Thermal States and Euclidean Black-Hole Saddles

At large N and strong coupling, a boundary thermal partition function is approximated by a sum over Euclidean bulk saddles with the same conformal boundary data. The leading phase is the saddle with smallest renormalized Euclidean action, provided its negative modes and integration contour are understood. A smooth black-hole cigar encodes the temperature through tip regularity; it does not turn a saddle approximation into an exact thermal trace.

Required background. Gauge Theory and String-Theory Preparation supplies large-N counting, operator/source couplings, and string-scale control used below.

Helpful background. Quantum Gravity frames the semiclassical expansion, Quantum Information distinguishes ensembles from pure states, Black-Hole Geometry and Mechanics supplies horizon thermodynamics, and Quantum Fields in Curved Spacetime clarifies Euclidean state preparation.

Euclidean smoothness fixes the ensemble temperature

For a nonextremal outer horizon in the static metric $ds^2=-f(r)dt^2+dr^2/f(r)+ds_H^2$, with the boundary time normalization fixed, continue $t=-i\tau$. Expand $f(r)=f'(r_h)(r-r_h)+\cdots$ and set $\rho=2\sqrt{(r-r_h)/f'(r_h)}$. The $(\tau,r)$ plane becomes

$$ds_E^2\simeq d\rho^2+\kappa^2\rho^2d\tau^2+ds_H^2, \qquad \kappa=\frac{f'(r_h)}{2}.$$

Absence of a conical singularity requires $\tau\sim\tau+\beta$ with

$$\beta=\frac{2\pi}{\kappa}, \qquad T=\beta^{-1}=\frac{\kappa}{2\pi}.$$

This is a boundary condition, not a freely adjustable period for a fixed smooth geometry. Fermions are antiperiodic around the contractible thermal circle. Gauge potentials must be regular at the tip, and angular identifications fix which grand-canonical potentials are held fixed.

The conformal boundary must also be matched before two actions are compared. A radial cutoff surface in the black-hole geometry and one in thermal AdS generally have different induced proper periods. Matching coordinate periods rather than proper boundary metrics leaves a finite error after subtraction. Holographic counterterms give the same physical difference while making clear which finite terms are scheme choices and which phase comparison is invariant.

The renormalized saddle sum

With the boundary metric, sources, spin structure, and ensemble fixed, the semiclassical expansion is

$$Z(\beta,J) \simeq \sum_s e^{-I_{E,\mathrm{ren}}[g_s,\Phi_s;J]}, \qquad F_s=\frac{I_{E,\mathrm{ren}}[s]}{\beta}.$$

The action includes the Einstein–Hilbert term, the Gibbons–Hawking–York term, matter boundary terms appropriate to the ensemble, and local counterterms. Comparing unrenormalized volumes is meaningless because the individual actions diverge. Hawking and Page performed the original saddle comparison with matched boundary geometries in Hawking and Page 1983, §§2–4, pp. 581–587; Witten supplied the large-N gauge-theory interpretation in Witten 1998, §§2–3, pp. 511–522.

Dominance is meaningful only among saddles allowed by the same boundary problem. Changing topology in the interior is permitted; changing the boundary spatial manifold, source, chemical potential, or spin structure is not. The exponential weight is parametrically sharp when the action scales like $L^{d-1}/G_{d+1}$, but one-loop determinants shift the crossing at subleading order and zero or negative modes can invalidate a naive Gaussian approximation.

For global AdS$_{d+1}$ Schwarzschild with $d\ge3$,

$$f(r)=1+\frac{r^2}{L^2}-\frac{\mu}{r^{d-2}}, \qquad T(r_h)=\frac{d r_h^2+(d-2)L^2}{4\pi L^2r_h}.$$

Below the minimum of $T(r_h)$ there is no smooth black-hole saddle with that boundary period. Above it there are small and large branches. Existence, local stability, and global dominance are therefore three different questions.

Thermodynamic derivatives check the action

In the canonical ensemble, a consistent on-shell action obeys

$$E=\frac{\partial I_E}{\partial\beta}, \qquad S=\beta E-I_E, \qquad C=T\frac{\partial S}{\partial T}.$$

For the spherical black hole, background subtraction or holographic renormalization gives the total free-energy difference,

$$F_{\mathrm{BH}}-F_{\mathrm{thermal\ AdS}} = \frac{\Omega_{d-1}r_h^{d-2}}{16\pi G_{d+1}} \left(1-\frac{r_h^2}{L^2}\right).$$

Here $\Omega_{d-1}$ is the area of the unit sphere. The sign changes at $r_h=L$. Independently differentiating the mass with respect to entropy reproduces $T(r_h)$; differentiating $F$ then reproduces the same energy and entropy. This agreement checks the boundary terms and normalization. The small branch has negative heat capacity even where it exists, while the large branch can be locally stable before it becomes globally preferred. The three-dimensional BTZ case is different: its nonrotating temperature increases monotonically with horizon radius, so there is no small/large turning point of this kind.

The boundary interpretation is correspondingly precise. On a compact sphere, the low-temperature phase has free energy of order the confined spectrum, whereas the large black hole carries entropy of order the large-N number of degrees of freedom. The order parameter is not literally a local field in every theory; Polyakov-loop behavior, entropy scaling, and center symmetry must be stated for the particular gauge theory. The gravitational saddle exchange nevertheless provides the leading thermodynamic distinction.

What saddle dominance does not establish

A subdominant saddle is not absent; it contributes with exponential suppression and may control metastability or late-time effects. A Euclidean saddle can also possess a negative mode, and a complex saddle requires a specified contour. For a finite-N theory on a compact spatial manifold with a discrete spectrum and a well-defined thermal trace, a single decaying black-hole correlator cannot reproduce arbitrarily late evolution. Finite N alone does not discretize an infinite-volume spectrum. The explicit free-energy crossing is developed next in the Hawking–Page worked example.

Exercises

Show that regularity fixes the period and explain why choosing a different period does not describe the same smooth black hole.

Solution

Near the horizon define a proper radial coordinate so that the Euclidean metric is $d\rho^2+\kappa^2\rho^2d\tau^2$. This is a plane in polar coordinates only when $\kappa\tau$ has period $2\pi$, hence $\beta=2\pi/\kappa$. A different period leaves a conical defect at $\rho=0$ and therefore changes the bulk configuration or introduces a singular source. As a check, the resulting temperature equals $f'(r_h)/(4\pi)$ and satisfies the first law for the renormalized saddle.

References

• Hawking, Stephen W., and Don N. Page. “Thermodynamics of Black Holes in Anti-de Sitter Space.” Communications in Mathematical Physics 87 (1983): 577–588. DOI.
• Witten, Edward. “Anti-de Sitter Space, Thermal Phase Transition, and Confinement in Gauge Theories.” Advances in Theoretical and Mathematical Physics 2 (1998): 505–532. DOI.