Pith. sign in

REVIEW 3 major objections 5 minor 7 references

Interface Energy and Phase Transformations: A Comparative Analysis of Cahn-Hilliard and CALPHAD-based Models in Ternary Substitutional Alloys

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Reactive diffusion and the Cahn-Hilliard model are the same variational problem, differing only in the molar free energy.

desk verdict A genuinely useful conceptual unification of Cahn-Hilliard and reactive diffusion, but the advertised stable continuous FEM is likely unstable with P2/P2 and the discrete variational forms contain a sign inconsistency. read the letter →

arxiv 2411.16430 v2 pith:IV3ASYWC submitted 2024-11-25 math.NA cond-mat.mtrl-scics.NAphysics.chem-ph

Wolfgang Flachberger , Thomas Antretter , Swaroop Gaddikere-Nagaraja , Silvia Leitner , Manuel Petersmann , Jiri Svoboda This is my paper · ORCID
classification math.NAcond-mat.mtrl-scics.NAphysics.chem-ph MSC 65M6035K5580A22
keywords Cahn-HilliardequationreactivediffusionCALPHADinterfaceenergyOnsagercoefficientsmixedfiniteelementmethodphase-fieldmodelingvacancy
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

This paper argues that two modeling traditions that materials scientists usually treat as separate, Cahn-Hilliard phase-field models and CALPHAD-based reactive diffusion models, are in fact one and the same mathematical object once both are written as a variational problem with the diffusional flux as an independent variable. The two differ only in what molar free energy $f_m$ is inserted: a double-well potential with a gradient term for Cahn-Hilliard, or its convex hull (the CALPHAD common-tangent energy) for reactive diffusion. On that basis the authors build a continuous finite element method, stabilized by solving explicitly for the flux and the chemical affinities, that remains stable for both smooth interfaces and sharp, regularization-free interfaces. If the unification holds, a single simulation framework can dial interface energy from zero to arbitrary strength and can treat phase transformations, ternary alloys, and vacancy-mediated diffusion without switching models.

What carries the argument

The load-bearing object is the Lagrangian functional with the flux as an independent variable, together with the mixed variational form that solves for flux explicitly. Because the affinity never appears under a differential operator in these forms, the solution no longer requires differentiating a free energy that is only piecewise smooth or $C^1$ (the convex hull), which is what previously forced either regularization or discontinuous Galerkin discretizations. The relation $\tilde{L}=\tilde{D}T/(k\Omega)$, obtained from in-phase quadratic free energies, is what converts measured diffusion coefficients into Onsager coefficients and produces the scaled Lagrangian in which the flux has units of velocity.

What would settle it

Run the sharp-interface model on a domain with many closely spaced interfaces, so that interface regions occupy a measurable fraction of the volume, and compare the resulting phase-growth kinetics against the same initial state modeled with Onsager coefficients evaluated separately inside each bulk phase; any significant discrepancy in interface velocity or concentration profiles would falsify the global use of Equation (39).

Watch

Extended reading notes

Core claim

The central claim is that reactive diffusion and the Cahn-Hilliard phase-field model are inherently identical, distinguishable only by their definitions of the molar free energy $f_m$. The proof runs through a Lagrangian in which the fluxes are independent variables and the chemical affinities enter as Lagrange multipliers that enforce mass conservation; varying the Lagrangian with respect to flux, mole fraction, and affinity reproduces the flux law, the identity that the affinity is the variational derivative of the molar free energy, and the mass-conservation equations. Combined for a binary system this gives $\partial c/\partial t = \nabla\cdot(-\tilde{L}/T\,\nabla(\delta f_m/\delta x))$, which is the Cahn-Hilliard equation when $f_m$ contains a double-well potential and the CALPHAD/reactive-diffusion equation when $f_m$ is its convex hull. The paper further proposes the relation $\tilde{L} = \tilde{D}T/(k\Omega)$ between Onsager coefficients, interdiffusion coefficients, the curvature $k$ of the quadratic phase free energies, temperature, and molar volume, and shows how the resulting scaled Lagrangian supports a continuous Galerkin finite element method that stays stable for both smooth and sharp interfaces.

Load-bearing premise

The load-bearing premise is that the relation $\tilde{L}=\tilde{D}T/(k\Omega)$, derived inside bulk phases where the molar free energy is quadratic, can be applied globally in the simulations, even though the paper itself notes it fails at interfaces where the convex free energy is linear.

Editorial extensions

If this is right

  • A single finite-element code can reproduce classical Cahn-Hilliard spinodal decomposition, regularized smooth-interface transformations, and sharp-interface CALPHAD-type reactive diffusion by switching the molar free energy between double-well, convex-hull with $\kappa>0$, and convex-hull with $\kappa=0$.
  • Interface energy strength can be varied continuously in the same framework, so morphology changes, such as the observed transition from spherical coarsening to non-area-minimizing growth, can be studied as a function of $\Delta F_{\mathrm{int}}$ rather than as a change of model.
  • The thermodynamic-consistency proof and the explicit flux solve guarantee nonnegative entropy production and mass conservation, so results are not hostage to the stability of a particular discretization of the chemical potential.
  • The Onsager-to-diffusion-coefficient relation turns experimentally measured interdiffusion coefficients and phase-curvature data into model inputs, enabling ternary and multi-component simulations with non-diagonal, coupled fluxes.
  • The vacancy model with sources and sinks, including non-conserved vacancy generation and annihilation and the Kirkendall effect, is realized within the same variational formalism, showing that the method reaches beyond binary diffusion.

Reading between the lines

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

  • An untested corollary of the claimed equivalence is that every Cahn-Hilliard simulation with vanishing gradient coefficient has a sharp-interface reactive-diffusion twin with the same macroscopic kinetics; running a known spinodal-decomposition benchmark in both limits would test the unification directly.
  • Because Equation (39) is exact only inside bulk phases, the error of kinetic predictions should grow with the volume fraction of interfaces; for lamellar or nanoprecipitate microstructures with dense interfaces, the global relation may need to be replaced by phase-local Onsager coefficients.
  • The Lagrangian's structure of free energy plus dissipation plus constraint invites direct coupling to mechanics, heat conduction, and other rate processes by adding terms, so the method could become a template for multiphysics phase-transformation simulation.
  • The convex-hull/double-well duality is not alloy-specific; the same variational identification might apply to any system where a free energy and its convex envelope describe different interface physics.
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

3 major / 5 minor

Summary. The paper proposes a unified variational framework for diffusional phase transformations, claiming that Cahn-Hilliard models and reactive diffusion (CALPHAD-based) models are the same physical model distinguished only by the choice of molar free energy fm. The authors derive a Lagrangian formulation with fluxes and chemical affinities as independent variables, prove thermodynamic consistency via nonnegative entropy production and a Gibbs-Duhem check, and introduce a relation between Onsager coefficients and interdiffusion coefficients. They then propose a continuous Galerkin mixed finite element discretization of degree two for the resulting variational forms and demonstrate the method on binary Cahn-Hilliard, sharp-interface, and ternary vacancy-diffusion examples. The central claims are the conceptual unification of the two modeling paradigms and the existence of a stable continuous finite element method for both smooth and sharp interface problems.

Significance. If the claims hold, the paper would make a useful conceptual contribution by showing that Cahn-Hilliard and reactive diffusion models are not fundamentally different, and by providing a single variational principle from which both can be recovered. The algebraic derivation of the Gibbs-Duhem consistency and the explicit link between Onsager and interdiffusion coefficients are valuable and clearly presented. The numerical stabilization claim, however, is the main advertised novelty and is currently not supported by the analysis or the experiments; the manuscript needs substantial revision before the stability and consistency of the discrete formulation can be accepted.

major comments (3)
  1. [Section 2.4, Eqs. (49)-(50)] The stabilization claim is not established. The paper cites [BDM85] as the origin of forms (49)-(50), but [BDM85] proves stability for H(div)-conforming mixed finite elements with appropriately selected pressure spaces, not for continuous P2/P2 equal-order elements as used here. For continuous P2 vector fields and P2 scalar affinities, the discrete inf-sup condition for b(j, \bar{\mu}) = \int \bar{\mu} \nabla\cdot j is not satisfied: the divergence of a P2 vector field lies in P1, so any P2 affinity component orthogonal to P1 belongs to the discrete kernel, giving an inf-sup constant of zero and permitting checkerboard modes. The oscillations visible in Figure 3 may therefore be the expected numerical artifact of this unstable discretization rather than physical interfacial fluctuations. This directly undermines the advertised stable continuous FEM for sharp interfaces.
  2. [Section 2.4, Eqs. (45)-(49) and Eq. (42)] There is a sign inconsistency between the continuous and discrete Lagrangians. The continuous Lagrangian (26)/(40) contains the constraint term with a minus sign, while the scaled discrete Lagrangian (45)/(46) uses a plus sign. As written, Eq. (48) enforces \bar{\mu} = -\delta F/\delta x rather than Eq. (31)'s \bar{\mu} = \delta f_m/\delta x, and Eq. (49), after integration by parts, corresponds to j = +(D/k)\nabla\bar{\mu} rather than Eq. (42)'s j = -(D/(k\Omega))\nabla\bar{\mu}. The two sign changes happen to cancel if \bar{\mu} is reinterpreted as the negative affinity, but the manuscript does not make this reinterpretation, and the variational forms are therefore not a discretization of the stated variational principle. The authors must correct the signs or redefine \bar{\mu} consistently, then rerun the simulations.
  3. [Section 2.2, Eq. (39), used in Section 2.3 and all simulations] The relation \tilde{L} = \tilde{D} T/(k\Omega) is derived under the assumption that the molar free energy is quadratic in the bulk phases, and the paper explicitly states that the relation fails at interfaces. Nevertheless, the scaled Lagrangian (41) and all numerical experiments apply this relation globally. Since the paper's new capability is specifically the treatment of sharp interfaces, the interface region—where this relation is invalid—may contribute appreciably to the kinetics. The manuscript needs a quantitative justification, such as a convergence study in interface width or an estimate of the interface contribution, before the kinetic predictions built on Eq. (39) can be considered reliable.
minor comments (5)
  1. [Eqs. (30)-(31)] The step from \partial_t(\delta f_m/\delta x - \bar{\mu})=0 to \bar{\mu}=\delta f_m/\delta x silently drops a time-independent integration constant; the constant should be fixed by initial or boundary conditions, or the argument should state why it is irrelevant for the subsequent flux law.
  2. [Figure 3 caption] The caption states that the solution "just behaves well when the phases are entirely separated," which directly contradicts the claimed stabilization of the sharp-interface case; the authors should clarify whether the observed fluctuations are numerical artifacts or physical features.
  3. [Throughout] There are numerous typographical errors, including "adressed" (p. 3), "Therby" (p. 9), "subsititutional" (p. 11), "entirely seperated" (Fig. 7 caption), and "Therefore" capitalized mid-sentence in Section 2.3.
  4. [Keywords] The keyword "Theorem of Minimum Entropy Production" is listed, but the paper does not state or prove such a theorem; either add a relevant discussion or remove the keyword.
  5. [Section 2.5] The use of \tilde{D}(x_t) in Eq. (51) is presented as a way to avoid variation of the dissipative term, but the paper does not discuss the consistency of this explicit-in-time treatment with the variational principle; a brief comment on the resulting time-discretization error would be helpful.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's derivation chain is self-contained, and its stated approximations are modeling assumptions, not fitted predictions.

full rationale

The central derivation is self-contained. Starting from the Lagrangian in Eq. (26), variations with respect to the flux, mole fraction, and affinity give the flux law (28), the affinity relation (31), and mass conservation (32). Combining these for a binary system gives Eq. (33), which reduces to the Cahn-Hilliard equation when fm contains the square-gradient term; this is a mathematical equivalence shown by the paper's own equations, not an input renamed as a prediction. The Onsager-to-diffusion relation in Eq. (39) is obtained by equating Fick's law with the thermodynamic flux under a quadratic free energy, and the paper explicitly states that this relation is valid only in bulk phases and not at interfaces (Section 2.2). Applying it globally in the scaled Lagrangian is therefore a stated approximation, not a circular fit. The stabilization claim cites Brezzi-Douglas-Marini (an external, machine-independent mathematical result) and the paper's own continuity assertion for the convex free energy can be checked directly from the analytical expression in Appendix C; no load-bearing step is justified solely by an author's prior uniqueness or existence theorem. The variational construction of the flux law to guarantee nonnegative entropy production is the standard form of linear irreversible thermodynamics and is a consistency check, not a derivational circularity. The numerical studies show behavior that follows from the chosen free-energy inputs, but they are not presented as independent predictions of those inputs. The paper's self-citations (e.g., [Fla+24], [SFF06]) are used as background or as previously published model ingredients, not as the logical basis forcing the central result. No fitted parameter is renamed as a prediction, and no known result is merely relabeled; the claimed unification is an explicit derivation. Accordingly, no circular step is present.

Assumptions & free parameters 8 free parameters · 8 assumptions · 0 invented entities

The central derivation rests on standard linear nonequilibrium thermodynamics and variational calculus, plus domain assumptions about substitutional diffusion, including equal molar volumes, no volume flow, and a sharp-interface limit. The main ad hoc elements are the quadratic phase free energy used to link Onsager and Fick diffusion coefficients, the global use of that bulk-only relation, the neglect of vacancy flux dissipation, and the assumed stability of the mixed finite element spaces. No new physical entities are postulated.

free parameters (8)
  • Phase-curvature parameter k = Not given; chosen per simulation for convergence
    Appears in the quadratic phase free energy, Eq. (34), and in the Onsager-diffusion relation, Eq. (39). The paper calls it phenomenological and states it does not affect the kinetics.
  • Gradient energy coefficient kappa = 0, 1, 5, or 100 J mm^2/mol depending on figure
    Controls interface width and regularization in the free energy functional, Eq. (1); varied across the simulations in Figures 2, 4, 6, 8, and 10.
  • Interface energy Delta F_int = Varied in Figure 6 sweep; set by double-well height
    Defined as the height difference between the double-well f0 and its convex hull (Figure 1); an input parameter controlling morphology, not measured.
  • Interdiffusion coefficient D_tilde = 1 mm^2/s in binary examples
    Set by hand for the demonstrations and used to set the Onsager coefficient via Eq. (39).
  • Component diffusion coefficients D1 and D2 = D1 = 2 mm^2/s, D2 = 1 mm^2/s, D0 tends to infinity
    Chosen to induce the Kirkendall effect in the vacancy example, Section 3; not calibrated to a specific alloy.
  • Equilibrium vacancy site fraction x0_eq = 0.001
    Set as the initial and equilibrium vacancy site fraction in the ternary example, Section 3.
  • Vacancy curvature k0 and source/sink resistance A_phi = Not specified
    Parameters in the vacancy free energy, Eq. (52), and dissipation functional, Eq. (53); the values needed for implementation are not reported.
  • Phase equilibrium free energies and compositions = Not specified explicitly
    The quantities f0^alpha, f0^beta, x_eq^alpha, and x_eq^beta define the quadratic phase free energies in Eq. (34) and the free energy plot in Figure 1; they are chosen by hand.
assumptions (8)
  • standard math Positive definiteness of the Onsager coefficient matrix
    Used to guarantee nonnegative entropy production in Eq. (21); the paper states both L and tilde-L must be positive definite.
  • domain assumption Constant molar volume and equal partial molar volumes for all components
    Invoked in Section 2.1, Eqs. (8) and (13), and used to simplify the scaled Lagrangian; variable molar volumes are only discussed later and assumed negligible for sharp interfaces.
  • domain assumption Constraints that mole fractions sum to one and fluxes sum to zero
    Eqs. (7) and (9); justified for substitutional diffusion in crystalline solids, but restricts the method to that diffusion mechanism.
  • domain assumption Quadratic molar free energy in each phase
    Eq. (34) is needed to derive the direct relation between Onsager and diffusion coefficients in Eq. (39); general CALPHAD free energies would break the proportionality.
  • ad hoc to paper Bulk-only Onsager-diffusion relation can be applied globally
    The paper acknowledges the relation fails at interfaces where fm is linear (Section 2.2) and assumes the small interface fraction makes the error negligible; this underpins the scaled Lagrangian.
  • ad hoc to paper Neglect of vacancy flux dissipation, D0 tends to infinity
    Section 3 states this choice is unlike the suggestion in [SFF06]; it changes vacancy kinetics and is not further justified.
  • ad hoc to paper Stability of the mixed continuous Galerkin formulation with degree-two spaces
    The paper asserts the variational form stabilizes Laplace-type problems and cites [BDM85], but no discrete inf-sup or convergence analysis is provided for this specific formulation.
  • domain assumption Sharp-interface idealization is legitimate when interface energy is absent
    Section 1.2 argues that continuum thermodynamics permits sharp interfaces at sufficiently large length scales; this underlies limiting case (ii).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Interface Energy and Phase Transformations: A Comparative Analysis of Cahn-Hilliard and CALPHAD-based Models in Ternary Substitutional Alloys." pith.science (2026). https://pith.science/paper/IV3ASYWC

@misc{pith2026241116430,
  author       = {Pith},
  title        = {Pith review of: Interface Energy and Phase Transformations: A Comparative Analysis of Cahn-Hilliard and CALPHAD-based Models in Ternary Substitutional Alloys},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IV3ASYWC}},
  note         = {Machine review of arXiv:2411.16430}
}
read the original abstract

There are various methods for modeling phase transformations in materials science, including general classes of phase-field methods and reactive diffusion methodologies, which most importantly differ in their treatment of interface energy. These methodologies appear mutually exclusive since the respective numerical schemes only allow for their primary use case. To address this issue, a novel methodology for modeling phase transformations in multi-phase, multi-component systems, with particular emphasis on applications in materials science and the study of substitutional alloys is introduced. The fundamental role of interface energy in the evolution of a material's morphology will be studied by example of binary and ternary systems. Allowing full control over the interface energy quantity enables more detailed investigations and bridges the gaps between known methods. We prove the thermodynamic consistency of the derived method and discuss several use cases, such as vacancy-mediated diffusion. Furthermore a scheme for relating Onsager and Diffusion coefficients is proposed, which allows us to study the intricate coupling that is observed in multicomponent systems. We hope to contribute to the development of new mathematical tools for modeling complex phase transformations in materials science.

Figures

Figures reproduced from arXiv: 2411.16430 by the authors.

Figure 1
Figure 1. f0 represents a general double-well potential as proposed by Cahn and Hilliard. The molar free energies of the individual phases are given by f α 0 and f β 0 respectively. In an equilibrium system without interface energy the molar free energy is given by the convex hull curve f ∗∗ 0 . Therefore, the maximal height difference ∆Fint between f0 and f ∗∗ 0 represents the interface energy. Hilliard model was not adresse… view at source ↗
Figure 2
Figure 2. System evolution of the variational forms given in (48) - (50). The contour visible is the distribution of the [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗
Figure 3
Figure 3. System evolution of the variational forms given in (48) - (50) for limiting case (ii). The contour visible is [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: System evolution of the variational forms given in (48) - (50) for limiting case (i). The contour visible is [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]
Figure 5
Figure 5. Figure 5: Direct comparison of the final states (t = 0.2[s]) of the limiting cases (i) and (ii) as well as the Cahn-Hilliard model. 14 [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]
Figure 6
Figure 6. Figure 6: System evolution of the variational forms given in (48) - (50) for different choices of the interface energy [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: The initial condition of the mole fraction [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: System evolution of the variational forms given in (48) - (50). The contour visible is the distribution of [PITH_FULL_IMAGE:figures/full_fig_p016_8.png]
Figure 9
Figure 9. Figure 9: System evolution of the variational forms given in (48) - (50) for limiting case (ii).The contour visible is the [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]
Figure 10
Figure 10. Figure 10: System evolution of the variational forms given in (48) - (50) for limiting case (i). The contour visible is [PITH_FULL_IMAGE:figures/full_fig_p017_10.png]
Figure 11
Figure 11. Figure 11: System evolution for the variational problem defined in equation (56). The contour visible is the distribution [PITH_FULL_IMAGE:figures/full_fig_p017_11.png]
Figure 12
Figure 12. Figure 12: System evolution for the variational problem defined in equation (56). The contour visible is the distribution [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

7 extracted references · 7 canonical work pages

  1. [1]

    Reciprocal relations in irreversible processes. I

    [Ons31] Lars Onsager. “Reciprocal relations in irreversible processes. I.” In: Physical review 37.4 (1931), p

    work page 1931

  2. [405]

    Free energy of a nonuniform system. I. Interfacial free energy

    [CH58] John W Cahn and John E Hilliard. “Free energy of a nonuniform system. I. Interfacial free energy”. In: The Journal of chemical physics 28.2 (1958), pp. 258–267. [Man71] John R Manning. “Correlation factors for diffusion in nondilute alloys”. In: Physical Review B4.4 (1971), p

    work page 1958

  3. [1111]

    Chemical thermodynamics of materials

    [Lup83] Claude HP Lupis. “Chemical thermodynamics of materials”. In: (No Title) (1983). [BDM85] Franco Brezzi, Jim Douglas, and L Donatella Marini. “Two families of mixed finite elements for second order elliptic problems”. In: Numerische Mathematik 47 (1985), pp. 217–235. [VRD+95] Guido Van Rossum, Fred L Drake, et al. Python reference manual. V ol

    work page 1983

  4. [1995]

    Molecular dynamics simulation of crack growth under cyclic load- ing

    [NM04] Kenji Nishimura and N Miyazaki. “Molecular dynamics simulation of crack growth under cyclic load- ing”. In: Computational Materials Science 31.3-4 (2004), pp. 269–278. [SFF06] J Svoboda, Franz Dieter Fischer, and P Fratzl. “Diffusion and creep in multi-component alloys with non-ideal sources and sinks for vacancies”. In: Acta materialia 54.11 (2006...

    work page 2004

  5. [2014]

    The FEniCS project version 1.5

    [Aln+15] Martin Alnæs et al. “The FEniCS project version 1.5”. In: Archive of numerical software 3.100 (2015). [LP16] Hans Petter Langtangen and Geir K Pedersen. Scaling of differential equations. Springer Nature,

    work page 2015

  6. [2017]

    Sharp limit of the viscous Cahn–Hilliard equation and thermo- dynamic consistency

    [DG17] Wolfgang Dreyer and Clemens Guhlke. “Sharp limit of the viscous Cahn–Hilliard equation and thermo- dynamic consistency”. In: Continuum Mechanics and Thermodynamics 29 (2017), pp. 913–934. 21 Interface Energy and Phase Transformations: A Comparative Analysis of Cahn-Hilliard and CALPHAD-based Models in Ternary Substitutional Alloys A PREPRINT [SF17]...

    work page 2017

  7. [3982]

    Construction of arbitrary order finite element degree-of-freedom maps on polygonal and polyhedral cell meshes

    [Scr+22b] Matthew W Scroggs et al. “Construction of arbitrary order finite element degree-of-freedom maps on polygonal and polyhedral cell meshes”. In: ACM Transactions on Mathematical Software (TOMS) 48.2 (2022), pp. 1–23. [Bar+23] Igor A Barrata et al. “DOLFINx: The next generation FEniCS problem solving environment”. In: (2023). [Fla+24] Wolfgang Flach...

    work page 2022

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.