Física del Cosmos: Lecture III

Tensors: the “alphabet” of GR

Kostas Glampedakis

2019-20
 This lecture

• Spacetime and coordinates.

• Index notation, summation convention.

• Tensors: definition and basic properties.

• The metric tensor.

• New derivatives: covariant and Lie derivative.

• Curvature and the Riemann tensor.

 Spacetime manifold
• Spacetime is the continuum of “events”.

• We assign coordinates at each event:

a 0 1
x = (x , x , . . . ) x↵

• These coordinates are just labels

(think of them as addresses in a city).

• We can relabel points without changing physical reality.

This coordinate transformation is of the form

X @ x̄↵
a a
x̄ = x̄ (x ) ) dx̄↵ = dx
@x
=0,1,..
 Index notation
• We are used to work with vectors, like the position and velocity.
An alternative “language” is that of coordinate components.
X
• For example: A·B! Aµ B µ
µ
• Einstein’s summation convention: repeated indices are summed
over all their values. In this example, µ is a dummy index:

X
µ
A Bµ = A Bµ µ Aµ B µ = A0 B 0 + A1 B 1 + A2 B 2 + A3 B 3
µ

• Note: Aµ Bµ 6= Aµ B⌫
X
• The same rule applies for objects with µ⌫ µ⌫
A⌫ B = A⌫ B
more than one index. For example:
⌫
 Vectors (I)
• Consider a curve given by xµ ( ) with λ “running” along the curve.
µ
• The function f (x µ
) is defined along the curve. dx
xµ ( )
µ = d
df @f dx⌫ u
⌫
= ⌫
= f ,⌫ u <latexit
sha1_base64="km0yhAIIHJKH1SUkAwrWjNy2qe4=">AAACCXicbVC7TsMwFHXKq5RXgJHFokJiqhKEBAtSBQtjkehDakLlOE5r1U4iPxBVlJWFX2FhACFW/oCNv8FpM0DLkSwdn3Pvte8JUkalcpxvq7K0vLK6Vl2vbWxube/Yu3sdmWiBSRsnLBG9AEnCaEzaiipGeqkgiAeMdIPxVeF374mQNIlv1SQlPkfDmEYUI2WkgQ31ncc1vIBeJBDOwofimmehx8yMEOVwYNedhjMFXCRuSeqgRGtgf3lhgjUnscIMSdl3nVT5GRKKYkbymqclSREeoyHpGxojTqSfTTfJ4ZFRQhglwpxYwan6uyNDXMoJD0wlR2ok571C/M/raxWd+xmNU61IjGcPRZpBlcAiFhhSQbBiE0MQFtT8FeIRMpEoE17NhODOr7xIOicN12m4N6f15mUZRxUcgENwDFxwBprgGrRAG2DwCJ7BK3iznqwX6936mJVWrLJnH/yB9fkDV5maGg==</latexit>

d @x d
µ
• The “velocity” u is tangent to the curve at each point.

µ ⌫
• Now change to new coordinates x̄ (x ) . The new tangent vector is:
µ µ ⌫ µ
dx̄ @ x̄ dx @ x̄
ūµ = ūµ = = u ⌫
d @x⌫ d @x⌫

@f µ @f ⌫
• It is easy to show that: µ
ū = ⌫
u = coordinate invariant
@ x̄ @x
 Vectors (II)
• A vector (or contravariant vector) is an object which under a coordinate
transformation behaves like:
µ
@ x̄
Āµ = A ⌫
@x⌫
• A dual vector (or covariant vector) is an object that behaves like:

@x⌫
Āµ = µ
A ⌫
@ x̄
• The scalar combination of a contravariant and a
µ µ
covariant vector is coordinate independent (=invariant): <latexit sha1_base64="rA+DwRjO0d/ozG465RytCu8A+5A=">AAAB7nicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE0ItQ9OIxgv2ANpTNdtIu3WzC7kYooT/CiwdFvPp7vPlv3LY5aOuDgcd7M8zMC1PBtXHdb6e0tr6xuVXeruzs7u0fVA+PWjrJFMMmS0SiOiHVKLjEpuFGYCdVSONQYDsc38389hMqzRP5aCYpBjEdSh5xRo2V2j1/xMkN6Vdrbt2dg6wSryA1KOD3q1+9QcKyGKVhgmrd9dzUBDlVhjOB00ov05hSNqZD7FoqaYw6yOfnTsmZVQYkSpQtachc/T2R01jrSRzazpiakV72ZuJ/Xjcz0XWQc5lmBiVbLIoyQUxCZr+TAVfIjJhYQpni9lbCRlRRZmxCFRuCt/zyKmld1D237j1c1hq3RRxlOIFTOAcPrqAB9+BDExiM4Rle4c1JnRfn3flYtJacYuYY/sD5/AEUA466</latexit>
sha1_base64="hP+6LrUf2d3tZaldqaQQvEKMXyw=">AAAB2XicbZDNSgMxFIXv1L86Vq1rN8EiuCozbnQpuHFZwbZCO5RM5k4bmskMyR2hDH0BF25EfC93vo3pz0JbDwQ+zknIvSculLQUBN9ebWd3b/+gfugfNfzjk9Nmo2fz0gjsilzl5jnmFpXU2CVJCp8LgzyLFfbj6f0i77+gsTLXTzQrMMr4WMtUCk7O6oyaraAdLMW2IVxDC9YaNb+GSS7KDDUJxa0dhEFBUcUNSaFw7g9LiwUXUz7GgUPNM7RRtRxzzi6dk7A0N+5oYkv394uKZ9bOstjdzDhN7Ga2MP/LBiWlt1EldVESarH6KC0Vo5wtdmaJNChIzRxwYaSblYkJN1yQa8Z3HYSbG29D77odBu3wMYA6nMMFXEEIN3AHD9CBLghI4BXevYn35n2suqp569LO4I+8zx84xIo4</latexit>
sha1_base64="4IaEcXDFmnvKhz7swkp3NkXx6nc=">AAAB43icbZBLSwMxFIXv1FetVatbN8EiuCozbnQjCG5cVrAPaIeSSe+0oZnMkNwRytAf4caFIv4nd/4b08dCWw8EPs5JyL0nypS05PvfXmlre2d3r7xfOageHh3XTqptm+ZGYEukKjXdiFtUUmOLJCnsZgZ5EinsRJP7ed55RmNlqp9ommGY8JGWsRScnNXpN8eS3bJBre43/IXYJgQrqMNKzUHtqz9MRZ6gJqG4tb3AzygsuCEpFM4q/dxixsWEj7DnUPMEbVgsxp2xC+cMWZwadzSxhfv7RcETa6dJ5G4mnMZ2PZub/2W9nOKbsJA6ywm1WH4U54pRyua7s6E0KEhNHXBhpJuViTE3XJBrqOJKCNZX3oT2VSPwG8GjD2U4g3O4hACu4Q4eoAktEDCBF3iDdy/zXr2PZV0lb9XbKfyR9/kD7TiNYQ==</latexit>
sha1_base64="I78jhN8II+nF/LbC7awfuqU9YwM=">AAAB7nicbVA9SwNBEJ3zM8avqKXNYhCswp2NNkLQxjKC+YDkCHubvWTJ3u6xOyeEIz/CxkIRW3+Pnf/GTXKFJj4YeLw3w8y8KJXCou9/e2vrG5tb26Wd8u7e/sFh5ei4ZXVmGG8yLbXpRNRyKRRvokDJO6nhNIkkb0fju5nffuLGCq0ecZLyMKFDJWLBKDqp3WuMBLkh/UrVr/lzkFUSFKQKBRr9yldvoFmWcIVMUmu7gZ9imFODgkk+Lfcyy1PKxnTIu44qmnAb5vNzp+TcKQMSa+NKIZmrvydymlg7SSLXmVAc2WVvJv7ndTOMr8NcqDRDrthiUZxJgprMficDYThDOXGEMiPcrYSNqKEMXUJlF0Kw/PIqaV3WAr8WPPjV+m0RRwlO4QwuIIArqMM9NKAJDMbwDK/w5qXei/fufSxa17xi5gT+wPv8ARLDjrY=</latexit>
= Ā B̄ µ = A Bµ

• In a n-dimensional spacetime a vector has n components. For example,

in special relativity’s 4-D Minkowski spacetime we have:
Aµ = (A0 , A1 , A2 , A3 )
 Tensors (I)
• Apart from vectors there are objects with more than one index
called tensors. A tensor of type ( k, l ) is an object:

T µ1 . . . µk⌫1 . . . ⌫l
µ
• Example: T⌫ is a type (1,2) tensor.
By definition, vectors are type (1,0) tensors
µ ⌫
and dual vectors are type (0,1) tensors. µ⌫ @ x̄ @ x̄ ⇢
T̄ = ⇢
T
@x @x
• Rank of a tensor is the sum k + l .
@x⇢ @x
T̄µ⌫ = µ ⌫
T ⇢
• Tensors are defined by the transformation rules: @ x̄ @ x̄

µ @ x̄µ @x ⇢
T̄ ⌫ = T
@x⇢ @ x̄⌫
 Tensors (II)
µ1 . . . µ k
• More generally, a tensor of type ( k,l ) is an object T ⌫1 . . . ⌫ l
which transform as:

@ x̄µ1 @ x̄µk @x 1 @x l ⇢1 . . . ⇢k
T̄ µ1 . . . µk⌫1 . . . ⌫l = . . . . . . T 1. ..
@x⇢1 @x⇢k @ x̄⌫1 @ x̄⌫l l

{
{
upper indices lower indices

• Location matters: the location of indices in a tensorial symbol is

important. For example:
a a
T 6= T
 Tensor algebra
• Tensors of the same type can be summed, at the same given point.
The sum is a tensor of the same type. Example:

↵ ↵ ↵ ↵ ↵
F +G =N F +A

• The product of a type (k,l) tensor with a type (m,n) tensor is a type
(k+m, l+n) tensor. Example (and exercise):

↵ ↵
A B =H
• Comment: in all tensorial equations we need to make sure that the same
free indices appear on both sides of the equation!

↵ ↵
A B =H
↵ ↵
A B =H
A↵ B
<latexit sha1_base64="+mM7l7C3Cg/+RZw5/cJHHjj0+8A=">AAACHXicbVDLSgMxFM3UV62vqks3wSK4KjMi6EaoFcRlBfuATjvcSdM2NJkZkoxQhvohbvwVNy4UceFG/BszbRVtPRA4nHMuN/f4EWdK2/anlVlYXFpeya7m1tY3Nrfy2zs1FcaS0CoJeSgbPijKWUCrmmlOG5GkIHxO6/7gIvXrt1QqFgY3ehjRloBewLqMgDaSlz8+bycu8KgPbg+EgBEut12favCSu2/lDF/+hFJr5OULdtEeA88TZ0oKaIqKl393OyGJBQ004aBU07Ej3UpAakY4HeXcWNEIyAB6tGloAIKqVjK+boQPjNLB3VCaF2g8Vn9PJCCUGgrfJAXovpr1UvE/rxnr7mkrYUEUaxqQyaJuzLEOcVoV7jBJieZDQ4BIZv6KSR8kEG0KzZkSnNmT50ntqOjYRef6uFAqT+vIoj20jw6Rg05QCV2hCqoigu7RI3pGL9aD9WS9Wm+TaMaazuyiP7A+vgDor6Jq</latexit>
= F↵
 Some tensor properties
• Contraction [in this example a (3,2) tensor becomes (3-1,2-1)]:

µ⌫ µ⌫
T a =K a

• Symmetric tensor: Tµ⌫ = T⌫µ Anti-symmetric tensor : Tµ⌫ = T⌫µ

• Symmetric and anti-symmetric parts:

1 1
T(µ⌫) = (Tµ⌫ + T⌫µ ) T[µ⌫] = (Tµ⌫ T⌫µ )
2 2
• An arbitrary tensor can be decomposed into symmetric and antisymmetric
parts. For example, a rank 2 tensor can be written as:

Tµ⌫ = T(µ⌫) + T[µ⌫]

 The metric tensor
• A fundamental rank 2 tensor is the metric gµ⌫ .

• In General Relativity this tensor represents the gravitational field.

• It also appears in the invariant spacetime line element:

2 a
ds = ga dx dx
• With the help of the metric we can raise/lower indices:

⌫ µ µ⌫
Aµ = gµ⌫ A A =g A⌫

aµ a
• The inverse metric is defined as: g gµ =
why?
• The metric is symmetric: ga = g a (=10 independent components in 4D)
 Worldlines & proper time
• A body’s worldline its trajectory in spacetime and is parametrised as:
↵
x ( )
• The proper time of a particle (or observer) is the time measured by an ideal
clock comoving with the particle. It is an invariant quantity given by
(note that this is valid for timelike worldlines only):

2 2 2 xµ (⌧ )
ds = c d⌧ dx↵ ↵
u =
d⌧
<latexit sha1_base64="UMpUo7NK/l0lO29zMVMMr9HZk5A=">AAACC3icbVC7SgNBFJ31GeNr1dJmSBCswq4I2ghBG8sI5gHZGO7OziZDZh/MQwzL9jb+io2FIrb+gJ1/4yTZQhMPXDiccy/33uOnnEnlON/W0vLK6tp6aaO8ubW9s2vv7bdkogWhTZLwRHR8kJSzmDYVU5x2UkEh8jlt+6Orid++p0KyJL5V45T2IhjELGQElJH6dkXfecDTIeAL7IUCSBY8FEqeBZ4CnfftqlNzpsCLxC1IFRVo9O0vL0iIjmisCAcpu66Tql4GQjHCaV72tKQpkBEMaNfQGCIqe9n0lxwfGSXAYSJMxQpP1d8TGURSjiPfdEaghnLem4j/eV2twvNexuJUKxqT2aJQc6wSPAkGB0xQovjYECCCmVsxGYJJRJn4yiYEd/7lRdI6qblOzb05rdYvizhK6BBV0DFy0Rmqo2vUQE1E0CN6Rq/ozXqyXqx362PWumQVMwfoD6zPH1fgmz8=</latexit>

• We can then parametrise

particle worldlines in terms
of proper time τ: x↵ (⌧ )

• Four-velocity normalisation (c=1):

2 µ ⌫ µ ⌫ µ
<latexit sha1_base64="I0tsf0+1hgmIQ1lcF/AmNx1ef1g=">AAACGnicbZDLSgMxFIYzXmu9VV26CRbBjWWmCLoRim5cVrEX6LQlk0nb0ExmSE7UMvQ53Pgqblwo4k7c+Daml4W2Hgj5+P9zSM4fJIJrcN1vZ2FxaXllNbOWXd/Y3NrO7exWdWwUZRUai1jVA6KZ4JJVgINg9UQxEgWC1YL+5civ3TGleSxvYZCwZkS6knc4JWClds47xqEPxLSK+Bx326kfGV+aIQ4fWhbHlzTYv+HdHhCl4nvczuXdgjsuPA/eFPJoWuV27tMPY2oiJoEKonXDcxNopkQBp4INs77RLCG0T7qsYVGSiOlmOl5tiA+tEuJOrOyRgMfq74mURFoPosB2RgR6etYbif95DQOds2bKZWKASTp5qGMEhhiPcsIhV4yCGFggVHH7V0x7RBEKNs2sDcGbXXkeqsWC5xa865N86WIaRwbtowN0hDx0ikroCpVRBVH0iJ7RK3pznpwX5935mLQuONOZPfSnnK8f3F6gKg==</latexit>
d⌧ = gµ⌫ dx dx ) g µ⌫ u u = u µ u =
<latexit sha1_base64="k+8TLyhSdDk4SUmmfddwwLsn2iE=">AAACEXicbVDLSgMxFM3UV62vUZdugkXoxjIjgm4KRTcuK9gHtOOQSdM2NMkMeQhl6C+48VfcuFDErTt3/o2ZdhbaeiDh5Jx7ubknShhV2vO+ncLK6tr6RnGztLW9s7vn7h+0VGwkJk0cs1h2IqQIo4I0NdWMdBJJEI8YaUfj68xvPxCpaCzu9CQhAUdDQQcUI22l0K0Mw7THTU+YKTT3lmW3MLAGTZi/uKnBUz90y17VmwEuEz8nZZCjEbpfvX6MDSdCY4aU6vpeooMUSU0xI9NSzyiSIDxGQ9K1VCBOVJDONprCE6v04SCW9ggNZ+rvjhRxpSY8spUc6ZFa9DLxP69r9OAySKlIjCYCzwcNDIM6hlk8sE8lwZpNLEFYUvtXiEdIIqxtiCUbgr+48jJpnVV9r+rfnpfrV3kcRXAEjkEF+OAC1MENaIAmwOARPINX8OY8OS/Ou/MxLy04ec8h+APn8wfVO5xU</latexit>
1
 The metric from the EP
• The existence of a dynamical spacetime metric can be viewed as a
consequence of the Equivalence Principle (EP). Locally, at any spacetime
point (=event) we can set up Minkowski coordinates ⇠ µ so that:

ds2 = ⌘µ⌫ d⇠ µ d⇠ ⌫

• The transformation to the “global” coordinates x↵ is:

µ
@⇠
⇠ µ (x↵ ) ) d⇠ µ = dx ↵
@x↵
• Given that ds is invariant, we have:
✓ µ ⌫
◆
2 @⇠ @⇠
ds = ⌘µ⌫ ↵ dx↵ dx = g↵ dx↵ dx
@x @x
{ a function of xµ
{
<latexit sha1_base64="RDBS5kvTUaa9O146kL+pJ8t3K/M=">AAAB7HicbVDLSgNBEOz1GeMr6tHLYBA8hV0R9Bj04jGCmwSSNcxOZpMh81hmZsWw5Bu8eFDEqx/kzb9xkuxBEwsaiqpuurvilDNjff/bW1ldW9/YLG2Vt3d29/YrB4dNozJNaEgUV7odY0M5kzS0zHLaTjXFIua0FY9upn7rkWrDlLy345RGAg8kSxjB1knh00NXZL1K1a/5M6BlEhSkCgUavcpXt69IJqi0hGNjOoGf2ijH2jLC6aTczQxNMRnhAe04KrGgJspnx07QqVP6KFHalbRopv6eyLEwZixi1ymwHZpFbyr+53Uym1xFOZNpZqkk80VJxpFVaPo56jNNieVjRzDRzN2KyBBrTKzLp+xCCBZfXibN81rg14K7i2r9uoijBMdwAmcQwCXU4RYaEAIBBs/wCm+e9F68d+9j3rriFTNH8Afe5w/ukY7A</latexit>

tensor rule for the metric

 The covariant derivative
• The usual differential is not tensorial Aµ (Q)
AµT (P )
Aµ (P )
a a a xµ ( )
dA ⌘ A (Q) A (P ) = Aa, dx
a a
Q: x + dx
• We need to build a coordinate-invariant
(tensorial) derivative. a
P: x

• Imagine that the vector is “parallel transported” from Q to P

so that the difference at P is tensorial:

DAa = AaT (P ) Aa (P )

• We assume:
Aa = AaT (P ) Aa (Q) = a
µ Aµ dx

Connection
 Covariant derivative & parallel transport
• Then,
a a a µ DAa a
DA = A, dx + µ A dx !
) =A; u
d
• The covariant derivative a a is:
A; =r A

a a a µ a a a µ
r A =@ A + µ A or A; = A, + µ A

• The covariant derivative is a tensor (although Aa, , aµ are not tensors).

• Parallel-transported tensor along a curve:

DAa a
=A; u =0
d
 More covariant derivatives
• Formulae for the covariant derivative of a dual vector and a scalar:

µ
F; = F , Aa; = Aa, a Aµ

• Covariant derivatives of rank 2 tensors:

Ta ; = Ta , + a
µ Tµ + µ T aµ
µ µ
T↵ ; = T↵ , ↵ Tµ T↵µ
a a a µ µ a
T ; =T , + µ T Tµ
 The Christoffel symbols
• So far the connection was arbitrary. A specific choice is made by demanding
that it be symmetric and metric compatible:

a a
= , ga ; =0

• This choice is dictated by the EP, more specifically from the fact that in a LIF :

⌘µ⌫, = 0

• The connection becomes the “Christoffel symbol” (exercise):

⇢ 1 ⇢
µ⌫ = g ( gµ ,⌫ + g⌫ ,µ gµ⌫, )
2
 The Lie derivative (I)
• Considering again the points P and Q Aµ (Q)
along a curve x↵ ( ) :
Aµ (P )
0
xµ ( )
↵ ↵ ↵ ↵ ↵ uµ
x = x + dx = x + u d
Q: xa + dxa
• Treat this as a coordinate transformation: a
P: x
0
µ
0
µ 0 @x µ µ
A (x ) = A = A (x) + u , A (x)d
@x
Then:
0 µ
µ µ
A (Q) = A (P ) + u, A (P )d ( “dragged-along” tensor )

µ µ µ
A (Q) = A (P ) + u A, (P )d
 The Lie derivative (II)
• Define the Lie derivative:
Aµ (Q) A0µ (Q)
Lu Aµ (P ) =
d

• This is: µ µ µ
Lu A = u A, u, A

• Normal derivatives can be changed to covariant derivatives (exercise):

L u Aµ = u r Aµ r uµ A

• Other cases: L u f = u r f = u f,
L u Bµ = u r Bµ + rµ u B
 The Killing equation & vectors
• For a rank-2 tensor:

Lu Tµ⌫ = u r Tµ⌫ + rµ u T ⌫ + r⌫ u Tµ
= u Tµ⌫, + u ,µ T ⌫ +u ,⌫ Tµ

↵
• The Lie-derivative of the metric (with respect to a displacement field ):
⇠

L⇠ gµ⌫ = rµ ⇠⌫ + r⌫ ⇠µ

• Killing equation/vectors:

L⇠ gµ⌫ = rµ ⇠⌫ + r⌫ ⇠µ = 0

• The Killing vectors are associated with spacetime symmetries and

conservation laws.
 Lie transportation
• A tensor is “Lie-transported” when its Lie derivative is = 0. For example:
↵
Lu A = 0

• Choose coordinates so that the curve of the previous example has x0 =

and xi = constant.

↵ ↵ ↵
Then: u = 0 !
) u, =0

• This means that:

@A↵
L u A↵ = 0 )
! 0
=0
@x
• Lie-transportation , conserved quantities (in this example A↵ is time-
independent).
 Curvature: the Riemann tensor
• Covariant derivatives do not commute:

µ µ µ ⌫
r⇢ r A r r⇢ A = R ⌫⇢ A

µ
• This relation can be used to define the Riemann curvature tensor R ⌫⇢ .

• Ricci tensor and Ricci scalar:

µ µ⌫
R⌫ = R ⌫µ R=g Rµ⌫
 The Riemann tensor
• From its defining relation we can write the Riemann tensor as:

µ µ µ µ  µ 
R ⌫⇢ = ⌫ ,⇢ ⌫⇢, + ⇢ ⌫  ⌫⇢

• Properties of the Riemann tensor:

Rµ⌫⇢ = R⌫µ⇢ = Rµ⌫ ⇢ = R⇢ µ⌫

{
{
Rµ⌫⇢ + Rµ ⌫⇢ + Rµ⇢ ⌫ =0
{

• The Bianchi identities: {

Rµ⌫⇢ ; + Rµ⌫⇢; + Rµ⌫ ;⇢
{ =0
{
{
 Why tensors?

• General Relativity is a theory designed to be invariant for all observers

(=coordinate frames). This is called “general covariance”.

• A tensorial equation has the key property of remaining unchanged in all

coordinate frames (although the components of a given tensor may
change in different frames). For example:

a a , a a
G = T $ Ḡ = T̄

• The mathematics of GR is similar to that of differential geometry of

Riemannian 3-D spaces. The spacetime of GR is a 4-D pseudo-
Riemannian (or “Lorentzian”) space.
 Textbook reading

• Hartle’s book: Chapters 7 (exercises 7-9, 14, 18) & 20 (exercises 10, 12, 13)

• D’Inverno’s book: Chapters 5 (exercises 3-5, 8-13), 6 (exercises 3-22, 25, 31)
, & 7 (exercises 8, 10, 11, 13, 14)