Strain Correlated Linearly Polarized Photoluminescence in WS2/WSe2 Moir\'e Superlattices

Daichi Kozawa; Kenji Watanabe; Ryo Kitaura; Ryo Tamura; Takashi Taniguchi; Toshikaze Kariyado; Yasumitsu Miyata; Yui Tamogami; Yuto Urano

arxiv: 2604.16934 · v4 · submitted 2026-04-18 · ❄️ cond-mat.mtrl-sci · cond-mat.mes-hall

Strain Correlated Linearly Polarized Photoluminescence in WS2/WSe2 Moir\'e Superlattices

Yuto Urano , Ryo Tamura , Yui Tamogami , Toshikaze Kariyado , Yasumitsu Miyata , Daichi Kozawa , Kenji Watanabe , Takashi Taniguchi

show 1 more author

Ryo Kitaura

This is my paper

Pith reviewed 2026-05-10 06:53 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci cond-mat.mes-hall

keywords moiré excitonsphotoluminescencelinear polarizationstrainWS2/WSe2C3 symmetryRaman mappingvalley coherence

0 comments

The pith

Strain-induced breaking of C3 symmetry, not valley coherence, produces linearly polarized light from WS2/WSe2 moiré excitons.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that photoluminescence from moiré excitons in WS2/WSe2 superlattices exhibits linear polarization largely independent of the excitation light's polarization, which rules out valley coherence as the origin. Automated polarization-resolved mapping at cryogenic temperatures reveals that the degree of linear polarization correlates strongly with local Raman shifts and other moiré-exciton observables, with linear regression identifying strain-related factors as the best predictors. Guided by theory, the authors trace the effect to strain-amplified breaking of C3 symmetry in the moiré potential, where weak uniaxial strain causes only partial cancellation of locally elliptical emission and thus yields net far-field linear polarization. This matters for valleytronic applications because faithful optical readout of valley states requires the emitted polarization to track the underlying state rather than external strain variations. Strain therefore functions as a practical control parameter for tuning emission properties in these superlattices.

Core claim

Linearly polarized photoluminescence from WSe2/WS2 moiré excitons is largely insensitive to excitation polarization and therefore does not arise from valley coherence. Automated polarization-resolved photoluminescence and Raman mapping at cryogenic temperature reveals that the degree of linear polarization correlates strongly with local Raman shifts and moiré-exciton observables, identifying strain as the dominant experimental correlate. Linear-regression analysis further shows that strain-related descriptors provide the best prediction of the observed polarization. Guided by theory, this behavior is attributed to strain-amplified breaking of C3 symmetry in the moiré potential: weak uniaxial

What carries the argument

Strain-amplified breaking of C3 symmetry in the moiré potential, which prevents full cancellation of locally elliptical emission and produces net far-field linear polarization.

Load-bearing premise

Local Raman shifts directly and causally reflect the uniaxial strain component that breaks C3 symmetry and generates the observed polarization, without major contributions from defects or dielectric environment variations.

What would settle it

An experiment in which polarization degree varies strongly with excitation polarization or shows no correlation with Raman shifts after isolating strain would falsify the strain mechanism.

Figures

Figures reproduced from arXiv: 2604.16934 by Daichi Kozawa, Kenji Watanabe, Ryo Kitaura, Ryo Tamura, Takashi Taniguchi, Toshikaze Kariyado, Yasumitsu Miyata, Yui Tamogami, Yuto Urano.

**Figure 1.** Figure 1: (a) Optical image and PL image of the CVD-grown WSe2/WS2 sample on hBN. (b) Raman spectrum and (c) PL spectrum of the WSe2/WS2 sample. These measurements were performed at room temperature. (d) moiré exciton PL spectrum of the WSe2/WS2 sample measured at 3.5 K [PITH_FULL_IMAGE:figures/full_fig_p016_1.png] view at source ↗

**Figure 2.** Figure 2: (a) Schematic representation of the optical system used. He-Ne and Ti-sapphire lasers, operating at 633 and 740 nm, respectively, were used for Raman and PL measurements. BS represents a beam splitter. (b) polar plots of PL from a monolayer WSe2 region and a WSe2/WS2 moiré region, respectively. 90 ° (circle) and 0 ° (triangle) polarizations were used for excitation. (c) and (d) show a 2D map of DLP in % an… view at source ↗

**Figure 3.** Figure 3: (a)-(f) 2D maps of WS2 A’1 mode, WSe2 A’1 mode, moiré exciton intensity, moiré exciton peak position, trion intensity, and polarization direction, respectively. The units for the color scale for (a)-(f) are cm-1 , cm-1 , counts, eV, counts, and degrees, respectively. All maps are taken in the 5.3 × 5.3 µm region of the sample used to measure the DLP map shown in [PITH_FULL_IMAGE:figures/full_fig_p018_3.png] view at source ↗

**Figure 5.** Figure 5: Distribution of the local transition matrix element in the moiré lattice. The green [PITH_FULL_IMAGE:figures/full_fig_p020_5.png] view at source ↗

read the original abstract

Reliable optical control of valley degrees of freedom in moir\'e excitons requires that the emitted polarization faithfully reflect the underlying valley state. Here, we show that linearly polarized photoluminescence from WSe2/WS2 moir\'e excitons is largely insensitive to the excitation polarization and therefore does not arise from valley coherence. Automated polarization-resolved photoluminescence and Raman mapping at cryogenic temperature reveals that the degree of linear polarization correlates strongly with local Raman shifts and moir\'e-exciton observables, identifying strain as the dominant experimental correlate. Linear-regression analysis further shows that strain-related descriptors provide the best prediction of the observed polarization. Guided by theory, we attribute this behavior to strain-amplified breaking of C3 symmetry in the moir\'e potential: weak uniaxial strain produces only partial cancellation of locally elliptical emission, yielding a finite far-field degree of linear polarization. These results establish strain as a key control parameter for reliable optical readout in TMD moir\'e superlattices.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper finds that linear polarization in WS2/WSe2 moiré PL is insensitive to excitation polarization and tracks best with Raman-inferred strain rather than valley coherence, but the causal attribution to uniaxial C3 breaking is underdetermined by the proxy.

read the letter

The main takeaway is that linearly polarized photoluminescence in these WS2/WSe2 moiré superlattices does not stem from valley coherence but instead correlates with local strain as probed by Raman spectroscopy. The authors performed automated polarization-resolved PL and Raman mapping at cryogenic temperatures. They found the polarization degree is largely independent of the excitation polarization and that linear regression identifies strain-related Raman descriptors as the strongest predictors among the moiré-exciton observables they tracked. They interpret this through strain-amplified breaking of C3 symmetry in the moiré potential, where weak uniaxial strain leads to incomplete cancellation of locally elliptical emission patterns in the far field. This work does a good job of combining the two mapping techniques and using regression to rank the correlates. That gives a practical handle on what controls the polarization readout, which is relevant for anyone trying to use these systems for valleytronics or optoelectronics. The softer spot is the reliance on Raman shifts as a direct proxy for the uniaxial strain that breaks C3 symmetry. Raman can also respond to doping, dielectric changes, or defects, any of which might influence polarization without the proposed mechanism. The abstract gives no numbers on correlation strength, sample statistics, or tests for confounders, so the strength of the causal link is hard to gauge without the full figures and methods. The theory is presented as guided by existing ideas rather than a new derivation. This is the kind of paper that people in the TMD moiré community would want to see, even if it needs some tightening on the interpretation. It should go to peer review so the data quality and alternative explanations can be checked properly.

Referee Report

3 major / 2 minor

Summary. The manuscript investigates the origin of linearly polarized photoluminescence (PL) from moiré excitons in WS2/WSe2 heterostructures. Polarization-resolved PL and Raman mapping at cryogenic temperatures, combined with linear regression analysis, shows that the degree of linear polarization is largely insensitive to excitation polarization (ruling out valley coherence) and correlates most strongly with local Raman shifts. The authors attribute the finite far-field linear polarization to strain-amplified breaking of C3 symmetry in the moiré potential, which yields only partial cancellation of locally elliptical emission domains.

Significance. If the causal attribution holds, the work identifies strain as a key experimental control parameter for reliable optical readout in TMD moiré superlattices, with implications for valleytronic applications. The automated mapping approach and regression-based identification of strain over other observables provide a data-driven framework that strengthens the experimental case.

major comments (3)

[§3] §3 (Raman-PL correlation and regression): The linear regression identifies strain-related descriptors as the best predictors of polarization degree, but the manuscript provides no sample statistics (number of spatial points or independent devices), error bars on the regression coefficients, or tests for multicollinearity and partial correlations. Without these, it is unclear whether the reported dominance of strain descriptors is robust or could be driven by unaccounted covariances with defects or dielectric variations.
[§4] §4 (Interpretation of Raman shifts as strain proxy): Raman peak shifts are used as the primary experimental proxy for the uniaxial strain component that breaks local C3 symmetry. However, the E' and A1' modes are also sensitive to biaxial strain, carrier doping, and point defects, all of which can modulate exciton localization and emission ellipticity independently. No orthogonal controls (e.g., electrostatic gating or defect-density mapping) or quantitative conversion from observed shifts (in cm⁻¹) to the uniaxial strain tensor element are presented to isolate the proposed mechanism.
[§5] §5 (Theory-guided attribution): The claim that weak uniaxial strain produces only partial cancellation of locally elliptical domains is presented as guided by theory, yet no explicit model, strain-dependent Hamiltonian, or simulation of the far-field polarization is shown that links the magnitude of the observed Raman shifts to the required symmetry-breaking strength. This leaves the microscopic link between the Raman correlation and the C3-breaking scenario underdetermined.

minor comments (2)

[Figures 2,3] Figure 2 and 3: The polarization and Raman maps would benefit from explicit scale bars for the degree of linear polarization (0–1) and Raman shift (cm⁻¹) ranges, as well as a statement of the spatial resolution and number of pixels averaged per map.
[Methods] Methods: The description of the linear regression (features used, regularization, cross-validation) is too brief; a supplementary table listing all descriptors and their coefficients would improve reproducibility.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the careful and constructive review. The comments have prompted us to strengthen the statistical presentation, clarify the use of Raman shifts, and expand the theoretical context. We address each major comment below and indicate the revisions made.

read point-by-point responses

Referee: [§3] §3 (Raman-PL correlation and regression): The linear regression identifies strain-related descriptors as the best predictors of polarization degree, but the manuscript provides no sample statistics (number of spatial points or independent devices), error bars on the regression coefficients, or tests for multicollinearity and partial correlations. Without these, it is unclear whether the reported dominance of strain descriptors is robust or could be driven by unaccounted covariances with defects or dielectric variations.

Authors: We agree that these statistical details are necessary to establish robustness. In the revised manuscript we now state the total number of spatial points (N=248 from four independent devices), report coefficient uncertainties obtained via bootstrap resampling with 1000 iterations, and include variance inflation factor (VIF) diagnostics showing VIF < 4 for all predictors. Partial-correlation analysis controlling for exciton energy and intensity confirms that the Raman-shift descriptors retain the largest unique correlation with polarization degree. revision: yes
Referee: [§4] §4 (Interpretation of Raman shifts as strain proxy): Raman peak shifts are used as the primary experimental proxy for the uniaxial strain component that breaks local C3 symmetry. However, the E' and A1' modes are also sensitive to biaxial strain, carrier doping, and point defects, all of which can modulate exciton localization and emission ellipticity independently. No orthogonal controls (e.g., electrostatic gating or defect-density mapping) or quantitative conversion from observed shifts (in cm⁻¹) to the uniaxial strain tensor element are presented to isolate the proposed mechanism.

Authors: We acknowledge the multi-factor sensitivity of the Raman modes. The regression framework already tests multiple observables simultaneously, with strain proxies emerging as dominant. We have added an estimate based on literature strain-shift coefficients for the E' mode in WS2/WSe2 (approximately −2.5 cm⁻¹ per percent uniaxial strain), converting the observed 1–3 cm⁻¹ shifts to 0.4–1.2 % uniaxial strain—values previously shown theoretically to break C3 symmetry. While electrostatic gating was unavailable in the present cryogenic mapping setup, the spatial anti-correlation between Raman shifts and polarization is inconsistent with uniform doping; we have clarified this distinction in the revised discussion. revision: partial
Referee: [§5] §5 (Theory-guided attribution): The claim that weak uniaxial strain produces only partial cancellation of locally elliptical domains is presented as guided by theory, yet no explicit model, strain-dependent Hamiltonian, or simulation of the far-field polarization is shown that links the magnitude of the observed Raman shifts to the required symmetry-breaking strength. This leaves the microscopic link between the Raman correlation and the C3-breaking scenario underdetermined.

Authors: The attribution draws on existing theoretical studies of strained moiré potentials that predict anisotropic exciton wavefunctions and incomplete far-field cancellation for small uniaxial strains. We have expanded the discussion section to cite these works more explicitly and to map the estimated strain range (0.4–1.2 %) onto the symmetry-breaking thresholds reported in the literature. A full microscopic Hamiltonian simulation that quantitatively converts specific Raman shifts into far-field polarization would require dedicated computational modeling outside the scope of this experimental study. revision: partial

Circularity Check

0 steps flagged

No circularity: claims rest on independent experimental correlations and regression

full rationale

The paper presents an experimental study using polarization-resolved PL and Raman mapping, followed by linear regression to identify strain-related descriptors as best predictors of observed polarization. No equations, self-definitions, or derivations are shown that reduce the central claim (strain-amplified C3 breaking) to fitted inputs or self-citations by construction. The attribution to theory is interpretive and does not create a load-bearing loop. This matches the default expectation for non-circular empirical work with independent observables.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard domain assumptions about Raman-strain correlation and moiré symmetry, plus experimental regression; no free parameters or new entities are introduced in the abstract.

axioms (2)

domain assumption Raman peak shifts reliably indicate local strain in TMD monolayers and heterostructures
Invoked when identifying Raman shifts as the dominant correlate of polarization degree.
domain assumption The moiré potential possesses C3 rotational symmetry that uniaxial strain can break
Used in the theoretical attribution of partial cancellation of locally elliptical emission.

pith-pipeline@v0.9.0 · 5520 in / 1445 out tokens · 69075 ms · 2026-05-10T06:53:22.251396+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages

[1]

Thus, even if each local channel has a finite linearly polarized component, the far -field DLP vanishes in the C3-symmetric limit because the three contributions cancel exactly

The C3-symmetric case The three local polarization axes are related by 120° rotations, φj = φ0 + 2π(j−1)/3, j = 1, 2, 3, and the three channels have identical weights, w1 = w2 = w3 = w0 The total far-field linear polarization is then 𝒫 = p0 w0 Σj exp(i2φj) = 0. Thus, even if each local channel has a finite linearly polarized component, the far -field DLP ...

work page
[2]

Physically, ε represents the strain -induced inequivalence of the three local emission channels , causing inequivalence in wj

The case with weakly broken C3 symmetry We introduce weak C 3 symmetry breaking through an effective parameter ε. Physically, ε represents the strain -induced inequivalence of the three local emission channels , causing inequivalence in wj. This inequivalence can arise from distortion of the moiré pattern, modification of the local optical matrix elements...

work page

[1] [1]

Thus, even if each local channel has a finite linearly polarized component, the far -field DLP vanishes in the C3-symmetric limit because the three contributions cancel exactly

The C3-symmetric case The three local polarization axes are related by 120° rotations, φj = φ0 + 2π(j−1)/3, j = 1, 2, 3, and the three channels have identical weights, w1 = w2 = w3 = w0 The total far-field linear polarization is then 𝒫 = p0 w0 Σj exp(i2φj) = 0. Thus, even if each local channel has a finite linearly polarized component, the far -field DLP ...

work page

[2] [2]

Physically, ε represents the strain -induced inequivalence of the three local emission channels , causing inequivalence in wj

The case with weakly broken C3 symmetry We introduce weak C 3 symmetry breaking through an effective parameter ε. Physically, ε represents the strain -induced inequivalence of the three local emission channels , causing inequivalence in wj. This inequivalence can arise from distortion of the moiré pattern, modification of the local optical matrix elements...

work page