Chapter 3 Cartesian Tensors 3.1
Suffix Notation and the Summation Convention
We will consider vectors in 3D, though the notation we shall introduce applies (mostly) just as well to n dimensions. For a general vector x = (x1 , x2 , x3 ) we shall refer to xi , the ith component of x. The index i may take any of the values 1, 2 or 3, and we refer to “the vector xi ” to mean “the vector whose components are (x1 , x2 , x3 )”. However, we cannot write x = xi , since the LHS is a vector and the RHS a scalar. Instead, we can write [x]i = xi , and similarly [x + y]i = xi + yi . Note that the expression yi = xi implies that y = x; the statement in suffix notation is implicitly true for all three possible values of i (one at a time!). Einstein introduced a convention whereby if a particular suffix (e.g., i) appears twice in a single term of an expression then it is implicitly summed. For example, in traditional notation 3 X x . y = x1 y1 + x2 y2 + x3 y3 = xi yi ; i=1
using summation convention we simply write x . y = xi yi . All we are doing is not bothering to write down the
P
!
The Rules of Summation Convention Summation convention does not allow any one suffix to appear more than twice within a single term; so xi yi zi is meaningless. We have to take care to avoid this: for example, 46
© R. E. Hunt, 2002
consider the vector relation y = (a . b)x. We have a . b = ai bi , but we cannot write yi = ai bi xi as this would be ambiguous. How can we correct this? Note that a . b = ai b i = aj b j – the suffix we use for the summation is immaterial. (Compare with the use of dummy R∞ R∞ variables in integrations: 0 e−x dx = 0 e−y dy.) So we can write y i = aj b j x i . In any given term, then, there are two possible types of suffix: one that appears precisely once, e.g., i in aj bj xi , which is known as a free suffix ; and one that appears precisely twice, e.g., j in aj bj xi , which is known as a dummy suffix. It is an important precept of summation convention that the free suffixes must match precisely in every term (though dummy suffixes can be anything you like so long as they do not clash with the free suffixes). So in the equation aj bj zk = xk + ai ai yk bj bj every term has a free suffix k, and all other suffixes are dummy ones. In vector notation, this equation reads (a . b)z = x + |a|2 |b|2 y. (Note that the order of variables in the final term of this equation in suffix notation is unimportant: we could equally well have written bj yk ai bj ai .) There need not be any free suffixes at all, as in the equation ai zi = (xi + yi )ai (which reads a . z = (x + y) . a in vector notation). Suffix notation can also be used with matrices. For a matrix A, we write aij to denote the entry in the ith row and j th column of A (for each i = 1, 2, 3 and j = 1, 2, 3). We write either A = (aij ) or [A]ij = aij – these equations are equivalent – to indicate this. (Sometimes the upper-case letter is used instead, in which case the matrix A would have entries Aij .)
Examples of Summation Convention (i) 2x + y = z ⇔ 2xi + yi = zi . Note that the RHS of this suffix notation equation does not mean z1 + z2 + z3 – no repeated suffix, no sum! (ii) (a . b)(x . y) = 0 ⇔ ai bi xj yj = 0. 47
© R. E. Hunt, 2002
(iii) In summation convention, y = Ax is written yi = [Ax]i = aij xj (check that this is correct by writing it out long-hand for each possible value of the free suffix i). (iv) The matrix multiplication C = AB (where A and B are 3 Ă— 3 matrices) is written cij = [AB]ij = aik bkj . (v) The trace of a matrix C may be written as Tr C = cii , i.e., c11 + c22 + c33 . Hence Tr(AB) = aik bki . Replacing two free suffixes (e.g. i, j in cij ) by a single dummy suffix (cii ) is known as contraction. Not all expressions written in suffix notation can be recast in vector or matrix notation. For example, aijk = xi yj zk is a valid equation in suffix notation (each term has three free suffixes, i, j and k), but there is no vector equivalent.
3.2
The Kronecker Delta and the Alternating Tensor
The Kronecker delta is defined by  1 i = j δij = 0 i = 6 j and the alternating tensor is defined by
ijk =
    1
if (i, j, k) is a cyclic permutation of (1, 2, 3)
−1 if (i, j, k) is an anti-cyclic permutation of (1, 2, 3)    0 if any of i, j, k are equal
(i.e., 123 = 231 = 312 = 1; 213 = 132 = 321 = −1; all others are zero). Note that δij = δji and that ijk = jki = kij = − jik etc. 1 0 0 If I is the identity matrix 0 1 0 then [I]ij = δij . We see that 0 0 1
xi = δij xj 48
Š R. E. Hunt, 2002
because (i) this is equivalent to x = Ix; or (ii) we can check for each value of i (e.g., when i = 1, RHS = δ1j xj = δ11 x1 + δ12 x2 + δ13 x3 = x1 = LHS). The Kronecker delta just “selectsâ€? entries: e.g., δik ajk is equal to aji . What is δii ? It is not 1. The alternating tensor can be used to write down the vector equation z = x Ă— y in suffix notation: zi = [x Ă— y]i = ijk xj yk . (Check this: e.g., z1 = 123 x2 y3 + 132 x3 y2 = x2 y3 − x3 y2 , as required.) There is one very important property of ijk : ijk klm = δil δjm − δim δjl . This makes many vector identities easy to prove. (The property may be proved by first proving the  δil δim  ijk lmn = det ďŁ Î´jl δjm δkl δkm
generalisation  δin  δjn  . δkn
Both sides clearly vanish if any of i, j, k are equal; or if any of l, m, n are. Now take i = l = 1, j = m = 2, k = n = 3: both sides are clearly 1. Finally consider the effect of swapping say i and j. Once we have proved this generalisation, contract k and n and simplify, noting that for example δjk δkm = δjm .) Example: prove that a Ă— (b Ă— c) = (a . c)b − (a . b)c. [a Ă— (b Ă— c)]i = ijk aj [b Ă— c]k = ijk aj klm bl cm = (δil δjm − δim δjl )aj bl cm = aj b i c j − aj b j c i = (a . c)bi − (a . b)ci = [(a . c)b − (a . b)c]i , as required. ijk can also be used to calculate determinants. The determinant of a 3 Ă— 3 matrix A = (aij ) is given by ijk a1i a2j a3k (check this by just expanding the product and sum in
49
Š R. E. Hunt, 2002
full). This can be written in several other ways; for example, det A = ijk a1i a2j a3k = jik a1j a2i a3k [swapping i and j] = − ijk a2i a1j a3k . This proves that swapping two rows of a matrix changes the sign of the determinant.
3.3
What is a Vector?
A vector is more than just 3 real numbers. It is also a physical entity: if we know its 3 components with respect to one set of Cartesian axes then we know its components with respect to any other set of Cartesian axes. (The vector stays the same even if its components do not.) For example, suppose that {e1 , e2 , e3 } is a right-handed orthogonal set of unit vectors, and that a vector v has components vi relative to axes along those vectors. That is to say, v = v 1 e 1 + v 2 e2 + v 3 e 3 = v j e j . What are the components of v with respect to axes which have been rotated to align with a different set of unit vectors {e01 , e02 , e03 }? Let v = v10 e01 + v20 e02 + v30 e03 = vj0 e0j . Now e0i . e0j = δij , so v . e0i = vj0 e0j . e0i = vj0 δij = vi0 but also v . e0i = vj ej . e0i = vj lij where we define the matrix L = (lij ) by lij = e0i . ej . Then vi0 = lij vj (or, in matrix notation, v0 = Lv where v0 is the column vector with components vi0 ). L is called the rotation matrix. This looks like, but is not quite the same as, rotating the vector v round to a different vector v0 using a transformation matrix L. In the present case, v and v0 are the same vector, just measured with respect
50
Š R. E. Hunt, 2002
to different axes. The transformation matrix corresponding to the rotation {e1 , e2 , e3 } 7→ {e01 , e02 , e03 } is not L (in fact it is L−1 ).
Now consider the reverse of this argument. Exactly the same discussion would lead to vi = ˆlij vj0 where ˆlij = ei . e0 j (we swap primed and unprimed quantities throughout the argument). We note that ˆlij = lji from their definitions; hence ˆ = LT L and so ˆ 0 = LT v 0 . v = Lv We can deduce that v = LT Lv, and furthermore, this is true for all vectors v. We conclude that LT L = I, i.e., LT = L−1 . (Hence LLT = I also.) L is therefore an orthogonal matrix. In suffix notation, the equation LT L = I reads lki lkj = δij , and LLT = I reads lik ljk = δij ; both of these identities will be useful. Another way of seeing that LLT = I (or, equivalently, LT L = I) is to consider the components of L. Since e0i . ej is just the j th component of e0i measured with respect to the first frame, we see that the ith row of L just consists of the components of e0i measured with respect to the first frame: e01 . e1 e01 . e2 e01 . e3 L = e02 . e1 e02 . e2 e02 . e3 e03 . e1 e03 . e2 e03 . e3 =
e01 T e02 T e03 T
[measured with respect to frame 1].
51
© R. E. Hunt, 2002
Alternatively, the ith column consists of the components of ei with respect to the second frame. To calculate the top left component of LLT , we find the dot product of the first row of L with the first column of LT . Both are simply e01 measured with respect to the first frame, so we obtain e01 . e01 , which is 1. Similarly, the top right component of LLT is e01 . e03 , which is zero. So, considering all possible combinations of rows and columns, we see that LLT = I as required.
3.4
Tensors
Tensors are a generalisation of vectors. We think informally of a tensor as something which, like a vector, can be measured component-wise in any Cartesian frame; and which also has a physical significance independent of the frame, like a vector.
Physical Motivation Recall the conductivity law, J = σE, where E is the applied electric field and J is the resulting electric current. This is suitable for simple isotropic media, where the conductivity is the same in all directions. But a matrix formulation may be more suitable in anisotropic media; for example, 5 0 0 J = 0 4 0 E 0 0 0 might represent a medium in which the conductivity is high in the x-direction but in which no current at all can flow in the z-direction. (For instance, a crystalline lattice structure where vertical layers are electrically insulated.) More generally, in suffix notation we have Ji = σij Ej where σ is the conductivity tensor. What happens if we measure J and E with respect to a different set of axes? We would expect the matrix σ to change too: let its new components be σij0 . Then Ji0 = σij0 Ej0 . But J and E are vectors, so Ji0 = lij Jj 52
© R. E. Hunt, 2002
and Ei = lji Ej0 from the results regarding the transformation of vectors in §3.3. Hence σij0 Ej0 = Ji0 = lip Jp = lip σpq Eq = lip σpq ljq Ej0 =⇒
(σij0 − lip ljq σpq )Ej0 = 0.
This is true for all vectors E0 , and hence the bracket must be identically zero; hence σij0 = lip ljq σpq . This tells us how σ transforms. Compare this argument with the corresponding argument for the case Ax = 0 where A is a matrix; if it is true for all x then A must be zero, though this is not the case if it is only true for some x’s.
σ is a second rank tensor, because it has two suffixes (σij ). Definition: In general, a tensor of rank n is a mathematical object with n suffixes, Tijk... , which obeys the transformation law 0 Tijk... = lip ljq lkr . . . Tpqr...
where L is the rotation matrix between frames. Note: for second rank tensors such as σ, the transformation law Tij0 = lip ljq Tpq can be rewritten in matrix notation as T 0 = LT LT – check this yourself!
Examples of Tensors (i) Any vector v (e.g., velocity) is a tensor of rank 1, because vi0 = lip vp . (ii) Temperature T is a tensor of rank 0 – known as a scalar – because it is the same in all frames (T 0 = T ). (iii) The inertia tensor. Consider a mass m which is part of a rigid body, at a location x within the body. If the body is rotating with angular velocity ! then the mass’s velocity is v = ! × x, and its angular momentum is therefore mx × v = mx × (! × x) = m |x|2 ! − (! . x)x . 53
© R. E. Hunt, 2002
Changing from a single mass m to a continuous mass distribution with density Ď (x), so that an infinitesimal mass element is Ď (x) dV , we see that the total angular momentum of a rigid body V is given by ZZZ h= Ď (x) |x|2 ! − (! . x)x dV, V
or, in suffix notation, ZZZ Ď (x)(xk xk ωi − ωj xj xi ) dV
hi = V
ZZZ Ď (x)(xk xk δij − xj xi )ωj dV
= V
= Iij ωj where
ZZZ Ď (x)(xk xk δij − xi xj ) dV
Iij = V
is the inertia tensor of the rigid body. Note that the tensor I does not depend on !, only on properties of the body itself; so it may be calculated once and for all for any given body. To see that it is indeed a tensor, note that both h and ! are vectors, and apply arguments previously used for the conductivity tensor. (iv) Susceptibility χ. If M is the magnetization (magnetic moment per unit volume) and B is the applied magnetic field, then for a simple medium we have M = (m) χ(m) B where χ(m) is the magnetic susceptibility. This generalises to Mi = χij Bj (m) where χij is the magnetic susceptibility tensor. Similarly for polarization density (e) (e) P in a dielectric: Pi = χij Ej where E is the electric field and χij is the electric susceptibility tensor. (v) The Kronecker delta itself. We have defined δij without reference to frame; i.e., its components are by definition the same in all frames (δij0 ≥ δij ). Surprisingly, then, we can show that it is a tensor: lip ljq δpq = lip ljp = δij = δij0 (from §3.3), which is exactly the right transformation law. We can also show that ijk is a tensor of rank 3. Both δij and ijk are isotropic tensors: that is, their components are the same in all frames.
54
Š R. E. Hunt, 2002
(vi) Stress and strain tensors. In an elastic body, stresses (forces) produce displacements of small volume elements within the body. Let this displacement at a location x be u; then the strain tensor is defined to be 1 ∂ui ∂uj + . eij = 2 ∂xj ∂xi The stress tensor pij is defined as the j th component of the forces within the body acting on an imaginary plane perpendicular to the ith axis. Hooke’s law for simple media says that stress ∝ strain. We can now generalise this to the tensor formulation pij = kijkl ekl where kijkl is a fourth rank tensor, which expresses the linear (but possibly anisotropic) relationship between p and e.
3.5
Properties of Tensors
Linear Combination of Tensors If Aij and Bij are second rank tensors, and α, β are scalars, then Tij = αAij + βBij is a tensor. Proof: Tij0 = α0 A0ij + β 0 Bij0 = αlip ljq Apq + βlip ljq Bpq = lip ljq (αApq + βBpq ) = lip ljq Tpq as required. This result clearly extends to tensors of rank n.
Contraction (also known as the Inner Product) If Tij is a tensor then Tii is a scalar. Proof: Tii0 = lip liq Tpq = δpq Tpq = Tpp = Tii , so Tii has the same value in all frames as required. 55
© R. E. Hunt, 2002
We can extend this result: if Tijk...lmn... is a tensor of rank n then Sjk...mn... = Tijk...imn... is a tensor of rank n − 2. Proof: 0 0 Sjk...mn... = Tijk...imn...
= lip ljq lkr . . . liι lmβ lnγ . . . Tpqr...ιβγ... = (lip liι )ljq lkr . . . lmβ lnγ . . . Tpqr...ιβγ... = δpι ljq lkr . . . lmβ lnγ . . . Tpqr...ιβγ... = ljq lkr . . . lmβ lnγ . . . Sqr...βγ... .
Outer Product If a and b are vectors then the outer product Tij defined by Tij = ai bj is a tensor of rank two. Proof: Tij0 = a0i b0j = lip ap ljq bq = lip ljq ap bq = lip ljq Tpq as required. Similarly (left as an exercise for the reader) we can show that if Aijk... is a tensor of rank m and Blmn... is a tensor of rank n, then Tijk...lmn... = Aijk... Blmn... is a tensor of rank m + n. Example: if a and b are vectors then a . b is a scalar. Proof: Tij = ai bj , being an outer product of two vectors, is a tensor of rank two. Then Tii = ai bi , being a contraction of a tensor, is a scalar, as required. Note that |a|2 = a . a and |b|2 are also scalars; hence a . b/|a| |b| = cos θ is a scalar, so that the angle between vectors is unaffected by a change of frame.
3.6
Symmetric and Anti-Symmetric Tensors
A tensor Tijk... is said to be symmetric in a pair of indices (say i, j) if Tijk... = Tjik... or anti-symmetric in i, j if Tijk... = −Tjik... . For a second rank tensor we need not specify the indices as there are only two to choose from! For example, δij is symmetric; ijk is anti-symmetric in any pair of indices.
56
Š R. E. Hunt, 2002
Note: if Aij is a symmetric second rank tensor then the matrix corresponding to A is symmetric, i.e. A = AT . Similarly for an anti-symmetric tensor. Suppose that Sij is a symmetric tensor and Aij an anti-symmetric tensor. Then Sij Aij = 0. Proof: Sij Aij = −Sij Aji = −Sji Aji = −Sij Aij =⇒
(swapping dummy i and j)
2Sij Aij = 0,
as required. Try to work out also how to see this “by inspectionâ€?, by considering appropriate pairs of components. Example: for any vector a, a Ă— a = 0 because [a Ă— a]i = ijk aj ak and ijk is anti-symmetric in j, k whilst aj ak is symmetric. The properties of symmetry and anti-symmetry are invariant under a change of frame: that is, they are truly tensor properties. For example, suppose that Tij is symmetric. Then Tij0 = lip ljq Tpq = lip ljq Tqp = ljq lip Tqp = Tji0 , so that Tij0 is also symmetric. (Alternative, and simpler, proof for second rank tensors: T 0 = LT LT
=⇒
T 0T = (LT LT )T = LT T LT = LT LT = T 0
using T T = T .) Symmetry and anti-symmetry occur frequently in practical applications. For example, the strain tensor eij = 21 (∂ui /∂xj + ∂uj /∂xi ) is clearly symmetric. In most situations the stress tensor is also symmetric; but in some circumstances (for instance in crystallography or geodynamics) it is forced to be anti-symmetric while the strain remains symmetric. Inertia tensors are always symmetric; conductivity and susceptibility tensors usually are.
57
Š R. E. Hunt, 2002
Decomposition into Symmetric and Anti-Symmetric Parts Any second rank tensor Tij can be uniquely expressed as the sum of a symmetric and an anti-symmetric tensor; for Tij = Sij + Aij where Sij = 12 (Tij + Tji ),
Aij = 12 (Tij − Tji )
are symmetric and anti-symmetric respectively. Exercise: prove that S and A are tensors. Furthermore, any anti-symmetric tensor Aij can be expressed in terms of a vector (sometimes known as the dual vector ) such that
!
Aij = ijk ωk . Proof: define
! by ωk = 12 klm Alm .
Then ijk ωk = 21 ijk klm Alm = 12 (δil δjm − δim δjl )Alm = 21 (Aij − Aji ) = Aij
! is a vector as it is a contraction of two tensors. This definition of ! actually corresponds to setting
as required.

 0 ω3 âˆ’Ď‰2   A = ďŁâˆ’ω3 0 ω1  . ω2 âˆ’Ď‰1 0 Example: suppose that two symmetric second rank tensors Rij and Sij are linearly related. Then there must be a relationship between them of the form Rij = cijkl Skl . It is clear that cijkl must be symmetric in i, j (for otherwise, Rij would not be). It is not necessarily the case that it must also be symmetric in k, l, but without loss of generality (s) we may assume that it is, by the following argument. Decompose cijkl into a part cijkl (a) which is symmetric in k, l and a part cijkl which is anti-symmetric. Then (s)
(a)
(s)
Rij = cijkl Skl + cijkl Skl = cijkl Skl because the second term is the contraction of an anti-symmetric tensor with a symmetric one, which we showed was zero above. Hence we can ignore any anti-symmetric part of cijkl . 58
Š R. E. Hunt, 2002
3.7
Diagonalization of Symmetric Second Rank Tensors
Suppose Tij is a symmetric second rank tensor. We shall show that there exists a frame such that, if we transform T to that frame, it has components given by λ1 0 0 T 0 = 0 λ2 0 . 0 0 λ3 This process is known as diagonalization. The values λ1 , λ2 and λ3 are known as the principal values of T , and the Cartesian coordinate axes of the corresponding frame are known as the principal axes. We will see that in fact the principal values are just the eigenvalues of the matrix corresponding to T , and the principal axes are the eigenvectors. Because T is symmetric, we know that there are 3 real eigenvalues and that we can find 3 corresponding eigenvectors which are orthogonal and of unit length. Let λ1 , λ2 , λ3 be the eigenvalues and e01 , e02 , e03 be the eigenvectors (arranged as a right-handed set of orthonormal vectors). Change frame to one in which the coordinate axes are aligned with {e01 , e02 , e03 }. What is T 0 ?
Recall that LT = e01 e02 e03 ; i.e., the three columns of LT are the vectors e01 , e02 and e03 (measured relative to the first frame). Hence in matrix notation,
T LT = T e01 e02 e03
= λ1 e01 λ2 e02 λ3 e03 . So e01 T e02 T e03 T
T 0 = LT LT =
0 λ 1 e 1
λ2 e02
λ3 e03
λ1 0 0 = 0 λ2 0 0 0 λ3 because, for example, the top LHS entry is given by e01 . λ1 e01 , and the top RHS entry is e01 . λ3 e03 .
λ1 0 There is another way of seeing that T 0 = 0 λ2 0 0 (because T is a tensor, e01 a vector and λ1 a scalar).
0 0 . The equation T e01 = λ1 e01 is true in any frame λ3 In particular it is true in the frame with {e01 , e02 , e03 }
59
© R. E. Hunt, 2002
as coordinate axes. But, measured in this frame, e01 is just (1, 0, 0)T , and T has components T 0 ; so     1 Îť1 T 0 ďŁ0 = ďŁ 0  0 0 which shows that the first column of T 0 is (Îť1 , 0, 0)T . Similarly for the other columns.
Note: the three principal values are invariants of T . That is, whatever frame we start from, when we diagonalize T we will obtain the same values of Îť. The eigenvalues are properties of the tensor, not of the coordinate system.
3.8
Isotropic Tensors
An isotropic tensor is one whose components are the same in all frames, i.e., 0 Tijk... = Tijk... .
We can classify isotropic tensors up to rank four as follows: Rank 0: All scalars are isotropic, since the tensor transformation law states that T 0 = T for tensors of rank zero. Rank 1: There are no non-zero isotropic vectors. Rank 2: The most general isotropic second rank tensor is Νδij where Îť is any scalar, as proved below. Rank 3: The most general isotropic third rank tensor is Îť ijk . Rank 4: The most general isotropic fourth rank tensor is Νδij δkl + ¾δik δjl + νδil δjk where Îť, Âľ, ν are scalars. What is the physical significance of an isotropic tensor? Consider the conductivity tensor Ďƒij in an isotropic medium. As the medium is the same in all directions, we expect that Ďƒij will be isotropic too. Hence Ďƒij = Νδij and Ji = Ďƒij Ej = Νδij Ej = ÎťEi , i.e., J = ÎťE. So we recover the “simple versionâ€? of the conductivity law, as we might expect. 60
Š R. E. Hunt, 2002
Isotropic Second Rank Tensors Consider a general tensor T of rank two, with components Tij with respect to some set of axes {e1 , e2 , e3 }. Suppose that T is isotropic. Its components should then be unaltered by a rotation of 90◦ about the 3-axis, i.e., with respect to new axes e01 = e2 ,
e02 = −e1 ,
The matrix of this rotation is
Using the matrix 0 T11 0 T21 0 T31
e03 = e3 .
0 1 0 L = −1 0 0 . 0 0 1
formulation of the transformation law for tensors, 0 0 0 1 0 T11 T12 T13 0 T13 T12 0 0 T22 T23 = −1 0 0 T21 T22 T23 1 0 0 0 0 1 T31 T32 T33 0 T32 T33 T22 −T21 T23 = −T12 T11 −T13 . T32 −T31 T33
we see that −1 0 0 0 0 1
But, because T is isotropic, Tij0 = Tij . Hence, comparing matrix entries, we have: T11 = T22 ; T13 = T23 = −T13
so that
T13 = T23 = 0;
T31 = T32 = −T31
so that
T31 = T32 = 0.
Similarly, considering a rotation of 90◦ about the 2-axis, we find that T11 = T33 and that T12 = T32 = 0, T21 = T23 = 0. Therefore all off-diagonal elements of T are zero, and all diagonal elements are equal, say λ. Thus λ 0 0 T = 0 λ 0 , 0 0 λ or in suffix notation, Tij = λδij . In summary, we have shown that any isotropic second rank tensor must be equal to λδij for some scalar λ.
61
© R. E. Hunt, 2002
3.9
Tensor Differential Operators
A tensor field is a tensor which depends on the location x. For example: (i) Temperature is a scalar field (a tensor field of rank zero), because T = T (x). (ii) Any vector field F(x), such as a gravitational force field, is a tensor field of rank one. In particular, x is itself a vector field, because it is a vector function of position! (iii) In a conducting material where the conductivity varies with location, we have Ďƒij = Ďƒij (x), a tensor field of rank two. We are interested here in calculating the derivatives of tensor fields; we start with scalars and vectors. We can rewrite the definitions of grad, div, and curl using suffix notation. Grad:
[âˆ‡ÎŚ]i =
âˆ‚ÎŚ ∂xi
Div:
∇.F=
∂F1 ∂F2 ∂F3 ∂Fi + + = ∂x1 ∂x2 ∂x3 ∂xi
Curl:
[∇ Ă— F]i = ijk
∂Fk ∂xj
There is another notation worth knowing: if u, v are vectors then we define the vector ∂ ∂ ∂ + u2 + u3 (u . ∇)v = u1 v. ∂x1 ∂x2 ∂x3 In suffix notation, [(u . ∇)v]i = uj
∂vi . ∂xj
Laplace’s equation ∇2 ÎŚ = 0 becomes ∂2ÎŚ =0 ∂xi ∂xi in suffix notation. Similarly, ∂ 2 Fi ∂xj ∂xj
[∇2 F]i =
(note that we only use Cartesian coordinates here). We sometimes find it useful to use the differential operator ∂i defined by ∂i =
∂ . ∂xi
62
Š R. E. Hunt, 2002
Then ∇ . F = ∂i Fi ;
[âˆ‡ÎŚ]i = ∂i ÎŚ;
[∇ Ă— F]i = ijk ∂j Fk .
It turns out that ∂i is in fact a tensor of rank one. We know that xj = lij x0i (from x = LT x0 ) so that ∂xj ∂x0k ∂ 0 ) = l = (l x = lkj δik = lij . kj kj k ∂x0i ∂x0i ∂x0i (This looks obvious but has to be proved very carefully!) Now let T be some quantity (perhaps a scalar or a tensor of some rank). Then ∂i0 T = =
∂T ∂x1 ∂T ∂x2 ∂T ∂x3 ∂T = + + 0 0 0 ∂xi ∂x1 ∂xi ∂x2 ∂xi ∂x3 ∂x0i ∂T ∂xj ∂xj ∂x0i
= lij
∂T = lij ∂j T. ∂xj
This is true for any quantity T , so ∂i0 = lij ∂j , i.e., ∂i transforms like a vector, and is hence a tensor of rank one. This result allows us to prove that âˆ‡ÎŚ, ∇ . F and ∇ Ă— F are scalars or vectors (as appropriate). For example, to show that if F is a vector field then ∇ Ă— F is a vector field: [∇ Ă— F]0i = 0ijk ∂j0 Fk0 = lip ljq lkr pqr ljs ∂s lkt Ft [ , ∂ and F are tensors] = lip (ljq ljs )(lkr lkt ) pqr ∂s Ft = lip δqs δrt pqr ∂s Ft = lip pqr ∂q Fr = lip [∇ Ă— F]p , as required. Alternatively, we can just state that ∇ Ă— F is a contraction of the tensor outer product Tijklm = ijk ∂l Fm (because [∇ Ă— F]i = Tijkjk ).
As an example of a tensor field of rank three, consider the derivative of the conductivity tensor, ∂i Ďƒjk . This cannot be written using ∇. 63
Š R. E. Hunt, 2002