Thermodynamics in Alternative Gravity

Black-hole thermodynamics in an alternative metric theory is not obtained by inserting a corrected area into the Einstein formula. The entropy is the horizon charge selected by the theory’s covariant Lagrangian, while the mass, angular momentum, and other potentials are defined by compatible surface terms at infinity. For a stationary bifurcate horizon this construction gives a first law and a decisive Einstein-gravity check; it does not, without more work, establish a second law or an entropy for an arbitrary dynamical slice.

Helpful background. Einstein Equations provides the on-shell background; Gravitational Actions and Boundary Terms fixes the variational and asymptotic surface terms; Formulations of General Relativity supplies the covariant phase-space language; Black-Hole Geometry and Mechanics supplies Killing horizons, surface gravity, and binormals; Black-Hole Perturbation Theory supplies linearized solutions; and Relativistic Matter and Compact Objects distinguishes horizon thermodynamics from stellar equilibrium.

From a covariant Lagrangian to a surface charge

Let the dynamical fields be collectively $\Phi$ and let $\mathbf L$ be a diffeomorphism-covariant $D$-form. Its first variation has the form

$$\delta\mathbf L=\mathbf E_{\Phi}\,\delta\Phi+d\boldsymbol\theta(\Phi,\delta\Phi),$$

where $\boldsymbol\theta$ is the symplectic potential. For a vector field $\chi^a$, define the Noether current

$$\mathbf J_{\chi} =\boldsymbol\theta(\Phi,\mathcal L_{\chi}\Phi) -\chi\mathbin{\lrcorner}\mathbf L.$$

On shell, $d\mathbf J_{\chi}=0$, so locally $\mathbf J_{\chi}=d\mathbf Q_{\chi}$. Iyer and Wald prove the general local form of this charge, including Lagrangians with curvature derivatives, in Iyer–Wald 1994, § 4, Proposition 4.1 and Eqs. (49)–(60), pp. 10–12 (PDF).

If the background obeys $\mathcal L_{\chi}\Phi=0$ and $\delta\Phi$ solves the linearized equations, the symplectic identity reduces to

$$d\!\left(\delta\mathbf Q_{\chi} -\chi\mathbin{\lrcorner}\boldsymbol\theta\right)=0.$$

Integrating this closed form over a hypersurface turns a bulk variational problem into a relation between its outer boundary and a horizon cross-section. The hypotheses matter: an arbitrary off-shell metric variation does not satisfy this identity, and an asymptotic charge exists only when the falloff and boundary terms make its Hamiltonian integrable.

Wald entropy and its normalization

For an action written as

$$I=\int d^Dx\,\sqrt{-g}\,\mathcal L$$

with no covariant derivatives of the Riemann tensor, define

$$P^{abcd}=\frac{\partial\mathcal L}{\partial R_{abcd}}$$

while respecting the algebraic symmetries of $R_{abcd}$. On a smooth bifurcation surface $B$, normalize the binormal by $\epsilon_{ab}\epsilon^{ab}=-2$. The stationary entropy is

$$S_{\mathrm W}=-2\pi\int_B d^{D-2}x\,\sqrt h\, P^{abcd}\epsilon_{ab}\epsilon_{cd}.$$

This is the stationary Noether-charge prescription introduced in Wald 1993, Eqs. (25)–(26), p. 12 (PDF).

The normalization check is immediate for Einstein–Hilbert gravity,

$$\mathcal L_{\mathrm{EH}}=\frac{R}{16\pi G}, \qquad P^{abcd}_{\mathrm{EH}} =\frac{g^{ac}g^{bd}-g^{ad}g^{bc}}{32\pi G}.$$

Contracting twice with the binormal gives $-4/(32\pi G)$, hence $S_{\mathrm W}=A_B/(4G)$. This check catches the two common normalization errors: differentiating $R$ as though $R_{abcd}$ lacked antisymmetries, or choosing a unit-normalized binormal instead of one with norm $-2$.

For Lagrangians containing $\nabla R_{abcd}$, $P^{abcd}$ must be replaced by the appropriate Euler derivative, with alternating covariant derivatives. Quoting the simple partial-derivative formula outside its domain silently omits entropy terms.

First law on a bifurcate stationary horizon

Let

$$\chi^a=t^a+\Omega_H^{(i)}\phi^a_{(i)}$$

be the horizon generator, with $t^a$ normalized as unit time translation at infinity. On $B$, $\chi^a=0$ and $\nabla_a\chi_b=\kappa\epsilon_{ab}$. The boundary identity then yields

$$\delta M-\sum_i\Omega_H^{(i)}\delta J_{(i)} -\sum_A\Phi_H^{A}\delta Q_A =\frac{\kappa}{2\pi}\,\delta S_{\mathrm W}, \qquad T_H=\frac{\kappa}{2\pi}.$$

The gauge-work sign shown here corresponds to moving $\Phi_H^A\delta Q_A$ to the left; a different definition of the canonical gauge charge may move the sign. What is invariant is the single covariant-phase-space identity with one consistent orientation and generator normalization. Iyer and Wald state the diffeomorphism-covariant theorem for asymptotically flat stationary black holes with bifurcate Killing horizons and arbitrary asymptotically flat linearized perturbations in Iyer–Wald 1994, § 6, Theorem 6.1 and Eqs. (91)–(103), pp. 17–18 (PDF). Explicit gauge-work terms require the corresponding gauge choice, regularity, and asymptotic analysis; they are not produced by dimensional analogy.

Notice also that the Noether charge of an asymptotic time translation is not generally the energy by itself. The Hamiltonian contains the compensating surface term determined by $\boldsymbol\theta$; in Einstein gravity it resolves the familiar factor-of-two difference between Komar and ADM masses Iyer–Wald 1994, § 6, Eqs. (79)–(90), pp. 15–16 (PDF).

Branches, boundaries, and effective-theory limits

Thermodynamic comparisons must specify the ensemble and branch. At fixed couplings, compare solutions with the same asymptotic sources and conserved charges; a shift in a scalar’s boundary value can add a conjugate work term. A Euclidean free energy additionally depends on the boundary action and counterterm prescription. Two roots at the same temperature need not be continuously connected, and a first-law check along one branch says nothing about the global dominance of that branch.

The covariant construction has controlled ambiguities. Adding an exact form to $\mathbf L$, shifting $\boldsymbol\theta$ by an exact form, or changing the representative of $\mathbf Q$ changes intermediate expressions. On a stationary bifurcation surface the entropy and its first variation are protected under the hypotheses of the theorem; on a general nonstationary cross-section the analogous Jacobson–Kang–Myers ambiguity is consequential. Iyer and Wald track these terms in Iyer–Wald 1994, §§ 3–4, Eqs. (43), (58), and (59), pp. 9–12 (PDF).

In an effective theory, evaluate both the solution and $S_{\mathrm W}$ consistently to the same order in the higher-curvature couplings. A large correction near a small horizon can signal that omitted operators are equally important. Topological densities may shift entropy by a coupling-dependent constant without changing local field equations; this can matter in comparisons involving topology change but not in a variation within a fixed topology. Finally, the first law is a linear variational identity. Positivity of entropy, linear stability, and a classical or generalized second law are independent questions.

Exercise: entropy in f(R) gravity

For

$$I=\frac{1}{16\pi G}\int d^Dx\sqrt{-g}\,f(R),$$

derive the stationary Wald entropy and recover Einstein gravity.

Solution

Because $R=g^{ac}g^{bd}R_{abcd}$ with the Riemann symmetries included,

$$P^{abcd} =\frac{f'(R)}{32\pi G} \left(g^{ac}g^{bd}-g^{ad}g^{bc}\right).$$

Its binormal contraction is $-f'(R)/(8\pi G)$. Therefore

$$S_{\mathrm W} =\frac{1}{4G}\int_B d^{D-2}x\sqrt h\,f'(R).$$

If $R$ is constant on $B$, this is $f'(R_H)A_B/(4G)$. Choosing $f(R)=R-2\Lambda$ gives $f'(R)=1$ and recovers $A_B/(4G)$, independent of $\Lambda$. If $f'(R)$ varies over the horizon, pulling it outside the integral is invalid.

References

• Iyer, Vivek, and Robert M. Wald, “Some Properties of Noether Charge and a Proposal for Dynamical Black Hole Entropy,” Physical Review D 50 (1994) 846–864. DOI; Open PDF.
• Wald, Robert M., “Black Hole Entropy Is the Noether Charge,” Physical Review D 48 (1993) R3427–R3431. DOI; Open PDF.