You are on page 1of 43

J. Fluid Mech. (2017), vol. 812, pp. 10761118.

c Cambridge University Press 2017 1076


doi:10.1017/jfm.2016.839

Effect of viscoelasticity on the soft-wall


transition and turbulence in a microchannel

S. S. Srinivas1 and V. Kumaran1,


1 Department of Chemical Engineering, Indian Institute of Science, Bangalore 560 012, India

(Received 19 May 2016; revised 6 December 2016; accepted 6 December 2016;


first published online 12 January 2017)

The modification of soft-wall turbulence in a microchannel due to small amounts


of polymer dissolved in water is experimentally studied. The microchannels are of
rectangular cross-section with height 160 m, width 1.5 mm and length 3 cm,
with three walls made of hard polydimethylsiloxane (PDMS) gel, and one wall made
of soft PDMS gel with an elasticity modulus of 18 kPa. Solutions of polyacrylamide
of molecular weight 5 106 and mass fraction up to 50 ppm, and of molecular weight
4 104 and mass fraction up to 1500 ppm, are used in the experiments. In all cases,
the solutions are in the dilute limit below the critical overlap concentration, and the
solution viscosity does not exceed that of water by more than 10 %. Two distinct types
of flow modifications are observed below and above a threshold mass fraction for
the polymer, wt , which is 1 ppm and 500 ppm for the solutions of polyacrylamide
with molecular weights 5 106 and 4 104 , respectively. At or below wt , there
is no change in the transition Reynolds number, but there is significant turbulence
attenuation, by up to a factor of 2 in the root-mean-square velocities and a factor
of 4 in the Reynolds stress. When the polymer concentration increases beyond wt ,
there is a decrease in the transition Reynolds number and in the intensity of the
turbulent fluctuations. The lowest transition Reynolds number is 35 for the solution
of polyacrylamide with molecular weight 5 106 and mass fraction 50 ppm (in
contrast to 260290 for pure water). The fluctuating velocities in the streamwise and
cross-stream directions are lower by a factor of 5, and the Reynolds stress is lower
by a factor of 10, in comparison to pure water.

Key words: complex fluids, instability, transition to turbulence

1. Introduction
The transition to turbulence in soft-walled tubes and channels, due to a dynamical
instability involving a coupling between the fluid flow and wall dynamics, could
take place at a Reynolds number significantly lower than the transition Reynolds
number for rigid tubes and channels (Kumaran 2000, 2003, 2015; Shankar 2015).
The transition was first reported by Lahav, Eliezer & Silberberg (1973) and Krindel
& Silberberg (1979) in the flow through a gel-walled tube. The authors observed a
transition from a laminar flow to a more complicated velocity profile, as evidenced

Email address for correspondence: kumaran@chemeng.iisc.ernet.in

Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1077
by the breakup of a stream of dye introduced at the centre of the tube. The pressure
difference required to drive the flow after transition was also larger than that required
for a laminar flow at the same flow rate. There have been some doubts raised (Yang
et al. 2000) that the observations of Krindel & Silberberg (1979) could be explained
by the tube deformation due to the applied pressure gradient instead of a dynamical
transition. However, the experiments of Krindel & Silberberg (1979) motivated many
subsequent studies on the flow through soft tubes and channels.

1.1. Stability and transition in soft-walled conduits


Linear stability studies predict that the transition Reynolds number depends on a
dimensionless parameter = Gh2 /2 , which is the ratio of the wall elasticity and
fluid viscosity, and is independent of the flow velocity. Here, and are the fluid
density and viscosity, G is the elasticity modulus of the wall material and h is the
characteristic length (tube diameter or channel height). The flow could go unstable
even in the limit of zero Reynolds number (Kumaran, Fredrickson & Pincus 1994;
Kumaran 1995; Shankar & Kumaran 2001b; Gkanis & Kumar 2003, 2005; Chokshi
& Kumaran 2008). Since the interface is deformable, there is an energy transfer from
the mean flow to the fluctuations due to the shear work done at the interface, and
the flow is destabilised when the parameter V/Gh exceeds a critical value. Here, V
is the average flow velocity. The transition Reynolds number is proportional to
for this low-Reynolds-number instability. At high Reynolds number, there are two
important modes of destabilisation. The high-Reynolds-number inviscid instability
(Kumaran 1996; Shankar & Kumaran 1999, 2000) is a modification of the instability
in rigid tubes/channels, and the transition Reynolds number is proportional to 1/2 .
For the high-Reynolds-number wall-mode instability (Kumaran 1998; Shankar &
Kumaran 2001a, 2002; Chokshi & Kumaran 2009; Gaurav & Shankar 2009, 2010),
there is a wall layer of thickness Re1/3 where viscous forces are important. The flow
is destabilised due to the shear work done by the mean strain at the interface, and
the transition Reynolds number scales as 3/4 .
Experimental results for the viscous instability (Kumaran & Muralikrishnan
2000; Muralikrishnan & Kumaran 2002; Eggert & Kumar 2004) are in quantitative
agreement with theoretical predictions. The low-Reynolds-number transition has also
been used to enhance mass transfer from the interface (Shrivastava, Cussler & Kumar
2008). More recently, the high-Reynolds-number transitions in a soft tube (Verma
& Kumaran 2012) and in a microchannel (Verma & Kumaran 2013) were studied
experimentally. The experimental results were found to be in quantitative agreement
with theoretical predictions for the wall-mode instability, if the change in the channel
shape due to the applied pressure, and the resulting change in the velocity profile and
streamwise pressure variation, are incorporated in the analysis (Verma & Kumaran
2013, 2015).
The agreement between the predictions of linear stability analysis and the
experimental results is contrary to the situation for the flow through rigid conduits.
In a rigid channel, the linear stability analysis predicts a transition at a Reynolds
number of 5771, but the transition is experimentally observed at a Reynolds number
of 1200. In a rigid tube, the linear stability analysis indicates that laminar flow
is stable at all Reynolds numbers, but there is a transition at a Reynolds number
of 2100 in experiments. This is because the transition is highly subcritical and
three-dimensional, and is not captured by a linear stability analysis. In contrast, based
on the evidence available so far, it appears the transition in a soft-walled channel/tube
can be quantitatively predicted by a linear stability analysis.
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1078 S. S. Srinivas and V. Kumaran
1.2. Soft-wall turbulence
It has recently been shown that the flow after transition in a soft-walled microchannel
(Srinivas & Kumaran 2015) shares many of the characteristics of the turbulence
in a rigid channel. There is a discontinuous change in the velocity profile from a
parabolic profile before transition to a plug-like flow with a higher gradient near the
walls and a smaller curvature at the centre, and there is a near-wall maximum in the
streamwise root-mean-square (r.m.s.) velocity. There are also significant differences:
The fluctuating velocities near the soft wall are 23 times higher than those near
the hard wall, reflecting the role of the soft wall in generating turbulence, and the
Reynolds stress appears to be non-zero near the soft wall. There appears to be a
logarithmic layer in soft-wall turbulence, where the mean velocity has a logarithmic
dependence on the distance from the wall. The amplitudes of the fluctuating velocities
in soft-wall turbulence for Reynolds numbers less than 400, when scaled by suitable
powers of the mean velocity, are larger than those in the turbulent flow in a rigid
channel at much higher Reynolds numbers of up to 20 000.

1.3. Flow of viscoelastic fluids in soft-walled conduits


There have been fewer studies on the effect of viscoelasticity on the transition in soft-
walled conduits. Shankar & Kumar (2004) used the upper convected Maxwell model
to study the effect of viscoelasticity on the stability of a plane Couette flow in the
limit of zero Reynolds number. It was found that fluid elasticity had a stabilising effect
on the low-Reynolds-number instability of a Newtonian fluid, though there were other
modes present in the flow of viscoelastic fluids past rigid surfaces that are destabilised
when the surface is made soft. Chokshi & Kumaran (2007) analysed the Couette flow
of a viscoelastic fluid, described by the Oldroyd model, past a soft surface in the limit
of zero Reynolds number, and found that an increase in the fluid elasticity stabilised
the flow. More recently, Chokshi, Bhade & Kumaran (2015) studied the wall-mode
instability in the flow of an Oldroyd B fluid past a soft surface, and found that the
transition Reynolds number for the wall modes could decrease by a factor of 10 due
to fluid elasticity. The effect of viscoelasticity on the viscous instability has been
studied by Neelamegam, Shankar & Das (2013), who found a suppression of the
viscous instability. However, it appears the suppression may be due to the compression
of the soft surface by the normal stress difference, and the resulting increase in the
channel height, rather than the fluid elasticity. The effect of added polymers on the
high-Reynolds-number instability has not been studied experimentally so far, and it
provides one of the motivations for the present analysis.

1.4. Drag reduction


Addition of small amounts of polymer, as small as 10 ppm, has been successfully
used to reduce drag by up to 30 % in oil pipelines (Toms 1977). Though there has
been extensive research in this area, the mechanism still remains poorly understood.
A remarkable feature is the maximum drag reduction (MDR) asymptote, which is
the maximum drag reduction that could be achieved for a specified geometry and
Reynolds number independent of the concentration or composition of the polymer
(Virk 1975). The observed MDR asymptote in the Prandtlvon Krmn plots relating
friction factor and Reynolds number has been associated with a modification in values
of the constants in the von Krmn logarithmic law for the near-wall mean velocity
profiles. Direct numerical simulations (Sureshkumar, Beris & Handler 1997; Wang
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1079
et al. 2014), usually using variants of the FENE model for viscoelastic fluids, do show
evidence of drag reduction even at relatively low Reynolds numbers. The observed
mechanism involves attenuation of the rollup and bursting of the vortices responsible
for turbulence generation, and the extension of the time periods of low turbulence
intensities.
Drag reduction is not expected in soft-wall turbulence, because the Reynolds stress
is a relatively small fraction of the total stress even when the turbulent fluctuating
velocities are comparable to the mean velocity. The reason is that the strain rates
are very large (up to 104 s1 ) due to the small cross-stream dimensions of the order
of hundreds of micrometres. For this reason, the Reynolds stress forms only a small
fraction, less than 20 %, of the total stress. Damping of these fluctuations may result
in some drag reduction, but this is compensated by the increase in viscosity due to
the addition of polymers. So the issue of interest here is turbulence modification itself,
rather than drag reduction.

1.5. Objective
For the flow of water through a soft-walled microchannel, our objective is to study
(i) the effect of polymer addition on the transition Reynolds number, and
(ii) the turbulence modification by polymer addition.
As summarised in the previous discussion, there are linear stability results (Chokshi
et al. 2015) which indicate that the addition of small amounts of polymers could
result in a significant decrease in the transition Reynolds number. Our first objective
is to specifically examine whether such reductions are observed in experiments. Drag
reduction due to dissolved polymers in turbulent wall-bounded flows is a well-studied
subject, but the phenomenon is still not well understood. A recent study (Srinivas
& Kumaran 2015) has shown that the turbulent flow in a soft channel has many
characteristic features in common with, but also significant differences from, turbulent
flows bounded by hard walls. However, the effect of added polymers on the turbulence
in the flow through soft channels, a subject of considerable interest in physiological
applications, has not been studied so far. The second objective is to examine the effect
of polymer addition on the turbulence, and compare this with the effect of polymers
on turbulent flows bounded by hard walls.
The experimental configuration and the different experimental techniques used
are presented in the next section. The fabrication of the microchannels and the
measurement of the wall deformation and fluid velocity profiles are only briefly
summarised, since they are discussed in detail in 2 of Verma & Kumaran (2013)
and 2 of Srinivas & Kumaran (2015). The preparation and characterisation of
the polymer solutions, and the dimensionless numbers used for characterising the
experiments are discussed in further detail. The main results are presented in 3,
beginning with the transition Reynolds number for different polymer molecular
weights and mass fractions in 3.1. Results from the visualisation of dye mixing,
indicative of cross-stream mixing in the spanwise direction, are presented, and the
effect of polymer addition on the cross-stream mixing is discussed in 3.2. After
this, the profiles of the mean and fluctuating velocities in the channel are analysed
in 3.3. The fluctuations in the soft wall induced by the fluid velocity fluctuations
are discussed in 3.4. The important conclusions of the present study are presented
in 4. The present results are placed in the context of previous studies, relating to
both the transition Reynolds number and the turbulence intensities, in 5.
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1080 S. S. Srinivas and V. Kumaran
2. Experimental set-up
The experimental set-up is identical to that used in Srinivas & Kumaran (2015), and
so the description here is brief.

2.1. Microchannel fabrication


The microchannel is a rectangular bore of width 1.5 mm and height 160 m in
polydimethylsiloxane (PDMS) gel, fabricated by soft lithography using the procedure
described in Verma & Kumaran (2013) and Srinivas & Kumaran (2015). Three of
the four walls of the microchannel are fabricated as a PDMS stamp using gel
with catalyst concentration of 10 % as prescribed in the Sylgard 184 kit from Dow
Corning, resulting in a shear modulus of 0.55 MPa. The stamp pattern consists
of two identical inlets, which converge into one channel of dimensions shown in
figure 1(a). Inlet holes are punched into the PDMS stamp, and these are connected to
syringe pumps using pipette tips. The fourth wall is prepared as a film of thickness
2 mm of much softer gel with a catalyst concentration 1.75 %, and a shear modulus
of 18 kPa. The resulting cross-sectional configuration is shown in figure 1(c). The
microchannel is designed so that there is a developing section of 1 cm, where all
four walls are made of hard gel, upstream of the test section of length 3 cm where
one of the walls is made of soft gel, so that the flow is fully developed before it
enters the test section. A side port is also fabricated in the stamp just upstream of
the test section, so that the difference in pressure between the entrance of the test
section and the outlet (which is open to the atmosphere) can be measured.
When a pressure gradient is applied, there is a deformation of the bottom soft wall
of the microchannel, as shown in figure 1(d). There is very little deformation of the
three hard walls made of PDMS with elasticity modulus 0.55 MPa, but there is
significant deformation of the bottom soft wall made of PDMS with elasticity modulus
18 kPa.

2.2. Experimental configuration


The microchannel is mounted on an optical table with the flow (x) and spanwise (z)
directions horizontal, and the cross-stream (y) direction (with the smallest height of
160 m) vertical, as shown in figure 2. Two different configurations are used for
imaging in the experiments. The configuration shown in figure 2(a), where the xy
plane is imaged from the side, is used for measuring the wall deformation discussed
in 2.3, and for the particle image velocimetry (PIV) measurements described in 2.5.
The configuration shown in figure 2(b), where the xz plane is imaged from above, is
used for the dye-stream experiments, where clear water is pumped though one inlet
and water coloured with a dye is pumped through the other inlet, and the mixing of
the two streams is observed in the microchannel. The configuration in figure 2(b) is
also used for measuring the motion in the bottom wall, by marking the wall with a
dye spot and observing the motion of the spot.

2.3. Wall deformation


When a pressure difference is applied between the inlet and outlet, there is a
deformation of the soft wall of the channel, as shown in figure 1(d). There is
very little deformation in the development section, because all four walls are made
of hard gel, but there is significant deformation in the test section. The deformation
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1081

(a) Pressure port I II III z

x
1.5 mm
Inlets
Development section Test section

1 cm 1.2 cm 0.8 cm 0.9 cm


3 cm

(b) Syringe
pump 3 cm
1 cm
y
Hard gel
x
Hard gel Soft gel

(c) 1.5 mm (d) 1.5 mm

Hard y 2 mm Hard
A B
z F y
E h
D zC
Soft 2 mm
Soft

F IGURE 1. The schematic of the top view (a) of the PDMS stamp bonded onto the bottom
polymer film, the side view (b) of the deformed channel when there is an applied pressure
difference, the channel cross-section in the absence of deformation (c) and when there is
deformation due to applied pressure gradient (d). The maximum height h is a function of
the downstream position, as shown in (b). The locations I, II and III, where the particle
image velocimetry measurements are carried out, are shown in (a).

of the wall is measured as a function of the downstream distance using images taken
from the side in the configuration shown in figure 2(a), using a Photron FastCam
SA-Z camera with an attached Navitar 48X Zoom Tube for magnification. The outline
of the channel in the images provides the maximum height along the central plane
of the channel shown by the dashed line in figure 1(d). The shape of the bottom
wall is then reconstructed using a cubic spline fit at each downstream location. In
appendix B, the results for the change in height of the channel due to the applied
pressure gradient are presented.
The deformed shape of the channel is reconstructed, and ANSYS FLUENT
simulations in the deformed channel are used to determine the velocity profile
for the laminar flow of a Newtonian fluid with the same viscosity as the polymer
solution. These are then compared with the experimental velocity profiles, to examine
whether the flow is laminar. The procedure is as follows.
(i) Multiple images of the side of the channel are captured at predefined locations
using a zoom tube, and these are combined as shown in figure 3. From this side
view, the maximum height of the channel along the central plane in the spanwise
direction, the vertical line DF in figure 1(d), is measured. The maximum height is
shown as a function of downstream distance in appendix B in figures 23 and 24.
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1082 S. S. Srinivas and V. Kumaran

(a) Laser head


Syringe (c)
y 0.2 mm Masking tape
pump
Laser sheet
z
x

Glass substrate h
Goniometers
Translation stages
Motorised stage
Zoom tube 1.5 mm
Imaging area
Camera

(b) Camera

Zoom tube
Syringe
pump y

z
x

Glass substrate
Goniometers
Translation stages
Motorised stage

Light source

F IGURE 2. The configuration used (a) for the PIV measurements and for determining
the channel deformation, and (b) for the dye-stream imaging and the wall oscillations; (c)
shows the cross-sectional view of the masking tape used for confining the laser sheet in
the PIV measurements.

(ii) The shape of the deformed bottom surface in figure 1(d) is reconstructed using
a cubic spline procedure, from the maximum height DF along the central plane,
the undeformed heights AE and BC along the side walls, and the condition that
the variation of the height in the z direction is zero by symmetry at the point D.
An example of the side view and the cross-section of the reconstructed channel
is shown in figure 4.
(iii) The deformed cross-section of the channel at each downstream location is then
used to reconstruct the three-dimensional channel shape using Autodesk Inventor.
The mesh generation in the channel was carried out using Gambit 2.4.6; the
smallest spatial dimension in the cross-stream direction was 20 m; this level
of discretisation required 106 grid points. The procedure and validation has been
discussed earlier in 3.1 of (Verma & Kumaran 2013).
(iv) The flow simulations were carried out using ANSYS FLUENT 14.0 computational
fluid dynamics software, using the laminar flow model for a Newtonian fluid,
and the pressure velocity formulation for solving the NavierStokes equations.
The boundary conditions at all the bounding surfaces were the zero-velocity
boundary conditions. The velocity profile was specified to be a plug flow at the
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1083

(a)

x (cm) 0 0.2 0.8 1.5 2.1 2.7 3.0

(b)

x (cm) 0 0.2 0.8 1.5 2.1 2.7 3.0

F IGURE 3. The side view, imaged using the configuration in figure 2(a), of the
undeformed microchannel (a) and the deformed microchannel for the flow of
polyacrylamide solution with molecular weight 5 106 , mass fraction 50 ppm and
Reynolds number 406 (b). The streamwise locations where the height was measured
are shown by the vertical lines, and x = 0 corresponds to the entrance of the test (soft)
section.

(a)

10 mm
x (cm) 0 0.2 0.8 1.5 2.1 2.7 3.0

(b) (c) (d) (e)

F IGURE 4. Side view (a) and cross-sectional views at different downstream locations in
the test section (b) and at locations I (c), II (d) and III (e) (in figure 1a) of the channel
reconstructed from the images in figure 3(b). In (be), the width of the channel, which is
1.5 mm, serves as a reference length.

inlet to the development section, zero pressure at the outlet and zero velocity
as the initial condition. The validation for this procedure is provided in 3.5 of
Srinivas & Kumaran (2015), and further details are provided in 2.2 of Srinivas
& Kumaran (2015). When compared with a parabolic velocity profile for a
laminar flow, the difference between the simulation and analytical results was
reported to be 2 % in Srinivas & Kumaran (2015).

2.4. Dye-stream mixing


In the dye-stream experiments, clear water/polymer solution is pumped through one
of the inlets, while water/polymer solution with black coloured Bril ink is pumped
through the other inlet, both at equal flow rates. The mixing of the two fluids is
recorded using a Photron FastCam SA-Z camera from above in the configuration
shown in figure 2(b). In a laminar flow, the two streams flow parallel to each other
and there is no motion of the interface. This is because the diffusion time across a
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1084 S. S. Srinivas and V. Kumaran
width W is much longer than the residence time in the channel. The diffusion time,
which scales as W 2 /D, is of the order of 2250 s across a width of 1.5 mm, even
if we consider a relatively high diffusion coefficient of D = 109 m2 s1 for small
molecules in water, while the residence time of the fluids in the microchannel is
much less than 1 s. Therefore, there is virtually no mixing between the fluid streams
in the microchannel in a laminar flow. Motion of the dye across the channel provides
evidence of mixing between the two streams.
The mixing is quantified by the following procedure used in Verma & Kumaran
(2013), following similar procedures used earlier in Lee, Choi & Park (2010). The
grey scale images of the channel taken from above are processed to determine the grey
scale intensity, Gij (tk ), at each pixel (i, j) at the time instant tk within a rectangular
region. The average intensity and the variance of the intensity at each time instant
are
1 X
Gav = Gij , (2.1)
N i,j
1 X
h(1G)2 i = (Gij Gav )2 , (2.2)
N i,j

where N is the total number of pixels in each frame. The segregation index is obtained
by taking the ratio of the r.m.s. of the grey scale intensity fluctuations and dividing it
by the maximum difference in the grey scale intensities in the dye and the clear fluid,
p
h(1G)2 i
SI = , (2.3)
Gw Gd
where Gw and Gd are the grey scale intensities of clear water stream and the dye
stream. The segregation index is close to zero if the two streams are fully mixed, and
has a maximum of 0.5 when the two streams are perfectly segregated with no lateral
mixing. The segregation index is used to quantify the effect of polymer addition on
the cross-stream mixing after transition.
For the dye concentrations we have used here, the grey scale intensity is not a linear
function of the concentration. When the dye that was initially restricted to one half of
the channel mixes throughout the channel, the dye concentration decreases by a factor
of 2, but there is virtually no change in the colour. Thus, the colour cannot be used
as a quantitative measure of the concentration. It can only be used to infer the spatial
extent to which the dye has mixed into the clear fluid, and to distinguish the extent
of mixing for different polymer concentrations. Quantitative measures of mixing have
been analysed in an earlier study of Kumaran & Bandaru (2016).
The segregation index is calculated at locations I, II and III in figure 1(a), using
images of the type shown in figures 57 presented in 3.2.

2.5. Fluid velocity fields


The fluid velocity fields in the xy plane (the central plane of the channel in the
spanwise (z) direction) are measured using PIV with the configuration in figure 2(a).
An IDT PIV system, which consists of a SharpVisionTM 1500-EX charge-coupled
device (CCD) camera with spatial resolution of 1360 1036 pixels, Nd:YAG laser
(Solo-III, New-Wave Research), laser sheet generating optics and a frame rate of
15 Hz, is used for the PIV measurements. The seed particles used for the PIV
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1085
measurements were prepared using a customised procedure, dispersion polymerisation
of styrene in a solution of ethanol (Srinivas & Kumaran 2015), adapted from
Lenzmann et al. (1994). This procedure produces spherical polymer particles of
size 2.12 0.41 m with circularity between 0.95 and 1.0. (4 area/perimeter2
of the projection of the shape onto a two-dimensional plane.) The size of the particles
was carefully selected to ensure that they follow the fluid streamlines (Srinivas &
Kumaran 2015). For particles with size below 1 m, the Pclet number is small and
Brownian diffusion is important, so the particles cross streamlines due to Brownian
diffusion. Particles with size greater than 10 m have high Stokes number, so
they cross streamlines due to inertial effects. For the particle size selected here, both
Brownian diffusion and inertial effects are small, and so the particles follow the fluid
streamlines and the fluid velocity is accurately captured.
For the PIV measurements, a laser sheet is directed downwards along the central
plane of the channel using the configuration as shown in figure 2(a), and images
are taken using a camera from the side. The laser sheet is confined to a width of
200 m along the central plane of the channel in the z direction using two masking
tapes placed on the channel, as shown in figure 2(c), in order to avoid averaging of
the velocity field in the spanwise direction.
The images are captured with a frame rate of 15 image pairs per second. The pulsed
laser illumination for each pair of images is separated by a time interval of 10100 ns
depending on the flow speed. The images obtained in the xy plane are divided into 60
equal intervals in the cross-stream (y) direction. The fluid velocities, calculated from
the peak of the autocorrelation function between successive images in each image pair,
are assigned to the midpoint of each interval. The mean and r.m.s. velocities are then
calculated by averaging over multiple frames. For each set of parameter values, 4000
pairs of images are captured in two sequences of 2000 pairs each. The mean velocity
is the average over all 4000 image pairs. In order to assess the variability in the data,
the 4000 image pairs are divided into eight subsequences of 500 image pairs each,
and the average over each of these subsequences is calculated. The standard deviation
is determined from the average of these eight subsequences, and the error bars in the
figures show one standard deviation above and below the mean value. The validation
of the PIV measurements with the ANSYS FLUENT simulations for the flow of
pure water in a deformed channel are shown in figure 7 of Srinivas & Kumaran
(2015). The validation of the turbulence measurements is shown in the appendix of
Srinivas & Kumaran (2015) at a Reynolds number of 3500 for the same channel
configuration.

2.6. Wall motion


The configuration in figure 2(b) has been used to monitor wall motion. The bottom
soft wall of the channel is marked with a dye spot at three different downstream
locations along the centreline of the channel in the spanwise direction. The dye spot
is irregular and has a linear dimension of between 0.025 and 0.05 mm. The images
of the dye spot are captured from above using the Photron FastCam SA-Z camera
with a frame rate of 1000 frame s1 , and the displacement of the centroid of the dye
spot in the streamwise (x) and spanwise (z) directions between successive frames is
determined. One pixel in the image corresponds to a linear displacement of 0.35 m,
and the resolution in the displacement measurements is 0.5 m.
When there is fluid flow in the channel, there is a mean wall displacement in the
spanwise direction, u x , due to the shear stress exerted by the fluid on the wall. This is
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1086 S. S. Srinivas and V. Kumaran

Molecular weight 4 104 Molecular weight 5 106


Mass frac. (ppm) Viscosity (mPa s) Mass frac. (ppm) Viscosity (mPa s)
100 1.002 0.5 1.003
300 1.003 1.0 1.006
500 1.005 2.0 1.009
700 1.008 10.0 1.040
1000 1.030 30.0 1.085
1500 1.050 50.0 1.095
TABLE 1. The mass fractions and the viscosities of the polymer solutions used in the
experiments. Addition of polymer causes only small changes in the solution viscosity, but
it has a large effect on transition and turbulence, as shown in 3.

measured from the displacement of the dye spot in the streamwise direction observed
using a camera above the channel as shown in figure 2(b). There are also displacement
fluctuations in the streamwise and spanwise directions due to the fluid turbulence,
and the r.m.s. values of these fluctuations are designated u0x and u0z , respectively. The
variation of the displacement with time is captured, and the Fourier transform of
this time series is used to determine the spectra of the displacement fields, u x ()
and u z (),
1 T
Z
u x () = dt exp(it)ux (t). (2.4)
T 0
Since the maximum frame rate is 1000 frame s1 , the Nyquist frequency is
500 frame s1 or 3000 rad s1 . The time period T in (2.4) was set equal to
1 s in most experiments, but longer runs up to 30 s have also been used to ensure
that there are no systematic low-frequency oscillations.

2.7. Polymer solutions


Solutions of polyacrylamide, with molecular weights 4 104 and 5 106 , procured
from Sigma-Aldrich, are used for the experiments. Since the monomer molecular
weight is 71, each polymer molecule comprises 560 repeat units for the polymer
with molecular weight 4 104 , and 70 500 repeat units for the polymer with
molecular weight 5 106 . The solutions are always freshly prepared by dissolving
polyacrylamide powder in de-ionised water, since it is known that the rheological
properties of polyacrylamide solution could change if it is stored for hours or days
(Kulicke, Kniewske & Klein 1982). The mass fractions and viscosities of the polymer
solutions are provided in table 1.
The viscosity measurement was carried out in a Brookfield capillary viscometer,
since it is too small to examine in a traditional rheometer. The accuracy depends
on the sensitivity of the pressure transducer used to calculate the pressure difference.
Even here, the pressure difference is calibrated to that of pure water, and then the
change in the pressure difference due to the added polymer is determined. With this,
we were able to calculate to within 0.002 of the viscosity of water. It should be
noted that the capillary viscometer is also not ideal for measuring the viscosity of
polymers, since there is likely to be polymer extension and contraction at the inlet
and outlet. For added verification, we have compared with the viscosity obtained from
the specific viscosity reported by Kulicke et al. (1982), which results in a solution
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1087
viscosity of 1.075 mPa s for the polymer with molecular weight 5 106 at the
highest concentration of 50 ppm. Also, Samanta et al. (2013) report a zero-shear
viscosity of 1.1 mPa s for a concentration of 50 ppm for the same molecular weight.
Therefore, our viscosity of 1.095 mPa s from the Brookfield viscometer is consistent
with these.
The relaxation time of the polymers in solution could not be measured directly,
since there is very little modification of the water viscosity. The relaxation time has
been estimated by different methods, and the values differ significantly depending
on the method. In shear-thinning fluids, the CarreauYasuda fit for the viscosity
versus strain rate is used to estimate the relaxation time. Small-amplitude oscillatory
shear measurements can be fitted to a Maxwell model, in order to obtain the longest
relaxation time. It is also possible to use the first normal stress difference to estimate
the relaxation time (Pan et al. 2013) or to use the capillary bridge extensional
rheometer, as was done by Samanta et al. (2013).
For polyacrylamide with molecular weight 5 106 , Samanta et al. (2013) report the
relaxation time in the range of 110 ms by the CaBER method. For the multi-mode
Maxwell model, Neelamegam et al. (2013) report relaxation times of 210 s, while
Poole et al. (2007) report relaxation times up to 20 s, so there is a range of 105
in the relaxation times reported by different authors. Zell et al. (2010) have analysed
the relaxation times calculated by different methods for, among other polymers,
polyacrylamide with molecular weight 5 106 . They have found that the relaxation
time from the Maxwell model is actually the smallest, in the range 110 ms.
The relaxation time from the viscosity measurements is the largest, in the range
1100 s, while that from the normal stress differences and capillary extension is in
between, in the range 0.11 s.
In order to avoid ambiguity, we have defined the relaxation time based on the Zimm
model for a polymer in a good solvent (Doi & Edwards 1988),
s R3g
= , (2.5)
kB T
where s is the solvent viscosity, Rg is the radius of gyration and kB T is the thermal
energy. Defined this way, there is no uncertainty in the definition of the relaxation
time, especially since we are in the dilute single-chain limit. Kulicke et al. (1982)
report that the end-to-end distance of polyacrylamide molecules of molecular weight
5 106 is 0.5 m. This can also be calculated if we consider the Flory characteristic
ratio (b/a)2 of 8.5 (Bohdanecky, Petrus & Sedldtek 1983), where a is the monomer
size and b is the Kuhn segment length. This implies that there are approximately
three monomers in each Kuhn segment, and that each Kuhn segment is 10 . The
relaxation time from (2.5) is then 0.03 s for the polymer with molecular weight
5 106 . For the smaller molecular weight polymer, the ratio of relaxation times scales
as M 1.8 , if we consider the radius of gyration to scale as M 0.6 . Based on this, the
relaxation time is 5 106 s for polyacrylamide with molecular weight 4 104 .

2.8. Dimensionless numbers


(1) The Reynolds number is defined as
Q
Re = , (2.6)
W
where Q is the flow rate, W is the channel width, and and are the fluid
density and viscosity. For a rectangular microchannel, this definition reduces
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1088 S. S. Srinivas and V. Kumaran
to vh/,
where v is the average velocity (flow rate divided by area of
cross-section) and h is the height. However, the advantage of the definition
of Re in (2.6) is that it is independent of height, and so the Reynolds number
does not change when there is channel deformation.
(2) The Weissenberg number is defined as
v
We = , (2.7)
h
where is the polymer relaxation time. The ratio of the Weissenberg number and
the Reynolds number, denoted as the elasticity number El, is independent of the
flow velocity,

El = 2 . (2.8)
h
The relaxation time is 0.03 s for polyacrylamide with molecular weight 5 106 ,
and approximately 5 106 s for polyacrylamide with molecular weight 4 104 .
The channel height varies approximately between 160 m when there is no flow,
to a maximum of 500 m at the largest flow rate used here. Based on this,
for channels of height 160500 m, the elasticity number varies between 1.2 and
0.12 for the polyacrylamide with molecular weight 5 106 , and between 2 105
and 2 104 for the polyacrylamide with molecular weight 4 104 . However,
it should be cautioned that the elasticity number could change by a factor of
10100 if other measures of the relaxation time based on the rheology are used.
There is deformation of the soft wall in the experiments, and the Weissenberg
and elasticity numbers vary with channel height. Therefore, the results are
parametrised by the polyacrylamide molecular weight and mass fraction instead
of the Weissenberg and elasticity numbers.
(3) The fluidwall interaction is characterised by the dimensionless parameter =
Gh2 /2 , where G is the shear modulus of the wall material. The soft wall has
a shear modulus of 18 kPa for all the experiments, and so the parameter is
4.6 105 to 4.6 106 for water and for h in the range 160500 m. The value of
decreases by 20 % for the most viscous polymer solution used here, because
the viscosity increases by 10 %.

3. Results
To place the results in context, the transition Reynolds numbers for polymer
solutions with different mass fractions at the three locations I, II and III (figure 1a),
are first provided in table 2. After this, the different indicators of transition, such
as dye-stream mixing, wall motion and fluid fluctuating velocities, are discussed in
detail.

3.1. Transition Reynolds number


Table 2 shows the highest Reynolds number at which the flow is observed to be
laminar, Rel , the lowest Reynolds number at which the laminar flow is observed
to be disrupted, Reh , and the average of the two Reynolds numbers, Ret , which is
designated the transition Reynolds number for the present purposes. At polymer mass
fractions below a threshold wt , which is 500 ppm and 1 ppm for the solutions of
polyacrylamide with molecular weight 4 104 and 5 106 , respectively, the transition
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1089

Polyacrylamide solution MW 5 106 Polyacrylamide solution MW 4 104


ppm Location I and II Location III ppm Location I and II Location III
Rel Reh Ret Rel Reh Ret Rel Reh Ret Rel Reh Ret
0 277.3 305.0 291.1 249.5 277.2 263.4 0 277.3 305.0 291.1 249.5 277.2 263.4
0.5 277.0 304.6 290.8 249.3 277.0 263.1 100 277.2 305.0 291.1 249.5 277.2 263.4
1 276.1 303.7 289.9 248.5 276.1 262.3 300 277.0 304.6 290.8 249.3 277.0 263.1
2 176.2 209.2 192.7 176.2 209.2 192.7 500 276.4 304.0 290.2 248.8 276.4 262.6
10 128.2 149.6 138.9 128.2 149.6 138.9 700 198.4 220.5 209.4 198.4 220.5 209.4
30 61.44 81.93 71.69 61.44 81.93 71.69 1000 151.0 172.6 161.8 151.0 172.6 161.8
50 30.44 40.59 35.52 30.44 40.59 35.52 1500 105.8 127.0 116.4 105.8 127.0 116.4
TABLE 2. The highest Reynolds number at which the flow is laminar, Rel , the lowest
Reynolds number at which there is flow disruption, Reh , and the arithmetic average of
these two, which is considered the transition Reynolds number, Ret , as a function of the
polymer mass fraction.

Reynolds number is almost the same as that of pure water. The small variation in
the transition Reynolds number is due to the slight increase in the viscosity due to
polymer addition, as shown in table 1. Above the threshold mass fraction, there is a
sharp decrease in the transition Reynolds number, and the transition is observed at a
Reynolds number as low as 116 for the solution of polyacrylamide with molecular
weight 4 104 and mass fraction 1500 ppm, and as low as 35 for the solution of
polyacrylamide with molecular weight 5 106 and mass fraction 50 ppm.
For pure water, transition is observed when the Reynolds number is increased
from 277 to 305 at the downstream locations I and II, and from 250 to 277 at
the downstream location III. This slight variation is consistent with the earlier
experimental results of Verma & Kumaran (2013) and Srinivas & Kumaran (2015)
for the flow of water in a microchannel, where it was observed that the disruption
of the laminar flow first takes place at the downstream locations when the Reynolds
number is increased, and then progresses upstream.
It is important to note that there is a significant increase in the height of the
microchannel due to the applied pressure gradient. As shown in figure 1(d), there
is no discernible deformation of the three hard walls of the microchannel, but there
is a significant compression of the soft wall. The height of the microchannel first
increases downstream from the start of the developing section (joint between the hard
and soft sections in figure 1b), and then decreases further downstream as the pressure
decreases. The deformation is discussed in detail in appendix B.

3.2. Dye-stream mixing


In order to assess the cross-stream mixing, one of the inlet streams (figure 1a) is
mixed with a small amount of black ink, while the other is transparent. The resulting
images are shown in figures 57 for pure water, for polymer solution of molecular
weight 5 106 with mass fraction 2 ppm (just above wt ) and for the maximum mass
fraction 50 ppm. In all cases, images are shown, from left to right, at the entrance
to the test section (to verify that there is no disturbance at the inlet) and the three
locations I, II and III in figure 1(a). The panels from top to bottom show the images
at a Reynolds number below transition, one above transition and a Reynolds number
significantly exceeding the transition Reynolds number.
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1090 S. S. Srinivas and V. Kumaran

(a)

1.5 mm

(b)

(c)

F IGURE 5. Images of, left to right, the entrance to the test section, location I, location
II and location III, in figure 1(a), for the flow of water at Reynolds numbers 250 (a),
305 (b) and 444 (c). The water pumped into the bottom inlet is mixed with black ink for
visualisation. The development section is to the left of the arrow, and the test section to
the right of the arrow.

The images in figure 5 show the results for the flow of pure water. No cross-stream
mixing is observed below transition. There is intense cross-stream mixing at the
downstream locations I, II and III at a Reynolds number of 305, which is higher than
the transition Reynolds number, and the extent of mixing increases with downstream
distance. The images for polymer solutions with mass fraction just above wt are shown
in figure 6. In this case, there is no mixing below the transition Reynolds number, but
cross-stream mixing is observed when the transition Reynolds number is exceeded.
However, the mixing is clearly imperfect, and the mixing intensities are significantly
lower than those for pure water. Complete cross-stream mixing is not observed even
at the highest experimental Reynolds numbers of 400. In figure 7, the polymer
mass fraction is increased to 50 ppm for the polyacrylamide with molecular weight
5 106 . In this case, there is very little cross-stream mixing. However, fluctuations at
the interface are clearly observed when the transition Reynolds number is exceeded,
and in video recordings, motion of the interface between the transparent and coloured
fluid is clearly visible, indicating that there is an instability of the laminar flow.
The segregation index SI , defined in (2.3), is shown as a function of Reynolds
number in figure 8 at two locations, I and III. For pure water, the segregation index
is close to 0.5 when the Reynolds number is below Ret , and it decreases steeply when
the Reynolds number crosses Ret . The segregation index is lower at the downstream
location III in comparison to the upstream location I, indicating greater mixing at the
downstream location, as observed in figure 5. The results for mass fractions below
wt are qualitatively similar to that for water; these are not shown in figure 8 in
order to enhance clarity. When the mass fraction increases beyond wt , two effects are
clearly visible. Firstly, there is a decrease in the transition Reynolds number at which
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1091

(a)

(b)

(c)

F IGURE 6. Images of, left to right, the entrance to the test section, location I, location II
and location III in figure 1(a), for the flow of a solution of polyacrylamide of molecular
weight 5 106 and mass fraction 2 ppm at Reynolds numbers 193 (a), 231 (b) and 441
(c). The solution pumped into the bottom inlet is mixed with black ink for visualisation.
The development section is to the left of the arrow, and the test section to the right of
the arrow.

SI decreases sharply. Secondly, the segregation index is higher than that for pure
water, and it increases with polymer mass fraction. At the highest mass fraction of
50 ppm for the polyacrylamide with molecular weight 5 106 , the segregation index
decreases to 0.35 at location I and 0.3 at location III, indicating that the mixing is
very imperfect; this is evident in the images in figure 7, where there is poor mixing
even at the downstream location III. There is better mixing for the polyacrylamide
with molecular weight 4 104 at the highest mass fraction of 1500 ppm, but, even
here, the segregation index decreases to 0.25 at location I and 0.2 at location III.
This is much higher than the minimum value of 0.05 at the location III for pure
water.

3.3. Flow dynamics


The results for the mean and fluctuating velocities are shown in figures 912. Results
are shown for four Reynolds numbers, two below transition, the third just above the
transition Reynolds number and the fourth much higher than the transition Reynolds
number. Only the results at location II are shown in all the figures; the results
at other locations are qualitatively similar. In all these figures, the bottom wall is
located at y = 0, while the location of the top wall is indicated by the vertical dashed
lines. The change in position of the top wall is due to the increase in height as
the Reynolds number increases. It is important to reiterate that we have not been
able to make velocity measurements within a distance of 15 m from the bottom
wall, due to reflection of the PIV laser from the bottom wall, and so we do not
have a direct measurement of the velocities at the bottom wall. Therefore, results
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1092 S. S. Srinivas and V. Kumaran

(a)

(b)

(c)

F IGURE 7. Images of, left to right, the entrance to the test section, location I, location II
and location III in figure 1(a), for the flow of a solution of polyacrylamide of molecular
weight 5 106 and mass fraction 50 ppm at Reynolds numbers 20 (a), 41 (b) and 101
(c). The solution pumped into the bottom inlet is mixed with black ink for visualisation.
The development section is to the left of the arrow, and the test section to the right of
the arrow.

for the fluctuating velocity within a distance of 15 m from the wall are not
shown, and the results for the mean velocity at the wall are extrapolated to zero
velocity. In the figures 912, the absolute mean and fluctuating velocities before
and after transition are discussed; this is not a measure of the relative magnitudes
of the fluctuations to the mean velocity. In the following figures 14 and 15, the
dimensionless measures of the mean and fluctuating velocities are shown and
discussed.
The mean velocity profiles at different Reynolds numbers are compared in figure 9
for the flow of pure water, and polymer solutions of molecular weight 5 106 with
mass fraction 2 ppm just above wt and the maximum mass fraction of 50 ppm. In this
figure, the velocity profiles for a laminar flow obtained using the ANSYS FLUENT
computational fluid dynamics software for the same channel shape and flow velocity
are shown by the dashed lines. The velocity profiles are in close agreement with the
computed velocity profiles for a laminar flow below the transition Reynolds number
in table 2, but there is a clear difference between the experimental velocity profile
and the laminar profile when the transition Reynolds number is exceeded. For flow
of pure water in figure 9(a), it is observed that the experimental velocity profile is in
quantitative agreement with that from the ANSYS FLUENT simulations for a laminar
flow up to a Reynolds number of 277. When the Reynolds number is increased to
305, the velocity profile is distinctly different from the laminar profiles obtained in the
simulations, with a smaller curvature at the centre and sharper gradients at the walls.
It is shown a little later that the quantitative measure of the departure from a laminar
profile, vdiff , defined in (3.1), exhibits a discontinuous increase at this transition point.
For the flow of a solution of polyacrylamide of molecular weight 5 106 and mass
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1093

(a) 0.5 (b) 0.5

0.4 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0 100 200 300 400 500 0 100 200 300 400 500
Re Re

F IGURE 8. The segregation index SI (2.3) as a function of the Reynolds number Re at


locations I (solid line) and III (dashed line) for (a) a solution of polyacrylamide with
molecular weight 4 104 and mass fraction 1500 ppm (E), 1000 ppm (A) and 700 ppm
(C), and (b) a solution of polyacrylamide with molecular weight 5 106 and mass fraction
50 ppm (u), 30 ppm (E), 10 ppm (A) and 2 ppm (C). The results for pure water are
shown by the 6 symbols in both panels. The vertical solid and dashed lines show the
transition Reynolds numbers at location I and III, respectively, for different polymer
concentrations.

fraction 2 ppm, figure 9(b) shows that the velocity profile is in close agreement with
the laminar profile for a Newtonian fluid for Reynolds number up to 178, but there is
a distinct departure from the laminar profile at a Reynolds number of 211 and higher.
It is also evident from figure 9(b) that the extent of departure from the laminar profile
is lower for the polymer solution than for pure water in figure 9(a). When the polymer
concentration is further increased to 50 ppm, the velocity profiles appear to be close
to the laminar profiles at all Reynolds numbers. However, a closer examination of
figure 9(c) suggests that there is a small but discernible departure of the profile from
the laminar profile when the Reynolds number is increased from 30 to 40. This
distinction will also be evident in the discussion of the measure vdiff in figures 14
and 15. Based on figure 9, it can be concluded that there is a significant decrease
in the Reynolds number for transition when the polymer mass fraction is 2 ppm or
higher, and there is also a significant decrease in the departure of the velocity profile
from the laminar profile.
The profiles of the r.m.s. of the velocity fluctuations, vx0 and vy0 , are shown in
figures 10 and 11. It is important to reiterate that, in the experiments, there is a
non-zero level of fluctuations even in the laminar flow, and this level of fluctuations
also decreases significantly as the polymer mass fraction is increased. For pure water,
the streamwise r.m.s. velocity vx0 is 15 % of the average velocity, but this decreases
to 5 % of the average velocity for the highest polymer mass fractions used here.
Similarly, the r.m.s. velocity in the cross-stream direction vy0 is 3 % of the average
velocity for pure water, but it decreases to 1 % of the average velocity for the
highest polymer mass fractions used here. Thus, the addition of polymer appears to
have a significant damping effect on the fluctuations in the laminar flow as well.
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1094 S. S. Srinivas and V. Kumaran

(a) 2.0 (b)


2.0

1.5
1.5

1.0
1.0

0.5 0.5

0 0.1 0.2 0.3 0 0.1 0.2 0.3


y (mm) y (mm)

(c) 1.0

0.5

0 0.1 0.2 0.3


y (mm)

F IGURE 9. The mean velocity vx as a function of the cross-stream distance y for


pure water (a), and for a solution of polyacrylamide with molecular weight 5 106
and mass fraction 2 ppm (b) and 50 ppm (c) at location II (in figure 1a). In (a),
the Reynolds numbers are Re = 133 (E),277 (A), 305 (C) and 388 (6). In (b), Re =
133 (E), 178 (A), 211 (C) and 333 (6). In (c), Re = 20.29 (E), 30.44 (A), 40.59 (C) and
121.8 (6). The symbols and solid lines are the experimental profiles and the dashed lines
are the results of ANSYS FLUENT simulations for the laminar flow of a Newtonian fluid
in a channel of the same shape at the same flow rate. The bottom wall is at y = 0, and the
dashed lines on the right show the location of the top wall at different Reynolds numbers.

Corresponding to the departure of the mean velocity profile from the laminar profile
in figure 9, there is also a distinct increase in the amplitude of the velocity fluctuations
in figures 10 and 11. The profiles are also asymmetric, and the amplitude of the
fluctuations near the bottom (soft) wall is higher by a factor of 2 in comparison to
that near the top (hard) wall. This indicates that the soft wall has a significant role to
play in the generation of turbulent fluctuations. Most importantly, there is a significant
increase in the amplitude of vx0 and vy0 , by more than a factor of 2, when the Reynolds
number exceeds the transition Reynolds number shown in table 2 at each polymer
mass fraction. This coincides with the Reynolds number at which the velocity profile
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1095

(a) (b)
0.4

0.10
0.3

0.2
0.05

0.1

0 0.1 0.2 0.3 0 0.1 0.2 0.3


y (mm) y (mm)

(c)
0.04

0.03

0.02

0.01

0
0 0.1 0.2 0.3
y (mm)

F IGURE 10. The r.m.s. of the fluctuating velocity vx0 as a function of the cross-stream
distance y for pure water (a), and for a solution of polyacrylamide with molecular weight
5 106 and mass fraction 2 ppm (b) and 50 ppm (c) at location II (in figure 1a). In
(a), the Reynolds numbers are Re = 133 (E), 277 (A), 305 (C) and 388 (6). In (b), Re =
133 (E), 178 (A), 211 (C) and 333 (6). In (c), Re = 20.29 (E), 30.44 (A), 40.59 (C) and
121.8 (6). The bottom wall is at y = 0, and the dashed lines on the right show the location
of the top wall at different Reynolds numbers.

shows a distinct departure from the laminar profile in figure 9, indicating that the
transition is accompanied by a distinct shift in the form of the velocity profile as well
as a significant increase in the amplitude of the fluctuations in both the streamwise
and the cross-stream directions.
There is a decrease in the amplitude of the fluctuations in both the streamwise and
cross-stream directions as the polymer concentration is decreased. The maximum of
vx0 decreases by a factor of 3 in vx0 when the polymer concentration is increased from
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1096 S. S. Srinivas and V. Kumaran

(a) (b)

0.05
0.10

0.05

0 0.1 0.2 0.3 0 0.1 0.2 0.3


y (mm) y (mm)

(c)
0.010

0.005

0 0.1 0.2 0.3


y (mm)

F IGURE 11. The r.m.s. of the fluctuating velocity vy0 as a function of the cross-stream
distance y for pure water (a), and for a solution of polyacrylamide with molecular weight
5 106 and mass fraction 2 ppm (b) and 50 ppm (c) at location II (in figure 1a). In
(a), the Reynolds numbers are Re = 133 (E), 277 (A), 305 (C) and 388 (6). In (b), Re =
133 (E), 178 (A), 211 (C) and 333 (6). In (c), Re = 20.29 (E), 30.44 (A), 40.59 (C) and
121.8 (6). The bottom wall is at y = 0, and the dashed lines on the right show the location
of the top wall at different Reynolds numbers.

0 ppm to 2 ppm and another factor of 3 when the polymer concentration is decreased
to 50 ppm.
The profiles of the correlation hvx0 vy0 i at different polymer concentrations are shown
in figure 12. This correlation is related to the Reynolds stress xyR by the relation
xyR = hvx0 vy0 i, where the Reynolds stress is the rate of transport of momentum due
to the turbulent velocity fluctuations. For pure water, when the Reynolds number is
below 277, there is a background level of Reynolds stress fluctuations even in the
laminar flow. However, at the Reynolds number of 305, there is a sharp increase
in hvx0 vy0 i by a factor of 3. When the polymer concentration is increased, there is a
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1097

(a) (b)
5 3

0 1

5 1

10 3

15 5
0 0.1 0.2 0.3 0 0.1 0.2 0.3
y (mm) y (mm)

(c)
2

3
0 0.1 0.2 0.3
y (mm)

F IGURE 12. The correlation hvx0 vy0 i as a function of the cross-stream distance y for
pure water (a), and for a solution of polyacrylamide with molecular weight 5 106
and mass fraction 2 ppm (b) and 50 ppm (c) at location II (in figure 1a). In (a), the
Reynolds numbers are Re = 133 (E), 277 (A), 305 (C) and 388 (6). In (b), Re =
133 (E), 178 (A), 211 (C) and 333 (6). In (c), Re = 20.29 (E), 30.44 (A), 40.59 (C)
and 121.8 (6). The bottom wall is at y = 0, and the dashed lines on the right show the
location of the top wall at different Reynolds numbers.

significant decrease in hvx0 vy0 i, as shown in figure 12, by a factor of 5 when the
polymer mass fraction is increased from 0 ppm to 2 ppm, and a factor of 10 when
the polymer mass fraction is increased from 2 ppm to 50 ppm. For each polymer
concentration, there is a sharp increase in hvx0 vy0 i, by a factor of 3 or more, when the
Reynolds number exceeds the transition Reynolds number listed in table 2. Thus, the
indicators of a transition are consistently observed in the mean velocity as well as the
r.m.s. fluctuating velocity at the same transition Reynolds number where cross-stream
mixing is observed in the dye-stream experiments.
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1098 S. S. Srinivas and V. Kumaran
The value of hvx0 vy0 i appears to extrapolate to zero at the top (hard) wall, as
required by the condition that the velocity fluctuations should decrease to zero at
the wall. However, at the bottom wall, it is difficult to extrapolate these to zero,
indicating that the fluctuations at the soft surface may be non-zero. A similar feature
was observed for pure water in Srinivas & Kumaran (2015). As noted in the first
paragraph of this section, we do not have direct measurements of the mean and
fluctuating velocities within a distance of 15 m from the walls. The fluctuating
velocities vx0 and vy0 in figures 10 and 11 could plausibly be extrapolated to zero at
the bottom wall. However, it is very difficult to extrapolate hvx0 vy0 i to zero at the wall.
More sophisticated measurements are required to determine the exact nature of the
fluctuations at the wall and in the near-wall region.
Figure 13 shows the viscous stress (dashed lines) and total stress (solid lines) as
a function of the cross-stream distance. For a proper interpretation of this figure,
it is important to clearly specify the procedure used for calculating these stresses.
The viscous stress is defined as xyv = (dvx /dy), where is considered a constant
equal to the viscosity of the polymer solution. Here, the velocity gradient (dvx /dy)
is measured by numerically taking the derivative of the mean velocity profile using
a second-order central difference scheme. The total stress is the sum of the viscous
and the Reynolds stresses, xy = (dvx /dy) hvx0 vy0 i. The error bars in the stress
calculations are primarily due to the error bars in the correlation hvx0 vy0 i, since the
errors in the derivative of the mean velocity are much smaller. For pure water
(figure 13a), there is very little difference between the total and viscous stresses
when the Reynolds number is below the transition Reynolds number. However,
there is a significant difference above the transition Reynolds number, due to the
contribution of the Reynolds stress. This Reynolds stress contribution decreases to
zero at the upper hard wall, but is significant at the lower soft wall. The stress
is close to a linear function of the cross-stream distance below the transition
Reynolds number, as expected for a rigid channel when the pressure gradient is
a constant. The stress does depart from a linear function at higher Reynolds numbers
due to the channel deformation. However, it should be noted that the viscous
stress departs significantly from a linear function of y. It is necessary to add the
Reynolds stress, which is non-zero at the soft wall, to the viscous stress in order
to obtain a total stress profile that is closer to a linear function of cross-stream
distance.
At the highest polymer mass fraction considered in figure 13(c), the stress profiles
are linear functions of cross-stream distance. This is because the transition Reynolds
number is low, and the channel deformation is small at the Reynolds numbers
under consideration. Moreover, the linear variation of the stress with cross-stream
distance indicates that the fluid viscosity can be considered a constant across the
channel. At the polymer mass fraction just above wt in figure 13(b), the difference
between the total stress and viscous stress is comparable to the experimental
error bars when the Reynolds number is less than Ret , but there is a difference
between the total and viscous stresses when the Reynolds number is greater than
Ret . Thus, there is a discernible contribution to the total stress due to velocity
fluctuations.
An important point to note is that the stress profiles are not linear functions of
cross-stream distance in figure 13(b). Moreover, the curvature of the stress profiles is
the opposite of that expected due to channel deformation, shown in figure 13(a) for
pure water. This is possibly related to the non-uniform viscosity of the fluid across the
channel, a phenomenon also observed in the flow through a rigid channel considered
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1099

(a) (b) 25
40
20
30 15
10
20
5
10
0
0 5
10
10
15
20
20
30 25
0 0.1 0.2 0.3 0 0.1 0.2 0.3
y (mm) y (mm)

(c)
10

10

0 0.1 0.2 0.3


y (mm)

F IGURE 13. The total stress xy (solid lines) and the viscous stress xy = (dvx /dy)
(dashed lines) as a function of the cross-stream distance y for pure water (a), and for
a solution of polyacrylamide with molecular weight 5 106 and mass fraction 2 ppm
(b) and 50 ppm (c) at location II (in figure 1a). In (a), the Reynolds numbers are Re =
133 (E), 277 (A), 305 (C) and 388 (6). In (b), Re = 133 (E), 178 (A), 211 (C) and 333 (6).
In (c), Re = 20.29 (E), 30.44 (A), 40.59 (C) and 121.8 (6). The bottom wall is at y = 0,
and the dashed lines on the right show the location of the top wall at different Reynolds
numbers. To enhance clarity, only the error bars for the highest Reynolds number are
shown.

in appendix A at polymer mass fractions just above wt . A close examination of


the velocity profiles for the flow through rigid and flexible channels suggests a
higher viscosity near the walls and a lower viscosity near the centre. This is the
opposite of a shear-thinning effect or shear-induced migration, where the viscosity
is expected to be lower at the walls (where the strain rate is higher) and higher at
the centre (where the strain rate is zero). One possible explanation is migration of
polymers from centre to the walls, resulting in slight variations of the viscosity. The
other possibility is shear-thickening due to finite extensibility or due to association
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1100 S. S. Srinivas and V. Kumaran

(a) (b) 0.4

0.02
0.3

0.2
0.01

0.1

0 100 200 300 400 500 0 100 200 300 400 500

(c) (d)
0.10

0.010

0.05
0.005

0 100 200 300 400 500 0 100 200 300 400 500
Re Re

F IGURE 14. The measure vdiff of the departure of the velocity profile from the
laminar profile (a), and the scaled maximum of the velocity fluctuations max(vx0 )/v (b),
max(vy0 )/v (c) and max(h|vx0 vy0 |i)/v 2 (d), all at location II (in figure 1a), for a solution
of polyacrylamide with molecular weight 4 104 and with mass fraction 1500 ppm (E),
1000 ppm (A), 700 ppm (C), 500 ppm (B), 100 ppm (D) and pure water (6). The
vertical dashed lines show the transition Reynolds numbers (table 2) at different polymer
concentrations.

between polymers at very high shear rates (Hatzikiriakos & Vlassopoulos 1996).
This effect, which requires further study possibly by direct measurement of polymer
mass fraction, is numerically small but discernible in both hard- and soft-walled
channels. Since the computation of the stress assumes a constant viscosity, we find
that the stress is no longer linear, but the magnitude of the slope of the xy y curve
seems to decrease near the walls. This subtle effect is similar to that of added
polymers on the solution viscosity, which is also small but discernible. However,
this has a significant effect on the transition, and the flow dynamics in a turbulent
flow.
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1101

(a) (b) 0.4

0.02
0.3

0.2
0.01

0.1

0 100 200 300 400 500 0 100 200 300 400 500

(c) 0.10 (d)


0.010

0.05
0.005

0 100 200 300 400 500 0 100 200 300 400 500
Re Re

F IGURE 15. The measure vdiff of the departure of the velocity profile from the
laminar profile (a), and the scaled maximum of the velocity fluctuations max(vx0 )/v (b),
max(vy0 )/v 2 (c) and max(h|vx0 vy0 |i)/v (d), all at location II (in figure 1a), for a solution
of polyacrylamide with molecular weight 5 106 and with mass fraction 50 ppm (u),
30 ppm (E), 10 ppm (A), 2 ppm (C), 1 ppm (B), 0.5 ppm (D) and pure water (6). The
vertical dashed lines show the transition Reynolds numbers (table 2) at different polymer
concentrations.

Next, we discuss quantitative measures of the departure from a laminar flow


as a function of the Reynolds number. The departure from the mean velocity is
characterised by the measure vdiff , defined as
s
h
1
Z
vdiff = ( p)2
dy (vx vx(l) )2 , (3.1)
vx h 0

where vx (y) is the experimental measurement, vx(l) (y) is the laminar velocity profile
from the ANSYS FLUENT simulations for a channel of the same shape and with the
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1102 S. S. Srinivas and V. Kumaran
same flow rate as in the experiments, and vx( p) is the profile-averaged velocity,
h
1
Z
vx( p) = vx dy. (3.2)
h 0

The fluctuations are characterised by the scaled fluctuation intensities, max(vx0 )/v,
max(vy0 )/v and max(|hvx0 vy0 i|)/v 2 , where max(?) is the maximum value of ? across the
channel, and the mean velocity v is the ratio of the flow rate and the cross-sectional
area.
Figures 14(a) and 15(a) show vdiff as a function of Reynolds number for different
polymer mass fractions. This measure vdiff is numerically small when the flow is
laminar, but it exhibits a sharp increase by a factor of 1.52 at transition, indicating
a qualitative change in the form of the velocity profile. The difference between the
experimental and laminar profiles decreases systematically with an increase in the
polymer mass fraction. However, a sharp change in the value of vdiff is evident at the
transition Reynolds number for all values of the polymer mass fraction.
Figures 14(bd) and 15(bd) show the maximum values of vx0 , vy0 and |hvx0 vy0 i|
across the channel, scaled by suitable powers of v as a function of the Reynolds
number. At transition, there is a discontinuous change in max(vx0 )/v, max(vy0 )/v and
max(|hvx0 vy0 i|)/v 2 for all values of the polymer mass fraction. For the solution of
polyacrylamide with molecular weight 4 104 , max(vx0 )/v and max(vy0 )/v decrease
by 30 % when the polymer mass fraction is increased from 0 ppm to 500 ppm,
and the correlation max(hvx0 vy0 i)/v 2 decreases by more than 50 %. For the solution of
polyacrylamide with molecular weight 5 106 , there is a larger decrease of 50 %
in max(vx0 )/v and max(vy0 )/v, and 80 % reduction in max(hvx0 vy0 i)/v 2 when the
polymer mass fraction is increased from 0 ppm to 1 ppm. When the polymer mass
fraction exceeds the threshold value, there is a further decrease in the amplitude of
the fluctuations, accompanied by a decrease in the transition Reynolds number. At
the highest polymer mass fractions, the values of max(vx0 )/v and max(vy0 )/v are lower
by a factor of 5, and the value of max(hvx0 vy0 i)/v 2 is lower by a factor of 10, in
comparison to that of pure water. Even at the highest polymer mass fractions studied
here, there is a sharp increase by a factor of 2 in max(vx0 )/v and max(vy0 )/v and a
factor of 4 in max(hvx0 vy0 i)/v 2 at transition. This increase is observed at a Reynolds
number as low as 139 for the polymer solution with molecular weight 4 104 and
mass fraction 1500 ppm, and as low as 35 for the polymer solution with molecular
weight 5 106 and mass fraction 50 ppm.

3.4. Wall motion


The displacement of a dye spot marked on the bottom wall at location II in the
streamwise direction, u x , is shown in figure 16. It should be noted that the mean
displacement is proportional to the mean shear stress at the wall, if linear elasticity is
used to model the displacement within the wall medium. Figure 16 shows that there is
a distinct change in the slope of the graphs of u x versus Reynolds number at transition
for all values of the polymer mass fraction considered here. This is consistent with
the slopes of the curves of the maximum height with Reynolds number in figure 25 in
appendix B. The mean displacement and the shear stress first decrease as the polymer
mass fraction increases, and then increase when the mass fraction increases beyond wt .
As the polymer mass fraction is increased, the height of the channel first increases,
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1103

(a) 10 (b) 10

5 5

0 100 200 300 400 500 0 100 200 300 400 500
Re Re

F IGURE 16. The mean wall displacement as a function of the Reynolds number Re
for (a) polymer with molecular weight 4 104 and mass fraction 1500 ppm (E),
1000 ppm (A), 700 ppm (C), 300 ppm (B) and 100 ppm (D), and (b) polymer with
molecular weight 5 106 and mass fraction 50 ppm (u), 30 ppm (E), 10 ppm (A),
2 ppm (C), 1 ppm (B) and 0.5 ppm (D). The results for pure water are shown by
the 6 symbols in both panels. The vertical dashed lines show the transition Reynolds
numbers (table 2) at different polymer concentrations as indicated by the symbols.

resulting in a lower strain rate for the same flow rate. As the mass fraction is further
increased, there is an increase both in the area of cross-section and in the viscosity;
the combination of these two effects results in a net increase in the wall shear stress
and consequently the wall displacement.
The r.m.s. of the wall displacement fluctuations in the streamwise direction, u0x , is
shown in figure 17. The r.m.s. of the displacement in the spanwise direction, u0z , is
25 % smaller than u0x , but it has the same qualitative characteristics, and so it is not
shown separately. For pure water and for low polymer mass fraction below wt , the
variation of u0x with the Reynolds number is in agreement with the earlier study of
Srinivas & Kumaran (2015). A baseline fluctuation level is observed even in a laminar
flow due to experimental noise, but there is a sharp increase in the level of fluctuations
at transition in both the streamwise and spanwise directions, confirming the onset of
wall motion along with the fluctuations in the fluid. As the polymer concentration was
increased, the magnitude of the fluctuations decreased below the detection limit in our
experiments, which is 0.5 m, and so we were not able to detect a change in the
fluctuations.
The frequency spectrum of the fluctuations has also been examined in the
range of 0500 rad s1 at low polymer concentrations. It should be noted that
the fluctuations were measured using a high-speed imaging system with a frame
rate of 1000 frame s1 , and so the Nyquist frequency for the measurements
is 500 frame s1 , or 3000 rad s1 . The measurements have been restricted
to 500 rad s1 in order to provide adequate oversampling. The low-frequency
structure in the spectrum is evident in figure 18(a) for the flow of pure water,
and is similar to the results reported in Srinivas & Kumaran (2015). There is
no sharp maximum in the range 0500 rad s1 , and there is no clear scaling in
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1104 S. S. Srinivas and V. Kumaran

(a) 1.5 (b)

1.0

1.0

0.5
0.5

0 100 200 300 400 500 0 100 200 300 400 500
Re Re

F IGURE 17. The r.m.s. of the fluctuations in the wall displacement as a function of
the Reynolds number Re for (a) polymer with molecular weight 4 104 and mass
fraction 1500 ppm (E), 1000 ppm (A), 700 ppm (C), 300 ppm (B) and 100 ppm (D), and
(b) polymer with molecular weight 5 106 and mass fraction 50 ppm (u), 30 ppm (E),
10 ppm (A), 2 ppm (C), 1 ppm (B) and 0.5 ppm (D). The results for pure water are
shown by the 6 symbols in both panels. The vertical dashed lines show the transition
Reynolds numbers (table 2) at different polymer concentrations as indicated by the
symbols.

with frequency, but the appearance of a broad spectrum at frequencies less then
100 rad s1 is observed in the experiments. When the polymer concentration is
increased just above the threshold concentration, the amplitude of the spectrum
decreases significantly, as shown in figure 18(b), and we do not observe any coherent
spectrum. This is probably because the amplitude of the velocity fluctuations
has decreased below 0.5 m, which is the limit of resolution of our optical
measurement technique. A similar result is obtained when the polymer concentration
is further increased. More advanced techniques for the measurement of displacement
fluctuations are required in order to detect the low-amplitude fluctuations for polymer
solutions.

4. Conclusions
The primary result of the experimental analysis is that polymer addition has two
types of effects on transition and turbulence in a soft-walled microchannel.
(1) When the polymer mass fraction is below a threshold value wt , there is a
reduction in the turbulence intensities after transition, but there is no change
in the transition Reynolds number. The value of wt is 1 ppm for the
polyacrylamide with molecular weight 5 106 , and 500 ppm for the polymer
with molecular weight 4 104 . The turbulence reduction is significant. When
scaled by suitable powers of the average flow velocity, the dimensionless
streamwise and cross-stream r.m.s. velocities decrease by a factor of 2, and
the correlation hvx0 vy0 i (scaled by the square of the average velocity) decreases by
a factor of 4, when the polymer weight fraction increases from 0 to wt .
(2) There is a systematic decrease in the transition Reynolds number with increasing
mass fraction when the polymer concentration increases beyond wt . For the
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1105

(a) 1000 (b) 60

50

40

500 30

20

10

0 5 10 15 20 0 10 20

F IGURE 18. The spectra of the streamwise wall displacement, u x , at different Reynolds
numbers for r.m.s. of the fluctuations in the wall displacement as a function of the
Reynolds number for pure water (a) and a solution of polyacrylamide with molecular
weight 5 106 and mass fraction 2 ppm (b). Vertical offsets are added to the baselines
at different Reynolds numbers to distinguish the different curves.

highest mass fraction of 1500 ppm for the polymer of molecular weight 4 104 ,
the transition Reynolds number is 139, and for the highest mass fraction of
50 ppm for the polymer of molecular weight 5 106 , the transition Reynolds
number is 35. This is in contrast to the transition Reynolds number of 291
when the mass fraction is below wt . The decrease in the transition Reynolds
number is accompanied by a significant decrease in the scaled turbulence
intensities, by a factor of 5 for the streamwise and cross-stream velocity
fluctuations, and by a factor of more than 10 for the correlation hvx0 vy0 i, in
comparison to pure water.
(3) The turbulence damping results in very poor cross-stream mixing even at very
small polymer concentrations.
There are three other incidental observations which may be of relevance in
microfluidics and in the flow of polymer solutions.
(i) It is observed that addition of small amounts of polymers causes a significant
damping of fluctuations even in the laminar flow. In the flow configurations
used here and in many other microfluidic devices, the inlets are perpendicular
to the microchannels and there are often sharp or gentle bends in the channels.
These generate significant velocity fluctuations even in the laminar flow. Here,
we observe that, even in the laminar flow, there is a significant damping of flow
disturbances due to the addition of small amounts of polymer, even when there
is very little change in the solution viscosity. Thus, polymer addition seems to
have a disproportionate effect on the fluctuations in comparison to the mean
flow.
(ii) For mass fractions around the threshold mass fraction wt , there is a small but
discernible change in the shape of the mean velocity profiles, for both rigid and
flexible channels. The strain rate appears to be smaller near the wall, and larger
near the centre, in comparison to that for a parabolic profile with the same flow
rate. This effect is not observed either for pure water or for the largest polymer
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1106 S. S. Srinivas and V. Kumaran
concentrations used here. This implies that the effective viscosity is higher near
the wall and lower near the centre the opposite of the effect expected either
due to shear-thinning or due to shear migration from high- to low-shear regions.
This is also not likely to be due to polymer adsorption on the wall, because
the radius of gyration of the molecules is only 0.5 m for the polymer with
molecular weight 5 106 (Kulicke et al. 1982), which is two orders of magnitude
smaller than the channel height. Further study is required to explain this
effect.
(iii) A transition is observed even in a rigid channel for relatively large polymer mass
fractions (30 and 50 ppm) of the solution of polyacrylamide with molecular
weight 5 106 at Reynolds numbers around 300, as discussed in appendix A.
In the dye-stream experiments, disturbances are observed at the interface
between the inlet streams, and there is a sharp increase in the magnitude of
the velocity fluctuations. This instability is not observed for the polyacrylamide
with molecular weight 4 104 even at the largest mass fraction of 1500 ppm.

5. Discussion
It should be emphasised that the polymer mass fractions used in this study are
sufficiently low that there is very little change in the viscosity of the polymer
solutions. For the solution of polyacrylamide of molecular weight 5 106 , even at
the maximum concentration of 50 ppm, the viscosity exceeds that of water by less
than 10 %. Thus, the modification of the steady shear viscosity due to the addition of
polymers is very small. However, as our study reveals, the polymers have a significant
and disproportionate effect on the soft-wall transition and turbulence.

5.1. Concentration regime


Polymer solutions are classified into dilute, semi-dilute and concentrated regimes
based on the polymer concentration. In the dilute regime, the concentration is
sufficiently low that the separation between polymer chains is much larger than
the radius of gyration of a chain, so that the polymer molecules are effectively
non-interacting. In the semi-dilute regime, the mass fraction of the polymer is much
smaller than 1, but there is overlap between different polymer molecules. The critical
concentration for the transition from the dilute to the semi-dilute regime is measured
in different ways. The critical concentration based on the volume fraction, c , is the
concentration at which polymer molecules start to physically overlap; this is calculated
from the volume occupied by each molecule based on the radius of gyration. The
critical concentration based on the solution viscosity c is the concentration at
which the polymer interactions are sufficiently strong to affect the transmission
of stress. Both of these critical concentrations scale as M 4/5 (de Gennes 1976),
where M is the molecular weight. For a solution of polyacrylamide with molecular
weight 5 106 , Kulicke et al. (1982) report that c = 1.5 104 g cm3 , which is
equivalent to 150 ppm, while c has a much larger value of 1.67 103 g cm3 ,
which is equivalent to 1670 ppm. For a solution of polyacrylamide with molecular
weight 4 104 , the critical concentrations obtained using the M 4/5 scaling law are
c = 7.14 103 g cm3 (mass fraction 7140 ppm) and c = 7.95 102 g cm3
(mass fraction 79 500 ppm). Thus, all the polymer concentrations used here are
clearly below any of these definitions of the critical concentration. The two distinct
turbulence modification effects, discussed at the beginning of this section, appear to
be entirely due to non-interacting polymer molecules. However, it must be cautioned
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1107
that we have not directly verified the homogeneity of the distribution of the polymer
molecules; this would require advanced techniques for directly imaging polymers in
high-speed flows at high frame rates and small scales.
Based on the limited data available, the threshold mass fraction wt decreases by a
factor of 500, from 500 ppm to 1 ppm, when the molecular weight is increased
by a factor of 125, from 4 104 to 5 106 . This decrease in wt is possibly indicative
of a decrease proportional to M 1 , subject to factors of O(1), but is too large to scale
as the inverse of the radius of gyration, which scales as M 0.6 , and too small to scale
as the inverse of the polymer relaxation time, which scales as M 1.8 (de Gennes 1976).

5.2. Polymer relaxation time


As discussed in 2.7, for the polyacrylamide with molecular weight 5 106 , the
Weissenberg number varies between 1.2 Re and 0.12 Re, and for the polymer with
molecular weight 4 104 , the Weissenberg number varies between 2 104 Re and 2
105 Re. Thus, the Weissenberg number changes by a factor of 3000 for the different
polymer molecular weights used here. The qualitative effect of added polymer on the
fluid turbulence does not seem to depend on the Weissenberg number, though the
threshold mass fraction for reduction in the transition Reynolds number does depend
on the polymer molecular weight. There is one effect that does seem to depend on the
Weissenberg number, which is the instability in a rigid channel reported in appendix A.
This instability is observed at a relatively high Reynolds number of 300 only for the
solution of polyacrylamide with molecular weight 5 106 and polymer concentrations
30 and 50 ppm. In this case, the Weissenberg numbers are very large, of the order
of 102 , and an instability in a rigid channel at such large Weissenberg number and
moderate Reynolds number does not seem to have been reported before.

5.3. Linear stability studies


The effect of polymer addition on the wall-mode instability of the interface between
a fluid and a soft solid for a Couette flow has been carried out by Chokshi et al.
(2015). The configuration is not the same as that in the present experiments, and wall
deformation is not incorporated in that study. So a quantitative comparison cannot be
made, but the predictions can be qualitatively compared with the experiments. Chokshi
et al. (2015) predict that polymer addition reduces the transition Reynolds number for
most parameter values, unless the solid is very thin and the polymer solution is very
dilute. In Chokshi et al. (2015), the parameter = Gh2 /2 was considered to be in
the range 106 107 , which is comparable to the 4.5 105 to 4.6 106 in the present
experiments. For this range of numbers, Chokshi et al. (2015) predicted a decrease
in the transition Reynolds number by a factor of 10 at the highest Weissenberg number
of 50 considered in the analysis; the authors anticipated a greater reduction when the
Weissenberg number is further increased.
For the polyacrylamide solution with molecular weight 4 105 , the Weissenberg
number is in the range 2 103 2 102 when the Reynolds number is 100 (see
2.7). For this, the transition Reynolds number decreases by a factor greater than
2 at the highest mass fraction of 1500 ppm. For the polyacrylamide solution with
molecular weight 5 106 , the Weissenberg number varies between 12 and 120 at a
Reynolds number of 100. The reduction in the Reynolds number, by a factor of 10,
is consistent with the prediction of Chokshi et al. (2015) at a Weissenberg number
of 50. While making the comparison, it is important to note that deformation is not
taken into account in Chokshi et al. (2015). Previous linear stability studies on the
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1108 S. S. Srinivas and V. Kumaran
flow of Newtonian fluids in soft channels (Verma & Kumaran 2013, 2015) have
shown that the transition Reynolds number could decrease by an order of magnitude
when deformation is incorporated. Subject to this caveat, a cautious conclusion is
that the experimental results are qualitatively in agreement with the predictions of
Chokshi et al. (2015), though the correct geometry and wall deformation have to be
incorporated to make a quantitative comparison.

5.4. Transition to turbulence in viscoelastic flows


Purely elastic turbulence in rotational flows of polymer solutions at low Reynolds
number is generated due to an instability caused by the stretching of polymer chains
along curved streamlines (Larson, Shaqfeh & Muller 1990). The resulting turbulent
flow shares many of the features of turbulence in Newtonian fluids, such as the
large velocity fluctuations and high diffusivities, but there were differences in salient
features such as the spectra for the decay of the velocity fluctuations. There have been
several studies on the possibility of turbulence in the absence of curved streamlines
in the base flow, where an instantaneous curvature of the streamlines could result
in a subcritical bifurcation due to a nonlinear stability mechanism (Morozov & van
Saarloos 2005). In the experiments of Pan et al. (2013), rather large perturbations
were required to trigger the subcritical instability. Pillars at the entrance were used
to generate perturbations, and the downstream progression of these perturbations
was analysed. Our observation of an instability in a rigid-walled channel at high
Weissenberg number of the order of 100 in appendix A appear to be different. We
observe the formation of a wavy interface even without imposing any perturbations.
This was an unexpected result while attempting to do clean experiments to validate
the experimental measurement technique, we found that we were unable to suppress
the instability even after reducing the level of fluctuations to the extent possible. So
the transition appears to be due to a linear instability of the base flow; this requires
further study.
At high Reynolds number, Samanta et al. (2013) considered the modification
of hard-wall transition and turbulence in a rigid channel. Linear stability studies
(Sadanandan & Sureshkumar 2002) do predict a decrease in the Reynolds number
for the linear stability below the value of 1000 for a rigid channel. Samanta et al.
(2013) find the transition Reynolds number to be 800 for the polyacrylamide solution
of 500 ppm used in their experiments. The von Krmn plots show that the flow
approaches the maximum drag reduction asymptote. There is a significant damping
in the turbulent fluctuations, and the velocity profile approaches the logarithmic law
proposed by Virk (1975) instead of the von Krmn log law for a Newtonian fluid.
In the present case, we study the soft-wall turbulence at much lower Reynolds
number in the range 250400. Here, the maximum concentration of the polyacrylamide
of molecular weight 5 106 is 50 ppm, which is the minimum concentration used
in Samanta et al. (2013), who explored the range 50500 ppm. The transition is due
a linear instability caused by dynamical interaction between the fluid and the wall,
and not the hard-wall transition. Linear stability analyses of the wall-mode instability
suggest a significant decrease in the transition Reynolds number. The present study
brings in one more aspect, which is the wall motion in the case of a soft wall. The
transfer of energy from the mean flow to the fluctuations is due to the shear work
done at the soft solidfluid interface. A previous study (Srinivas & Kumaran 2015)
has shown that the turbulent energy production is a maximum at the wall itself, and
not in the near-wall region. Therefore, this mechanism of turbulence generation is
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1109
distinct. However, the effect of added polymers is similar there is a significant
damping in the intensity of the turbulent fluctuations. Though the effect of polymer
dynamics on the transition has been predicted by Chokshi et al. (2015) using the
Oldroyd B model, the effect on turbulence generation is not yet fully understood, and
has to be explored in future.

5.5. Comparison with observations on drag reduction


As discussed in 1.4, significant drag reduction is not expected in soft-wall turbulence
in a microchannel, since the Reynolds stress forms a relatively small part of the total
stress. However, the modification of the turbulence features could be compared with
those reported in drag reduction studies.
A central idea in polymer drag reduction is the maximum drag reduction (MDR)
asymptote, which has been attributed to a modification in the constants in the von
Krmn log law for the mean velocity in the inertial sublayer (Virk 1975). The results
from numerical simulations indicate that the MDR log law is not perfect (Graham
2014), but it appears to be a good approximation in an intermediate region. The
near-wall logarithmic velocity profile is also a striking feature of soft-wall turbulence
(Srinivas & Kumaran 2015), but the constants in the von Krmn log law turn out
to be very different from those for a rigid channel. Moreover, the extent of the
logarithmic layer in soft-wall turbulence, 3 . y+ . 20, is also much closer to the
wall than the extent 30 . y+ . 200 reported for hard-wall turbulence. Here, y+ is
the distance from the wall scaled by /v , where v is the friction velocity and is
the kinematic viscosity. Despite the differences, it is pertinent to ask whether there
is a systematic variation of the logarithmic profile when small amounts of polymer
are dissolved in the fluid. The present studies indicate that, when polymer is added,
there is no observable logarithmic layer close to the wall. This is a negative result,
and so we have not provided details. However, the mean velocity profiles shown in
figure 9(b,c) are clearly well approximated by a parabolic profile, and so these would
not fit a logarithmic profile in the near-wall region.
The proposed mechanism of polymer drag reduction is the disruption of the rollup
and bursting of near-wall streaks in hard-wall bounded flows. Our previous study on
soft-wall turbulence (Srinivas & Kumaran 2015) has suggested that the mechanism
of turbulence generation may be different from that in hard-wall turbulence. In soft-
wall turbulence, the turbulence production rate is a maximum at the soft wall itself; in
contrast, the turbulence generation is a maximum near the wall in hard-wall turbulence
in the region where there is rollup and bursting of vortices. In the profiles of the
fluctuating velocities, it is observed that there is turbulence attenuation throughout the
channel, and not just in the near-wall region. Therefore, it is likely that the turbulence
generation mechanism here is different from that for hard-wall turbulence. Despite this,
a significant reduction in the turbulence intensities is observed upon addition of small
amounts of polymer.
The most astonishing feature observed here is that the effect of polymers on the
turbulent fluctuations appears to be disproportionate to that on the mean flow. Even
though the fluid viscosity changes by no more than 10 % with the addition of a
few ppm of polymer, the turbulence intensities decrease by an order of magnitude.
It would require more fundamental work to understand whether this is specific to
turbulence at small scales and high strain rates, or whether it is more general.

Acknowledgements
The authors would like to thank the Department of Science and Technology,
Government of India, for financial support. The authors gratefully acknowledge
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1110 S. S. Srinivas and V. Kumaran
the experimental facilities made available and advice provided by Professors O. N.
Ramesh, J. Dey, R. Govardhan, J. H. Arakeri, A. Ghosh, S. Basu, S. Bose and
S. V. Kailas. The authors are grateful to Professor A. C. Mandal, Dr M. K. S. Verma,
Mr J. Suryanarayana Murthy, Mr P. Kumar, Mr N. Jha, Mr V. Swamybabu, Mr S. Jaju,
Mr A. Tyagi and Mr U. Abbasi for instructive discussions and help with the
experiments.

Appendix A. Rigid channel


The validation of the laminar flow profiles was carried out in a rigid microchannel
of rectangular cross-section with height 160 m, width 1.5 mm and length 4 cm,
with the same Y inlet configuration as that shown in figure 1(a). In this case,
all four walls are made of PDMS gel with elasticity modulus 0.55 MPa. The
PIV measurements are carried out using the same configuration as those used for
the soft-walled microchannel described in 2. The parabolic velocity profiles are
recovered for the flow of pure water through the microchannel, as shown in figure 19.
There is no departure from the parabolic profile even when the Reynolds number
is 444, which is the highest Reynolds number that could be attained. Subject to
experimental error, a parabolic profile is also obtained for the highest polymer
mass fractions used here, which is 50 ppm for the solution of polyacrylamide with
molecular weight 5 106 , as shown in figure 20(b). For the intermediate polymer
mass fraction of 2 ppm shown in figure 20(a), there is good agreement between
the experimental and parabolic profiles, but there is also a discernible systematic
deviation. The strain rate appears to be lower at the wall and higher near the centre
in comparison to that for a parabolic profile. This seems to indicate a higher viscosity
near the wall and a lower viscosity near the centre, as discussed in the 3.
The measure of the difference between the mean and fluctuating velocities vdiff
(3.1), as well as the magnitudes of vx0 , vy0 and hvx0 vy0 i, scaled by suitable powers or the
mean velocity v for a rigid channel, are shown as a function of polymer mass fraction
for solutions of polyacrylamide with molecular weight 5 106 in figure 21. Here, v is
the ratio of the flow rate and cross-sectional area. The magnitudes of the fluctuations
are comparable to those observed for a laminar flow in a soft-walled channel in
figure 15. However, there is no sharp increase observed at any Reynolds numbers up
to 400, and the magnitudes are significantly smaller than those for the turbulent flow
in figure 15. The systematic decrease in the magnitude of the fluctuating velocities
with polymer concentration, observed in a soft-walled channel in the laminar flow, is
also observed in a rigid channel in figure 21. Qualitatively similar results are obtained
for the solution of polyacrylamide with molecular weight 4 104 .
An interesting feature is observed for the solution of polyacrylamide with molecular
weight 5 106 in figure 21. There is a sharp increase in the magnitude of the
fluctuating velocities for the solution with mass fraction 50 ppm between Reynolds
numbers of 267 and 311, and for the solution with mass fraction 30 ppm between
Reynolds numbers 311 and 355. This transition takes place at a Reynolds number
that is significantly higher than that for a flexible channel (which is 35 for the
solution with mass fraction 50 ppm and 72 for the solution with mass fraction
30 ppm). The average of the Reynolds numbers before and after the transition are
shown by the vertical dashed lines in figure 21. The transition is also reflected in
the images of dye-stream mixing taken from above the microchannel. Disturbance at
the interface between the two inlet streams is observed above the transition Reynolds
numbers, as shown in figure 22. It should be noted that this transition occurs at very
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1111

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

0.5

0 0.05 0.10 0.15


y (mm)

F IGURE 19. The mean velocity profiles for the flow of pure water through a rigid
microchannel at different Reynolds numbers of 133 (6), 200 (D), 333 (B) and 444 (E).
The solid lines are the velocity profiles along the central plane in the spanwise direction
from experiments, and the dashed lines are parabolic profiles with the same flow rate.

(a) 4.5 (b) 4.5


4.0 4.0
3.5 3.5
3.0 3.0
2.5 2.5
2.0 2.0
1.5 1.5
1.0 1.0
0.5 0.5

0 0.05 0.10 0.15 0 0.05 0.10 0.15


y (mm) y (mm)

F IGURE 20. The mean velocity profiles for the flow of a solution of polyacrylamide with
molecular weight 5 106 and (a) mass fraction 2 ppm at Reynolds numbers of 132 (6),
198 (D), 331 (B) and 441 (E) and (b) mass fraction 50 ppm at Reynolds numbers of
162 (6), 244 (C), 325 (A) and 406 (E). The solid lines are the velocity profiles along
the central plane in the spanwise direction from experiments, and the dashed lines are
parabolic profiles with the same flow rate.

Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1112 S. S. Srinivas and V. Kumaran

(a) (b) 0.2

0.05

0.1

0 100 200 300 400 500 0 100 200 300 400 500

(c) 0.05 (d) 0.004

0.04
0.003

0.03
0.002
0.02

0.001
0.01

0 100 200 300 400 500 0 100 200 300 400 500
Re Re

F IGURE 21. The measure vdiff of the departure of the velocity profile from the laminar
profile (a), the scaled maximum of the velocity fluctuations max(vx0 )/v (b), max(vy0 )/v (c)
and max(h|vx0 vy0 |i)/v 2 (d), in a rigid microchannel for a solution of polyacrylamide with
molecular weight 5 106 and with mass fraction 50 ppm (u), 30 ppm (E), 10 ppm (A),
2 ppm (C), 1 ppm (B), 0.5 ppm (D) and pure water (6). The vertical dashed lines show
the transition Reynolds numbers for polymer concentrations 50 ppm (u) and 30 ppm (E).

high Weissenberg number, of the order of 250, which are not normally accessible in
the flow of dilute polymer solutions. This transition is not observed for molecular
weight 4 104 , possibly because the Weissenberg number is too low. More work is
necessary to determine the reasons for this instability.

Appendix B. Channel deformation


The height along the central plane in the spanwise direction, shown by the vertical
line DF in figure 1(d), is plotted as a function of Reynolds number for pure water in
figure 23. There is no change in the height of the channel from the value of 160 m
in the upstream development section for x < 0, but there is a large increase at the
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1113

(a)

(b)

(c)

F IGURE 22. Images from, left to right, the entrance to the test section, location I, location
II and location III (in figure 1a), for the flow of a solution of polyacrylamide of molecular
weight 5 106 and concentration 50 ppm in a rigid microchannel of height 160 m at
Reynolds numbers 266 (a), 311 (b) and 333 (c). The clear solution is pumped into the
top inlet, and solution mixed with dye is pumped into the bottom inlet. The developing
section is to the left of the arrow, and the test section is to the right of the arrow.

entrance to the soft test section. The maximum height increases from 300 m at a
Reynolds number of 200, to a maximum of 550 m at the largest Reynolds number
of 388 considered here. It is noteworthy that there is no sharp change in the height
of the channel between the Reynolds numbers 250277 (where transition is observed
at location III) and 277305 (where transition is observed at locations I and II). Thus,
the transition in the flow characteristics is not accompanied by a discontinuous change
in the channel height. Another important observation is that the slope of the wall is
small even for the highest Reynolds numbers considered here. At the entrance to the
test section, the maximum wall slope is less than 8 % at the Reynolds number is 388,
and the slope decreases with downstream distance. At the locations I, II and III where
the results are reported, the maximum slope is 3 % or less at the highest Reynolds
number of 388.
Figure 24 shows the height profiles along the central plane in the spanwise direction
for solutions of polyacrylamide of molecular weight 5 106 with mass fraction just
above wt , 2 ppm, and for the maximum mass fraction, 50 ppm. There is an increase
in the maximum height with polymer mass fraction, from 550 m for pure water
to a little above 600 m when the polymer mass fraction is 2 ppm, and then a larger
increase to 800 m when the mass fraction is 50 ppm. This is expected due to the
normal stress difference in polymer solutions, which tends to push the channel walls
apart, and which increases as the polymer mass fraction increases. The maximum
slope of the wall at the entrance to the test section also increases to a maximum of
12 % for the polymer with molecular weight 5 106 and mass fraction 50 ppm.
However, at the locations I, II and III where the flow and wall characteristics are
measured, the maximum slope is only 5 %.
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1114 S. S. Srinivas and V. Kumaran

0.6 I II III

0.5

0.4
h (mm)

0.3

0.2

0.1

0
0 0.5 1.0 1.5 2.0 2.5 3.0
x (cm)

F IGURE 23. The channel height h as a function of streamwise distance x measured from
the joint between the hard and soft sections (figure 1a) for pure water at Reynolds number
200 (E), 222 (A), 250 (C), 277 (B), 305 (D), 333 (6) and 388 (). The locations I, II
and III, where the PIV measurements are carried out, are shown by the dashed vertical
lines.

The maximum channel height is shown as a function of Reynolds number in


figure 25. When the polymer mass fraction is below wt , there is a small but
discernible increase in the height. However, above the threshold mass fraction, there
is a significant increase in the maximum height of the channel. There is no apparent
discontinuity in the channel height at transition. However, it is also observed that
the slope of the hmax versus Re curves also appears to increase significantly at the
transition Reynolds number for each value of the polymer mass fraction, and then
decreases as the Reynolds number is further increased. This suggests a significant
enhancement of the first normal stress difference at the transition point.
Owing to the channel deformation, the Reynolds number for the velocity profile
along the central plane in the spanwise direction (line DF in figure 1d) is different
from the Reynolds number based on the flow rate and channel width, equation (2.6).
There are two reasons for this difference: the increase in the height of the channel
along the central plane in the spanwise direction, and the greater velocity along the
centreline of the channel in comparison to the region near the side walls, where the
no-slip condition retards the flow. For the mean velocity profile determined by the PIV
measurements in figure 9, the profile-averaged Reynolds number can be defined as

vx( p) hmax
Rep = , (B 1)

where vx( p) is the profile-averaged velocity in (3.2), and is the kinematic viscosity.
The Reynolds number Rep is shown as a function of the Reynolds number Re (2.6)
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1115

(a) I II III (b) 0.9 I II III


0.6
0.8

0.5 0.7
0.6
0.4
h (mm)

0.5
0.3 0.4

0.2 0.3
0.2
0.1
0.1
0 0
0 0.5 1.0 1.5 2.0 2.5 3.0 0 0.5 1.0 1.5 2.0 2.5 3.0
x (cm) x (cm)

F IGURE 24. The channel height as a function of streamwise distance x measured from
the joint between the hard and soft sections (figure 1a) for a solution of polyacrylamide
of molecular weight 5 106 and mass fraction 2 ppm (a) and 50 ppm (b). In (a), the
Reynolds numbers are 110 (E), 154 (A), 209 (C), 231 (B), 264 (D), 308 (6) and 396 ().
In (b), the Reynolds numbers are 20.3 (E), 30.4 (A), 40.6 (C), 81.2 (B), 121.8 (D), 244 (6)
and 365 (). The locations I, II and III, where the PIV measurements are carried out, are
shown by the dashed vertical lines.

(a) (b) 1.0


0.7
0.9
0.6 0.8
0.7
0.5
0.6
0.4
0.5
0.3 0.4
0.3
0.2
0.2
0.1
0.1

0 100 200 300 400 500 0 100 200 300 400 500
Re Re

F IGURE 25. The maximum height as a function of the Reynolds number Re for
(a) polymer with molecular weight 4 104 and mass fraction 1500 ppm (E),
1000 ppm (A), 700 ppm (C), 300 ppm (B) and 100 ppm (D) and (b) polymer with
molecular weight 5 106 and mass fraction 50 ppm (u), 30 ppm (E), 10 ppm (A),
2 ppm (C), 1 ppm (B) and 0.5 ppm (D). The results for pure water are shown by
the 6 symbols in both panels. The vertical dashed lines show the transition Reynolds
numbers (table 2) at different polymer concentrations as indicated by the symbols.

in figure 26. At location I, where the deformation is largest, the profile-averaged


Reynolds number could exceed that based on the flow rate by 100200 when the
Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1116 S. S. Srinivas and V. Kumaran

(a) 600 (b) 600

500 500

400 400

300 300

200 200

100 100

0 100 200 300 400 500 0 100 200 300 400 500
Re Re

F IGURE 26. The profile-averaged Reynolds number Rep (B 1) as a function of the


Reynolds number Re (2.6) at location I (solid line) and location III (dashed line) (see
figure 1a). In (a), the polyacrylamide molecular weight is 5 106 and the mass fractions
are 2 ppm (A) and 50 ppm (C). In (b), the polyacrylamide molecular weight is 4 104
and the mass fractions are 700 ppm (A) and 1500 ppm (C). In both panels, the results for
pure water are shown by the E symbol. Re and Rep are equal along the thick dotted line.

Reynolds number is 400, due to the large increase in the height. However, at location
III, the profile-averaged Reynolds number is quite close to that based on flow rate
even when the Reynolds number is as large as 400, since the deformation of the
channel is relatively small.

REFERENCES

B OHDANECKY, M., P ETRUS , V. & S EDLDTEK , B. 1983 Estimation of the characteristic ratio of
polyacrylamide in water and in a mixed theta-solvent. Makromol. Chem. 184, 20612073.
C HOKSHI , P., B HADE , P. & K UMARAN , V. 2015 Wall-mode instability in plane shear flow of
viscoelastic fluid over a deformable solid. Phys. Rev. E 91, 023007.
C HOKSHI , P. P. & K UMARAN , V. 2007 Stability of the flow of a viscoelastic fluid past a deformable
surface in the low Reynolds number limit. Phys. Fluids 19, 104103.
C HOKSHI , P. P. & K UMARAN , V. 2008 Weakly nonlinear analysis of viscous instability in flow past
a neo-Hookean surface. Phys. Rev. E 77, 056303.
C HOKSHI , P. P. & K UMARAN , V. 2009 Weakly nonlinear stability analysis of a flow past a neo-
Hookean solid at arbitrary Reynolds numbers. Phys. Fluids 21, 014109.
D OI , M. & E DWARDS , S. F. 1988 The Theory of Polymer Dynamics. Oxford Science Publications.
E GGERT, M. D. & K UMAR , S. 2004 Observations of instability, hysteresis and oscillations in low-
Reynolds-number flow past polymer gels. J. Colloid Interface Sci. 274, 238242.
G AURAV & S HANKAR , V. 2009 Stability of fluid flow through deformable neo-Hookean tubes. J. Fluid
Mech. 627, 291322.
G AURAV & S HANKAR , V. 2010 Stability of pressure-driven flow in a deformable neo-Hookean channel.
J. Fluid Mech. 659, 318350.
DE G ENNES , P. G. 1976 Dynamics of entangled polymers solutions. I. The Rouse model.
Macromolecules 9, 587593.
G KANIS , V. & K UMAR , S. 2003 Instability of creeping Couette flow past a neo-Hookean solid. Phys.
Fluids 15, 28642871.

Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
Effect of viscoelasticity on soft-wall turbulence 1117
G KANIS , V. & K UMAR , S. 2005 Stability of pressure driven creeping flows in channels lined with a
nonlinear elastic solid. J. Fluid Mech. 524, 357375.
G RAHAM , M. D. 2014 Drag reduction and the dynamics of turbulence in simple and complex fluids.
Phys. Fluids 26, 101301.
H ATZIKIRIAKOS , S. G. & V LASSOPOULOS , D. 1996 Brownian dynamics simulations of shear-
thickening in dilute polymer solutions. Rheol. Acta 35, 274287.
K RINDEL , P. & S ILBERBERG , A. 1979 Flow through gel-walled tubes. J. Colloid Interface Sci. 71,
3950.
K ULICKE , W.-M., K NIEWSKE , R. & K LEIN , J. 1982 Preparation, characterisation, solution properties
and rheological behaviour of polyacrylamide. Prog. Polym. Sci. 8, 373468.
K UMARAN , V. 1995 Stability of the viscous flow of a fluid through a flexible tube. J. Fluid Mech.
294, 259281.
K UMARAN , V. 1996 Stability of an inviscid flow in a flexible tube. J. Fluid Mech. 320, 117.
K UMARAN , V. 1998 Stability of wall modes in a flexible tube. J. Fluid Mech. 362, 115.
K UMARAN , V. 2000 Classification of instabilities in the flow past flexible surfaces. Curr. Sci. 79,
766773.
K UMARAN , V. 2003 Hydrodynamic stability of flow through compliant channels and tubes. In
Proceedings of IUTAM Symposium on Flow in Collapsible Tubes and Past Other Highly
Compliant Boundaries (ed. P. W. Carpenter & T. J. Pedley). Kluwer Academic Publishers.
K UMARAN , V. 2015 Experimental studies on the flow through soft tubes and channels. Sadhana 40,
911923.
K UMARAN , V. & BANDARU , P. 2016 Ultra-fast microfluidic mixing by soft-wall turbulence. Chem.
Engng Sci. 149, 156168.
K UMARAN , V., F REDRICKSON , G. H. & P INCUS , P. 1994 Flow induced instability at the interface
between a fluid and a gel at low Reynolds number. J. Phys. France II 4, 893911.
K UMARAN , V. & M URALIKRISHNAN , R. 2000 Spontaneous growth of fluctuations in the viscous
flow of a fluid past a soft interface. Phys. Rev. Lett. 84, 33103313.
L AHAV, J., E LIEZER , N. & S ILBERBERG , A. 1973 Gel-walled cylindrical channels as models for the
microcirculation: dynamics of flow. Biorheology 10, 595604.
L ARSON , R. G., S HAQFEH , E. S. G. & M ULLER , S. J. 1990 A purely elastic instability in Taylor
Couette flow. J. Fluid Mech. 218, 573600.
L EE , M. G., C HOI , S. & PARK , J.-K. 2010 Rapid multivortex mixing in an alternately formed
contractionexpansion array microchannel. Biomed. Microdevices 12, 10191026.
L ENZMANN , F., L I , K., K ITAI , A. H. & S TOVER , H. D. H. 1994 Thin-film micropatterning using
polymer microspheres. Chem. Mater. 6, 156159.
M OROZOV, A. N. & VAN S AARLOOS , W. 2005 Subcritical finite-amplitude solutions for plane couette
flow of viscoelastic fluids. Phys. Rev. Lett. 95, 024501.
M URALIKRISHNAN , R. & K UMARAN , V. 2002 Experimental study of the instability of the viscous
flow past a flexible surface. Phys. Fluids 14, 775780.
N EELAMEGAM , R., S HANKAR , V. & DAS , D. 2013 Suppression of purely elastic instabilities in the
torsional flow of a viscoelastic liquid past a soft solid. Phys. Fluids 25, 124102.
PAN , L., M OROZOV, A., WAGNER , C. & A RRATIA , P. E. 2013 Nonlinear elastic instability in channel
flows at low Reynolds numbers. Phys. Rev. Lett. 110, 174502.
P OOLE , R. J., E SCUDIER , M. P., A FONSO , A. & P INHO , F. T. 2007 Laminar flow of a viscoelastic
shear-thinning liquid over a backward-facing step preceded by a gradual contraction. Phys.
Fluids 19, 093101.
S ADANANDAN , B. & S URESHKUMAR , R. 2002 Viscoelastic effects on the stability of wall-bounded
shear flows. Phys. Fluids 14, 4148.
S AMANTA , D., D UBIEF , Y., H OLZNER , M., S CHFER , C., M OROZOV, A. N., WAGNER , C. & H OF ,
B. 2013 Elasto-inertial turbulence. Proc. Natl Acad. Sci. USA 110, 1055710562.
S HANKAR , V. 2015 Stability of fluid flow through deformable tubes and channels: an overview.
Sadhana 40, 925943.
S HANKAR , V. & K UMAR , S. 2004 Instability of viscoelastic plane couette flow past a deformable
wall. J. Non-Newtonian Fluid Mech. 116, 371393.

Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839
1118 S. S. Srinivas and V. Kumaran
S HANKAR , V. & K UMARAN , V. 1999 Stability of non-parabolic flows in a flexible tube. J. Fluid
Mech. 395, 211236.
S HANKAR , V. & K UMARAN , V. 2000 Stability of non-axisymmetric modes in a flexible tube. J. Fluid
Mech. 407, 291314.
S HANKAR , V. & K UMARAN , V. 2001a Asymptotic analysis of wall modes in a flexible tube revisited.
Eur. Phys. J. B 19, 607622.
S HANKAR , V. & K UMARAN , V. 2001b Weakly nonlinear stability of viscous flow past a flexible
surface. J. Fluid Mech. 434, 337354.
S HANKAR , V. & K UMARAN , V. 2002 Stability of wall modes in the flow past a flexible surface.
Phys. Fluids 14, 23242338.
S HRIVASTAVA , A., C USSLER , E. L. & K UMAR , S. 2008 Mass transfer enhancement due to a soft
boundary. Chem. Engng Sci. 63, 43024305.
S RINIVAS , S. S. & K UMARAN , V. 2015 After transition in a soft-walled microchannel. J. Fluid
Mech. 780, 649686.
S URESHKUMAR , R., B ERIS , A. N. & H ANDLER , R. A. 1997 Direct numerical simulation of the
turbulent channel flow of a polymer solution. Phys. Fluids 9, 743755.
T OMS , B. A. 1977 On the early experiments on drag reduction by polymers. Phys. Fluids 20,
S3S8.
V ERMA , M. K. S. & K UMARAN , V. 2012 A dynamical instability due to fluidwall coupling lowers
the transition Reynolds number in the flow through a flexible tube. J. Fluid Mech. 705,
322347.
V ERMA , M. K. S. & K UMARAN , V. 2013 A multifold reduction in the transition Reynolds number,
and ultra-fast mixing, in a micro-channel due to a dynamical instability induced by a soft
wall. J. Fluid Mech. 727, 407455.
V ERMA , M. K. S. & K UMARAN , V. 2015 Stability of the flow in a soft tube deformed due to an
applied pressure gradient. Phys. Rev. E 91, 043001.
V IRK , P. S. 1975 Drag reduction fundamentals. AIChE J. 21, 625656.
WANG , S.-N., G RAHAM , M. D., H AHN , F. J. & X I , L. 2014 Time-series and extended Karhunen
Love analysis of turbulent drag reduction in polymer solutions. AIChE J. 60, 14601475.
YANG , C., G RATTONI , C. A., M UGGERIDGE , A. H. & Z IMMERMAN , R. M. 2000 A model for
steady laminar flow through a deformable gel-coated channel. J. Colloid Interface Sci. 226,
105111.
Z ELL , A., G IER , S., R AFAI , S. & WAGNER , C. 2010 Is there a relation between the relaxation time
measured in caber experiments and the first normal stress coefficient? J. Non-Newtonain Fluid
Mech. 165, 12651274.

Downloaded from https:/www.cambridge.org/core. Indian Institute of Science, on 17 Mar 2017 at 17:11:36, subject to the Cambridge Core terms of use, available
at https:/www.cambridge.org/core/terms. https://doi.org/10.1017/jfm.2016.839

You might also like