Pith. sign in

REVIEW 4 major objections 6 minor 50 references

From MLIPs to Microstructure: A High-Throughput Computational Framework to Design Spinodal Alloys in High-Dimensional Composition Spaces via Analytic Derivatives of CALPHAD Model Predictions

T0 review · 4 major / 6 minor · reviewed 2026-08-01 · deepseek-v4-flash

Pith's one-line read An open-source workflow builds CALPHAD Gibbs energies from a machine-learned interatomic potential, computes Hessians analytically via Jansson derivatives, and adds coherency strain to map spinodal decomposition and microstructure across th

desk verdict A genuinely useful open-source integration of analytic Hessians, MLIP-derived CALPHAD, and elasto-chemical phase field, but the quaternary validation fails on the exact quantities the coherency correction is meant to predict, and the paper is honest about it. read the letter →

arxiv 2607.20077 v1 pith:7RB5SVMI submitted 2026-07-22 cond-mat.mtrl-sci

classification cond-mat.mtrl-sci
keywords spinodaldecompositionCALPHADmachinelearninginteratomicpotentialanalyticHessianJanssonderivativescoherencystrainHf-Nb-Ti-Vphasefield
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 aims to show that two bottlenecks to high-throughput spinodal-alloy design — the lack of reliable thermodynamic databases for many refractory multicomponent systems and the cost of numerical Hessian evaluation — can both be removed. It replaces DFT-based database construction with a machine-learned interatomic potential that generates CALPHAD-style Gibbs energies, and replaces finite-difference second derivatives with analytic Jansson derivatives of the equilibrium chemical potentials. To these chemical driving forces it adds a Cahn coherency-strain correction built from machine-learned elastic constants, and feeds the result into elasto-chemical phase-field simulations. Applied to the BCC phase of Hf-Nb-Ti-V, the workflow predicts that coherency systematically shrinks the spinodal region — closing the miscibility gap at symmetric Nb-V and confining accessible ternary compositions to Hf ≤ 0.25 — while matching the experimentally observed precipitate phase fraction (≈1/3) and modulation wavelength (50–100 nm). The paper is candid about three quantitative misses: Hf partitions to the matrix instead of the precipitate, the coherency misfit is overestimated (8.7% vs 1.6%/0.54%), and the predicted <100> modulation direction disagrees with the observed <110>.

What carries the argument

The load-bearing object is the analytic Hessian of the molar Gibbs energy, computed on the composition simplex by the Jansson derivative method — an adjoint-like technique that differentiates single-phase equilibrium chemical potentials directly, avoiding finite-difference truncation error near the simplex edges. To this chemical Hessian the paper adds the Cahn coherency correction H^coh_ab = ∂²G/∂x_a∂x_b + 2 η_a η_b Y V_m, where η are lattice-parameter derivatives from a bond-length model and Y is a cubic-elasticity factor built from MLIP-computed C11, C12, C44. The gauge-invariant eigenvalues of this coherent Hessian locate the spinodal; the same Hessian, stored as smooth surrogates, drive

What would settle it

A direct DFT calculation of the cubic elastic constants of the CALPHAD tie-line matrix composition (about Nb30Hf37Ti31V2) at 0 K would settle the central elastic claims: if the Zener ratio A_Z comes out below 1, the predicted <100> modulation direction for the quaternary is an artifact of the MLIP; if A_Z exceeds 1, the framework's elastic machinery is corroborated, and the experimental <110> direction must be explained by thermal history. Equally decisive would be an equilibrated (long-annealed, not as-cast) specimen of equiatomic HfNbTiV, measuring whether Hf partitions to the precipitate as

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that spinodal stability and microstructure evolution in a quaternary refractory alloy can be predicted from a single open-source pipeline in which (i) a universal machine-learned interatomic potential supplies formation energies for special quasirandom structures, (ii) these are parameterized into a CALPHAD Gibbs energy model, (iii) gauge-invariant Hessian eigenvalues are computed analytically via the Jansson derivative method rather than by finite differences, and (iv) the Cahn coherency strain term, built from MLIP elastic constants and analytic bond-based lattice-parameter derivatives, is added to obtain the coherent spinodal. The authors show that

Load-bearing premise

The whole prediction rests on the untested assumption that the universal machine-learned interatomic potential yields quantitatively correct formation energies, bond lengths, and cubic elastic constants for this specific BCC Hf-Nb-Ti-V chemistry — the paper's Section 3.4 notes the quaternary SQS elastic constants have not been validated against DFT or experiment, and errors there flow directly into the coherency misfit and Zener-ratio predictions.

Editorial extensions

If this is right

  • Coherency strain can close a miscibility gap entirely at symmetric compositions (Nb50V50), so purely chemical spinodal maps overestimate instability in elastically stiff alloys; design maps must include the elastic term.
  • Designers can now screen high-dimensional composition spaces for spinodal regions without commercial CALPHAD databases, using affine projections to visualize all instabilities on one 2D plot.
  • Hf acts as a dual-action lever in Nb-V-based alloys: it deepens the chemical driving force while raising the coherency penalty, defining a composition window (x_Hf ≤ 0.25) in which spinodal microstructures are accessible.
  • The analytic-Hessian phase-field surrogate eliminates finite-difference errors in spinodal wavelength (λ* ∝ |H|^{-1/2}) and stable time-step estimates, making predicted microstructural length scales more trustworthy.
  • Three testable quantitative disagreements remain, each traced to a specific input: Hf partitioning, coherency misfit, and the Zener-anisotropy-driven modulation direction.

Reading between the lines

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

  • If the Zener-anisotropy inversion is confirmed by DFT (A_Z > 1 at the tie-line matrix composition), Ti content becomes a design control for switching spinodal morphology between <110> lamellae and <100> basket patterns in refractory high-entropy alloys.
  • Fine-tuning the universal potential on BCC Hf and Ti, or adding temperature-dependent elastic corrections from short molecular-dynamics runs, would likely bring the misfit and modulation-direction predictions into agreement without changing the framework's architecture.
  • The same Hessian machinery could be extended to third-order derivatives via second-order Jansson derivatives, enabling analytic critical-point (consolute) searches in high-dimensional composition spaces.
  • Because the workflow is open-source and whitebox, it invites inversion: instead of screening compositions for a given microstructure, one could optimize composition directly against coherent-Hessian eigenvalues as a target property.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. Kunselman et al. present an open-source workflow for predicting spinodal decomposition in multicomponent alloys. The workflow (i) constructs CALPHAD Gibbs-energy models from ORB-v2 MLIP SQS energies via PhaseForge; (ii) computes analytic Gibbs-energy Hessians using Jansson derivatives; (iii) adds a Cahn-type coherency correction using MLIP bond lengths and cubic elastic constants; (iv) visualizes stability via affine projections; and (v) feeds the CALPHAD model and elastic data into a surrogate-based elasto-chemical Cahn-Hilliard phase-field solver. The method is demonstrated on BCC Hf-Nb-Ti-V, progressing from binary Nb-V to the equiatomic quaternary, with comparison to the experiments of An et al. The authors report agreement in phase fraction (~1/3) and modulation wavelength (50-100 nm), while acknowledging discrepancies in Hf partitioning, coherency misfit (8.7% vs 1.6%/0.54%), and modulation direction (<100> vs <110>).

Significance. The paper's main strengths are the open-source implementation, the clear exposition of gauge-invariant Hessian analysis, and the use of analytic Jansson derivatives to sidestep finite-difference Hessian errors. The MLIP-to-CALPHAD coupling and the coherency-strain augmentation address a real gap in high-throughput spinodal design. If the method were quantitatively validated, it would be a useful tool for refractory multi-principal-element alloys. However, the decisive quaternary validation currently fails on the very quantities that the new coherency physics controls, and the surrogate 'exactness' demonstration is circular. The significance as a predictive design tool is therefore not yet established; the paper is more convincing as a methodological framework with an honest stress test.

major comments (4)
  1. [Sec. 2.3.5, Fig. 1, Table 1] The RMSE=0 result for Jansson/hybrid is circular. The 'CALPHAD data' markers are the same stored points used to construct the G, μ, and Jansson-derivative surrogates; any interpolant that passes through its training data yields zero error at those points by construction. This validates interpolation consistency, not derivative accuracy at off-grid compositions. Since the phase-field simulations (Eqs. 24-25) evaluate the surrogates at arbitrary grid points, the claim that the Jansson mode 'eliminates' derivative error needs a held-out or analytic-derivative comparison. Without it, the surrogate error budget for the phase-field results is unknown.
  2. [Sec. 3.4, Table 6, Eq. (16)] All three quantities controlled by the Cahn coherency correction disagree with experiment in the quaternary: coherency misfit 8.7% vs 1.6% (XRD)/0.54% (TEM), Zener ratio A_Z=1.52 predicting <100> modulation vs observed <110>, and Hf partitioned to the matrix instead of the precipitate. Section 3.4 explicitly states that the SQS/ORB-MLIP elastic constants 'have not been validated against DFT or experimental measurements for the specific quaternary compositions predicted by the CALPHAD tie-line.' Since these inputs enter the coherent Hessian, the stability maps (Fig. 3), eigenvalue curves (Fig. 4), and phase-field morphologies (Fig. 7) all shift with this unvalidated input. The central predictive claim is therefore unsupported at the decisive validation point; independent DFT validation of lattice parameters and elastic constants, or a substantial softening of the predictive claims, is req
  3. [Sec. 2.3.1, Figs. 5-7] The gradient-energy coefficients κ_i and mobilities M_ij are input parameters set to round numbers (e.g., κ=1e-12 J m^2/mol, M=5e-21 m^5 J^-1 s^-1) without calibration or sensitivity analysis. The predicted modulation wavelength scales as λ*=4π√(κ/|H|), so the reported 50-100 nm agreement depends directly on the arbitrary κ value. The phase-field morphology comparison therefore does not independently validate the thermodynamic/elastic model unless the sensitivity of the results to κ and M is quantified.
  4. [Sec. 3.4] The comparison is made between 500 °C equilibrium simulations and an as-cast experimental alloy. The paper attributes the Hf partitioning and possibly other discrepancies to non-equilibrium solidification, but this explanation is not tested. If the as-cast microstructure is not equilibrated at 500 °C, the partial agreement in phase fraction and wavelength may be partially fortuitous, and the disagreements do not cleanly test the model. A controlled isothermal-aging experiment or a solidification-aware initial condition would be needed to make the validation well-posed.
minor comments (6)
  1. [Table 1] The row entries list three numerical RMSE values (e.g., 302, 442, 219) but the header has only two RMSE columns (RMSE μ, RMSE H). Clarify the third quantity or correct the table.
  2. [Eq. (14)] The symbols C11, C12, C44 are used before being introduced; define the elastic constants at first use.
  3. [Eq. (29)] The stable time step uses |λ_max(H)| evaluated 'in a neighborhood of x0'; for a spinodal system the Hessian eigenvalues vary strongly with composition. State the precise sampling or use a global bound.
  4. [Fig. 8] The three composition paths are not fully specified; state the paths explicitly in the caption (e.g., exactly which compositions vary for each curve).
  5. [Sec. 2.3.2] The unconditional-stability claim for Eq. (33) is based on a local quadratic convex splitting with A_i taken at x0; off-diagonal bulk terms are explicit, so the scheme is not proven unconditionally stable in the global nonlinear setting. Rephrase as 'linearly stable' or provide a global construction.
  6. [Minor typographical issues] 'V oigt-Reuss-Hill' in Section 2.2, 'Anet al.' spacing throughout, and 'T emperature' in Figure 4 axis label should be corrected.

Circularity Check

1 steps flagged · score 4.0 of 10

RMSE=0 'validation' of the Jansson Hessian reduces by construction; the central spinodal/microstructure predictions remain independent and are openly tested against experiment.

  1. fitted input called prediction [Section 2.3.5, Fig. 1 and Table 1 (surrogate construction in Section 2.3.3)]
    "Panel (b) shows the relative chemical potential µNb−µV: the fd mode deviates visibly from the CALPHAD reference (RMSE=302 J mol−1) while jansson and hybrid reproduce the stored values exactly (RMSE=0 J mol−1). Panel (c) ... the fd mode yields an RMSE of 442 kJ mol−1 relative to the CALPHAD reference, entirely due to truncation error amplification near the simplex boundaries, whereas jansson matches exactly."

    Section 2.3.3 states that at each grid point the molar Gibbs energy, chemical potentials, and Jansson second derivatives are evaluated once with PyCalphad and 'stored in a structured HDF5 file'; these same stored values are then plotted as the 'CALPHAD reference' in Fig. 1/Table 1. The jansson/hybrid modes are interpolants (Clough-Tocher, RBF) built on exactly those grid points, so they reproduce the reference values at the nodes by construction. RMSE=0 therefore demonstrates interpolation self-consistency, not accuracy of the analytic Hessian or of the underlying MLIP/CALPHAD model. No independent reference outside the fitted data is used; the claim 'jansson matches exactly' restates how the surrogate was built.

full rationale

The main prediction chain — ORB-v2/SQS formation energies, PhaseForge CALPHAD TDB, Hessian eigenanalysis, coherent Cahn correction, Cahn-Hilliard phase field — is not circular: the spinodal maps and morphologies are model outputs compared against the experiments of An et al., not quantities fitted to those experiments. The paper explicitly reports three failures (Hf partitioned to matrix instead of precipitate; misfit overpredicted 8.7% vs 1.6%/0.54%; AZ=1.52 predicting <100> instead of observed <110>) and states in Section 3.4 that the SQS/ORB-MLIP elastic constants 'have not been validated against DFT or experimental measurements for the specific quaternary compositions predicted by the CALPHAD tie-line.' That is an acknowledged validation gap and a correctness risk, not circularity. The workflow leans heavily on the authors' own prior tools (Jansson [15], PhaseForge [16], MaterialsFramework [19], MLIP-CALPHAD benchmarks [22,23]), but these are open-source methods/benchmarks and no uniqueness or ansatz is smuggled in through them. The single genuine circularity is the RMSE=0 validation in Fig. 1/Table 1, where the 'CALPHAD reference' is identical to the data used to build the Jansson surrogates. This supports a score of 4 rather than 0 or 2: one constructed validation plus heavily self-cited infrastructure, while the central materials prediction retains independent content.

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

The central predictions rest on (a) fitting Redlich-Kister parameters to MLIP/SQS formation energies, (b) assuming ORB v2 transferability to BCC HfNbTiV, (c) standard CALPHAD/coherency/phase-field model choices, and (d) hand-set mobility and gradient coefficients. No new physical particles or fields are introduced; the paper's 'new' objects are computational surrogates, affine projections, and an open-source workflow.

free parameters (5)
  • Redlich-Kister interaction parameters L^(ν)_A,B and ternary interaction parameters = not reported; least-squares fitted to MLIP/SQS formation energies
    Eq. 3 and Section 2.1: the thermodynamic database is built by fitting RK polynomials to ORB-MLIP SQS formation energies; these parameters determine all subsequent Hessians and spinodal predictions.
  • gradient energy coefficients κ_i = κ = 1×10^-12 J m^2/mol in the reported simulations
    Section 2.3.1 and figure captions: phase-field morphology and spinodal wavelength scale directly with κ; it is set by hand, not derived or fitted to experiment.
  • bare mobility coefficients M̊_ij = 5×10^-21 m^5 J^-1 s^-1 for Nb/V; Hf mobility roughly 10× lower
    Section 2.3.1 and Figs. 6-7: absolute evolution timescales and Hf-halo morphologies depend on the chosen mobility tensor, which is not measured.
  • composition-vector regularization offset = 1×10^-4
    Section 2.2: small offset added to zero composition components when computing Hessians to mitigate numerical ill-conditioning; hand-chosen but minor.
  • convex-splitting stabilization constant A_i = A_i ≥ max(0, -H_ii(x0))
    Eq. 32: lower bound chosen to preserve convexity of the split free energy; not fitted to data, but a free algorithmic parameter.
assumptions (6)
  • domain assumption SQS supercells faithfully represent random solid-solution thermodynamics across the composition space
    Section 2.1: the entire MLIP-to-CALPHAD construction assumes SQS short-range-order matching yields representative formation energies; no finite-size convergence study is shown.
  • domain assumption ORB v2 universal MLIP gives accurate formation energies, lattice parameters, and elastic constants for BCC Hf-Nb-Ti-V, including metastable BCC Hf and Ti
    Sections 2.1, 2.2, and 3.4: all thermodynamic and elastic inputs come from ORB v2; the paper explicitly notes the elastic constants have not been validated for the quaternary.
  • domain assumption CALPHAD/CEF formalism with SGTE unary data and ideal configurational entropy captures thermal contributions
    Eq. 1 and Section 2.1: thermal effects beyond 0 K are added through SGTE unaries with static MLIP mixing energies; no vibrational or magnetic excess terms are included.
  • domain assumption Jansson derivative method correctly computes analytic partial derivatives of PyCalphad equilibrium chemical potentials
    Section 2.2: the analytic Hessians rely entirely on the adjoint-like Jansson algorithm as formalized in ref [15]; it is not re-derived here.
  • domain assumption Cahn coherency-strain correction with cubic Y and bond-based Vegard-type lattice-parameter derivatives applies to the BCC phase
    Section 2.2, Eqs. 13-16: linear-elastic coherency theory with composition-independent stiffness tensors and a bond-length lattice-parameter model.
  • domain assumption Phase-field gradient-energy and mobility parameters are sufficient and transferable for morphology prediction
    Section 2.3: κ and mobility are chosen by hand; no experimental interfacial energies or tracer diffusivities are used to anchor them.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From MLIPs to Microstructure: A High-Throughput Computational Framework to Design Spinodal Alloys in High-Dimensional Composition Spaces via Analytic Derivatives of CALPHAD Model Predictions." pith.science (2026). https://pith.science/paper/7RB5SVMI

@misc{pith2026260720077,
  author       = {Pith},
  title        = {Pith review of: From MLIPs to Microstructure: A High-Throughput Computational Framework to Design Spinodal Alloys in High-Dimensional Composition Spaces via Analytic Derivatives of CALPHAD Model Predictions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7RB5SVMI}},
  note         = {Machine review of arXiv:2607.20077}
}
read the original abstract

Identifying regions of design space subject to spinodal decomposition is a critical component of alloy design in high-dimensional composition spaces. In cases where designers are seeking to exploit spinodal microstructures to tailor alloy properties, prediction of microstructure evolution and morphology is also needed. In this work, we present a Machine Learning Interatomic Potential (MLIP)-trained, CALPHAD-based, open-source workflow for high-throughput microstructure stability analysis and visualization. In this workflow, coherent strain contributions are captured via high-throughput MLIP elastic constant calculations. To predict microstructure morphology for compositions of interest, MLIP-generated thermodynamic models are fed into an elasto-chemical phase field simulation. Both stability analyses and phase-field simulations utilize analytically-derived Gibbs energy Hessians to improve computational efficiency and accuracy over finite difference approximations. We demonstrate this workflow by investigating microstructure stability in the Hf-Nb-Ti-V quaternary system.

Figures

Figures reproduced from arXiv: 2607.20077 by the authors.

Figure 1
Figure 1. Comparison of the three derivative methods for the Nb [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Derivative-method comparison (fd vs. jansson) for Nb V BCCA2 at T = 500 ◦C (c0 = xV = 0.50, L = 1 µm). Top row (2D, 256 × 256 grid, t = 800 s, 80 000 steps): (a) finite-difference and (b) Jansson direct composition fields xV; (c) absolute difference map |x fd V − x jansson V | (RMSE = 1.31 × 10−5 , max = 2.42 × 10−4 ); table: composition statistics and wall times (FD: 279 s; Jansson: 199 s). Bottom row (3D, 643 grid… view at source ↗
Figure 3
Figure 3. Chemical (a, c, e) and coherent (b, d, f) Hessian stability maps for BCC Hf [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Gauge-invariant chemical first and second and coherent first eigenvalues of the BCC Hessian as a function of temperature for (a) Hf2 [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Simulated xNb microstructures from Cahn-Hilliard simulations of binary Nb V BCCA2 at 500 °C (128 × 128 grid, L = 1 µm, κ = 1 × 10−12 J m2 mol−1 , M = 5 × 10−21 m5 J −1 s −1 ). Top row: chemical only. Bottom row: elasto-chemical (tensorial Vegard coupling, αNb = 0.087, …
Figure 6
Figure 6. Figure 6: Hf-content sweep of xNb microstructures in ternary Nb Hf V BCCA2 at 500 °C (128 × 128 grid, L = 1 µm, κ = 1 × 10−12 J m2 mol−1 , MNb = 5 × 10−21 m5 J −1 s −1 ). (a) Chemical-only simulations, all compositions after 6000 s of isothermal aging. (b) Elasto-chemical simula…
Figure 7
Figure 7. Figure 7: Phase field simulations of equiatomic Nb [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: Zener anisotropy AZ = C44/C ′ from SQS/ORB-v2 MLIP elastic constants along three composition paths plotted against xV (mole fraction). Bi￾nary Nb50−xVx (green dashed): AZ < 1 throughout, ⟨110⟩ favored. Ternary Nb50Hf50−xVx (red dash-dot): AZ briefly exceeds unity near …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 2 canonical work pages

  1. [1]

    J. W. Cahn, On spinodal decomposition, Acta Metallur- gica 9 (9) (1961) 795–801

  2. [2]

    S. Wang, J. Wang, D. Shu, P. Shi, M. Wu, D. Wang, C. Yang, G. Zhu, B. Sun, Spinodal decomposition in- duced brittleness of Zr-Ta containing medium-entropy al- loys, Materials Characterization 205 (2023) 113330

  3. [3]

    Dasari, A

    S. Dasari, A. Sharma, V . Soni, Z. Kloenne, H. Fraser, R. Banerjee, Crystallographic and compositional evolu- tion of ordered B2 and disordered BCC phases during isothermal annealing of refractory high-entropy alloys, Microscopy and Microanalysis 29 (1) (2023) 303–313

  4. [4]

    Z. An, S. Mao, T. Yang, C. T. Liu, B. Zhang, E. Ma, H. Zhou, Z. Zhang, L. Wang, X. Han, Spinodal-modulated solid solution delivers a strong and ductile refractory high- entropy alloy, Materials Horizons 8 (3) (2021) 948–955

  5. [5]

    S. W. Lee, H. Park, R. E. Kim, J. Kim, D. W. Lee, J. H. Lee, S. I. Hong, H.-S. Jang, Y .-U. Heo, H. S. Kim, Spinodal-assisted multiscale immiscibility enables supe- rior strength–ductility synergy in (CuFeMnNi) 96Al2Ti2 high-entropy alloy, Materials Science and Engineering: A (2025) 149582

  6. [6]

    Zhang, Z

    W. Zhang, Z. Ma, H. Zhao, L. Ren, Refinement strengthening, second phase strengthening and spinodal microstructure-induced strength-ductility trade-offin a high-entropy alloy, Materials Science and Engineering: A 847 (2022) 143343

  7. [7]

    Z. Rao, B. Dutta, F. Körmann, W. Lu, X. Zhou, C. Liu, A. K. da Silva, U. Wiedwald, M. Spasova, M. Farle, et al., Beyond solid solution high-entropy alloys: tailoring mag- netic properties via spinodal decomposition, Advanced Functional Materials 31 (7) (2021) 2007668

  8. [8]

    P. Dong, L. Zhang, L. Huang, Q. Yang, L. Li, L. Ma, Z. Zhong, J. Li, Cooperative regulation of mechanical properties and magnetoresistance effect in high-entropy alloys by spinodal decomposition, Journal of Alloys and Compounds 970 (2024) 172547

Show all 50 references
  1. [9]

    S. Duan, X. Bao, J. Huang, R. Shi, L. Fei, W. Xue, H. Yao, X. Li, J. Wang, X. Liu, et al., Spinodal decomposition pro- moting high thermoelectric performance in half-Heusler, Joule 9 (4) (2025)

  2. [10]

    Usanmaz, P

    D. Usanmaz, P. Nath, J. J. Plata, G. L. Hart, I. Takeuchi, M. B. Nardelli, M. Fornari, S. Curtarolo, First principles thermodynamical modeling of the binodal and spinodal curves in lead chalcogenides, Physical Chemistry Chemi- cal Physics 18 (6) (2016) 5005–5011

  3. [11]

    Divilov, H

    S. Divilov, H. Eckert, C. Toher, R. Friedrich, A. C. Zettel, D. W. Brenner, W. G. Fahrenholtz, D. E. Wolfe, E. Zurek, J.-P. Maria, et al., A priori procedure to establish spin- odal decomposition in alloys, Acta Materialia 266 (2024) 119667

  4. [12]

    S. R. Koneru, K. Kadirvel, Y . Wang, High-throughput de- sign of multi-principal element alloys with spinodal de- composition assisted microstructures, Journal of Phase Equilibria and Diffusion 43 (6) (2022) 753–763

  5. [13]

    Kadirvel, S

    K. Kadirvel, S. R. Koneru, Y . Wang, Exploration of spinodal decomposition in multi-principal element alloys (MPEAs) using CALPHAD modeling, Scripta Materialia 214 (2022) 114657

  6. [14]

    Cao, S.-L

    W. Cao, S.-L. Chen, F. Zhang, K. Wu, Y . Yang, Y . Chang, R. Schmid-Fetzer, W. Oates, PANDAT software with Pa- nEngine, PanOptimizer and PanPrecipitation for multi- component phase diagram calculation and materials prop- erty simulation, Calphad 33 (2) (2009) 328–342. 17

  7. [15]

    Kunselman, B

    C. Kunselman, B. Bocklund, A. van de Walle, R. Otis, R. Arróyave, Analytically differentiable metrics for phase stability, Calphad 86 (2024) 102705

  8. [16]

    Sarıtürk, S

    D. Sarıtürk, S. Zhu, R. Arróyave, PhaseForge (Jun. 2025). doi:10.5281/zenodo.15730911

  9. [17]

    J. W. Cahn, Coherent fluctuations and nucleation in isotropic solids, Acta Metallurgica 10 (10) (1962) 907– 913

  10. [18]

    B. Vela, T. Hastings, M. Allen, R. Arróyave, Visualizing high entropy alloy spaces: methods and best practices, Digital Discovery 4 (1) (2025) 181–194

  11. [19]

    Sarıtürk, R

    D. Sarıtürk, R. Arróyave, MaterialsFramework (Jun. 2025).doi:10.5281/zenodo.15731044

  12. [20]

    Kunselman, S

    C. Kunselman, S. Zhu, D. Sarıtürk, R. Arróyave, Construction and Tuning of CALPHAD Models Using Machine-Learned Interatomic Potentials and Experimen- tal Data: A Case Study of the Pt–W System, Journal of Phase Equilibria and Diffusion (2025) 1–9

  13. [21]

    Bocklund, R

    B. Bocklund, R. Otis, A. Egorov, A. Obaied, I. Roslyakova, Z.-K. Liu, ESPEI for efficient ther- modynamic database development, modification, and uncertainty quantification: application to Cu–Mg, MRS Communications 9 (2) (2019) 618–627

  14. [22]

    S. Zhu, D. Sarıtürk, R. Arróyave, Accelerating CALPHAD-based phase diagram predictions in complex alloys using universal machine learning potentials: Opportunities and challenges, Acta Materialia 286 (2025) 120747

  15. [23]

    S. Zhu, D. Sarıtürk, R. Arróyave, Machine learning poten- tials for alloys: a detailed workflow to predict phase dia- grams and benchmark accuracy, npj Computational Mate- rials 11 (2025) 340

  16. [24]

    Zunger, S.-H

    A. Zunger, S.-H. Wei, L. G. Ferreira, J. E. Bernard, Special quasirandom structures, Physical Review Letters 65 (3) (1990) 353–356

  17. [25]

    van de Walle, M

    A. van de Walle, M. Asta, G. Ceder, The alloy theoretic automated toolkit: A user guide, Calphad 26 (4) (2002) 539–553

  18. [26]

    A. T. Dinsdale, SGTE data for pure elements, Calphad 15 (4) (1991) 317–425

  19. [27]

    Neumann, J

    M. Neumann, J. Gin, B. Rhodes, S. Bennett, Z. Li, H. Choubisa, A. Hussey, J. Godwin, Orb: A Fast, Scalable Neural Network Potential (2024).arXiv:2410.22570

  20. [28]

    B. Deng, P. Zhong, K. Jun, J. Riebesell, K. Han, C. J. Bar- tel, G. Ceder, CHGNet: Pretrained Universal Neural Net- work Potential for Charge-Informed Atomistic Modeling (2023).arXiv:2302.14231

  21. [29]

    Schmidt, N

    J. Schmidt, N. Hoffmann, H.-C. Wang, P. Borlido, P. J. M. A. Carriço, T. F. T. Cerqueira, S. Botti, M. A. L. Marques, Machine-Learning-Assisted Determination of the Global Zero-Temperature Phase Diagram of Materi- als, Advanced Materials (2023).doi:10.1002/adma. 202210788

  22. [30]

    Otis, Z.-K

    R. Otis, Z.-K. Liu, pycalphad: CALPHAD-based Compu- tational Thermodynamics in Python, in: Zentropy, Jenny Stanford Publishing, 2017, pp. 373–392

  23. [31]

    J. W. Gibbs, On the equilibrium of heterogeneous sub- stances, American Journal of Science 3 (96) (1878) 441– 458

  24. [32]

    De Fontaine, A computer simulation of the evolution of coherent composition variations in solid solutions, Ph.D

    D. De Fontaine, A computer simulation of the evolution of coherent composition variations in solid solutions, Ph.D. thesis, Northwestern University (1967)

  25. [33]

    De Fontaine, An analysis of clustering and ordering in multicomponent solid solutions-I

    D. De Fontaine, An analysis of clustering and ordering in multicomponent solid solutions-I. Stability criteria, Jour- nal of Physics and Chemistry of Solids 33 (2) (1972) 297– 310

  26. [34]

    J. E. Morral, S. Chen, Stability of high entropy alloys to spinodal decomposition, Journal of Phase Equilibria and Diffusion 42 (5) (2021) 673–695

  27. [35]

    Singh, A

    P. Singh, A. V . Smirnov, D. D. Johnson, Atomic short- range order and incipient long-range order in high-entropy alloys, Physical Review B 91 (22) (2015) 224204

  28. [36]

    D. De Fontaine, An analysis of clustering and ordering in multicomponent solid solutions—II fluctuations and ki- netics, Journal of Physics and Chemistry of Solids 34 (8) (1973) 1285–1304

  29. [37]

    Tandoc, L

    C. Tandoc, L. Qi, Y .-J. Hu, A bond-based model for accu- rate prediction of lattice parameters of bcc solid solution alloys, Materialia 40 (2025) 102410

  30. [38]

    J. W. Cahn, J. E. Hilliard, Free energy of a nonuniform system. i. interfacial free energy, The Journal of chemical physics 28 (2) (1958) 258–267

  31. [39]

    L. Q. Chen, J. Shen, Applications of semi-implicit fourier-spectral method to phase field equations, Com- puter Physics Communications 108 (2-3) (1998) 147–158

  32. [40]

    D. J. Eyre, Unconditionally gradient stable time march- ing the cahn-hilliard equation, MRS online proceedings library (OPL) 529 (1998) 39

  33. [41]

    J. Shen, X. Yang, Numerical approximations of allen-cahn and cahn-hilliard equations, Discrete Contin. Dyn. Syst 28 (4) (2010) 1669–1691

  34. [42]

    R. W. Clough, Finite element stiffness matricess for analy- sis of plate bending, in: Proc. of the First Conf. on Matrix Methods in Struct. Mech., 1965, pp. 515–546. 18

  35. [43]

    Virtanen, R

    P. Virtanen, R. Gommers, T. E. Oliphant, M. Haber- land, T. Reddy, D. Cournapeau, E. Burovski, P. Peterson, W. Weckesser, J. Bright, et al., Scipy 1.0: fundamental al- gorithms for scientific computing in python, Nature meth- ods 17 (3) (2020) 261–272

  36. [44]

    De Fontaine, Configurational thermodynamics of solid solutions, Solid state physics 34 (1979) 73–274

    D. De Fontaine, Configurational thermodynamics of solid solutions, Solid state physics 34 (1979) 73–274

  37. [45]

    Gururajan, T

    M. Gururajan, T. Abinandanan, Phase field study of pre- cipitate rafting under a uniaxial stress, Acta Materialia 55 (15) (2007) 5015–5026

  38. [46]

    Moulinec, P

    H. Moulinec, P. Suquet, A numerical method for com- puting the overall response of nonlinear composites with complex microstructure, Computer methods in applied mechanics and engineering 157 (1-2) (1998) 69–94

  39. [47]

    W. B. Pearson, A handbook of lattice spacings and struc- tures of metals and alloys: International series of mono- graphs on metal physics and physical metallurgy, V ol. 4, V ol. 4, Elsevier, 2013

  40. [48]

    De Jong, W

    M. De Jong, W. Chen, T. Angsten, A. Jain, R. Notes- tine, A. Gamst, M. Sluiter, C. Krishna Ande, S. Van Der Zwaag, J. J. Plata, et al., Charting the complete elastic properties of inorganic crystalline compounds, Scientific data 2 (1) (2015) 150009

  41. [49]

    Söderlind, O

    P. Söderlind, O. Eriksson, J. Wills, A. Boring, Theory of elastic constants of cubic transition metals and alloys, Physical Review B 48 (9) (1993) 5844

  42. [50]

    Ikeda, B

    Y . Ikeda, B. Grabowski, F. Körmann, Ab initio phase sta- bilities and mechanical properties of multicomponent al- loys: A comprehensive review for high entropy alloys and compositionally complex alloys, Materials Characteriza- tion 147 (2019) 464–511. 19

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.