Mass estimates of the SU(2) $0^{++}$ glueball from spectral methods

Basic Information

• Paper ID: 2112.06785
• Title: Mass estimates of the SU(2) $0^{++}$ glueball from spectral methods
• Authors: David Dudal (KU Leuven & Ghent University), Orlando Oliveira (University of Coimbra), Martin Roelfs (KU Leuven)
• Classification: hep-lat (lattice field theory), hep-ph (phenomenology)
• Publication Time/Conference: The 38th International Symposium on Lattice Field Theory, LATTICE2021 (July 26-30, 2021)
• Paper Link: https://arxiv.org/abs/2112.06785

Abstract

This paper proposes a method for estimating the Källén-Lehmann spectral density from gauge-invariant lattice QCD two-point correlation functions through an inversion strategy based on Tikhonov regularization. The authors test the method on a meson toy model, demonstrating its competitiveness with the traditional maximum entropy method. As a proof of concept, the spectrum of SU(2) glueballs with quantum numbers $J^{PC}=0^{++}$ is investigated using data at different lattice spacings. The ground state mass estimates are consistent with traditional methods based on the large-time exponential decay behavior of correlation functions, and the spectral density also reveals signatures of excited states.

Research Background and Motivation

1. Problem to be Addressed: Traditional lattice QCD particle mass calculations rely on analyzing the large-time decay behavior of two-point correlation functions. This paper proposes direct inversion of the Källén-Lehmann spectral representation to extract particle mass information.
2. Importance of the Problem:
   • Glueballs are bound states in pure gauge theory, crucial for understanding non-perturbative properties of QCD
   • The scalar glueball ($0^{++}$) is the lightest glueball state with important theoretical and experimental significance
   • Provides an alternative analysis method independent of smearing techniques
3. Limitations of Existing Methods:
   • Traditional methods require smearing techniques to improve signal-to-noise ratio
   • Depend on exponential fitting of large-time behavior, potentially missing excited state information
   • Maximum entropy method, while widely used, may lack stability in certain cases
4. Research Motivation: Develop a method to directly extract spectral information from momentum-space propagators, avoiding technical complexity of traditional methods while enabling excited state identification.

Core Contributions

1. Proposed a spectral density inversion method based on Tikhonov regularization, combined with non-negative least squares (NNLS) to ensure physically reasonable positive spectral density
2. Validated the method's effectiveness on a meson toy model, demonstrating comparable performance with the maximum entropy method
3. Successfully applied to SU(2) $0^{++}$ glueball mass estimation, obtaining ground state masses consistent with traditional methods across multiple lattice spacings
4. Identified the presence of excited states, providing additional information for glueball spectroscopy
5. Provided an analysis pathway without smearing techniques, simplifying the lattice QCD analysis procedure

Detailed Methodology

Task Definition

Extract the Källén-Lehmann spectral density $\rho(\omega)$ from lattice QCD two-point correlation functions $G(p)$, thereby determining the mass spectrum of glueballs. Input: momentum-space correlation function data; Output: spectral density function in frequency domain.

Model Architecture

1. Källén-Lehmann Spectral Representation

The relationship between correlation function and spectral density is: $G(p) = \int_0^{\infty} \frac{2\omega\rho(\omega) d\omega}{\omega^2 + p^2}$

2. Subtraction Scheme Treatment

To handle divergences, a subtraction is introduced: $G(p^2) = \sum_{k=0}^{n-1} a_k (p^2 - \bar{p}^2)^k + (-p^2 + \bar{p}^2)^n \int_0^{\infty} \frac{2\omega\tilde{\rho}(\omega) d\omega}{\omega^2 + p^2}$

3. Schwinger Function Method

Avoid direct numerical subtraction through Fourier transformation: $C(\tau) = \mathcal{F}\{G(p)\}(\tau) = \int_{-\infty}^{\infty} G(p^2)|_{\vec{p}=0, p_4\neq 0} e^{-ip_4\tau} dp_4$

For $\tau \neq 0$, we have $C(\tau) = \mathcal{L}\{\rho(\omega)\}(|\tau|)$, which is a Laplace transform relationship.

4. Tikhonov Regularization Inversion

Convert the Laplace inversion problem into a regularized least squares problem: $J_\alpha = \|\mathbf{K}\boldsymbol{\rho} - \mathbf{C}\|_2^2 + \alpha^2\|\boldsymbol{\rho} - \boldsymbol{\rho}^*\|_2^2$

where $\mathbf{K}_{k\ell} = e^{-\omega_\ell \tau_k}\Delta\omega$, and $\boldsymbol{\rho}^*$ is a prior estimate.

5. Non-negativity Constraint

Ensure $\rho(\omega) \geq 0$ through NNLS solver: $\mathbf{A} = \begin{pmatrix} \mathbf{K} \\ \alpha\mathbf{1} \end{pmatrix}, \quad \mathbf{b} = \begin{pmatrix} \mathbf{C} \\ \alpha\boldsymbol{\rho}^* \end{pmatrix}$

Minimize $J_\alpha = \|\mathbf{A}\boldsymbol{\rho} - \mathbf{b}\|_2^2$ subject to $\boldsymbol{\rho} \geq 0$.

Technical Innovations

1. Avoids Numerical Subtraction: Cleverly circumvents direct numerical subtraction through the Schwinger function method
2. Physical Constraints: Enforces positivity constraints to ensure physical reasonableness of spectral density
3. Parameter Selection: Employs Morozov discrepancy principle for automatic selection of regularization parameter $\alpha$
4. Logarithmic Sampling: Uses logarithmically uniform sampling in frequency domain, covering wide frequency range $[10^{-5}, 10^5]$ GeV

Experimental Setup

Dataset

Uses SU(2) pure Yang-Mills theory data published by Yamanaka et al. Phys. Rev. D 102, 054507 (2020):

Evaluation Metrics

• Accuracy of ground state mass estimates (comparison with traditional methods)
• Physical reasonableness of spectral density function (positivity, mass gap, etc.)
• Capability for excited state identification

Comparison Methods

• Traditional large-time exponential decay analysis method
• Maximum entropy method (in toy model testing)

Implementation Details

• Number of frequency sampling points: $N_\omega = 1000$
• Frequency range: $[10^{-5}, 10^5]$ GeV (logarithmically uniform distribution)
• Regularization parameter determined by Morozov discrepancy principle
• Uses raw unsmeared data

Experimental Results

Main Results

1. Toy Model Validation

Testing on the meson decay model shows that mass reconstruction improves with increased data points and reduced noise. The method performs comparably to the maximum entropy method.

2. SU(2) $0^{++}$ Glueball Mass Estimates

Ground state mass estimates are highly consistent with traditional methods, with differences within 5% in most cases.

3. Excited State Identification

Except for $\beta = 2.4$, all conditions show evidence of a first excited state in the 3.3-3.9 GeV range:

Ablation Studies

• Without Positivity Constraint: Produces significant negative oscillations and overfitting
• With Positivity Constraint: Eliminates infrared oscillations, producing physically reasonable spectral density

Experimental Findings

1. Mass Gap: All spectral densities show a clear mass gap corresponding to approximately 1.4 GeV ground state
2. Excited State Structure: Excited state peaks are relatively broad, potentially requiring better signal quality
3. Lattice Effects: Anomalous behavior at $\beta = 2.4$ may be related to UV cutoff from finite lattice spacing

Related Work

Major Research Directions in the Field

1. Lattice QCD Glueball Spectroscopy: Using traditional correlation function analysis methods
2. Spectral Function Extraction: Maximum entropy method, Bayesian inference, etc.
3. Källén-Lehmann Representation: Applications in lattice field theory

Relationship to Related Work

• Builds upon the foundation of lattice QCD glueball research, particularly work by Yamanaka et al.
• Extends spectral function extraction methods, providing Tikhonov regularization as an alternative approach
• Complements the maximum entropy method

Advantages of This Work

• Avoids dependence on smearing techniques in traditional methods
• Provides excited state information
• Relatively straightforward numerical implementation

Conclusions and Discussion

Main Conclusions

1. Method Effectiveness: The spectral density extraction method based on Tikhonov regularization performs well in glueball mass estimation
2. Result Consistency: Ground state mass estimates are highly consistent with traditional methods
3. Excited State Information: Successfully identifies excited state structure, providing additional insights for glueball spectroscopy
4. Technical Advantages: Eliminates need for smearing techniques, simplifying the analysis procedure

Limitations

1. Data Quality Dependence: Method is sensitive to statistical errors, requiring high-quality data
2. Excited State Uncertainty: Excited state peaks are relatively broad with significant position variation
3. Finite Volume Effects: Impact of finite lattice volume not sufficiently discussed
4. UV Cutoff Effects: Effects of lattice spacing require further investigation

Future Directions

1. Extension to glueball states with other quantum numbers
2. Application in SU(3) gauge theory
3. Cases including dynamical fermions
4. Techniques to improve excited state resolution

In-Depth Evaluation

Strengths

1. Methodological Innovation: Provides an effective alternative to traditional lattice analysis
2. Technical Rigor: Correctly addresses the ill-posed nature of numerical inversion
3. Comprehensive Validation: Dual verification through toy model and actual data
4. Physical Insights: Provides excited state information, enriching understanding of glueball spectrum
5. Practical Value: Avoids smearing techniques, simplifying analysis procedure

Weaknesses

1. Insufficient Error Analysis: Statistical error analysis of spectral density is relatively simple
2. Systematic Errors: Systematic errors introduced by the method itself not sufficiently discussed
3. Parameter Sensitivity: Sensitivity analysis to regularization parameter selection lacks depth
4. Physical Interpretation: Physical interpretation of excited state peak widths needs strengthening

Impact

1. Theoretical Contribution: Provides new tools for lattice QCD spectroscopy analysis
2. Methodological Value: Generalizable to other hadron spectroscopy studies
3. Technical Impact: Successful application of Tikhonov regularization in lattice field theory

Applicable Scenarios

• Analysis of high-statistics lattice QCD data
• Spectroscopy studies requiring excited state information
• Cases where smearing techniques are inconvenient
• Studies of glueballs and other pure gauge states

References

This paper is primarily based on the following important works:

1. Yamanaka et al., Phys. Rev. D 102, 054507 (2020) - Provides SU(2) glueball data
2. Asakawa et al., Prog. Part. Nucl. Phys. 46, 459 (2001) - Maximum entropy method review
3. Dudal et al., Nucl. Phys. B 952, 114912 (2020) - Prior work on spectral representation methods

This paper provides a valuable new method in the field of lattice QCD glueball spectroscopy. While there is room for improvement in certain technical details, its basic approach and preliminary results are promising, laying a solid foundation for further development in this field.