⚛️ lattice

Spectral densities from Euclidean correlators via integral transforms: theoretical framework

This paper presents a theoretical framework using integral transforms to derive analytic formulae for extracting spectral densities from Euclidean correlators, addressing both continuum and lattice scenarios while providing rigorous bounds to control errors arising from finite temporal extents.

Original authors: Leonardo Giusti, Matteo Saccardi, Diego Toniolo

Published 2026-06-29

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Leonardo Giusti, Matteo Saccardi, Diego Toniolo

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: Listening to a Symphony in a Noisy Room

Imagine you are trying to understand a complex symphony (the Spectral Density), which tells you exactly which instruments are playing and how loud they are. This symphony represents the fundamental rules and particles of a quantum universe.

However, you cannot hear the symphony directly. Instead, you are in a room where you can only hear a muffled, fading echo of the music (the Correlator). This echo is recorded on a computer simulation (the Lattice) that has two major problems:

The Room is Small: The recording stops after a certain amount of time because the computer simulation runs out of memory or time.
The Floor is Grid-Like: The recording isn't smooth; it's taken in tiny, discrete steps (like a digital video with low resolution), which can make the sound sound "jagged."

The goal of this paper is to provide a mathematical "decoder ring" that allows physicists to take that short, jagged echo and perfectly reconstruct the original symphony, while also telling them exactly how much they can trust the result.

The Core Problem: The "Inverse" Puzzle

In physics, there is a known rule: if you know the symphony (the spectral density), you can mathematically predict what the echo will sound like. This is easy. It's like knowing the ingredients of a cake and baking it.

The hard part is the Inverse Laplace Transform. This is the process of looking at the echo and figuring out what the ingredients were. It is notoriously difficult because:

The echo fades away very quickly.
Small errors in the echo get blown up into huge errors in the recipe.
If you only have a short recording, you are missing the "tail" of the echo, which makes it impossible to know for sure what the full symphony sounds like.

The Solution: A Special Kind of "Magic Glasses"

The authors propose using specific mathematical tools called Integral Transforms (specifically Mellin, Kontorovich–Lebedev, and Mehler-Fock transforms).

The Analogy of the Prism:
Imagine the echo is a beam of white light. To see the colors (the spectral density), you need to pass it through a prism.

Old methods were like trying to guess the colors by squinting. They were unstable and often gave wrong answers.
This paper's method provides a perfect, custom-made prism. It takes the echo and mathematically splits it into its component frequencies using a special set of "basis functions" (the lenses of the prism).

The paper shows that if you use these specific lenses, the math becomes much simpler. Instead of a messy, impossible puzzle, it turns into a straightforward algebraic equation. You take the "transformed" echo, divide by a known number, and you get the "transformed" symphony. Then, you just reverse the process to get the final answer.

Handling the "Short Recording" (Incomplete Data)

Since the computer simulation stops at a certain time ( $t_{max}$ ), the authors have to deal with a recording that is cut off.

The Analogy of the Fading Scream:
Imagine someone screaming a note that gets quieter and quieter until it vanishes. If you stop recording while they are still screaming (but very quietly), you miss the very end.

The authors prove that if the note fades away fast enough (which it does in these physics theories), the part you missed is tiny.
They introduce a concept called "Incomplete Spectral Densities." This is a version of the symphony reconstructed only from the part of the echo you do have.
Crucially, they provide a Rigorous Bound. This is like a safety net. They calculate a mathematical "error bar" that says: "We are missing the very end of the recording, but we know for a fact that the missing part cannot change the result by more than X amount."

This allows physicists to say, "We have reconstructed the symphony with 99% certainty, and here is the exact math proving it."

Handling the "Grid Floor" (Discrete Lattice)

The computer simulation doesn't measure time smoothly; it measures it in steps (like a clock ticking). This creates "discretization errors."

The Analogy of the Staircase:
Imagine walking up a smooth ramp (continuous time) versus walking up a staircase (discrete time).

The authors show that if the staircase is built correctly (using a technique called "O(a)-improvement"), the difference between the ramp and the stairs becomes negligible as the steps get smaller.
They prove that their mathematical decoder works perfectly even on the staircase, provided the steps are small enough. The "jaggedness" of the grid doesn't ruin the final symphony.

Smearing: Listening to a "Blurry" Version

Sometimes, physicists don't need the exact, sharp note of a single instrument. They want to know the "blurry" sound of a whole section (like the whole string section playing together). This is called a Smeared Spectral Density.

The Analogy of the Soft Focus Lens:

Trying to hear a single, sharp note from a muffled echo is very hard.
Trying to hear the "blurry" sound of a whole section is easier.
The paper shows that if you ask for a "blurry" version (using a specific mathematical filter called a kernel, like a Breit-Wigner shape), the missing parts of the recording (the tail you didn't capture) become even less important.
The math proves that if you choose your "blur" correctly, the error from the missing data drops off very quickly, making the result very stable and reliable.

Summary of Claims

New Formulas: The paper provides explicit mathematical formulas to turn a short, jagged computer echo into a smooth, physical spectrum.
Error Control: It doesn't just give an answer; it gives a strict mathematical guarantee of how much error exists due to the recording stopping early.
Grid Independence: It proves these formulas work correctly even when the data comes from a "stepped" computer grid, as long as the grid is fine enough.
Smearing: It shows that asking for "blurred" or "smeared" answers is a smart strategy because it makes the missing data less dangerous.

What the paper does NOT do:
The paper is purely theoretical. It does not apply these formulas to real-world data yet, nor does it claim to solve specific medical or engineering problems. It sets up the "decoder ring" and proves it works in theory; a companion paper (mentioned in the text) will show how to use it on actual supercomputer data.

Technical Summary: Spectral densities from Euclidean correlators via integral transforms: theoretical framework

Problem Statement
Spectral densities are fundamental in quantum field theory (QFT), linking dynamical properties to experimental observables such as inclusive hadronic cross sections and semi-leptonic decay rates. In non-perturbative regimes, particularly in Quantum Chromodynamics (QCD), these densities are encoded in the Euclidean time-dependence of correlation functions via the Källén-Lehmann spectral decomposition, which relates them through a Laplace transform. Extracting the spectral density from lattice QCD data requires solving the Inverse Laplace Transform (ILT), a notoriously ill-posed problem.

Current challenges include:

Discretization Effects: Lattice correlators are defined at discrete time steps, introducing $O(a)$ or $O(a^2)$ artifacts.
Finite Temporal Extent: Practical lattice simulations are limited to a finite Euclidean time $t_{max}$ , leading to a lack of knowledge of the correlator at large times. This truncation prevents the exact reconstruction of the spectral density.
Noise and Resolution: As time-distance increases, the signal-to-noise ratio deteriorates, and the spectrum of states becomes denser in "master-field" simulations, complicating the extraction of continuous spectral functions from discrete data.

Methodology
The authors propose a theoretical framework based on integral transforms to analytically invert the Laplace transform relation. The core strategy involves:

Integral Transform Formulation: Instead of direct numerical inversion, the authors map the correlation function $C(t)$ and the spectral density $\rho(\omega)$ into a conjugate space using specific integral transforms (Mellin, Kontorovich–Lebedev, and Mehler-Fock).
Algebraic Relation: By choosing basis functions that are eigenfunctions of (quasi-)Carleman operators, the integral equation relating $C(t)$ and $\rho(\omega)$ is converted into a simple algebraic relation in the transform domain: $C_s = \lambda_s \hat{\rho}_s$ .
Inverse Reconstruction: The spectral density is recovered via an inverse transform integral: $\rho(\omega) = \int (C_s / \lambda_s) \hat{u}_s(\omega) ds$ . The authors explicitly address the "delicate cancellation" between the numerator and denominator in this integral, which typically decays exponentially.
Handling Discretization and Truncation:
- Discrete Sampling: The framework is adapted for lattice data by discretizing the integral transforms. The authors demonstrate that if the lattice correlator is $O(a)$ -improved, the extracted spectral density is also $O(a)$ -improved, with leading discretization effects appearing only at $O(a^2)$ .
- Incomplete Transforms: For correlators known only up to $t_{max}$ , the authors introduce "incomplete" integral transforms. They rigorously bound the systematic error introduced by the unknown tail of the correlator ( $t > t_{max}$ ).
Smeared Spectral Densities: To mitigate the instability of reconstructing the exact density, the authors focus on smeared spectral densities (integrals of the density against a kernel $\kappa(\omega)$ , e.g., Breit-Wigner). They show that if the integral transform of the smearing function decays sufficiently fast with the conjugate variable $s$ , the contribution from the unknown tail can be bounded and controlled.

Key Contributions and Results

Analytic Formulae: The paper derives explicit analytic formulae to extract regulated and/or smeared spectral densities from both continuous and discrete Euclidean correlators.
Rigorous Error Bounds: A major contribution is the derivation of rigorous upper bounds for the systematic uncertainty arising from the finite temporal extent of the lattice. The error is shown to be exponentially suppressed with $t_{max}$ provided the smearing function's transform decays fast enough.
Discretization Analysis: The authors prove that the proposed lattice solution converges to the continuum counterpart with $O(a^2)$ effects, provided the input lattice correlator is $O(a)$ -improved.
Numerical Validation: The framework is tested on single exponential and multi-exponential correlators. The results demonstrate that:
- The reconstructed smeared spectral density converges to the exact result as $t_{max}$ increases.
- The error bounds derived theoretically match the observed deviations in numerical tests.
- Preconditioning (subtracting known leading exponential states) significantly improves the reconstruction by effectively increasing the relevant mass gap for the residual correlator.
Generalizability: The formalism is presented in a general context, applicable to Mellin, Kontorovich–Lebedev, and Mehler-Fock transforms, and extends to regulated spectral densities (where $\rho(\omega)$ may diverge at low $\omega$ ).

Significance and Claims
The paper claims to provide a robust, theoretically grounded framework for the direct extraction of spectral information from lattice QCD data. Its significance lies in:

Systematic Control: It offers a method to rigorously quantify and bound the systematic errors associated with the finite time extent of lattices, a common limitation in current simulations.
Precision: By leveraging $O(a)$ -improved correlators and integral transforms, the method avoids introducing additional $O(a)$ discretization errors during the inversion process.
Practical Utility: The focus on smeared spectral densities aligns with physical observables (like cross-sections) that are naturally inclusive, making the method directly applicable to phenomenology.
Foundation for Future Work: The authors state that the practical implementation of this strategy, including its application to real multi-level Monte Carlo data, will be detailed in a companion paper. The current work establishes the necessary theoretical machinery and error analysis.

The authors remain modest regarding immediate experimental applications, positioning this work as a theoretical framework to enable more precise non-perturbative calculations in the future, particularly as lattice techniques (like multi-level integration) continue to improve the quality of Euclidean correlator data.