Toward a Refinement of Krylov Complexity: Scrambling, Classical Operator Growth and Replicas

arXiv:2603.19359 · hep-th, cond-mat.stat-mech, quant-ph · Submitted 2026-03-19 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: High Energy Physics: Theory. The day in high-energy theory: gravity, strings, fields and the questions left open.

Kai: We're looking at the paper: "Toward a Refinement of Krylov Complexity".

Mira: Here's the gist of the paper. The authors propose and test a new complexity measure called logarithmic Krylov complexity (logK) and its exponentiated form (elogK), which is derived by applying a replica trick to higher-order generalizations of conventional Krylov complexity.

Kai: The motivation for this proposal is to resolve "fake" chaos signatures observed in saddledominated scrambling systems, where conventional measures might incorrectly suggest exponential growth due to unstable saddles.

Mira: The manuscript details the construction and properties of logK and elogK.

Kai: They demonstrate that in finite-dimensional quantum systems exhibiting both classical Liouville integrability and quantum mechanical integrability (but not necessarily Bethe-Ansatz integrable), such as the Lipkin–Meshkov–Glick (LMG) model and the mixed-field Ising model at the chaotic point, logK discriminates effectively between genuine and saddle-dominated scrambling at early times.

Paper summary: Mira: Looking at the whole piece, "Toward a Refinement of Krylov Complexity: Scrambling, Classical Operator Growth and Replicas," the authors are essentially proposing logK as this new tool to distinguish between real chaos and spurious signals from saddle points in scrambling systems.

Kai: The main thrust is that by using this replica approach to a higher-order generalization of Krylov complexity, they get a measure that should avoid those exponentially growing contributions from unstable saddles when probing early-time operator scrambling.

Lev: What this means for us on the hardware side is that if we use logK instead of standard Krylov complexity, we might be getting a cleaner look at how the system scrambles itself in the very first moments before it settles into late-time behavior.

Kai: The paper also touches on classical systems and shows that these issues aren't just quantum artifacts; they don't show those false positives when you look at the classical Krylov complexity or elogK complexity, because those are sub-exponential in those cases.

Mira: The authors conclude that elogK is a viable probe for early-time scrambling without generating false positives, and it’s a solid foundation for defining a more refined quantum complexity indicator.

Lev: So the implication is we have a better way to characterize the initial dynamics of these complex systems than what we might be seeing with just standard Krylov measures right now.

Conclusion: Kai: So, we’re wrapping up on this piece about "Toward a Refinement of Krylov Complexity: Scrambling, Classical Operator Growth and Replicas." It really boils down to finding a better way to measure how fast things scramble in quantum systems.

Mira: Exactly. The authors are introducing logK and elogK as these new measures because the old ones sometimes give you the wrong answer when dealing with systems that have classical structure mixed with quantum chaos.

Lev: It’s about correcting those spurious signals, right? They're using a replica trick to derive these new forms of complexity, which helps them separate the real scrambling behavior from things that just look exponential because of unstable points in the system.

Kai: From my side on the hardware, what this means is if we're building systems—like those superconducting qubits we’ve been working on—we can use this to check if our initial dynamics are truly chaotic or just getting tricked by a saddle point in the model.

Mira: The core finding is that elogK works in both cases. It correctly tracks the exponential growth when it happens, but it stops growing at late times, which is what we expect for most systems.

Lev: And they showed that even in classical versions of these measures, like those looking at classical phase space dynamics, there aren't these false positive signals from unstable saddles either; those are just sub-exponential growth.

Kai: So the implication is that elogK might be a more reliable starting point for defining what we mean by quantum complexity indicators in scrambling research.

Mira: It sounds like they’re providing a tool that’s robust enough to handle both the chaotic and the integrable regimes, which is exactly what you need to understand these complex dynamics.

Physics Division, National Center for Theoretical Sciences, National Taiwan University · Department of Physics, National Sun Yat-sen University, Kaohsiung 80424, Taiwan · Department of Physics and Photon Science, Gwangju Institute of Science and Technology · Research Center for Photon Science Technology, Gwangju Institute of Science and Technology

hep-th, cond-mat.stat-mech, quant-ph

Submitted: 2026-03-19

Updated: 2026-11-10

Comments: 32 pages, 7 figures, 1 table;v2:minor changes, references updated;v3:typos fixed, matches published version in PRD

Journal ref: Phys. Rev. D 114, 106004 (2026)

DOI: 10.1103/s95g-g577

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 91/100

The gist: The authors propose and test a new complexity measure called logarithmic Krylov complexity (logK) and its exponentiated form (elogK)…

Key concepts

Logarithmic Krylov Complexity ($ ext{logK}$)
A new complexity measure derived using a replica trick applied to higher-order Krylov complexity. It is designed to accurately track true chaotic scrambling without being misled by unstable saddle points in systems that have classical integrability.
Exponentiated Logarithmic Krylov Complexity ($ ext{elogK}$)
The exponential form of $ ext{logK}$. This measure shows exponential growth characteristic of chaotic behavior in finite-dimensional systems with unstable saddles, but both $ ext{logK}$ and $ ext{elogK}$ eventually saturate at late times.
Saddle-Dominated Scrambling
A phenomenon observed in certain systems where conventional complexity measures incorrectly suggest exponential growth. This occurs because unstable classical saddle points can artificially inflate the measured complexity, leading to 'fake' chaos signatures.
Replica Trick
A mathematical technique used by the authors to derive $ ext{logK}$ and higher-order generalizations of Krylov complexity. It involves applying this trick to modify the spreading operator, allowing for a refined analysis in both quantum and classical contexts.

Terminology

Summary

The authors propose and test a new complexity measure called logarithmic Krylov complexity (logK) and its exponentiated form (elogK), which is derived by applying a replica trick to higher-order generalizations of conventional Krylov complexity. p. 1 The motivation for this proposal is to resolve fake chaos signatures observed in saddledominated scrambling systems, where conventional measures might incorrectly suggest exponential growth due to unstable saddles.

The manuscript details the construction and properties of logK and elogK. p. 1 They demonstrate that in finite-dimensional quantum systems exhibiting both classical Liouville integrability and quantum mechanical integrability (but not necessarily Bethe-Ansatz integrable), such as the Lipkin–Meshkov–Glick (LMG) model and the mixed-field Ising model at the chaotic point, logK discriminates effectively between genuine and saddle-dominated scrambling at early times. p. 0 Specifically, it correctly avoids the exponential contribution arising from unstable saddles in saddle-dominated cases while closely tracking conventional Krylov complexity in truly chaotic systems. p. 0

Numerical comparisons show that for finite-dimensional many-body systems containing classically unstable saddles, elogK exhibits a window of exponential growth characteristic of chaotic behavior, and both logK and elogK show late-time saturation. p. 8 The authors conclude that elogK successfully captures the correct dynamical behavior in both saddle-dominated scrambling and truly chaotic finite-dimensional systems, suggesting it serves as a reliable starting point for a refined definition of quantum complexity indicator for scrambling. p. 2

In related analyses:

  1. In integrable quantum systems with infinite-dimensional Krylov subspaces (like the SYK2 model and the quantum inverted harmonic oscillator), modifying the Krylov spreading operator via analytic continuation in the replica trick allows logK to be refined to capture integrable properties. p. 0

  2. The authors extend this formalism to classical dynamical systems, defining classical versions of these measures, showing that false positives arising from unstable saddles in classical phase space are non-existent, as both classical Krylov and elogK complexity exhibit sub-exponential growth in these cases. p. 0

  3. In the conformal limit of the q-body Sachdev–Ye–Kitaev (SYK) model, both Krylov and elogK complexities show early-time exponential growth regardless of q. p. 2 Furthermore, in the inverted harmonic oscillator, while conventional Krylov complexity resembles the SYK case with eta = 1/2, higher-order Krylov and elogK complexities match the SYK case for eta = 1. p. 1

The manuscript is organized to review basic Krylov complexity (Section II), define logK and elogK (Section III), analytically compute it in the low-energy limit of SYK (Section IV), examine it numerically in LMG and mixed-field Ising models (Section V), extend the analysis to the inverted harmonic oscillator (Section VI), formulate classical Krylov approaches (Sections VII and VIII), define a refined spreading operator with universal information retention (Section IX), and conclude with results discussion in Sections X and XI. p. 1

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved AI systems can distinguish between genuine scrambling and saddle-dominated scrambling in finite-dimensional quantum systems by calculating logK complexity, which discriminates between genuine and saddle-dominated scrambling at early times.

  2. The improved system can be refined to capture integrable properties of theories like the SYK2 model by modifying the Krylov spreading operator through a generalized analytic continuation procedure in the replica trick.

  3. The AI can analyze classical dynamical systems to define classical versions of Krylov and logK complexity, showing that the false positives arising from unstable saddles in classical phase space are non-existent.

  4. For finite-dimensional systems with classically unstable saddles, the system can use elogK complexity to show a significant distinction between elogK complexity from the conventional Krylov complexity compared to the conventional Krylov complexity.

  5. The AI can analyze the behavior of logK and its exponentiated form in integrable models like the inverted harmonic oscillator, identifying that elogK is suppressed at early times compared to Krylov complexity.

  6. The system can utilize a q-sensitive measure of operator growth, Q(q) O(t), which accurately distinguishes between integrable q=2 and chaotic q>=4 cases in the SYKq model.

  7. The AI can use the refined definition of the Krylov spreading operator, which includes information about the system and operator while retaining the universal information from Krylov complexity, to resolve tensions found in integrable models like q=2 SYK.

  8. The system can be designed to compute a regularized logK complexity by subtracting a logarithmic divergence at n=0, leading to the renormalized expression LK(t) which is equivalent to the renormalized expression and shows that log(K) ∼ LK(t) at late times.

  9. The AI can analyze classical observables in phase space to find that the simplest observable with a non-trivial classical Krylov complexity is f(delta+ = delta+ + A-delta-, leading to a Krylov complexity Kq2 (t) that is independent of delta and, moreover, is also independent of the choice of normalization factor 1/mu(Sdelta(t)) in the phase space average.

Sources

Related papers