Berry's 1981 Appendix I conjecture: suppression of level-density oscillations by fractal billiard boundaries

In the closing appendix of his 1981 Annals of Physics paper on quantizing the Sinai billiard, M. V. Berry conjectured that a two-dimensional drum whose boundary is a fractal curve of Hausdorff dimension $D$ almost completely suppresses the oscillatory part of its spectral counting function: modeling the boundary locally by a Weierstrass–Mandelbrot curve, the effective radius of curvature of the $k$-smoothed boundary shrinks as $R(k)\sim k^{-D}$, and his closed-orbit formula for the dominant two-reflection orbit then gives the oscillatory counting function as $\tilde{N}(k)\sim(\mathrm{const}/k^{D})\sin 2kL$, predicting nearly regularly spaced levels. Berry stated the argument is 'hardly rigorous' and that the desirable test — the spectrum and its density oscillations for an exactly soluble model, for which he suggested the Koch snowflake boundary — 'to my knowledge no such model exists to date'. The objective is to establish or refute the predicted $k^{-D}$ suppression law for drums with fractal boundaries, determining which fractal dimension and mechanism control the oscillatory spectral asymptotics, and test the prediction on an exactly or numerically tractable fractal-boundary model such as the Koch snowflake drum. No resolution of this law is known in the literature to date: the adjacent lines (the Weyl–Berry–Lapidus conjecture for the smooth counting term, the fractal Weyl law for resonances of open systems, wave localization on fractal drums, complex-dimension tube and heat asymptotics) all address different claims.

Background

A quantum billiard is a wave system confined by perfectly reflecting walls — equivalently a membrane or acoustic cavity — whose allowed wave numbers $k$ are the eigenvalues of the Laplacian with Dirichlet boundary conditions. The counting function $N(k)$, the number of eigenvalues below $k$, splits into a smooth 'Weyl' term growing proportionally to the enclosed area, plus an oscillatory part $\tilde{N}(k)$ that semiclassics expresses as a sum over classical periodic orbits ('closed' orbits): each repeating classical trajectory contributes an oscillating term whose amplitude is set by the orbit's length and stability. In integrable systems, closed orbits appear in continuous families and produce strong oscillations; in fully chaotic (ergodic, meaning that typical trajectories cover the accessible phase-space region uniformly) systems, closed orbits are isolated and unstable, and the resulting weak level fluctuations agree with random-matrix statistics (universal fluctuation laws modeled on the eigenvalues of large random matrices).

Berry's 1981 Annals of Physics paper 'Quantizing a classically ergodic system: Sinai's billiard and the KKR method' studied the paradigmatic chaotic case: a disc at the center of a periodic unit cell (the Sinai billiard, a classically ergodic system). He computed exact eigenvalues via the KKR method (the Korringa–Kohn–Rostoker multiple-scattering determinant of solid-state band theory), then analytically transformed this exact determinant into the semiclassical oscillatory level density as a sum over classical closed orbits. The derivation required cancelling 'ghost' contributions — terms of the multiple-scattering expansion on the unfolded disc lattice corresponding to paths that would pass through intermediate discs; one cancellation step rests on a combinatorial identity (his Eq. (7.35)) and on a statistical assertion (his Eq. (7.39), $\int db\,n(b,\rho_0)^{-1}=-1/\rho_0$ for primitive lattice vectors $\rho_0$ with intermediate ghosts) about which Berry wrote that he had checked particular cases and had a statistical argument, but 'not found a general proof'. Other steps (Debye/WKB phase shifts, i.e. asymptotic approximations to the partial-wave scattering phases; stationary-phase integrals; Ewald resummation of the lattice sums, a convergence-acceleration technique borrowed from solid-state physics) were justified asymptotically and numerically, not with uniform error bounds; and the number of isolated orbits grows exponentially with orbit length, so the orbit sums converge only slowly.

The question addressed here is the one Berry posed as an explicit open problem, in Appendix I. Motivated by auditorium acoustics — choosing a wall shape that suppresses clustering of resonance frequencies — he asked what happens when the boundary itself is a fractal: a continuous but nowhere-differentiable curve of Hausdorff dimension $D$ (the Hausdorff dimension measures how the measure of the curve scales under magnification; a smooth curve has $D=1$, and a plane fractal curve can have $1<D<2$). Classical reflection is then undefined, but the modes exist, and Berry gave a heuristic estimate of their clustering: a wave of wave number $k$ cannot resolve boundary detail below its wavelength, so the boundary may be replaced by a '$k$-smoothed' boundary $B_k$ whose effective radius of curvature $R(k)$ shrinks as increasing $k$ reveals the fractality. Modeling the boundary locally by the Weierstrass–Mandelbrot function (an explicitly constructed fractal graph with tunable dimension, introduced by Berry, Lewis and Nye), whose curvature he estimated as growing like $k^{D}$, Berry obtained $R(k)\sim k^{-D}$ and, substituting into his two-reflection 'bounce' orbit formula, his Eq. (I.5): $\tilde{N}(k)\sim(\mathrm{const}/k^{D})\sin 2kL$. He concluded that fractal boundaries should produce 'almost complete suppression of the oscillations in level density, so that the levels themselves should be very regularly spaced', emphasized that 'these arguments are hardly rigorous', and called for a check by computing the spectrum and its density oscillations for an exactly soluble model — 'To my knowledge no such model exists to date' — suggesting the Koch snowflake curve (a standard exactly self-similar fractal boundary) as a candidate.

This conjecture must be distinguished from two adjacent, better-developed lines. The Weyl–Berry–Lapidus conjecture concerns the smooth second term of the counting function itself, predicting a correction of order $k^{D}$ from the fractal boundary; it was partially resolved by Lapidus (with the relevant dimension being the Minkowski, i.e. box-counting, dimension rather than the Hausdorff dimension) and confirmed numerically for Koch-curve perimeters above a crossover frequency by Hobiki, Yakubo and Nakayama. The 'fractal Weyl law' for open chaotic systems, proved in important cases by Nonnenmacher, Sjöstrand and Zworski, counts scattering resonances governed by the fractal dimension of a trapped set — resonances of open systems, a different spectral object from the eigenvalue oscillations of a closed drum. Berry's Appendix-I prediction concerns the oscillatory, closed-orbit part of the spectrum of a closed drum with a fractal boundary, for which no exact model and no rigorous asymptotic law is known; the modern theory of 'complex dimensions' (poles of fractal zeta functions, which produce log-periodic oscillations in tube volumes and heat content of self-similar boundaries) has so far treated the geometric and heat-kernel sides, not the eigenvalue oscillations.

Problem Statement

Berry's Appendix I conjecture (1981) on fractal drums: for a two-dimensional drum whose boundary is a fractal curve of Hausdorff dimension $D$ with $1<D<2$, the amplitude of the oscillatory part of the spectral counting function is predicted to be suppressed as $\tilde{N}_{\mathrm{osc}}(k)\sim k^{-D}$ (Berry's Eq. (I.5): $\tilde{N}\sim(\mathrm{const}/k^{D})\sin 2kL$ for the dominant two-reflection orbit, following from $R(k)\sim k^{-D}$ for the effective radius of curvature of the $k$-smoothed boundary), so that a fractal boundary almost completely suppresses level-density oscillations and the levels themselves should be very regularly spaced. Berry stated this argument is 'hardly rigorous' and that no exactly soluble model exists to check it, suggesting the Koch snowflake boundary. The problem is to establish or refute this conjecture: (i) determine the true high-$k$ asymptotic behavior of the oscillatory part of the counting function (equivalently, the level density) for drums with fractal boundaries, including which fractal dimension (Hausdorff, Minkowski, or a spectral quantity such as the real parts of complex dimensions) and which mechanism controls the decay, and how log-periodic self-similar oscillations interplay with the predicted suppression; and (ii) test the prediction on a controlled model — ideally by computing, exactly or with controlled asymptotic or numerical methods, the spectrum and its density oscillations for an exactly treatable fractal-boundary drum such as the Koch snowflake drum, as Berry proposed. A determinate asymptotic law for the oscillatory counting function, together with its governing dimension (Hausdorff, Minkowski, or a real part of a complex dimension) and mechanism, either confirms or refutes the conjecture; a law governed by a non-Hausdorff dimension refutes the specific Hausdorff-dimension form of the prediction. The two parts serve one objective: deciding whether boundary fractality suppresses closed-orbit spectral oscillations in the predicted universal way.

References

1. M. V. Berry, 'Quantizing a classically ergodic system: Sinai's billiard and the KKR method', Annals of Physics 131 (1981) 163-216. DOI: 10.1016/0003-4916(81)90189-5, https://doi.org/10.1016/0003-4916(81)90189-5
2. M. V. Berry, Z. V. Lewis and J. F. Nye, 'On the Weierstrass-Mandelbrot fractal function', Proceedings of the Royal Society of London A 370 (1980) 459-484. DOI: 10.1098/rspa.1980.0044, https://doi.org/10.1098/rspa.1980.0044
3. P. Cvitanović and B. Eckhardt, 'Periodic-orbit quantization of chaotic systems', Physical Review Letters 63 (1989) 823-826. DOI: 10.1103/PhysRevLett.63.823, https://doi.org/10.1103/PhysRevLett.63.823
4. H. Primack and U. Smilansky, 'Quantization of the Three-Dimensional Sinai Billiard', Physical Review Letters 74 (1995) 4831-4834. DOI: 10.1103/PhysRevLett.74.4831, https://doi.org/10.1103/PhysRevLett.74.4831
5. H. Primack and U. Smilansky, 'The quantum three-dimensional Sinai billiard - a semiclassical analysis', Physics Reports 327 (2000) 1-107. DOI: 10.1016/S0370-1573(99)00093-9, https://doi.org/10.1016/S0370-1573(99)00093-9
6. P. Dahlqvist and G. Vattay, 'Periodic orbit quantization of the Sinai billiard in the small scatterer limit', Journal of Physics A 31 (1998) 6333-6345. DOI: 10.1088/0305-4470/31/30/003, https://doi.org/10.1088/0305-4470/31/30/003 (preprint arXiv:chao-dyn/9710012, https://arxiv.org/abs/chao-dyn/9710012)
7. B. Sapoval, Th. Gobron and A. Margolina, 'Vibrations of fractal drums', Physical Review Letters 67 (1991) 2974-2977. DOI: 10.1103/PhysRevLett.67.2974, https://doi.org/10.1103/PhysRevLett.67.2974
8. Y. Hobiki, K. Yakubo and T. Nakayama, 'Spectral distribution of drums with fractal perimeters: The Weyl-Berry-Lapidus conjecture', Physical Review E 52 (1995) R1310-R1312. DOI: 10.1103/PhysRevE.52.R1310, https://doi.org/10.1103/PhysRevE.52.R1310
9. M. L. Lapidus, 'Fractal drum, inverse spectral problems for elliptic operators and a partial resolution of the Weyl-Berry conjecture', Transactions of the American Mathematical Society 325 (1991) 465-529. DOI: 10.1090/S0002-9947-1991-0994168-5, https://doi.org/10.1090/S0002-9947-1991-0994168-5
10. M. L. Lapidus and E. P. J. Pearse, 'A tube formula for the Koch snowflake curve, with applications to complex dimensions', Journal of the London Mathematical Society 74 (2006) 397-414. DOI: 10.1112/S0024610706022988, https://doi.org/10.1112/S0024610706022988 (preprint arXiv:math-ph/0412029, https://arxiv.org/abs/math-ph/0412029)
11. M. L. Lapidus, J. W. Neuberger, R. J. Renka and C. A. Griffith, 'Snowflake harmonics and computer graphics: numerical computation of spectra on fractal drums', International Journal of Bifurcation and Chaos 6 (1996) 1185-1210. DOI: 10.1142/S0218127496000680, https://doi.org/10.1142/S0218127496000680
12. J. M. Neuberger, N. Sieben and J. W. Swift, 'Computing eigenfunctions on the Koch Snowflake: a new grid and symmetry', Journal of Computational and Applied Mathematics 191 (2006) 126-142. DOI: 10.1016/j.cam.2005.03.075, https://doi.org/10.1016/j.cam.2005.03.075 (preprint arXiv:1010.0775, https://arxiv.org/abs/1010.0775)
13. M. L. Lapidus and R. G. Niemeyer, 'The current state of fractal billiards', arXiv:1210.0282 (2012), https://arxiv.org/abs/1210.0282
14. W. E. Hoffer and M. L. Lapidus, 'On Complex Dimensions and Heat Content of Self-Similar Fractals', arXiv:2508.09512 (2025), https://arxiv.org/abs/2508.09512
15. S. Nonnenmacher, J. Sjöstrand and M. Zworski, 'Fractal Weyl law for open quantum chaotic maps', Annals of Mathematics 179 (2014) 179-251. DOI: 10.4007/annals.2014.179.1.3, https://doi.org/10.4007/annals.2014.179.1.3 (preprint arXiv:1105.3128, https://arxiv.org/abs/1105.3128)
16. R. S. Strichartz and S. C. Wiese, 'Spectral Properties of Laplacians on Snowflake Domains and Filled Julia Sets', Experimental Mathematics 31 (2022) 1014-1025. DOI: 10.1080/10586458.2020.1743213, https://doi.org/10.1080/10586458.2020.1743213 (preprint arXiv:1903.08259, https://arxiv.org/abs/1903.08259)
17. S. Kombrink and L. Schmidt, 'Eigenvalue Counting Functions and Parallel Volumes for Examples of Fractal Sprays Generated by the Koch Snowflake', in 'Recent Developments in Fractals and Related Fields' (Trends in Mathematics, Birkhäuser, 2025) 237-256. DOI: 10.1007/978-3-031-80453-3_10, https://doi.org/10.1007/978-3-031-80453-3_10 (preprint arXiv:2312.12331, https://arxiv.org/abs/2312.12331)