Schematic of the flow


geometry and the coordinate
system, for a fluid falling
downwards a on the outer
surface or b on the inner
surface of a long cylinder
1034 Rheol Acta (2009) 48:1031–1048

the Reynolds number is not relevant in fully developed fluid elasticity, λ̃, in dimensionless form becomes the
rectilinear flows. As discussed, combining the resulting Deborah number De = λ̃Ũ/ H̃, while the fraction of
value and the definition of the Stokes number with the solvent viscosity is β ≡ η̃
s /η̃o , where η̃0 = η̃ p + η̃s .
the imposed flow rate determines the film thickness. The stress function, Y tr τ , takes either the expo-
p
The assumption of fully developed, one-dimensional
nential or its linearized form
flow makes only the axial component of the velocity
vector non-zero and its dependence is reduced to the ⎧ 

⎪ ε De
radial direction only. Hence, the continuity equation ⎪

⎨ exp (1 − β) tr τ p

is satisfied identically and only the axial momentum
Y tr τ =
, (8)
balance is non-trivial, which in dimensionless form is: p ⎪
⎪ ε De

⎩1 + tr τ
1 ∂ (1 − β) p
(rτrz ) + St = 0, (3)
r ∂r
where τrz is the only non-trivial extra stress component where ε is a PTT model parameter that controls the
arising in the momentum balance, but it is coupled elongational and shear-thinning behavior of the mater-
with other non-trivial stress components through the ial, while ξ affects the slip of the polymeric chains with
constitutive model. respect to the medium and tr(τ ) is the trace of the
p
This equation is independent of the fluid model and polymeric stress tensor defined as
can be integrated subject to the no-shear boundary
condition τrz = 0 at the free surface (r = R ± 1), to give

the radial variation of the total shear stress. tr τ ≡ τ p,rr + τ p,zz + τ p,θ θ . (9)
p

St  
τrz = (R ± 1)2 − r2 (4)
2r For the fully developed film flow, γ̇ reduces to the
velocity gradient du z
and the non-trivial components of
Next, we split the total extra stress τrz into a purely dr
the most general PTT constitutive model reduce to
viscous part β dudr
z
and a polymeric contribution τ p,rz ,
duz
i.e., τrz = β dr + τ p,rz , and derive an expression for the

duz
latter in terms of the velocity gradient to be used in the Y tr τ τ p,rr = −ξ Deτ p,rz , (10)
p dr
constitutive model
St   duz
τ p,rz = (R ± 1)2 − r2 − β (5)

2r dr De duz
Y tr τ τ p,rz = (2 − ξ ) τ p,rr
The PTT model is used to describe the viscoelastic
p 2 dr
behavior of the material. Its more general version, De duz duz
−ξ τ p,zz + (1 − β) , (11)
which includes the non-affine chain motion (ξ = 0) 2 dr dr
and the presence of a Newtonian solvent (β = 0), in
dimensionless form becomes

duz

Y tr τ τ + De τ p = 2 (1 − β) γ̇ , (6) Y tr τ τ p,zz = De(2 − ξ ) τ p,rz , (12)
p p p dr
where τ and γ̇ are the part of the extra stress ten- while τ p,θ θ = 0. Therefore, the only additional poly-
p
sor due to the polymer and the rate of deformation meric stress components that must be considered in

tensors, respectively, and τ p is the Gordon–Schowalter addition to the shear stress are the normal stress in the
convected derivative of τ defined as streamwise and the shear direction. The latter is non-
p zero only when ξ = 0.



p
τ = − τ · L − LT · τ + ξ τ · γ̇ + γ̇ · τ ,
p Dt p p p p
Simplification of the general equations
(7) and numerical solution

where L ≡ ∇ u and γ̇ = 12 L + LT , u is the veloc- Equations 5 and 10–12 must be solved for the axial
ity vector and ∇ denotes the gradient operator. The velocity and the polymeric stress components subject
Rheol Acta (2009) 48:1031–1048 1035

to the no-slip condition at the wall of the cylinder and which is real and positive since 0 ≤ ξ , β ≤ 1 and sub-
the no traction boundary condition on the free surface. stitute Eq. 5 in Eq. 16 to obtain an expression for axial
First, we eliminate the normal stress component in the polymeric stress in terms of the velocity gradient
shear direction by dividing Eq. 10 by Eq. 12. Fortu-
nately, this eliminates also the term involving the trace 1−β
τ p,zz =
of the stress tensor and the velocity gradient yielding a 2Deξ
simple relation between the two normal stresses. ⎛  ⎞
 2
St   duz
× ⎝1− 1−α 2 (R ± 1)2 −r2 −β ⎠
ξ 2r dr
τ p,rr =− τ p,zz , (13)
2−ξ
(18)
while the trace of the stress tensor reduces to
Finally, we introduce Eqs. 5 and 18 in Eq. 12 and,

2 (1 − ξ ) using the linear variation of the PTT, we obtain
tr τ = τ p,zz . (14)
2−ξ ⎛ ⎛  ⎞⎞
 2
1 St   du
⎝1+ ⎝1− 1 − α 2 (R ± 1)2 −r2 −β
z ⎠⎠
Next, we divide Eq. 10 by Eq. 11 and after introducing χ 2r dr
Eq. 13 we obtain an expression of the axial polymeric
⎛  ⎞
stress in terms of the shear polymeric stress,  2
St   duz
× ⎝1− 1−α 2 (R ± 1)2 − r2 −β ⎠
Deξ τ p,zz
2
− (1 − β) τ p,zz + (2 − ξ ) Deτ p,rz
2
=0 (15) 2r dr
 
This expression is either linear or quadratic in τ p,zz , 1−β 2 St   duz duz
depending on whether the affinity parameter is zero or = α (R ± 1)2 − r2 − β (19)
2 2r dr dr
not, respectively. This suggests that we need to adopt
the following two different procedures. where χ = ξε(1−ξ
(2−ξ )
is another positive parameter, also
)
used by Alves et al. (2001), which combines the two
Simplification of equations when ξ = 0 model parameters ε and ξ . Similarly, using the expo-
nential variation of the PTT model, we obtain
In this case, we adopt to the present problem the proce-
⎛ ⎛  ⎞⎞
dure described by Alves et al. (2001). Now, Eq. 15 has  2
1 St   du
two possible solutions for τ p,zz , exp ⎝ ⎝1− 1−α 2 ⎠⎠
z
(R ± 1)2 −r2 −β
χ 2r dr
   ⎛  ⎞
 2
1−β 4ξ (2 − ξ )
De2 St   duz
τ p,zz = 1± 1− τ p,rz
2 , (16) × ⎝1− 1−α 2 (R ± 1)2 −r2 − β ⎠
2Deξ (1 − β)2 2r dr
 
1−β 2 St   duz duz
neither one of which can be rejected immediately. Con- = α (R ± 1)2 −r2 −β (20)
2 2r dr dr
sidering that (1) the air exerts no normal force, (2)
capillarity is negligible, (3) the radial velocity is zero, Either one of these two equations must be solved for
and (4) setting the datum of the film pressure to zero the axial velocity subject to the no-slip boundary condi-
shows that τ p,rr = 0 at the film surface. Then, Eq. 10 tion at the wall, i.e., uz (r = R) = 0.
requires that either the velocity gradient or τ p,rz is zero,
which coupled with the zero total shear stress condition
on the free surface shows that τ p,rz = τ p,zz = du z
=0 Simplification of equations when ξ = 0
dr
there. The requirement of zero polymeric stresses at the
free surface can be satisfied only when the minus sign is In this case, we adopt to the present problem the
retained in Eq. 16. Next, we introduce the parameter α procedure described by Cruz et al. (2005). Now, Eq. 15
for compactness, provides a single expression for τ p,zz in terms of τ p,rz ,

√ 2
2Deτ p,rz
2De ξ (2 − ξ )
α= , (17) τ p,zz = (21)
1−β (1 − β)
1036 Rheol Acta (2009) 48:1031–1048

Now we can substitute Eqs. 14, 21 into the simplified τ p,rz = 0 in Eq. 5, directly integrating it and applying
Eq. 12 yielding the no-slip boundary condition,
 
2 2
2Deτ p,rz 2Deτ p,rz duz  
Y = 2Deτ p,rz (22) St 1  2  r
(1 − β) (1 − β) dr uz = R − r2 + (R ± 1)2 ln . (24)
2 2 R
which involves only the polymeric shear stress and the The corresponding Stokes number is determined by
velocity gradient. The former can be eliminated by computing the average value of the axial velocity from
using Eq. 5 to obtain a single equation for the velocity this expression and applying Eq. 2,
gradient.
  2   
2De St   du 1
Y (R ± 1)2 − r2 − β
z −4 ±R +
(1 − β) 2r 2
dr St =    ,
  3 7 1
St   duz duz ± 3R + R2 ± R3 − (±1 + R)4 ln 1 ±
× (R ± 1)2 − r2 − β = (1 − β) (23) 4 2 R
2r dr dr
(25)

This equation also will have two forms depending on


whether we introduce the linearized or the exponential where, as before, the plus/minus sign stands for a
form for the function Y. A similar procedure can be film flowing on the outer/inner surface of the cylinder.
followed for viscoelastic film flow over a planar wall. Clearly, the Stokes number and, consequently, the film
thickness is a function of the cylinder radius. It may
be shown analytically that in the limit R → ∞, the
Numerical solution Stokes number St → 3, which is the well-known result
for flow over a flat plate, mentioned as the Nusselt
Equations 19 and 20 or 23 admit analytical solutions solution; see Bird et al. (2002), Chang et al. (2002), and
only in limited cases, which will be presented below. In Bontozoglou and Serifi (2008). The same tendency as R
general, they must be solved numerically. To this end, increases can be observed in Fig. 2, depicting the Stokes
we employ the finite element/Galerkin method. The number as a function of the cylinder radius for a film
axial velocity is approximated by three-node quadratic flowing either on the inner or the outer surface of the
basis functions (Lagrangian polynomials). A typical dis- cylinder (Eq. 25). Furthermore, for a film flowing on
cretization consists of 800 one-dimensional elements. the outer surface, no mathematical lower limit in the
The Stokes number that enters these equations de-
pends on the average velocity and is in fact an un-
known. Therefore, the profile of the axial velocity
is calculated simultaneously with the Stokes number
using the Newton–Raphson method. The additionally
required equation stems from the definition of the aver-
age velocity as a characteristic velocity for making the
variables dimensionless (Eq. 2). The iterations of the
Newton–Raphson method are terminated using 10−9
as a tolerance for the absolute error of the residual
vector. The code was implemented in Fortran 90 and
each calculation typically required 15–20 s to complete
in a dual core Xeon at 2.8 GHz.

Results and discussion

Film flow of a Newtonian fluid


Fig. 2 Dependence of St on the cylindrical radius for a
For a Newtonian fluid flowing down a vertical substrate, Newtonian fluid. Solid and dashed lines correspond to coating of
the velocity profile can be obtained by setting β = 1 and the outer and the inner surface, respectively
Rheol Acta (2009) 48:1031–1048 1037

cylinder radius can be imposed and the Stokes number Film flow when both β = 0 and ξ = 0
decreases down to St ≈ 1.25 at R = 0.2 and eventually
to zero as R → 0, as can be shown using Eq. 25 with In this case, Eq. 23 is greatly simplified yielding an
the plus sign. On the contrary, for a film flowing on the explicit expression for the velocity gradient
inner surface, the cylinder radius cannot be decreased  3
below R = 1, because in that limit R̃ ≈ H̃ and the duz 1 St  
= ε De2 (R ± 1)2 − r2
free surface of the film would have reached the axis dr 4 r
of symmetry of the cylinder invalidating the present St  
analysis. This can be deduced also from Eq. 25 with + (R ± 1)2 − r2 (27)
2r
the minus sign, showing that St → ∞ in that limit. The
opposite but monotonic trend of the Stokes number as This equation has the following analytic solution, after
R decreases for a film on the opposite sides of a cylinder applying the no-slip boundary condition,
is noteworthy. Physically, it can be explained as follows:
1
For a film on the outer surface, as R̃/ H̃ decreases while uz = St
keeping H̃ constant, the surface area per unit axial 16
  
length (2π R̃) on which the retarding viscous force acts 4R4 +ε De2 St2 R6 −6R4 (R±1)2 +2 (R±1)6
on the fluid decreases faster than the fluid volume per ×
R2
unit axial length π(2 R̃ H̃ + H̃ 2 ) on which gravity acts,  
leading to a larger dimensional average velocity and, 4r4 +ε De2 St2 r6 −6r4 (R ± 1)2 +2 (R±1)6

hence, a lower St. For a fluid on the inner surface, r2
the same surface area decreases slower than the fluid  
  R
volume π(2 R̃ H̃ − H̃ 2 ), which can decrease only down + 4 (R±1) −2+3ε De St (R±1) ln
2 2 2 2
r
to its lower limiting value of π H̃ 2 , leading to a smaller
average velocity and a larger St. (28)
It is very simple to assume the film thickness and
proceed to calculate R, St, and Ũ from the previous Setting ε = 0 in Eq. 28 reduces this case of the PTT fluid
expressions and figure, but this is not easy to enforce in model to the UCM fluid and the velocity profile to the
an experiment. On the other hand, it is more realistic to Newtonian one given in Eq. 24. This simplification is a
know the fluid properties, the flow rate, and the radius direct result of the decoupling of the stress components
of the cylindrical surface. Then the above solution can in the constitutive law when ε = 0. Equation 11 now
be completed only after the film thickness and the aver- yields an expression for τ p,rz which when inserted in
age velocity are determined. First, the film thickness is Eq. 5 shows that, even when β = 0, the momentum bal-
calculated by trial and error. For example, we assume a ance reduces to that of a Newtonian fluid. This verifies
value for H̃, determine R and, from it, we calculate the the well-known result that the velocity profile of the
Stokes number (Eq. 25). Then, we combine St with the UCM or Oldroyd-B fluids in simple flows is identical
flow rate, Q, to eliminate the average velocity, to the Newtonian one. In Fig. 3, we depict the effect of
fluid elasticity measured by combining the two relevant
parameters, De and ε, in the particular way they appear
St Q = ρ̃ g̃ H̃ 4 π (2R ± 1)/η̃. (26) in Eq. 27 or 28, ε1/2 De, used also by Oliveira and Pinho
(1999), on velocity profiles for a l-PTT fluid. In this
and all subsequent simulations, we will take R = 5 as a
This equation gives a new value for H̃ and we iterate typical dimensionless cylinder radius. The two families
until convergence. Finally, the average velocity is cal- of curves depict velocity profiles on the outer or the
culated from Ũ = Q/π(2 R̃ H̃ ± H̃ 2 ). inner surface of the cylinder. Having also computed the
line for ε1/2 De = 0.01, we have observed that it cannot
be distinguished from the one for a Newtonian fluid.
Linear PTT model (l-PTT) The main observation is that the velocity profile flattens
monotonically as ε1/2 De increases. This is caused by
Next, we will present results for the l-PTT fluid model the increased shear-thinning effect of the PTT fluid
starting with the simpler cases that admit an analytical model generated by the presence of ε1/2 De in the Y
solution and proceeding with more complex ones that function. The shear-thinning results in smaller shear
require a numerical solution. This solution is typically viscosity, but a larger velocity gradient at the wall.
validated by confirming the preceding analytical results. Combining these with the requirement that the average
1038 Rheol Acta (2009) 48:1031–1048

The polymeric stresses for this case are given in


Appendix 1. It is a bit more involved to complete the so-
lution by determining H̃ in the presence of viscoelastic-
ity because, additionally, the Deborah number involves
both H̃ and Ũ. Extending the procedure we described
above for a Newtonian fluid, we first eliminate Ũ by
combining De and St,
 
De St = λ̃ρ̃ g̃ H̃ η̃ p + η̃s (32)

Then we assume a value for H̃, determine R and the


product De St from Eq. 32, and iterate between Eqs. 26
and 29 to determine the converged values of De, St, and
H̃. Finally, the average velocity is calculated as before.
Fig. 3 Effect of ε1/2 De on the velocity profile in a l-PTT film,
when coating occurs on the outer or inner surface of the cylinder. In the limit R → ∞ making the substrate curva-
The other parameter values are (β, ξ, R) = (0, 0, 5) ture negligible, the above Eqs. 28 and 29 are greatly
simplified:

r
1  
dimensionless velocity is unity decreases the maximum uz = St r 1 − − ε De2 St3r −4 + 6r − 4r2 + r3 ,
velocities at the film free surface. 2 2
Equation 28 can be integrated analytically to obtain (33)
the average velocity and by imposing Eq. 2 we can
derive an explicit solution for the Stokes number in where now r stands for the distance from the wall and
terms of S1 , S2 , which are complicated functions of the setting either ε = 0 or De = 0 yields the Newtonian
cylinder radius R and the De and ε fluid parameters, profile, as it should. The second term in Eq. 33 is
introduced by the PTT model. The expression for the
 1/3 Stokes number is also much simpler
2
S2
3 52/3 51/3 Z
St =  1/3 St = − + , where
√  3 3 3 21/3 Z 3 22/3 ε De2
9S1 + 3 4S1 S2 + 27S1
2 4

1/3

 1/3 Z = 27 ε2 De4 + ε De3 10 ε + 729 ε2 De2 . (34)
√ 
9S21 + 3 4S31 S32 + 27S41
Although it is not obvious from Eq. 34, taking the
+ , (29)
21/3 32/3 S1 limit of this equation as ε → 0 or De → 0 yields St = 3,
the result for a Newtonian film over a flat plate. The
where polymeric stresses are given in the Appendix 2.
     Film flow when β = 0 and ξ = 0
S1 = ε De2 ±1 + 2R 3 + R ± 6 + R 16 + R ± 52

 

In this case, Eq. 23 reduces to the following cubic


+ R 73 ± 42R + 9R2 + 6R2 (±1 + R)2
equation for the velocity gradient,
  
R  
× ln ±1 ± 3 (±1 + R) 4    
±1 + R duz 3 2 St
  duz 2
2ψβ 3
− 3ψβ (R ± 1) − r
2 2
  dr r dr
24R2 (±1 + 2R) , (30)  
 
3 St   2 duz
+ ψβ (R ± 1)2 − r2 +1
2 r dr
    
1 4 (±1+ R)4 R 1 St   3 St  
S2 = −3−2R (±3+ R) ∓ ln . − ψ (R ± 1)2 − r2 − (R ± 1)2 − r2 = 0,
8 (±1+2R) ±1+ R 4 r 2r
(31) (35)
Rheol Acta (2009) 48:1031–1048 1039

Fig. 5 Velocity profile of l-PTT fluids scaled with the Newtonian


average film velocity, for (β = 0, 0.1, 0.5, 0.9), at (De, ε, ξ , R) =
(2, 0.25, 0, 5), when the film is in contact with the outer cylindrical
Fig. 4 Effect of β on velocity profile in a l-PTT film when surface
coating occurs either on the outer or on the inner surface of
a long cylinder. The parameter values are (De, ε, ξ , R) =
(5, 0.01, 0, 5)
film of the same thickness so that the Stokes number
remains equal to the corresponding for a Newtonian
ε De 2
where we have introduced the parameter ψ = (1−β) 2 fluid, which solely depends on the cylinder radius R,
to simplify the expression. Although this equation can as depicted in Fig. 2. Since here R = 5, this value is
be solved for the velocity gradient using the standard St = 2.74. Moreover, in Fig. 5 we use De = 2 and ε =
formulae for solving a cubic equation, the resulting 0.25 and depict only the case with the film on the outer
expression can be solved only numerically for the axial surface of the cylinder. We observe that decreasing β
velocity, as described previously. Setting ε = ψ = 0 in from 0.9 (i.e., when the Newtonian solvent dominates)
Eq. 35 reduces this case of the PTT fluid to the Oldroyd- down to zero produces an almost 7-fold increase in
B fluid and the velocity profile to the Newtonian one the maximum film velocity. In other words, the shear-
given in Eq. 24. The final outcome of this simplification thinning effect of the fluid, which is intensified by the
has been explained above. Figure 4 represents velocity larger value of ε used in the figure, increases its flow
profiles scaled so that the average film velocity remains rate dramatically over what would be expected for a
unity for different values of the parameter β. This Newtonian fluid. A similar picture is obtained when
parameter varies from zero, which is more appropriate other parameters affecting shear thinning are examined
for polymer melts, up to unity, when the Newtonian and also for a film flowing in the inner surface of the
fluid behavior is recovered. Note that in this figure cylinder.
De = 5 and ε = 0.01 and the two families of curves cor- The variation of St, calculated as discussed in
respond to film on the inner or outer cylinder surface. “Numerical solution”, with De is shown in Fig. 6 for
We can see that as β decreases from unity, the fluid the same parameter values as in Fig. 5. As explained
departs from its Newtonian (solvent) behavior and, for a Newtonian fluid, the Stokes number is larger
as in Fig. 3, the velocity profiles become flatter, their when the fluid is on the inner cylindrical surface, all
maxima smaller, and their slopes at the wall larger. All else being equal. For both geometries, we observe the
these result from the increasing contribution from the following: The St remains nearly constant to its value
polymer, which has a shear-thinning behavior. for a Newtonian fluid up to De ≈ 1 for all values of
Using the corresponding to the particular fluid- β . As De increases passed unity, the Stokes number
average velocity as a characteristic variable helps in decreases and more sharply as β → 0 despite the fact
understanding the previous parametric effects but hin- that the parameter considered the primary contributor
ders a quantitative comparison when the characteristic to shear thinning, ε = 0.01, is quite low. On the other
flow rate or average velocity remain constant between hand, when the solvent viscosity ratio tends to its upper
simulations. Figure 5 depicts the velocity profile when limit β → 1, the Stokes number is hardly affected by
the dimensionless velocity indicated by u∗z has been nor- the increase in De by three orders of magnitude. This
malized by the average film velocity for a Newtonian sharp decrease of the Stokes number is caused by the
1040 Rheol Acta (2009) 48:1031–1048

In the limit R → ∞ making the curvature negli-


gible, Eq. 35 can be greatly simplified. Nevertheless,
it is more straightforward to develop the correspond-
ing equations in Cartesian coordinates and following
the previous procedure integrate the velocity gradi-
ent to yield the axial velocity profile over a flat plate
analytically,

1  
uz = 3W −M− (W) + M+ (W)
8Q1
 
− 3W0 −M− (W0 ) + M+ (W0 )
 
+ 9Q1 (−1 + r) M− (W) + M+ (W)
 ! St r

Fig. 6 Dependence of St on De in a l-PTT film for (β = 0, + 9Q1 M− (W0 )+ M+ (W0 ) + r 1−


β 2
0.1, 0.5, 0.9). The parameter values are (ε, ξ , R) = (0.01,
0, 5). Lines/symbols correspond to coating of the outer/inner (36)
surface

where
increased average fluid velocity, when the fluid exhibits
stronger shear thinning as De increases or β decreases, 1 St (1−β)3 1 (1−β)2
as explained above. In fluids with larger ε, the value Q1 = − , Q2 = ,
4 εβ 4 De2 6 εβ 3 De2
of De for which the Stokes number starts to decrease 
will decrease as well, because ε and De affect the shear W(r) = Q32 + Q21 (−1+r)2 ,M± (W) = (Q1 (1−r)±W)1/3
thinning of the fluid in the same direction through the
Y function. (37)
Analytical expressions for the viscoelastic stresses
are presented in the Appendix 2. It is noteworthy and
that these expressions do not involve the axial velocity

profile; only the fluid properties, St and De, which can
W0 ≡ W (r = 0) = Q32 + Q21 , M± (W0 ) ≡ (Q1 ± W0 )1/3 ,
be calculated as described above. Both the shear and
normal polymeric stress components are zero at the film (38)
free surface and monotonically increase towards the
cylindrical wall. Moreover, they decrease as the solvent where r is the coordinate normal to the wall. The
viscosity ratio β is increased, becoming zero when the last term in Eq. 36 is the Newtonian velocity profile.
polymer fraction in the fluid is zeroed, as expected. The Unfortunately, it was not possible to obtain the average
curve depicting the radial variation of τ p,rz is convex velocity from Eq. 36 and compute from it an expression
whereas that of τ p,zz is concave for small values of β, for the Stokes number. This was performed numeri-
both becoming linear as β increases. When the film cally. The analytic expressions for the stresses are given
resides in the inner surface of the cylinder, the same in the Appendix 2.
variation of the stresses in the radial direction and the
same dependence on β arises, except that now the shear
polymeric stress assumes negative values away from the Film flow with β = 0 and ξ = 0
free surface, as expected for Eq. 5. The corresponding
plots are not given for conciseness but are available When β = 0, Eq. 19 is greatly simplified yielding an
from the authors upon request. explicit expression for the velocity gradient

⎛  ⎞
   2 2  
  St   2
4 1+
2 ⎝
1 − 1 − α2
St
(R ± 1)2 − r2 ⎠− α (R ± 1)2 − r2
χ 2r χ 2r
duz
=   (39)
dr α 2 St (R ± 1)2 − r2
Rheol Acta (2009) 48:1031–1048 1041

Although this equation is linear in the velocity gradi- small values of ξ lead to much smaller in magnitude
ent it can be solved only numerically for the axial veloc- values of τ p,rr compared to those of τ p,zz , while its
ity, as described previously. The two families of velocity relative variation with ξ is the most noticeable among
profiles are presented in Fig. 7 with ξ as a parameter, the three polymeric stress components.
while keeping De = 1 and ε = 0.2. As the molecular Figure 8a–b presents the dependence of the Stokes
slip parameter increases, the velocity profiles become number on the Deborah number for different values
flatter, with a smaller maximum and a larger slope at of the affinity parameter ξ for a film on the outer
the wall. In other words, we observe the same trends cylindrical surface. We present two cases: one with a
attributed earlier to the other fluid properties inducing low value of the shear-thinning parameter, ε = 0.01,
shear thinning (De, ε). Our code did not converge to a and another one with a larger value, ε = 0.2. All lines
solution at larger values of ξ , so the variation observed representing the dependence of St on De terminate at a
in both families of curves is not as pronounced as critical value of De, which increases as ε increases or ξ
before. In fact, the maximum value of ξ for which a con- decreases. Beyond these values of De, our code did not
verged solution can be obtained depends on the other converge to solution. The line connecting the terminal
two parameters (De, ε), as we will explain next. Critical points has also been drawn in both figures. Steady, one-
conditions arise and a steady-state solution ceases to dimensional solutions cannot be computed below this
exist, when the term inside the square root in Eq. 39 line. We observe that increasing either ξ or ε decreases
vanishes. Further increase of α would lead to complex the Stokes number for the same value of De, a direct
values of the shear rate. The critical conditions occur result of the larger flow rates achieved by the intensified
first at the substrate wall, where the shear rate is always shear thinning induced by both these parameters. As in
maximized. Consequently, the critical conditions arise Fig. 6, at low De, the Stokes number retains its New-
first when α = 1±2R2R 1
St
. Comparing this expression with tonian value depending solely on R, but, at a particular
the wall shear stress, derived from Eq. 4, we find that break point which depends on both ξ and ε, it decreases
the critical (maximum) shear stress is τrz |r=R = α1 while abruptly and more so as ξ increases. Similar are the
the corresponding critical
shear rate is deduced from curves of St vs. De with ξ and ε as parameters for a fluid
" 4+7 ξε(1−ξ )
z" (2−ξ ) (1±2R)St on the inside of the cylindrical surface.
Eq. 39, du " = . Unfortunately, as
dr max 4R2 In the limit R → ∞ making the curvature negligible,
we mentioned already, it is not possible to integrate Eq. 39 can be greatly simplified. However, it is more
Eq. 39 and obtain an analytical expression relating the convenient to work directly in Cartesian coordinates
critical values of the Stokes and Deborah numbers, and eventually show that the velocity profile is given
although this is possible for the pipe flow as discussed analytically by
by Alves et al. (2001). The analytical expressions for   
the stress components are given in Appendix 1. The 2St δ 2 2
uz = 2 r (r − 2) − 1 +
corresponding figures are available upon request from δ χ χ
the authors. We will only mention here that the very  
1 + F (r; δ)
× −F (r; δ) + F (0; δ) + ln
1 + F (0; δ)
(40)
where

 
F (r; δ) = 1−δ 2 (−1+r)2 and δ = 2DeSt ξ (2−ξ ) > 0.
(41)

Equation 40 can be integrated to finally derive an ex-


plicit expression for the Stokes number as a function of
the dimensionless parameters δ, χ defined above.

3 δ3χ
St = − √
!
Fig. 7 Effect of ξ on the velocity profile in a l-PTT film, when the δ 2δ 2 +3 −2+ 1−δ 2 (2+χ) +3 (2+χ)arcsin (δ)
film is either on the outer or on the inner cylindrical surface. The
parameter values are (De, ε, β, R) = (1, 0.2, 0, 5) (42)
1042 Rheol Acta (2009) 48:1031–1048

Fig. 8 Dependence of St on De for (ξ = 0.001, 0.01, 0.05, 0.1, 0.2), in a l-PTT film, when coating occurs on the outer cylindrical surface.
The parameter values are (β, R) = (0, 5), and for (a) ε = 0.01, while for (b) ε = 0.2

Both Eqs. 40 and 42 reveal that the positive para- expressions for the stress components are given in the
meter δ must also be less or equal to unity, at the Appendix 2.
planar wall, r = 0, where the shear stress is maxi-
mized. This requirement necessitates that either 2De
St ≤ 1, in which case all physically permissible ξ val- Film flow with β = 0 and ξ = 0
 make
ues δ ≤ 1, otherwise
 ξ has an upper limit ξ ≤

This is the most general case for a l-PTT fluid model.
1 − 1 − 1/(2De St) /2. Similar limitations on the
2
Now Eq. 19 governs the velocity profile, which can
allowable values of the parameters ξ , De, and ε should be computed as already described. This presents no
exist for the flow around a cylindrical surface and additional problems, but results showing its paramet-
were computed and discussed in relation to Fig. 8. ric dependence will not be included herein, because
In fact, the curves connecting the terminal computed they are not qualitatively different from those already
points approximately follow a relation St = 1/De, even presented.
though the geometry is cylindrical. All these results and
observations are directly related to the non-monotonic
behavior of the shear stress vs. shear rate curves in
simple shear flow of the PTT fluid model, when ξ = 0,
i.e., when the lower convective derivative is present
in the constitutive law. The maximum of such curves
defines a critical shear rate value above which a steady
solution to the flow problem does not exist (Alves et al.
2001). This critical shear rate signifies the onset of
the so-called constitutive instability, which has been
studied by Kolkka et al. (1988) and Georgiou and
Vlassopoulos (1998), among others for the Johnson–
Segalman fluid model to which the non-affine PTT
model reduces, when ε = 0. This type of instability has
been suggested as a mechanism that leads to common
instabilities found in polymer processing flows. When a
solvent viscosity is added to the model, a second critical
value (local minimum) of the shear stress arises. This
would be the case of a PTT fluid when both β = 0 Fig. 9 Effect of ε1/2 De on the velocity profile in an e-PTT film,
and ξ = 0. Shear banding instabilities exist between when coating occurs on the outer or inner cylindrical surface. The
these two extreme values of shear rate. The analytical remaining parameter values are (β, ξ , R) = (0, 0, 5)
Rheol Acta (2009) 48:1031–1048 1043

Fig. 10 Dependence of St on a the Deborah number with R = 5 β, ξ ) = (0.01, 0, 0). Lines/symbols correspond to coating of the
and b the cylinder radius when De = 5, for either a linear or an outer/inner cylindrical surface
exponential PTT model. The remaining parameter values are (ε,

Exponential PTT model (e-PTT) cylinder radius becomes about 30 times greater than
the film thickness, the Stokes number approaches its
Film flow when both β = 0 and ξ = 0 asymptotic value, which is the one attained for flow
over a flat surface. Of course, this value of St is different
Now the velocity gradient is deduced from Eq. 20 and from the Newtonian one and different between the
given by exponential and the linear version of the PTT model.
   As explained for a Newtonian fluid, flow on the outer
duz St   2 cylindrical surface can be modeled for arbitrarily small
= exp 2ε De 2
(R ± 1) − r
2 2
dr 2r values of R and gives an ever decreasing St, whereas on
the inner surface an asymptote arises as R → 1, when
St  
× (R ± 1)2 − r2 (43) the motion is strongly retarded increasing St without
2r bound. Analytical expressions for the viscoelastic stress
This equation is solved numerically. In Fig. 9, we show components are given in the Appendix 1. Increasing De
the influence of elasticity on the velocity profile for the
same parameter values as in Fig. 3. All the changes in
the velocity profiles as ε1/2 De increases reported for
the l-PTT model are observed here as well, but they
are much more pronounced; see for example the nearly
plug velocity for half the film thickness when ε1/2 De =
10. This is a consequence of the stronger influence of
both ε and De on fluid shear thinning through the ex-
ponential form of the Y function. This implies a larger
increase in the average film velocity and decrease of the
Stokes number due to the enhanced shear thinning of
the e-PTT fluid.
In Fig. 10a, we depict the effect of De on St. Increase
in the Deborah number intensifies shear thinning and
facilitates higher flow rates, which through the average
film velocity decrease St and more so in the exponential
form of the model. The influence of the cylinder radius
Fig. 11 Dependence of St on ε, when either a linear or an
on St is shown in Fig. 10b for a relatively large value
exponential PTT model is employed. The parameter values are
of De. Again, the e-PTT model leads to smaller Stokes (De, β, ξ , R) = (5, 0, 0, 5). Lines/symbols correspond to coating
values than the linear one irrespective of R. When the of the outer/inner cylindrical surface
1044 Rheol Acta (2009) 48:1031–1048

The effect of ε on St is presented in Fig. 11 for


both the linear and the exponential versions of the
PTT model. Clearly, the e-PTT leads to lower Stokes
values than the l-PTT. When ε increases above 0.1,
shear thinning is very pronounced and the geometry of
the wall not as important. Therefore, St decreases with
the same rate in both models and interestingly assumes
almost the same values irrespective of whether the film
is in the inner or the outer cylindrical surface. On the
other hand, as ε decreases, ε → 0.001, the deviation in
St between e-PTT and l-PTT decreases and the values
of St approach those of the Newtonian fluid for R =
5 on the inner or outer cylindrical surface. Since the
dependence of both stress components on ε arises only
through the Y function it remains monotonic without
Fig. 12 Velocity profile of the e-PTT fluid scaled with the
Newtonian average film velocity, for (β = 0.1, 0.5, 0.9), when (De,
any renormalization irrespective of the version of the
ε, ξ , R) = (2, 0.25, 0, 5), when coating occurs on the outer surface PTT model and is not discussed further.
of a cylinder Similarly, for a film flowing over a flat surface and
following the e-PTT model we have shown that its
velocity has the following analytic solution
monotonically decreases the shear stress due to shear
thinning for both PTT models. This effect is slightly     
exp 2ε De2 St2 − exp 2ε De2 St2 −1 + r2
more pronounced for the exponential version of the uz = (44)
4ε De2 St
model. Increasing De increases the axial normal stress
due to fluid elasticity but the trend is not monotonic This expression can be integrated to obtain the average
for the e-PTT. This unexpected non-monotonic depen- film velocity and from it the Stokes number. The result-
dence of τ p,zz on De is attributed to the dimensionaliza- ing expression is not explicit in St, but one can obtain its
tion of the stresses coupled with the fact that inspection values numerically as a function of the fluid parameters,
of the stress equations shows that both stresses depend ε and De,
on the dimensionless group ε1/2 De but τ p,zz depends
  √ √

additionally on De. The monotonic dependence of τ p,zz


4ε1/2 DeSt exp 2ε De2 St2 − 2πer f i 2ε DeSt
on De is recovered if stresses are measured with
 respect

to the value of shear stress at the substrate τ p,rz r=R . − 16ε3/2 De3 St2 = 0, (45)

Fig. 13 a Dependence of St on De for various solvent viscosity (5, 0.01, 0.1, 0) when both linear and exponential PTT models
ratios (β = 0, 0.1, 0.5, 0.9) in an e-PTT film with (ε, ξ , R) = are employed, while lines/symbols correspond to coating of the
(0.01, 0, 5). b Dependence of St on R with (De, ε, β, ξ ) = outer/inner cylindrical surface
Rheol Acta (2009) 48:1031–1048 1045

where the imaginary error function is defined as is in excess, to 0.1, where the polymer is in excess,
erfi(z) = erf(iz)/i. increases the maximum velocity at the free film surface
by a factor of almost 7. The effect of De on St is
Film flow when β = 0 and ξ = 0 shown in Fig. 13a with β as a parameter. We observe
a similar behavior to that in Fig. 6, except that the
Again the velocity gradient is deduced from Eq. 20 and decrease in St is sharper with the e-PTT model. This is
it is given by depicted quantitatively in Fig. 13b, where we also give
  2  the dependence of St on R for a particular value of the
St   du
exp 2ψ (R ± 1) − r2 − β
2 z
solvent viscosity β = 0.1. The asymptotic increase of St
2r dr as the radius decreases to unity and the film is in the
  inner cylindrical surface is observed again. The radial
St   duz duz
× (R ± 1)2 − r2 − β − (1 − β) =0 profiles of the polymeric stress components are similar
2r dr dr
to those for the l-PTT fluid model, except that, with
(46) the e-PTT fluid, the stress magnitudes are smaller in
This equation is solved numerically for the velocity general.
profile, which is depicted in Fig. 12 when velocity is
normalized with the Newtonian average film velocity Film flow with β = 0 and ξ = 0
for the same R and a Newtonian fluid so that the Stokes
number remains constant. In this way, the effect of Under these parameter values, Eq. 20 is greatly sim-
shear thinning on the flow rate becomes clearer. We plified yielding an explicit expression for the velocity
observe that decreasing β, from 0.9, where the solvent gradient

⎛ ⎛  ⎞⎞ ⎛  ⎞
 2  2
1 St   St  
2 exp ⎝ ⎝1 − 1 − α 2 (R ± 1)2 − r2 ⎠⎠ ⎝1 − 1 − α 2 (R ± 1)2 − r2 ⎠
χ 2r 2r
duz
= (47)
dr St  
α2 (R ± 1)2 − r2
2r

In Fig. 14, we present velocity profiles with ξ as necessary also in the l-PTT model. Following a similar
a parameter keeping its value relatively low in order derivation, one may deduce that under critical

condi-
" ε(1−ξ )
z"
to obtain a converged solution. This was found to be exp ξ (2−ξ ) (1±2R)St
tions the wall shear stress is du
dr "
= R
.
max
In general, increasing the affinity parameter makes the
velocity profiles flatter and increases their slopes at the
solid wall, i.e., it acts in the same way as increasing ε
or De. Exact solutions for the viscoelastic stresses are
presented in the Appendix 1. The non-zero value of ξ
gives rise to a stress component in the shear direction,
although it remains fairly small due to small values of ξ ,
while the axial normal stress has the largest magnitude
which decreases faster as De, ε, or ξ increases. The
profiles of the three stress components are qualitatively
similar to those already presented and discussed, but
increasing the shear thinning of the fluid affects them
more than with the l-PTT fluid. It is noteworthy that ξ
contributes to decrease the deviation in the predictions
Fig. 14 Effect of ξ on the velocity profile in an e-PTT film, when
between the two forms of the PTT model.
the film is either on the outer or on the inner cylindrical surface. Figure 15a–b shows the variation of St with De hav-
The parameter values are (De, ε, β, R) = (1, 0.2, 0, 5) ing ξ as a parameter for either low value of ε (ε = 0.01)
1046 Rheol Acta (2009) 48:1031–1048

Fig. 15 Dependence of St on De for (ξ = 0.001, 0.01, 0.05, 0.1, 0.2), in an e-PTT film, when coating occurs on the outer cylindrical
surface. The parameter values are (β, R) = (0, 5) and for (a) ε = 0.01, while for (b) ε = 0.2

or a relatively large one (ε = 0.2). In both figures, the Integrating this equation to derive the average ve-
film is in contact with the outer cylindrical surface, locity, and from there the Stokes number, yields the
and similar are the corresponding results for a film in following implicit expression for St:
contact with the inner cylindrical surface. As with the l-  √ 
PTT model, the ξ values for which converged solutions 1 − 1 − α2 
2 χ St exp −1 + α 2 + 1 − α 2 + χ
are possible decrease as De increases or ε decreases, χ
although with the e-PTT fluid converged solutions are 

possible with larger values of De. All curves exhibit −2 χ 1 − α 2 − 2 χ 2 + 2 χ 2 St (1 + 2χ) + α 2 = 0


nearly constant values of St for small De and at a
critical De, St decreases abruptly. As ξ increases, the (49)
St vs. De lines attain a steeper slope. For a film in the
inner cylindrical wall, the computed Stokes numbers Film flow with β = 0 and ξ = 0
are somewhat larger than the corresponding ones for
a film in the outer wall, especially for low De. The This is the most general case for an e-PTT fluid model.
terminal points in all curves have been connected with Now the entire Eq. 20 must be integrated to determine
a curve, below which steady, one-dimensional solutions the velocity profile. Since this presents no additional
cannot be computed. problems and produces qualitatively similar results, it
In the case of a planar wall, R → ∞, an e-PTT fluid will not be discussed further.
with β = 0 and ξ = 0 admits the following analytical
solution:
Conclusions
#    
2St 1
uz = −χ exp 1 − 1 − α (1 − r)
2 2
Analytical and numerical solutions have been derived
α 2 χ
 for the one-dimensional, fully developed flow of a film
1 
  
2
+ χ exp 1− 1−α 2 + exp in contact with the inner or the outer cylindrical surface
χ χ or a planar wall using the PTT fluid either in its lin-
     ear or exponential form. We have presented indicative
1
× −Ei − 1 + 1 − α 2 (1 − r)2 results for the velocity profiles, the stress profiles, and
χ
 the Stokes number and studied how they are affected
1 
$
+Ei − 1 + 1 − α2 , (48) by the flow geometry and the constitutive model pa-
χ rameters. The shear thinning induced by any of the
parameters ε, ξ , and De results in higher flow rate and,
%∞ consequently, lower Stokes numbers. Results for either
e−t
where Ei (ω) = − −ω t dt. cylindrical configuration approach those of the planar,
Rheol Acta (2009) 48:1031–1048 1047

as the ratio of the cylinder radius to film thickness R DeSt2  


increases. As R decreases, the Stokes number either τ p,zz = 2
(R ± 1)2 − r2 (51)
2r
decreases to zero as R → 0 for a film on the outer
cylindrical surface or increases and approaches infinity Linear PTT model for β = 0 and ξ = 0
as R → 1+ for a film on the inner cylindrical surface,
irrespective of the fluid model. In all geometries, when In this case, explicit expressions exist only for the l-PTT
ξ = 0, an upper limit in De exists above which a steady, model.
one-dimensional solution cannot be computed. This St   β
critical value increases with ε and decreases with ξ and τrz = (R ± 1)2 − r2 −
2r 12r3
is related to the critical value of shear stress at which ⎡
 2 
the shear stress is maximized in simple shear flow, i.e., ⎢ 6r −r + (±1 + R) St
2 2
432/3r6
the constitutive instability phenomena. ×⎣ −
1/3
β
± +K±
Acknowledgement This paper is part of the 03ED276 research ⎤
project, implemented within the framework of the “Reinforce-
1

1/3 ⎥
ment Programme of Human Research Manpower” (PENED)
+ 3 231/3 ± +K± ⎦
and co-financed by Greek (20%) and EU Funds (80%). β ψ

(52)
Appendix 1: Analytical expressions for the polymeric #
2De St  
stresses τ p,zz = (R ± 1)2 − r2
1−β 2r
 2 2 
Cylindrical geometry β 6r −r + (±1 + R)2 St

12r3 β
Linear and exponential models for β = 0 and ξ = 0 $
432/3r6 1 1/3  ± 
2
± 1/3
− + 23  +K
The same expressions hold for both the l-PTT and the (± +K± )1/3 β 3 ψ
e-PTT models (53)
St  
τ p,rz = (R ± 1)2 − r2 (50) where K± and ± are defined as
2r

√   
K± = 3 r16 β 9 ψ 3 8r2 + 27 (1 ∓ r ± R)2 (±1 + r + R)2 St2 (−1 + β)2 βψ (54)

  
± = −9r8 (∓1 + r − R) (±1 + r + R) St (−1 + β) β 5 ψ 2 1− St2
1 − α 2 4r 2 (R ± 1) − r
2 2 2
τ p,rr = − (58)
(55) 2De (2 − ξ )

Linear and exponential models for β = 0 and ξ = 0


Appendix 2: Cartesian geometry (limiting case of
R → ∞)
For both the l-PTT and the e-PTT models, the following
expressions hold
Linear and exponential models for β = 0 and ξ = 0
St  
τ p,rz = (R ± 1)2 − r2 (56) The expressions for the viscoelastic stresses coincide:
2r
τ p,rz = St (1 − r) (59)
  
St2
1 − 1 − α 2 4r 2 2
2 (R ± 1) − r
2
τ p,zz = (57)
2Deξ τ p,zz = 2DeSt2 (1 − r)2 (60)
1048 Rheol Acta (2009) 48:1031–1048

Linear PTT model for β = 0 and ξ = 0 Bird RB, Stewart WE, Lightfoot EN (2002) Transport phenom-
ena, 2nd edn. Wiley, New York
Bontozoglou V, Serifi K (2008) Falling film flow along steep two-
In this case, explicit expressions exist only for the l-PTT
dimensional topography: the effect of inertia. J Multiphase
model. Flow 34(8):734–747
 ,  Borkar A, Tsamopoulos J, Gupta S, Gupta R (1994) Spin coating
τ p,rz = −β 3 Q1 (1 − r) + Q32 + Q21 (1 − r)2 of viscoelastic and nonvolatile fluids over a planar disk. Phys
Fluids 6(11):3539–3553
,   Chang HC, Demekhin E, Saprikin S (2002) Noise-driven wave
transitions on a vertically falling film. J Fluid Mech 462:255–
+ 3
Q1 (1 − r) − Q32 + Q21 (1 − r)
2
(61) 283
Cruz DOA, Pinho FT, Oliveira PJ (2005) Analytical solutions
for fully developed laminar flow of some viscoelastic liquids
,  with a Newtonian solvent contribution. J Non-Newton Fluid
β
τ p,zz = −2 De 3
Q1 (1−r)+ Q32 + Q21 (1−r)2 Mech 132:28–35
1−β Dimakopoulos Y, Tsamopoulos J (2004) On the gas-penetration
,   in straight tubes completely filled with a viscoelastic fluid. J
Non-Newton Fluid Mech 117/2–3:117–139
+ 3
Q1 (1−r)− Q32 + Q21 (1−r) 2
(62) Georgiou GC, Vlassopoulos DV (1998) On the stability of the
simple shear flow of a Johnson–Segalman fluid. J Non-
Newton Fluid Mech 75:77–97
where Q1 and Q2 are given by Eq. 37. Karapetsas G, Tsamopoulos J (2008) Steady extrusion of vis-
coelastic materials from an annular die. J Non-Newton Fluid
Linear and exponential models for β = 0 and ξ = 0 Mech 154:136–152
Kistler SF, Schweitzer PM (eds) (1997) Liquid film coating.
Chapman and Hall, London
Finally, when β = 0 and the affinity parameter be- Kolkka RW, Malkus DS, Hansen MG, Ierley GR, Worthing RA
comes non-zero ξ = 0, analytical solutions exist for the (1988) Spurt phenomena of the Johnson–Segalman fluid and
polymeric stresses employing either the linear or the related models. J Non-Newton Fluid Mech 29:303–335
exponential variation of the PTT model, which coincide Oliveira PJ, Pinho FT (1999) Analytical solutions for fully devel-
oped channel and pipe flow of Phan-Thien Tanner fluids. J
to each other: Fluid Mech 387:271–280
Phan-Thien N (1978) A nonlinear network viscoelastic model. J
τ p,rz = St (1 − r) (63) Rheol 22:259–283
Phan-Thien N, Tanner RI (1977) A new constitutive equation
 derived from network theory. J Non-Newton Fluid Mech
1− 1 − α 2 (1 − r)2 2:353–365
τ p,zz = (64) Quere D (1999) Fluid coating on a fiber. Annu Rev Fluid Mech
2Deξ 31:347–384
Quinzani LM, Armstrong RC, Brown RA (1995) Use of coupled
 birefringence and LDV studies of flow through a planar
1− 1 − α 2 (1 − r)2 contraction to test constitutive equations for concentrated
τ p,rr = − (65)
2De (2 − ξ ) polymer solutions. J Rheol 34(5):1201–1228
Ruschak KJ (1985) Coating flows. Annu Rev Fluid Mech 17:65–
89
Ruyer-Quil C, Treveleyan P, Giogiutti-Dauphiné F, Duprat C,
References Kalliadasis S (2008) Modeling film flows down a fibre. J Fluid
Mech 603:431–462
Alves MA, Pinho FT, Oliveira PJ (2001) Study of steady pipe and Saprykin S, Koopmans RJ, Kalliadasis S (2007) Free-surface
channel flows of a single-mode Phan-Thien–Tanner fluid. J thin-film flows over topography: influence of inertia and vis-
Non-Newton Fluid Mech 101:55–76 coelasticity. J Fluid Mech 578:271–293
Baaijens FPT (1993) Numerical analysis of start-up planar and Saramito P, Piau JM (1994) Flow characteristics of viscoelastic
axisymmetric contraction flows using multi-mode differen- fluids in an abrupt contraction by using numerical modeling.
tial constitutive models. J Non-Newton Fluid Mech 48:147– J Non-Newton Fluid Mech 52:263–288
180 Sunwoo KB, Park SJ, Lee SJ, Ahn KH, Lee SJ (2001) Numerical
Baaijens FPT, Selen SHA, Baaijens HPW, Peters GWM, Meijer simulation of three-dimensional viscoelastic flow using the
HEH (1997) Viscoelastic flow past a confined cylinder of a open boundary condition method in coextrusion process. J
low-density polyethylene melt. J Non-Newton Fluid Mech Non-Newton Fluid Mech 99:125–144
68:173–203 Tsamopoulos J, Chen MF, Borkar A (1996) On the spin-coating
Baloch A, Townsend R, Webster MF (1996) On vortex develop- of viscoplastic Fluids. Rheol Acta 35(6):597–615
ment in viscoelastic expansion and contraction flows. J Non- Vlachogiannis M, Bontozoglou V (2001) Observations of solitary
Newton Fluid Mech 65:133–149 wave dynamics of film flows. J Fluid Mech 435:191–215

View publication stats

You might also like