arXiv:2611.03188
Universality and Emergent Classicality in Krylov Space
Debarghya Chakraborty
Abstract
Krylov complexity grows linearly at late times for many systems. In this regime, we give a general spectral formula for the leading growth coefficients of Krylov complexity and its higher moments, without requiring detailed knowledge of the Lanczos coefficients. For states with a narrow energy distribution compared to the bandwidth, dynamics in the Krylov basis resembles motion along a classical trajectory. We demonstrate our results for the critical XY chain and double-scaled SYK. For the latter, we compute the leading late-time behavior of the chord-number moments through one-loop order.
AI Summary
The paper studies what determines late-time ballistic spreading in a Krylov basis and when that spreading admits an effectively classical description. For a smooth, gapless spectral measure $\Phi(E)$ supported on $[-E_0,E_0]$, it develops the relation $\langle n^k(t)\rangle=v_k t^k+O(t^{k-1})$, with $v_k=\int dE\,\Phi(E)(E_0^2-E^2)^{k/2}$. This formula, also obtained in cited and concurrent work, determines leading growth without reconstructing the Lanczos sequence. Its basis is orthogonal-polynomial asymptotics; a complementary argument treats the asymptotic Krylov chain as a constant-hopping lattice. When the spectral distribution is narrow relative to the bandwidth and away from its edges, relative number fluctuations are suppressed, while the phase inferred from the Krylov shift operator concentrates near $p=\arccos(\langle E\rangle/E_0)$. The resulting trajectory has $n\simeq\sqrt{E_0^2-\langle E\rangle^2}\,t$. For operator spectra with exponential tails of scale $\alpha$ and cutoff $\omega_m$, a phenomenological model estimates $t_{\rm on}\simeq(2\alpha)^{-1}\log(\omega_m/\alpha)$ and $t_{\rm cl}\sim\omega_m^2/\alpha^3$. Critical-XY-chain calculations test these estimates. In double-scaled SYK, thermal spectral integrals yield all leading chord-number moments and their one-loop corrections. The stated remainder requires additional assumptions. Finite-system applicability depends on Lanczos fluctuations not obstructing propagation; the small nonintegrable-chain calculations do not resolve a clear linear regime. The classicalization timescale relies partly on empirical subleading corrections, narrow spectra near an edge need not classicalize, and the SYK loop expansion requires staying away from spectral edges.
Error found: The abstract overstates emergent classicality: an energy distribution narrow compared with the bandwidth can retain order-one relative Krylov fluctuations if concentrated sufficiently close to a spectral edge, an exception acknowledged in the main text.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 10 pages, 6 figures