ME 563 - Intermediate Fluid Dynamics - Su

Lecture 2 - More math, plus some basic physics

In the first lecture we went over some basic math concepts, in particular the operator ∇. We’ll
finish up with basic math by going over tensors and index notation, then talk about some basic
physical concepts.

1 Tensors and index notation

In the last lecture we considered the dot product, where two vectors result in a scalar, and the
cross product, where the two vectors yield a third vector. There is another way to multiply vectors
together that gives rise not to a scalar or a vector, but to a tensor. To deal with those it’s most
convenient to use index notation (also called tensor notation).

1.1 Index notation

Scalars and vectors are actually specific cases of tensors. A scalar is a tensor of order (or rank)
zero, and a vector is a tensor of order one. (Generally, however, if we say that something is a tensor
without specifying its order, we will mean that it is a tensor of order two.) The order of a tensor
tells you the number of indices necessary to describe it. In n-dimensional space, a tensor of order
m has nm components.
We can repeat some of the results of the last lecture using index notation. Consider a vector f
defined in three-dimensional space. Let the three orthogonal unit vectors be ê1 , ê2 , and ê3 . (We’re
not using x, y and z because the coordinate system is not necessarily Cartesian, or it could be
rotated from x, y and z, etc.) Then we can write f as

f = f1 ê1 + f2 ê2 + f3 ê3 = fi êi (1)

This expression illustrates two key aspects of index notation:

• An index (in this case i) takes on values corresponding to the number of dimensions in the
space being considered.

• If an index is repeated in a term, then that term is summed over that index (this is the
summation convention, which physicists call the Einstein summation convention).

The dot product and cross product of two vectors can be expressed conveniently in index
notation, with the aid of two new operators. The dot product of two vectors a and b is
(
1 if i = j
a · b = ai bj δij , where δij = (2)
0 if i 6= j.

This δij is called the Kronecker delta. (Of course, we could also have written a · b = ai bi using the
summation convention.) The cross product of a and b is


1 if ijk is cyclic, i.e. 123, 231, or 312
a × b = ijk aj bk êi , where ijk = −1 if ijk is anti-cyclic, i.e. 321, 213, 132 (3)


0 if any of the two indices are identical.

The ijk is called the permutation tensor or permutation operator.

1
1.2 Tensors
Consider two vectors a = a1 ê1 + a2 ê2 + a3 ê3 = ai êi and b = b1 ê1 + b2 ê2 + b3 ê3 = bi êi . If we multiply
them together, not as a dot product or cross product, but just ab, we get ab = T, or
 
a1 b1 a1 b2 a1 b3
ab = ai bj = a2 b1 a2 b2 a2 b3  = Tij = T (4)
a3 b1 a3 b2 a3 b3
where, by convention, the first index, here i, represents the row of the tensor and the second index,
j, represents the column. Also, a quick notational point. The order of a tensor is sometimes
identified by underlining. So a vector a is often written a, and a tensor T is often written T .
The underlining convention is common with written work; in printed work, vectors and tensors are
usually just represented in boldface (some authors use lowercase bold for vectors and uppercase
bold for tensors, but this isn’t universal).
Because the first index in a tensor represents the row, and a vector is a tensor of order one, the
vector a can be written
 
a1
a = ai êi = a2  .
 (5)
a3
The matrix representation of a is not written with the unit vectors êi because in tensor form, the
unit vectors are assumed to go with the components i (this can sometimes be confusing).
The transpose of a tensor is a tensor with its rows and columns switched. For a second-order
tensor T, this means the transpose TT is given by
   
T11 T12 T13 T11 T21 T31
for T = Tij = T21 T22 T23  , the transpose is TT = Tji = T12 T22 T32  . (6)
T31 T32 T33 T13 T23 T33

In the expression for the transpose, when we write TT = Tji , the first index still corresponds to the
row even though we use j instead of i. A tensor T is called symmetric if T = TT , i.e. if Tij = Tji .
A tensor T is anti-symmetric if T = −TT , or Tij = −Tji .
For a vector a, taking the transpose just means
 
a1

for a = a2  , the transpose is aT = a1 a2 a3 . (7)
a3
The standard rules of matrix multiplication can be applied to tensors. Matrix multiplication
corresponds to the dot product. It is possible to take the dot product of a tensor T and a vector
a–
    
T11 T12 T13 a1 T11 a1 + T12 a2 + T13 a3
T · a = Tij aj = T21 T22 T23  a2  = T21 a1 + T22 a2 + T23 a3  (8)
T31 T32 T33 a3 T31 a1 + T32 a2 + T33 a3

(Notice the application of the summation convention.) The dot product in this case doesn’t yield
a scalar, but instead a vector. The dot product essentially gives a result one order reduced from
the highest order tensor in the product. This leads to the other form of multiplication we’ll be
interested in, the double-dot product (also known as the scalar product for tensors). This takes
two second-order tensors and yields a scalar, as

T : T = Tij Tij . (9)

2
1.3 Velocity gradient tensor
The tensor we will be most interested in is the velocity gradient tensor, ∇u. For Cartesian coordi-
nates, with u = uêx + vêy + wêz , ∇u is defined as
 ∂u ∂v ∂w 
∂uj ∂x ∂x ∂x
∇u = =  ∂u
∂y
∂v
∂y
∂w 
∂y (10)
∂xi ∂u ∂v ∂w
∂z ∂z ∂z

The Laplacian of the velocity, ∇2 u, will also appear often in our discussions. This is found by
 ∂u ∂v ∂w 
  ∂x ∂x ∂x
∇2 u = ∇ · (∇u) = ∂ ∂ ∂  ∂u ∂v ∂w 
∂x ∂y ∂z ∂y ∂y ∂y
∂u ∂v ∂w
∂z ∂z ∂z
 ∂2u 2 ∂2u

∂x2
+ ∂∂yu2 + ∂z 2
 ∂2v ∂2v ∂2v 
=  ∂x2 + ∂y2 + ∂z 2 
(11)
∂2w 2 ∂2w
∂x2
+ ∂∂yw2 + ∂z 2
 2 
∇ u
=  ∇2 v 
∇2 w

so taking the Laplacian of the velocity vector is equivalent to taking the Laplacian of each of the
components.
(We’ll talk about the physical meaning of ∇u in detail later.)

2 Physical concepts
Let’s now discuss some basic physical ideas. One of the goals of this class is to develop the ability
to think through problems intuitively before having to bring math tools to bear. Suppose I have
two airplanes. They are identical except one of them has winglets on the wingtips, like you see on
some airliners (747’s, A340’s etc.).

Figure 1: Two wings: wing 1, no winglet, wing 2, winglet.

The first wing shows a tip vortex, while the second doesn’t.

Figure 2: Tip vortex on wing 1.

Why does this tip vortex arise? First, we know that wings are designed to provide elevated
pressures on the lower surface, and reduced pressures on the upper surface. However, if we pick a
point just outside of the wingtip, the pressure has to be the same from above and below, because

3
the pressure can’t have a discontinuity. As a result, the pressure on the underside of the wing is
gradually reduced as we approach the tip, and the pressure above the wing is gradually increased
as we approach the tip, so the pressures become equal outside the tip. Because fluids move in the

Figure 3: Pressure distribution around a wing.

direction of decreasing pressure, the airflow on the lower surface of the wing moves toward the tip,
and the airflow on the lower surface moves toward the plane’s fuselage. Put together, the result
is a tip vortex with the sense of rotation shown in Fig. 2. (The winglet in wing 2 is designed to
prevent the formation of this vortex.) Basically, two principles (and no math) were necessary to
show this - the idea that pressure fields are continuous, and that flows are driven in the direction
of decreasing pressure.
Why, though, would a designer want to get rid of the tip vortex? Let’s consider what effect
the tip vortex has by imagining that we’re stationary in space, and can somehow see the velocity
of the air. A plane equipped with winglets passes, leaving the air essentially undisturbed. A plane
without winglets passes, and in contrast we see the air spinning off the wingtips. The subsequent
reasoning goes like this:

• The flow with the vortical structures has more kinetic energy than the undisturbed flow.

• The kinetic energy has to be provided by the plane’s engines.

• The energy provided by the engines to the vortices is energy not provided to the plane’s
forward motion.

• The vortices thus correlate with an effective drag force on the plane.

We’ve thus shown that the tip vortices cause drag, and we needed to apply only the conservation
of energy, again with no math. Unfortunately...not all problems can be solved with just physical
intuition, but physical intuition is still very handy.