Extremal, Near-Extremal, and Late-Time Limits

An extremal horizon is not obtained safely by erasing $\kappa$ from every nonextremal formula. The horizon becomes degenerate, the throat becomes infinitely long, the exponential affine map changes character, and the state and late-time limits can fail to commute. The reliable procedure is to state the family of geometries, construct the state at finite nonextremality or directly at extremality, and only then take the declared time, frequency, and extremal limits.

Required background. Surface gravity and redshift fixes the generator and affine maps, while Boulware, Hartle–Hawking, and Unruh states distinguishes the boundary conditions whose limits are being compared.

Helpful background. Rotating and charged horizons supplies shifted energies near extremality, and multiple-horizon state obstructions illustrates why separate horizon periods need not define one global state.

Degenerate versus simple horizons

For Reissner–Nordström,

$$f(r)=\left(1-\frac{r_+}{r}\right) \left(1-\frac{r_-}{r}\right), \qquad r_\pm=M\pm\sqrt{M^2-Q^2},$$

and the outer-horizon surface gravity is

$$\kappa_+=\frac{r_+-r_-}{2r_+^2}.$$

At $|Q|<M$, $r_+$ is a simple zero and the tortoise coordinate has a logarithmic divergence,

$$r_*=r+\frac{r_+^2}{r_+-r_-}\log|r-r_+| -\frac{r_-^2}{r_+-r_-}\log|r-r_-|+C.$$

The regular outgoing Kruskal coordinate is proportional to $-e^{-\kappa_+u}$. At $|Q|=M$, however,

$$f(r)=\left(1-\frac{M}{r}\right)^2, \qquad r_*=r+2M\log\left|\frac{r}{M}-1\right| -\frac{M^2}{r-M}+C.$$

The pole term dominates the logarithm. An exponential coordinate with $\kappa_+=0$ is neither defined by substitution nor regularizing; a direct extremal affine map contains power-law and logarithmic structure determined by the collapse history.

The Euclidean argument changes for the same reason. A simple horizon is locally a polar origin and smoothness fixes $\beta=2\pi/\kappa_+$. The extremal horizon lies at infinite proper distance down the Euclidean throat, so eliminating a conical tip does not select a unique period. The statement $T\to0$ is a limit along nonextremal equilibrium states, not a temperature fixed by smoothness of the exact extremal Euclidean section.

The structure map therefore forks at its first box. “Near-extremal simple horizon” and “exact degenerate horizon” require distinct state constructions before any thermal or flux comparison.

Two routes through the extremal limit. The diagram is schematic and not to scale; taking $\kappa\to0$ after constructing a regular nonextremal state is not the same operation as quantizing directly on the degenerate geometry.

The failure map highlights the lost information: setting $\kappa=0$ early removes the coordinate transformation, scaled low-frequency sector, and transients whose combined limit may keep the stress tensor regular.

Order-of-limits failure for extremal horizon QFT. This schematic, not-to-scale map requires the geometry family, state, frequency scaling, and observation time to be fixed before a zero-temperature or zero-flux conclusion is licensed.

Application: take the state limit in two orders

Use a neutral massless scalar to avoid charged superradiance. For each $\kappa>0$, construct the horizon-regular equilibrium two-point function on the nonextremal bifurcate geometry. Along one outgoing chiral sector, differentiating the two-point function removes the scalar infrared constant and gives the local thermal form

$$C_\kappa(\Delta u) =-\frac{\kappa^2}{16\pi} \frac{1}{\sinh^2[\kappa(\Delta u-i\epsilon)/2]}.$$

At fixed exterior separation $\Delta u$,

$$\lim_{\kappa\to0}C_\kappa(\Delta u) =-\frac{1}{4\pi(\Delta u-i\epsilon)^2},$$

the zero-temperature chiral vacuum correlator. Equivalently, for every fixed positive Killing frequency,

$$n_\kappa(\omega)=\frac{1}{e^{2\pi\omega/\kappa}-1} \longrightarrow0.$$

That convergence is not uniform. If $y=\omega/\kappa$ is fixed, then

$$n_\kappa(\kappa y)=\frac{1}{e^{2\pi y}-1},$$

so modes whose Killing frequency scales with the shrinking temperature retain nontrivial occupation. Geometrically, those modes probe the growing throat. A pointwise exterior limit can therefore miss a finite scaled sector even though every fixed-$\omega$ occupation vanishes.

Now reverse the order. Set $Q=M$ first and quantize on the exact extremal collapse geometry. There is no bifurcation surface and no Euclidean condition selecting the preceding KMS family. One must choose initial data, derive the nonexponential ray map, and test the resulting two-point function and stress tensor directly. The state obtained this way need not be a KMS state and need not equal the global limit of the nonextremal Hartle–Hawking family. Liberati, Rothman, and Sonego’s collapse calculation finds nonthermal extremal particle production rather than a naive equilibrium limit (2000, §§ II–IV).

The comparison thus has four reported controls: fixed versus scaled frequency, fixed exterior point versus throat-scaled point, finite versus late retarded time, and state constructed before versus after extremality. Calling either route simply “the extremal state” suppresses the data that determine the answer.

Late-time transients can become leading

For nonextremal collapse, transient contributions often decay exponentially and the stationary Hawking term dominates. At extremality the affine map decays only algebraically, so the same separation is nonuniform. In a two-dimensional reduction of a family of regular charged collapses, Balbinot, Fabbri, Farese, and Parentani find that transients required for horizon regularity survive the extremal limiting procedure; dropping them before the limit produces a spurious singular static stress (2007, §§ 3–5). This is an analytic controlled-model result, not a theorem for every four-dimensional interacting field.

The converse overgeneralization is also unsafe. A modern four-dimensional mode-sum calculation gives numerical evidence that the Boulware-state scalar stress tensor is regular at an exact extremal Reissner–Nordström horizon for the masses and curvature couplings studied (Arrechea et al. 2025, abstract and §§ IV–V). That result is state-, field-, geometry-, and renormalization-specific. It neither turns the Boulware state into a collapse state nor proves uniformity of the near-extremal Hartle–Hawking or Unruh limits.

Exterior late-time tails introduce another order. The limits

$$\lim_{\kappa\to0}\lim_{u\to\infty} \quad\text{and}\quad \lim_{u\to\infty}\lim_{\kappa\to0}$$

need not agree because the decay time grows like $\kappa^{-1}$ and the throat supports increasingly low frequencies. A calculation should use finite-time wavepackets, then demonstrate uniform convergence before exchanging the limits.

Adversarial test: set κ to zero too early

Start from $U=-\kappa^{-1}e^{-\kappa u}$ and $n_\kappa(\omega)$. Substituting $\kappa=0$ makes $U$ indeterminate and sets $n$ to zero only for fixed $\omega>0$. It says nothing about the exact extremal affine coordinate, about $\omega/\kappa$ fixed, or about transients at $u\sim1/\kappa$. Those are precisely the sectors that become nonuniform.

The strongest conclusion that survives is conditional: the fixed-frequency thermal population of a specified nonextremal equilibrium family vanishes as $\kappa\to0$. A zero-flux claim for an exact extremal collapse state requires a direct ray map and stress-tensor calculation. A global regularity claim additionally requires control of the throat and both horizons.

Domain and failure conditions

See the chapter domain and failure-conditions table. This page distinguishes a near-extremal family from an exact degenerate background, and fixed from scaled time and frequency limits. It does not decide microscopic near-extremal thermodynamics, nonlinear stability, or evaporation endpoints. Evidence was checked through 10 August 2026: analytic collapse models and state-specific numerical stress tensors establish controlled examples, while a state-independent equivalence of extremal and near-extremal late-time limits remains unsupported.

Exercise

Show explicitly why the nonextremal Kruskal coordinate has no useful $\kappa\to0$ limit at fixed $u$, even after subtracting a constant.

Solution

Expand

$$U_\kappa=-\frac{1}{\kappa}e^{-\kappa u} =-\frac{1}{\kappa}+u-\frac12\kappa u^2+O(\kappa^2u^3).$$

Subtracting the divergent constant leaves $u$ at fixed retarded time, which is still singular at the extremal future horizon. Moreover, the expansion fails when $u\sim1/\kappa$, exactly the late-time scaling relevant to the long throat. A direct extremal regular coordinate must be derived from the extremal geometry and collapse map.

Handoff

The next conceptual step is the semiclassical backreaction chapter, where a renormalized state-dependent stress tensor and the geometry are evolved together rather than related by a quasistatic substitution.

References

• Arrechea, Julio, Cormac Breen, Adrian Ottewill, and Peter Taylor. “Renormalized Stress-Energy Tensor for Scalar Fields in the Boulware State with Applications to Extremal Black Holes.” Physical Review D 111 (2025): 085009. doi:10.1103/PhysRevD.111.085009.
• Balbinot, Roberto, Alessandro Fabbri, Sara Farese, and Renaud Parentani. “Hawking Radiation from Extremal and Non-Extremal Black Holes.” Physical Review D 76 (2007): 124010. doi:10.1103/PhysRevD.76.124010.
• Liberati, Stefano, Tony Rothman, and Sebastiano Sonego. “Nonthermal Nature of Incipient Extremal Black Holes.” Physical Review D 62 (2000): 024005. doi:10.1103/PhysRevD.62.024005.