arxiv: 2605.12194 · v1 · submitted 2026-05-12 · ❄️ cond-mat.mtrl-sci · cs.LG

Recognition: 1 theorem link

· Lean Theorem

Probing Non-Equilibrium Grain Boundary Dynamics with XPCS and Domain-Adaptive Machine Learning

Mouyang Cheng , Bowen Yu , Chu-Liang Fu , Nina Andrejevic , Matthias T. Agne , Riley Hanus , Qiwei Wan , Nathan C. Drucker

show 9 more authors

Thanh Nguyen Andrei Fluerasu Lutz Wiegart Xiaoqian M Chen Daniel Pajerowski Yongqiang Cheng Joshua J Turner G. Jeffrey Snyder Mingda Li

Authors on Pith no claims yet

Pith reviewed 2026-05-13 04:51 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci cs.LG

keywords grain boundary dynamicsX-ray photon correlation spectroscopydomain-adaptive machine learningnanocrystalline siliconnon-equilibrium dynamicskinetic parametersmaterial stability

0 comments

The pith

A combination of X-ray photon correlation spectroscopy and domain-adaptive machine learning extracts kinetic parameters of grain boundary motion from experimental data.

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

Grain boundaries control the stability of nanocrystalline materials, but their slow non-equilibrium motion has been hard to measure directly. This paper combines X-ray photon correlation spectroscopy with a domain-adaptive machine learning framework to turn high-dimensional fluctuation maps into quantitative kinetic parameters. The approach transfers labels from simulations to experimental data in nanocrystalline silicon, revealing that grain boundary relaxation stays far from equilibrium. A reader would care because this turns indirect signals into usable numbers for material behavior.

Core claim

The central claim is that XPCS measurements combined with domain-adaptive ML enable extraction of bulk diffusivity, GB stiffness, and effective GB concentration from experimental two-time correlation maps, which show pronounced departures from time-translation invariance indicating far-from-equilibrium dynamics.

What carries the argument

Domain-adaptive representation alignment in a semi-supervised learning framework that maps simulation-derived physical parameters onto unlabeled experimental XPCS maps.

If this is right

Quantitative parameters of grain boundary dynamics become accessible from standard XPCS experiments.
Non-equilibrium relaxation in nanocrystalline materials can be characterized over lab timescales.
The method provides a route to study defect motion in solids using fluctuation data and machine learning.
Key material properties like stiffness and concentration of grain boundaries can be measured directly.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

This technique could be applied to monitor grain boundary evolution in real time during material processing.
Similar domain adaptation might help other experimental techniques where simulations can label data.
Improved understanding of non-equilibrium states may help design more stable nanocrystalline materials.

Load-bearing premise

The domain-adaptive representation alignment transfers physical parameter labels from continuum simulations to experimental XPCS maps without introducing significant bias or artifacts.

What would settle it

Comparing the extracted parameters to values obtained from independent experiments like atomic-scale imaging or separate diffusion studies on the same samples, and finding systematic disagreements.

Figures

Figures reproduced from arXiv: 2605.12194 by Andrei Fluerasu, Bowen Yu, Chu-Liang Fu, Daniel Pajerowski, G. Jeffrey Snyder, Joshua J Turner, Lutz Wiegart, Matthias T. Agne, Mingda Li, Mouyang Cheng, Nathan C. Drucker, Nina Andrejevic, Qiwei Wan, Riley Hanus, Thanh Nguyen, Xiaoqian M Chen, Yongqiang Cheng.

**Figure 1.** Figure 1: Integrated workflow with experiment, theory, and AI for grain-boundary dynamics. Experimentally, X-ray photon correlation spectroscopy (XPCS) measures the two-time correlation function g2(q,t1,t2), capturing non-equilibrium dynamics of the grain boundary. Theoretically, stochastic differential equation (SDE)-based simulations incorporate thermal diffusion and curvature-driven grain-boundary migration to ge… view at source ↗

**Figure 2.** Figure 2: Experimental temperature-dependent XPCS dynamics and non-equilibrium behavior. a. Two-time correlation functions g2(t1,t2) measured at 299 K, 371 K, 421 K, and 471 K for the same nano-crystalline silicon sample. The results reveal increasingly non-equilibrium features in the dynamics as temperature rises. Here, g2(t1,t2) denotes g2(q,t1,t2) evaluated with in-plane momentum transfer qr = 0.045A˚ −1 and out-… view at source ↗

**Figure 3.** Figure 3: Theoretical SDE predictions and mapping of non-equilibrium dynamics across parameter space. a. Simulated two-time correlation functions g2(t1,t2) for four representative parameter sets, showing the progression from nearly time-translation-invariant dynamics to strongly non-equilibrium behavior. Here g2(t1,t2) denotes g2(q,t1,t2) with in-plane qr = 0.045A˚ −1 . b. Corresponding intensity-intensity correlati… view at source ↗

**Figure 4.** Figure 4: Semi-supervised domain adaptation for mapping experimental XPCS onto the theoretical parameter manifold. a. Prediction results before domain adaptation using the vanilla model trained only on simulated XPCS. Left and middle panel: predicted versus real D and λGB for simulated test data, with a test-set R 2 = 0.995 and 0.893, respectively. Right panel: inferred experimental XPCS locations overlaid on the th… view at source ↗

read the original abstract

Grain-boundary (GB) dynamics control the stability, mechanical, and functional response of nanocrystalline materials, but direct experimental access to their slow non-equilibrium motion has been limited. Here we establish X-ray photon correlation spectroscopy (XPCS), combined with domain-adaptive machine learning, as a quantitative probe of GB dynamics. Temperature- and grain-size-dependent two-time XPCS measurements in nanocrystalline silicon reveal pronounced departures from time-translation invariance, showing that GB relaxation can remain far from equilibrium over experimental timescales. However, direct extraction of quantitative physical information from these high-dimensional, noisy fluctuation maps faces a significant challenge. To overcome this barrier, we develop a semi-supervised learning framework that transfers physical parameter labels from continuum simulations to unlabeled experimental XPCS maps through domain-adaptive representation alignment. This AI-augmented approach enables the extraction of key kinetic parameters, including bulk diffusivity, GB stiffness, and effective GB concentration, directly from experimental XPCS measurements. Our results show how machine learning can transform indirect fluctuation signals into quantitative materials dynamics, providing a general route to study non-equilibrium defect motion in solids.

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 XPCS measurements show real non-equilibrium GB behavior in nano silicon, but the domain-adaptive ML extraction of diffusivity and stiffness lacks any independent experimental check.

read the letter

The key takeaway is that the authors combine XPCS with domain-adaptive machine learning to extract grain boundary kinetic parameters from experimental data in nanocrystalline silicon, but the lack of independent validation leaves the quantitative results on shaky ground. They do solid work on the experimental side. The two-time XPCS measurements at different temperatures and grain sizes show clear departures from time-translation invariance. This indicates that grain boundary relaxation stays out of equilibrium over the timescales of the experiment, which is a useful observation for understanding stability in these materials. What is new is the semi-supervised learning setup that transfers parameter labels from continuum simulations to the unlabeled XPCS maps using representation alignment. This addresses the problem of high-dimensional noisy data where direct fitting is hard. The paper handles the integration reasonably, and the approach could be general for other fluctuation-based measurements. The soft spots are around validation. The description suggests they extract bulk diffusivity, GB stiffness, and effective concentration, but there are no cross-checks with other techniques like tracer diffusion or microscopy-based migration rates. Without those, or at least error estimates and sensitivity tests, it's difficult to trust the specific numbers. The domain adaptation might introduce bias if the simulations don't match the experiment closely enough. The citation pattern seems standard, pulling from XPCS literature and ML transfer learning papers. This paper is aimed at materials physicists and engineers working on nanocrystalline systems, particularly those interested in non-equilibrium defect dynamics or applying ML to experimental signals. A reader in that area would find the experimental results worthwhile even if they skip the ML details. It deserves a serious referee because the experimental data is real and the method is a plausible step forward, though it will need revisions on the validation front. I would send it to peer review.

Referee Report

2 major / 1 minor

Summary. The manuscript claims that XPCS two-time correlation measurements on nanocrystalline silicon, combined with a semi-supervised domain-adaptive machine learning framework, can extract quantitative kinetic parameters (bulk diffusivity, GB stiffness, and effective GB concentration) from experimental fluctuation maps by transferring labels from continuum simulations. It reports pronounced departures from time-translation invariance, indicating persistent non-equilibrium GB relaxation, and positions the AI-augmented method as a general route to study slow defect dynamics.

Significance. If the domain-adaptive alignment proves robust, the work would provide a valuable new capability for quantifying non-equilibrium GB kinetics from indirect XPCS signals where conventional analysis is intractable. The integration of continuum modeling with representation alignment is a clear strength that could generalize to other fluctuation spectroscopies and materials systems lacking ground-truth labels.

major comments (2)

[Abstract] Abstract: The central claim that the method 'enables the extraction of key kinetic parameters, including bulk diffusivity, GB stiffness, and effective GB concentration, directly from experimental XPCS measurements' is presented without any reported quantitative validation (recovered values with uncertainties, comparison to independent experiments such as tracer diffusion, or ablation studies on the alignment step). This is load-bearing because experimental maps carry no ground-truth labels, so correctness rests entirely on unshown fidelity of the transfer.
[Methods / Results] Simulation-to-experiment transfer section: The workflow assumes continuum simulations generate XPCS signals faithful to experiment across the explored parameter ranges, yet no sensitivity analysis, mismatch quantification, or cross-validation against known GB migration rates is described. Any systematic discrepancy would propagate directly into the reported diffusivity, stiffness, and concentration values.

minor comments (1)

[Abstract] Abstract: Adding one sentence specifying the neural-network architecture and loss function used for domain-adaptive alignment would improve immediate clarity and reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their positive evaluation of our work and for the constructive major comments. We address each point below and outline revisions that will strengthen the manuscript without altering its core claims.

read point-by-point responses

Referee: [Abstract] Abstract: The central claim that the method 'enables the extraction of key kinetic parameters, including bulk diffusivity, GB stiffness, and effective GB concentration, directly from experimental XPCS measurements' is presented without any reported quantitative validation (recovered values with uncertainties, comparison to independent experiments such as tracer diffusion, or ablation studies on the alignment step). This is load-bearing because experimental maps carry no ground-truth labels, so correctness rests entirely on unshown fidelity of the transfer.

Authors: We acknowledge that the abstract states the central claim without accompanying quantitative metrics. The main text reports domain-adaptation performance via held-out simulation accuracy and qualitative consistency with experimental trends, but does not include explicit recovered parameter values with uncertainties from the experimental maps, ablation studies, or direct comparisons to independent measurements. In the revised version we will add a validation subsection that reports the extracted bulk diffusivity, GB stiffness, and effective concentration values together with their uncertainties, includes an ablation study on the domain-alignment component, and discusses consistency with literature values for silicon self-diffusion. Direct tracer-diffusion data on the identical nanocrystalline samples are unavailable, which we will note explicitly. revision: yes
Referee: [Methods / Results] Simulation-to-experiment transfer section: The workflow assumes continuum simulations generate XPCS signals faithful to experiment across the explored parameter ranges, yet no sensitivity analysis, mismatch quantification, or cross-validation against known GB migration rates is described. Any systematic discrepancy would propagate directly into the reported diffusivity, stiffness, and concentration values.

Authors: The referee is correct that the current manuscript does not present a dedicated sensitivity analysis or quantitative mismatch metrics between simulated and experimental fluctuation maps. The continuum model follows established GB evolution equations, and parameter ranges were chosen to bracket experimental conditions, but explicit propagation of simulation-experiment discrepancies was not quantified. We will revise the Methods and Results sections to add: (i) a sensitivity study showing how variations in simulation inputs (mobility, stiffness, concentration) affect the downstream experimental parameter estimates; (ii) latent-space distribution distances as a mismatch measure; and (iii) cross-validation against published GB migration rates for silicon where comparable data exist. These additions will make the robustness of the label transfer explicit. revision: yes

Circularity Check

0 steps flagged

No significant circularity; transfer-learning pipeline is externally grounded

full rationale

The paper's core workflow generates labeled XPCS maps from independent continuum simulations of GB dynamics, then applies domain-adaptive representation alignment to map those labels onto unlabeled experimental two-time correlation maps. This is a standard semi-supervised transfer procedure whose inputs (simulation physics and real experimental data) are not defined in terms of the extracted outputs (bulk diffusivity, GB stiffness, effective GB concentration). No equation or method step reduces a prediction to a fitted parameter by construction, no self-citation chain is load-bearing for the central claim, and the derivation remains self-contained against external simulation benchmarks and experimental measurements.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim depends on the fidelity of continuum simulations as label sources and on the effectiveness of domain adaptation; no new physical entities are postulated.

axioms (2)

domain assumption Grain-boundary relaxation in nanocrystalline silicon remains far from equilibrium on experimental timescales, as evidenced by departures from time-translation invariance in two-time XPCS data.
Stated directly in the abstract as the key physical observation.
domain assumption Continuum simulations generate physically accurate labels for bulk diffusivity, GB stiffness, and GB concentration that can be aligned with experimental fluctuation maps.
Required for the semi-supervised transfer step to produce meaningful extracted parameters.

pith-pipeline@v0.9.0 · 5563 in / 1401 out tokens · 88215 ms · 2026-05-13T04:51:59.954180+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

semi-supervised learning framework that transfers physical parameter labels from continuum simulations to unlabeled experimental XPCS maps through domain-adaptive representation alignment... extraction of key kinetic parameters, including bulk diffusivity D, GB stiffness Γ, and effective GB concentration λGB

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

40 extracted references · 40 canonical work pages · 1 internal anchor

[1]

& Hemker, K

Rupert, T., Gianola, D., Gan, Y . & Hemker, K. Experimental observations of stress-driven grain boundary migration.Science326, 1686–1690 (2009). 2.Wang, L.et al.Tracking the sliding of grain boundaries at the atomic scale.Science375, 1261–1265 (2022)

work page 2009
[2]

Adv.8, eabn3785 (2022)

Fang, Z.et al.Atomic-scale observation of dynamic grain boundary structural transformation during shear- mediated migration.Sci. Adv.8, eabn3785 (2022)

work page 2022
[3]

Zhang, B., Tang, Y ., Mei, Q., Li, X. & Lu, K. Inhibiting creep in nanograined alloys with stable grain boundary networks.Science378, 659–663 (2022)

work page 2022
[4]

M.et al.Irradiation-induced grain boundary facet motion: In situ observations and atomic-scale mechanisms.Sci

Barr, C. M.et al.Irradiation-induced grain boundary facet motion: In situ observations and atomic-scale mechanisms.Sci. Adv.8, eabn0900 (2022)

work page 2022
[5]

V ., Suter, R

Bernier, J. V ., Suter, R. M., Rollett, A. D. & Almer, J. D. High-energy x-ray diffraction microscopy in materials science.Annu. Rev. Mater. Res.50, 395–436 (2020). 7.Zewail, A. H. Four-dimensional electron microscopy.science328, 187–193 (2010)

work page 2020
[6]

Science386, 49–54 (2024)

Tian, Y .et al.Grain rotation mechanisms in nanocrystalline materials: Multiscale observations in pt thin films. Science386, 49–54 (2024)

work page 2024
[7]

& Harder, R

Robinson, I., Clark, J. & Harder, R. Materials science in the time domain using bragg coherent diffraction imaging.J. Opt.18, 054007 (2016)

work page 2016
[8]

W., Stephenson, G

Yau, A., Cha, W., Kanan, M. W., Stephenson, G. B. & Ulvestad, A. Bragg coherent diffractive imaging of single-grain defect dynamics in polycrystalline films.Science356, 739–742 (2017)

work page 2017
[9]

Mater.33, 2104659 (2021)

García-Moreno, F.et al.Tomoscopy: Time-resolved tomography for dynamic processes in materials.Adv. Mater.33, 2104659 (2021)

work page 2021
[10]

M., Ludwig, W

McKenna, I., Poulsen, S., Lauridsen, E. M., Ludwig, W. & V oorhees, P. W. Grain growth in four dimensions: A comparison between simulation and experiment.Acta materialia78, 125–134 (2014). 13.Zhang, J.et al.Grain boundary mobilities in polycrystals.Acta Materialia191, 211–220 (2020)

work page 2014
[11]

& Shvindlerman, L

Gottstein, G., Ma, Y . & Shvindlerman, L. Triple junction motion and grain microstructure evolution.Acta Materialia53, 1535–1544 (2005). 12/14

work page 2005
[12]

& Gottstein, G

Upmanyu, M., Srolovitz, D., Shvindlerman, L. & Gottstein, G. Misorientation dependence of intrinsic grain boundary mobility: simulation and experiment.Acta materialia47, 3901–3914 (1999)

work page 1999
[13]

E., Cahn, J

Taylor, J. E., Cahn, J. W. & Handwerker, C. A. Overview no. 98 i—geometric models of crystal growth.Acta Metall. et Materialia40, 1443–1474 (1992)

work page 1992
[14]

Mishin, Y . M. Atomistic theory and computer simulation of grain boundary diffusion. InDefect and Diffusion Forum, vol. 143, 1357–1372 (Trans Tech Publ, 1997)

work page 1997
[15]

Kobayashi, R., Warren, J. A. & Carter, W. C. A continuum model of grain boundaries.Phys. D: Nonlinear Phenom.140, 141–150 (2000)

work page 2000
[16]

& Mishin, Y

Trautt, Z. & Mishin, Y . Grain boundary migration and grain rotation studied by molecular dynamics.Acta Materialia60, 2407–2424 (2012)

work page 2012
[17]

Leheny, R. L. Xpcs: Nanoscale motion and rheology.Curr. opinion colloid & interface science17, 3–12 (2012)

work page 2012
[18]

X-ray photon correlation spectroscopy for the characterization of soft and hard condensed matter

Bikondoa, O. X-ray photon correlation spectroscopy for the characterization of soft and hard condensed matter. X-ray Neutron Tech. for Nanomater. Charact.95–156 (2016)

work page 2016
[19]

& Sprung, M

Lu, X., Mochrie, S., Narayanan, S., Sandy, A. & Sprung, M. Temperature-dependent structural arrest of silica colloids in a water–lutidine binary mixture.Soft Matter6, 6160–6177 (2010)

work page 2010
[20]

& Narayanan, T

Zinn, T., Sharpnack, L. & Narayanan, T. Phoretic dynamics of colloids in a phase separating critical liquid mixture.Phys. Rev. Res.2, 033177 (2020)

work page 2020
[21]

Porter, Z.et al.Understanding the superconductivity and charge density wave interaction through quasi-static lattice fluctuations.Proc. Natl. Acad. Sci.121, e2412182121 (2024)

work page 2024
[22]

Shen, L.et al.Interplay between relaxational atomic fluctuations and charge density waves in la 2- x sr x cuo 4. Phys. Rev. B108, L201111 (2023)

work page 2023
[23]

Adv.11, eadt5680 (2025)

Tumbleson, Z.et al.Thermodynamic phase transitions of nematic order in magnetic helices.Sci. Adv.11, eadt5680 (2025)

work page 2025
[24]

P.et al.Ai-nerd: Elucidation of relaxation dynamics beyond equilibrium through ai-informed x-ray photon correlation spectroscopy.Nat

Horwath, J. P.et al.Ai-nerd: Elucidation of relaxation dynamics beyond equilibrium through ai-informed x-ray photon correlation spectroscopy.Nat. Commun.15, 5945 (2024)

work page 2024
[25]

K.et al.Nanostructured bulk silicon as an effective thermoelectric material.Adv

Bux, S. K.et al.Nanostructured bulk silicon as an effective thermoelectric material.Adv. Funct. Mater.19, 2445–2452 (2009)

work page 2009
[26]

Siegert, A.On the fluctuations in signals returned by many independently moving scatterers(Radiation Laboratory, Massachusetts Institute of Technology, 1943)

work page 1943
[27]

Berne, B. J. & Pecora, R.Dynamic light scattering: with applications to chemistry, biology, and physics (Courier Corporation, 2000). 31.Allen, M. P. & Tildesley, D. J.Computer simulation of liquids(Oxford university press, 2017)

work page 2000
[28]

& Kulkarni, Y

Chen, D. & Kulkarni, Y . Atomistic modeling of grain boundary motion as a random walk.Phys. Rev. Mater.2, 093605 (2018)

work page 2018
[29]

& Saenko, K

Sun, B. & Saenko, K. Deep coral: Correlation alignment for deep domain adaptation. InEuropean conference on computer vision, 443–450 (Springer, 2016)

work page 2016
[30]

UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction

McInnes, L., Healy, J. & Melville, J. Umap: Uniform manifold approximation and projection for dimension reduction.arXiv preprint arXiv:1802.03426(2018). 35.Qiu, C.et al.Grain boundaries are brownian ratchets.Science385, 980–985 (2024). 36.Carrasquilla, J. Machine learning for quantum matter.Adv. Physics: X5, 1797528 (2020)

work page internal anchor Pith review Pith/arXiv arXiv 2018
[31]

Chen, Z.et al.Machine learning on neutron and x-ray scattering and spectroscopies.Chem. Phys. Rev.2(2021). 13/14

work page 2021
[32]

Cheng, M., Okabe, R., Chotrattanapituk, A. & Li, M. Machine learning detection of majorana zero modes from zero-bias peak measurements.Matter7, 2507–2520 (2024)

work page 2024
[33]

Han, X.-Q.et al.Ai-driven inverse design of materials: Past, present, and future.Chin. Phys. Lett.42, 027403 (2025)

work page 2025
[34]

Mater.37, 2505642 (2025)

Fu, C.-L.et al.Ai-driven defect engineering for advanced thermoelectric materials.Adv. Mater.37, 2505642 (2025)

work page 2025
[35]

Mater.1–17 (2026)

Cheng, M.et al.Artificial intelligence-driven approaches for materials design and discovery.Nat. Mater.1–17 (2026)

work page 2026
[36]

Cheng, M.et al.A foundation model for non-destructive defect identification from vibrational spectra.Matter 102728 (2026)

work page 2026
[37]

Chen, X.et al.Detecting thermodynamic phase transition via explainable machine learning of photoemission spectroscopy.Newton1(2025)

work page 2025
[38]

& Krishnapriyan, A

Nelson, K., Kreiman, T., Levine, S. & Krishnapriyan, A. S. Bridging the simulation-to-experiment gap with generative models using adversarial distribution alignment.arXiv preprint arXiv:2604.01169(2026)

work page arXiv 2026
[39]

Nye, A., Leff, A., Barr, C. M. & Taheri, M. Direct observation of recrystallization mechanisms during annealing of cu in low and high strain conditions.Scripta Materialia146, 308–311 (2018)

work page 2018
[40]

Zhang, L., Han, J., Srolovitz, D. J. & Xiang, Y . Equation of motion for grain boundaries in polycrystals.npj Comput. Mater.7, 64 (2021). 47.Rohrer, G. S.et al.Grain boundary migration in polycrystals.Annu. Rev. Mater. Res.53, 347–369 (2023). 14/14

work page 2021