MIT-CTP/4068, NSF-KITP-09-136

An analytic Lifshitz black hole

Koushik Balasubramanian and John McGreevy
Center for Theoretical Physics, MIT, Cambridge, Massachusetts 02139, USA
arXiv:0909.0263v1 [hep-th] 2 Sep 2009

Abstract

A Lifshitz point is described by a quantum field theory with anisotropic scale invariance (but
not Galilean invariance). In 0808.1725, gravity duals were conjectured for such theories. We
construct analytically a black hole which asymptotes to a vacuum Lifshitz solution; this black
hole solves the equations of motion of some simple (but somewhat strange) extensions of the
models of 0808.1725. We study its thermodynamics and scalar response functions. The
scalar wave equation turns out to be exactly solvable. Interestingly, the Green’s functions
do not exhibit the ultralocal behavior seen previously in the free Lifshitz scalar theory.

August 2009
1 Introduction

A great deal of progress has been made in the study of quantum field theories and
their holographic duals. The possible scope of this enterprise is not yet clear; for example,
the correspondence seems to extend to some systems without Lorentz invariance. Recently,
attempts have been made to apply the holographic principle to study condensed matter
systems near a critical point (for reviews, see [1, 2]). There are many scale-invariant field
theories that are not Lorentz invariant, which are of interest in studying such critical points.
In such a theory, time and space can scale differently i.e., t → λz t, ~x → λ~x under dilata-
tion. The relative scale dimension of time and space, z, is called the ‘dynamical exponent’.
Such a scale invariance is exhibited by a Lifshitz theory, which we will take to mean an
anisotropic scale-invariant theory which is not Galilean-invariant. The following Gaussian
action provides a simple example of a (free) Lifshitz theory in d space dimensions:
Z h
d 2 2

2 i
S[χ] = d xdt (∂t χ) − K ∇ χ (1.1)

This action describes a fixed line parametrized by K, and the dynamical exponent is z = 2.
This theory describes the critical behavior of e.g. quantum dimer models [3]. In many ways,
the d = 2, z = 2 version of the theory (1.1) is like a relativistic boson in 1+1 dimensions1 .
The scaling behavior of the ground-state entanglement entropy for this class of theories was
studied recently in [4, 5]. This analysis also supports the similarity with 2d CFT, in that a
universal leading singular behavior is found.
In the free theory, the boson has logarithmic correlators
1
Z
~
hχ(x)χ(0)i ∼ dωd2 k 2 eik·~x−iωt ∼ ln x. (1.2)
ω − k4
As in the familiar d = z = 1 case, the operators of definite scaling dimension are not the
canonical bose field itself, but rather its exponentials and derivatives. In the connection
with quantum dimer models, the bose field is a height variable constructed from the dimer
configuration, and the exponentials of the bose field are order parameters for various dimer-
solid orderings [3]. At zero temperature, the logarithmic behavior of the correlator of the
bose fields implies that the two-point function of the order parameter decays as a power
1
Similar statements apply whenever z = d. However, constructing a rotation-invariant, local spatial
kinetic operator that scales like p2d is tricky for d 6= 2k for integer k. We note in passing that the existence
of such theories seem to be suggested by the calculations of [6].

1
law. However, the equal-time correlators at finite temperature are ultra-local in the infinite-
volume limit [7]: they vanish at any nonzero spatial separation. In [7], it was suggested that
this might be a mechanism for the kind of local criticality (scaling in frequency, but not
momentum) seen in the strange metal phase of the cuprates and in heavy fermion materials.
One is led to wonder whether this property should is shared by interacting Lifshitz theories,
and whether the Lifshitz scaling is sufficient to produce this behavior. In [7] the addition of
perturbative interactions was shown to lead to a finite correlation length; these perturbations
violate the Lifshitz scaling. Below we will show that interactions which preserve the Lifshitz
scaling need not give ultralocal behavior.
Gravity solutions with Lifshitz-type scale invariance were found in [8]. They found that
the following family of metrics, parametrized by z, provide a geometrical description of
Lifshitz-like theories (with z as the dynamical exponent):

~ 2 + dr 2
!
dt2 dx
ds2 = L2 − 2z + , (1.3)
r r2

where ~x denotes a d−dimensional spatial vector. For d = 2, this metric extremizes the
folowing action:

1 1
Z Z Z
4


S= d x (R − 2Λ) − F(2) ∧ ⋆F(2) + F(3) ∧ ⋆F(3) − c B(2) ∧ F(2) , (1.4)
2 2

where F(2) = dA(1) , F(3) = dB(2) and Λ is the four dimensional cosmological constant. They
computed the two-point function for the case when z = 2 and showed that it exhibits power
law decay. They also studied the holographic renormalization group flow for this case and
found that AdS4 is the only other fixed point of the flow. Lifshitz vacuum solutions were
shown to be stable under perturbations of the bulk action in [9].
In this paper we shall study a black hole solution which asymptotes to the Lifshitz
spacetime with d = 2, z = 2. In section 2, an analytical solution for a black hole that
asymptotes to the planar Lifshitz spacetime is written down. We present several actions
whose equations of motion it solves; they all involve some matter sector additional to (1.4).
Section 3 presents an analysis of the thermodynamics of this black hole. In section 4, we solve
the wave equation for a massive scalar field in this background; surprisingly, this equation
is exactly solvable. We use this solution to calculate the two-point functions of boundary
operators in section 5.

2
 Since we found the solution described in this paper, some related work has appeared. [10]
constructs a black hole solution in a related background with slightly different asymptotics.
Danielsson and Thorlacius [11] found numerical solutions of black holes in global Lifshitz
spacetime. Interestingly, these are solutions to precisely the system studied by [8], with
no additional fields. Related solutions were found by [12, 27]. [13] found solutions of type
IIB supergravity that are dual to Lifshitz-like theories with spatial anisotropy and z = 3/2;
these solutions have a scalar field which breaks the scaling symmetry. To our knowledge, a
string embedding of z = 2 Lifshitz spacetime is still not known; obstacles to finding such an
embedding are described in [14].

2 Black hole solution

2.1 Vacuum solution

The tensor fields in [8] can be rewritten as one massive gauge field. The Chern-Simons-like
coupling is responsible for the mass. A familiar example is that of a 2-form field strength F
and a 3-form field strength H in five dimensions with L = F ∧ ⋆F + H ∧ ⋆H + F ∧ H: this
gives the same equation of motion as L = F ∧ ⋆F + A2 . In the four dimensional case studied
in [8], the dual of the 3-form field strength in four dimensions is a scalar field ϕ. Then

√
B2 ∧ F2 = −F3 ∧ A1 + bdy terms = − ⋆ dϕ ∧ A1 = − g∂ µ ϕAµ . (2.1)

The action then reduces to

F2 ∧ ⋆F2 + (∂ϕ + A)2 , (2.2)

and ϕ shifts under the A gauge symmetry, and we can fix it to zero, and this is just a massive
gauge field2 . Hence, the zero-temperature Lifshitz metric

2 dt2 d~x2 + dr 2
ds = − 2z + , (2.3)
r r2
is a solution of gravity in the presence of cosmological constant and a massive gauge field,
and the gauge field mass is m2 = dz. The bulk curvature radius has been set to one here
2 +(d−1)z+d2
and throughout the paper; in these units, the cosmological constant is Λ = − z 2
.
The gauge field profile is A = Ωr −z dt (in the r coordinate with the boundary at r = 0), and
2
This was also observed in [10].

3
the strength of the gauge field is (for d = 2)

z2 + z − 2
Ω2 = 8 .
z(z + 2)

We note in passing that the Schrödinger spacetime is a solution of the same action with a
different mass for the gauge field and a different cosmological constant [16, 17]. Therefore we
find the perhaps-unfamiliar situation where the same gravitational action has solutions with
very different asymptopia. Another recent example where this happens is ‘chiral gravity’ in
three dimensions, which has asymptotically AdS solutions as well as various squashed and
smushed and wipfed solutions [15].
Given this fact, one might expect that the Lifshitz spacetime can be embedded into the
same type IIB truncations as the Schrödinger spacetime (see [18, 19] and especially [20]).
However, the scalar equation of motion is not satisfied by the Lifshitz background since F 2
is non-zero.

2.2 Black hole solution

We shall now study a black hole in four dimensions that asymptotically approaches the
Lifshitz spacetime with z = 2. We first observe that there is such a black hole in a system
with a strongly-coupled scalar (i.e. a scalar without kinetic terms). The action is
 −2Φ
m2 2

1 e
Z Z
4 4 2 −2Φ


S1 = d x (R − 2Λ) − d x F + A + e −1 . (2.4)
2 4 2

A solution of this system is

1
Φ = − log 1 + r 2 /rH
2
A = f /r 2dt


,
2
dt2 d~x2 dr 2
ds2 = −f + 2 + 2, (2.5)
r 2z r fr
with
r2
f =1− 2
.
rH
Note that the metric has the same simple form as in the RG flow solution (eqn (4.1)) of [8].
We can get the same contributions to the stress tensor as from the scalar without kinetic
terms from several more-reasonable systems. One such system is obtained by adding a

4
second massive gauge field B which will provide the same stress-energy as the scalar. It has
a slightly unfamiliar action:
1 2 2 m2A 2 1 2 m2B
 
1
Z Z
4 4 2
S2 = d x (R − 2Λ) − d x B dA + A + dB − (1 − B ) (2.6)
2 4 2 4 2
where A, B are one-forms, and m2A = 4 and m2B = 2 . The solution looks like B =
B(r)dr, A = A(r)dt and the metric is same as (2.5). In the solution, the scalar functions
take the form s
r2
 
B(r) = grr 1 + 2 , A(r) = Ωf r −z dt
rH
Note that B(r) isn’t gauge-trivial (even though its field strength vanishes) because of the
mass term. Since B(r) asymptotes to 1, the effective gauge coupling of the field A is not
large at the boundary.
The system with a strongly-coupled scalar in (2.4) is not equivalent to the system (2.6)
with two gauge fields. For example, there are solutions of (2.4) where the scalar has a profile
that depends both on r and x; such configurations do not correspond to solutions of (2.6).
It is not clear whether the solution written above is stable. We leave the analysis of the
stability of such solutions to small perturbations to future work. As weak evidence for this
stability, we show in the next section that these black holes are thermodynamically stable.
Another action with this Lifshitz black hole (2.5) as a solution is
 
1 1 2 1 −2Φ 2
Z
4 2 2
S3 = d x R − 2Λ − dB − (∂Φ − B) − mA A − e F − V (Φ) (2.7)
2 2 2
where V (Φ) = 2e−2Φ − 2. In the solution, the metric and gauge field A take the same form
as in (2.5). The other fields are

−2Φ r2
e =1+ 2 , B = dΦ.
rH

Note that the action (2.7) is not invariant under the would-be gauge transformation

B → B + dΛ, Φ → Φ + Λ,

because of the coupling to F 2 − 4 (the sum of the gauge kinetic term and the potential
term)3 . We are not bothered by this: it means that in quantizing the model, mass terms for
the fluctuations B will be generated; however, such a mass term is already present.
3
We note that this quantity does vanish on the solution of interest.

5
 We would also like to point out that in the three systems S1,2,3 described above, the stress-
energy tensor of the fields with local propagating degrees of freedom satisfy the dominant
(Φ,A,B)
= Rµν − ( 21 R + Λ)gµν satisfies the following


energy condition4 , i.e. Tµν

Ttt Ttt Ttt

= > −1 and > −1.
Txx Tyy Trr
Hence, there are no superluminal effects in the bulk. This is basically a consequence of the
fact that the squared-masses of the gauge fields are positive.

3 Lifshitz black hole thermodynamics

The Hawking temperature and entropy can be calculated using the near horizon geometry.
The Hawking temperature is the periodicity of the Euclidean time direction in the near
κ
horizon metric (proportional to the surface gravity) i.e., T = |
2π r=rH
, with

1
κ2 = − ∇a v b ∇b v a
2

where v = ∂t . Hence,
1
T = 2
. (3.1)
2πrH
The entropy of the black hole is

Area of Horizon Lx Ly
S= = 2
. (3.2)
4GN 4G4 rH

We shall now evaluate the free energy, internal energy and pressure by calculating the
on-shell action and boundary stress tensor. In order to renormalize the action, it is essential
to add counterterms which are intrinsic invariants of the boundary (see [21]).
Consider the following gravitational action:

e−2Φ 2 m2A 2
 
1 4 √
Z
S= d x g R − 2Λ − F − A − V (Φ)
2 M 4 2

√
Z
d3 x γ K + cN e−2Φ nµ Aν Fµν


− (3.3)
∂M
1 √
1 √
Z Z
3
d x γ 2c0 − c1 Φ − c2 Φ2 + d3 x γ (c3 + c4 Φ) A2 + c5 A4 .


+
2 ∂M 2 ∂M
4
We would like to thank Allan Adams, Alex Maloney and Omid Saremi for bringing this criterion to our
attention.

6
The second line of (3.3) contains extrinsic boundary terms: the Gibbons-Hawking term, and
a ‘Neumannizing term’ which changes the boundary conditions on the gauge field. The last
line of (3.3) describes the intrinsic boundary counterterms5 . In the above expression, we
have set 8πG = 1. We have written the analysis in terms of S1 (2.4); the analysis can be
adapted for S2 (2.6) by simply replacing Φ in (3.3) by − 12 log B 2 . If Neumann boundary
conditions are imposed on the gauge field, then cN = 1 and ci = 0 for i ≥ 3. Similarly,
cN = 0, if Dirichlet boundary condition is imposed on the gauge field.
The boundary stress tensor resulting from (3.3) is

e−2Φ r
 
1 1 2
Tµν = Kµν − K − c0 + c1 Φ + c2 Φ γµν + (n Aµ ∂r Aν + nr Aν ∂r Aµ − nr Aα ∂r Aα γµν )
2 2 2
1
+ c3 + c4 Φ + 2c5 A2 Aµ Aν − c3 + c4 Φ + c5 A2 A2 γµν



(3.4)
2
The values of ci are determined by demanding that the action is “well-behaved”. The
action is well-behaved if the variation of the action vanishes on-shell and if the residual
gauge symmetries of the metric are not broken. The values of ci which makes the action
well-defined also render finite the action and boundary stress tensor (please see appendix
A). Implementing this procedure, we find for the energy density, pressure and free energy

1 Lx Ly
E = P = −F = T S = 4
(3.5)
2 2rH

Satisfying the first law of thermodynamics (in the Gibbs-Duhem form E + P = T S) is a nice
check on the sensibility of our solution, since it is a relation between near-horizon (T, S) and
near-boundary (E, P) quantities.
Recently, [28] have described an alternative set of boundary terms for asymptotically
Lifshitz theories. They do not include the Neumannizing term, but instead include an
p
intrinsic but nonanalytic Aµ Aµ term.
5
The most general combination of counterterms, which do not vanish at the boundary, is
1 √
1 √
Z Z
d3 x γ 2c′0 + c′1 Φ + c′2 Φ2 + d3 x γ (c′3 + c′4 Φ) (A2 − 1) + c′5 (A2 − 1)2


2 ∂M 2 ∂M

which has the same form as (3.3).

7
4 Scalar response

In this section, we study a probe scalar in the black hole background (2.5). The scalar can
be considered a proxy for the mode of the metric coupling to Tyx .

4.1 Exact solution of scalar wave equation

Consider a scalar field φ of mass m in the black hole background (2.5)6 .

r2
Let u ≡ 2 .
rH
Fourier expand:

~
X
φ= φk (u)e−iωt+ik·~x
k

The wave equation takes the form:

u (−f k 2 + uω 2) + m2 f 1 ′
0= 2 2
φk (u) − φk (u) + φ′′k (u)
4f u fu

where k 2 ≡ ~k 2 . Near the horizon, the incoming (−) and outgoing (+) waves are

φk ∼ (1 − u)±iω/2 .

The solutions near the boundary at u = 0 are

1
√
4+m2
φk ∼ u1± 2

1
√
m2 +4
The exact solution to the wave equation is φk (u) = f −iω/2 u1− 2 Gk (u) with
√
m2 +4
Gk (u) = A1 2 F1 (a+ , b+ ; c+ , u)u + A2 2 F1 (a− , b− ; c− , u) (4.1)

and

(a± , b± ; c± ) ≡
 √
m2 +4
√ √
m2 +4
√ √ 
− iω
2
± 2
− 1
2
−k 2 − ω 2 + 1 + 12 , − iω
2
± 2
+ 1
2
−k 2 − ω 2 + 1 + 12 ; 1 ± m2 + 4; u

We emphasize that this is the exact solution to the scalar wave equation in this black hole;
such a solution is unavailable for the AdSd>3 black hole. The difference is that the equation
6
In the following we have set both the bulk radius of curvature and the horizon radius to one. This means
that frequencies and momenta are ‘gothic’ [22], i.e. measured in units of rH . Note that since z = 2, ω needs
two factors of rH to make a dimensionless quantity.

8
here has only three regular singular points, whereas the AdS5 black hole wave equation has
four. This is because in the AdS5 black hole, the emblackening factor is f = 1 − u2 which
has two roots, whereas ours is just f = 1 − u.
The other example of a black hole with a solvable scalar wave equation is the BTZ black
hole in AdS3 [25]7 . The origin of the solvability in that case is the fact that BTZ is an
orbifold of the zero-temperature solution. This is not the origin of the solvability in our case
– this black hole is not an orbifold of the zero-temperature solution. This may be seen by
comparing curvature invariants: they are not locally diffeomorphic. More simply, if the black
hole were an orbifold, it would solve the same equations of motion as the vacuum solution.
The fact that we were forced to add an additional matter sector (such as Φ or Bµ ) to find
the black hole solution immediately shows that they are not locally diffeomorphic.
Now we ask for the linear combination of (4.1) which is ingoing at the horizon. In terms
√ √
of ν ≡ 4 + m2 , γ ≡ 1 − ω 2 − k 2 , this is the combination with
1 1



A1 Γ(ν) Γ (1 − iω − ν − γ) Γ (1 − iω − ν + γ)
= −(−1)ν 2 2
. (4.2)
Γ(−ν) Γ 12 (1 − iω + ν − γ) Γ 21 (1 − iω + ν + γ)


A2

In the massless case, one of the hypergeometric functions in (4.1) specializes to a Meijer
G-function, and the solution is φk = u2 f −iω/2 Gk (u) with

Gk (u) =
√ √
c2 2 F1 − iω 1
−k 2 − ω 2 + 1 + 32 , − iω 1
−k 2 − ω 2 + 1 + 32 ; 3; u +


2
− 2 2
+ 2

 
1
√
1
√

2
iω − −k 2 − ω2 +1−1 , 2
iω + −k 2 − ω2 +1−1 
2,0 

c1 G2,2 u


−2, 0

In this solution, the coefficient of c1 (the Meier-G function) is purely ingoing at the horizon.

4.2 Correlators of scalar operators

In the previous section we wrote the solution for the wave equation in this black hole for a
scalar field with an arbitrary mass. As mentioned earlier, the BTZ black hole also shares
this property of having a scalar wave equation whose solutions are hypergeometric. Hence,
7
Another example, in two dimensions, is [26].

9
one might expect that the two-point function of scalar operators in a Lifshitz-like theory to
have a form that is similar to that of 2D CFTs.
The momentum space correlator for a scalar operator of dimension ∆ = ∆− is deter-
mined from the ratio of the non-normalizable and normalizable parts of the solution. The
asymptotic behavior of the solution in (4.1) is
∆+ ∆−
φ∼u 2 (A1 + O(u)) + u 2 (A2 + O(u)) (4.3)

Hence, the retarded Green’s function (two-point function) is

1 1



A1 Γ(ν) Γ (1 − iω − ν − γ) Γ (1 − iω − ν + γ)
Gret (ω, ~k) = − = (−1)ν 2
1

2
1

(4.4)
A2 Γ(−ν) Γ 2 (1 − iω + ν − γ) Γ 2
(1 − iω + ν + γ)
with ν and γ defined above equation (4.2). Note that the correlator has a form very similar
to that of a 2D CFT. It would be nice to know the precise connection between z = 2 Lifshitz-
like theories in 2 + 1 D with 2D CFTs that is responsible for this similarity. Note that the
poles of the retarded Green’s function do not lie on a straight line in the complex frequency
plane, as they do for 2D CFTs.
Next, we would like to see whether the correlators exhibit ultra local behavior at finite
temperature as observed in the free scalar Lifshitz theory [7]. We find that the Green’s
function is not ultra-local – this removes the possibility that Lifshitz-symmetric interactions
require ultralocal behavior.
We will now calculate the two-point function of a scalar operator of dimension ∆ = 4
at finite temperature. In this case, the correlator is given by the coefficient of r 4 in the
asymptotic expansion of the solution near r = 0. Kachru et. al. [8] showed that the correlator
exhibits a power law decay at zero temperature.
We can evaluate this correlator by extracting the coefficient of the u2 term (note that
u ∝ r 2 ) in the asymptotic expansion of the solution of the massless scalar wave equation.
The behavior of the solution near u = 0 is
 u2  2 "   q 
~ u 
~ 2 ~ 2 2 1 ~ 2 2
φ(u, k, ω) = 1− k + 2iω − k + 4ω −3+2ψ −1 + iω − 1 − k − ω
4 64 2
  q  #
1
−1 + iω + 1 − ~k 2 − ω 2 + 2γE + 2 ln u + O u3


+ 2ψ (4.5)
2
where γE is Euler’s constant, ψ is the digamma function. The behavior of the solution in the
Euclidean black hole can be obtained by replacing ω by −i|ω|. The choice of the negative sign

10
gives the solution which is ingoing at the horizon, as appropriate to the retarded correlator
[22]. Henceforth, we shall work with the solution for the Euclidean case. The correlator is
the sum of the two digamma functions. All other terms in the coefficient of u2 are contact
terms. Hence, the correlator in momentum space is
  "   q 
~ ~ ~ 2
2
2 1 ~ 2 2
hO(−ω, −k)O(ω, k)i ∝ k − 4ω ψ −1 + |ω| − 1 − k + ω +
2
  q  #
1
ψ −1 + |ω| + 1 − ~k 2 + ω 2 (4.6)
2
After dropping the contact terms, the above expression can be written as follows
∞
hO(−ω, −~k)O(ω, ~k)i ∝
X
An
n=1

where
(2n − 3) + |ω| an + | ω |
An = = 2
(2n − 3)2 2
+ 2|ω|(2n − 3) + k − 1 an + 2an | ω | +k 2 − 1
We can now calculate the correlators in coordinate space by performing the Fourier transform
of the above expression8 . This is given by
h iX
D(|~x|, t) = (4∂t2 − (∇2 )2 ) Fn (4.7)
n

where, D(|~x|, t) is the two-point function and

X XZ an + | ω |
Fn = kdkdωdθ 2 2 − 1/4
eik|~x| cos θ+iωt
n n
a n + 2an | ω | +k

The short distance (r ≪ rH ) behavior of the equal time correlator is

h
2 2 2
i 1
D(|~x| ≪ rH , 0) = (4∂t − (∇ ) )F ∝ 8. (4.8)
t=0,|~
x|→0 |~x|
As a check, we note that, the short distance behavior of this expression reproduces the
zero-temperature answer |~x|−8 found in [8].
The long distance (|~x| ≫ rH ) behavior is
√
h i e− 2|~x|/rH
D(|~x| ≫ rH , 0) = (4∂t2 − (∇2 )2 )F ∝ . (4.9)
t=0,|~
x|→∞ |~x|3/2
The correlator is not ultra-local, unlike the thermal correlator in free scalar Lifshitz theory.
8
We would like to thank Shamit Kachru and Mike Mulligan for sharing the Mathematica file that computes
the spatial two-point function derived in [8].

11
5 Outlook

An important defect of our work which cannot have avoided the reader’s attention is the
fact that the matter content which produces the stress-energy tensor for this black hole is
unfamiliar and contrived. There is no physical reason why terms such as A2 B 2 should not
be added. In our defense, a perturbation analysis in the coefficient of such terms indicates
that a corrected solution can be constructed.
It is not clear how to embed such solutions in a UV-complete gravity theory; a stringy
description is not known yet even for the zero temperature case. Such a description would
help in finding specific Lifshitz-like field theories with gravity duals. It would be nice to
understand the connection (if it exists) between the Lifshitz spacetime and non-Abelian
Lifshitz-like gauge theories [23, 24].

Acknowledgements We thank Shamit Kachru and Mike Mulligan for correspondence

on attempts to construct more familiar actions with Lifshitz black hole solutions and for
helpful advice on Fourier transforms. We also thank Allan Adams, Sean Hartnoll, Alex
Maloney, Omid Saremi, T. Senthil and Larus Thorlacius for useful discussions and comments.
This work was supported in part by funds provided by the U.S. Department of Energy
(D.O.E.) under cooperative research agreement DE-FG0205ER41360 and by the National
Science Foundation under Grant No. NSF PHY05-51164.

A Regularizing the action and boundary stress tensor

In this appendix, we will show that the on-shell action and boundary stress tensor can be
rendered finite by making the action well-behaved, i.e. the action is stationary on-shell under
an arbitrary normalizable variation of the bulk fields, and the boundary terms in the action
must not break the residual gauge symmetries of the metric.
We will first find the constraints imposed by finiteness of the free energy, internal energy
and pressure on ci .

12
 The free energy of the boundary theory is
"
Sonshell 1 64cN − 8c0 + 16c1 + 8c2 + 6c3 + 16c4 − 15c5
−F = = Lx Ly 4
β 2 32rH
#
32 + 4c1 + 8c0 + 6c3 + 2c4 − 5c5 24 + 2c3 − c5 − 8cN + 8c0
− 2
+ (A.1)
16ǫ2 rH 8ǫ4
where β is inverse temperature. We must set −c5 + 24 − 8cN + 8c0 + 2c3 = 0 and −c1 −
8 − 2c0 − 3/2c3 − 1/2c3 + 5/4c5 = 0 to get rid of the divergences in the on-shell action.
Further, finiteness of the boundary stress tensor and conformal ward identities impose more
constraints on the counterterms.
The internal energy of the boundary theory is

√ 16 + 8c0 − 2c3 + 3c5 + 8cN

E = −Lx Ly γTtt = −Lx Ly (A.2)
8ǫ4
!
32 + 8c0 + 4c1 − 6c3 − 2c4 + 15c5 8c0 − 16c1 − 8c2 + 6c3 + 16c4 − 45c5 + 64cN
− 2 2
− 4
(A.3)
16rH ǫ 64rH

Similarly, the expression for pressure is

"
1 √ i √ x 1 64cN − 8c0 + 16c1 + 8c2 + 6c3 + 16c4 − 15c5
P = Lx Ly γTi = Lx Ly γTx = Lx Ly 4
2 2 32rH
#
32 + 4c1 + 8c0 + 6c3 + 2c4 − 5c5 24 + 2c3 − c5 − 8cN + 8c0
− 2
+ (A.4)
16ǫ2 rH 8ǫ4
Note that F = −P, as expected in the grand canonical ensemble. Hence, the condition
for the divergences in pressure to cancel is same as the condition for divergences in the
on-shell action to cancel. However, finiteness of energy imposes additional constraints on
the counterterms. In the case of Schrödinger black hole, it is not possible to get rid of the
divergence in the energy without the Neumannizing term [19].
The conformal Ward identity for conservation of the dilatation current requires zE = dP,
and in our discussion d = z = 2. The residual gauge freedom of the metric is broken if this
condition is not satisfied (see [16]). Note that making the boundary stress tensor finite does
not ensure this condition. We must set c2 = 7/2 for the conformal Ward identity to hold.
After imposing these conditions, we find
15 − 2c1 − 26cN
E = P = −F = Lx Ly 4
(A.5)
16rH

13
 In order to have a well-defined variational principle, we must ensure that δS = 0 onshell.
We shall now determine the value of c1 using this condition9 . The variation of the action is
"
1 3 √
Z Z
d x γ Tνµ δγµν + (cN − 1)e−2Φ nν F νµ + c3 + c4 Φ + 2c5 A2 Aµ δAµ +



δS = EOM+
bulk 2 bdy
#
1
cN Aµ δ nν e−2Φ Fνµ − c1 + 2c2 Φ − c4 A2 − 4cN Aµ nν Fνµ e−2Φ δΦ



(A.6)
2
The first term vanishes onshell. Therefore, the boundary terms must also vanish onshell.
Let us assume, for convenience that Dirichlet boundary condition is imposed on the gauge
field (cN = 0). Prescribing boundary conditions is equivalent to prescribing the coefficient
of the non-normalizable mode of the solution. The allowed variations at the boundary fall
faster than the non-normalizable part of the solution, i.e.,

µ µ
δγνµ = δγν(1) r 2 + δγν(2) r4 + . . .

(A.7)


δAµ = r −2 δAµ(1) r 2 + δAµ(2) r 4 + . . .

δΦ = r 2 δΦ(1) + r 4 δΦ(2) + . . .

Substituting these expressions in (A.6) and using the conditions on ci for energy and pressure
to be finite10 , we find
   
√ µ 2 ν c2 − c1
Z
3 2 2




δS = d x γTν r δγµ(1) + O r δAµ(1) + 2
δΦ1 + O r (A.8)
rH
Since E and P are finite, the first term in the integrand vanishes at the boundary. Hence,
c1 = c2 = 7/2 for the variation of the action to vanish on-shell. Using the values of ci found
above in (A.5) we get
Lx Ly
E = P = −F = 4
2rH
After restoring factors of 8πG,
Lx Ly 1 ∂F 1
E = P = −F = 4
=− T = TS
16πGrH 2 ∂T 2
We have shown that the stress tensor and on-shell action can be regularized by making the
action well-behaved, i.e. δS must vanish on-shell and the counterterms should not break any
residual gauge symmetry.
9
We have determined the value of c1 for the case where Dirichlet boundary condition is imposed on the
gauge field. However, the method is general and can be used for other boundary conditions as well.
10
c0 = −(17 − c1 )/8, c3 = −5 − c1 , c4 = −2c1 and c5 = −3 − c1 , when cN = 0.

14
References

[1] S. Sachdev and M. Mueller, “Quantum Criticality and Black Holes,” arXiv:0810.3005
[[Link]-el].

[2] S. A. Hartnoll, “Lectures on holographic methods for condensed matter physics,”

arXiv:0903.3246 [hep-th].

[3] E. Ardonne, P. Fendley and E. Fradkin, “Topological order and conformal quantum
critical points,” Annals Phys. 310, 493 (2004) [arXiv:cond-mat/0311466].

[4] B. Hsu, M. Mulligan, E. Fradkin, E. A. Kim, “ Universal entanglement entropy in 2D

Quantum critical points” arXiv:0812.0203 [[Link]-mech].

[5] E. Fradkin, “Scaling of Entanglement Entropy at 2D Quantum Lifshitz fixed points and
topological fluids” arXiv:0906.1569 [[Link]-el].

[6] D. S. Park, “Graviton and Scalar Two-Point Functions in a CDL Background for General
Dimensions,” JHEP 0906, 023 (2009) [arXiv:0812.3172 [hep-th]].

[7] P. Ghaemi, A. Vishwanath, T. Senthil, “Finite temperature properties of quantum

Lifshitz transitions between valence bond solid phases: An example of ‘local’ quantum
criticality,” Phys. Rev. B 72, 024420 (2005) arXiv:cond-mat/0412409v1 [[Link]-
el].

[8] S. Kachru, X. Liu and M. Mulligan, “Gravity Duals of Lifshitz-like Fixed Points,” Phys.
Rev. D 78, 106005 (2008) [arXiv:0808.1725 [hep-th]].

[9] A. Adams, A. Maloney, A. Sinha and S. E. Vazquez, “1/N Effects in Non-Relativistic

Gauge-Gravity Duality,” JHEP 0903, 097 (2009) [arXiv:0812.0166 [hep-th]].

[10] M. Taylor, arXiv:0812.0530 [hep-th].

[11] U. H. Danielsson and L. Thorlacius, “Black holes in asymptotically Lifshitz spacetime,”

arXiv:0812.5088 [hep-th].

[12] N. Lashkari, A. Maloney, O. Saremi, unpublished.

15
[13] T. Azeyanagi, W. Li and T. Takayanagi, “On String Theory Duals of Lifshitz-like Fixed
Points,” arXiv:0905.0688 [hep-th].

[14] W. Li, T. Nishioka and T. Takayanagi, “Some No-Go Theorems for String Duals of
Non-Relativistic Lifshitz-Like Theories,” arXiv:0908.0363 [hep-th].

[15] W. Li, W. Song and A. Strominger, “Chiral Gravity in Three Dimensions,” JHEP
0804 (2008) 082 [arXiv:0801.4566 [hep-th]]; D. Anninos, W. Li, M. Padi, W. Song and
A. Strominger, “Warped AdS3 Black Holes,” JHEP 0903 (2009) 130 [arXiv:0807.3040
[hep-th]]; A. Maloney, W. Song and A. Strominger, “Chiral Gravity, Log Gravity and
Extremal CFT,” arXiv:0903.4573 [hep-th].

[16] D. T. Son, “Toward an AdS/cold atoms correspondence: a geometric realization of the

Schroedinger symmetry,” Phys. Rev. D 78, 046003 (2008) [arXiv:0804.3972 [hep-th]].

[17] K. Balasubramanian and J. McGreevy, “Gravity duals for non-relativistic CFTs,” Phys.
Rev. Lett. 101, 061601 (2008) [arXiv:0804.4053 [hep-th]].

[18] C. P. Herzog, M. Rangamani and S. F. Ross, “Heating up Galilean holography,” JHEP

0811, 080 (2008) [arXiv:0807.1099 [hep-th]].

[19] A. Adams, K. Balasubramanian and J. McGreevy, “Hot Spacetimes for Cold Atoms,”
JHEP 0811, 059 (2008) [arXiv:0807.1111 [hep-th]].

[20] J. Maldacena, D. Martelli and Y. Tachikawa, “Comments on string theory backgrounds

with non-relativistic conformal symmetry,” JHEP 0810, 072 (2008) [arXiv:0807.1100
[hep-th]].

[21] V. Balasubramanian and P. Kraus “A Stress Tensor For Anti-de Sitter Gravity,”
arXiv:9902121 [hep-th].

[22] D. T. Son and A. O. Starinets, “Minkowski-space correlators in AdS/CFT correspon-

dence: Recipe and applications,” JHEP 0209, 042 (2002) [arXiv:hep-th/0205051].

[23] M. Freedman, C. Nayak, K. Shtengel, Phys. Rev. Lett. 94, 147205 (2005)
[arXiv:cond-mat/0408257]

16
[24] P. Horava, “Quantum Criticality and Yang-Mills Gauge Theory,” arXiv:0811.2217 [hep-
th].

[25] D. Birmingham, I. Sachs and S. Sen, “Exact results for the BTZ black hole,” Int. J.
Mod. Phys. D 10, 833 (2001) [arXiv:hep-th/0102155].

[26] A. W. Peet, L. Susskind and L. Thorlacius, “Tachyon hair on two-dimensional black

holes,” Phys. Rev. D 48, 2415 (1993) [arXiv:hep-th/9305030].

[27] R. B. Mann, JHEP 0906 (2009) 075 [arXiv:0905.1136 [hep-th]]; G. Bertoldi, B. A. Bur-
rington and A. Peet, arXiv:0905.3183 [hep-th], arXiv:0907.4755 [hep-th].

[28] S. F. Ross and O. Saremi, “Holographic Stress Tensor for Non-Relativistic Theories,”
arXiv:0907.1846 [hep-th].

17