Professional Documents
Culture Documents
Raghu Mahajan
Contents
1 Bending of light 3
2 Pound-Rebka experiment 3
3 Basic GR 4
5 Geodesics 9
7 Killing Vectors 12
8 ADM Mass 13
1
11 Energy conditions 16
12 Frame formulation 16
13 Hamiltonian Formulation 18
15 Gravitational Waves 25
16 Geodesic Congruences 26
17.1 2D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
17.2 3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
19 Vaidya Metric 31
20 Bondi Mass 33
22 Induced Gravity 34
23.1 Schwarzschild . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2
1 Bending of light
How much velocity does light pick up downwards as it coasts by the sun in the horizontal direction?
2 Pound-Rebka experiment
https://en.wikipedia.org/wiki/Pound%E2%80%93Rebka_experiment
Retreived on 2017-11-02
1. Consider two copies of an electron-atom system, one in the excited state (the emitter), the
other in the lower energy state (the receiver).
2. The emitter was placed at the top of tower with the receiver at the bottom. General relativity
predicts that the gravitational field of the Earth will cause a photon emitted downwards
(towards the Earth) to be blueshifted.
3. To counteract the effect of gravitational blueshift, the emitter was moved upwards (away from
the receiver) causing the photon frequency to be redshifted, according to the Doppler shift.
5. Mössbauer spectroscopy. All atoms in a solid lattice absorb the recoil energy when a single
atom in the lattice emits a gamma ray. Therefore, the emitting atom will move very little
(just as a cannon will not produce a large recoil when it is braced, e.g. with sandbags).
3
6. The result confirmed that the predictions of general relativity were borne out at the 10%
level. This was later improved to better than the 1% level by Pound and Snider. Another
test involving a space-borne hydrogen maser increased the accuracy to about 10−4 .
3 Basic GR
r
~G
`p =
c3
κ2 = 8πG
1
Γµab = g µc ∂a gbc + ∂b gac − ∂c gab
2
R bcd = ∂c Γabd + Γµbd Γaµc − (c ↔ d)
a
Rab = R∗a∗b
1
Gab = Rab − gab R = (8πG)Tab
2
2 2
Wabµν = Rabµν − ga[µ Rν]b − (b ↔ a) + g g R
D−2 (D − 1)(D − 2) a[µ ν]b
W ∗b∗ν = 0
Wabµν = 0 when D = 3
The only independent contractions that one needs to check for the Riemann and the Weyl tensors
is between the first and the third indices.
Ra[bµν] = 0
D[c Rab]µν = 0
4
D2 (D2 − 1)
Independent Riemann Components:
12
The Weyl tensor is zero in three dimensions, since the number of components of Riemann is six,
which is the same as the number of components of the Ricci tensor. The Riemann tensor Rabcd has
to do with two planes (ab) and (cd), which in three dimensions, is the same as two axes. When
we parallel transport, one of these two planes is the one that contains the two directions of the
parallelogram, and the other two have to do with the plane in which the vector got rotated. This
also means that in the three-dimensional vacuum, the field equation Rab = 0 implies Rabcd = 0, and
thus there is no gravitational radiation.
R
Maximally Symmetric ⇐= Rabµν = gaµ gbν − (µ ↔ ν)
D(D − 1)
R
=⇒ Rbν = gbν
D
R
Einstein Manifold ⇐= Rbν = gbν
D
δV 3 = R3∗12 V ∗ A1 B 2
D1 , D2 V 3 = R3∗12 V ∗
δθ = R δa,
where δθ is the angle by which the vector rotates as you take it around a small loop of oriented area
δa. Note that this also determines the sign of the cruvature, so it would be different to a sphere vs
a hyperboloid. The simplest place to look for this is on a cone, where all the curvature is at the
tip. Cones with angular excess have negative curvature, and cones with angle deficit has positive
curvature.
D2 µ
Geodesic deviation: 2
S = Rµabc T a T b S c
dt
5
p
1 δ |g| 1
p µν
= − gµν
|g| δg 2
∂a g
g ∗? ∂a g∗? =
g
∂a g
Γ∗∗a =
2g
1
D∗ V ∗ = p ∂∗ |g| V ∗
p
|g|
−2 δS
Tµν = p
|g| δg µν
For Euclidean signature, flip overall sign
Z Z
1 p p
(8πG)S = |g| (R − 2Λ) + |γ| K
2 M ∂M
1
Rµν − gµν R − 2Λ = κ2 Tµν
2
2 2 1
Rµν − Λgµν = κ Tµν − gµν T
D−2 D−2
D−2
− R + DΛ = κ2 T
2
One way to motivate the Einstein equations is to use the Newtonian equation ∇2 φ = 4πGρ and note
that in the weak-field limit φ is related to gtt , and ρ is Ttt . Thus, this says that the second-derivative
of the metric must be related to the stress tensor.
6
Z
D
p 1 2
S[φ] = − d x |g| (∂φ) + V (φ)
2
For Euclidean signature, flip overall sign
1 2
Tµν [φ] = ∂µ φ∂ν φ − gµν (∂φ) + V (φ)
2
D−2
Z
1 p h i
S[g, φ] = − dD x |g| (∂φ)2 + Bφ2 R , B=
2 4(D − 1)
7
4 Some algebraic manipulations
1
∂µ g = g ab ∂µ gab
g
∇a ξb + ∇b ξa = ∂a ξb + ∂b ξa − 2Γ? ab ξ?
8
5 Geodesics
Z
1 1 µ 2
S= dτ Ẋ Ẋµ − η m
2 η
Z q
= −m dτ −Ẋ µ Ẋµ (if m 6= 0)
Ẋ 2 = −η 2 m2
dX ∗
d 2 1 ∂g∗? ∗ ?
gµ∗ = Ẋ Ẋ
dτ η dτ η ∂X µ
Affine parameterization: η̇ = 0
1
Massive affine parameterization: η =
m
In a weak gravity field when particles are moving slowly, the only Christoffel term that is important
is Γi00 . This tells us that the gravitational acceleration is given by these Christoffel symbols. As
expected, they are the first derivatives of the gravitational potential which is encoded in the metric.
9
6 Black Hole metrics
10
dr2
ds2 = −f dt2 + , (dr = f dr∗ )
f
= −f (dt2 − dr∗2 ), (v = t + r∗ )
= −f dv 2 + 2 dv dr
κ f0
T = , κ=
2π 2
A
SBH =
4`2p
Near the horizon of a general Schwarzschild type black hole, the emblackening factor has a simple
zero.
f (r) ≈ 2κ(r − rH )
1
r∗ ≈ log(r − rH )
2κ
U V = − exp[2κr∗ ] ≈ −(r − rH )
ds2 = −f dudv
f
= dU dV
κ2 U V
11
7 Killing Vectors
Dµ ξ ν + Dν ξ µ = 0
D1 D2 ξ 3 = R321∗ ξ ∗
D∗ D2 ξ ∗ = R2∗ ξ ∗
Z Z
c
Qξ (Σ) = 2 dS∗? D∗ ξ ? = dS∗ J ∗
2κ ∂Σ Σ
µ c µ∗
J (ξ) = 2 R ξ∗
κ
D∗ J ∗ = 0
12
8 ADM Mass
Dµ k ν + Dν k µ = O(r−3 )
Let us give more detail about how to go from the first expression to the second.
dSµν Dµ k ν = DSi0 Di k 0
Di k 0 ≈ g ij Γ0 0j ≈ Γ0 0j
1
Γ0 0j = g 0µ (−g0j,µ + g0µ,j + gµj,0 ) only µ = 0 is relevant
2
1
≈ − (−g0j,0 + g00,j + g0j,0 )
2
1
= − ∂j h00
2
d 2GM 2GM
ni ∂i h00 = =− 2
dr r r
1 2 2GM
EADM =− × (4πr ) × − 2 =M
8πG r
The final expression in terms of the spatial components of the metric can be obtained by looking
at the 00-component of the Einstein equation, together with the trace of Einstein’s equations.
An even more down-to-earth way of thinking about the ADM mass is to think of the Poisson
equation ∇2 φ = 4πG ρ, and recall that h00 = −2φ. This gives
Z Z
1
4πG M = dΩ ni ∂i φ = − dΩ ni ∂i h00 .
2
13
9 Null hypersurfaces and Killing horizons
ω[ 1 D 2 ω3 ] = 0
For a null hypersurface N , the normal nµ is null. This implies that nµ is also a tangent vector
to N . One can show that the integral curves of nµ are geodesics.
If it so happens that a Killing vector ξ is normal to N , we say that N is a Killing horizon for
ξ. Now we can pick a coordinate v on N such that ξ = ∂/∂v. The relation of v to affine parameter
V of the geodesics is V = eκv .
ξ · D ξ µ =: κξ µ .
This relation holds because the integral curves of ξ are geodesics. If these geodesics were affinely
parameterized via v, the right hand side would be zero; but they are not affinely parameterized by
v, the affine parameter is V . Let us check this for Schwarzschild, where the Killing vector is ξ = ∂v .
M M (r − 2M )
Dv ξ µ = ∂v + ∂r
r2 r3
1
= ∂v (on the horizon)
4M
14
10 Light-like particle motion
Coordinates (U, V, . . .)
U =T −X T = (V + U )/2
V =T +X X = (V − U )/2
∂U = (∂T − ∂X )/2 ∂T = ∂V + ∂U
∂V = (∂T + ∂X )/2 ∂X = ∂V − ∂U
0 −1/2 0 −2
g?? = , g ?? =
−1/2 0 −2 0
Surface: U = constant
n? = (1, 0, 0, 0) = dU = dT − dX
15
11 Energy conditions
1. Strong energy condition. tRt ≥ 0, or (Tµν − gµν T /(D − 2))tµ tν ≥ 0. Violated by positive cc,
or a scalar field.
3. Dominant energy condition. Weak energy condition, and also that −T µ ? k ? should be a future
pointing vector which is timelike or null.
4. Null energy condition. kT k ≥ 0 for all future-pointing null vectors k. Violated by quantum
effects in a black hole background. Also violated by the Casimir effect.
12 Frame formulation
Frame 1-forms ea
If Γ(g) is the usual Christoffel connection that we use in torsion-free general relativity, then
Γ = Γ(g) − K
Here Γ is the actuall connection under which the metric is covariantly constant.
h i h i
R12ab := ∂1 ω2ab − 1 ↔ 2 + ω1a∗ ω2?b η ∗? − 1 ↔ 2
16
de + ω ∧ e = T
dω + ω ∧ ω = R
17
13 Hamiltonian Formulation
Denote the spacetime manifold by M. Define a function t on M such that the surfaces of constant
t are spacelike hypersurfaces. Denote these hypersurfaces by Σt . Also define a time like vector
field ta which represents the flow of time. The vector field ta is such that ta ∇a t = 1. This is a
normalization condition that says that the rate of change of the time function t along ta is unity.
P ab = δ ab − σna nb , σ = −1
N := σ n · t (1)
The logic for determining the various vector fields in component form is as follows
1
• n? = 1 2 3
N (1, −N , −N , −N ). This is implied by (2), na = (ta − N a )/N .
• hab nb = 0 implies that ha0 = 0. This implies that g a0 = −na n0 . This allows us to fix uniquely
18
the inverse metric g ab = hab + σna nb .
−1/N 2 N i /N 2 0 0
g ?? = , h?? =
hij − N i N j /N 2 hij
• To get the shift with indices lowered, use (3), Na = hab tb = ha0 = (Ni N i , Ni ).
19
The extrinsic curvature can (almost) be defined as the covariant derivative of the normal vector.
But this is not good, because we want the extrinsic curvature tensor to have only indices that are
tangent to the hypersurface. So we project.
Kab := ha c ∇c nb
= ∇a nb + na nc ∇c nb
This is not explicitly symmetric in a and b. We would like to show that. The idea is to use Frobenius
theorem (contracted once with the normal). We will also use the fact that na na = 1, and hence
n∗ ∇b n∗ = 0, and also n∗ ∇b n∗ = 0.
−nc na ∇c nb + nc nb ∇c na − ∇a nb + ∇b na = 0
∇a nb + nc na ∇c nb = ∇b na + nc nb ∇c na
Kab = Kba
Ln na = nc ∇c na + nc ∇a nc = nc ∇c na
Ln hab = Ln gab + na Ln nb + nb Ln na
= (∇a nb + ∇b na ) + na nc ∇c nb + nb nc ∇c na
Now, let us deal with the extrinsic curvature tensor. Let us denote by D the covariant derivative
built out of g and by ∇ the covariant derivative built out of h.
20
The Einstein-Hilbert term can be massaged into the following form
p √ h i
|g| R = h N R + Kab K ab − K 2 + total derivative.
When we substitute for Kij from the above expression, we see that the time derivatives of neither
the lapse N , nor the shift Ni appear.
∂L √
π ab = = h (K ab − Khab ).
∂ ḣab
If one wants to use boundary conditions that preserve asymptotic flatness, then the boundary term
in the Hamiltonian is precisely the ADM energy.
21
We should take the configuration space of general relativity to be the set if equivalence classes
of Riemannian metrics on Σt . This configuration space is known as superspace. The momentum
constraints are eliminated by the choice of superspace as the configuration space.
The energy constraint, however, remains. This is the sense in which GR is different from gauge
theory. Thus, it does not appear possible to find a choice of configuration space for GR such that
only the “true dynamical degrees of freedom” are present in its phase space. This provides a serious
obstacle to the formulation of quantum gravity by the canonical quantization approach.
22
14 The Gibbons-Hawking term
1
δΓabc = g aµ Db δgµc + Dc δgµb − Dµ δgbc
2
n2 = σ ∈ {+1, −1}
n1 n2 g 12 = σ =⇒ δnµ = αnµ
σ σ
=⇒ α = − na nb δg ab = na nb δgab
2 2
P ab = δ ab − σna nb
δK = Kab δg ab + g ab δKab
Now we write σnb nα = δ αb − P αb . The term with the Kroncker delta gives us two terms. One
is K 12 δg12 , and the other combines with the second term. The guy with P αb is a total intrinsic
covariant derivative, so it vanishes when integrated over ∂M. Thus, we get
1 1 1
δK = Kab δg ab + K ab δgab − na Db δgab + nα Dα g ab δgab
2 2 2
1 1 1 α
= Kab δg + na Db δg − n Dα gab δg ab
ab ab
(5)
2 2 2
23
So we see that if the K term is taken with twice the coefficient, the second and third terms in (5)
√
cancel those in (4). The first term combined with the variation of the γ gives us the Brown-York
stress tensor.
√
Z
1
total variation = dD−1 x γ Kab − γab K δγ ab
16πG δM
24
15 Gravitational Waves
R = ∂∗ ∂? h∗? − h
k ? = (ω, 0, 0, ω)
Rµν = 0
h i
k 2 δµa δνb + kµ kν η ab − kµ k a δνb − kν k a δµb ab = 0 (6)
There are four gauge transformations which can shift ∗∗ by the following matrices
−2 1 −1 −1 −1
−1 1
, , ,
−1 1
1 1 1 −1 2
Use the first and the fourth gauge transformation to set a = 0 and j = 0. Use the second to set
b − g = 0 and the third to set c − i = 0. Thus the most general form of the polarization after fixing
25
the gauge is
0 b c d
e f b
∗∗ =
h c
0
Just like we did in the E&M case, the equations of motion imply that
• Using the first two items, look at the original equation of motion (6). Get = 0 =⇒ e+h = 0.
(D − 1)(D − 2)
In D dimensions number of graviton modes = −1
2
D= 2 3 4 5 6 7 8 9 10 11
0 0 2 5 9 14 20 27 35 44
16 Geodesic Congruences
Four vector fields t, ηα with α ∈ {0, 1, 2}. They all commute with each other.
t·Dη = η ·Dt
26
For the case of null geodesics pick t, n, ηα with α ∈ {0, 1}. The choice of the null n is such that
n · t = −1. Projector onto the space spanned by the η’s.
Pµν = gµν + nµ tν + tµ nν
t·Dt = t·Dn = 0
θ2
θ̇ = − − σ ∗? σ∗? + ω ∗? ω∗? − R∗? t∗ t?
2
θ2
≤− (assuming WEC)
2
17.1 2D
√
g 0 = e2ω g, g 0 R0 = g (R − 2∇2 ω)
p
R
Rabcd = (gac gbd − gad gbc )
2
R
Rbd = gbd
2
2
R= for S 2 of radius a
a2
27
Z Z
1 1
χ= R+ k
4π M 2π ∂M
χ=2 for S 2
Consider a cone with half-angle θ. Let ρ be the coordinate radially outward from the tip.
α := sin θ
p
x = ρα cos φ, y = ρα sin φ, z = ρ 1 − α2
= dρ2 + α2 ρ2 dφ2
p
z = fa (ρ) 1 − α2 (7)
Here a is a regulator and could be, for example, that we deform the cone into a hyperboloid
1 − α2 2
z2 − (x + y 2 ) = a2
α2
The function f should have the following two properties. The first one recovers the cone in the limit
that the regulator goes away. The second one just says that the new tip should be a minimum.
Basically the derivative of f has to go from zero to one very quickly in the interval ρ ∈ (0, a).
28
The new coordinate embedding (7) makes the metric
=: u(ρ)dρ2 + α2 ρ2 dφ2
u(ρ a) = 1, u(ρ = 0) = α2
u0
R=
ρu2
√
Z Z
du
g R = 2πα
u3/2
= −4πα(1 − 1/α)
= 4π(1 − α)
17.2 3D
In three dimensions, the Riemann tensor can be expressed in terms of the Ricci tensor, since the
Weyl tensor is known to be identically zero.
gb = e2ω g
h i
b = e−2ω R − 2(D − 1)∇2 ω − (D − 2)(D − 1)∂µ ω∂ µ ω
R
29
18 Spherically Symmetric Metrics
−a(t, r)
b(t, r)
g∗∗ =
r2
r2 sin2 θ
Points to note:
• Grr = 0 implies that a0 /a + b0 /b = 0 which means that ab = constant. This constant can be
set to 1 after rescaling time.
• Solutions to Gtt and Grr equations automatically solve the Gθθ equation.
30
19 Vaidya Metric
We want to throw in a shell, which is moving at the speed of light. The only non-zero component
of the stress tensor is Tvv and it is a delta function. So it is convenient to work in the Eddington-
Finkelstein coordinates. The basic algebra is below.
31
Note that the Gvr = 0 is easily solved to give f = 1 − α(v)/r. This means that we basically have
a Schwarzschild solution, but the mass is now a function of v. We expect it to jump from 0 to the
final black hole mass that is formed by the shell. The equation Gvv = 8πGN Tvv achieves this since
Gvv is proportional to α0 (v); the parameter α(v) is proportional to a theta function.
32
20 Bondi Mass
In general relativity, there are two notions of mass of an asymtptocially flat spacetime. The
ADM mass is defined at spatial infinity. It measures the total mass of an initial data set on
an asymptotically flat slice, i.e. a slice of constant t in an asymptotically Cartesian coordinate
system. The Bondi mass is defined at null-infinity. It measures the total mass associated with an
asymptotically hyperboloidal slice, i.e. a slice approaching constant retarded time u = t − R, in an
asymptotically Cartesian coordinate system. The ADM mass is independent of t, but the Bondi
mass is in general a function of u, or more precisely, of a cut of null-infinity. Its change reflects an
outgoing flux of gravitational radiation.
Problems:
• For lattice QCD, we more or less know the UV Lagrangian, and that is how we know what
to write down for the lattice action. For gravity, we do not know the full-blown UV action.
String theory provides a candidate UV completion. So, in principle, if we know what String
Field Theory was, we could use that and latticize it. Counterargument: in four dimensions
(for example), maybe you could by hand write down all the invariant terms with dimensions
less than or equal to four and study that.
• What about the holographic principle? A local statistical system on the lattice has volume law
entropy, whereas black-holes are supposed to have area law entropy. Counterargument: the
reasoning in string theory that estalishes connections between black-hole entropy and string
states can be used for 3D building blocks also. So it seems that one could get area law entropy
from this kind of a setup.
• How do you get flat space as a solution? (Emphasized vehemently by Nathan Seiberg at a
lunch at the IAS, late September 2015)
33
22 Induced Gravity
Integrating out matter fields in a fixed background geometry will generate kinetic terms for the
metric.
log det + m2 + ξR
Z ∞
b ds h −as i
log = e − e−bs
a 0 s
Need to cutoff the s integral near s = 0. Put a cutoff s = 1/κ2 , assuming κ large. In four-dimensions,
the divergent terms will be proportional to κ4 , κ2 and log κ. The quantity a0 is equal to one, a1 (g)
is equal to the Ricci scalar, etc. You will also generate higher-curvature terms, etc.
23.1 Schwarzschild
Imagine calculating the Euclidean on-shell action with periodicity of the τ direction fixed. We are
working in the canonical ensemble. The log Z should be proportional to 1/GN and so by dimensional
analysis, we get
1
log Z ∼ −
GN T 2
dF d 1
S=− = (T log Z) ∼ ,
dT dT GN T 2
34
and the energy
d d 1
E=− log Z = T 2 log Z ∼
dβ dT GN T
35