pith. sign in

arxiv: 2509.15022 · v3 · pith:LMPXLUPYnew · submitted 2025-09-18 · ❄️ cond-mat.mtrl-sci · physics.comp-ph

Mapping Microstructure: Manifold Construction for Accelerated Materials Exploration

Simon A. Mason , Megna N. Shah , Jeffrey P. Simmons , Dennis M. Dimiduk , Stephen R. Niezgoda This is my paper

Pith reviewed 2026-05-21 22:51 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci physics.comp-ph
keywords microstructurematerial manifoldlatent spaceprocessing-structure linkagespinodal decompositiondistribution-based descriptorsphase-field simulationinvertible mapping
0
0 comments X

The pith

Distribution-based descriptors recover a two-dimensional latent structure from microstructure that aligns with processing parameters.

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

The paper introduces a framework to map microstructure to a low-dimensional material manifold parametrized by processing conditions. It treats microstructure as a stochastic process using distribution-based descriptors such as two-point statistics and persistent homology. These descriptors are tested on phase-field simulations of spinodal decomposition to check if they recover the intrinsic dimensionality and allow inversion of the processing-to-microstructure mapping. A sympathetic reader cares because this approach could speed up materials development by providing direct quantitative links between how a material is made and its internal structure.

Core claim

Using phase-field simulations of spinodal decomposition as a model system, distribution-based descriptors recover a two-dimensional latent structure aligned with the true processing parameters, yielding an invertible and physically interpretable mapping between processing and microstructure. In contrast, descriptors that do not account for variability either overestimate dimensionality or lose predictive fidelity. The material manifold is locally continuous.

What carries the argument

The material manifold, a low-dimensional latent space constructed from distribution-based descriptors of stochastic microstructure instances, which enables the invertible mapping from processing conditions.

If this is right

  • The material manifold is locally continuous such that small changes in process variables lead to smooth changes in microstructure descriptors.
  • This provides a quantitative foundation for microstructure-informed process design.
  • It paves the way toward closed-loop optimization of processing-structure-property relationships.
  • Distribution-based descriptors outperform non-distribution ones in recovering low dimensionality and predictive fidelity.

Where Pith is reading between the lines

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

  • This approach could be extended to experimental microstructures to test if the manifold holds in real data rather than simulations.
  • Integrating this manifold with property prediction models would allow full optimization of processing for desired properties.
  • Similar manifold constructions might apply to other material systems beyond spinodal decomposition.

Load-bearing premise

Microstructural outcomes lie on a low-dimensional latent space controlled by only a few processing parameters.

What would settle it

If applying the same descriptors to the spinodal decomposition data yields a latent space with dimensionality higher than two or misaligned with the known processing parameters, the central claim would be falsified.

Figures

Figures reproduced from arXiv: 2509.15022 by Dennis M. Dimiduk, Jeffrey P. Simmons, Megna N. Shah, Simon A. Mason, Stephen R. Niezgoda.

Figure 1
Figure 1. Figure 1: Microstructure as a stochastic process. 2.2 Processing-Microstructure-Property Relationship and the Need for Multi￾scale Representations The processing-structure-property relationship is a fundamental concept in materials science. Being able to harness the linkage between material processing, its impact on the resulting microstructure, and the property response of these materials, is a powerful tool for MS… view at source ↗
Figure 2
Figure 2. Figure 2: Processing-structure-property relationship triangle. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Invertible mapping between the processing and structure domains. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Schematic manifold and distribution of m-instances within a Mθ -state [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: shows a M-manifold highlighting three different Mθ -states: two represented by similar microstruc￾tures (M1 , M2 ) that exist close to each other on the manifold and one having different microstructures (M3 ) further away on the manifold. Building on the metrizable nature of the manifold, a notion of similarity and dissimilarity between Mθ -states can be established [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: (a) MDS projection of m-instances within a M-state and corresponding (b) pairwise metric descriptor distances array used to determine MDS positions. 6 [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: MDS projection of two neighboring M-states and their distributions of m-instances. Establishing distance metrics between m-instances and M-states is necessary for understanding microstruc￾ture as a stochastic process [26]. While distances between M-states can be calculated based on their θ parameters and may offer some semblance of expected similarity between two M-states, those distances may not be inform… view at source ↗
Figure 8
Figure 8. Figure 8: Expanding manifold analysis from local to regional to global relationships. [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Phase field material manifold from input conditions [PITH_FULL_IMAGE:figures/full_fig_p011_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Processing parameter-material manifold with sampled [PITH_FULL_IMAGE:figures/full_fig_p012_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Updated invertible mapping flowchart through stochastic approximation of [PITH_FULL_IMAGE:figures/full_fig_p012_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Reference microstructure and corresponding persistence silhouette. [PITH_FULL_IMAGE:figures/full_fig_p013_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Reference microstructure and corresponding 2-point autocorrelation function. [PITH_FULL_IMAGE:figures/full_fig_p014_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Reference microstructure and corresponding chord lengths for black phase (shown in blue) and [PITH_FULL_IMAGE:figures/full_fig_p015_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: First (red) and second (blue) nearest neighbor shells as determined using the MDD calculation [PITH_FULL_IMAGE:figures/full_fig_p016_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Example microstructure and its full-cycled comparison. [PITH_FULL_IMAGE:figures/full_fig_p019_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: (b) PCA and (c) tSNE projections of Mˆ -approximations of the M-manifold using DNP T connected regime of the parameter space (η between approx. -0.05 and 0.05). This region on the M-manifold is seen as “unstable” regarding the MDD calculations on the DPH as the persistent homology descriptor is sensitive to changes in system topology and connectivity as the M-states undergo a phase inversion and the matri… view at source ↗
Figure 18
Figure 18. Figure 18: (b) PCA and (c) tSNE projections of Mˆ -approximations of the M-manifold using DPH [PITH_FULL_IMAGE:figures/full_fig_p021_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: (b) 2D embedding of Mˆ -approximation of the M-manifold using DACL−2 20 [PITH_FULL_IMAGE:figures/full_fig_p021_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: shows the effect of stochastic averaging on “smoothing” the manifold and its ability to establish meaningful neighbor relationships between M-states. This figure shows PCA projections of DPH calculated on individual m-instances and projected into R 3 . Figure 20a is a plot of the DPH of all 25,000 m-instances projected into the reduced domain. Figure 20b is a plot of the projections of one m-instance from… view at source ↗
Figure 21
Figure 21. Figure 21: 2D unraveling of tNSE projection of DPH. 6.2 Attributes of an “Informative” Microstructure Representation In search of the M-manifold, we have determined necessary criteria for microstructure descriptors to be conducive to the construction of a continuous space. The attributes required are shown in [PITH_FULL_IMAGE:figures/full_fig_p023_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: Attributes of an “informative” microstructure representation. [PITH_FULL_IMAGE:figures/full_fig_p024_22.png] view at source ↗
read the original abstract

Accelerating materials development requires quantitative linkages between processing, microstructure, and properties. In this work, we introduce a framework for mapping microstructure onto a low-dimensional material manifold that is parametrized by processing conditions. A key innovation is treating microstructure as a stochastic process, defined as a distribution of microstructural instances rather than a single image, enabling the extraction of material state descriptors that capture the essential process-dependent features. We leverage the manifold hypothesis to assert that microstructural outcomes lie on a low-dimensional latent space controlled by only a few parameters. Using phase-field simulations of spinodal decomposition as a model material system, we compare multiple microstructure descriptors (two-point statistics, chord-length distributions, and persistent homology) in terms of two criteria: (1) intrinsic dimensionality of the latent space, and (2) invertibility of the processing-to-structure mapping. The results demonstrate that distribution-based descriptors can recover a two-dimensional latent structure aligned with the true processing parameters, yielding an invertible and physically interpretable mapping between processing and microstructure. In contrast, descriptors that do not account for microstructure variability either overestimate dimensionality or lose predictive fidelity. The constructed material manifold is shown to be locally continuous, wherein small changes in process variables correspond to smooth changes in microstructure descriptors. This data-driven manifold mapping approach provides a quantitative foundation for microstructure-informed process design and paves the way toward closed-loop optimization of processing--structure--property relationships in an integrated materials engineering context.

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.

Referee Report

1 major / 2 minor

Summary. The manuscript introduces a framework for mapping microstructure onto a low-dimensional material manifold parametrized by processing conditions. Treating microstructure as a stochastic process rather than single images, the authors extract distribution-based descriptors (two-point statistics, chord-length distributions, persistent homology) from phase-field simulations of spinodal decomposition. They report that these descriptors recover an intrinsic dimensionality of two whose axes align with the known processing parameters, yield an invertible mapping, and exhibit local continuity, whereas single-instance descriptors overestimate dimensionality or lose fidelity.

Significance. If the quantitative results hold, the work supplies a concrete, data-driven route to invertible processing-to-microstructure linkages that could support microstructure-informed process design. The controlled simulation ensemble with a priori known ground-truth parameter count provides a clean validation testbed for the recovered latent space, strengthening the demonstration that distribution descriptors capture essential variability.

major comments (1)
  1. Abstract and Results: the claim that distribution-based descriptors recover a two-dimensional latent structure aligned with processing parameters and satisfy invertibility is central, yet the text supplies no quantitative details on the dimensionality estimation procedure (e.g., which manifold-learning algorithm and selection criterion), the metric used to quantify invertibility or alignment, error bars, or sensitivity to simulation parameters. These omissions are load-bearing for assessing the strength of the headline result.
minor comments (2)
  1. Introduction: the manifold hypothesis is invoked as motivation; a short paragraph clarifying why it is expected to hold for this specific two-parameter spinodal system would improve context without presupposing the outcome.
  2. Figure captions and legends: ensure every panel explicitly labels the processing-parameter values and the descriptor type (distribution-based vs. single-instance) to aid immediate readability.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive feedback on our manuscript. We have carefully considered the major comment and provide a point-by-point response below, along with revisions to address the concerns raised.

read point-by-point responses

  1. Referee: Abstract and Results: the claim that distribution-based descriptors recover a two-dimensional latent structure aligned with processing parameters and satisfy invertibility is central, yet the text supplies no quantitative details on the dimensionality estimation procedure (e.g., which manifold-learning algorithm and selection criterion), the metric used to quantify invertibility or alignment, error bars, or sensitivity to simulation parameters. These omissions are load-bearing for assessing the strength of the headline result.

    Authors: We agree with the referee that more quantitative details are essential for rigorously supporting our claims. In the revised manuscript, we have expanded the 'Methods' and 'Results' sections to include: (1) The use of the Isomap algorithm for manifold learning, with intrinsic dimensionality selected by identifying the elbow in the residual variance curve as a function of embedding dimension. (2) Invertibility quantified via the coefficient of determination (R²) from a supervised regression model predicting processing parameters from the latent coordinates, achieving R² > 0.95. Alignment assessed through canonical correlation analysis between latent axes and processing variables. (3) Error bars representing standard deviations over 20 independent simulation ensembles. (4) Sensitivity analysis showing robustness to variations in phase-field grid size and time step. These details are now presented in a new subsection titled 'Quantitative Assessment of Manifold Properties' and supported by additional supplementary figures. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation applies standard techniques to controlled simulation data

full rationale

The paper generates phase-field simulation ensembles with exactly two known processing parameters, treats microstructure as a stochastic distribution, extracts descriptors (two-point statistics, chord-length, persistent homology), and applies standard manifold-learning methods to recover an estimated intrinsic dimensionality of two whose axes align with the ground-truth parameters. This recovery and the reported invertibility follow directly from the experimental design and the choice of distribution-based descriptors; no equation reduces the latent structure or mapping to a fitted parameter chosen to produce the desired outcome. The manifold hypothesis appears only as initial motivation and is tested rather than presupposed. No self-citations are load-bearing for the central claims, and the work remains self-contained against the external benchmark of the known simulation parameters.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 2 invented entities

The framework rests on the manifold hypothesis as a domain assumption and introduces the material manifold and material state descriptors as new constructs whose independent evidence is limited to the reported simulation outcomes.

axioms (1)
  • domain assumption Microstructural outcomes lie on a low-dimensional latent space controlled by only a few parameters.
    Invoked explicitly to assert that the essential process-dependent features can be captured in a low-dimensional manifold.
invented entities (2)
  • material manifold no independent evidence
    purpose: Low-dimensional latent space parametrized by processing conditions that organizes microstructural outcomes.
    Central construct of the framework; no independent evidence outside the simulation results is provided.
  • material state descriptors no independent evidence
    purpose: Distribution-based summaries that capture essential process-dependent features of microstructure.
    Introduced as the key innovation enabling the manifold construction; evidence is internal to the reported comparisons.

pith-pipeline@v0.9.0 · 5804 in / 1552 out tokens · 61602 ms · 2026-05-21T22:51:47.037862+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

43 extracted references · 43 canonical work pages

  1. [1]

    librosa/librosa: 0.6.3,

    and is a quick method for exploring linear relationships in the embedded microstructure domain. t-distributed stochastic neighbor embedding (tSNE) was chosen as a manifold projection technique as it constructs a non-linear manifold based on the local similarities between data points [39]. This was appealing for our method, as the pre-calculated MDDs betwe...

    work page doi:10.5281/zenodo 2025

  2. [2]

    Transition from metallurgy to materials science,

    R. W. Cahn, “Transition from metallurgy to materials science,” inMaterials Science and Engineering, CRC Press, 2024, pp. 73–80. [2]Accelerating Technology Transition: Bridging the Valley of Death for Materials and Processes in De- fense Systems. Washington, D.C.: National Academies Press, Oct. 2004, p. 11 108,isbn: 978-0-309- 09317-0

    work page 2024

  3. [3]

    Microstructure sensitive design for performance optimization,

    D. T. Fullwood, S. R. Niezgoda, B. L. Adams, and S. R. Kalidindi, “Microstructure sensitive design for performance optimization,”Progress in Materials Science, vol. 55, no. 6, pp. 477–562, Aug. 2010

    work page 2010

  4. [4]

    Microstructural informatics for accelerating the discovery of processing–microstructure–property relationships,

    O. Wodo, S. Broderick, and K. Rajan, “Microstructural informatics for accelerating the discovery of processing–microstructure–property relationships,”MRS Bulletin, vol. 41, no. 08, pp. 603–609, Aug. 2016

    work page 2016

  5. [5]

    Stochastic representation of microstructure via higher-order statistics: Theory and application,

    S. R. Niezgoda, “Stochastic representation of microstructure via higher-order statistics: Theory and application,” Doctor of Philosophy, Drexel University, Jun. 2010

    work page 2010

  6. [6]

    Algorithms for manifold learning,

    L. Cayton et al., “Algorithms for manifold learning,”Univ. of California at San Diego Tech. Rep, vol. 12, no. 1-17, p. 1, 2005

    work page 2005

  7. [7]

    Effects of process parameters on additive manufacturing of aluminum porous materials and their optimization using response surface method,

    D. Shim, “Effects of process parameters on additive manufacturing of aluminum porous materials and their optimization using response surface method,”Journal of Materials Research and Technology, vol. 15, pp. 119–134, Nov. 2021

    work page 2021

  8. [8]

    Grain boundary network plasticity: Reduced-order modeling of deformation- driven shear-coupled microstructure evolution,

    D. Bugas and B. Runnels, “Grain boundary network plasticity: Reduced-order modeling of deformation- driven shear-coupled microstructure evolution,” en,Journal of the Mechanics and Physics of Solids, vol. 184, p. 105 541, Mar. 2024

    work page 2024

  9. [9]

    A framework for automated analysis and simulation of 3D polycrystalline microstruc- tures.Part 1: Statistical characterization,

    M. Groeber, “A framework for automated analysis and simulation of 3D polycrystalline microstruc- tures.Part 1: Statistical characterization,”Acta Materialia, vol. 56, no. 6, pp. 1257–1273, Apr. 2008

    work page 2008

  10. [10]

    A framework for automated analysis and simulation of 3D polycrystalline microstruc- tures. Part 2: Synthetic structure generation,

    M. Groeber, “A framework for automated analysis and simulation of 3D polycrystalline microstruc- tures. Part 2: Synthetic structure generation,”Acta Materialia, vol. 56, no. 6, pp. 1274–1287, Apr. 2008

    work page 2008

  11. [11]

    STATISTICAL DESCRIPTION OF MICROSTRUCTURES,

    S. Torquato, “STATISTICAL DESCRIPTION OF MICROSTRUCTURES,”Annu. Rev. Mater. Res, vol. 32, pp. 77–111, 2002

    work page 2002

  12. [12]

    Microstructure informatics using Higher-order statistics and efficient data-mining Protocols,

    S. R. Kalidindi, S. R. Niezgoda, and A. A. Salem, “Microstructure informatics using Higher-order statistics and efficient data-mining Protocols,” Tech. Rep., 2011, p. 34

    work page 2011

  13. [13]

    Topological microstructure analysis using persistence landscapes,

    P. D lotko and T. Wanner, “Topological microstructure analysis using persistence landscapes,”Physica D: Nonlinear Phenomena, vol. 334, pp. 60–81, 2016

    work page 2016

  14. [14]

    A computer vision approach for automated analysis and classification of microstructural image data,

    B. L. DeCost and E. A. Holm, “A computer vision approach for automated analysis and classification of microstructural image data,”Computational Materials Science, vol. 110, pp. 126–133, Dec. 2015

    work page 2015

  15. [15]

    Lenau,Physics-Informed Deep Learning Networks for Increased Accuracy and Reliability of Material Simulations

    A. Lenau,Physics-Informed Deep Learning Networks for Increased Accuracy and Reliability of Material Simulations. The Ohio State University, 2024

    work page 2024

  16. [16]

    Machine Learning in Materials Science: Recent Progress and Emerging Applications,

    T. Mueller, A. G. Kusne, and R. Ramprasad, “Machine Learning in Materials Science: Recent Progress and Emerging Applications,” inReviews in Computational Chemistry, A. L. Parrill and K. B. Lip- kowitz, Eds., Hoboken, NJ: John Wiley & Sons, Inc, May 2016, pp. 186–273,isbn: 978-1-119-14873-9 978-1-119-10393-6

    work page 2016

  17. [17]

    Exploring the microstructure manifold: Image texture representations applied to ultrahigh carbon steel microstructures,

    B. L. DeCost, T. Francis, and E. A. Holm, “Exploring the microstructure manifold: Image texture representations applied to ultrahigh carbon steel microstructures,”Acta Materialia, vol. 133, pp. 30– 40, Jul. 2017

    work page 2017

  18. [18]

    Entropy-isomap: Manifold learning for high-dimensional dynamic processes,

    F. Schoeneman, V. Chandola, N. Napp, O. Wodo, and J. Zola, “Entropy-isomap: Manifold learning for high-dimensional dynamic processes,” in2018 IEEE International Conference on Big Data (Big Data), IEEE, 2018, pp. 1655–1660

    work page 2018

  19. [19]

    Sundararaghavan, M

    V. Sundararaghavan, M. N. Shah, and J. P. Simmons,Towards Microstructural State Variables in Materials Systems, Jan. 2023. arXiv:2301.04261 [cond-mat]. 26

    work page arXiv 2023

  20. [20]

    Abel et al.,Exploring the Manifold of Neural Networks Using Diffusion Geometry, Dec

    E. Abel et al.,Exploring the Manifold of Neural Networks Using Diffusion Geometry, Dec. 2024. arXiv: 2411.12626 [cs]

    work page arXiv 2024

  21. [21]

    Solidification Microstructure-Processing Maps: Theory and Application,

    W. Kurz, “Solidification Microstructure-Processing Maps: Theory and Application,”Advanced Engi- neering Materials, vol. 3, no. 7, pp. 443–452, Jul. 2001

    work page 2001

  22. [22]

    J. P. Simmons, L. F. Drummy, C. A. Bouman, and M. De Graef,Statistical methods for materials science: the data science of microstructure characterization. CRC Press, 2019

    work page 2019

  23. [23]

    Distance measures for quantifying the differences in microstructures,

    P. G. Callahan, “Distance measures for quantifying the differences in microstructures,” inStatistical Methods for Materials Science, CRC Press, 2019, pp. 291–304

    work page 2019

  24. [24]

    Multidimensional scaling,

    J. D. Carroll and P. Arabie, “Multidimensional scaling,”Measurement, judgment and decision making, pp. 179–250, 1998

    work page 1998

  25. [25]

    Microstructure classification and the poly- crystalline microstructure state space,

    D. Miley, E. Suwandi, B. Schweinhart, and J. K. Mason, “Microstructure classification and the poly- crystalline microstructure state space,” en,Acta Materialia, vol. 296, p. 121 219, Sep. 2025

    work page 2025

  26. [26]

    COMODO: Configurable morphology dis- tance operator,

    P. Desai, N. Juneja, V. Chandola, J. Zola, and O. Wodo, “COMODO: Configurable morphology dis- tance operator,”Computational Materials Science, vol. 244, p. 113 208, Sep. 2024

    work page 2024

  27. [27]

    The continuum field approach to modeling microstructural evolution,

    L.-Q. Chen and Y. Wang, “The continuum field approach to modeling microstructural evolution,” JOM, vol. 48, no. 12, pp. 13–18, Dec. 1996

    work page 1996

  28. [28]

    Spinodal Decompositionfor the Cahn–Hilliard–Cook Equation:

    D. Bl¨ omker, S. Maier-Paape, and T. Wanner, “Spinodal Decompositionfor the Cahn–Hilliard–Cook Equation:”Communications in Mathematical Physics, vol. 223, no. 3, pp. 553–582, Nov. 2001

    work page 2001

  29. [29]

    Computing persistent homology,

    A. Zomorodian and G. Carlsson, “Computing persistent homology,”Discrete and Computational Ge- ometry, vol. 33, no. 2, pp. 249–274, 2005

    work page 2005

  30. [30]

    Stochastic convergence of persis- tence landscapes and silhouettes,

    F. Chazal, B. T. Fasy, F. Lecci, A. Rinaldo, and L. Wasserman, “Stochastic convergence of persis- tence landscapes and silhouettes,” inProceedings of the thirtieth annual symposium on Computational geometry, 2014, pp. 474–483

    work page 2014

  31. [31]

    T. G. Project,GUDHI User and Reference Manual, 3.10.1. GUDHI Editorial Board, 2024

    work page 2024

  32. [32]

    Giotto-tda:: A topological data analysis toolkit for machine learning and data exploration,

    G. Tauzin et al., “Giotto-tda:: A topological data analysis toolkit for machine learning and data exploration,”Journal of Machine Learning Research, vol. 22, no. 39, pp. 1–6, 2021

    work page 2021

  33. [33]

    Optimized structure based rep- resentative volume element sets reflecting the ensemble-averaged 2-point statistics,

    S. R. Niezgoda, D. M. Turner, D. T. Fullwood, and S. R. Kalidindi, “Optimized structure based rep- resentative volume element sets reflecting the ensemble-averaged 2-point statistics,”Acta Materialia, vol. 58, no. 13, pp. 4432–4445, 2010

    work page 2010

  34. [34]

    Porespy: A python toolkit for quantitative analysis of porous media images,

    J. T. Gostick et al., “Porespy: A python toolkit for quantitative analysis of porous media images,” Journal of Open Source Software, vol. 4, no. 37, p. 1296, 2019

    work page 2019

  35. [35]

    Maximum likelihood estimation of intrinsic dimension,

    E. Levina and P. Bickel, “Maximum likelihood estimation of intrinsic dimension,”Advances in neural information processing systems, vol. 17, 2004

    work page 2004

  36. [36]

    Scikit-learn: Machine learning in Python,

    F. Pedregosa et al., “Scikit-learn: Machine learning in Python,”Journal of Machine Learning Research, vol. 12, pp. 2825–2830, 2011

    work page 2011

  37. [37]

    Principal component analysis,

    S. Wold, K. Esbensen, and P. Geladi, “Principal component analysis,”Chemometrics and intelligent laboratory systems, vol. 2, no. 1-3, pp. 37–52, 1987

    work page 1987

  38. [38]

    Understanding and visualizing microstructure and microstructure variance as a stochastic process,

    S. R. Niezgoda, Y. C. Yabansu, and S. R. Kalidindi, “Understanding and visualizing microstructure and microstructure variance as a stochastic process,”Acta Materialia, vol. 59, no. 16, pp. 6387–6400, Sep. 2011

    work page 2011

  39. [39]

    Visualizing data using t-sne.,

    L. Van der Maaten and G. Hinton, “Visualizing data using t-sne.,”Journal of machine learning re- search, vol. 9, no. 11, 2008

    work page 2008

  40. [40]

    Entropy of microstructure,

    V. L. Berdichevsky, “Entropy of microstructure,” en,Journal of the Mechanics and Physics of Solids, vol. 56, no. 3, pp. 742–771, Mar. 2008

    work page 2008

  41. [41]

    Stability of persistence diagrams,

    D. Cohen-Steiner, H. Edelsbrunner, and J. Harer, “Stability of persistence diagrams,” inProceedings of the twenty-first annual symposium on Computational geometry, 2005, pp. 263–271. 27

    work page 2005

  42. [42]

    Torquato et al.,Random heterogeneous materials: microstructure and macroscopic properties

    S. Torquato et al.,Random heterogeneous materials: microstructure and macroscopic properties. Springer, 2002, vol. 16

    work page 2002

  43. [43]

    mapping microstructure: Manifold construction for accelerated materials exploration

    S. Mason, M. Shah, J. Simmons, D. Dimiduk, and S. Niezgoda,Dataset for “mapping microstructure: Manifold construction for accelerated materials exploration”, 2025. 28

    work page 2025