Prove the BKL locality conjecture in the general inhomogeneous context

Prove the BKL locality conjecture in the general inhomogeneous context

Definitions and mathematical statement

Near a spacelike singularity, a Kasner epoch has a schematic local metric \[ ds^2\simeq-dt^2+\sum_{i=1}^3t^{2p_i}(\omega^i)^2,\qquad \sum_i p_i=\sum_i p_i^2=1. \] Vacuum Bianchi IX (Mixmaster) behavior consists of successive anisotropic epochs and transitions in a spatially homogeneous model. BKL locality predicts that, for generic inhomogeneous collapse, time evolution asymptotically dominates coupling between distinct spatial points, with corresponding oscillatory local behavior. The catalogue does not specify a data topology or measure defining "generic," a gauge, or a convergence mode. Consequently this is an asymptotic geometric program, not a fully quantified theorem determined by the single sentence alone.

Formulation and concept references: [S1].

Short English statement

Near a generic gravitational singularity, does each spatial point evolve almost independently through oscillatory Mixmaster-like dynamics?

Research attempt

Current frontier and formulation audit.

Section 3.1 of the supplied 2025 review discusses oscillatory vacuum models, velocity-dominated special cases, numerical evidence, and spatial spikes. It does not supply a topology defining generic initial data or a norm in which arbitrary inhomogeneous solutions shadow Bianchi IX evolution. The vacuum and matter settings must also be distinguished. Thus the original editorial warning is substantive, not a missing notational convention. No choice of a convenient meaning of "generic" below is substituted for the catalogue's target. [S1]

Ringström proves an attractor theorem for spatially homogeneous Bianchi IX solutions; this does not estimate spatial derivatives in an inhomogeneous Einstein evolution. Andersson--Rendall construct analytic, nonsymmetric, quiescent singularities with scalar-field or stiff-fluid matter. Those are important warnings about matter dependence, not counterexamples to a properly specified generic vacuum assertion. [E1][E2]

Attack route.

Try to separate three claims often compressed into the word locality: shrinking causal neighborhoods, small spatial-derivative terms in the field equations, and shadowing by an autonomous homogeneous trajectory. The first admits an explicit integrated criterion; the second requires derivative estimates that the criterion cannot provide; the third is sensitive to instability near exceptional epochs. The calculations below deliberately stress-test all three implications. They are auxiliary statements about specified models, not a formulation of the entire BKL conjecture.

Attempt: an integrated horizon criterion.

Consider the diagonal, spatially homogeneous kinematic metric \begin{equation}\label{wsref-0-33-30-30-30-36-37-33-36-3a-6d-65-74-72-69-63} ds^2=-dt^2+\sum_{i=1}^3 a_i(t)^2(dx^i)^2, \qquad 0<t\le1, \end{equation} and put $\tau=-\log t$ and \[ a_i(e^{-\tau})=\exp\!\left(-\int_0^\tau p_i(s)\,ds\right), \qquad D_i(s)=\int_0^s(1-p_i(r))\,dr. \] We do not yet impose the vacuum evolution equations. A causal curve, parametrized by $t$, satisfies $\sum_i a_i(t)^2(dx^i/dt)^2\le1$. Consequently its displacement in direction $i$, between two times tending to zero and a time $e^{-\tau}$, is bounded by \begin{equation}\label{wsref-0-33-30-30-30-36-37-33-36-3a-68-6f-72-69-7a-6f-6e} d_i(\tau):=\int_0^{e^{-\tau}}\frac{dt}{a_i(t)} =\int_\tau^\infty e^{-D_i(s)}\,ds. \end{equation} Thus finiteness of all three integrals gives shrinking coordinate causal neighborhoods, with coordinate diameter at most $2\sum_i d_i(\tau)$. The shrinking follows from convergence of the improper integrals, not from an interchange of an uncontrolled limit with integration.

For a fixed Kasner exponent $p_i<1$, this is the familiar exact expression \[ d_i(\tau)=\frac{e^{-(1-p_i)\tau}}{1-p_i}. \] A uniform bound $1-p_i(s)\ge\delta>0$ is sufficient but unnecessarily strong. The following weaker integrated estimate allows arbitrarily close approaches to a Taub exponent $p_i=1$.

Lemma (logarithmic Taub-excursion budget). If, for all sufficiently large $s$, \begin{equation}\label{wsref-0-33-30-30-30-36-37-33-36-3a-62-75-64-67-65-74} D_i(s)\ge (1+\delta)\log(s+2)-C, \qquad \delta>0, \end{equation} then for all sufficiently large $\tau$, \begin{equation}\label{wsref-0-33-30-30-30-36-37-33-36-3a-74-61-69-6c} d_i(\tau)\le\frac{e^C}{\delta}(\tau+2)^{-\delta}. \end{equation} Indeed, insert (30006736:budget) into (30006736:horizon) and integrate $(s+2)^{-1-\delta}$. This proves the lemma, including its constants. A bound only of the form $D_i(s)\ge\log(s+2)-C$ does not guarantee convergence: the comparison integral can be harmonic. $\square$

This calculation isolates a potentially useful quantity: accumulated time away from a Taub direction, not a pointwise lower bound on every epoch. For actual rotating, inhomogeneous frames, one would need corresponding bounds on the full frame coefficients along causal curves. Equation (30006736:horizon) alone does not justify that extension.

An exact test that defeats a pointwise shortcut.

Let $\epsilon(\tau)=1/(\tau+4)$ and define \begin{equation}\label{wsref-0-33-30-30-30-36-37-33-36-3a-65-78-70-6f-6e-65-6e-74-73} p_1=1-\epsilon,\qquad p_2=\frac{\epsilon+\sqrt{4\epsilon-3\epsilon^2}}2,\qquad p_3=\frac{\epsilon-\sqrt{4\epsilon-3\epsilon^2}}2. \end{equation} At every time, \[ \sum_i p_i=1,\qquad \sum_i p_i^2=1,\qquad p_i<1. \] Nevertheless, \[ D_1(\tau)=\log\frac{\tau+4}{4},\qquad d_1(\tau)=\int_\tau^\infty\frac{4\,ds}{s+4}=\infty. \] Thus instantaneous membership in the non-Taub part of the Kasner circle does not imply a shrinking horizon. The amount of time spent increasingly near its Taub point matters.

Crucially, this is not a vacuum counterexample. The Einstein equations can be checked rather than presumed. For (30006736:metric), write $H_i=\dot a_i/a_i=p_i(\tau)/t$. Then \[ \dot H_i=\frac{-p_i'(\tau)-p_i(\tau)}{t^2}, \quad R_{00}=-\sum_i(\dot H_i+H_i^2)=0, \quad \frac{R_{ii}}{a_i^2}=\dot H_i+H_i\sum_jH_j =-\frac{p_i'(\tau)}{t^2}. \] Since the $p_i$ in (30006736:exponents) are not constant, the spatial vacuum equations fail. Even though the dimensionless residuals $t^2R_{ii}/a_i^2=-p_i'$ tend to zero, approximate satisfaction is not exact satisfaction. The example invalidates a proposed kinematic inference, not the conjecture about genuine Einstein solutions.

Why a shrinking coefficient is not a PDE estimate.

In a schematic normalized evolution \begin{equation}\label{wsref-0-33-30-30-30-36-37-33-36-3a-73-63-68-65-6d-61-74-69-63} \partial_\tau U=F(U)+\sum_a E_a^{\ i}(\tau,x)\partial_iU+\mathcal R, \end{equation} one cannot discard the derivative term merely because $E_a^{\ i}\to0$. For instance, for any $\delta>0$, \[ E(\tau)=e^{-\delta\tau},\qquad U(\tau,x)=\sin(e^{\delta\tau}x) \quad\Longrightarrow\quad E\partial_xU=\cos(e^{\delta\tau}x). \] The coefficient vanishes while the weighted derivative has supremum norm one. This is another elementary test function, not an Einstein solution. It explains the exact additional estimate a proof needs, for example $\|E_a^{\ i}\partial_iU\|\to0$ in a specified norm on a specified set of spatial points. Spikes are especially relevant when assessing which such norm and exceptional set could be appropriate. [S1]

A shadowing stress test.

Even small or integrable remainders in (30006736:schematic) need not give tracking by the homogeneous solution with the same initial state. On an interval on which $F$ has Lipschitz bound $L(\tau)$, subtraction from $z'=F(z)$ and Gronwall give only \begin{equation}\label{wsref-0-33-30-30-30-36-37-33-36-3a-67-72-6f-6e-77-61-6c-6c} |U(\tau)-z(\tau)|\le e^{\int_{\tau_0}^\tau L(r)dr}|U(\tau_0)-z(\tau_0)| +\int_{\tau_0}^\tau e^{\int_s^\tau L(r)dr}|R(s)|\,ds, \end{equation} where $R$ denotes the full discarded term. The exponential amplification cannot be replaced by one without a stability argument.

There is an exact discrete illustration. For the doubling map $T(x)=2x\pmod1$, the unperturbed orbit from zero is zero. Perturb only the first step, setting $y_1=1/(3\cdot2^K)$, and then set $y_{j+1}=T(y_j)$. The sum of all forcing errors is the arbitrarily small number $1/(3\cdot2^K)$, but $y_{K+1}=1/3$, after which the orbit alternates between $1/3$ and $2/3$. Tracking the zero orbit fails permanently. This does not exclude shadowing by a different orbit with adjusted initial data; that is precisely why a correctly formulated shadowing theorem is needed. Nor is this expanding map being identified with the complete Bianchi IX flow.

Stress test against the exact target.

None of these arguments proves that a generic singularity is spacelike. The first lemma starts with the particular metric (30006736:metric); it cannot establish that a general Einstein solution admits this form. The Kasner-circle example fails the spatial Einstein equations, the oscillatory function fails to be a constructed gravitational solution, and the doubling map is a stability test rather than a relativistic counterexample. Homogeneous attractor theorems supply neither the missing spatial estimates nor an automatic identification of individual inhomogeneous trajectories. [E1]

A viable continuation would have to specify the vacuum data class and gauge, prove an integrated excursion bound or another causal-locality estimate for actual solutions, control the spatial derivative terms including their exceptional sets, and establish a suitable nonuniform shadowing statement. Quiescent matter solutions must remain outside any claim specifically about vacuum oscillations; suppressing that distinction would change the problem. [E2]

Outcome.

Outcome: Exploratory approach with unresolved gap.

The attempt gives an explicit integrated horizon criterion and its sharp logarithmic comparison threshold, together with three falsified shortcut lemmas. The varying-Kasner test has its vacuum residual calculated exactly, so it is not misreported as a counterexample. The companion script checks the algebraic identities and the discrete instability example.

No generic inhomogeneous Einstein estimate is established. In addition to those analytic gaps, the supplied formulation still lacks quantifiers for genericity and convergence. The notebook preserves that limitation rather than declaring a resolution of a newly chosen special formulation.

Sources

• S1 Josu C. Aurrekoetxea, Katy Clough, and Eugene A. Lim, Cosmology using numerical relativity, Section 3.1 $2025$. https://link.springer.com/article/10.1007/s41114-025-00058-z. Catalogue formulation source.
• E1 Hans Ringström, The Bianchi IX attractor, Annales Henri Poincaré 2 $2001$, 405--500; arXiv:gr-qc/0006035. Homogeneous attractor theorem and scope of the matter cases. https://arxiv.org/abs/gr-qc/0006035.
• E2 Lars Andersson and Alan D. Rendall, Quiescent cosmological singularities, Communications in Mathematical Physics 218 $2001$, 479--511; arXiv:gr-qc/0001047. Analytic scalar-field and stiff-fluid constructions. https://arxiv.org/abs/gr-qc/0001047.