---
title: Scale Echoing in Critical Collapse and Beyond
url: https://www.emergentmind.com/topics/scale-echoing
type: topic
---

# Scale Echoing in Critical Collapse and Beyond

Searching arXiv for the cited papers to ground the article in current paper records.
arXiv search: "1807.10342 Aspherical deformations of the Choptuik spacetime"
Scale echoing denotes recurrent structure under a change of scale, but the literature does not use the term uniformly. In critical gravitational collapse it refers primarily to discrete self-similarity, where fields become periodic in logarithmic time and the near-critical solution repeats with a fixed echoing period \(\Delta\). In related gravitational-wave and braneworld settings it denotes delayed pulses generated by repeated reflections in an effective cavity, so that echo times encode an internal geometric length scale. In recent large-language-model work, “echoing” has been extended to prompt restatement and to identity mirroring in agent-agent dialogue, where recurrence is semantic rather than geometric. The shared motif is recurrence with a characteristic scale, but the underlying mechanisms differ substantially [1807.10342] [2112.04065] [2508.20937] [2602.06600] [2511.09710] [2602.10185].

## 1. Terminological scope

| Domain | Manifestation | Characteristic scale |
|---|---|---|
| Critical collapse | Periodicity in logarithmic time and fine structure in scaling laws | \(\Delta\), \(\gamma\) |
| Thick-brane perturbations | Delayed pulses from repeated reflections in a double-barrier potential | \(\Delta t_{\rm echo}\), QNM spectrum |
| LLM reasoning and AxA | Prompt restatement or role mirroring | \(\Delta \mathcal{L}\), echoing rate, onset turn |

In the Choptuik problem and its generalizations, scale echoing is tied to a discretely self-similar critical solution at the threshold of black-hole formation. The relevant scale is logarithmic: one introduces \(T=-\ln(\tau_*-\tau)\), and the critical solution becomes periodic in \(T\) with period \(\Delta\). This periodicity induces the oscillatory fine structure observed in mass or curvature scaling laws and is therefore inseparable from the critical exponent \(\gamma\).

A distinct usage arises in wave propagation on backgrounds with internal structure. In a five-dimensional thick brane, scalar perturbations encounter an effective double-barrier potential when the brane splits into two sub-branes. Echoes then appear as repeated delayed pulses produced by successive reflections between the barriers, and the observable scale is the round-trip travel time inside the cavity rather than a logarithmic self-similarity period [2508.20937].

Recent LLM work employs “echoing” in a third sense. “Echo of Prompt” denotes the spontaneous front-loaded restatement of the task at the head of an internal reasoning trace, while multi-agent “echoing” denotes identity failure in which one agent mirrors the role of another. These usages preserve the idea of structured repetition, but they do not invoke DSS or wave trapping [2602.06600] [2511.09710].

## 2. Discrete self-similarity in critical collapse

For massless scalar collapse, one considers a one-parameter family of initial data labeled by a strength parameter such as \(\eta\). Small \(\eta\) yields dispersion to Minkowski spacetime, large \(\eta\) yields black-hole formation, and the threshold \(\eta_*\) defines the critical solution. The characteristic scale shrinks according to
\[
R_{\rm sc} \simeq \tau_*-\tau,
\]
and the logarithmic time coordinate
\[
T=-\ln(\tau_*-\tau)
\]
maps the accumulation event \(\tau\to\tau_*\) to \(T\to\infty\). Near criticality, the black-hole mass obeys
\[
M \simeq |\eta-\eta_*|^\gamma,
\]
while for subcritical data the maximum central energy density satisfies
\[
\rho_c^{\rm max}\simeq |\eta_*-\eta|^{-2\gamma}.
\]
Because the critical solution is discretely self-similar rather than continuously self-similar, these power laws acquire a periodic correction,
\[
\ln \rho_c^{\rm max}= C-2\gamma \ln|\eta_*-\eta| + f(\ln|\eta_*-\eta|),
\]
with angular frequency
\[
\omega=\frac{\Delta}{2\gamma}.
\]
In practice the periodic term is often fitted as a sinusoid, which permits independent extraction of \(\gamma\) and \(\Delta\) from subcritical data [1807.10342].

Discrete self-similarity means that fields repeat after a finite shift in \(T\), up to a rescaling of spatial coordinates. In the Choptuik spacetime this implies
\[
R_{\rm sc}(T+\Delta)=e^{-\Delta}R_{\rm sc}(T),
\]
so each echo occurs at a length scale smaller by a factor \(e^\Delta\). In spherical symmetry the canonical values for a massless scalar field are
\[
\gamma \simeq 0.374, \qquad \Delta = 3.445452402(3),
\]
and numerical evolutions reproduce these values through scaling-law fits, zero-crossing methods based on the central scalar field, and direct overlap of rescaled radial profiles. The central field \(\phi_c(T)\) is approximately periodic during the intermediate phase in which the solution tracks the Choptuik attractor, and radial profiles at every other zero crossing overlap when the radius is rescaled by \(e^{N\Delta}\), providing a direct operational definition of echoing [1807.10342].

## 3. Robustness and deformation of echoing beyond the spherical GR limit

Once exact spherical symmetry is relaxed, the main question is whether the DSS pattern persists unchanged or whether the effective \(\Delta\) and \(\gamma\) drift. In axisymmetry, initial data of the form
\[
\phi = \eta \exp\left( - (r / r_0)^2 \left[\sin^2 \theta + (1 - \epsilon^2) \cos^2 \theta \right] \right)
\]
excite even \(\ell\) modes, especially \(\ell=2\). For \(\epsilon^2=0.01\), the measured critical exponent and echoing period agree with the spherical values within numerical error, and aspherical deviations behave as damped oscillations in logarithmic time,
\[
u(T)\simeq e^{\kappa T}\cos(\omega_p T+\phi_{\rm ph})\, g(T),
\]
with \(\kappa_{\rm pert}=-0.07/\Delta\), consistent with the even \(\ell=2\) mode of linear perturbation theory. For larger deformations, however, the effective parameters drift: \(\epsilon^2=0.5\) gives \(\gamma\approx0.369\) and \(\Delta\approx3.37\)–\(3.39\), while \(\epsilon^2=0.75\) gives \(\gamma\approx0.306\) and \(\Delta\approx2.8\)–\(2.9\), together with growth of the aspherical mode and evidence for bifurcation into two collapsing centers on the symmetry axis. In that regime, central echoing ceases to be an adequate global descriptor of the dynamics [1807.10342].

A different deformation is non-minimal coupling. With
\[
F(\phi)=1+8\pi \xi G_0 \phi^2, \qquad G_{\text{eff}}=\frac{G_0}{F(\phi)},
\]
small couplings \(\xi=10^{-3},10^{-2},10^{-1}\) leave the critical behavior essentially unchanged: \(\gamma\) remains near \(0.374\) and \(\Delta\) near \(3.44\). For strong coupling, standard \(1+\log\) slicing develops gauge pathologies, and the simulations require the shock-avoiding Bona-Massó slicing
\[
F_{BM}(\alpha)=1+\frac{1}{\alpha^2}.
\]
In that regime the exponents decrease: for \(\xi=1\), \(\gamma=0.368\pm0.001\) and \(\Delta=3.386\pm0.017\) from scaling; for \(\xi=10\), \(\gamma=0.365\pm0.006\) and \(\Delta=3.109\pm0.007\). The residual modulation in \(\ln R_{\max}\) is no longer well described by a single sine and must be fitted with at least two harmonics,
\[
\ln R_{\max}= C - 2\gamma\ln|\phi_0^*-\phi_0|
+ a_0\sin(\omega x+\varphi_0)
+ a_1\sin(2\omega x+\varphi_1),
\]
which is why the strong-coupling echoing is described as richer than a single harmonic [2112.04065].

Einstein-Gauss-Bonnet gravity deforms the problem more radically because the Gauss-Bonnet coupling introduces a microscopic length scale and breaks the scale invariance that supports DSS. Near criticality, the scalar field at the origin no longer oscillates with ever-decreasing period; instead it approaches a constant physical period
\[
T_{(n)} \propto \tilde{\alpha}^{\beta_{(n)}},
\]
with \(\beta_{(5)}=0.34\pm0.05\) and \(\beta_{(6)}=0.24\pm0.08\). In five dimensions the horizon-radius scaling suggests a radius gap and the curvature scaling suggests a finite maximum \(T^*\), whereas in six dimensions no radius gap is evident and the system instead crosses into a new GB-dominated scaling regime with modified exponents below the GB scale [1208.5250].

## 4. Continuous dimensions and critical spacetime crystals

A recent reformulation treats the critical DSS solution itself as a one-parameter family indexed by continuous spacetime dimension \(D>3\). In this construction, the physical metric is written as
\[
g_{\mu\nu}(\tau,x^i)=e^{-2\tau}\,\tilde g_{\mu\nu}(\tau,x^i),\qquad
\tilde g_{\mu\nu}(\tau+\Delta,x^i)=\tilde g_{\mu\nu}(\tau,x^i),
\]
so the discrete scale symmetry becomes periodicity in \(\tau\). The resulting solutions are called critical spacetime crystals. Their fundamental domain lies between the center \(x=0\) and the self-similar horizon \(x=1\), where the crystal vector \(\partial_\tau\) becomes null. The unstable perturbation spectrum is Floquet-like,
\[
\delta Z(\tau,x)\sim e^{\lambda\tau}\,\delta Z_{\rm periodic}(\tau,x),
\qquad
\gamma(D)=\frac{1}{\lambda(D)},
\]
and the black-hole mass scales as
\[
M_{\rm BH}=M_0\,(p-p_*)^{(D-3)\gamma}.
\]
The mass fine structure remains periodic, with period \(\Delta/(2\gamma)\) in \(\ln(p-p_*)\) [2602.10185].

The main result is that \(\Delta(D)\) and \(\gamma(D)\) become smooth functions of \(D\). The echoing period has a maximum at
\[
D_{\rm crit}=3.755726\pm10^{-6},\qquad
\Delta(D_{\rm crit})=3.466772\pm10^{-6}.
\]
The same computation recovers
\[
D=4:\ \Delta=3.445453,\ \gamma=0.373961,
\qquad
D=5:\ \Delta=3.22176,\ \gamma=0.41322.
\]
Analytical expansions in \(1/D\) and \(D-3\) support the conjectures
\[
\lim_{D\to 3^+}\Delta=0^+,\qquad \lim_{D\to 3^+}\gamma=0^+,
\qquad \lim_{D\to\infty}\gamma=\frac12,
\]
while also suggesting that \(\Delta(D)\to0\) as \(D\to\infty\). This recasts scale echoing as a dimension-dependent order parameter of the critical solution rather than as a fixed four-dimensional constant [2602.10185].

## 5. Delayed echoes from internal geometry

In braneworld perturbation theory, “echoing” refers not to DSS but to repeated delayed signals in the time-domain response. For scalar perturbations of a five-dimensional thick brane, the KK-reduced master field satisfies a Schrödinger-like equation
\[
-\psi''(z)+U(z)\psi(z)=m^2\psi(z),\qquad m^2=\omega^2-a^2.
\]
When the brane splits into two sub-branes, the effective potential becomes a double barrier with a cavity between the peaks. Wave packets entering the cavity undergo repeated partial reflections, and each leakage pulse appears as an echo. The echo separation is approximately
\[
\Delta t_{\rm echo}\approx 2\int_{z_-}^{z_+}dz,
\]
namely twice the light-travel time between the effective barriers in the conformal coordinate \(z\) [2508.20937].

The internal structure is controlled by the parameters \(s\) and \(\delta\). For \(s=1\), the energy density has a single peak and the potential is a single barrier, so no cavity-induced echoes occur. For \(s>1\), the energy density splits into two peaks, and for \(s>1,\delta>1\) the double barrier becomes pronounced enough to produce clear echo signals. The observed scaling is geometric: increasing sub-brane separation increases the effective cavity size \(L_{\rm eff}\), which in turn increases \(\Delta t_{\rm echo}\), decreases the spacing of cavity frequencies, and makes the imaginary parts of the quasinormal frequencies very small. The waveform also depends strongly on observer position in the extra dimension: observers located on a sub-brane detect clean periodic signals, whereas observers between sub-branes observe more complex, modulated waveforms [2508.20937].

This usage is sometimes conflated with critical-collapse echoing, but the mechanisms are different. DSS echoing is periodicity in logarithmic scale near a critical attractor, whereas braneworld echoes arise from cavity leakage in a linear wave problem. What they share is that an internal length scale becomes measurable through a repeated signal.

## 6. Echoing in LLM reasoning and multi-agent systems

In reasoning models, a front-loaded semantic restatement of the task is termed the Echo of Prompt. Formally, if \(x\) is the prompt and \(y\) a full generated trace, one partitions traces into an echo-free subset \(\mathcal{Y}_{\rm trim}\) and an echo-containing subset, and defines a trimmed distribution by rejection-style conditioning,
\[
T_\theta^{\rm trim}(y\mid x)=
\frac{T_\theta(y\mid x)\,\mathbf{1}_{y\in\mathcal{Y}_{\rm trim}}}{Z_x}.
\]
The associated Echo Likelihood Gap is
\[
\Delta\mathcal{L}=\mathcal{C}(y_{\rm raw})-\mathcal{C}(y_{\rm trim}),
\]
with \(\mathcal{C}(y)\) the per-token average log-likelihood. On GSM8K with DeepSeek-R1-Distill-Llama-8B, large EOPs have mean echo length \(\approx 219\) tokens, \(\Delta\mathcal{L}\) is positive across echo-length bins, and logistic regression finds \(\beta_1\approx0.24\) with odds ratio \(\exp(0.24)\approx1.27\). Layer-wise attention analysis shows that correct traces exhibit stronger answer-to-answer-prefix attention in middle layers, motivating Echo-Distilled SFT and Echoic Prompting as ways to exploit prompt echoing as a compute-shaping mechanism [2602.06600].

In agent-agent systems, echoing denotes a different pathology: agents abandon their assigned identities and mirror their conversational partners. This is formalized in a two-agent partially observable stochastic game with agents \(A_i=(I_i,O_i,T_i,U_i,\pi_i)\), and echoing is detected by an evaluator \(\text{EchoEvalLM}(H_T,I_i,I_j)=\{\sigma,a_e,m_e\}\). Across \(60\) AxA configurations, \(3\) domains, and \(2000+\) conversations, echoing rates range from \(5\%\) to \(70\%\) depending on model and domain. The failure scales with interaction length: the average onset is \(7.6\) turns, the median onset is \(8.0\), and conversations with echoing average \(9.6\) turns versus \(8.7\) without echoing. Increased reasoning effort does not materially reduce the problem, but a protocol-level mitigation that requires each response to include an explicit `role` field reduces echoing to about \(9\%\) in the evaluated models [2511.09710].

## 7. Comparative interpretation

Across these literatures, scale echoing is best understood as a family of recurrence phenomena rather than a single invariant concept. In critical collapse, the essential structure is a universal DSS attractor with logarithmic periodicity, an echoing period \(\Delta\), and a scaling exponent \(\gamma\). In modified theories or deformed initial data, the central issue is whether those parameters remain universal or become effective, theory-dependent quantities; the answer depends on the deformation, with small asphericity leaving \(\Delta\) and \(\gamma\) unchanged, strong non-minimal coupling introducing multi-harmonic modulation, and Gauss-Bonnet terms breaking scale invariance altogether [1807.10342] [2112.04065] [1208.5250].

A common misconception is to treat every “echo” as evidence of DSS. The thick-brane problem shows that delayed pulses can arise from a double-barrier cavity with no critical scaling at all, while LLM echoing can describe either a beneficial semantic restatement or a failure of role stability. The unifying idea is that repeated structure exposes a hidden scale: the accumulation scale of a critical spacetime, the cavity size of a split brane, the probability geometry of a reasoning trace, or the interaction horizon of an agent-agent dialogue [2508.20937] [2602.06600] [2511.09710].

Within gravitational collapse itself, the recent continuous-\(D\) program sharpens that perspective by turning \(\Delta\) and \(\gamma\) into continuous functions of dimension and by showing that the four-dimensional Choptuik solution is one point on a larger manifold of critical spacetime crystals. That result suggests that scale echoing is neither an accidental numerical artifact nor a fixed constant of nature, but a sharply defined property of a universality class whose value depends on the theory, the matter content, and, in a continuous-dimensional sense, the ambient spacetime dimension [2602.10185].

Source: https://www.emergentmind.com/topics/scale-echoing