You are on page 1of 16

3-weeks course report, s022973

The dependence of the


cross-sectional shape on the
hydraulic resistance of
microchannels

Hatim Azzouz

Supervisor: Niels Asger Mortensen and Henrik Bruus

MIC Department of Micro and Nanotechnology


Technical University of Denmark

5 July 2004
Contents

1 Introduction 3

2 Theory 4
2.1 Poiseuille Flow in a circular pipe . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1.1 Velocity field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1.2 Flow rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Poiseuille Flow in an elliptic pipe . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.1 Velocity field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.2 Flow rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Poiseuille Flow in a rectangular pipe - Fourier Transformation . . . . . . . . 8
2.3.1 Velocity field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3.2 Flow rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4 Hydraulic resistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4.1 Elliptic cross-section . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4.2 Rectangular cross-section . . . . . . . . . . . . . . . . . . . . . . . . 11

3 FEMLAB simulations 13

4 Conclusion 16

2
Chapter 1

Introduction

For a steady incompressible flow through a channel the hydraulic resistance of the channel
depends on the cross-section shape of the channel. The objective of this project is to
investigate the dependence between the hydraulic resistance and the cross-section of the
channel in detail as well as to investigate the hydraulic resistance for various cross-section
geometries. In the following, Chapter 2, a description of the Poiseuille flow through a
channel with a series of cross-section geometries will be given and a definition of the
hydraulic resistance will be introduced. The flow through channels will be simulated by
use of FEMLAB. The results obtained in the FEMLAB simulation will be compared with
those predicted by theory and discussed in Chapter 3. Finally, conclusions are given.

3
Chapter 2

Theory

2.1 Poiseuille Flow in a circular pipe


2.1.1 Velocity field
We consider an incompressible fluid flowing through a pipe (cylindrical tube) of length L
in the x direction and a circular cross section with radius a in the yz plane; see Figure 2.1.
The pressure drop over the length L is p, i.e p(0)=p0 +p and p(L)=p0 . We introduce
cylindrical polar coordinates in the yz-plane, y=rcos and z=rsin, where r is the radial
distance from the cylinder axis and is the azimuthal angle (the angle around from the z
axis).

Figure 2.1: A cylindrical tube with radius a.

We study a flow along the axis of the tube, so that

v = vx (r)ex = (vx (r), 0, 0). (2.1)

This field depends only on r and it is invariant under translations along the cylinder axis
as well as rotations.

4
2.1. POISEUILLE FLOW IN A CIRCULAR PIPE CHAPTER 2. THEORY

In applying the Navier-Stokes equation 1 we assume that there is neither external forces
nor instantaneous acceleration affecting the flow i.e f =0 and v
t =0, respectively, giving

2
p = v (2.3)

where is the viscosity of the fluid.


2 2 2
Since v=ex vx (r) it follows from projecting of the Navier-Stokes equation p= v
on er and e that
p p
= = 0. (2.4)
r
This means that pressure only depends on the longitudinal coordinate p = p(x). Calcu-
lating the Laplacian term acting on a function of r in the Navier-Stokes equation

2 dvx (r) d2 vx (r) 1 dvx
vx (r) = (vx (r)) = ex = + (2.5)
dr dr2 r dr
we finally obtain
dp(x) d2 vx (r) 1 dvx
= + . (2.6)
dx dr2 r dr
The left hand side of this equation depends only on x whereas the right hand side depends
only on r. Since the pressure difference over the length of the pipe is given by p
L it follows
that
dp(x) p
= (2.7)
dx L
where the minus sign is due to the assumption that if the fluid is supposed to run in the
positive x-direction the pressure must be higher at the beginning (x = 0) than at the end
(x = L). Thus we obtain, with a little rewriting, the following Navier-Stokes equation
equation
1 d dvx (r) p
r = . (2.8)
r dr dr L
The solution to this equation with no-slip boundary conditions, i.e vx (0)=vx (a)=0, is

r2
vx (r) = v0 1 2 (2.9)
a
where v0 is the integration constant. This can easily be shown to be the solution by
inserting
dvx (r) 2v0 r
= 2 (2.10)
dr a
1
Navier-Stokes equation is a partial differential equation expressing the local balance of momentum in
a fluid around any point of space at any time. It is given by
v 2
= p + v + f (2.2)
t
2
where the terms v t
is the instantaneous acceleration, p is the dynamical pressure gradient, v is
the shear viscosity term and f is the external body force.

5
2.2. POISEUILLE FLOW IN AN ELLIPTIC PIPE CHAPTER 2. THEORY

in the finally obtained Navier-Stokes. The integration constant is then determined to be


1 p 2
v0 = a . (2.11)
4 L
Thus, the solution to our Navier-Stokes equation is

1 p 2 r2
vx (r) = a 1 2 . (2.12)
4 L a

2.1.2 Flow rate


The total flow rate Q (volume pr. time) carried through the pipe is given by
I
Q= dydz vx (r). (2.13)
circle

In cylindrical polar coordinates it can be rewritten as


Z a Z 2
Q= rdrd vx (r) (2.14)
0 0

to get
p 4
Q= a . (2.15)
8 L

2.2 Poiseuille Flow in an elliptic pipe


2.2.1 Velocity field
For an elliptic pipe, as shown in Figure 2.2, with an elliptic cross section having axes a and
b along the y and the z axis, respectively, the same assumptions as for the circular pipe
are valid. Instead of using the cylindrical polar coordinates we introduce the cartesian
coordinates. The velocity field is again assumed longitudinal of the form

v = vx (y, z)ex = (vx (y, z), 0, 0). (2.16)

The Navier-Stokes equation is given by the same equation as in a circular pipe. Choosing
the cartesian coordinates the Navier-Stokes equation can be expressed as
2
vx (y, z) = p (2.17)

where the Laplacian term can be calculated by the same formalism of section 2.1 to give
2
2 vx (y, z) 2 vx (y, z)
vx (y, z) = + . (2.18)
y 2 z 2
Thus we obtain the following Navier-Stokes equation
2
1 p vx (y, z) 2 vx (y, z)
= + (2.19)
L y 2 z 2

6
2.2. POISEUILLE FLOW IN AN ELLIPTIC PIPE CHAPTER 2. THEORY

a b

Figure 2.2: A pipe with an elliptic cross-section having axes a and b

with the solution, satisfying the no-slip condition, given by



y2 z2
vx (y, z) = v0 1 2 . (2.20)
a2 b
2 2 vx (y,z) 2v0
Inserting vdy
x (y,z)
2 = 2v
a2 and
0
dz 2
= b2 in the Navier-Stokes equation we obtain the
integration constant
1 p (ab)2
vo = . (2.21)
2 L a2 + b2
Thus, the longitudinal velocity profile in an elliptic pipe becomes

1 p (ab)2 y2 z2
vx (y, z) = 1 2 . (2.22)
2 L a2 + b2 a2 b
Notice that for a = b the integration constant for the elliptic pipe becomes the one for the
circular pipe indicating that circular pipe is a special case of the elliptical.

2.2.2 Flow rate


The flow rate is given by I
Q= dydz vx (y, z). (2.23)
ellipse

Introducing the transformation variables Y = 1b z and Z=


1 y, and inserting them in the
a
integral for the flow rate we obtain
Z Z 1
Q = abv0 dY dZ (1 y2 z2 ) = 2abv0 rdr (1 r2 ) (2.24)
ellipse 0

After the integration we finally obtain the flow rate for the elliptic pipe

p (ab)3
Q= (2.25)
4 L a2 + b2

7
2.3. POISEUILLE FLOW IN A RECTANGULAR PIPE - FOURIER
TRANSFORMATION CHAPTER 2. THEORY

2.3 Poiseuille Flow in a rectangular pipe - Fourier Transfor-


mation
2.3.1 Velocity field
The solution of the flow in the rectangular pipe is not simple to obtain. The no-slip
boundary condition does not in this case allow for a separation of the 2D problem into
two independent 1D problems. Instead we solve the problem by Fourier analysis, where
only sine functions are involved since they guarantee fulfilment of the no-slip boundary
condition. We Fourier-expand in the z-axis, so that vx (y, 0) = vx (y, h) = 0. The solution
to the Navier-Stokes equation is of the form

X z
vx (y, z) = fn (y) sin(n ) (2.26)
n=1
h

In applying the Navier-stokes equation


P
2 vx (y, z) = (2.27)
L
we get
2
X
2 d fn (y) n2 2 z
vx (y, z) = 2 fn (y) sin(n ) (2.28)
n=1
dy 2 h h
In the Navier-Stokes equation we also have to Fourier transform the constant P/L,
or simpler just the constant 1 which can be multiplied by any constant. The Fourier
transform of the constant P/L thus becomes

P P P 4 X 1 z
= 1= sin(n ). (2.29)
L L L n,odd n h

The solution is only possible if

fn (y) = 0 for n even (2.30)


d2 fn (y) n2 2 P 4
fn (y) = for n odd (2.31)
dy 2 h2 L n
The first step to solve this nonhomogeneous equation is to find a general solution of the
homogeneous equation fn (y)h . The characteristic equation of the homogeneous equation
is
2 (n/h)2 = 0. (2.32)
It has the roots 1 = 0 and 2 = n/h. Hence a real general solution of the homogenous
equation is
fn (y)h = Aey1 + Bey2 . (2.33)

8
2.3. POISEUILLE FLOW IN A RECTANGULAR PIPE - FOURIER
TRANSFORMATION CHAPTER 2. THEORY

W H

Figure 2.3: A rectangular pipe of height h and width w.

Using the definitions of the hyperbolic functions

ex ex
sinh x = (2.34)
2
ex + ex
cosh x = (2.35)
2
the general solution of the homogeneous equation can be rewritten as

y y
fn (y)h = B cosh n + B sinh n (2.36)
h h
The second step is to find a particular solution fn (y)p of the nonhomogeneous equation.
Since the equation on the right has the derivative zero, we say

fn (y)p = C. (2.37)
4h2 1
Substitution gives (n h )2 C = P 4 P
L n . By comparison, C = L 3 n3
. Hence a general
solution of the nonhomogeneous equation is

fn (y) = fn (y)h + fn (y)p (2.38)



y y P 4h2 1
= B cosh n + B sinh n . (2.39)
h h 3 n3
P 4h2 1 1
From the boundary condition fn (/2) = fn (/2) = 0, we get B = w
3 n3 cosh(n 2h )
resulting in
" y #
P 4 h2 cosh n z
fn (y) = 1 . (2.40)
3 n3 cosh n w
2h

Thus the solution of the flow in the rectangular pipe is


" #

P 2 4 X 1 cosh n yz y
vx (y, z) = h 3 1 sin n . (2.41)
3 n w z
n,odd cosh n 2h

9
2.4. HYDRAULIC RESISTANCE CHAPTER 2. THEORY

2.3.2 Flow rate


The flow rate for a rectangular pipe is given by
Z w Z h
2
Q= dy dz vx (y, z). (2.42)
0 0
Using the following
Z y y
w
2 2h
2 dy cosh n = sinh n (2.43)
0 h n 2h
Z h z 2h
dz sinh n = for n odd (2.44)
0 h n
X 1 4
= . (2.45)
n
n,odd 4
96

we get " #
P
X 192 h 1 w
3
Q= wh 1 tanh n . (2.46)
12 n,odd
5 w n5 2h

2.4 Hydraulic resistance


In analogy with Ohms law for an electric resistor, it is natural to define the hydraulic
resistance Rhyd of a pipe carrying some liquid as

p = Rhyd Q (2.47)

Since we are interested in the relation between the hydraulic resistance and the cross-
section geometry of the pipe it is appropriate to reexpress the hydraulic resistance as

R= (2.48)
A2
where is the correction coefficient. We introduce the compactness which the measure
of the square of the perimeter ` over the area A of the cross-section geometry ,
Z 2 Z 1
`2
= = dl dr (2.49)
A

2.4.1 Elliptic cross-section


Expressing the correction coefficient as a function of the compactness we get for the circular
cross-section that () = 8 = 2.

10
2.4. HYDRAULIC RESISTANCE CHAPTER 2. THEORY

DdW

Figure 2.4: A domain . The red boundary is the perimeter l of the domain and the blue
subdomain is the area A.

For the elliptic cross-section the compactness is given by


q 2
R

0
2
d 1 (1 b2 /a2 ) sin2
= (2.50)
ab
Zq 2
16 a 2
= d 1 (1 b2 /a2 ) sin2 (2.51)
b 0
16 a
' , a b. (2.52)
b
Thus, the correction coefficient for the elliptic cross-section is given by
a2 + b2 2
() = 4 ' . (2.53)
ab 4

2.4.2 Rectangular cross-section



w
Assuming that w > h then tanh n 2h ' 1 and we get the flow rate to

P 186 w
Q' wh3 1 5 (5) . (2.54)
12 h
The correction coefficient is then given by
12 5 2
' , (2.55)
5 186(5)
where is the width-to-hight ratio w/h.
Setting the compactness = 8 + 4 + 4/ for a rectangle in the expression of we get
12 5 (/4 2 1/)2
= . (2.56)
5 (/4 2 1/) 186(5)
From the expression of it can be shown that for

lim=3 (2.57)

11
2.4. HYDRAULIC RESISTANCE CHAPTER 2. THEORY

we have the asymptotic dependence ' 3 + c. Since

2232(5)
lim ( 3) = 24 + (2.58)
5
we get the correction coefficient for the rectangular class to

= 3 16.44. (2.59)

12
Chapter 3

FEMLAB simulations

FEMLAB is a partial differential equation solver that builds on MATLAB. FEMLAB -


which stands for Finite Element Modelling LABoratory- is an advanced software package
for modeling and simulation of any physical process that can be described with partial
differential equations. The differential equation must be specified in some geometry, the
so-called computational domain, together with the boundary conditions along the surface
of the domain as well as the initial values at the starting time. The computational domain
is divided into a huge number of finite-sized elements forming the basis computation.
Having introduced the analytic description of the Poiseuille flow through a pipe with
various cross-sections, we are now interested in simulating the flow in FEMLAB for the
same cross-section geometries. The numerical calculations are then compared to those
predicted by the theory.
Before starting the simulation, a differential equation must be specified. Since the
2
equation expressing the velocity field is of the form P = v, the the simulation differ-
ential model was chosen to be a Piossons equation in the 2D-space. To draw the geometry
in which the partial differential equation must be specified, we click on Draw Mode on
the tool bar. In this window the desired geometry can now be designed. Hereafter, the
no-slip boundary is specified by clicking on Boundary Settings under Physics. Specifying
the subdomain settings is done by clicking on Subdomain Settings under Physics. The
geometry is then refined and the PDE solved. The solution plotted, displays the velocity
field; see Figure 3.1. The velocity field of the flow has its maximum in the center of the
pipe, but begins to decrease when moving radially from the center. When in contact with
the cross-section boundary the velocity of the flow is at its minimum, fulfilling the no-slip
condition.

Since we also are interested in calculating the the hydraulic resistance, the correction
coefficient1 and the compactness is determined by making subdomain A, boundary l and
velocity field integration. The results depicting the correction coefficient versus the com-
1
We saw Rthat the hydraulic resistance was defined as P/L = Rhyd Q. This can be rewritten as
P
L
= A2 dydzvx (y, z). Since all the values are required dimensionless in FEMLAB we set the factor
A2
P/L = 1 giving us the correction coefficient: = R .
dydzvx (y,z)

13
CHAPTER 3. FEMLAB SIMULATIONS

Figure 3.1: A diagram showing the simulation results of the velocity field for a circular, elliptical,
rectangular and triangular cross-section.

pactness for various cross-sections is shown in Figure 3.2. It emerges clearly from Figure
3.2 that the relation between the correction coefficient and the compactness is linear for
the three cross-section classes: circular, elliptic and rectangular. The numeric propor-
tionality factor for all cross-section groups is in good agreement with those predicted by
theory. For the elliptic it is 2.55, close to the theoretical value of 2.47 The deviation is due
to the assumption: a b, made in our calculation. For the rectangular class, the value of
the proportionality factor is found to be 3. This value agree completely with the theory.
Since the analytic calculations are not quite possible for triangle classes, only simulation
was made. The results for this cross-section class is also shown in Figure 3.2. Also here,
the relation between and is linear. The proportionality factor is less than the other
cross-section classes, namely 1.46. For cross-section classes more complicated than the
elliptical (circular), increases for small values of , leading to an increase of the hydraulic
resistance.

14
CHAPTER 3. FEMLAB SIMULATIONS

200
Correction coefficient vs. Compactness

Theory for rectangular class.


Theory for ellipse class.
Linear fit to triangular class.
150

100
a
For circular

50

0
0 20 40 60
2
l /A

Figure 3.2: Correction coefficient versus compactness for the elliptical, rectangular and triangular
class. The result for the circle is also indicated.

15
Chapter 4

Conclusion

We have demonstrated a close to linear relation between the hydraulic resistance and the
compactness for the following cross-section geometries: circular, elliptic, triangular and
rectangular. The results obtained were in good agreement with theory.
For geometries more complicated than the circular the resistance increases, since a big
part of the fluid is in contact with the inner surface of the channel.

16

You might also like