A Schwinger-Keldysh Formulation of Semiclassical Operator Dynamics

This paper develops a real-time Schwinger-Keldysh formulation of Krylov dynamics that treats Krylov complexity as an in-in observable, revealing an emergent phase-space description where semiclassical chaos and integrability-chaos crossovers are characterized by the behavior of Lanczos coefficients and their fluctuations.

The Gist

The Big Picture: Tracking a "Fuzzy" Particle

Imagine you have a single drop of ink dropped into a glass of water. Over time, that drop spreads out, mixing with the water until it's everywhere. In quantum physics, scientists study how "information" (like a specific quantum operator) spreads through a complex system, similar to how that ink spreads.

For a long time, scientists have used a method called Krylov complexity to measure how far this information has traveled. Think of it like measuring how many steps a traveler has taken down a long, winding path. The standard way to calculate this involves a mathematical recipe (the Lanczos algorithm) that is very good at giving a number, but it's like looking at a map without understanding the terrain. It tells you where the traveler is, but not why they are moving that way or what the landscape looks like.

This paper introduces a new way to look at the problem. Instead of just counting steps, the authors build a dynamic movie of the journey. They use a tool from physics called the Schwinger-Keldysh formalism (which is usually used to study systems that are changing over time, like a cup of coffee cooling down) to create a "path integral."

The Analogy:
Imagine the standard method is like taking a photo of a runner at the finish line and calculating their average speed. The new method described in this paper is like putting a camera on the runner's chest and filming the whole race in slow motion, showing every stumble, every sprint, and every turn.

The New Tool: The "Closed Time Loop"

To get this "movie," the authors use a clever trick. In physics, to measure what happens inside a system (rather than just the start and end), you have to imagine time running forward and then backward simultaneously, like a loop.

• The Forward Path: Represents the system evolving normally.
• The Backward Path: Represents the system "un-evolving" to check the math.
• The Loop: By connecting these two, they create a closed loop that captures the full story of the system's behavior, including all the tiny fluctuations and "jitters" that usually get averaged out.

This allows them to treat the spreading of information not just as a list of numbers, but as a particle moving through a landscape.

The Landscape: A Hilly Path

In this new view, the "path" the information travels is a one-dimensional chain (like a ladder). The "Lanczos coefficients" (which were just numbers in the old method) now act like hills and valleys on this path.

• The Effective Hamiltonian: The authors show that these numbers create an invisible "force field" or a landscape. The information particle rolls down this landscape.
• The Saddle Point: In the middle of this landscape, there is a specific shape (a saddle) that determines how fast the particle moves.

The Discovery: Why Chaos Happens

The paper explains why chaotic systems (systems that are very sensitive to changes) behave the way they do.

1. The "Hyperbolic" Slide: When a system is chaotic, the landscape has a specific shape called a "hyperbolic trajectory." Imagine a slide that gets steeper and steeper the further you go. Once the information particle starts sliding down this specific path, it accelerates exponentially. This explains why complexity grows so fast in chaotic systems.
2. The Universal Fixed Point: The authors found that no matter how you tweak the system (as long as it's chaotic), the landscape eventually looks the same at the bottom. It's like how all rivers eventually flow into the ocean; they might start differently, but they all end up following the same "chaotic fixed point" rules.
3. Classifying the Tweaks: The paper categorizes different ways to change the system:
   • Irrelevant: Small changes (like shifting the starting point) don't change the final speed. The particle still slides down the same exponential slide.
   • Marginal: Changes that are just on the edge. They don't stop the slide, but they make the particle speed up or slow down very slowly over time.
   • Relevant: Big changes that flatten the slide. If the landscape isn't steep enough, the particle stops accelerating exponentially and just walks at a normal, slow pace. This signals that the system is not chaotic.

The Secret Weapon: Listening to the Noise

The most exciting part of this paper is what it reveals about fluctuations.

In the old method, scientists only looked at the "average" path. If you have a crowd of people walking, the average might show a smooth line. But the new method looks at the noise—the fact that some people run ahead, some lag behind, and some get stuck.

The authors show that even when the "average" path looks smooth and boring (like when a system is transitioning from being orderly to chaotic), the fluctuations (the noise) scream the truth.

• The Analogy: Imagine a crowd of people crossing a bridge. If the bridge is safe, everyone walks at a steady pace. If the bridge is shaky (chaotic), everyone jitters. The paper shows that by measuring how much the people are jittering (variance), you can detect a "shaky bridge" even if the average walking speed hasn't changed yet.

Summary

This paper takes a complex mathematical tool (Krylov complexity) and gives it a physical body. It turns a static calculation into a dynamic story of a particle rolling down a landscape.

• It explains chaos as a particle sliding down a steep, exponential hill.
• It explains order as a particle walking on flat ground.
• It proves that by listening to the noise (fluctuations) rather than just the average, we can spot the transition between order and chaos much more clearly than before.

This doesn't just give a number; it gives a geometric and physical reason for why quantum systems behave the way they do.

Technical Summary

Technical Summary: A Schwinger-Keldysh Formulation of Semiclassical Operator Dynamics

Problem Statement
Krylov complexity has emerged as a robust diagnostic for operator growth in quantum many-body systems, complementing out-of-time-order correlators (OTOCs) and spectral probes. The standard formulation relies on the Lanczos algorithm, which maps operator dynamics to a tight-binding problem on a semi-infinite chain characterized by Lanczos coefficients ($b_n$). While computationally efficient, this recursion-based approach obscures the underlying dynamical principles governing operator spreading. It offers limited insight into universality classes, struggles to systematically classify finite-size effects and fluctuations, and lacks a natural framework for extending to open quantum systems. The authors identify a need for an explicit formulation of Krylov complexity as an effective dynamical theory, rather than merely a computational construct.

Methodology
The authors develop a real-time Schwinger-Keldysh (SK) formulation of Krylov dynamics. The core methodology involves:

1. Tight-Binding Mapping: Recasting the Heisenberg evolution of an operator in the Krylov basis as a single-particle quantum mechanics problem on a semi-infinite lattice with position-dependent hopping amplitudes $b_n$.
2. Closed-Time-Contour Path Integral: Treating Krylov complexity as an "in-in" observable. The authors construct a generating functional $Z_{SK}[J_+, J_-]$ on a closed time contour, coupling sources to the Krylov position operator on forward ($+$) and backward ($-$) branches.
3. Phase-Space Representation: By introducing conjugate momentum states, the authors derive a phase-space path integral. This yields an effective action $S_{SK}$ where the Lanczos coefficients define an emergent effective Hamiltonian, $H_{\text{eff}}(n, p)$, governing motion in Krylov phase space.
4. Semiclassical Analysis: The authors analyze the saddle-point equations of the SK action, which correspond to classical Hamilton's equations. They classify perturbations to the linear growth of Lanczos coefficients ($b_n \sim \alpha n$) using renormalization group (RG) terminology (irrelevant, marginal, relevant) to determine their impact on the stability of the dynamics.
5. Fluctuation Analysis: Beyond the saddle point, the framework allows for a systematic loop expansion to compute higher cumulants and large-deviation statistics of the Krylov complexity.

Key Contributions and Results

• Emergent Phase-Space Description: The formulation reveals that Krylov dynamics is governed by a classical Hamiltonian $H_{\text{eff}}(n, p) = 2b(n)\cos p + \dots$ in the semiclassical limit. The Lanczos coefficients $b_n$ act as the position-dependent potential in this effective theory.
• Hyperbolic Trajectories and Chaos: For chaotic systems where $b_n$ grows linearly ($b_n \sim \alpha n$), the effective Hamiltonian generates hyperbolic trajectories in phase space. The exponential growth of Krylov complexity is identified as a manifestation of classical instability (Lyapunov behavior) along these trajectories, with a growth rate $\lambda_K = 2\alpha$.
• Universality and RG Classification: The paper classifies deviations from linear Lanczos growth:
  • Irrelevant Perturbations: Constant shifts ($b_n = \alpha(n+c)$) or logarithmic corrections ($b_n = \alpha n + \beta \ln n$) do not alter the exponential growth rate or the hyperbolic nature of the fixed point. They only modify prefactors or induce slow drifts. This explains the robustness of exponential growth in models like $SU(1,1)$ and conformal systems.
  • Marginal Perturbations: Deformations of the form $b_n = \alpha n (1 + \epsilon/\ln n)$ induce a slow, universal running of the effective instability exponent, resulting in power-law corrections to the exponential growth ($n(t) \sim t^\epsilon e^{2\alpha t}$).
  • Relevant Perturbations: Sublinear growth ($b_n \sim n^\gamma$ with $\gamma < 1$) destroys the hyperbolic instability, causing the angular velocity to vanish and leading to polynomial (rather than exponential) complexity growth.
• Fluctuations and Integrability-Chaos Crossovers: The SK framework provides access to the full probability distribution $P(n, t)$ and its cumulants. The authors demonstrate that while the mean complexity $K(t)$ may vary smoothly across an integrability-chaos crossover, the fluctuations (variance and large-deviation rate functions) exhibit sharp signatures. In regimes where the system transitions from integrable to chaotic, the "escape time" from a weakly hyperbolic region diverges, leading to parametrically enhanced fluctuations even when the mean growth appears smooth.
• Exact Solutions: The formalism is validated against exactly solvable models, including square-root hopping (harmonic oscillator) and $SU(1,1)$ Liouvillian models. In these cases, the SK generating functional reproduces known results for $K(t)$ and the full distribution $P(n, t)$ without relying on explicit recursion solutions, instead deriving them from algebraic properties of the effective Hamiltonian.

Significance
The paper claims to reorganize Krylov complexity from a recursion-based construct into a dynamical field-theoretic framework. Its primary significance lies in:

1. Conceptual Clarity: It provides a geometric and physical origin for operator growth, identifying linear Lanczos growth as a universal chaotic fixed point in an emergent phase space.
2. Systematic Classification: It offers a systematic way to classify universality classes of operator growth based on the asymptotic behavior of Lanczos coefficients, distinguishing between robust chaotic behavior and integrable-like dynamics.
3. New Diagnostics: By elevating the focus from the mean complexity to the full distribution and its fluctuations, the formulation identifies new diagnostics (such as large-deviation functions) capable of detecting sharp dynamical crossovers that are invisible to mean-field observables.
4. Extensibility: The closed-time-contour approach naturally extends to open quantum systems and non-equilibrium dynamics, providing a unified description of Krylov complexity in both unitary and dissipative regimes.

The authors position this work as a step toward applying standard non-equilibrium field theory tools (semiclassical analysis, fluctuation theory, large deviations) to the study of quantum chaos and thermalization, moving beyond the limitations of the standard Lanczos algorithm.