Action Formulation in General Relativity

Ashley Chraya

Department of Physics
IISER Mohali

May 25, 2021

Organized at Indian Institute of Science Education and Research, Mohali

Ashley Chraya Action Formulation in General Relativity May 25, 2021 1 / 32
Overview

1 Introduction Addition of derivative term

2 Timelike and Null-like Congruences Remarks on Boundary term
Basics 5 Gravitational field Action
Theorems Introduction
3 Hypersurfaces Variation of the action
Basics Well-Posed Gravitational Action
Integeration on Hypersurfaces Boundary term in Electrodynamics
Tangent Vector Fields 6 3+1 Decomposition
Extrinsic Curvature Line Element
4 Well-posed problem in classical particle Schematic of Decomposition
dynamics 3+1 Decomposed Action

Ashley Chraya Action Formulation in General Relativity May 25, 2021 2 / 32

Basics
The discussion throughout the talk will be on arbitrary spacetime (M,g) with connections defined on it.
For affinely parameterized curve, geodesic equation is : U α ∇α U β = 0
For any two vector fields intersecting (as shown below) the lie derivative of one vector field w.r.t other
vanishes.
Lu ξ α = Lξ u α = 0 ⇒ α β
ξ;β α β
u = u;β ξ
For affinely parameterized geodesic : u α ξα = 0

Figure: Deviation vector between two neighbouring geodesics

Ashley Chraya Action Formulation in General Relativity May 25, 2021 3 / 32

Basics of Timelike Congruence
A congruence is a family of curves such that through each point there passes only one curve from this family.
For affine geodesics, the following relationships holds:
u α uα = 1, u;β
α β
u = 0, u;β α β
ξ = ξ;βα β
u , u α ξα = 0
Deviation vector points in the directions transverse to the flow of the congruence.
Spacetime metric gαβ can be decomposed into longitudinal and transverse part as hαβ = gαβ − uα uβ
α β
Evolution of Deviation vector: ξ;β u = Bβα ξ β
Bαβ is purely transverse : Bαβ u β = u α Bαβ = 0
h θ
Hence we can decompose it as : Bαβ = αβ 3
+ σαβ + ωαβ
Expansion Parameter : θ = ∇i u i
This parameter describes the fractional rate of change of congruence’s volume.
Shear Tensor σαβ : This is a symmetric and traceless quantity, which describes how the shape of
the congruence changes.
Rotation Tensor ωαβ : This is the antisymmetric part of Bij which describes how the congruence of
geodesics will rotate.

Ashley Chraya Action Formulation in General Relativity May 25, 2021 4 / 32

Theorems for Timelike Geodesics

Frobenius Theorem: Congruence of Curves is Hypersurface Orthogonal iff

∇[i Uj Uk] = 0
For timelike geodesic, rotation parameter should vanish ωab = 0
Raychaudhuri’s Equation: The evolution equation for the expansion scalar:
2
dθ
dτ
= − θ3 + σ αβ σαβ − ω αβ ωαβ − Rij u i u j
Focusing Theorem: For matter following Strong energy condition Rαβ u α u β ≥ 0 and the geodesic
congruence be hypersurface orthogonal, the expansion must decrease during its’ evolution.
dθ
dτ
≤0
3
For θ = θ0 < 0 under these conditions θ goes to −∞ along the geodesic within the proper time τ ≤ |θ0 |
The congruence will develop a caustic within finite proper time.

Ashley Chraya Action Formulation in General Relativity May 25, 2021 5 / 32

Basics of Null Geodesics

For affine geodesics, the following relationships holds:

k α kα = 0, k;βα β
k = 0, k;β α β α β
ξ = ξ;β k , k α ξα = 0
Last condition fails to remove component of ξ α in the direction of k α and hence, deviation vector doesn’t
points in the directions transverse to the flow of the congruence.
The transverse metric is 2-Dimensional, hence to isolate transverse part of the metric we need two null
vectors. Spacetime metric gαβ can be decomposed as
gαβ = hαβ + Kα Nβ + Nα Kβ
where N Nα = 0 and k Nα = 1 do not determine N α uniquely.
α α

α β
Evolution of Deviation vector: ξ;β k = Bβα ξ β
Bαβ is not purely transverse : Bαβ k β = k α Bαβ = 0 but Bαβ is not orthogonal to N α .
By isolating transverse component we get:
ξ˜α ≡ hα µ ξ µ = α

ξ +α (Nµ ξ µ ) k α
¯α β ∼
ξ;β k = B̄β ξ¯β
where B̄βα = hµα hβv Bvµ

Ashley Chraya Action Formulation in General Relativity May 25, 2021 6 / 32

Continued

B̄βα ξ¯β can be interpreted as the transverse relative velocity between the neighbouring geodesics.
Therefore, decomposing it as we did earlier:
h θ
B˜αβ = αβ + σαβ + ωαβ
2
i
Expansion Parameter : θ = ∇i k
This parameter describes the fractional rate of change of null-congruence’s cross sectional area.

Figure: Physical interpretation of Parameters

Ashley Chraya Action Formulation in General Relativity May 25, 2021 7 / 32

Theorems for Null Geodesics

Frobenius Theorem: Congruence of Curves is Hypersurface Orthogonal iff

∇[i Uj Uk] = 0
For Null geodesics, rotation parameter should vanish ωab = 0
Raychaudhuri’s Equation: The evolution equation for the expansion scalar:
2
dθ
dτ
= − θ2 + σ αβ σαβ − ω αβ ωαβ − Rij k i k j
using B˜αβ B˜αβ = Bαβ Bαβ
Weak Null Energy Condition ⇐⇒ Strong Energy Condition
Focusing Theorem: For matter following Strong /Weak Null energy condition Rαβ k α k β ≥ 0 and the
geodesic congruence be hypersurface orthogonal, the expansion must decrease during its’ evolution.
dθ
dλ
≤0
For θ = θ0 < 0 under these conditions θ goes to −∞ along the geodesic within the affine parameter
λ ≤ |θ2 |
0
The congruence will develop a caustic within finite proper time.

Ashley Chraya Action Formulation in General Relativity May 25, 2021 8 / 32

Theorems: Continued
For Null Hypersurface, Null congruences K α is tangent to the hypersurface, and thus are the null
generators of the hypersurface.
These null congruences satisfies the geodesic equation
k α ∇α k β = kk β
Therefore, null hypersurface is generated by null geodesics.

Figure: Family of Hypersurfaces orthogonal to congruence of null geodesics

Ashley Chraya Action Formulation in General Relativity May 25, 2021 9 / 32

Basics
In a four-dimensional spacetime manifold, a hypersurface is a three-dimensional submanifold that can be
either timelike, spacelike, or null.
εΦ,α
Timelike Hypersurface: Normal can be normalized nα = |g µν Φ,µΦ,v |1/2
where  is 1 if hypersurface is spacelike and -1 if hypersurface is timelike.
∂x α
Induced Metric or First fundamental form: Induced metric is 3 dimensional hab = gαβ eaα ebβ where eaα = ∂y a

Normal nα is defined such that nα eaα = 0

Null Hypersyrface : Unit normal is not defined. We let kα = Φ,α making the hypersurface generated by the
null geodesics.
As K α is tangent to hypersurface, we can take null geodesics parameter as one of the coordinates of
hypersurface y a = λ, θA

Induced Metric or First fundamental form: Induced metric is 2 dimensional as h becomes degenerate
 α
σAB = gαβ eAα eBβ , eAα = ∂x
∂θ A λ

Normal Nα is defined such that Nα eAα = 0, Nα K α

=1

Ashley Chraya Action Formulation in General Relativity May 25, 2021 10 / 32

Integeration on Hypersurfaces

Non-Null case: Invariant three dimensional volume (surface element) on the hypersurface is
dΣ ≡ |h|1/2 d3 y
dΣµ = nα dΣ is the directed surface element that points in the direction of increasing φ
Null case: As hab is degenerate h = 0 and nα does not exist. Therefore, the above expression should
be generalized.
Directed Surface element
dΣµ = εµαβγ e1α e2β e3γ d3 y
This expression holds for both null and non-null hypersurfaces.
Taking intrinsic coordinate y 1 as λ we can show dΣµ = k v dSµv dλ where 2-dimensional surface
element dSµv = εµv βγ e2β e3γ d2 θ
Further evaluating, we get
√
dSαβ = 2k[α Nβ] σd2 θ
Gauss Theorem
√ 4
Aα α
R H
V ;α −g d x = ∂V A dΣα

We can obtain conservation of charge if the divergence of Aα vanishes.

Ashley Chraya Action Formulation in General Relativity May 25, 2021 11 / 32

Tangent Vector Fields

Tangent vector fields Aα purely tangent to hypersurface admit the decomposition

Aα := Aa eaα
Instrinsic Covariant Derivative: Defined as the component of the projected Aα;β , we get
∇a Ab = ∇α Aβ eaα ebβ = Aa,b − Γcab Ac
where we have defined Γcab = ecγ eaγ;β ebβ
Assuming this connection to be metric compatible and this connection to be symmetric in its last two
argument we get the expression for connection:
1
Γcab = 2
(hca,b + hcb,a − hab,c )

Ashley Chraya Action Formulation in General Relativity May 25, 2021 12 / 32

Extrinsic Curvature

Extrinsic Curvature or Second fundamental form 1. Defined as the normal components of ∇β Aα ebβ
or 2. Defined as the component of the tagential covariant derivative of ∇β nα we find,
Kab = nα;β eaα ebβ
We can get kαβ from the above expression: kαβ = ∇α nβ − nα nγ ∇γ nβ
Kab is a symmetric tensor. Therefore,
Kab = n(α;β) eaα ebβ = 1
2
(Ln gαβ ) eaα ebβ
Physical Interpretation : Kab is related to normal derivative of the metric
We further note that K ≡ hab Kab = n;α
α

Physical Interpretation : K is equal to the expansion of the congruence of geodesics that are hypersurfa
orthogonal. Therefore, k = θ
hab is concerned with purely intrinsic aspects of hypersurface.
kab is concerned with extrinsic aspects.

Ashley Chraya Action Formulation in General Relativity May 25, 2021 13 / 32

Equation of Motion and well Posed Problem

The trajectory of the system between two instants of time is the extremum of the action integral.
Rt
A[qi1 , qi2 , t1 , t2 ] = t 2 L(qi , q̇i , t)dt
1

By varying the trajectory, the change in action is

h i t2 R h  i
t d
∂L
δA = ∂∂Lq̇ δqi + t 2 dtδqi ∂q − dt
∂L
∂ q̇i
i t1 1 i

By setting the end points fixed δq = 0 and making δA = 0 we get

∂L d ∂L
∂q i
− dt ∂ q̇ i
=0
Upon expanding the time derivative we get 2nd order DE
∂2L ∂2L
∂ q̇ i ∂ q̇ j
q̈ j = ∂L
∂q i
− ∂ q̇ i ∂q j
q̇ j

Remark : Action Principle tells us what need to be fixed at the boundary without we assuming anything.

Ashley Chraya Action Formulation in General Relativity May 25, 2021 14 / 32

Addition of derivative term

Let L = L(q, q̇, q̈), to get the 2nd order EOM, we have to set δqi = 0 and δ q̇i = 0. etting δq = 0 at the
endpoints is no longer sufficient to kill the boundary term. We are setting 4 boundary conditions for 2nd
order EOM. For most choices of boundary data EOM would not render any solution. Addition of total
derivative can make the variational principle ill-posed.
Let Lagrangian be linear in q̈
L1 = L + dtd f (q, q̇, t)
δA1 = δA + δf (q2 , q̇2 , t2 ) − δf (q1 , q̇, t1 )
Example
R2
R2 R2
1 2
A = 1 dt − 12 q q̈ = 1 dt dt dtd 1



2
q̇ − 1 2
q q̇
Varying the action we obtain
Z 2
1
δA = − dtδq q̈ + (q̇δq)21 − (δq q̇ + qδ q̇)21
1 2
Z 2
1
=− dtδq q̈ + (q̇δq − qδ q̇)21
1 2

Ashley Chraya Action Formulation in General Relativity May 25, 2021 15 / 32

Remarks on Boundary term

Z 2
1
δA = − (q̇δq − qδ q̇)21
dtδq q̈ +
1 2
Bulk equation is still 2nd order but we have to fix both δ q̇ = 0 and δq = 0 as the boundary conditions
which makes the problem ill-posed.
Solution: To make this problem well posed we add boundary terms so that we are left with Lagrangian
which does not depend on higher derivatives in the action.
We will see ahead how this is relevant in the action for gravitational field. This is exactly in analogy with
Gravitational action where we add GHY term to make the action well posed.

Ashley Chraya Action Formulation in General Relativity May 25, 2021 16 / 32

Introduction

Motivation : Just as the action for scalar/ vector field, action for gravitation can be dependent on
dynamical variable and derivative of dynamical variable but no non trivial scalar Lagrangian can be
constructed from the metric and its first derivative because in local inertial frame gαβ = ηαβ and
∂γ gαβ = 0 . Only choice left with us is to use ∂ν ∂µ gαβ in the action but we are further constrained to get
2nd order field equations.
To get 2nd order differential equation the 2nd derivative of gαβ should be linear in action. By doing this,
we get some boundary term which we can cancel by adding another term to the action, which in this
case is GHY boundary term.
Therefore, the most simple scalar that can be constructed which has second derivative of metric is Ricci
scalar which is constructed from the Riemann tensor, which contains second derivatives of the metric.
Action for gravitational field is
√
d 4 x −gR g, ∂g, ∂ 2 g
R

16πGA = V

Ashley Chraya Action Formulation in General Relativity May 25, 2021 17 / 32

Variation of the action

Assuming gαβ the dynamical variable, varying the action w.r.t. it

√ R4 √
δAEH = V d 4 x −gGαβ δg αβ + V x −g∇α δnα
R
√
= V d 4 x −gGαβ δg αβ + ∂V δnα dΣα
R R

where δnγ = g αβ δΓγαβ − g γµ δΓα

αµ

Assuming the boundary is timelike, variation of metric at the boundary vanishes δgαβ = 0, due to which
variation of the tagential derivative of metric also vanishes δ∂µ gαβ eamu = 0 we get :
Z  
1
d 4 x −g Rαβ − Rgαβ δg αβ
p
16πGδA =
2
ZV
d 3 y |h|nµ hαβ (∂µ δgαβ )
p
−
∂V

Ashley Chraya Action Formulation in General Relativity May 25, 2021 18 / 32

Gauss Theorem: Illustration

Figure: The Gauss theorem in a spacetime volume is illustrated

Ashley Chraya Action Formulation in General Relativity May 25, 2021 19 / 32

Solution - I

To get the bulk equation we have to fix normal derivatives also but fixing that makes our action principle
ill-posed and hence the field equations are not consistent with boundary data.
Therefore to make gravitational action well posed we have to add a boundary term so that normal
derivative part cancels away.
√
AGHY = 2 ∂M d3 y  hK
R

where K is the trace of extrinsic curvature

Remark : As extrinsic curvature is related to normal derivative of the metric and boundary term
is normal derivative of the metric, therefore, trace of extrinsic curvature is not a bad guess.
Varying the GHY term we can cancel our boundary term and obtain the bulk term.

Ashley Chraya Action Formulation in General Relativity May 25, 2021 20 / 32

Well-Posed Gravitational Action

Now, we will derive the field equations without assuming a priori that variation of the metric vanishes.
Remark : In earlier calculation we expanded tensor δΓ to variation of metric terms, and then assumed
the variation of the metric vanishes. Here, we will convert δΓ to covariant derivatives using
 
δ (∇α nβ ) = ∇α δnβ − δΓγαβ nγ
δ (∇α nα ) = ∇α δnα + nγ δΓα
αγ

By varying Einstein Hilbert Action we get:

√ √ h i
δAEH = V d 4 x −g Gαβ δg αβ + V d 4 x −g∇γ g αβ δΓγαβ − g γµ δΓα
R   R
αmu
√ √
≡ V d 4 x −g Gαβ δg αβ + ∂V d 3 x hB [nγ ]
R   R
h i
where B [nγ ] = nγ g αβ δΓγαβ − g γµ δΓα
αµ

B [nγ ] = ∇α (δu α ) − δ (2∇α nα ) + (∇α nβ ) δg αβ

where δu α ≡ δnα + g αβ δnβ

Ashley Chraya Action Formulation in General Relativity May 25, 2021 21 / 32

Well Posed Gravitational Action

We can easily see δu α lie on the hypersurface. Using the definition of intrinsic covariant derivative, we
get ∇a δu a = ∇α δu α − aα δu α where aα = nβ ∇β nα ; and ∇a δu a is the intrinsic covariant derivative.
Further, by property of aα δnα = 0 we get, B [nγ ] = ∇a (δu a ) − δ (2∇α nα ) + (∇α nβ − nα aβ ) δg αβ
where we used the definition of extrinsic curvature kαβ = ∇α nβ − nα nγ ∇γ nβ
Finally using the properties of extrinsic curvature we get the boundary term
Z √ Z √ Z √
d 3 x hB [vc ] = d 3 x h∇a δu a + δ d 3 x2K h


∂V ∂V ∂V
Z
3
√
+ d x h (Khαβ − Kαβ ) δhαβ
∂V
√
We have to add 2K h term,which is the GHY term, to cancel the boundary term and get the field
equation.
At the boundary we need not have to fix whole metric but just induced metric hij

Ashley Chraya Action Formulation in General Relativity May 25, 2021 22 / 32

Boundary term in Electrodynamics

Electrodynamics is the vector field theory where Ai is the 4-vector for which the action for the free
Electrodynamic field is given by
1
F F ik d 4 x
R
A = − 16π V ik
where Fik = ∂i Ak − ∂k Ai
By assuming Ai as dynamical variable and varying the action w.r.t it, we get:
 Z 
1
δA = δ − d 4 xFab F ab
16π V
Z Z
1 1
=− d 4 x∂k F ik δAi − d 3 xE.δA
4π V 4π t
Bulk term will lead to Maxwell equations and to vanish boundary term we have to fix spatial part of
vector potential A at the t = constant surfaces. Further, we can show that we have to fix just magnetic
field at the t = constant surfaces.

Ashley Chraya Action Formulation in General Relativity May 25, 2021 23 / 32

Introduction

The purpose of 3+1 decompostion is to express the action in terms of the Hamiltonian it is necessary to
foliate V with a family of spacelike hypersurface.

Figure: Foliation of Spacetime by Spacelike Hypersurfaces

Assuming a scalar field t(x α ) such that t = constant, and on each hypersurface we assume coordinates
y a . Also assuming y a is constant along the flow of the curve, which defines the mapping of the point P
to P 0 and so on.

Ashley Chraya Action Formulation in General Relativity May 25, 2021 24 / 32

Line Element
Therefore, the transformation exists x α = x α (t, y a ).
 α
Tangent vector along the curve is t α = ∂x∂t
y a α 
α
Tangent vectors on the hypersurface is ea = ∂x ∂y a
t
Lapse function N, and shift vector N a can be defined as
t α = Nnα + N a eaα and nα = −N∂α t
where nα is the normal vector.

Figure: Decomposition of t α into lapse and shift vector

Ashley Chraya Action Formulation in General Relativity May 25, 2021 25 / 32

Line Element

The metric in the coordinates (t, y a ) can be expressed as

ds2 = −N 2 dt 2 + hab ( dy a + N a dt) dy b + N b dt


√ √
Further, we can easily get −g = N h
For displacement along the curve, increment in proper time is related to the increment in coordiante
time as
dτ 2 = N 2 + hab N a N b dt 2

If the congruence is Hypersurface orthogonal, Shift vectors N a vanishes and hence proper time would
be related to coordinate time by lapse function as:
dτ 2 = N 2 dt 2

Ashley Chraya Action Formulation in General Relativity May 25, 2021 26 / 32

Schematic of Decomposition

Figure: Decomposition of t α into lapse and shift vector

Foliation of V by Σt
St embedded in Σt
St embedded in V
B embedded in Spacetime
Foliation of B by St

Ashley Chraya Action Formulation in General Relativity May 25, 2021 27 / 32

3+1 Decomposed Action
Surface Σt St B
Unit normal nα rα rα
Coordinates ya θA zi
Tangent vectors eaα eAα eiα
Induced metric hab σAB γij
Extrinsic curvature Kab kAB Kij

3+1 Decomposition of the action

1
Z t2 Z   √
3
SG = dt R + K ab Kab − K 2 N h d3 y
16π t1 Σt
√ 2
I 
+2 (k − k0 ) N σd θ
St

Ashley Chraya Action Formulation in General Relativity May 25, 2021 28 / 32

Remarks

Let ḣab ≡ Lf hab , we can show

ḣab = 2NKab + Na|b + Nb|a
Therefore, the action does not involve Ṅ, nor Ṅ a . The dynamical variable is induced metric only, lapse
and shift only serve to specify the foliation of V into spacelike hypersurfaces.
Because the foliation is arbitrary, we are free to choose lapse and Shifts.
√
Momentum conjugate to induced metric is defined as pab = ∂ ḣ∂ ( −gLG ) where LG is the bulk part of
ab
lagrangian. It can be shown that
√
(16π)pab = h K ab − Khab

Ashley Chraya Action Formulation in General Relativity May 25, 2021 29 / 32

Applications

Singularity Theorems
Boundary term for Null Hypersurfaces
Numerical Relativity
Relativistic Hydrodynamics

Ashley Chraya Action Formulation in General Relativity May 25, 2021 30 / 32

References

1. Eric Poisson. A relativist’s toolkit: the mathematics of black-hole mechanics. Cambridge university
press, 2004.
2. Thanu Padmanabhan. Gravitation: foundations and frontiers. Cambridge University Press, 2010.
3. T Padmanabhan. A short note on the boundary term for the hilbert action. Modern Physics Letters A,
29(08):1450037, 2014.
4. Ethan Dyer and Kurt Hinterbichler. Boundary terms, variational principles, and higher derivative
modified gravity. Physical Review D, 79(2):024028, 2009.
5. Krishnamohan Parattu, Sumanta Chakraborty, Bibhas Ranjan Majhi, and T Padmanabhan. A boundary
term for the gravitational action with null boundaries. General Relativity and Gravitation, 48(7):1–28,
2016.

Ashley Chraya Action Formulation in General Relativity May 25, 2021 31 / 32

 Thank You

Ashley Chraya Action Formulation in General Relativity May 25, 2021 32 / 32