Kinematics - 2

Harish N Dixit
ME5310: Incompressible Fluid Flow
Department of Mechanical & Aerospace Engineering
Indian Institute of Technology Hyderabad
www.iith.ac.in/~hdixit

1 Analysis of relative motion near a point

The force exerted by one portion of a fluid on another depends on the way the fluid deformed by the motion.

Let u(x, t) be the velocity at position x and time t. Then the velocity at a neighbouring point (x + r) is
u + δu. Using Taylor series expansion about x, we get
∂ui
ui (x + r) = ui (x) + rj + O(r2 ). (1)
∂xj

Therefore

δui = ui (x + r) − ui (x), (2)

∂ui
= rj , we assume linearity in δui vs r. (3)
∂xj

We can decompose the velocity gradient tensor into a symmetric, S, and antisymmetric, ω, part.
   
∂ui 1 ∂ui ∂uj 1 ∂ui ∂uj
= + + − , (4)
∂rj 2 ∂xj ∂xi 2 ∂xj ∂xi
= Sij + Ωij . (5)

Here Sij is the rate-of-strain tensor and Ωij is the vorticity tensor.

Therefore, the relative velocity becomes

δui = rj Sij + rj Ωij , (6)

(s) (a)
= δui + δui . (7)

The two contributions are distinct and we will examine each one of them carefully.

(s)
2 Symmetric contribution, δui
Sij can be further decomposed into a diagnol tensor (with non-zero trace) and a symmetric traceless
tensor. Hence  
1 1
Sij = Skk δij + Sij − Skk δij . (8)
3 3
Hence we have1
∂Φ
rj Sij = , (9)
∂xi

1
 1
where Φ1 = rk rl Skl . Note that Φ = constant form a family of quadrics (look up conic sections). Now
2
∂Φ
gives the normal direction to the quadrics. Thus, the existence of such a scalar potential implies that
∂xi
the corresponding velocity is directed normal to the constant Φ surfaces.
The nature of δu(s) contribution to δu becomes clearer if we choose the orthogonal axes of reference in
such a way that the off-diagnol elements of Sij become zero. The axes of reference then coincide with the
principal axes of Sij and the family of quadrics becomes

1  02 2 2

Φ= ar1 + br20 + cr30 , (15)
2
where ri0 , i = 1, 2, 3 are the components of r in the principal direction and a, b, c are the diagonal compo-
nents of the tensor Sij 0 such that

0 ∂rk ∂rl
Sij = Skl . (16)
∂x0i ∂x0j
Hence  
a 0 0
S0 =  0 b 0  .
0 0 c
∂ui
Note that a + b + c = Sii0 = Sii = .
∂xi
The contribution δu(s) to the relative velocity therefore has three components, (ar10 , br20 , cr30 ) with reference
to the new axes.
0 ∂u1 ∆u1
Take the case of a line element initially parallel to x1 axis. We know that S11 = ≈ . Consider
∂x1 ∆x1
two points initially on this line element separated by a distance ∆x1 . We therefore have

1
Check: We assumed δu to be linear in r, i.e., we neglected O(r2 ) terms in δu. Hence
 
∂Φ 1 ∂rk ∂rl ∂Skl
= rl Skl + rk Skl + rk rl , (10)
∂xi 2 ∂xi ∂xi ∂xi
1
rl Sil + rk Ski + O(r2 ) ,


= (11)
2
1
= (rl Sil + rk Sik ) , (12)
2
1
= (2rk Sik ) , since l and k are dummy variables, (13)
2
= rj Sij . (14)

2
 ∆u1 = u1 (x1 + ∆x1 ) − u1 (x1 ), (17)
∂u1
= ∆x1 , (18)
∂x1
0
= ∆x1 S11 , (19)
= a∆x1 . (20)
If a > 0, then the velocity at (x1 + ∆x1 is higher than the velocity at x1 . So if u1 (x1 ) is positive, then the
relative velocity contribution is such that the point (x1 + ∆x1 ) moves further away from x1 thus stretching
the line element. The rate of this stretching depends on the magnitude of a.
The contribution of δu(s) is said to represent pure straining motion.

Extending the above argument to all the three directions, we say

say that a cube of each side equal to unit length stretches into
a cuboid with sides of lengths a, b, c respectively. The volume
change in this operation is then equal to 1 − abc. For an in-
compressible fluid, since ∆ · u = Sii = 0, a + b + c = 0. In
such a case, pure straining motion is a volume preserving opera-
tion.

The same pure straining motion also extends a sphere of unit radius into
an ellipsoid of minor axes of lengths a, b, c. A classic example of pure
straining motion is an extensional flow. Experimentally, such a flow is generated with a four roll mill.
This is how stretching experiments on bubbles, drops and polymers are carried out in reality.

It is customary to distinguish between straining motions which result in

volume changes and those that don’t result in any volume change. The
latter part would be identical to the case of an incompressible fluid. So,
for a general compressible fluid, we can write
Φ = Isotropic expansion accommodating all volume changes
+ pure straining motion without change in volume. (21)
This is accomplished by decomposing Sij into an isotropic part with a
non-zero trace and a traceless part, i.e.,
 
1 1
Sij = Skk δij + Sij − Skk δij . (22)
3 3
We therefore have  
1 2 1 1
Φ = Skk r + Skl − Smm δkl rk rl . (23)
6 2 3
It is important to remember that in the above analysis, we have analyzed the straining motions only in
the principal directions. We achieved this by tilting the coordinate axes to coincide with the principal
directions. As a result, only the diagonal entries of Sij survived. To analyze the role of off-diagonal entries,
we will need a a more careful analysis which will be postponed to a later section. But at this point, it
suffices to state that Sij contributes to a pure straining motion with or without volume change.

(a)
3 Anti-symmetric contribution, δui
We see that Ωij is anti-symmetrical. We can therefore write
1
Ωij = − ijk ωk . (24)
2

3
The negative −1/2 is there for convenience. We therefore have

(a) 1
δui = rj Ωij = − ijk rj ωk , (25)
2
1
= (ω × r)i . (26)
2
(a)
Clearly, δui is the tangential velocity produced at a point with position vector r
resulting in a solid-body rotation with angular velocity ω/2.
Comparing the components of Ωij with the last expression, we can write the cor-
responding components of ω.
∂u3 ∂u2
ω1 = − , (27)
∂x2 ∂x3
∂u1 ∂u3
ω2 = − , (28)
∂x3 ∂x1
∂u2 ∂u1
ω3 = − . (29)
∂x1 ∂x2
The vector ω is called the local vorticity. In symbolic notation, we can also write

ω = ∇ × u. (30)

If ∇ × u = 0 everywhere in the flow, we refer to such a flow as an irrotational flow since the local rotation
vanishes at every point. We will return to this point later.

3.1 Additional physical interpretation

It is easy to see why ∇ × u appears as twice the local angular velocity of the fluid. By using Kelvin-Stokes
theorem (also simply stated as the Stokes theorem), we have
Z I
(∇ × u) · ndA = u · dr. (31)

Let the area be a small circle of radius a. We therefore have

I
1
Tangential velocity averaged over circumference = u · dr, (32)
2πa
I
1
= a u · dr, (33)
2πa2
1
= a (∇ × u) · n. (34)
2
In the last expression, we have assumed (∇ × u) to be constant since it is reasonable to assume that the
vorticity field is constant over a very tiny area of size πa2 .

But tangential velocity at the edge of a circle of radius a is given by ut = a × (angular velocity). Hence
1
angular velocity = (∇ × u). (35)
2

4
4 Summary
In summary, we have seen that, to first order in linear dimensions of a small region surrounding the
position x, the velocity in this region consists, in effect, of a superposition of

1. a uniform translation with velocity u(x),

2. a pure straining motion characterized by a rate-of-strain tensor, Sij , which itself can be further
decomposed into an isotropic expansion and a straining motion without change in volume,
ω
3. a rigid-body rotation with angular velocity .
2
In analytical terms, the conclusion is that the velocity at the position (x + r) may be written approximately
as  
∂ 1 1
ui (x + r) = ui (x) + rj rk Sjk + ijk ωj rk + O(r2 ), (36)
∂xi 2 2
or in symbolic notation as
∂Φ 1
u(x + r) = u(x) + + (ω × r) + O(r2 ), (37)
∂x 2
where Sij and ωj are evaluated at the point x.

5 Vorticity and vortex

We defined the local vorticity of a fluid as ω = ∇ × u. We briefly discuss the difference between vorticity
and vortex here by considering three different flow cases.

5.1 Unidirectional shear flow

Consider a simple viscous flow through a channel of width 2h. The velocity is given by

y2
   
[u, v, w] = u0 1 − 2 , 0, 0 . (38)
h

5
The only non-zero velocity component is
∂v ∂u
ωz = − , (39)
∂x ∂y
 
−2y
= 0 − u0 , (40)
h2
2u0 y
= . (41)
h h
Therefore |ωz | is maximum at y = ±h and ωz = 0 at y = 0. Here the streamlines are straight, yet vorticity
is non-zero.

5.2 Stagnation point flow

Now let us consider another flow as shown schematically in figure XXXX. The velocity components are
given by
(u, v, w) = (cx, −cy, 0), (42)
where c is an arbitrary positive constant. It is easy to verify that all the
vorticity components are zero, i.e.,

ωx = ωy = ωz = 0. (43)

The streamlines in this flow are curved, yet the vorticity is zero. The
present example clearly demonstrates that vorticity is not directly con-
nected with the curvature of streamlines.

5.3 Point vortex

A point vortex is an irrotational vortex with vorticity at just one single
point in the entire flow. The velocity components in the cylindrical polar
coordinates are given by
 
Γ
(ur , uθ , uz ) = 0, , 0 . (44)
r

The z component of vorticity in cylindrical coordinates is given by

1 ∂(ruθ ) 1 ∂ur
ωz = − , (45)
r ∂r r ∂θ
1 ∂Γ
= = 0, except at r = 0. (46)
r ∂r
Vorticity is identically zero everywhere except at r = 0 where the derivative operation is not well defined.
This is an example of a vortex with zero vorticity, and a flow with closed streamlines with zero vorticity.
We will return to the subject of irrotational vortex later when dealing with the concept of circulation.

6 Decomposition of straining motion

In the previous sections, we decomposed the total deformation into a translation, a pure straining motion
and a solid-body rotation. The pure straining motion was analyzed by rotating the coordinate axes to
coincide with the principal axes such that only the diagonal entries of Sij survive.

6
  
1 ∂ui ∂uj
The rate-of-strain tensor2 , Sij = + , is a second order tensor. Written in a matrix form, we
2 ∂xj ∂xi
have  
S11 S12 S13
S =  S12 S22 S23  .
S13 S23 S33
We will show that the diagonal entries in S are associated with extensional strains and off-diagonal extras
are associated with shear strains. We know that the straining component of relative velocity is given by
(s)
δui = rj Sij . (47)

Let r = αds where α is a unit vector in the r

direction and ds is the length. Hence
(s)
δui = αj Sij ds = di ds, (48)

where d = α · S is the strain vector as shown

schematically in the figure.
For a point P 0 in the αi direction from P , the vec-
tor di is the strain rate of P 0 with respect to P .
It indicates the direction and velocity with which
P 0 moves away from P . Clearly, as shown in the
figure, di is not necessarily along αi . As suggested
by vector decomposition, we can write
(s) (es) (ss)
δui = δui + δui . (49)

The first term on RHS is the extensional strain which denotes deformation in the αi direction, whereas
the second term is the shearing strain which denotes deformation perpendicular to the αi direction.

6.1 Extensional strain

The extensional strain is proportional to α · d pointing in the α direction. This can be written as

δu(es) = α(α · d)ds, (50)

(es)
δui = αi αj dj ds. (51)

For example, let P be at the origin and P 0 be on the x axis at a distance ds. The vector r = αds is
therefore pointing in the direction of x axis and we have

α1 = 1, α2 = α3 = 0.

Hence
(es)
δu1 = d1 ds,
= (α · S)1 ds,
= S11 ds. (52)

Therefore S11 is the extension rate of two particles separated in the x1 direction.
(es)
Similarly, it can be shown that δu2 = α2 (αj dj )ds = 0.
2
Sij is also referred to as the rate-of-deformation tensor. In literature, this tensor is also referred to with other symbols,
εij or γ̇ij .

7
6.2 Shear strain
The shear strain is the component of strain vector, di perpendicular to the αi direction. This can be
accomplished by the cross product of α and (α × d). Therefore, we have

δu(ss) = (α × d) × α ds. (53)

Alternately, we can also calculate the shear strain as

δu(ss) = δu(s) − δu(es) . (54)

If P = (0, 0, 0) and P 0 = (ds, 0, 0), we have α = (1, 0, 0). Hence

i j k
α×d= 1 0 0 = −j d3 + k d2 .
d1 d2 d3
and
i j k
(α × d) × α = 0 −d3 d2 = j d2 + k d3 .
1 0 0
Moreover, since di = αj Sij , we have
X
d1 = αj S1j = S11 , (55)
j
X
d2 = αj S2j = S12 . (56)
j

Putting everything together, we have

(ss)
δu1 = 0, (since there is no i-component in (α × d) × α), (57)
(ss)
δu2 = d2 = S12 , (58)
(ss)
δu3 = d3 = S13 . (59)

Clearly, there is no relative velocity in the x1 -direction. S12 gives the shearing velocity in the x2 -direction
of a particle P 0 which is originally separated from P only in the x1 -direction. Figure XXXXX

In general, the off-diagonal entries of Sij give the shearing velocity in the j-direction for two particles
initially separated in the i-direction.

7 Simple Shear Flow

To see the application of all the previous analysis, let us consider a standard flow and determine the
components of the relative velocity in the neighborhood of single point, P . Let the velocity components
be given by

(u1 , u2 , u3 ) = (cx2 , 0, 0), (60)

where c > 0 determines the velocity gradient of the shear flow. This flow is called the simple shear flow.
The rate-of-strain tensor becomes  
0 c/2 0
S =  c/2 0 0  .
0 0 0

8
Using di = αj Sij , we have
c
d1 = α1 S11 + α2 S12 + α3 S13 = α2 , (61)
2
c
d2 = α1 S21 + α2 S22 + α3 S33 = α1 , (62)
2
d3 = α1 S31 + α2 S32 + α3 S33 = 0. (63)
Vorticity, ω = (0, 0, −c).

Now we are in a position to calculate the elementary motions of P 0 with respect to P . Specifically, we
consider five points around P = (0, 0, 0) whose vertices and α components are given below:
At 1: P0 = (1, 0, 0); (α1 , α2 , α3 ) = (1, 0, 0),
At 2: P0 = ( √12 , √12 , 0); (α1 , α2 , α3 ) = ( √12 , √12 , 0),
At 3: P0 = (0, 1, 0); (α1 , α2 , α3 ) = (0, 1, 0),
At 4: P0 = (− √12 , √12 , 0); (α1 , α2 , α3 ) = (− √12 , √12 , 0),
At 5: P0 = (−1, 0, 0); (α1 , α2 , α3 ) = (−1, 0, 0).
.

7.1 The velocity due to solid-body rotation

(a) 1
δui = − ijk rj ωk , (64)
2
1
= − ijk αj ωk ds, (65)
2
1
= − ij3 αj ω3 ds (since only 3-component of ω is non-zero.) (66)
2
Hence
(a) 1
δu1 = − 123 α2 ω3 ds, (67)
2
1
= − α2 (−c)ds, (68)
2
c
= α2 ds. (69)
2
(a) c
δu2 = −α1 ds, (70)
2
(a)
δu3 = 0. (71)

9
At each of the points P 0 , we have,
(a) (a)
At 1: δu1 = 0, δu2 = − 2c ds,
(a) (a)
At 2: δu1 = 2√c 2 ds, δu2 = − 2√c 2 ds,
(a) (a)
At 3: δu1 = 2c ds, δu2 = 0,
(a) (a)
At 4: δu1 = 2√c 2 ds, δu2 = 2√c 2 ds,
(a) (a)
At 5: δu1 = 0, δu2 = 2c ds.
.
The relative velocity at points around 0 is shown schematically in figure below.

7.2 The velocity due to straining (deforming) motion

The relative velocity due to straining is given by
(s)
δui = rj Sij = αj dsSij = αj Sij ds = di ds. (72)

Hence
(s) c
δu1 = d1 ds = α2 ds, (73)
2
(s) c
δu2 = d2 ds = α1 ds, (74)
2
(s)
δu3 = d3 ds = 0. (75)

We can further decompose the straining motion into an extensional part and a shear part.

7.2.1 Extensional strain

Recall that
(es)
δui = αi αj dj ds. (76)
Hence
(es)
X
δu1 = α1 αj dj ds = α12 α2 c ds, (77)
j
(es)
X
δu2 = α2 αj dj ds = α1 α22 c ds, (78)
j
(es)
X
δu3 = α3 αj dj ds = 0. (79)
j

At each of the points P 0 , we have,

(es) (es)
At 1: δu1 = 0, δu2 = 0,
(es) (es)
At 2: δu1 = 2√c 2 ds, δu2 = 2√c 2 ds,
(es) (es)
At 3: δu1 = 0, δu2 = 0,
(es) (es)
At 4: δu1 = 2√c 2 ds, δu2 = − 2√c 2 ds,
(es) (es)
At 5: δu1 = 0, δu2 = 0.
.
The relative velocity at points around 0 is shown schematically in figure below.

10
7.2.2 Shear strain
The shear strain can be obtained by removing the extensional part of the total strain:
(ss) (s) (es)
δui = δui − δui . (80)

Therefore,
 
(ss) 1 2
δu1 = − α1 α2 c ds, (81)
2
 
(ss) 1 2
δu2 = − α2 α1 c ds, (82)
2
(ss)
δu3 = 0. (83)
√
√ happens at the coordinate axis, i.e. when α1 = 0 or α2 = 0. When both α1 = ±1/ 2◦
Maximum shearing
and α2 = ±1/ 2, shearing motion is zero. This forms a set of axis (the principal axes) rotated at 45
from the coordinate axes.

At each of the points P 0 , we have,

(ss) (ss)
At 1: δu1 = 0, δu2 = 2c ds,
(ss) (ss)
At 2: δu1 = 0, δu2 = 0,
(ss) (ss)
At 3: δu1 = 2c ds, δu2 = 0,
(ss) (ss)
At 4: δu1 = 0, δu2 = 0,
(ss) (ss)
At 5: δu1 = 0, δu2 = − 2c ds.
.
The relative velocity at points around 0 is shown schematically in figure.

7.2.3 Principal axes

The principal axes of Sij are defined such that only the diagonal entries of Sij survive. The contribution
of relative motion with such a rate-of-strain tensor suppresses shear strain contribution as we have already

11
seen that the shear strain contribution arises from the off-diagonal entries of Sij . Clearly, the shear strain
contribution goes to zero at points 2 and 4. At both these points, the total strain is identical to the
extensional strain. Therefore, we expect that principal axes to coincide with directions 02 ~ and 04.
~
To determine the principal axes of Sij , we first calculate the eigenvalues and then its eigenvectors. The
characteristic equation of Sij is given by

λ3 − I (1) λ2 − I (2) λ − I (3) = 0, (84)

where I (1) , I (2) and I (3) are the invariants of matrix Sij .

I (1) = tr(S) = Sii , (85)

1  1
I (2) = tr(S2 ) − (tr(S))2 = (Sij Sji − Sii Sjj ), (86)
2 2
I (3) = Det[S]. (87)

In the case of simple shear flow, S12 = S21 = c/2 and all other entries are zeros. The invariants then take
the values
c2
I (1) = I (3) = 0, I (2) = . (88)
4
The characteristic equation becomes  2
3 c
λ − λ = 0, (89)
4
which gives
c
λ=0 or λ = ± . (90)
2
The first eigenvalue is redundant as the base flow is two-dimensional. The eigenvectors, β, are obtained
from the equation
βi (Sij − λδij ) = 0. (91)
Inserting λ = c/2, we get
c (1) c (1)
− β1 + β2 = 0, (92)
2 2
   
which is satisfied with β (1) = √12 , √12 . Similarly, the second eigenvector becomes β (2) = − √12 , √12 .
Both these vectors in the ±45◦ directions. These principal directions are consistent with the directions of
maximum and minimum extensional strain together with zero shear strain.

12