Module II: Relativity and Electrodynamics

Lecture 5: Metric and higher-rank 4-tensors

Amol Dighe TIFR, Mumbai

Outline

Metric and invariant scalar products

Second rank 4-tensors: symmetric and antisymmetric

Higher-rank 4-tensors

Coming up...

Metric and invariant scalar products

Second rank 4-tensors: symmetric and antisymmetric

Higher-rank 4-tensors

Relating covariant and contravariant components

As observed in all examples, the change of signs of the space components of a covariant vector converts it to a contravariant one, and vice versa. The covariant and contravariant components thus contain the same information, and describe the same 4-vector. If a 4-vector X has covariant components Xm and contravariant components Xk , they can be related through Xm = xm k X , xk Xk = xk Xm . xm (1)

These matrices, which lower or raise the indices of the 4-vectors are the metrics: gkm = xk , xm g km = xk . xm (2)

Here in Special Relativity, gkm = g km = Diag(1, 1, 1, 1). At the moment, the metrics may be just considered as matrices. Later we shall discuss their tensorial nature.

Scalar products in terms of the metric

We saw earlier that the inner product (product with all indices summed over) of a covariant and a contravariant vector is invariant under Lorentz transformation: X m Ym = Xk Yk . Using Ym = gmk Yk , one may write this as X Y = gmk Xm Yk (3)

Thus one can talk about a scalar product X Y of two vectors X and Y, without referring to their components explicitly. The scalar product of two 4-vectors is invariant under Lorentz transformations, and hence is a frame-independent physical quantity.

Proper distance in special relativity

s2 = xm xm = gmk xm xk (4)

is the square of the length of the 4-vector x. The proper distance between two events x1 and x2 is s, with (s)2 = (x)i (x)i = (c t )2 (x )2 (y )2 (z )2 where x = x2 x1 . When (s)2 > 0, the two events are said to have a timelike separation: they are causally connected, i.e. one can inuence the other, and the time ordering of these events is the same in all frames. When (s)2 < 0, the two events are said to have a spacelike separation: they are not causally connected, and one cannot inuence the other. The time ordering of these events is frame-depedent. (s)2 = 0 is the null cone, or the light cone. Electromagnetic waves in vacuum travel on this cone. (5)

More Lorentz-invariant scalar products

Many Lorentz-invariant scalar products of 4-vectors will play an important role in our discussions. m m = is the DAlembertian. We have earlier explicitly veried its frame-dependence. pm pm = (E /c )2 |p|2 = m2 c 2 , clearly invariant since it is the mass of the particle. pm xm = Et p x = , the phase of a plane wave. We used this quantity while determining the aberration and Doppler shift.
(c ) m J m = (ct ) + J = 0 is the continuity condition / conservation of charge. (/c ) m Am = (ct ) + A = 0 is the Lorentz gauge condition, which was also a frame-independent statement, though we did not bother about it then.

Coming up...

Metric and invariant scalar products

Second rank 4-tensors: symmetric and antisymmetric

Higher-rank 4-tensors

Denition of a 4-tensor
A second rank 4-tensor B is the object that transforms under frame changes like the direct product of two 4-vectors. The contravariant components Bmn of B should transform as
n k B mn = m k B

(6)

and its covariant components Bmn should transform as km n Bk . Bmn = (7)

One can also express B in terms of its mixed components Bm n n or Bm , which satisfy appropriate transformation conditions. Bm m , the trace of B, is a Lorentz invariant. The s here are composed of proper rotations and boosts. Space reections are not included. However we know how elements of direct products of 4-vectors behave under space reections, since we know how 4-vectors behave under space reections. The tensors that behave like these are called proper tensors, the ones that do not are pseudotensors

, g Symmetric second-rank tensors: ,

Till now, we have treated to be simply a matrix. However note m n that m n = x / x transforms like mixed components of a 4-tensor, and hence is a 4-tensor. The dening equation for a 4-vector, X = X is then not simply a matrix equation, but a tensor equation.
m n m Similarly, n = x / x transform like mixed components of a is also a 4-tensor. 4-tensor, hence

Also, gmn = xm / xn and g mn = xm / xn transform like covariant and contravariant components of a 4-tensor, g. The m n m mixed components of this tensor are gm n = x / x = n . Since g is a tensor, the operation of raising and lowering indices in a 4-vector (or, by extension, a 4-tensor) is a tensor operation. All second rank tensors can be written as a sum of a symmetric and an antisymmetric tensor. The tensor g is always symmetric, are symmetric for only boosts, but are not so when while , rotations are involved. Now let us look at antisymmetric tensors.

Antisymmetric second-rank 4-tensor

An antisymmetric second-rank tensor B may be written in terms of its contravariant components as 0 B01 B02 B03 B01 0 B12 B13 . Bmn = B02 B12 0 B23 B03 B13 B23 0 There are six independent components, which transform as
n k B mn = m . k B

If we look at the transformation properties of the components B0 ( {1, 2, 3} are the space indices), one obtains
B 0 = 0 . B

Consider space rotations, for which 0 = 1 when = 0 and vanishes otherwise. Then one gets
0 B 0 = . B

Thus, the components B0 behave like a 3-vector under rotation. (Note: B is a proper tensor, not a pseudotensor.)

Antisymmetric second-rank tensor: general properties

Now let us look at the other components, of the form B . With the advantage of hindsight, consider the quantity 1 where is the completely antisymmetric, W = 2 B Levi-Civita 3-tensor. The transformation properties of W under space rotations can be obtained as: 1 1 W = B = B 2 2 1 1 B = = B 2 2 W . = Thus, W acts like a vector under proper rotations. Moreover, since B is a proper tensor and is a pseudotensor, W is a pseudovector / axial vector. Thus, the six independent components of the antisymmetric second-rank tensor can be separated into a 3-vector V = B0 = (Vx , Vy , Vz ), and a 3-axial vector W = B = (Wx , Wy , Wz ).

Vector axial vector antisymmetric tensor

In terms of the vector V and the axial vector W , one can write the components of Bij as 0 Vx Vy Vz Vx 0 Wz Wy . Bmn = Vy Wz 0 Wx Vz Wy Wx 0 This should remind us of the electromagnetic eld tensor F, whose components are written in terms of E, a vector, and B, an axial vector: 0 Ex Ey Ez Ex 0 Bz By . (8) Fmn = Ey Bz 0 Bx Ez By Bx 0 We shall later devote a complete lecture to the exploration of F.

Coming up...

Metric and invariant scalar products

Second rank 4-tensors: symmetric and antisymmetric

Higher-rank 4-tensors

Rank-n tensors and completely antisymmetric

Along the same lines as before, an nth rank tensor is an object whose components transform as the direct product of n 4-vectors. The tensors may be represented in terms of their contravariant, convariant, or mixed components. The 4-d analogue of the Levi-Civita tensor is , whose components are dened through 0123 = +1, and k mn = 1 depending on whether (k , , m, n) is an even or odd permutation of (0, 1, 2, 3). In all the other cases (i.e. when any two or more of k , , m, n are equal), k mn = 0. Components of do not change with proper Lorentz transformations. Moreover, they do not change sign even under space reections, which is what a proper tensor would be expected to do. Therefore, is a pseudotensor. Quantities that are made up from a product of tensors and a single are pseudotensors. We shall come across a few of them later in this course.

Dual tensors with

Because of its completely antisymmetric nature, plays an important role in creating dual tensors. For example,
k mn If X is a vector, X dened through Xk m = 1 Xn is a rank-3 2 pseudotensor, which is completely antisymmetric in its three indices.

If X is an antisymmetric rank-2 tensor, X dened through 1 k mn Xk = 2 Xmn is a rank-2 antisymmetric pesudotensor. If X is a completely antisymmetric rank-3 tensor, X dened k mn through Xk = 1 X mn is an axial vector. 2 The operation of forming a dual projects out the antisymmetric parts of the original tensors, and convert rank-n tensors to rank-(4 n) pseudotensors and vice versa. If the original tensor is completely antisymmetric, the dual contains the same information as the original tensor, but written in a form that transforms differently (tensor vs. pseudotensor, change of rank).

Take-home message from this lecture

Covariant and contravariant components of a 4-vector may be transformed into each other by the metric. Scalar products of 4-vectors are invariant under Lorentz transformations, and hence are frame-independent quantities. and the The Lorentz-transformation matrices and index-raising/lowering matrix g are actually symmetric rank-2 tensors. Totally antisymmetric rank-4 pseudotensor can be employed to create dual tensors.