Spectral densities from Euclidean correlators via integral transforms: theoretical framework Preprint: CERN-TH-2026-150

​​ L. Giusti, ​​, M. Saccardi, ​​, and D. Toniolo · 2026

Read Spectral densities from Euclidean correlators via integral transforms: theoretical framework Preprint: CERN-TH-2026-150 by ​​ L. Giusti, ​​, M. Saccardi, ​​, and D. Toniolo (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-28167.txt.

Introduction

L. Giusti, M. Saccardi, and D. Toniolo lay out a theoretical framework for recovering spectral densities from lattice simulations of quantum field theory. A spectral density links what experiments measure to a theory's inner dynamics, and it is hidden inside correlation functions that decay over imaginary, or Euclidean, time.

Giusti and colleagues give analytic formulas, built from integral transforms, that undo the Laplace transform and extract the density from continuous or sampled correlator data, including smeared versions used to tame noise. They show the lattice answer approaches the continuum one as the grid spacing shrinks. Because real lattices end after finite time, they add incomplete transforms and prove rigorous bounds on the unknown tail's influence, which also tell simulators how large a lattice a target precision needs.

Catalog record

CREATOR​​ L. Giusti, ​​, M. Saccardi, ​​, and D. Toniolo
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE102 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.28167

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Spectral densities from Euclidean correlators via integral transforms: theoretical framework Preprint: CERN-TH-2026-150 ​​ L. Giusti, ​​, M. Saccardi, ​​, and D. Toniolo Abstract Spectral densities link experimental measurements to dynamical properties of a quantum field theory which, in turn, can be resolved non-perturbatively from the Euclidean time-dependence of correlation functions. By making extensive use of integral transforms, we present analytic formulae to carry out the inverse Laplace transform so as to extract spectral densities from either the continuum or the discrete sampling of correlation functions in the Euclidean time. Formulae extend to regulated and/or smeared spectral densities as well. We explicitly show that the proposed lattice solution tends to its continuum counterpart up to O ⁡ ( a 2 ) O(a^{2}) effects in the lattice spacing a a if the lattice correlator is O ⁡ ( a ) O(a) -improved. In practical computations, lattices have necessarily a finite Euclidean temporal extent, a lack of knowledge which suggests to introduce incomplete integral transforms and the corresponding incomplete smeared spectral densities. The contribution from the unknowns to a smeared spectral density can then be rigorously bound and kept under control if the integral transform of the smearing function decays fast enough with the conjugate variable. Conversely, the bound can be used to plan lattices so as to achieve a given target precision on the reconstructed spectral density of interest. The formulae presented here in the context of lattice field theory can be easily applied or extended to other areas of research. 1 Introduction Spectral densities play a central rôle in field theories. They link dynamical properties of a theory to observables that can be measured in experiments. In the presence of strongly interacting fields, such as in Quantum Chromodynamics (QCD), non-perturbative computations of spectral densities from first principles are essential for a detailed comprehension of the dynamics of the theory, e.g. its particle and resonance content and their properties, and for the comparison against experimental results. Smeared spectral densities from physics-motivated kernels are interesting as well. These classes of observables give access to inclusive hadronic cross sections, semi-leptonic decay rates, or non-static properties of the quark-gluon plasma, to name a few (see for instance Refs. 45 ; 51 ; 26 ; 20 ; 30 ; 41 ). Correlation functions of (composite) fields in Euclidean space-time are the primary quantities that are computed non-perturbatively on the lattice by Monte Carlo evaluations of path-integrals. Thanks to the Källén–Lehmann decomposition, two-point correlators are related to the corresponding spectral densities by a Laplace transform. The direct extraction of (smeared) spectral densities from lattice data thus entails the Inverse Laplace Transform (ILT) of correlators, a notorious difficult inverse problem. The conceptual, algorithmic and technological progress in lattice QCD, however, unlocked unprecedented levels of precision in modern numerical calculations 7 ; 5 . Correlation functions at larger and larger time-distances are becoming accessible thanks to more efficient statistical estimators and to the development of multi-level integration techniques in the presence of fermions 13 ; 14 ; 15 . Larger and larger lattices with a denser and denser spectrum of states are being simulated in master-field simulations 38 ; 22 ; 19 . Therefore, it has become interesting and timely to reconsider the direct extraction of (smeared) spectral densities from lattice data 26 ; 25 ; 8 . Recently, we have proposed explicit analytic formulae to extract smeared spectral densities from the Euclidean time-dependence of correlation functions 10 . By making extensive use of integral transforms, the aim of this paper is to generalize and extend these results so as to be able to extract regulated and/or smeared spectral densities from either the continuous or discrete sampling of correlation functions in the Euclidean time. We pay particular attention to discretization effects and to rigorously bind the systematic uncertainties due to the finite extension of the time direction. The practical implementation of the strategy proposed here, and its first application to real lattice QCD data generated by a multi-level Monte Carlo will be presented in full details in a companion paper 21 . 2 Spectral densities from integral transforms In this paper we focus on two-point connected correlation functions in Euclidean space-time 1 1 1 In the main sections of this paper we assume the time-momentum representation of correlators, omitting the momentum dependence from the notation, and consider the zero momentum case. Our results, however, are fully generic and applicable to non-zero momentum correlators defined in Appendix A . Note also that our derivation applies straightforwardly to higher n -point functions, involving multiple Euclidean-time separations 11 ; 44 . C(t)=\int\,\langle{\mathcal{O}}_{1}(t,\mathbf{x}){\mathcal{O}}_{2}(0,\mathbf{0})\rangle_{\rm c}\;d^{3}\mathbf{x}\,, (1) where {\mathcal{O}}_{1} and {\mathcal{O}}_{2} are two renormalized scalar fields 2 2 2 Generalization to non-zero spin can be found in many textbooks, see for instance Ref. 51 . . Thanks to the Källén–Lehmann spectral decomposition, see Appendix A for the notation, two-point correlators are related to the corresponding spectral densities \rho(\omega) by the Laplace transform C(t)=\int_{0}^{\infty}\rho(\omega)\,e^{-\omega t}\,d\omega\,. (2) The spectral density is a function of the energy \omega , with support over [\omega_{0}\!>\!0,\infty) in channels with a mass gap \omega_{0} , which encodes the dynamical information of the theory in the channel identified by the quantum numbers of the fields in Eq. ( 1 ). When combined with the short-distance (operator product expansion) analysis, these properties fix the asymptotic behaviour of correlation functions at the origin and at infinity to be \lim_{t\rightarrow 0^{+}}C(t)=O(t^{-p})\,,\qquad\lim_{t\rightarrow+\infty}C(t)=O(e^{-\omega_{0}t})\,, (3) where p is close to an integer in QCD. This, together with Eq. ( 2 ), in turn implies that \rho(\omega)\Big|_{\omega\in[0,\omega_{0})}=0\,,\qquad\lim_{\omega\rightarrow+\infty}\rho(\omega)=O(\omega^{p-1})\,. (4) To determine the spectral density from the correlator, the Eq. ( 2 ) has to be inverted, or equivalently the ILT of the correlator has to be computed. For the clarity of the presentation, in the first part of this section we present a rather general scheme of how integral transforms can lead to the extraction of the spectral density. The underlying assumptions and the details of the particular transforms (e.g. basis functions, domain of integration, etc.) are fully spelled out in the following subsections. For the time being, given the asymptotic behaviours in Eqs. ( 3 ) and ( 4 ), let us assume that the correlator and the spectral density in Eq. ( 2 ) admit integral transforms, e.g. one of those in Appendix C , defined as {\overline{\rm C}}_{s}=\int C(t)\,\bar{\rm u}_{s}(t)\,dt\,,\qquad\hat{\uprho}_{s}=\int\rho(\omega)\,\hat{\rm u}_{s}(\omega)\,d\omega\,, (5) and their inverses, and that the basis functions \bar{\rm u}_{s}(t) and \hat{\rm u}_{s}(\omega) are related by \text{\char 21\relax}_{s}\,\hat{\rm u}_{s}(\omega)=\int e^{-\omega t}\,\bar{\rm u}_{s}(t)\,dt\,, (6) where \text{\char 21\relax}_{s} is a numerical coefficient and s is the real conjugate variable, see below. By inserting Eq. ( 6 ) into the second of Eqs. ( 5 ), and then by using Eq. ( 2 ), it immediately follows that {\overline{\rm C}}_{s}=\text{\char 21\relax}_{s}\,\hat{\uprho}_{s}\,. (7) As a result, the basic integral equation ( 2 ) becomes an algebraic relation between the integral transforms of the correlator and of the spectral density provided \text{\char 21\relax}_{s}\neq 0 for all s . This is the main advantage for considering integral transforms in this context. If we require the orthogonality condition \int\hat{\rm u}^{*}_{s}(\omega)\,\hat{\rm u}_{s^{\prime}}(\omega)\,d\omega=\delta(s-s^{\prime})\,, (8) then Eq. ( 6 ) leads to \int\int\bar{\rm u}^{*}_{s}(t)\,A(t+t^{\prime})\,\bar{\rm u}_{s^{\prime}}(t^{\prime})\,dt\,dt^{\prime}=|\text{\char 21\relax}_{s}|^{2}\delta(s-s^{\prime})\,, (9) where A(t+t^{\prime}) is the kernel of a (quasi) Carleman

Read the complete text