0 = f ( 1 , 2 ,..., n r )
where the k are (n - r) dimensionless variables of the form
k = v1 k 1 v 2 k 3
Lv
kn
n
Units
We have to pick m fundamental units Fi - examples of these are
length L, time T, and mass M.
We can write one of our variables vj as: v j = j [v j ]
m
[v j ] = Fi
Then write
aij
i =1
F
i =1
xj
= 1 x1 = x 2 =
L= x
= 0.
L a
M O M
a
L a
a11
A =
m1
mn
1n
Dimensionless quantities
What are the dimensionless combinations of our variables?
[ k ] = [v1
k1
v 2 k 3
Lv ] = 1
kn
a
m
i1
k1
+ ai 2 k 2 +
L+ a
in
kn
=0
A k = 0
Buckingham Pi theorem
Suppose we can write a physical relation as (v1 ,
L, v ) = 0
n
1 = g ( 2 ,..., n r )
, 0, m, l, g, t.
Thats 6 variables.
How many dimensions do we need to worry about?
[]
dimensionless
[0]
dimensionless
[m]
mass
[l]
length
[g]
length/time2
[t]
time
three dimensions.
Simple pendulum
So, we should be able to find 3 independent dimensionless parameters.
Note that m is the only variable to contain the mass unit. Therefore it
cannot be part of any dimensionless parameters - theres no way to
cancel out the mass dependence.
Common sense and inspection suggest the following dimensionless
parameters:
1 =
2 = 0
g 2
t
l
g
3 =
Simple pendulum
g
l
= f 0 , t 2
f ( 0 ,0 ) f 0 ,
g 2
P =0
l
We can solve this equation in principle for P, the period, and find,
in terms of another dimensionless function (0):
P = ( 0 )
g
l
So, knowing only the units involved and the fact that pendulum
motion is periodic, weve found:
The period is independent of pendulum mass, and scales as the
square root of g/l.
The period in general depends on initial position.
Caveats:
In exactly the same way that a basis set of vectors is not unique
(a new basis may be formed by an appropriate linear combination
of an old basis), our choice of dimensionless parameters is not
unique!
The dimensionless parameters must be independent.
Start with your initial variable list. If you try to use too few of
them, youll find that you cant make the right number of
dimensionless parameters.
Can go through and find dimensionless quantities by brute force,
but intuition and inspection are generally used.
pipe diameter
ML-3
ML-1T-1
LT-1
vD
e
D
DP
v 2
Reynolds number
relative roughness
(no name)
DP
= f (Re, )
v 2
2 v 2
or, playing with numerical factors, P =
f (Re, )
D
P =
Moody plot
2 v 2
f (Re, )
D
particle radius
ML-3
particle density
LT-2
gravitational acceleration
ML-3
fl
density of fluid
ML-1T-1
viscosity of fluid
LT-1
1 =
p
fl
2 =
v2
Rg
3 =
Froude number2
fl vR p
v2
= f
,
fl
Rg
fl vR
Reynolds number
v=
R 2 fl g
fl
R 2 fl g
fl
Better even than that, we know that if the two densities are equal
the particle should remain motionless!
So, knowing that what were really doing is buoyancy, we know
that g only shows up as:
g ( p fl )
The only way for this to be true is if we can write:
v=C
R 2 fl g p
fl
bomb yield
shock radius
time
M/L3
air density
Et 2
r 5
10
Reynolds number:
( v 2 d 2 )
vd
~
( v / d )(d 2 )
inertial forces
viscous forces
v2
~
Rg
v 2
gR
kinetic energy
gravitational potential energy
Mach number
v
cs
speed
speed of sound
Prandtl number
c p
=
c p
11
Summary
Dimensional consistency places significant (if subtle) constraints
on the relationships between quantities described by dimensioned
variables.
Dimensional analysis can give us an impressive amount of
information when supplemented with a little physical intuition and
insight.
Understanding dimensionless quantities and their relationships is
essential for similitude tests.
12
Viscosity
Defined by considering the stresses involved in shearing a
volume of fluid:
l
l
(l / t ) u
=
=
y
y
t
y
xz =
Viscosity
Thinking in terms of velocity fields,
u(x, y, z) = x-velocity component
shear stress, xz
u
y
=
Several types:
, t =
, t
Pseudoplastic - viscosity drops with increasing shear rate (shearthinning; e.g. mayonnaise, toothpaste)
Dilatant - viscosity increases with increasing shear rate (shearthickening; e.g. corn starch in water, silly putty)
Thixotropic - viscosity drops with time under constant shear
Rheopoxic - viscosity increases with time under constant shear (rare)
NonNewtonian fluids can be important for microfluidics! Colloids and
other suspensions (e.g. blood) are often pseudoplastic / thixotropic.
Turbulence
t
steady turbulent flow
u
laminar flow
Laminar flow
Most micro- and nano-fluidic devices are in the laminar regime.
This regime is typified by low Reynolds number.
Reynolds number:
vd
( v 2 d 2 )
~
( v / d )(d 2 )
inertial forces
viscous forces
y
x
Continuity
u + du
A + dA
x
For an incompressible fluid,
Au = ( A + dA)(u + du )
In integral form,
0=
dV + u dA
cs
t cv
( u ) = 0
Momentum equation
u + du
A + dA
x
Now consider the momentum flux in and out of the control volume.
x-momentum in left side:
A( u )u
( A + dA)( (u + du ))(u + du )
cv
(u )dV + cs ( u )u dA
Au s = ( A + dA)(u s + du s )
An example of a force acting on the fluid: hydrostatic pressure.
Assume for now that the pressure is uniform across each end.
Pressure force term:
Fs = pA ( p + dp)( A + dA) + ( p +
dp
)(dA)
2
Fs + Fb = u s ( u s A) + (u s + du s )( (u s + du s )( A + dA))
Recall,
Au s = ( A + dA)(u s + du s )
1
1
dpdA gAdz gdAdz = u s Adu s
2
2
Bernoulli equation
Regrouping terms, we get
or
dp
+ gdz + d u s2 = 0
2
dp
gdz = u s du s
+ gdz + u s2 = const
Eulerian analysis
Instead of a fixed volume, consider following an individual fluid
particle as it passes through the velocity field assumed to exist
everywhere.
That particle has to obey Newtons laws, and it can be deformed
(rotated, sheared, stretched).
So, at time t have particle at position x,y,z, with a velocity u(x,y,z,t).
An instant later, its velocity should be given by
u( x + x, y + y , z + z , t + t )
Eulerian analysis
If we call the components of the velocity field u, v, w, then
du p u
Du
u
u
u
u+
v+
w+
ap =
=
Dt
dt
t
z
y
x
This kind of derivative is called a total or convective or advective
derivative.
Du
In differential form,
= u (u) + u
Dt
t
local acceleration
Convective or advective acceleration of particle
Fsx = xx +
dx dydz xx
dx dydz
2 x
+ yx
2 x
1 yx
1 yx
+
dy dxdz yx
dy dxdz
2 y
2 y
+ zx +
1 zx
1 zx
dz dxdy zx
dz dxdy
2 z
2 z
Fsx =
yy
Fsy =
zz
Fsz =
+
+
+
yx
y
xy
x
+
+
zx
dxdydz
z
zy
z
dxdydz
xz yz
dxdydz
+
x
y
xy = yx =
yz = zy =
zx = xz =
xx = p u + 2
yy
zz
2
u
3
x
2
v
= p u + 2
3
y
2
w
= p u + 2
3
z
u = 0
t
2
2
2
u u u
p
+ 2 + 2 + 2
x
y
z
x
Fy = g y
2v 2 v 2v
p
+ 2 + 2 + 2
y
y
z
x
Fz = g z
2w 2w 2w
p
+ 2 + 2 + 2
z
y
z
x
Du
= g p + 2u
Dt
10
1 p
dx 2rdr
2 x
p +
1 p
dx 2rdr
2 x
rx
1
1 rx
dr 2 r dr dx
2
2 r
rx
1
1 rx
dr 2 r + dr dx
2
2 r
p rx rx 1 (r rx )
=
+
=
r
r r
x
r
Integrating w.r.t. r:
1 p
r rx = r 2
+ c1
2 x
u
rx =
r
1 p c1
u
r +
=
r 2 x r
u=
1 2 p c1
r
+ ln r + c2
4 x
11
u=
R 2 p r
1
4 x R
Q = u (r ) 2rdr =
0
R 4 p
8 x
This kind of fully developed laminar flow is also called Poiseuille flow.
This is the flow regime most commonly encountered in long, closed
microfluidic channels.
Note: Could have written Navier-Stokes eqn. in cylindrical coords.,
and solved with appropriate boundary condition, getting same answer.
u ( y = 0) = 0,
u
y
=0
y =d
12
Boundary layer
U
u( y = 0) = 0,
u 2 u
u
=
+v
y y 2
x
u( y = ) = U
Boundary layer
(x)
5x
Ux /
0.332 U 2
Ux /
13
r 2 sin
u =
1
r sin r
2 sin 1
+ 2
r 2
r sin
2 sin 1
+ 2
r 2
r sin
u r (r = a) = 0
FD = 6Ua
14
Surface interactions
Details of the interactions between fluids and surfaces can be of
crucial importance in micro- and nanofluidic environments.
Obviously, for small volumes of material, the surface to volume
ratio can become large, leading to surface forces swamping out
body forces.
Surface interactions often determined by surface chemistry
energetic competition between liquid cohering to itself and wetting
the surface.
W = pdV
W = dA
15
4r 2
3
=
3
r
(4 / 3)r
Clearly surfaces become very expensive at small size scales.
All other things being equal, two small drops will merge to
minimize the surface area.
Similarly, small bubbles are energetically very expensive, and
will tend to collapse if pressure forces from the vapor inside
cant balance the surface tension force.
This bubble collapse can be very intense in terms of energy
density - cavitation, sonoluminescence.
Water: LV = 0.072 J/m2
Surface interactions
Contact angle determined by energetics. Consider moving the
interface slightly.
Defining solid-liquid, liquid-air, and solid-air surface tensions as
sl, lv, and sv, respectively,
sv sl = lv cos c
Surface interactions (surface tensions) can be tailored chemically.
Under certain circumstances, they can also be tuned on the fly!
16
Wetability
sv sl = lv cos c
Spreading coefficient is defined as
S sv sl lv
Wetability
So, if the interaction potential is what determines surface
affinities, we can write:
sl = sv + lv V SL
where weve defined an attractive interaction to give a positive VSL.
Similarly, two identical liquid regions that then touch have no
surface energy associated with that interface:
0 = lv + lv V LL
So, combining these we can see that the wetting coefficient is then
S = VSL V LL
This is also pretty much why teflon is effectively water-proof: the
hydrogen bonding interaction between water molecules is much
more strong than the VdW interaction between the water and the
fluoropolymer.
17
Wetability
If were really talking about Van der Waals, the potential between
two different materials is related to their polarizabilities:
VLL ~ L2
V SL ~ S L
Capillary action
The interaction between the liquid and the solid surface can be
powerful enough to draw liquid up against gravity: this is
capillary action.
g (r 2 h) = 2r sl cos c
h=
2 sl cos c
gr
18
Self-assembled monolayers
Using soft lithography /
microcontact printing with
PDMS stamps, can pattern
surfaces into hydrophilic
and hydrophobic regions.
Also, optical patterning can
be done to modify the SAM
once its in place.
At right, dye in aqueous
solution only goes where
hydrophobic surface
termination has been
removed.
Zhao et al., Science 291, 1023 (2001).
19
Temperature-dependent layer
20
21
cos (V ) cos (V = 0) +
1 0 V 2
2 lv d
22
Solution electrostatics
Now that weve introduced electrically conductive liquids as a
possibility, we have to contend with the actual electrostatic
interactions between the liquid and the surrounding materials.
23