Two-Dimensional Backreaction Models

Two-dimensional dilaton gravity turns a difficult semiclassical initial-value problem into a controlled laboratory: the dilaton records an effective area, conformal matter supplies a calculable anomaly, and null constraints tie the outgoing flux to the evolving geometry. The laboratory is not a dimensional reduction theorem. It shows how backreaction, Bondi balance, and horizon motion can coexist, but its constant late-time flux, absence of angular barriers, and endpoint behavior do not automatically survive in four dimensions.

Helpful background. Quantum Fields in Curved Spacetime supplies anomaly-induced stress tensors and state choice; Einstein Equations supplies constraint evolution; Black-Hole Geometry and Mechanics supplies event and apparent horizons; Semiclassical Gravity supplies renormalized backreaction; and Black-Hole Dynamics supplies Bondi mass and trapping-horizon balance.

Dilaton gravity and its semiclassical data

In one common CGHS normalization, let $\lambda\gt0$ be an inverse-length scale. The metric $g_{ab}$, dimensionless dilaton $\phi$, and $N$ massless scalars $f_i$ have action

$$S=\frac{1}{2\pi}\int d^2x\sqrt{-g}\left[ e^{-2\phi}\!\left(R+4(\nabla\phi)^2+4\lambda^2\right) -\frac12\sum_{i=1}^{N}(\nabla f_i)^2\right].$$

The field $e^{-\phi}$ can be compared with a transverse radius, but it is a dynamical field of this two-dimensional theory. In conformal gauge, $g_{+-}=-e^{2\rho}/2$, the metric equations include two null constraints, while the matter fields obey $\partial_+\partial_-f_i=0$. Collapse can be specified by an incoming pulse on the linear-dilaton vacuum. The state must still be fixed: the integration functions $t_\pm(x^\pm)$ in the anomaly-induced stress tensor encode boundary conditions such as no incoming quantum flux. These ingredients, including the classical collapse solution, fixed-background flux, and large-$N$ effective action, are developed explicitly in Callan et al. 1992, Eqs. (1)–(23), PDF.

For a single conformal scalar, their asymptotic outgoing stress component $\langle T^f_{--}\rangle$ approaches $\lambda^2/48$ in their asymptotically inertial retarded coordinate $\sigma^-$. This is the CGHS stress normalization: for the matter action above its $T^f_{--}$ is $\pi$ times the usual stress defined by metric variation. Thus the physical flux is $N\lambda^2/(48\pi)=N\pi T^2/12$ at $T=\lambda/(2\pi)$. $N$ identical fields multiply the CGHS component by $N$ Callan et al. 1992, Eq. (19), PDF. Extending that fixed-background constant forever would radiate more than the initial mass. The contradiction is the signal to solve the anomaly-corrected geometry and flux together, not an endpoint prediction.

What the CGHS and RST analyses establish

At leading large $N$, the matter anomaly adds an order-$N$ Liouville term in conformal gauge. This makes a semiclassical backreaction regime accessible while $e^\phi$ remains small, but it does not remove the constraints or state functions. The original CGHS paper proposed a qualitative continuation into a Liouville-dominated region and explicitly described that continuation as speculative rather than a solved evolution Callan et al. 1992, Eqs. (20)–(29) and concluding discussion, PDF.

Russo, Susskind, and Thorlacius then analyzed the same one-loop, anomaly-corrected system along the collapse trajectory. Their calculation finds a curvature singularity where $Ne^{2\phi}/12=1$, a receding apparent horizon for a large hole, and leading energy-loss rate $N\lambda^2/48$. The approximation is controlled for $M\gg N\lambda$ but loses validity when the apparent horizon approaches the singularity and higher corrections matter Russo, Susskind, and Thorlacius 1992, §§III–IV, especially Eqs. (11)–(21), PDF. That paper says the general coupled solution is not known in closed form; it must not be confused with a later model carrying the same authors’ initials.

In End Point of Hawking Radiation, the same authors deliberately modify the one-loop action by a local covariant $\phi R$ term accompanying the nonlocal anomaly action. New fields $\Omega$ and $\chi$ reduce the semiclassical equations to an exactly soluble Liouville form. For super-threshold incoming flux, the apparent horizon recedes until it meets the singularity; future evolution beyond the emerging naked singularity is not supplied by the bulk equations. Matching to the linear-dilaton vacuum or imposing reflection at a timelike critical line is additional boundary data, not a derived uniqueness theorem Russo, Susskind, and Thorlacius 1992, §1, Eqs. (1.7)–(1.9); §2; and §4, Eqs. (4.4)–(4.9), PDF.

The modification can be stated without leaving conformal gauge. Define $\kappa_{\mathrm{2D}}=N/12$, an anomaly coefficient unrelated to black-hole surface gravity. The conformal-gauge form of their Eq. (1.7) is

$$S_{\mathrm{RST}} =S_{\mathrm{CGHS}} -\frac{\kappa_{\mathrm{2D}}}{\pi} \int dx^+dx^-\, \left(\partial_+\rho\,\partial_-\rho +\phi\,\partial_+\partial_-\rho\right).$$

The first added term is the matter anomaly; the second is the local RST modification. With the curvature convention of this page, $R=8e^{-2\rho}\partial_+\partial_-\rho$, the latter is $-\kappa_{\mathrm{2D}}(4\pi)^{-1}\int\sqrt{-g}\,\phi R\,d^2x$. It changes the dynamical theory rather than merely relabeling a solution of the original anomaly-corrected CGHS equations.

Comparison axis	Original CGHS analysis	hep-th/9201074 reassessment	Exactly solvable hep-th/9206070 model
Dynamical system	anomaly-corrected dilaton gravity at large $N$	the same coupled equations	a modified action with a local $\phi R$ term
Controlled claim	weak-coupling large-$N$ regime	large-mass recession before breakdown	exact semiclassical bulk solution of the modified model
Horizon result	qualitative evaporation scenario	receding apparent horizon	horizon meets the singularity for super-threshold flux
Endpoint status	conjectural continuation	approximation fails near the singular curve	continuation depends on boundary conditions at the exposed singularity

The robust content is methodological: evolve constraints and flux together, identify the state and renormalization convention, and stop at approximation breakdown. The model-specific content includes the mass-independent asymptotic flux, the anomaly coefficient, the dilaton definition of trapped regions, and the detailed singularity curve. Four-dimensional partial waves introduce greybody barriers, while the transverse area and tensor constraints carry information absent from a two-dimensional conformal model.

A flux-balance surrogate, not a CGHS solver

The retained computation isolates one necessary diagnostic. In dimensionless retarded time $u$, it evolves

$$\frac{dM}{du}=-F, \qquad \tau\frac{dF}{du}=F_\infty-F, \qquad \frac{dE_{\mathrm{rad}}}{du}=F,$$

with inputs

$$M(0)=2, \qquad F(0)=E_{\mathrm{rad}}(0)=0, \qquad F_\infty=0.08, \qquad \tau=0.7, \qquad u_f=3.$$

This exponential switch-on is a flux-balance surrogate. It contains no dilaton, metric constraint, anomaly coefficient, horizon, singularity, or greybody factor, so numerical agreement cannot validate either physical paper. Classical RK4 uses 6, 12, 24, and 48 uniform null-time steps. The independent benchmark is the exact solution

$$\begin{aligned} F(u)&=F_\infty\left(1-e^{-u/\tau}\right),\\ E_{\mathrm{rad}}(u)&=F_\infty\left[u-\tau \left(1-e^{-u/\tau}\right)\right],\\ M(u)&=M(0)-E_{\mathrm{rad}}(u). \end{aligned}$$

Environment, rerun, and convergence

The reference environment was CPython 3.9.6, NumPy 2.0.2, SciPy 1.13.1, and IEEE-754 binary64 arithmetic. The fixed-step method evaluates only the displayed ODEs, so interpolation is not applicable. Download the MIT-licensed benchmark, then use the standalone download-and-run recipe:

curl -fL https://blackhole.org/code/semiclassical-gravity/semiclassical-gravity-benchmarks.py -o semiclassical-gravity-benchmarks.py
python3 semiclassical-gravity-benchmarks.py --case two-dimensional-backreaction-models

“State error” is the Euclidean endpoint error in $(M,F,E_{\mathrm{rad}})$ relative to the exact solution. The balance residual is $\lvert M+E_{\mathrm{rad}}-M(0)\rvert$.

null-time resolution $N_u$	$\Delta u$	surrogate mass $M_f$	flux result $F_f$	state error	balance residual
6	5.0000000e-01	1.8152161e+00	7.8880111e-02	2.6434954e-05	2.2204460e-16
12	2.5000000e-01	1.8152286e+00	7.8898034e-02	1.2142816e-06	0.0000000e+00
24	1.2500000e-01	1.8152292e+00	7.8898851e-02	6.5316415e-08	4.4408921e-16
48	6.2500000e-02	1.8152292e+00	7.8898894e-02	3.7885989e-09	8.8817842e-16

The state error decreases rapidly under every refinement and approaches fourth-order behavior. Balance is preserved at binary64 roundoff because the same RK4 stages update $M$ and $E_{\mathrm{rad}}$ with opposite derivatives. That conservation cross-check is logically separate from comparison with the analytic exponential.

Failure modes of a physical calculation

A genuine two-dimensional rerun would evolve $\rho$, $\phi$, and matter data while enforcing both null constraints, fixing $t_\pm$ from a declared quantum state, and refining both null directions. It should monitor constraint norms, Bondi loss versus radiation at infinity, horizon location, curvature, and $Ne^{2\phi}$. Mixed null-coordinate conventions can change reported flux components; an incorrect anomaly coefficient changes the physical rate. Constraint-violating initial data, unresolved pulses, boundary reflections, stiff approach to the singular curve, and extrapolation beyond $M\sim N\lambda$ are rejection conditions. Even a converged two-dimensional endpoint is evidence about that matter model and counterterm choice, not a four-dimensional result.

Exercise

Derive the exact surrogate solution and explain why the balance residual can be near roundoff even when the state error is nonzero. Estimate the observed order from the 12-step and 24-step state errors.

Solution

Solving the linear flux equation gives $F=F_\infty(1-e^{-u/\tau})$. Integration yields the displayed $E_{\mathrm{rad}}$, and $d(M+E_{\mathrm{rad}})/du=0$ gives $M=M(0)-E_{\mathrm{rad}}$. RK4 applies identical weighted stage fluxes with opposite signs to $M$ and $E_{\mathrm{rad}}$, so their discrete sum is conserved apart from floating-point rounding even though both variables differ from their exact values.

For a step-size ratio of two, the observed order is

$$p=\frac{\log(1.2142816\times10^{-6}/6.5316415\times10^{-8})}{\log 2} \approx4.22.$$

The result is consistent with fourth-order RK4; the slight excess reflects the pre-asymptotic mixture of error components.

References

• Callan, C. G. Jr., S. B. Giddings, J. A. Harvey, and A. Strominger (1992), “Evanescent Black Holes,” Physical Review D 45, R1005–R1009. DOI. Open PDF.
• Russo, J. G., L. Susskind, and L. Thorlacius (1992), “Black Hole Evaporation in 1+1 Dimensions,” Physics Letters B 292, 13–18. DOI. Open PDF.
• Russo, J. G., L. Susskind, and L. Thorlacius (1992), “End Point of Hawking Radiation,” Physical Review D 46, 3444–3449. DOI. Open PDF.