Download Dyadic Tensor Notation

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Oscillator representation wikipedia , lookup

Equations of motion wikipedia , lookup

Angular momentum operator wikipedia , lookup

Frame of reference wikipedia , lookup

Fictitious force wikipedia , lookup

Stress (mechanics) wikipedia , lookup

Inertia wikipedia , lookup

Spinor wikipedia , lookup

Velocity-addition formula wikipedia , lookup

Deformation (mechanics) wikipedia , lookup

Photon polarization wikipedia , lookup

Theoretical and experimental justification for the Schrödinger equation wikipedia , lookup

Classical central-force problem wikipedia , lookup

Finite strain theory wikipedia , lookup

Derivations of the Lorentz transformations wikipedia , lookup

Minkowski space wikipedia , lookup

Moment of inertia wikipedia , lookup

Centripetal force wikipedia , lookup

Symmetry in quantum mechanics wikipedia , lookup

Laplace–Runge–Lenz vector wikipedia , lookup

Rigid body dynamics wikipedia , lookup

Cauchy stress tensor wikipedia , lookup

Kinematics wikipedia , lookup

Bra–ket notation wikipedia , lookup

Relativistic angular momentum wikipedia , lookup

Four-vector wikipedia , lookup

Tensor operator wikipedia , lookup

Hooke's law wikipedia , lookup

Transcript
Here are some notes on vector and dyadic tensor notation similar to
what I will be using in class, with just a couple of changes in
notation.
Notes = Dowling
∧ = × British notation for Cross-Product vs. American (Dowling’s)

A = A Notation for Vector vs. Standard (Dowling’s).

B=B
Notation for Dyadic Tensor vs. Standard (Dowling’s).
With these notes you should be able to make sense of the expressions
in the vector identities pages below that involve the dyadic tensor T.
(Notes courtesy of Peter Littlewood, University of Cambridge.)
The vector identity pages are from the NRL Plasma Formulary – order
your very own online or download the entire thing at:
http://wwwppd.nrl.navy.mil/nrlformulary/
--JPD
VECTOR IDENTITIES4
Notation: f, g, are scalars; A, B, etc., are vectors; T is a tensor; I is the unit
dyad.
(1) A · B × C = A × B · C = B · C × A = B × C · A = C · A × B = C × A · B
(2) A × (B × C) = (C × B) × A = (A · C)B − (A · B)C
(3) A × (B × C) + B × (C × A) + C × (A × B) = 0
(4) (A × B) · (C × D) = (A · C)(B · D) − (A · D)(B · C)
(5) (A × B) × (C × D) = (A × B · D)C − (A × B · C)D
(6) ∇(f g) = ∇(gf ) = f ∇g + g∇f
(7) ∇ · (f A) = f ∇ · A + A · ∇f
(8) ∇ × (f A) = f ∇ × A + ∇f × A
(9) ∇ · (A × B) = B · ∇ × A − A · ∇ × B
(10) ∇ × (A × B) = A(∇ · B) − B(∇ · A) + (B · ∇)A − (A · ∇)B
(11) A × (∇ × B) = (∇B) · A − (A · ∇)B
(12) ∇(A · B) = A × (∇ × B) + B × (∇ × A) + (A · ∇)B + (B · ∇)A
(13) ∇2 f = ∇ · ∇f
(14) ∇2 A = ∇(∇ · A) − ∇ × ∇ × A
(15) ∇ × ∇f = 0
(16) ∇ · ∇ × A = 0
If e1 , e2 , e3 are orthonormal unit vectors, a second-order tensor T can be
written in the dyadic form
(17) T =
P
i,j
Tij ei ej
In cartesian coordinates the divergence of a tensor is a vector with components
(18) (∇·T )i =
P
j
(∂Tji /∂xj )
[This definition is required for consistency with Eq. (29)]. In general
(19) ∇ · (AB) = (∇ · A)B + (A · ∇)B
(20) ∇ · (f T ) = ∇f ·T +f ∇·T
4
Let r = ix + jy + kz be the radius vector of magnitude r, from the origin to
the point x, y, z. Then
(21) ∇ · r = 3
(22) ∇ × r = 0
(23) ∇r = r/r
(24) ∇(1/r) = −r/r 3
(25) ∇ · (r/r 3 ) = 4πδ(r)
(26) ∇r = I
If V is a volume enclosed by a surface S and dS = ndS, where n is the unit
normal outward from V,
(27)
Z
(28)
Z
(29)
Z
(30)
Z
(31)
Z
(32)
Z
dV ∇f =
Z
V
dSf
S
dV ∇ · A =
Z
V
dS · A
S
dV ∇·T =
Z
V
dS ·T
S
dV ∇ × A =
Z
V
dS × A
S
2
2
dV (f ∇ g − g∇ f ) =
V
Z
dS · (f ∇g − g∇f )
S
dV (A · ∇ × ∇ × B − B · ∇ × ∇ × A)
V
=
Z
dS · (B × ∇ × A − A × ∇ × B)
S
If S is an open surface bounded by the contour C, of which the line element is
dl,
(33)
Z
S
dS × ∇f =
I
dlf
C
5
(34)
Z
(35)
Z
(36)
Z
S
dS · ∇ × A =
I
dl · A
C
(dS × ∇) × A =
I
S
S
dl × A
C
dS · (∇f × ∇g) =
I
f dg = −
C
I
C
6
gdf
EXAMPLES CLASS IN MATHEMATICAL PHYSICS
V. Introduction to Tensors (3.2.99)
Commentary: This is the rst of two sheets on tensors (the second will come in class
VIII). Tensors have many applications in theoretical physics, not only in polarization problems like the ones described below, but in dynamics, elasticity, uid mechanics, quantum
theory, etc.. The mathematical apparatus for dealing with tensor problems may not yet
be familiar, so this sheet contains quite a high proportion of formal material.
Section A introduces the concept of a tensor (in this case, a linear operator that
transforms one vector into another) in the physical context of studying the polarisation of
an aspherical molecule.
Section B summarizes various mathematical and notational ideas that are useful in
physics
these include sux notation, and concepts such as the outer product of two vectors,
and projection operators. (These are mathematical objects which take the projection or
component of a vector along a particular direction.)
Section C returns to a physical problem (also about polarization) and leads to a discussion of principal axes, eigenvectors and eigenvalues of tensor quantities.
Section D introduces the properties of the antisymmetric third-rank tensor ijk which
can be used to derive many results in vector calculus much more rapidly than by other
methods. This ground will be covered again in sheet VIII so it doesn't matter if you run
out of time.
A. Introduction
We often nd vectors that are linearly related to other vectors, for instance the polarisation P induced in a medium by applying a eld E
P = 0 E
(1)
Here P and E are parallel. Their cartesian components are related to each other by
the same constant of proportionality 0 for each component. Here we have denoted
vectors by underlinings as one would do if writing this out by hand. Usually books denote
vectors by bold type
in the rest of this sheet we treat the two notations as completely
interchangeable.]
It is possible to imagine situations where the medium is anisotropic and the response
factor 0 in (1) is dierent for components in dierent directions. The two vectors P and
E are no longer necessarily parallel and what connects them is called a tensor. We will
use simple model of molecular polarisability to show the underlying physics and introduce
the ideas of tensors.
A1. An Electrostatics Problem. The diacetylene molecule, H-C C-C C-H is
highly polarisable along its long axis, pk = kEk being the induced dipole for a eld Ek
applied along the molecule. The response to perpendicular elds is p? = ?E? where
? k. Let the cylindrical molecule point in the (1 1 0) direction and apply a eld in
the x direction: E = Exx^
(i) Prove:
px = (k + ?)Ex=2
(2)
py = (k ; ?)Ex=2
(3)
pz = 0
(4)
2
Hence one sees that p and E are not parallel and in general we must write
p=E
(5)
where is the polarisability tensor. Tensors are written with a double underline (as
matricies often are) or in books sometimes as \doubly bold" (bold sans serif typeface)
characters. In component form (5) is
px = xxEx + xy Ey + xz Ez
(6)
py = yxEx + yy Ey + yz Ez
(7)
pz = zxEx + zy Ey + zz Ez
(8)
(ii) What are the coecients ij explicitly for the above example? Write as a matrix of its
components. Note that, although the tensor is the same physical entity in all coordinate
systems (it relates p with E in whatever basis these are expressed), its components ij
depend on the coordinate system used.
Remark: If many molecules are present with the same orientation, P is p where there
are molecules per unit volume. A (hypothetical) diacetylene solid might consist of such
rods arranged in an array, all directed along (1 1 0). Then the macroscopic (to replace
in (1)) would (neglecting internal eld corrections) be .
B. Miscellaneous Ideas and Notations
B1. Sux notation: Commonly, suces such as i and j are dummy symbols and
when repeated in an expression they are to be summed over
this is the Einstein Convention.
For instance (6) to (8) can be succinctly written as
pi = ij Ej
3
which is Eq.(5) rewritten in sux notation. (Note the order of the indices on the right.)
Thus we have written the vector p simply as pi and it will be clear from context that a
vector is intended and not simply one of its components. Likewise can be denoted ij .
In sux notation the dot product (also called scalar product, or inner product) of two
vectors a b is written as aibi. The operation of a tensor on a vector also involves an inner
product (sum over repeated indices) as described above. The unit tensor ij by denition
transforms any vector into itself: ij vj = vi.
B2. The outer product: Consider the object Qij dened as
Qij ai bj
(9)
where a and b are vectors.
(i) Show using sux notation that if p is a vector, Q p is also a vector
nd its magnitude
and direction. Note that Qp Qij pj .] We see that Qij is a linear operator that transforms
one vector p into another
it is therefore a tensor. This denes the outer product or dyadic
product of two vectors.
In non-sux notation (sometimes called dyadic notation), Eq.(9) is written as
Q = ab
(10)
or Q = a b, where there is no dot between the two vectors. In this notation we have
(a b) c = a(b c): prove this using sux notation if you did not do so already above.
This notation is used in various areas of mathematics and physics, but in general sux
notation is more versatile and can always be used instead.
(ii) We can write the tensor in Q. A1 in the following form:
= xx x^ x^ + xy x^ y^ + xz x^ z^ + 6 other terms
(11)
Check that (11) indeed agrees with (6) to (8) by nding the result of applying a eld in
the x-direction, E = Exx^ :
p = E x^ Ex = (xx yx zx) Ex :
4
(iii) Projection operators: In dyadic notation, the identity tensor ij is often written
as the \unit operator" 1. Consider the action on a vector v of the operators u^ u^ and
1 ; u^ u^ , where u^ is a unit vector. Show that these perform the job of resolving the vector
v into vectors vk and v?, respectively parallel and perpendicular to u^ . Prove that these
\projection operators" are idempotent. The denition of an idempotent operator G is that
GG = G.] Show that the polarizability tensor in problem A1 can be written
p
= ku^ u^ + ?(1 ; u^ u^ )
where u^ = (1 1 0)= 2.
C. Use of Principal Axes.
C1. A xed object is placed in a uniform electric eld of 1 kV/ m. When the eld
is in the x direction, it acquires a dipole moment with components 4, 2, and 1 10;12 C
m in the x, y, z directions. For the same eld strengths in the y, z directions, the dipole
moments are (2 4 1) 10;12 and (1 1 4) 10;12 C m respectively.
(i) Write down the components of the tensor that describes the polarisability of the object
(using the x y z coordinate frame).
p
(ii) What is the torque T = p ^ E on the object when it is placed in an electric eld of 3
kV/m in the (1,1,1) direction? Ans: (1,-1,0) 10;9 Nm.]
(iii) Suppose there are eld directions u^ ,^v, w^ for which there is no torque on the object
what relation do E and p then have? Show that u^ , v^ , and w^ are the eigenvectors of .
(iv) Show that, if we use as a coordinate system the unit vectors u^ ,^v, w^ there are no
o-diagonal elements of the tensor when it is written as a matrix of components. The
three diagonal elements, 1 2 and 3 are called the principal polarisabilities of the object
and are the eigenvalues of . For this problem, one eigenvalue is = 2 10;15: nd the
5
other two. What are the vectors u^ ,^v, w^ in terms of x^, y^, z^?
(v) In this example, is symmetric and hence has real eigenvalues and orthogonal eigenvectors. An energy argument shows this to be true for the polarizability of any object
the u^ ,^v, w^ coordinate system can hence always be found and is called the principal frame
- often a very convenient frame to think about the physics in. The same applies to any
problem involving a symmetric tensor.
(vi) In a general coordinate system, the decomposition of into principal directions u^ ,^v,
w^ can be accomplished using projection operators:
= 1u^u^ + 2v^v^ + 3w^ w^
(12)
This is a relation between tensors, so if it holds in one co-ordinate system it also holds in
any other. Prove it for a convenient choice of co-ordinates system (and hence generally).
(vii) Eq.(12) expresses in a very direct way the operation of a symmetric tensor on a
vector b. To nd b, the recipe is: (1) resolve b into components along the principal
axes
(2) multiply each component by the corresponding eigenvalue
(3) add to form the
resultant. Conrm from Eq.(12) that this is precisely what operating with on b does.
C2. Force on a Dipole.
(i) By using sux notation, and the ordinary chain rule for derivatives, prove the vector
relation
r (
A) = r A + A r
Derive a similar result for r(A B). It's best to leave the answer in sux notation.]
(ii) In a uniform electric eld, E, a dipole p experiences a couple p E but no net force.
Show that in a non-uniform eld, a dipole experiences a net force
ip
Fi = @E
@x j
j
(13)
and explain why @E
@x is a tensor. This is very general
the gradient of a vector is a secondrank tensor.]
i
j
6
(iii) A more general expression for the force on a permanent dipole, derived from the
energy, is F =grad(p:E), or:
Fj = @x@ piEi :
j
Show that this is equivalent to Eq(13) only in the absence of a time-varying magnetic eld.
(iv) A speck of matter has no permanent dipole moment, but the eld E induces a moment
p in it. Under what conditions can we write (a) pi = ij Ej (b) pi = Ei? Write down
the force in each case. Under what conditions can this be written F = grad(E 2=2) ?
D. The antisymmetric three-tensor
ijk .
A tensor with two indices such as ij is called a tensor of second rank, and its linear
operation on a vector gives another vector. A vector is technically a tensor of rank one
a
scalar of rank zero. Tensors of higher rank can also be dened. For example a tensor of
rank four, Qijkl acts on a rank two tensor to give another rank two tensor, Bij = Qijkl Akl
(with summation over repeated indices). For tensors higher than rank two, dyadic notation
becomes extremely cumbersome and sux notation is essential.
D1. An important and useful tensor of rank three is the fully antisymmetric three
tensor denoted ijk . This operates on either a second rank tensor or two vectors to give a
vector. For i 6= j 6= k it is dened as ijk = 1 if ijk is a cyclic permutation of x y z (e.g.,
yzx) and = ;1 if the order is anticyclic (e.g., xzy). If any two of ijk are the same then
ijk = 0 (e.g., xxy = 0). The tensor ijk is very useful for writing down a vector product
in sux notation. Show that
(A ^ B)i = ijk Aj Bk
(14)
ijk klm = il jm ; im jl
(15)
The following relation
7
is well worth remembering (you should come across this in maths lectures) . It saves a
great deal of time when manipulating vector identities such as
A ^ (B ^ C ) = ijk Aj klm Bl Cm
= Aj Bl Cm (il jm ; im jl )
= Bi Aj Cj ; Ci Aj Bj = (A C )B ; (A B)C
which is a familiar result normally derived by far more tedious means.
The result Eq.(15) is especially helpful when dealing with vector identities involving
the r operator. Because ijk is a constant tensor, it commutes with r, and expressions
can be re-ordered accordingly. For example
r (E ^ H ) ri( ijk Ej Hk )
(16)
= ijk ri(Ej Hk ) = ijk Ej riHk + Hk riEj ]
(17)
= ;Ej jik riHk + Hk kij riEj
(18)
= ;E r ^ H + H r ^ E
(19)
which is a result used in in connection with Poynting's vector in electromagnetism.
(i) Make sure you understand each step of this derivation.
(ii) Show, using similar methods, that
r ^ (r ^ A) = r(r A) ; r2A
r ^ (
A) = r ^ A ; A ^ r
These relations are useful in electromagnetism and other topics in IB, and in the parts II
and III courses.
8
EXAMPLES CLASS IN MATHEMATICAL PHYSICS
VIII. Further Applications of
Tensor and Vector Calculus (24.2.99)
Commentary: This is the second sheet on tensors. It follows on from Sheet 5. Section D
of that sheet (on uses of the antisymmetric 3-tensor, ijk ) is reprinted here as an appendix,
and you should complete it before starting the main part of this sheet, if you didn't do so
last time.
As on sheet 5 we use underlining and bold notation interchangeably. Sux notation
and the summation convention are also used. Carets (^) denote unit vectors.
In Sections A and B tensor ideas are used to work through some results in rigid
body dynamics. You should have covered this material in IB dynamics courses, and the
formal development given here is good practice and will help deepen your understanding
of moments of inertia, etc., as well as of tensors.
Section C introduces the concepts of stress and strain tensors in a solid. The linear
relation between these is governed by a fourth rank tensor, whose properties are investigated.
Section D contains problems on the transformation properties of tensors: in mathematical courses, you may have seen tensors dened in terms of these transformation properties.
A. Rigid Body Rotations
A rigid body rotates with angular velocity ! about an axis through its centre of mass
at the origin, 0. The vector ! points along the axis of rotation and determines the velocity
of rotation at r according to v = ! ^ r. Make sure you can derive this if you are not sure
where it comes from.]
(i) Consider the body as a collection of point masses mr with position vectors r relative
to the centre of mass 0.
The tensor I is called the inertia tensor of the body. It is given by:
I=
X
r
;
r2 1 ; ^r ^r mr
(1)
Show, using Eq.(1), that the angular momentum L and the kinetic energy T of the body
are:
L=I!
(2)
T = 21 ! I ! 12 !i Iij !j
(3)
You will need to expand the vector triple product f ^ (g ^ h) { see Appendix.]
Thus I species the linear dependence of angular momentum Li on the angular velocity
!j (via Li = Iij !j , Eq.(2)) and replaces the single number, the moment of inertia, that
would be enough for a very symmetrical body, such as a sphere.
(ii) Find the components of Iij in a Cartesian system, that is the usual basis set x^, y^, ^z.
P
P
Answer: Ixx = r (y2 + z2 )mr, Ixy = ; r(xy)mr, etc..]
(iii) Iij is a symmetric tensor, that is I:v = v:I for any vector v, which should be clear
from the denition (1). Deduce it also from the matrix of components of Iij found in (ii).
(iv) Take the continuum limit in Eq.(1) to obtain formally, as an integral, the inertia tensor
of a body of variable density (r).
(v) Write down the inertia tensor of a body consisting of eight masses M at the corners of
2
a light cubic frame of side a. Choose the coordinate system most convenient to you. What
is the moment of inertia of this cube? Why can we talk of the moment of inertia when we
strictly speaking have a tensor?
In fact, even for an asymmetric body one sometimes speaks of the moment of inertia about
some specic axis, n^ , say. This is written as n^ :I:n^ and gives the ratio of the n^ component
of angular momentum (L:n^ ) to the n^ component of angular velocity (!:n^ ). Since in general
L has components which are not parallel to n^ , this does not oer a complete description.]
(vi) Angular velocity: A body rotates about an axis through its centre of mass with an
angular velocity !. The velocity of a point in the body is v = ! ^ r. Show that one can
equally well express it as v = ! r where ! is an antisymmetric tensor with components:
0 ;!z !y !
!z 0 ;!x
;!y !x
0
(4)
(vii) Show that any vector W, constant with respect to axes xed in the body, changes
with time in a space-xed frame, by an amount dW in time dt:
dW = (!dt) ^ W
and hence that if W is not body-xed but has a rate of change (dW=dt)body , the vector
W in the space-xed frame obeys
(dW=dt)space = (dW=dt)body + ! ^ W :
(5)
B. Principal Axes and Precession
We now use the above results to solve simply a dicult problem, the rotational motion
of an asymmetric body.
3
(i) What is the inertia tensor of four point masses M xed at the corners of a massless
square frame of side a? What are the principal axes, and principal moments of inertia?
Why are two of the axes indeterminate?
This is an example of a nontrivial body (unlike the sphere or cube) where the principal
moments of inertia are not all the same. Now L need not be parallel to ! for a body
whose angular momentum is conserved the result is free precession of the angular velocity
(see (iii) below).
(ii) It is clear from (1) that in general I for a body is constant in body-xed axes, and so
for an arbitrary motion the inertia tensor Iij(sp) in space-xed axes is not constant (unless,
trivially, it is proportional to ij ). Show that in general the motion of a nontrivial body
whose angular momentum L(sp) is conserved (in a space-xed frame) involves an angular
velocity ! which changes in time. Hint: consider L(isp) = Iij(sp)!j .]
(iii) The problem of nding ! for a body of xed angular momentum (not parallel to a
principal axis) is best solved by adopting a body frame where the inertia tensor is constant.
Show from (5) that in the body frame L obeys
dL=dt = ;! ^ L
(6)
(iv) Deduce that
Iij !_ j = ; ijk !j Ikl !l
and show that, by taking as the body-xed frame the principal frame of Iij , this can be
simplied to
I1 !_ 1 = ;!2 !3(I3 ; I2)
I2 !_ 2 = ;!3 !1(I1 ; I3)
I3 !_ 3 = ;!1 !2(I2 ; I1)
where I1, I2, I3 are the principal moments of inertia.
(v) For the inertia tensor of the square frame found in question (i), I1 = I2 (such a body
4
is known as a \symmetric top"). Prove for this case that ! precesses in the body frame
according to
!3 = const. !1 = A cos(t) !2 = A sin(t)
where = (I3I;1I1) !3, and so conrm that ! is not a constant vector unless it happens to
point along one of the principal axes.
C. Stress and strain tensors.
C 1. Consider a small area element dS embedded at a point r in an isotropic liquid.
If the material is subject to pressure P , there is a normal force on each side of the surface
element PdS. These point in opposite directions, so the net force on the surface element
is zero { which is as it must be, since the element has no mass and would otherwise be
subject to innite acceleration.
In the case of a solid, there can also be a shear force exerted in equal and opposite
directions on the two sides of the element such a force lies in the plane of the element
rather than normal to it. In the general case, we can write
dF = dS
(7)
where is the stress tensor at the point r. For example the element xz of determines
the x-force per unit area on an element of surface with its normal pointing in the zdirection: dFx = xz dSz . For a surface element of general orientation dFx = xx dSx +
xy dSy + xz dSz .
(i) Show that a small box of sides dx dy dz experiences a couple (xz ; zx)dx dy dz about
the y axis. The corresponding moment of inertia is of order (dx)2(dz)2dy. Explain why
this means that must be symmetric. Show also that, if (r) is a function of position,
5
then the net force on a nite piece of material is given by
Fi =
Z
@ij
@rj dV :
(8)
where (in a conventional notation) @r@ rj .
(ii) Strain: Suppose a solid material is subjected to a small deformation r ! r + u(r). So
long as the deformation depends continuously on position, we may write
j
i
dui = @u
@r drj :
j
(9)
Let us now decompose the tensor Uij @u
@r into its symmetric and anti-symmetric parts:
i
j
Uij = eij + Rij
;
eij = 12 Uij + Uji
;
Rij = 21 Uij ; Uji
Show that e is symmetric and R antisymmetric. Show also that R corresponds to a pure
rotation (i.e. without distortion) of the bit of matter near r. Hint: compare Eq.(4) in
section A.] What type of deformation does e represent? Why may we assume that the
stress tensor depends on the strain tensor e, but has no dependence on R?
(iii) Elasticity: For small deformations, we expect the stress and strain to be linearly
related, but since both are described by second-rank tensors, the relation between them
generally involves a tensor of the fourth rank:
ij = Eijkl ekl
(10)
which is the generalization of Hooke's law to an arbitrary three-dimensional material.
Since is symmetric, Eijkl must be symmetric under exchange of i and j various other
symmetries can be found from considering the form of the stored elastic energy, but even
when these are taken into account, Eijkl has 21 independent components.
6
(iv) Isotropic elastic solids are much simpler however, and have only two independent
constants. These are the Young's modulus Y and Poisson's ratio . The denition of these
is that in the principal axes of ij , the following relation applies
Y exx = xx ; (yy + zz )
(11)
with similar relations for eyy ezz all o-diagonal elements of e are zero in this coordinate
system. By rearranging Eq.(11) obtain the result
Y exx = (1 + )xx ; ii
(where x indicies are not summed over but ii is summed) with similar relations for yy
and zz components. Note that ii = Trace( ) under the summation convention.] Show
that, by adding Eq.(11) to its analogues,
ii = 1Y;e2ii
Combine the above two results to give expressions for xx etc in terms of the elements of
e, and hence derive the following result in the principal frame:
Y e + Tr(e) :
ij = 1 +
ij 1 ; 2 ij
Bear in mind that eij = 0 for i 6= j , in this frame.] Hence, derive the general (frame
independent) form for the elasticity tensor in an isotropic solid:
ik jl + il jk Y
+ (1 ; 2) ij kl
Eijkl = 1 + 2
C 2. (i) Apply Eijkl to suitable deformations to recover expressions for the bulk
modulus and shear modulus of an isotropic solid in terms of Y and .
(ii) Show that the displacement u inside a homogeneous isotropic solid, subject to
external forces only at its surface, obeys
grad divu + (1 ; 2)r2u = 0:
7
D. Transformation Laws.
D 1. Here we consider transformations of vectors and tensors between two sets of base
vectors, each of which is orthonormal. We call our two sets of base vectors e1 e2 e3 and
e01 e02 e03, so that we can use sux notation, e.g., the fact that the e basis is orthonormal
can be written
ei ej = ij
(12)
The unit vectors e0i are linear combinations of the ei:
e0i = Tij ej :
(13)
Suppose that x = xiei = x0ie0i is any vector, and y = Ax. Prove that
(i) e0i ek = Tik
(ii) The inverse transformation ei = Tik;1e0k has Tik;1 = Tki
(iii) x0i = Tik xk
(iv) xi = Tkix0k
(v) yi0 = Tij Tlk Ajk x0l
This means you have found the components of the tensor A in the rotated coordinate
system:
A0il = Tij Tlk Ajk
Indeed, this property under orthogonal coordinate transformations T is, mathematically,
often taken as the dening property of a tensor A. For tensors of higher rank than 2,
one factor of the transformation matrix T is needed for each index. Transformations with
det(T ) = 1 are usually called proper rotations and those with det(T ) = ;1, improper rotations. Show that these are the only possible values of the determinant for transformations
between orthonormal bases as dened above.
8
Appendix: The antisymmetric three-tensor
ijk .
Reprise of Sheet V section D.]
A tensor with two indices such as ij is called a tensor of second rank, and its linear
operation on a vector gives another vector. A vector is technically a tensor of rank one a
scalar of rank zero. Tensors of higher rank can also be dened. For example a tensor of
rank four, Qijkl acts on a rank two tensor to give another rank two tensor, Bij = Qijkl Akl
(with summation over repeated indices). For tensors higher than rank two, dyadic notation
becomes extremely cumbersome and sux notation is essential.
1. An important and useful tensor of rank three is the fully antisymmetric three tensor
denoted ijk . This operates on either a second rank tensor or two vectors to give a vector.
For i =
6 j 6= k it is dened as ijk = 1 if ijk is a cyclic permutation of x y z (e.g., yzx)
and = ;1 if the order is anticyclic (e.g., xzy). If any two of ijk are the same then ijk = 0
(e.g., xxy = 0). The tensor ijk is very useful for writing down a vector product in sux
notation. Show that
(A ^ B)i = ijk Aj Bk
(14)
ijk klm = il jm ; im jl
(15)
The following relation
is tedious to prove (using components { you should come across this in maths lectures) but
is worth remembering. It saves a great deal of time when manipulating vector identities
such as
A ^ (B ^ C ) = ijk Aj klm Bl Cm
= Aj Bl Cm (il jm ; im jl )
= Bi Aj Cj ; Ci Aj Bj = (A C )B ; (A B)C
9
which is a familiar result normally derived by more tedious means.
The result Eq.(15) is especially helpful when dealing with vector identities involving
the r operator. Because ijk is a constant tensor, it commutes with r, and expressions
can be re-ordered accordingly. For example
r (E ^ H ) ri ( ijk Ej Hk )
(16)
= ijk ri(Ej Hk ) = ijk Ej riHk + Hk riEj ]
(17)
= ;Ej jik riHk + Hk kij riEj
(18)
= ;E r ^ H + H r ^ E
(19)
which is a result used in in connection with Poynting's vector in electromagnetism.
(i) Make sure you understand each step of this derivation.
(ii) Show, using similar methods, that
r ^ (r ^ A) = r(r A) ; r2 A
r ^ (
A) = r ^ A ; A ^ r
These relations are useful in electromagnetism and other topics in the IB and parts II and
III courses.
10