Quantitative Benchmarking of Spectroscopic Homogeneous and Inhomogeneous Linewidth Separation

arXiv:2607.05564 · physics.optics, cond-mat.mes-hall, physics.chem-ph · Submitted 2026-07-06 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: Today's paper: "Quantitative Benchmarking of Spectroscopic Homogeneous and Inhomogeneous Linewidth Separation".

Mira: Separating contributions from homogeneous dephasing and inhomogeneous broadening in spectral linewidths is essential for connecting optical spectra to microscopic dissipation and disorder mechanisms.

Kai: First, who's behind it and why it matters.

Paper summary: Kai: So to summarize what this paper is about, they are addressing a real difficulty in connecting optical spectra to microscopic dissipation and disorder by trying to separate homogeneous dephasing from inhomogeneous broadening.

Mira: The central thesis is that standard linear Voigt fits applied to 1D spectra yield results where the Gaussian and Lorentzian widths are so correlated that you can't independently determine either one with certainty <ref:2607.05564#pg0>.

Lev: That difficulty in separation means that when we try to model noise for quantum hardware, our models are built on shaky assumptions because we're mixing up two fundamentally different types of physical effects.

Kai: The authors claim their 2DCS method solves this degeneracy by using orthogonal spectral slices—diagonal and cross-diagonal—which each provide unique sensitivity to the underlying broadening mechanisms <ref:2607.05564#pg0,orthogonal spectral slices—diagonal and>.

Mira: They demonstrate that when you perform a joint fit on these 2DCS data, the resulting parameters for sigma and gamma are significantly less correlated than what is found from fitting just one PL spectrum linearly <ref:2607.05564#pg0>.

Lev: This is significant because it means that theoretical models describing noise in real quantum hardware can be parameterized with much greater precision because the parameters aren't being artificially constrained by the measurement technique itself.

Kai: The paper uses quantitative benchmarks to show exactly how much this degeneracy is reduced, providing metrics like Pearson correlation coefficients and axis ratios to compare different spectroscopic methods.

Mira: The paper shows that for instance, linear PL analysis results in a correlation of zero point nine three plus or minus zero point zero one, while their joint 2DCS analysis brings that down to about zero point seven three plus or minus zero point zero three <ref:2607.05564#pg1>.

Lev: That difference in the correlation coefficient is what matters for us on the ground; a lower correlation means less uncertainty in the parameters we use to design error correction protocols for systems exhibiting this kind of noise.

Kai: Essentially, they are showing that 2DCS provides an extra layer of information, giving us a more robust way to separate these two components that standard methods miss <ref:2607.05564#pg0>.

Mira: The paper's importance lies in providing a practical tool for experimentalists and theorists who need to move beyond the limitations imposed by simple one-dimensional spectral analysis when characterizing system noise.

Lev: If this method works as claimed, it opens up a path for more reliable characterization of noise in actual quantum devices because we get better constraints on the physical parameters driving that noise.

Conclusion: Kai: Wrapping up this discussion on the "Quantitative Benchmarking of Spectroscopic Homogeneous and Inhomogeneous Linewidth Separation," I think it boils down to how we interpret the results from these different spectroscopic techniques.

Mira: Exactly, Kai; the authors, Adam Alfrey, Cole Tait, Joshua Hendrickson, and Steven T. Cundiff, are demonstrating that the choice of measurement technique matters when you're trying to extract fundamental physical parameters like dephasing rate and disorder strength.

Lev: From a hardware perspective, this means that if we are building systems where these noise sources are relevant—say in solid-state quantum emitters—we need to employ techniques like 2DCS rather than relying solely on simpler 1D measurements <ref:2607.05564#pg0>.

Kai: The implication is that for experimentalists, the paper suggests moving towards multi-dimensional data acquisition when characterizing spectra if they want reliable estimates of noise components.

Mira: And for theorists, it validates that incorporating higher-dimensional spectroscopic information into your models will lead to parameter estimates with much tighter confidence regions rather than just broad, uncertain ones.

Lev: That precision in parameter estimation is what we need to move from theoretical predictions about noise characteristics to actually designing functional error correction protocols that work reliably on physical hardware.

Kai: So, in simple terms, this paper shows that 2DCS gives us a better handle on the noise sources because it doesn't get stuck in the same confusing parameter trade-off that linear fits do <ref:2607.05564#pg0>.

Mira: That’s right; by providing those orthogonal constraints, they move away from relying on highly correlated estimates and toward having more compact and less elongated confidence regions for those important physical parameters.

Lev: For quantum error correction researchers, this means we can start designing codes based on noise models that are much more constrained and less prone to catastrophic failure because the underlying noise parameters have been better quantified.

Kai: It's a practical methodological improvement for characterizing real systems, showing how to get cleaner separation of the two processes at play.

Department of Physics, University of Michigan · Air Force Research Laboratory · Quantum Research Institute, University of Michigan

physics.optics, cond-mat.mes-hall, physics.chem-ph

Submitted: 2026-07-06

Updated: 2026-10-07

License: http://creativecommons.org/licenses/by-nc-nd/4.0/

Importance score: 92/100

The gist: Separating contributions from homogeneous dephasing and inhomogeneous broadening in spectral linewidths is essential for connecting optical spectra to microscopic dissipation and disorder mechanisms.

Key concepts

Homogeneous Broadening ($\gamma$)
This represents the intrinsic dephasing rate of a spectral line, caused by microscopic dissipation within the system. It is modeled as a Lorentzian width in the spectrum. A higher $\gamma$ indicates faster loss of phase coherence in the system.
Inhomogeneous Broadening ($\sigma$)
This describes the distribution of local environments or disorder within an ensemble, causing lines to have different natural widths. It is modeled as a Gaussian width. A larger $\sigma$ means a wider spread of energy levels due to variations in the local environment.
Voigt Profile
The Voigt profile is the mathematical function that describes the resulting spectral line shape when homogeneous broadening (Lorentzian) and inhomogeneous broadening (Gaussian) are combined through convolution. It is essential for accurately modeling real spectroscopic data.
Two-Dimensional Coherent Spectroscopy (2DCS)
2DCS provides two orthogonal spectral slices—diagonal and cross-diagonal—of the same measurement. By simultaneously fitting both slices, researchers gain an additional independent constraint that helps decouple the effects of $\sigma$ and $\gamma$, improving parameter separation.

Terminology

Summary

Separating contributions from homogeneous dephasing and inhomogeneous broadening in spectral linewidths is essential for connecting optical spectra to microscopic dissipation and disorder mechanisms.

How it works

The paper addresses the difficulty in independently determining homogeneous dephasing rate (represented by Lorentzian width, γ) and inhomogeneous broadening parameter (represented by Gaussian width, σ) when they are both present in a spectral line shape. Standard linear Voigt fits yield strongly correlated Gaussian and Lorentzian widths, making independent determination unreliable. Two-dimensional coherent spectroscopy (2DCS) is introduced as a method to reduce this degeneracy by providing orthogonal spectral slices: diagonal and cross-diagonal.

Key Mechanisms and Modeling

The linewidth of a spectral line encodes information about coherence dynamics and disorder. When an ensemble exhibits inhomogeneous broadening, the lineshape becomes a convolution of the homogeneous response (Lorentzian) and the inhomogeneous distribution (Gaussian), resulting in a Voigt profile [1]. Specifically:

  1. Homogeneous broadening yields a Lorentzian with half-width γ (the dephasing rate).

  2. The inhomogeneous distribution is typically assumed Gaussian with width σ.

The paper models the photoluminescence (PL) spectrum as:

(1)

y(ω) = A(ω - ωX; σ, γ) + B, where the Voigt profile is a convolution (Δ; σ, γ) = ∫ G(Δ - ω'; σ) L(ω'; γ).

Comparison of Spectroscopic Methods

The study compares the constraints provided by one-dimensional PL spectroscopy versus two-dimensional coherent spectroscopy (2DCS).

  1. Linear Voigt analysis of PL spectra shows that the parameters are strongly covariant, with a Pearson correlation coefficient ρσσ = 0.93 ± 0.01, indicating that an increase in σ can be compensated by a decrease in γ at nearly constant total linewidth, leaving the homogeneous–inhomogeneous partition poorly constrained.

  2. In contrast, simultaneous fitting of diagonal and cross-diagonal slices from 2DCS yields significantly less covariance between σ and γ, resulting in smaller confidence intervals.

Quantifying Parameter Coupling

The researchers quantify the relationship between parameters using statistical metrics derived from profiling the joint uncertainty-weighted landscape:

(10)

Δχ22 (σ, γ) = min p [y i - (A cross(ω i; σ, γ) + B cross)] / ε12 + [y i - (A diag(ω i; σ, γ) + B diag)] / ε22, where the minimization is performed over nuisance parameters.

The resulting covariance matrix C and metrics are extracted:

  1. Pearson correlation coefficient ρσσ = Cσσ/√(CσσC), which is found to be substantially reduced in 2DCS (0.73 ± 0.03) compared to linear PL (0.93 ± 0.01).

  2. The eigenvalue axis ratio rr, which measures parameter coupling and confidence-ellipse elongation, is also smaller in the joint 2DCS analysis (rr = 2.6 ± 0.2) versus linear PL (rr = 5.1 ± 0.2), indicating more compact and less elongated confidence regions.

Conclusion on Applicability

The paper demonstrates that jointly fitting diagonal and cross-diagonal slices of a 2DCS measurement, where the cross-diagonal width is governed primarily by γ and the diagonal width reflects the full Voigt convolution of σ and γ, provides an additional orthogonal constraint. This approach reduces the degeneracy inherent in linear Voigt analysis, yielding confidence regions with markedly reduced correlation and elongation, confirming that claims of near-homogeneous linewidths based solely on linear fits should be interpreted with caution for TMD excitons. The method is described as general: applicable to any system whose 2DCS lineshape can be modeled.

Summary of Key Findings

(12)

The joint 2DCS analysis substantially reduces the degeneracy, giving ρσσ = 0.73 ± 0.03 and rr = 2.6 ± 0.2, i.e., more compact and less elongated confidence regions across all measured locations compared to linear Voigt analysis of PL (ρσσ = 0.93 ± 0.01, rr = 5.1 ± 0.2).

(13)

The resulting best-fit linewidth partitions also differ systematically between modalities: PL gives σ = 0.56 ± 0.07 meV and γ = 1.00 ± 0.07 meV, whereas 2DCS yields σ = 0.91 ± 0.08 meV and γ = 0.58 ± 0.

Improvements for AI systems

Based on the provided scientific paper, here are specific improvements for AI systems and what those improved systems could achieve:

  1. The paper demonstrates a method to quantitatively separate homogeneous broadening (Lorentzian, characterized by dephasing rate gamma) from inhomogeneous broadening (Gaussian, characterized by distribution width sigma) in exciton linewidths using Two-Dimensional Coherent Spectroscopy (2DCS).

  2. The core improvement is the development of a joint fitting framework that utilizes orthogonal spectral slices: the diagonal slice (sensitive to both σ and gamma via Voigt convolution) and the cross-diagonal slice (sensitive primarily to gamma).

  3. The improved AI system can perform highly accurate, statistically grounded parameter extraction from complex spectroscopic data, moving beyond the limitations of linear Voigt fitting.

The improved AI system can do the following:

  1. Perform non-degenerate decomposition of spectral linewidths in materials science experiments (e.g., photoluminescence, absorption) where both homogeneous and inhomogeneous broadening mechanisms are present and comparable in magnitude.

  2. Quantitatively assess the reliability of existing single-parameter fits (like those based on linear Voigt analysis) by calculating statistical metrics such as the Pearson correlation coefficient (rhosigmasigma) and eigenvalue axis ratio (rr).

  3. Identify when spectral features are poorly constrained due to parameter covariance, allowing researchers to distinguish between true physical linewidths and artifacts of model degeneracy.

  4. Accurately determine the fundamental microscopic parameters—the dephasing rate gamma (homogeneous broadening) and the distribution width sigma (inhomogeneous broadening)—with significantly smaller confidence regions compared to traditional methods.

  5. Distinguish between different physical mechanisms by analyzing systematic differences in extracted parameters across varying experimental conditions or sample locations, as demonstrated by comparing PL results with 2DCS results.

Abstract

Separating homogeneous and inhomogeneous contributions to a spectral linewidth is essential for connecting optical spectra to microscopic dephasing and disorder. We quantitatively compare linewidth separation by one-dimensional Voigt analysis and two-dimensional coherent spectroscopy (2DCS). Using photoluminescence and 2DCS measurements of the same hBN-encapsulated monolayer MoSe2 sample, we construct uncertainty-weighted squared landscapes for the Gaussian inhomogeneous width and Lorentzian homogeneous half-width. Although a Voigt profile accurately reproduces the photoluminescence spectrum, the two widths are strongly anticorrelated and therefore poorly determined independently. By contrast, joint fitting of diagonal and cross-diagonal 2DCS slices produces more compact confidence regions with reduced parameter correlation and elongation. These results quantitatively demonstrate the improved linewidth separation provided by 2DCS and illustrate the limitations of assigning homogeneous and inhomogeneous contributions from one-dimensional Voigt fits alone.

Sources

Related papers