Mapping recrystallization trajectories in GaAs using latent space diffraction analysis

Ellis Rae Kennedy; Erik Thiede; Kwanghwi Je

REVIEW 5 major objections 5 minor 48 references

Mapping recrystallization trajectories in GaAs using latent space diffraction analysis

T0 review · 5 major / 5 minor · reviewed 2026-08-02 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Latent-space diffraction tracking reveals two distinct recrystallization regimes in irradiated GaAs, separated near 250°C.

desk verdict Latent-space pipeline for in situ 4D-STEM is a real novelty with a credible 250°C regime switch in GaAs, but the transition trajectories and precursor claim rest on alignment and CCF inconsistencies that need fixing. read the letter →

arxiv 2607.13779 v1 pith:PV4SVSJY submitted 2026-07-15 cond-mat.mtrl-sci

Ellis Rae Kennedy , Kwanghwi Je , Erik Thiede This is my paper · ORCID

classification cond-mat.mtrl-sci

keywords 4D-STEMconvolutionalautoencoderlatentspaceanalysisrecrystallizationgalliumarsenidephasetransitionpathwaystwinningamorphousprecursors

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that recrystallization of ion-irradiated GaAs is not a single smooth ordering process but splits into two distinct regimes on either side of roughly 250°C, and that the transition can be read directly from in situ diffraction data if the data are embedded in a learned latent space. Using a convolutional autoencoder to compress each 4D-STEM diffraction pattern, the authors cluster recurring structural states—amorphous, crystalline, polycrystalline, twinned, and hybrid intermediates—and track how each nanometer-scale pixel changes label between temperature steps. The resulting transition matrices show a low-temperature regime in which amorphous and crystalline states simply persist, and a high-temperature regime in which the amorphous phase becomes unstable and hybrid, faulted states act as precursors that funnel material into twinned end states. The paper also claims the latent space exposes amorphous regions with weak angular symmetry that begin ordering before any Bragg peaks appear, offering an early warning of nucleation. If correct, this gives a general, descriptor-free way to map phase-transformation pathways in disordered solids from microscopy data.

What carries the argument

The central mechanism is the latent-space trajectory model: a convolutional autoencoder compresses each 256×256 diffraction pattern into a 218-dimensional vector; PCA projection and k-means clustering (50 clusters, over-clustered deliberately) assign each pattern to a structural macrostate; and per-pixel transition matrices between consecutive temperature steps summarize the probability of moving from one macrostate to another. The load-bearing piece is the identification of hybrid clusters—A/C, C/T, C/P, T/P, R/T—that occupy boundary regions in latent space and act as the intermediates whose transition fates (dissolve, crystallize, twin, coarsen) define the recrystallization pathways. A log

What would settle it

Re-run the analysis with full-image cross-correlation registration of the bright-field frames and with sub-minute acquisition at 25°C steps around 250°C; if the regime boundary and the amorphous-to-twinned pathway do not survive, the trajectory model is an alignment or time-averaging artifact.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that a convolutional autoencoder trained on in situ 4D-STEM diffraction patterns from ion-irradiated GaAs yields a physically interpretable latent space in which unsupervised clustering recovers the known structural phases plus hybrid boundary states. Tracking cluster labels of individual scan pixels across temperature steps produces transition probabilities that reveal a sharp reorganization near 250°C: below it, amorphous and crystalline states are deep, self-retaining basins; above it, amorphous regions become unstable and hybrid, faulted, and polycrystalline states emerge as metastable intermediates that feed primarily into twinned structures, whic

Load-bearing premise

Every pixel trajectory assumes the same scan position corresponds to the same physical location at every temperature—via fiducial-based alignment with only global translations—and that each ~12-minute acquisition is a quasi-static snapshot at the nominal temperature; if the sample drifts, rotates, or recrystallizes significantly during acquisition, the transition probabilities are artifacts.

Editorial extensions

If this is right

Below roughly 250°C, amorphous and crystalline regions are kinetically trapped; the amorphous phase self-retains with probability 0.91 and the ordered basin with 0.79.
Above 250°C, the amorphous population collapses (A→A = 0.00) and hybrid, polycrystalline, and twinned states emerge rapidly.
Twinned structures are the dominant terminal state, fed primarily by faulted crystalline-hybrid intermediates (C/T→T probabilities of 0.46–0.57), not directly from amorphous material.
Amorphous patterns that later crystallize show elevated angular cross-correlation harmonics (notably six-fold), a precursor invisible to standard diffraction descriptors.
The latent space provides a continuous reaction coordinate (logistic-regression committor) that separates amorphous regions destined to dissolve from those destined to crystallize.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The same pipeline could be applied to other disordered systems (silicon, germanium, metallic glasses) where the crystalline orientation is unknown; the latent space would need no manual cluster labeling if phase identities are validated against known reference patterns.
The dominance of six-fold angular symmetry in incipient amorphous patterns suggests the amorphous matrix first forms tetrahedrally coordinated clusters before aligning with the surrounding crystal; this could be tested with fluctuation microscopy or by simulating diffraction patterns from paracrystalline models.
With faster detectors, the temperature-indexed transition matrices could become true time-resolved Markov models, allowing quantitative rate constants and validation of the proposed twinning pathway against kinetic Monte Carlo or phase-field simulations.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: the paper is worth engaging. The new thing is the application — a temperature-resolved in situ 4D-STEM series run through a convolutional autoencoder, clustered into phases, and turned into transition matrices. That combination is new for heating data. The central observation, a regime switch near 250°C where the amorphous population collapses and twinned and polycrystalline states take over, is credible: it shows up in the raw phase counts and matches known GaAs behavior. The data are on Zenodo, and the authors decline to label their matrices a Markov state model, which is the right call.

The problems are real but fixable, and the reader's report locates them correctly. First, registration. Every transition probability comes from following a fixed (y,x) index across temperatures, but the alignment is translation-only, with the advancing recrystallization front used as a fiducial. If the lamella rotated or deformed during the 12-minute acquisitions — common on MEMS heating chips — the interface-region transitions, exactly the A→H and H→T pathways the story leans on, become artifacts. Aligning to the moving front can cancel the very interface motion the trajectories are meant to measure. The conclusions acknowledge time-averaging but never quantify registration error. That is the load-bearing assumption and needs to be addressed head-on.

Second, the precursor claim has a numerical contradiction. The text says incipient amorphous patterns show elevated harmonic content dominated by six-fold symmetry, but the stated values show the four-fold harmonic dropping from 4.36e4 to 1.40e3 and the six-fold from 3.75e4 to 1.63e3 — an order-of-magnitude decrease. Only the two-fold increases. The numbers say the opposite of the prose. That has to be fixed before the precursor result can be assessed.

Smaller items: no error bars on the transition probabilities; the 290-sample logistic fit is called a committor and it is not one; the hybrid story has built-in circularity (hybrids are defined as lying between basins, so finding they funnel into neighbors is partly semantics — though the C/T→twinned pathway is not tautological and is the interesting bit); and the regime language flips between growth-dominated and nucleation-dominated across abstract, results, and captions, with one phrase, 'growth-limited (nucleation-dominated),' contradicting itself. The text says eight amorphous clusters; the SI table lists nine, and the stated counts sum to 49, not 50.

Bottom line: this deserves a serious referee, not a desk reject. The regime switch is defensible and the method is novel enough to matter. I would send it out with instructions to fix the CCF values, add uncertainty estimates, confront the registration limitation, and standardize the terminology.

Referee Report

5 major / 5 minor

Summary. The paper presents a latent-space workflow for analyzing in situ 4D-STEM data from ion-irradiated GaAs during thermal annealing. A convolutional autoencoder compresses normalized diffraction patterns into 218-dimensional latent vectors; PCA and k-means clustering (50 clusters) assign each pattern a microstate, and these clusters are manually grouped into phases (amorphous, crystalline, recrystallized, polycrystalline, twinned, and several hybrid classes). Pixel-level trajectories across eight temperature steps are converted into transition matrices, which are interpreted as coarse-grained recrystallization pathways. The central claim is that the data reveal two distinct regimes separated near 250°C: a low-temperature regime in which amorphous and crystalline basins persist, and a high-temperature regime in which hybrid and faulted states act as precursors to twinning and twinned structures emerge as terminal states. A secondary claim is that a subset of amorphous patterns carry weak angular-correlation symmetry signatures that predict their subsequent crystallization, visible only through the latent-space/CCF analysis. The paper also introduces a logistic-regression 'committor' on PCA-reduced latent vectors for A→C* events. The data are openly deposited, and the computational pipeline is described in sufficient detail to be reproduced.

Significance. If the central claims hold, this is a meaningful methodological contribution. The idea of using a learned latent manifold rather than hand-picked diffraction descriptors to define metastable states in 4D-STEM data is timely, and the application to a nontrivial solid-state transformation (amorphous-to-crystalline recrystallization with twinning) is a good showcase. The manuscript is also unusually transparent about its choices (over-clustering, manual labels, aggregation of temperature steps) and provides open data. The two-regime picture itself—persistent A/C basins at low temperature and a hybrid-mediated, twin-favoring landscape at high temperature—is physically plausible and qualitatively consistent with the known low stacking-fault energy of GaAs. However, the quantitative pathway claims rest on a pixel-registration assumption that is not demonstrated, and several supporting analyses lack uncertainty quantification. With appropriate validation and revised wording, the framework could become a useful bridge between in situ 4D-STEM experiments and trajectory-based analyses of phase transformations.

major comments (5)

[Methods, '4D-STEM dataset alignment'; Tables II/III; Fig. 5] The paper's central quantitative output is a set of pixel-resolved transition probabilities computed by tracking the cluster label at a fixed (y,x) index across consecutive temperatures. The Methods report only global translation alignment based on fiducial features that include the advancing recrystallization front and a carbon spot; no rotation, shear, affine, or local-distortion correction is described, and only scan positions present at every temperature are retained. With a 5 nm step and ~12 min acquisitions on a MEMS chip, thermal drift and lamella deformation can move features by a fraction of a step, generating spurious transitions at the amorphous/crystalline interfaces—precisely the regions that dominate the A→H (0.45) and H→T (0.47) pathways in Table II. Using the moving recrystallization front as a fiducial is also circular, since aligning to it can cancel the interface motio
[Results, 'Manual labeling'; Supplemental Table 1] Phase labels are assigned manually to unsupervised k-means clusters, and hybrid states are explicitly defined as clusters lying in boundary regions between major basins (A/C, C/T, C/P, T/P, R/T). The subsequent finding that hybrids 'funnel' into their neighboring phases—e.g., H→T=0.47, C/T clusters 34/43/44→T=0.46–0.57, C/P cluster 37→P=0.81—is therefore partly predetermined by the labeling rule. The paper should provide an independent validation of the phase labels, for example quantitative comparison of cluster-averaged diffraction signatures to kinematical/dynamical simulations, or a blinded/rule-based assignment, and should report how label uncertainty propagates into the transition probabilities. Without this, the claim that the latent space 'reveals' these pathways is weakened, because the same data that define the states are used to infer their dynamics.
[Abstract; Results, Table I; Discussion] The abstract and Discussion locate a 'transition near 250°C'. Table I shows that the qualitative change actually occurs between the 250°C step and the 300°C step: at 250°C the amorphous fraction is still 0.63 with H=0.10, P=0.00, T=0.00; at 300°C A=0.00, H=0.37, P=0.10, T=0.19. With 50°C increments, the data constrain the crossover only to the interval 250–300°C, not to 'near 250°C.' Please soften the claim or provide finer temperature sampling. This does not change the existence of two regimes, but it is a quantitative overstatement of the temperature resolution.
[Tables II/III and Fig. 6a] Several quantitative claims lack uncertainty estimates. The transition probabilities are point estimates without bootstrap confidence intervals or sensitivity analyses with respect to the number of clusters (50), the PCA truncation (64 components), and the network thresholds (3% probability, 10 transitions). Some counts are small (e.g., cluster 14 appears only at the final temperature step and has no outgoing transitions), so the reported probabilities are not obviously distinguishable from sampling noise. Add bootstrap or posterior intervals and report sensitivity to the clustering hyperparameters.
[Results, 'To quantify the fate of A/C hybrid patterns'] The logistic-regression 'committor' is fitted on 290 A/C* samples with no cross-validation, no reported classification accuracy, and no confidence interval on the coefficient vector or the boundary ξ=0. The claim that this 'functions analogously to a committor' and provides a 'quantitative, low-dimensional reaction coordinate' would be substantially stronger with out-of-sample prediction of A→C* outcomes and a comparison against a null model that uses only distance from the amorphous centroid. This is a supporting claim, but it is presented as one of the two main findings.

minor comments (5)

[Abstract and Introduction] The sentence 'not captured by conventional descriptors give new insights' is grammatically incomplete; adjust to '...and give new insights...' or similar.
[Results, angular cross-correlation paragraph] The text says the four-fold harmonic 'increases' from 4.36×10^4 to 1.40×10^3 and the six-fold from 3.75×10^4 to 1.63×10^3, but the numbers shown are decreases. Please correct the values or the wording.
[Table I] The counts assigned to C versus R swap abruptly across temperature (R=647 at 25°C, C=915 at 100°C, R=1498 at 350°C with C=0), even though the text states these two labels are structurally identical except for a <1° orientation difference. Since C and R are later combined into C*, this is not fatal, but the abrupt swapping should be acknowledged as a labeling-instability effect or justified as a true orientation reorientation.
[Discussion, second paragraph] The phrase 'our matrices the matrices should be interpreted' contains a duplication; please edit.
[Introduction/Discussion, novelty claims] The claim of 'first nanometer-resolved, in situ observation of high-temperature recrystallization pathways in an amorphous solid' is broad. Please temper it or place it in a more precise context, given the extensive in situ TEM recrystallization literature.

Circularity Check

0 steps flagged · score 2.0 of 10

No load-bearing circularity; central trajectory analysis is self-contained, with only minor self-citation and a mild definitional flavor in hybrid labels.

full rationale

The central derivation chain is self-contained: diffraction patterns are encoded by a CAE, clustered into 50 static microstates, manually grouped into phases from representative patterns and latent-space geometry, and transition probabilities are then measured by tracking fixed (y,x) positions across temperature steps. The two-regime claim, hybrid-mediated twinning, and incipient amorphous precursors are empirical outputs of this pipeline rather than identities. The hybrid labels (A/C, C/T, C/P, etc.) are assigned from latent-space geometry before the trajectory analysis; although a name like C/T names T as one endpoint, the measured directional probabilities (e.g., C/T to T vs. other outcomes) are not forced by the label, so the pathway claims have independent empirical content. The logistic-regression 'committor' is fit to the same A-to-C* outcomes it is used to summarize, so it functions as a descriptive classifier rather than an out-of-sample prediction; this is a validation limitation, not a circular derivation. Self-citations to Kennedy et al. [22] supply the dataset and prior structural characterization, but the latent-space trajectory topology does not reduce to those citations. Registration and quasi-static acquisition concerns are experimental validity risks, not circularity. Overall, no step of the derivation reduces to its own input.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

No new physical objects, forces, particles, or dimensions are introduced. 'Hybrid states' and 'commitment probability' are analytical labels rather than independently evidenced entities. The load-bearing ingredients are analysis choices (cluster count, thresholds, k-band), manual labels, and the registration/snapshot assumptions.

free parameters (6)

Number of k-means clusters = 50
Deliberately chosen to over-partition the latent space; determines which clusters become 'hybrid' states and thus shapes the transition-pathway statistics (Methods: Dimensionality reduction and clustering).
PCA retained components = 64
Reduces the 218-dimensional latent vectors before clustering and logistic regression; changes the geometry in which distances and clusters are computed.
CAE latent dimension = 218
Architecture choice for the autoencoder bottleneck; all latent-space distances, including the 29.57 vs 19.59 precursor analysis, depend on this coordinate system.
Logistic regression reaction-coordinate parameters = b = -3.53; w unspecified
Fitted to 290 A/C* patterns to predict whether an amorphous pattern crystallizes; the resulting ξ = w^T z + b is presented as a committor-like coordinate without held-out validation.
Transition-network thresholds = probability ≥ 3%, count ≥ 10
Used to prune the visualized transition networks; changing these thresholds changes the apparent topology of Figure 5.
CCF k-space band = 4.1–8.4 mrad
Chosen to cover the amorphous ring and {111} Bragg region; the reported 2-, 4-, and 6-fold harmonic amplitudes depend on this selection.

assumptions (5)

domain assumption Real-space registration across temperature is valid
Alignment uses global translations from fiducial features; if drift, rotation, or deformation occurs, pixel trajectories and transition matrices are invalid (Methods: 4D-STEM dataset alignment).
domain assumption Each 12-minute acquisition is a quasi-static snapshot at the nominal temperature
The sample is stabilized for ~1 minute then acquired over ~12 minutes; structural evolution during acquisition and cumulative thermal history are ignored (Methods: 4D-STEM data acquisition).
domain assumption Manual cluster-to-phase assignment is correct
The 50 clusters are grouped into A, C, R, P, T, H by human inspection of representative patterns (Methods; SI Table 1); all phase-level claims inherit this labeling.
domain assumption Euclidean distance in latent space measures structural stability and ordering propensity
The precursor analysis interprets the larger mean latent distance of incipient amorphous patterns as evidence of imminent crystallization; this requires the latent metric to be physically meaningful.
domain assumption CCF harmonic amplitudes in the chosen k-band report short-range order rather than noise
The 'weak symmetry signatures' claim relies on 2-, 4-, and 6-fold harmonic amplitudes extracted from angular cross-correlations with no noise model.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Mapping recrystallization trajectories in GaAs using latent space diffraction analysis." pith.science (2026). https://pith.science/paper/PV4SVSJY

@misc{pith2026260713779,
  author       = {Pith},
  title        = {Pith review of: Mapping recrystallization trajectories in GaAs using latent space diffraction analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PV4SVSJY}},
  note         = {Machine review of arXiv:2607.13779}
}

read the original abstract

Recrystallization in disordered solids proceeds through a sequence of local structural rearrangements that are difficult to resolve using conventional diffraction analysis. In amorphous and partially ordered materials, subtle variations in diffuse scattering, short-range order, and defect-mediated symmetry emergence encode the pathways through which ordering initiates and propagates. Here, we introduce a latent space framework for mapping these pathways directly from \textit{in situ} 4D-STEM diffraction data. A convolutional autoencoder provides a compact representation of structural motifs, and unsupervised clustering identifies recurring microstructural states, including amorphous, paracrystalline, crystalline, twinned, and hybrid intermediates. By tracking these states across temperature, we construct phase trajectory models that reveal the topology of the recrystallization landscape, including metastable basins, branching pathways, hybrid states, and temperature-dependent reorganizations of accessible states. Applied to ion irradiated GaAs, this approach uncovers two distinct recrystallization regimes separated by a transition near 250\textdegree{}C. At low temperature, recrystallization is growth-dominated and dominated by the persistence of amorphous and crystalline states. At high temperature, the transformation landscape reorganizes: hybrid and faulted states become metastable precursors to twinning, polycrystalline regions stabilize, and twinned structures emerge as dominant end states. The latent space representation also identifies amorphous patterns with weak symmetry signatures that precede recrystallization. This reveals structural precursors to ordering that are not captured by conventional descriptors give new insights into how recrystallization is initiated.

Figures

Figures reproduced from arXiv: 2607.13779 by the authors.

**Figure 2.** FIG. 2. Workflow for extracting structural phases and transition pathways from [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

**Figure 5.** Figure 5: FIG. 5 [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

**Figure 6.** Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

**Figure 7.** Figure 7: FIG. 7 [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 2 canonical work pages

[1]

C. Zhao, H. Batiz, B. Yasar, H. Kim, W. Ji, M. C. Scott, D. C. Chrzan, and A. Javey, Tellurium Single-Crystal Arrays by Low-Temperature Evaporation and Crystal- lization, Advanced Materials33, 2100860 (2021)

2021
[2]

Veli¸ sa, E

G. Veli¸ sa, E. Zarkadoula, D. Iancu, M. D. Mihai, A. Boulle, Y. Tong, D. Chen, Y. Zhang, and W. J. We- ber, Revealing a Pathway for Low-Temperature Recrys- tallization in Germanium, Advanced Sciencen/a, e07630 (2025)

2025
[3]

M. M. J. Treacy and K. B. Borisenko, The Local Struc- ture of Amorphous Silicon, Science335, 950 (2012), pub- lisher: American Association for the Advancement of Sci- ence

2012
[4]

L. A. M. Rosset, D. A. Drabold, and V. L. De- ringer, Signatures of paracrystallinity in amorphous sil- icon from machine-learning-driven molecular dynamics, Nature Communications16, 2360 (2025), publisher: Na- ture Publishing Group

2025
[5]

G. Wu, C. Liu, Y.-Q. Yan, S. Liu, X. Ma, S. Yue, and Z.- W. Shan, Elemental partitioning-mediated crystalline-to- amorphous phase transformation under quasi-static de- formation, Nature Communications15, 1223 (2024)

2024
[6]

K. F. Adam and D. P. Field, Analyzing recrystallization behavior of heterogeneous structures single-phase Al al- 16 loys, Materialia , 101190 (2021)

2021
[7]

N. G. Rudawski, A. G. Lind, and T. P. Martin, Chap- ter Four - Defective Solid-Phase Epitaxial Growth of Si, inDefects in Semiconductors, Semiconductors and Semimetals, Vol. 91, edited by L. Romano, V. Privitera, and C. Jagadish (Elsevier, 2015) pp. 123–163

2015
[8]

P. M. Voyles, N. Zotov, S. M. Nakhmanson, D. A. Drabold, J. M. Gibson, M. M. J. Treacy, and P. Keblin- ski, Structure and physical properties of paracrystalline atomistic models of amorphous silicon, Journal of Ap- plied Physics90, 4437 (2001)

2001

Show all 48 references

[9]

S. S. Sørensen, C. A. N. Biscio, M. Bauchy, L. Fajstrup, and M. M. Smedskjaer, Revealing hidden medium-range order in amorphous materials using topological data analysis, Science Advances6, eabc2320 (2020), publisher: American Association for the Advancement of Science

2020
[10]

Kennedy, E

E. Kennedy, E. Hollingworth, A. Ceballos, D. O’Mahoney, C. Ophus, F. Hellman, and M. Scott, Exploring structural anisotropy in amorphous Tb-Co via changes in medium-range ordering, Microscopy and Microanalysis , ozae113 (2024)

2024
[11]

Gammer, C

C. Gammer, C. Ophus, T. C. Pekin, J. Eckert, and A. M. Minor, Local nanoscale strain mapping of a metal- lic glass during in situ testing, Applied Physics Letters 112, 171905 (2018)

2018
[12]

Ophus, Four-Dimensional Scanning Transmission Electron Microscopy (4D-STEM): From Scanning Nan- odiffraction to Ptychography and Beyond, Microscopy and Microanalysis25, 563 (2019)

C. Ophus, Four-Dimensional Scanning Transmission Electron Microscopy (4D-STEM): From Scanning Nan- odiffraction to Ptychography and Beyond, Microscopy and Microanalysis25, 563 (2019)

2019
[13]

P. M. Voyles and D. A. Muller, Fluctuation microscopy in the STEM, Ultramicroscopy93, 147 (2002)

2002
[14]

T. C. Pekin, J. Ding, C. Gammer, B. Ozdol, C. Ophus, M. Asta, R. O. Ritchie, and A. M. Minor, Direct mea- surement of nanostructural change during in situ defor- mation of a bulk metallic glass, Nature communications 10, 2445 (2019)

2019
[15]

B. H. Savitzky, S. E. Zeltmann, L. A. Hughes, H. G. Brown, S. Zhao, P. M. Pelz, T. C. Pekin, E. S. Barnard, J. Donohue, L. R. DaCosta,et al., py4DSTEM: A soft- ware package for four-dimensional scanning transmission electron microscopy data analysis, Microscopy and Mi- croanal...

2021
[16]

J. W. Allen, Gallium Arsenide as a Semi-insulator, Na- ture187, 403 (1960), publisher: Nature Publishing Group

1960
[17]

X.-Y. Liu, C. N. Singh, E. R. Kennedy, S. Huang, S. P. Fluckey, C. Matthews, and B. P. Uberuaga, Review of radiation-induced defects in GaAs, Journal of Applied Physics138, 070701 (2025)

2025
[18]

Srour and J

J. Srour and J. McGarrity, Radiation effects on micro- electronics in space, Proceedings of the IEEE76, 1443 (1988)

1988
[19]

S. J. Pearton, A. Aitkaliyeva, M. Xian, F. Ren, A. Khachatrian, A. Ildefonso, Z. Islam, M. A. Ja- far Rasel, A. Haque, A. Y. Polyakov, and J. Kim, Review—Radiation Damage in Wide and Ultra-Wide Bandgap Semiconductors, ECS Journal of Solid State Science and Technology10, 055008...

2021
[20]

J. M. Gibson and M. M. J. Treacy, Diminished medium- range order observed in annealed amorphous germa- nium, Physical Review Letters78, 1074 (1997), pub- lisher: American Physical Society

1997
[21]

Gerthsen and C

D. Gerthsen and C. B. Carter, Stacking-Fault En- ergies of GaAs, physica status solidi (a)136, 10.1002/pssa.2211360104 (1993)

1993 doi
[22]

E. R. Kennedy, A. Jones, Y. Wang, M. Pena, H. Kim, C. Song, F. Selim, B. P. Uberuaga, and S. Greer, Critical disconnect between structural and electronic recovery in amorphous GaAs during recrystallization, Communica- tions Materials 10.1038/s43246-026-01235-5 (2026)

2026 doi
[23]

Ivanov and J

M. Ivanov and J. Pereiro, Autoencoder latent space sen- sitivity to material structure in convergent-beam low en- ergy electron diffraction, Ultramicroscopy266, 114021 (2024)

2024
[24]

J. Chen, S. Sathe, C. Aggarwal, and D. Turaga, Outlier detection with autoencoder ensembles: 17th SIAM In- ternational Conference on Data Mining, SDM 2017, Pro- ceedings of the 17th SIAM International Conference on Data Mining, SDM 2017 Proceedings of the 17th SIAM Internation...

2017
[25]

T. Yoo, E. Hershkovitz, Y. Yang, F. da Cruz Gallo, M. V. Manuel, and H. Kim, Unsupervised machine learning and cepstral analysis with 4D-STEM for characterizing com- plex microstructures of metallic alloys, npj Computa- tional Materials10, 223 (2024), publisher: Nature Pub- li...

2024
[26]

D. A. Porter, K. E. Easterling, and K. E. Easterling, Phase Transformations in Metals and Alloys (Revised Reprint), 3rd ed. (CRC Press, 2009)

2009
[27]

R. D. Doherty, Nucleation and growth kinetics of differ- ent recrystallization texture components, Scripta Metal- lurgica19, 927 (1985)

1985
[28]

L. Shao, C. C. Wei, J. Gigax, A. Aitkaliyeva, D. Chen, B. H. Sencer, and F. A. Garner, Effect of defect imbal- ance on void swelling distributions produced in pure iron irradiated with 3.5 MeV self-ions, Journal of Nuclear Ma- terials453, 176 (2014)

2014
[29]

M. T. Janish, M. M. Schneider, J. A. Valdez, K. J. McClellan, D. D. Byler, Y. Wang, D. Chen, T. G. Holesinger, and B. P. Uberuaga, In-situ re-crystallization of heavily-irradiated Gd2Ti2O7, Acta Materialia194, 403 (2020)

2020
[30]

Metzner, C

P. Metzner, C. Sch¨ utte, and E. Vanden-Eijnden, Tran- sition Path Theory for Markov Jump Processes, Mult. Mod. Sim.7, 1192 (2008)

2008
[31]

B. Lin, Y. Jin, C. M. Hefferan, S. F. Li, J. Lind, R. M. Suter, M. Bernacki, N. Bozzolo, A. D. Rollett, and G. S. Rohrer, Observation of annealing twin nucleation at triple lines in nickel during grain growth, Acta Materialia 99, 63 (2015)

2015
[32]

H. Chen, B. Roux, and C. Chipot, Discovering Reaction Pathways, Slow Variables, and Committor Probabilities with Machine Learning, Journal of Chemical Theory and Computation19, 4414 (2023)

2023
[33]

R. A. Talmazan, J. Gamper, I. Castillo, T. S. Hofer, and M. Podewitz, Coupling causality and interpretable ma- chine learning to reveal the reaction coordinate of C–N coupling with a supramolecular Cu-calix[8]arene catalyst, Digital Discovery4, 2954 (2025)

2025
[34]

A. C. Y. Liu, G. R. Lumpkin, T. C. Petersen, J. Etheridge, and L. Bourgeois, Interpretation of angu- lar symmetries in electron nanodiffraction patterns from thin amorphous specimens, Acta Crystallographica Sec- tion A: Foundations and Advances71, 473 (2015), pub- lisher: Inte...

2015
[35]

Wochner, C

P. Wochner, C. Gutt, T. Autenrieth, T. Demmer, V. Bugaev, A. D. Ortiz, A. Duri, F. Zontone, G. Gr¨ ubel, and H. Dosch, X-ray cross correlation analysis uncovers hidden local symmetries in disordered matter, Proceed- ings of the National Academy of Sciences106, 11511 (2009), pu...

2009
[36]

E. R. Kennedy, S. M. Ribet, I. S. Winter, C. A. Kohnert, Y. Wang, K. C. Bustillo, C. Ophus, and B. K. Derby, Mapping strain and structural heterogeneities around bubbles in amorphous ionically conductive Bi2O3, Ma- terials & Design256, 114282 (2025)

2025
[37]

A. C. Y. Liu, M. J. Neish, G. Stokol, G. A. Buckley, L. A. Smillie, M. D. de Jonge, R. T. Ott, M. J. Kramer, and L. Bourgeois, Systematic mapping of icosahedral short- range order in a melt-spun Zr36Cu64 metallic glass, Phys- ical Review Letters110, 205505 (2013), publisher: A...

2013
[38]

F. J. Humphreys and M. Hatherly, Chapter 8 - Recrys- tallization of Ordered Materials, inRecrystallization and Related Annealing Phenomena (Second Edition)(Else- vier, Oxford, 2004) pp. 269–283

2004
[39]

X.-Y. Cui, S. P. Ringer, G. Wang, and Z. H. Stachurski, What should the density of amorphous solids be?, The Journal of Chemical Physics151, 194506 (2019)

2019
[40]

T. B. Light, Density of ”Amorphous” Ge, Physical Re- view Letters22, 999 (1969), publisher: American Physi- cal Society

1969
[41]

G. R. Bowman, V. S. Pande, and F. No´ e,An introduc- tion to Markov state models and their application to long timescale molecular simulation(Springer Science & Busi- ness Media, 2013)

2013
[42]

Sarich, J.-H

M. Sarich, J.-H. Prinz, and C. Sch¨ utte,An Introduction to Markov State Models and Their Application to Long Timescale Molecular Simulation, edited by G. R. Bow- man, V. S. Pande, and F. No´ e (Springer Netherlands, Dordrecht, 2014) pp. 23–44

2014
[43]

A. D. Rollett, D. J. Srolovitz, R. D. Doherty, and M. P. Anderson, Computer simulation of recrystallization in non-uniformly deformed metals, Acta Metallurgica37, 627 (1989)

1989
[44]

M. P. Anderson, D. J. Srolovitz, G. S. Grest, and P. S. Sahni, Computer simulation of grain growth—I. Kinetics, Acta Metallurgica32, 783 (1984)

1984
[45]

Tikare, E

V. Tikare, E. A. Holm, D. Fan, and L. Q. Chen, Com- parison of phase-field and Potts models for coarsening processes, Acta Materialia47, 363 (1998)

1998
[46]

K. A. Jackson, Monte Carlo Simulation of the Rapid Crystallization of Bismuth-Doped Silicon, Physical Re- view Letters75, 2530 (1995)

1995
[47]

Pedregosa, G

F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cour- napeau, M. Brucher, M. Perrot, and E. Duchesnay, Scikit-learn: Machine learning in Python, Journal of Ma- chine Lear...

2011
[48]

Mapping recrystallization trajectories in GaAs using latent space diffraction analysis

T. J. Boerner, S. Deems, T. R. Furlani, S. L. Knuth, and J. Towns, Access: Advancing innovation: Nsf’s ad- vanced cyberinfrastructure coordination ecosystem: Ser- vices & support, inPractice and Experience in Advanced Research Computing 2023: Computing for the Common Good, PEA...

2023

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.

[1] [1]

C. Zhao, H. Batiz, B. Yasar, H. Kim, W. Ji, M. C. Scott, D. C. Chrzan, and A. Javey, Tellurium Single-Crystal Arrays by Low-Temperature Evaporation and Crystal- lization, Advanced Materials33, 2100860 (2021)

2021

[2] [2]

Veli¸ sa, E

G. Veli¸ sa, E. Zarkadoula, D. Iancu, M. D. Mihai, A. Boulle, Y. Tong, D. Chen, Y. Zhang, and W. J. We- ber, Revealing a Pathway for Low-Temperature Recrys- tallization in Germanium, Advanced Sciencen/a, e07630 (2025)

2025

[3] [3]

M. M. J. Treacy and K. B. Borisenko, The Local Struc- ture of Amorphous Silicon, Science335, 950 (2012), pub- lisher: American Association for the Advancement of Sci- ence

2012

[4] [4]

L. A. M. Rosset, D. A. Drabold, and V. L. De- ringer, Signatures of paracrystallinity in amorphous sil- icon from machine-learning-driven molecular dynamics, Nature Communications16, 2360 (2025), publisher: Na- ture Publishing Group

2025

[5] [5]

G. Wu, C. Liu, Y.-Q. Yan, S. Liu, X. Ma, S. Yue, and Z.- W. Shan, Elemental partitioning-mediated crystalline-to- amorphous phase transformation under quasi-static de- formation, Nature Communications15, 1223 (2024)

2024

[6] [6]

K. F. Adam and D. P. Field, Analyzing recrystallization behavior of heterogeneous structures single-phase Al al- 16 loys, Materialia , 101190 (2021)

2021

[7] [7]

N. G. Rudawski, A. G. Lind, and T. P. Martin, Chap- ter Four - Defective Solid-Phase Epitaxial Growth of Si, inDefects in Semiconductors, Semiconductors and Semimetals, Vol. 91, edited by L. Romano, V. Privitera, and C. Jagadish (Elsevier, 2015) pp. 123–163

2015

[8] [8]

P. M. Voyles, N. Zotov, S. M. Nakhmanson, D. A. Drabold, J. M. Gibson, M. M. J. Treacy, and P. Keblin- ski, Structure and physical properties of paracrystalline atomistic models of amorphous silicon, Journal of Ap- plied Physics90, 4437 (2001)

2001