Home Search Collections Journals About Contact us My IOPscience

Pressure and volume in the first law of black hole thermodynamics

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2011 Class. Quantum Grav. 28 235017

(http://iopscience.iop.org/0264-9381/28/23/235017)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 128.59.222.12
This content was downloaded on 16/02/2015 at 21:40

Please note that terms and conditions apply.

IOP PUBLISHING CLASSICAL AND QUANTUM GRAVITY
Class. Quantum Grav. 28 (2011) 235017 (13pp) doi:10.1088/0264-9381/28/23/235017

Pressure and volume in the first law of black hole

thermodynamics
Brian P Dolan
Department of Mathematical Physics, National University of Ireland, Maynooth, Ireland
and
Dublin Institute for Advanced Studies, 10 Burlington Rd, Dublin, Ireland

E-mail: [email protected]

Received 17 August 2011, in final form 27 October 2011

Published 17 November 2011
Online at stacks.iop.org/CQG/28/235017

Abstract
The mass of a black hole is interpreted, in terms of thermodynamic potentials,
as being the enthalpy, with the pressure given by the cosmological constant.
The volume is then defined as being the Legendre transform of the pressure,
and the resulting relation between volume and pressure is explored in the case
of positive pressure. A virial expansion is developed and a van der Waals like
critical point determined. The first law of black hole thermodynamics includes
a PdV term which modifies the maximal efficiency of a Penrose process. It
is shown that, in four-dimensional spacetime with a negative cosmological
constant, an extremal charged rotating black hole can have an efficiency of up
to 75%, while for an electrically neutral rotating black hole this figure is reduced
to 52%, compared to the corresponding values of 50% and 29% respectively
when the cosmological constant is zero.

PACS numbers: 04.60.−m, 04.70.Dy

(Some figures may appear in colour only in the online journal)

1. Introduction

The thermodynamics of black holes is a rich and fascinating area of research which continues
to yield surprises. The first law of black hole thermodynamics is usually written as
dM = T dS +  dJ +  dQ, (1)
κ
where T = 2π
is the Hawking temperature of the black hole (with κ being the surface gravity),
S = A4 is the entropy (with A being the area in Planck units),  is the angular velocity, J is
the angular momentum,  is the electrostatic potential difference between infinity and the
horizon, Q is the electric charge and M is the mass. The mass is usually interpreted as the
internal energy, in the thermodynamic sense, of the black hole, but it was suggested in [1]

0264-9381/11/235017+13$33.00 © 2011 IOP Publishing Ltd Printed in the UK & the USA 1
Class. Quantum Grav. 28 (2011) 235017 B P Dolan

that it is more correctly interpreted as the enthalpy. In this context, it is notable that there
is no PdV term in (1), corresponding to a change in volume at ambient pressure P. When a

cosmological constant, , is included there is a natural candidate for a pressure, P = − 8π , and
it was proposed in [2] that the volume of the black hole can be defined as the thermodynamic
variable conjugate to P. Interpreting the mass as the enthalpy, equation (1) should then be
modified to
dM = T dS + V dP +  dJ +  dQ, (2)

∂M 
where the thermodynamic volume is defined to be V = ∂P S,J,Q
[1]. The idea that  should be
thought of as a thermodynamic variable that can be varied is not new and has been considered
by a number of authors [3–9]. Equation (2), with Q = 0, was studied in [8], in the context of
varying , without any particular physical interpretation being given to the thermodynamic
conjugate of , often denoted .
One may question whether it is appropriate to identify  with a thermodynamic
pressure. While a cosmological constant gives a pressure term in Einstein’s equations, a fluid
dynamical pressure is not necessarily the same as a thermodynamic pressure. In equilibrium
situations, however, it is presumably correct to identify the fluid dynamical pressure with the
thermodynamic pressure and we shall do so here.
In general, the enthalpy, H, is the heat energy, beloved of chemists; it is not the internal
energy, U, of the first law of thermodynamics. That distinction goes to the Legendre transform
of the enthalpy,
U = H − PV, (3)
where H is a function of S, P, J and Q, while U (S, V, J, Q) is a function of purely extensive
variables. Then we get the usual form of the first law,
dU = T dS − PdV +  dJ +  dQ. (4)
This equation was written down in [9], using  −  notation, but its consequences were not
pursued.
When  is non-zero, we should expect the PdV term to contribute to the mechanical energy
that can be extracted from a black hole by a Penrose process for example. For a negative 
(positive pressure), the PdV term gives a positive contribution to dU if the black hole shrinks,
and the PdV term reduces the amount of energy available for extraction as mechanical work
W , with dW = −dU, hence reducing the efficiency. However, we shall show that the maximal
efficiency actually increases, relative to the  = 0 case, when  < 0, because the maximal
angular momentum of a black hole in anti-de Sitter (AdS) is greater than that of with  = 0
and this can outweigh the reduction in efficiency; the engine may not be as efficient at a given
J, but it can be pushed to a higher J. Conversely, one would expect that for a positive  more
energy becomes available at a given J, relative to  = 0, as the black hole shrinks.
Of course, there are no pistons pushing against a gas for a black hole, but a negative
cosmological constant contributes a negative energy density to spacetime, so a shrinking black-
hole exposes negative energy, thus increasing the black hole’s internal energy and decreasing
the amount of energy available for mechanical work. Conversely, a positive cosmological
constant would presumably release extra energy as the black hole shrinks, which can be used
to do work, hence increasing the efficiency at a given J.
The most efficient way to extract energy from a black hole is in an isentropic process,
with dS = 0 and the area of the event horizon constant. So
dU  dUmin = −P dV +  dJ +  dQ. (5)

2
Class. Quantum Grav. 28 (2011) 235017 B P Dolan

We shall see that, for a rotating black hole, it is possible to reduce V while keeping S constant.
The maximum amount of mechanical work that can be extracted in passing from an initial
state i to a final state f is
 f
Wmax = − dUmin . (6)
i
The efficiency is defined to be the ratio of the mechanical energy extracted to the initial heat
energy (enthalpy),
Wmax
η= , (7)
Mi
where the initial enthalpy is identified with the initial mass, Mi . It will be shown that, for
 < 0, this can be as high as 52% for a rotating black hole and 75% for a charged black hole.
The volume of a black hole has only recently been considered as a thermodynamic variable
[2, 13]. At the simplest level, there is a natural tendency to assume that the area and the volume
are related geometrically and are not independent. For a Schwarzschild black hole with radius
rh , for example, the area is of course 4π rh2 and indeed the thermodynamic volume works out
to be 4π r3 , but this seems to be coincidental and no particular significance should be attached
3 h
to it. At a deeper level, it is not even clear how to define a volume geometrically, as the
metric is not static in the interior of a black hole. For a Schwarzschild black hole, which is
not rotating, we shall see that the thermodynamic volume and the area are not independent;
fixing S fixes V so dV = 0 in an isentropic process, and the PdV term does not contribute to
the first law. But for a rotating black hole the area of the event horizon does not determine the
thermodynamic volume uniquely, and it is possible to vary the volume keeping the entropy
constant by changing the angular momentum and/or the charge.
The properties of the thermodynamic volume and its contribution to the first law of black
hole thermodynamics are explored in detail in this paper for a rotating charged black hole in
four-dimensional spacetime with a negative . In section 2, thermodynamic potentials and the
equation of state are discussed and the Legendre transform from the enthalpy to the internal
energy is given explicitly. In section 3, the efficiency of a Penrose type process is analysed and
section 4 contains a discussion and outlook. Two appendices are dedicated to the technicalities
of deriving some results used in the paper.

2. The internal energy

Including a pressure term in the first law gives

dU = T dS − PdV +  dJ +  dQ,
where the internal energy, U (S, V, J, Q), is a function of extensive variables. The
thermodynamic volume, V , is the conjugate variable to the pressure and is obtained from
the mass, which is identified in [1] with the enthalpy, M = H(S, P, J, Q), by

∂H 
V = . (8)
∂P  S,J,Q
It was proposed in [2] that (8) can be defined to be the thermodynamic volume of the black
hole.
The line element for a charged rotating black hole in four-dimensional AdS space is [10]
 2  2

a sin2 θ ρ2 2 ρ 2 2
θ sin2 θ r2 + a2
ds = − 2 dt −
2
dφ + dr + dθ + a dt − dφ ,
ρ

θ ρ2
(9)

3
Class. Quantum Grav. 28 (2011) 235017 B P Dolan

where
(r2 + a2 )(L2 + r2 ) a2

= − 2mr + q2
,
θ = 1 − cos2 θ ,
L2 L2 (10)
a2
ρ = r + a cos θ ,
2 2 2 2
= 1 − 2,
L
and the cosmological constant is  = − L32 , which is related to the pressure by L12 = 8πP
3
.
The physical properties of this spacetime are well known [11], and the first law, applied
to this metric, was discussed in [12], but without a PdV term. The metric parameters m and q
are related to the mass and charge by
m q
M = 2, Q= . (11)

The event horizon, r+ , lies at the largest root of
(r) = 0, so
 2 
r+ + a2 L2 + r+ 2
+ q2 L2
M= , (12)
2r+ L2 2
and the area of the event horizon is
 2
r+ + a2
A = 4π . (13)

The temperature is
 2 
L + 3r+ 2 2
r+ − a2 L2 − r+ 2
− q2 L2
T =  2 . (14)
4π L2 r+ r+ + a2
The angular momentum J = aM, and the relevant thermodynamic angular velocity is

a L 2 + r+
2
 = 2 2 . (15)
L r+ + a2
The electrostatic potential is
qr+
= 2 . (16)
r+ + a2
One can scale L out from all the above expressions by defining dimensionless variables
M = M/L, ā = a/L, r̄+ = r+ /L, etc, but we prefer to keep L explicit to expose more clearly
the role of the pressure, and make the comparison with the L → ∞ limit clear.
Under the assumptions made here, the thermodynamic volume for the Kerr–Newman–AdS
black hole works out to be

  2 2 
2π r+2
+ a2 2r+ L + a2 L2 − r+ a + L2 q2 a2
2 2
V = , (17)
3 L 2 2 r+
which is a simple generalization of the Kerr–AdS volume derived in [13]. A direct derivation
of (17) from (12) is most easily achieved by first writing the mass as a function of (S, P, J, Q),
differentiating with respect to P, and then transforming back to (r+ , a, q, L).
When a = 0 the area (13) and volume (17) are not independent and the area determines
the volume uniquely, but when a is non-zero the area and the volume become independent.
For asymptotically flat space, with L → ∞, one has
 2  2
2π r+ + a2 2r+ + a2 + q2 a2
V = . (18)
3 r+
4πr3
While this reduces to the naı̈ve result, 3 + , for the Schwarzschild black hole, a geometrical
interpretation when a is non-zero is not so clear.

4
Class. Quantum Grav. 28 (2011) 235017 B P Dolan

A correct description of the thermodynamics of the black hole, in terms of thermodynamic

potentials, requires replacing the geometric variables (r+ , L, a, q) with thermodynamic
variables (S, P, J, Q). The relevant expression for the mass, and hence the enthalpy, was
derived in [9],

 2 2 
1 S + π Q2 + 8PS + 4π 2 1 + 8PS J2
H(S, P, J, Q) := 3 3
. (19)
2 πS
This generalizes the Christodoulou–Ruffini formula [14] for the mass of a rotating black hole
in terms of its irreducible mass, Mirr . The irreducible mass for a black hole with entropy S is
the mass of a Schwarzschild black hole with the same entropy, Mirr 2
= 4πS
.
In terms of thermodynamic variables the temperature is [9]
     2 
∂H  1 π Q2 8PS π Q2 2 J
T = = 1+ + 1− + 8PS − 4π (20)
∂S J,Q,P 8π H S 3 S S
and the thermodynamic volume is
    
∂H  2 8PS2
V = = S S + πQ +
2
+ 2π J ,
2 2
(21)
∂P S,J,Q 3π H 3
which is manifestly positive.
The Legendre transform U = H − PV gives the thermal energy, a function of purely
extensive variables. The transform is evaluated in appendix A to be
 π 3  3V   S   S  
U (S, V, J, Q) = +Q +J
2 2
S 4π 2π π

  ⎤
 3  12  2  12
3V 2 S SQ
−|J| − + J2 ⎦ . (22)
4π π π
Dimensional analysis implies that U is only a function of three independent variables, since
U → λU when S → λ2 S, V → λ3V , J → λ2 J and Q → λQ.
One must be careful taking the J → 0 limit of these potentials. In this limit, the enthalpy
  
1 S π Q2 8SP
H(S, P, 0, Q) = 1+ + (23)
2 π S 3
is linear in P and the Legendre transform is singular,
  
∂H  4π S 3/2
V = = , (24)
∂P  S,Q 3 π
is independent of P and so cannot be inverted to obtain P(V ). Conversely, when J is zero, the
Legendre transform of (23) is
  
1 S π Q2
U= 1+ , (25)
2 π S
which is independent 3 the volume and is not equal to the J → 0 limit of (22), unless a
 3V 2  S of
constraint, 4π = π , is imposed. Indeed, (25) gives the wrong J = 0 temperature (unless
Q = P = 0, in which case U = H and T = ∂U ∂S
= 4πr1
+
is the correct Hawking temperature
for a Schwarzschild black hole). To get the correct temperature from (22) in the J → 0 limit,
we must take the partial
 derivative with respect to S before setting J to zero, and take note of
3V 2 3
the fact that 4π = πS when J = 0. This constraint can be derived from (22) directly by
 3V 2  S 3  2 SQ2
observing that U is not differentiable with respect to J at J = 0 unless 4π − π J + π
vanishes there. It can also be seen directly when a = 0 in (13) and (17).

5
Class. Quantum Grav. 28 (2011) 235017 B P Dolan

To derive the relation between the pressure and the volume in general, we first define
v := 4π
3V
and s := πS . Then (22) becomes
1 v 2  
U (s, v, J, Q) = 3 (s + sQ2 + 2J 2 ) − |J| (v 2 − s3 )(J 2 + sQ2 ) (26)
s 2
with temperature
1 ∂U |J|{3J 2 (2v 2 − s3 ) + sQ2 (5v 2 − 2s3 )} v(s2 + 2sQ2 + 6J 2 )
T = =  − (27)
π ∂s 2π s4 (v 2 − s3 )(J 2 + sQ2 ) 2π s4
and pressure

3 ∂U 3v|J| J 2 + sQ2 3
P=− = − (s2 + sQ2 + 2J 2 ). (28)
4π ∂v 4π s3 v 2 − s3 8π s3
√
In the J → 0 limit, |J| and v 2 − s3 must vanish together for finite T and P. The equation of
state, in the form of the relation between the pressure, the temperature and the volume can be
obtained by eliminating s between (27) and (28). In the limit J → 0, the temperature is
8π v 4/3 P + v 2/3 − Q2
TJ→0 = , (29)
4π v
which is the correct temperature as a function of P, Q and s, as can be seen from (14), with
a = 0, using (11), (13), (17) and J = aM. When Q = 0, (29) reduces to the equation of state
in [2].
The small J correction to (29) is
 4 
8π Pv 2 + v 4/3 − Q2 v 2/3 − 6J 2 J J 2 Q2
T = + O , . (30)
4π v 5/3 v 9/3 v 9/3
The angular momentum term dominates at small volumes for any J > 0. The first few terms
in an expansion for the pressure are
 4 
T 1 Q2 3J 2 J J 2 Q2
P = 1/3 − + + + O , . (31)
2v 8π v 2/3 8π v 4/3 4π v 2 v 10/3 v 10/3
(The virial expansion for Q = 0 is developed in more detail in appendix B.)
The P–V diagram is plotted in this order in figure 1 with Q = 0 and J = 1 (for clarity
the horizontal axis in figure 1 is v 1/3 rather than v). There is a critical point at Tc ≈ 0.0413,
Pc ≈ 0.002 80 and vc1/3 ≈ 3.08. This critical point represents a second-order phase transition,
similar in nature to that of a van der Waals gas—it is the same critical point that was found in
[9] by fixing the pressure and varying J (the authors of [9] use P = 8π 3
= 0.119, corresponding
to L = 1, and find a critical value of J at Jc = 0.0236 giving 8π = 0.002 82). The corrections
3Jc

in equation (31) become large when v is very small, but at the critical point the error in the
pressure is of order ∼ v10/31
∼ 10−5 , or about 1% of Pc . When v 1/3  2, the corrections are
of the same order as P and the virial expansion breaks down. In the limit, when v is small
3J 2
relative to J, a large J expansion shows that the leading-order behaviour is P ≈ 8πv 2 , which is

the same low v behaviour as that in figure 1 but reduced by a factor 2.

Thus by taking the J → 0 limit carefully the temperature can be kept finite when v 2 → s2 ,
but the non-analyticity in (22) is still present in the second derivative with respect to S, which
diverges when J = 0 and v 2 = 3
 s . This has an important consequence that the heat capacity
at constant volume, CV = T/ ∂T ∂S V,J,Q
, tends to zero when J = 0, as observed in [2]. CV can
be non-zero for J = 0, and it is plotted in figure 2 for Q = 0 as a function  of S and J, with
V = 1. For comparison, the heat capacity at constant pressure, CP = T/ ∂T ∂S P,J,Q
, is plotted in
figure 3 for Q = 0 as a function of S and J, with P = 1. CP vanishes when T = 0 and diverges
when ∂T ∂S
= 0.

6
Class. Quantum Grav. 28 (2011) 235017 B P Dolan

0.005

0.004

0.003

0.002

0.001

0
2 4 6 8 10

Figure 1. P–V diagram for J = 1 and Q = 0, plotted using the approximation in equation (31).
1
P is plotted as a function of v 3 for T = 0.02, 0.025, 0.3, 0.35, 0.04, 0.045 and 0.05. The critical
point for this value of J is Tc ≈ 0.0413, Pc ≈ 0.002 80 and vc ≈ 3.08. J > 0 causes the rapid rise
in P at low values of v.

A full stability analysis was carried out in [9] and figure 4 shows the phase diagram, when
Q = 0, in terms of S/L2 and J/L2 . The top curve, labelled I, is the T = 0 limit; the region above
and to the left of this curve is unphysical because T < 0 there. CP in figure 3 vanishes on curve
I. Curve II is the curve on which  = 1/L, below which the black hole can be in equilibrium
with radiation rotating at infinity [15]. On this curve, asymptotic spacetime is conformal to
the (2 + 1)-dimensional Einstein universe rotating at the speed of light; the region below and
to the right of this curve is of relevance to conformal field theories. Curve III is the curve
marking the edge of local stability, the determinant of the Hessian of the Gibbs free energy
is divergent on this curve. The heat capacity, CP , diverges on curve IV, which is blown up on
the right-hand picture (this curve was not included in the analysis of reference [9]). Curve V
marks the Hawking–Page phase transition; below this curve the black hole is globally stable.
The three curves marking the Hawking–Page transition, the edge of local stability  and the
2
 = 1/L curve all have similar asymptotic forms for large S/L2 , LJ2 ≈ 2πS2 L4 + O LS2 .
In the following section, it will be more convenient to use the variables S, P, J and Q,
rather than S, V , J and Q, in terms of which
 2  2
(S + π Q2 ) S + π Q2 + 8PS + 4π 2 1 + 4PS J
U =   3

3
 , (32)
2 2
2 π S S + π Q2 + 8PS 3
+ 4π 2 1 + 8PS J 2
3
which is manifestly positive. In terms of geometrical variables,
 2 
r+ + a2 2L4 + a2 r+ 2
− a2 L2 + q2 L2 (2L2 − a2 )
U= . (33)
4L4 2 r+

7
Class. Quantum Grav. 28 (2011) 235017 B P Dolan

Figure 2. CV as a function of S/L2 and J/L2 , with the fixed volume set to L3 .

3. Extracting energy from rotating black holes

For an electrically neutral black hole, we set Q = 0 in (32) and the internal energy U = H −PV ,
written in terms of S, L2 = 8πP 3
and J, is
 
S 1 + πLS 2 + 4π 2 J 2 1 + 2πL
2 S
2
U=     . (34)
2 π S 1 + πLS 2 S2 1 + πLS 2 + 4π 2 J 2
In an isentropic, isobaric process the black hole can yield mechanical work by decreasing the
angular momentum. If J is reduced from some finite value to zero the efficiency is
U (J) − U (0)
η=
H(J)
  2
S2 1 + πLS 2 + 4π 2 1 + 2πL S
2 J S
=    −    . (35)
1 + πL2 S 1 + πL2 + 4π J
S 2 S 2 2
1+ S S2 1 + S + 4π 2 J 2
πL2 πL2
For a given S and L there is a maximal value of J determined by demanding that T > 0, which
requires
  
S2 S 3S
J <
2
1+ 1+ . (36)
4π 2 π L2 π L2
The greatest efficiency is for extremal black holes, when the bound (36) is saturated,
(4π L2 + 3S) π L2
η= − √ . (37)
2(2π L2 + 3S) (π L2 + S) 2π L2 + 3S
8
Class. Quantum Grav. 28 (2011) 235017 B P Dolan

Figure 3. CP as a function of S/L2 and J/L2 , with the fixed pressure set to 3
8π , i.e. L = 1. CP
diverges along curve IV in figure 4 and vanishes along curve I.

The asymptotically flat case ( = 0) is obtained by taking the limit L → ∞, η = 1 − √12 =

0.29 . . ., which is the familiar result [16]. For finite L, the efficiency is greater than the
asymptotically flat value. It has a maximum value of 0.5184... when S/L2 = 15.39... (obtained
by solving a quartic equation for S), and asymptotes to 0.5 for large S.
Differentiating (19) with respect to P, we see that the thermodynamic volume is a
monotonically increasing function of J; hence, the volume decreases as J is lowered, keeping
the area constant. The black hole also heats up during the process.
When the black hole is charged there are various possibilities for extracting work through
an isentropic, isobaric process; one could decrease J keeping Q constant or decrease Q keeping
J constant or decrease both simultaneously. Since all that matters are the initial and final values
of U, the initial and final values of J and Q uniquely determine the efficiency in such a process.
If both J and Q are both non-zero initially and are both decreased to zero, the efficiency
is

 S2
 2
(S + π Q2 ) S + π Q2 + πL 2 + 4π 2 1 + 2πL
S
2 J
η=  
S2 2
S + π Q2 + πL 2 + 4π 2 1 + πLS 2 J 2
S
−  . (38)
2 2 
S + π Q2 + πL2 + 4π 2 1 + πLS 2 J 2
S

9
Class. Quantum Grav. 28 (2011) 235017 B P Dolan

2 2 0.2
J/L
2
J/L

1.5 0.15

I
1 I 0.1 II

II
III III
0.5 0.05
V
IV

IV
0 0
0 1 2 3 4 0 0.2 0.4 0.6 0.8 1 1.2 1.4
2 2
S/L S/L
Figure 4. Phase diagram for q = 0, plotted in terms of S/L2 and J/L2 . The region above I is
forbidden because T < 0; in the region above curve II the three-dimensional Einstein universe
at infinity rotates faster than the speed of light; curve III bounds the region of local stability
(an analysis of the Gibbs free energy shows that the black hole is locally unstable above curve
III); CP diverges on curve IV; in the region above curve V the black hole is unstable due to the
Hawking–Page phase transition; in the region below curve V the black hole is stable.

The optimal efficiency is for extremal black holes, and the extremal value of J is given by
      2  2 
1 S 2 S 3S S Q
Jmax =
2
1+ 1+ +2 − Q , (39)
4
4 π π L2 π L2 π L
2
when T in (20) vanishes. Positivity of Jmax then imposes a restriction on the allowed range of
2
Q,
  
S 3S
0  Q2  1+ . (40)
π π L2
The greatest efficiency is for an extremal black hole with maximal charge,
  
S 3S
Q2max = 1+ , (41)
π π L2
when
π L2 + 3S
η= . (42)
2(π L2 + 2S)
For large S efficiencies of up to 75% are possible in principle for extremal charged, rotating
AdS black holes, though as we saw in the previous section this is reduced to 51.8% for
extremal rotating, but electrically neutral, black holes. For comparison, the equivalent figures
in asymptotically flat spacetime, with  = 0, are 50% and 29.3% respectively.

4. Conclusions

The thermodynamics of rotating black holes in four-dimensional anti-de Sitter (AdS) spacetime
has been discussed in detail, with particular attention paid to the role of pressure and volume.

10
Class. Quantum Grav. 28 (2011) 235017 B P Dolan

The negative cosmological constant is interpreted as a positive pressure and treated as

a thermodynamic variable whose conjugate thermodynamic variable is a thermodynamic
volume. For rotating black holes the thermodynamic volume is independent of the geometric
area, and hence independent of the entropy, allowing a full thermodynamic treatment.
The black hole mass is associated with the enthalpy H, as suggested in [1], and the internal
energy U = H − PV is lowered relative to the mass by a positive pressure. Vanishing angular
momentum is a non-analytic point of the thermodynamics and must be treated with some care;
in particular, one cannot fully understand black hole thermodynamics simply by focusing on
the non-rotating, Schwarzschild case—rotation is essential for a complete analysis.
The black hole equation of state has been analysed in terms of pressure and volume and
a virial expansion has been developed. Non-zero angular momentum causes a rapid rise in
pressure as the volume is reduced at constant temperature, a rise that is not present at zero
angular momentum. The critical point found in [9] is found to be very similar in form to that
of a van der Waals gas.
In a Penrose process, the volume decreases as the black hole looses angular momentum
and the PdV term in the first law reduces the amount of energy available to do work. However,
a negative cosmological constant also extends the range of the angular momentum, pushing
the maximum allowed value beyond that of the  = 0 case, so that in fact more work can
be extracted from an extremal black hole in a spacetime that is asymptotically AdS than from
the one in a spacetime that is asymptotically flat. Efficiencies of up to 75% are theoretically
achievable for a charged black hole in asymptotically AdS spacetime, compared to 50% in
asymptotically flat spacetime. For electrically neutral black holes, the corresponding figures
are 51.8% compared to 29.3%.
The analysis here has focused on the  < 0 case for three reasons. Firstly, there is an
ambiguity in the definition of thermodynamic variables in the  > 0 case; there are two
relevant event horizons to be considered and hence two different temperatures in general. For
a general choice of black hole parameters, it is not possible to make the Euclidean geometry
regular by a unique periodic identification of Euclidean time. Secondly,  > 0 implies negative
pressure and consequent thermodynamic instabilities. This latter problem is not necessarily a
fatal objection; black holes are well known to be unstable for  = 0 because the heat capacity,
more precisely CP , is negative, but we can still consider metastable situations in which a
temperature can be defined for a period of time much shorter than the timescale of thermal
instability. Indeed, there are regions of parameter space for  < 0 in which CP < 0, and
we can still learn something about physical properties of black holes in this regime. Negative
pressures can be useful in metastable situations [17]. Lastly, it is not even clear how to define
the mass of a black hole in a  > 0 spacetime as there is no accessible asymptotic regime in
which one can compare with the zero mass solution.

Appendix A

To calculate the internal energy U = H − PV , we write the enthalpy in the form


H = α + βP + γ P2 , (A.1)
where

 2 
π 1 S
α := +Q 2
+J 2
S 4 π
   
4π S S
β := + Q2 + 2J 2
3 π π
11
Class. Quantum Grav. 28 (2011) 235017 B P Dolan
   
4π 2 S 3
γ := . (A.2)
3 π
Note that the discriminant
 
64π 2 2 2 SQ2
β − 4αγ =
2
J J + (A.3)
9 π
is positive.
Now,

∂H  β + 2γ P 2HV − β
V = = ⇒ P= . (A.4)
∂P S,J,Q 2H 2γ
This allows us to re-express H as a function of V ,

1 β 2 − 4αγ
H= . (A.5)
2 V2 − γ
We can immediately conclude that
 2  3
4π S
V2 > , (A.6)
3 π
in agreement with the observation in [13].
It is now straightforward to determine
  
HV 2 βV βV (V 2 − γ )(β 2 − 4αγ )
U = H − PV = H − − = − , (A.7)
γ 2γ 2γ 2γ
which immediately gives (22) in the text.

Appendix B

In this appendix, the virial expansion is developed. For simplicity, we set Q = 0, but the
same techniques can be applied to the case of non-zero Q. To develop the expansion, we use
dimensionless variables
v s 8π PJ
y := 3/2 , x := , p := , t := 2π T J 1/2 , (B.1)
J J 3
in terms of which equations (27) and (28) can be written:
3(2y2 − x3 ) y(x2 + 6)
t=  − (B.2)
x4 y2 − x3 x4

2y x2 + 2
p=  − . (B.3)
x3 y2 − x3 x3
When y and x are large let y2 ≈ x3 , with y2 − x3 = z2 , then t is finite provided z ≈ 3
y2/3 t
in
which case p ≈ 3y2t1/3 . Replacing x with z in equations (B.2) and (B.3) gives
3(y2 + z2 ) = {t(y2 − z2 )4/3 + y(y2 − z2 )2/3 + 6y}z, (B.4)

(y2 − z2 )zp = 2y − z{(y2 − z2 )2/3 + 2}

2 1
⇒ p = − . (B.5)
z(y + z) (y2 − z2 )1/3
12
Class. Quantum Grav. 28 (2011) 235017 B P Dolan
2
Now, we expand in powers of u = y1/3
1
. Let z = 3ut ζ for some ζ (t, u), in terms of which (B.2)
becomes
 
 4/3  2/3 
9u10 ζ 2 9u10 ζ 2 9u10 ζ 2
t 1+ = t 1− +u 1− + 6u ζ ,
5
(B.6)
t2 t2 t2
from which we can immediately conclude that
t
ζ = + ω(t, u), (B.7)
t + u + 6u5
where an expansion of ω in u starts at order 10. Any desired order can be obtained by further
expanding ω:
∞
ω = u10 an (t )un , (B.8)
n=0
with the coefficients an (t ) to be determined. Putting (B.8) and (B.7) into (B.6), and equating
coefficients of powers of u, gives an iterative procedure for evaluating the coefficients an which
can then be used to show that
u10 u11 u12 u13 u14 2 + 3 t 4 15
ω = 7 2 − 16 3 + 27 4 − 40 5 + 55 6 − 36 u + O(u16 ). (B.9)
t t t t t t7
Finally, using this expansion in (B.5) gives the required virial expansion

2tu 1 2 u u2 u3 u4
p= − u + 2u − u 6 10
8 − 9 2 + 10 3 − 11 4
3 3 t t t t

u 5
1 + 6t 6
4
+ 12 5 − 13 u + O(u17 ), (B.10)
t t6
the first three terms of which are used in the text.

References

[1] Kastor D, Ray S and Traschen J 2009 Class. Quantum Grav. 26 195011 (arXiv:0904.2765 [hep-th])
[2] Dolan B P 2011 Class. Quantum Grav. 28 125020 (arXiv:1008.5023[gr-qc])
[3] Henneaux M and Teitelboim C 1984 Phys. Lett. B 143 415
Henneaux M and Teitelboim C 1989 Phys. Lett. B 222 195
[4] Teitelboim C 1985 Phys. Lett. B 158 293
[5] Wang S, Wu S-Q, Xie F and Dan L 2006 Chin. Phys. Lett. 23 1096 (arXiv:hep-th/0601147)
[6] Sekiwa Y 2006 Phys. Rev. D 73 084009 (arXiv:hep-th/0602269)
[7] Larrañaga Rubio E A 2007 Stringy generalization of the first law of thermodynamics for rotating BTZ black
hole with a cosmological constant as state parameter arXiv:0711.0012 [gr-qc]
[8] Wang S, Wu S-Q, Xie F and Dan L 2006 Chin. Phys. Lett. 23 1096 (arXiv:hep-th/0601147)
[8] Wang S 2006 Thermodynamics of high dimensional Schwarzschild–de Sitter spacetimes: variable cosmological
constant arXiv:gr-qc/0606109
[9] Caldarelli M M, Cognola G and Klemm D 2000 Class. Quantum Grav. 17 399 (arXiv:hep-th/9908022)
[10] Carter B 1968 Commun. Math. Phys. 10 280 (available at http://projecteuclid.org/euclid.cmp/1103841118)
[11] Henneaux M and Teitelboim C 1985 Commun. Math. Phys. 98 391
[12] Gibbons G W, Perry M J and Pope C N 2005 Class. Quantum Grav. 22 1503 (arXiv:hep-th/0408217)
[13] Cvetic M, Gibbons G W, Kubiznak D and Pope C N 2010 Black hole enthalpy and an entropy inequality for
the thermodynamic volume arXiv:1012.2888[gr-qc]
[14] Christodoulou D 1970 Phys. Rev. Lett. 25 1596
[14] Christodoulou D and Ruffini R 1971 Phys. Rev. D 4 3552
[15] Hawking S W, Hunter C J and Taylor-Robinson M M 1999 Phys. Rev. D 59 064005 (arXiv:hep-th/9811056)
[16] Wald R M 1984 General Relativity (Chicago, IL: University of Chicago Press)
[17] Landau L D and Lifshitz E M 1980 Statistical Physics: Part 1 3rd edn (Amsterdam: Elsevier) p 44

13