You are on page 1of 6

MENG 463 / ENAS 704

Fall 2010

Problem Set 1 Solutions


1. Suppose that Im floating in a boat on a lake carrying a brick. If I throw the brick overboard,
will the level of the lake change, and if so, how?
Solution:
When the brick is on the bottom of the lake, it displaces its own volume of water: Vbrick =
bottom . When it is the boat, however, it displaces its weight of water, by Archimedess
Vwater
Principleso
boat
gwater Vwater
= gbrick Vbrick .
Rearranging, we have
bottom
Vwater
=

water boat
V
.
brick water

Since this density ratio is less than unity, the volume of the displaced water is larger when
the brick is in the boat, and so when I throw it overboard the level of the lake will decrease.
2. Consider a scalar field (x, t) and a velocity field u(x, t) defined in an inertial reference frame.
Now suppose that we want to work in the non-inertial frame where
X x + (t)
T t
In this new frame, the velocity and scalar fields are given by U(X, t) and (X, t).
(a) Express and U in terms of and u.
(b) What is the rate of change following a fluid element in the (X, T ) frame? That is, what
is the relationship between the material derivatives
D

+u
Dt
t
and

+ U X ?
DT
T

Comment on your result.


(c) Write the acceleration of a fluid element a = Du/Dt in the (X, T ) frame.
Solution:
(a) Scalars are by definition independent of reference frame. Therefore
(x, t) = (X, T ).
To transform the velocity, use its definition:
u(x, t) =

dx
dt

U(X, T ) =

dX
dT

We also have that


dX = dx +

d
dt.
dt

Therefore,
U(X, T ) = u(x, t) +

d
.
dt

(b) Consider the rate of change of the scalar field along a fluid element trajectory. The total
derivatives are



d
d =
dT + dX X =
dT + dx +
dt X
T
T
dt
and

dt + dx .
t
But since is a scalar, we must have d = d. Therefore,



d
dt + dx =
dT + dx +
dt X ,
t
T
dt
d =

or

D
=
Dt







d
D
+u =
+ u+
X =
.
t
T
dt
DT

We see that the material derivatives (acting on appropriately written fields) are the
same, as they must be: the rate of change along the trajectory of a fluid element is
independent of reference frame.
(c) Were looking for the acceleration written in the (X, T ) frame, namely A(X, T ). We can
use our result from part (a):


DU
D
d
d2
A=
=
u+
=a+ 2,
DT
Dt
dt
dt
since = 0.
3. Show that ij is an isotropic second-rank tensor; that is, ij transforms as a tensor and
0 for any coordinate systems E and E 0 related by rotations or reflections.
ij = ij
i e
j . Therefore, we have
Solution: In general, ij = e
0
0i e
0j
ij
=e

m ) (Cnj e
n )
= (Cmi e
= Cmi Cnj mn ,
and so ij transforms as a tensor. We can also write
0
ij
= Cmi Cnj mn = Cmi Cni = ij ,

and so ij is an isotropic tensor whose components are independent of coordinate system. It


turns out that ij is in fact the only second-rank isotropic tensor.

4. Consider two second-rank tensors Aij and Bij . Suppose that Aij = Aji and Bij = Bji ;
then Aij is called symmetric and Bij is antisymmetric. Show that Aij Bij = 0. Note that
this is a general result regardless of tensor rank: the contraction of a symmetric and and
antisymmetric tensor vanishes.
Solution:
Aij Bij = Aji Bij

(Aij is symmetric)

= Aji Bji

(Bij is antisymmetric)

= Aij Bij

(Relabeling dummy indices)

=0

(Only scalar that is its own negative!)

5. Let us explore some properties of the Levi-Civita symbol ijk .


(a) Consider two coordinate systems E and E 0 that are related by a reflection through each
i =
axis; that is, e
e0i for i = 1, 2, 3. Use these two coordinate systems to show that ijk
is not a third-rank tensor.
(b) Prove the epsilondelta rule: ijk `mk = i` jm im j` . Explain why the right-hand
side is a proper tensor while ijk is a pseudotensor.
(c) Show that ijk ijk = 6.
Solution:
(a) For this transformation, the direction cosines are
i e
0j = e
i (
Cij = e
e0j ) = ij .
Suppose that ijk were a third-rank tensor. We would then expect
0ijk = C`i Cmj Cnk `mn = li mj nk `mn = ijk .
But we should have 0ijk = ijk , since the ordering of i, j, k has not changed under the
reflection: Recall that the definition of ijk is

if i, j, k is a cyclic permutation of 1, 2, 3
1
1 if i, j, k is an anticyclic permutation of 1, 2, 3
ijk =

0
if any index is repeated
Nothing in this definition says anything about a reference frame; only the ordering of
the indices matters.
We thus have a contradiction. The only assumption we have made is that ijk is a tensor.
Therefore, ijk must not be a tensor, since it transforms differently.
With a little more work (noticing that we can write the determinant of matrix as
det(A) = ijk A1i A2j A3k and that the determinant of a direction-cosine matrix is (1)r
where r is the number of coordinate reflections), you can show the correct transformation
rule for the Levi-Civita symbol is
0ijk = det(C)C`i Cmj Cnk `mn .
This is the general transformation rule for pseudotensors.

(b) Consider the possible combinations of i, j, `, m:


If i = j, then iik `mk = 0, since an index is repeated. The same is true if ` = m.
If i = `, ijk imk is zero if j 6= m and 1 if j = m; so ijk imk = jm . Similarly, if
j = m, ijk `jk = i` .
If i = m, we have ijk `ik = ijk i`k = jl , where we have used our result from
above and the antisymmetry of the Levi-Civita symbol under acyclic permutations.
Similarly, if j = `, ijk jmk = im .
If i 6= ` or m, i = k and the sum vanishes due to the repeated index. The same is
true if j 6= ` or m.
Combining these results, we have that

if i = ` and j = m
1
1 if i = m and j = `
ijk `mk =

0
else
Alternatively, we can express this relation as
ijk `mk = i` jm im j` .
The reason that the right-hand side is a proper tensor is that there is a product of
Levi-Civita symbols here: since ijk only differs from a proper tensor under coordinate
reflections, where it picks up a negative sign, products of an even number of Levi-Civita
symbols will be proper tensors.
(c) Using the epsilondelta rule proved above, we have
ijk ijk = ii jj ij ij = ii jj ii .
Since ii = 3 (in three dimensions), we then have
ijk ijk = 9 3 = 6.
6. A very useful quantity is fluid mechanics is the vorticity, given by = u, or
k = ijk

uj
.
xi

Note that the vorticity is a pseudovector. We can, however, express the vorticity in terms of
a proper tensor.
(a) Write the vorticity in terms of the rate-of-rotation tensor


uj
1 ui
ij =

.
2 xj
xi
Note that it is a general result that a pseudovector may be related to an antisymmetric
(i.e., ij = ji ) tensor of higher rank.
(b) Invert the expression you found in (b); that is, write ij in terms of the vorticity.
Solution:

(a) Cleverly adding zero, we can write


k = ijk i uj
1
= ijk [(i uj + j ui ) + (i uj j ui )]
2
= ijk (Sij ij ),
where Sij = (1/2)(i uj + j ui ); we will see later that Sij is the rate-of-strain tensor.
For now, note that Sij is symmetric. Since the contraction of a symmetric (Sij ) and an
antisymmetric (ijk ) tensor vanishes, we are left with
k = ijk ij .
(b) A good way to eliminate a Levi-Civita symbol from your equations is to contract it with
another Levi-Civita symbol and use the epsilondelta rule. So, starting with our result
from part (a),
`mk k = `mk ijk ij
= (`i mj `j mi )ij
= (`m m` )
= 2`m .
Rearranging, and relabeling dummy indices, we have
1
ij = ijk k .
2
7. We have seen pseudovectors and pseudotensors in class. It should not be surprising, then,
that one can also construct pseudoscalars. As an example, show that the scalar triple product
a b c, where a, b, and c are proper vectors, is actually a pseudoscalar. What happens to
the scalar triple product under coordinate reflections?
Solution:
In index notation, we can write the scalar triple product as ijk ak bi cj . There are no free
indices, so this is a scalar. Consider its transformation under an odd number of coordinate
reflections. In that case, ak , bi , and cj will all transform as expected for vectors: they will
each pick up a negative sign. ijk , however, will not (see above). Therefore, the scalar triple
product changes sign under an odd number of coordinate reflections, and cannot be a true
scalar.
8. Which of these could be valid tensor equations? For those that are not, state why not.
(a) aij = bik` cj` dk
(b) = mii + nii xii
(c) Lijk = imn rjm pkn
(d) x = y
(e) dn = `kn q` ijk ri uj
(f) ak = rijk ij + Sijk`m fi gj`
(g) y3 = qij3 zij

Solution:
(a) Valid.
(b) Not valid. The term nii xii is ambiguous, and could mean any of nij xij , nij xji , or
nii xjj , all of which are (in principle) different.
(c) Not valid. The right-hand-side contains imn , which is not a tensor.
(d) Valid.
(e) Valid. The two Levi-Civita symbols are contracted, and can be rewritten in terms of
Kronecker deltas.
(f) Not valid. There is an extra free m on the right-hand side.
(g) Valid.

You might also like