Efficient method for calculation of low-temperature phase boundaries

Babak Sadigh; Christine Wu; Lucas Svensson; Paul Erhart

arxiv: 2603.09804 · v2 · submitted 2026-03-10 · ❄️ cond-mat.mtrl-sci · physics.chem-ph· physics.comp-ph

Efficient method for calculation of low-temperature phase boundaries

Lucas Svensson , Babak Sadigh , Christine Wu , Paul Erhart This is my paper

Pith reviewed 2026-05-15 13:29 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci physics.chem-phphysics.comp-ph

keywords phase boundariesquasi-harmonic approximationClausius-Clapeyronsilicamachine learning potentialdensity functional theoryfree energy

0 comments

The pith

A method combining the Clausius-Clapeyron equation with the quasi-harmonic approximation computes low-temperature phase boundaries with minimal calculations.

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

The paper presents an efficient framework for determining phase boundaries at low temperatures by integrating the Clausius-Clapeyron relation with the quasi-harmonic approximation. This approach accounts for thermal, quantum, and low-order anharmonic effects while requiring only a small number of computations. It is demonstrated on silica, producing phase boundaries from -2 to 12 GPa and up to 1750 K that match results from free-energy integration methods. The framework uses a machine-learned potential trained on DFT data for efficient sampling.

Core claim

By coupling the Clausius-Clapeyron equation to quasi-harmonic free-energy differences, the method traces phase boundaries with few DFT evaluations, naturally including internal degrees of freedom and quantum effects, and yields silica phase lines consistent with full thermodynamic integration.

What carries the argument

The Clausius-Clapeyron equation integrated with quasi-harmonic approximation free energies, updated via machine-learned potentials.

If this is right

Phase boundaries at finite but low temperatures become accessible without extensive sampling.
Internal degrees of freedom are automatically included in the free-energy differences.
Quantum effects are captured through the harmonic approximation.
Low-order anharmonic contributions are incorporated naturally.
The approach scales to larger systems via machine-learned potentials.

Where Pith is reading between the lines

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

The method could extend to higher temperatures if anharmonic corrections are added.
Similar frameworks might apply to other materials like metals or perovskites.
It reduces the barrier for computing full phase diagrams in materials discovery.
Validation against experiments could test the accuracy beyond DFT.

Load-bearing premise

The quasi-harmonic approximation must remain sufficiently accurate over the low-temperature range of the phase boundary, and the machine-learned potential must accurately reproduce the underlying DFT free-energy differences.

What would settle it

A direct comparison of the computed phase boundary for silica against experimental measurements or against results from a fully anharmonic free-energy calculation at a point along the boundary would falsify the claim if significant deviations appear.

read the original abstract

Understanding phase stability and phase transformations is central to predicting material behavior under varying thermodynamic conditions. One of the earliest and most influential applications of density functional theory in materials science has been the prediction of pressure-induced phase transitions at 0 K. Extending these calculations to finite temperatures, however, requires accounting for thermal, quantum, and anharmonic contributions to the free energy, often at significant computational cost. In this work, we present a general and efficient framework for calculating low-temperature phase boundaries by combining the Clausius-Clapeyron equation with the quasi-harmonic approximation. This methodology requires a minimal number of calculations, while naturally incorporating internal degrees of freedom as well as quantum and low-order anharmonic effects. We illustrate the accuracy and efficiency of the approach by constructing the phase diagram of silica in the pressure range from -2 to 12 GPa and temperatures up to 1750 K. To this end, we employ a machine-learned interatomic potential trained on density functional theory reference data, enabling well-converged free energy estimates via efficient thermodynamic sampling and a rigorous comparison between the proposed framework and free energy integration.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper gives a practical shortcut for tracing low-T phase boundaries by stepping the Clausius-Clapeyron equation with quasi-harmonic free energies from an ML potential, and the silica test shows agreement with full integration.

read the letter

The main point is that this method steps along the phase boundary using the Clausius-Clapeyron slope supplied by quasi-harmonic Delta F and Delta V values computed from a machine-learned potential. It starts from a known low-T point and updates the pressure-temperature line with far fewer full free-energy integrations than a grid-based approach would require. The silica demonstration covers -2 to 12 GPa up to 1750 K and reports agreement with direct thermodynamic integration on the same potential. That is the concrete advance: a reusable algorithmic framework that folds in quantum and low-order anharmonic contributions through the QHA while keeping the number of expensive calculations small. The ML potential is key because it makes the phonon sampling cheap enough to do repeatedly along the path. The math is standard and the comparison to full integration is the right check to run. The soft spot is that the accuracy still rests on the quasi-harmonic approximation holding for the entropy differences over the whole segment. At 1750 K even silica can have enough higher-order anharmonicity to shift the integrated boundary, and any volume-dependent bias in the ML potential would accumulate through the slope updates. The abstract claims agreement, but without seeing the step-size choices, error propagation, or how they handled the internal degrees of freedom in the stepping, it is hard to judge how tight the match really is. The method is not circular because the free energies are computed independently of the boundary location itself. This paper is for groups already running ML-potential workflows for phase stability who need low-T boundaries without the full cost of thermodynamic integration at every condition. It is an engineering improvement rather than a fundamental result, but the numerical test is concrete enough that the details are worth checking. I would bring it to a reading group to look at the stepping implementation. I would cite it in my own work if I were extending similar calculations. It deserves peer review because the core idea is straightforward to verify and the efficiency claim is testable on other systems.

Referee Report

2 major / 1 minor

Summary. The manuscript presents an efficient framework for low-temperature phase boundaries that integrates the Clausius-Clapeyron equation using free energies and volumes computed in the quasi-harmonic approximation (QHA) from a machine-learned interatomic potential trained on DFT data. The approach is illustrated on the silica phase diagram between -2 and 12 GPa up to 1750 K and is claimed to agree with full free-energy integration while incorporating quantum and low-order anharmonic effects with minimal calculations.

Significance. If the numerical agreement holds under scrutiny, the method offers a computationally economical route to finite-temperature phase boundaries that naturally includes volume-dependent phonons and quantum statistics via QHA, which could be broadly useful for materials where exhaustive anharmonic sampling remains prohibitive.

major comments (2)

[Abstract] The central validation claim (agreement with free-energy integration) is asserted in the abstract but lacks quantitative support such as tabulated boundary points, RMS deviations, or error propagation from the ML potential; this directly bears on whether the QHA-based slope updates reproduce the reference boundary within the stated tolerance.
[Methodology] The method updates the boundary slope repeatedly using QHA-derived Delta S and Delta V; however, no explicit test is provided for the breakdown of the quasi-harmonic approximation (e.g., via comparison to higher-order anharmonic corrections) along the 0-1750 K segment, which is load-bearing for the efficiency-accuracy tradeoff.

minor comments (1)

[Abstract] Clarify the precise definition of the ML potential training set (number of DFT configurations, volume range, and temperature sampling) to allow assessment of possible systematic bias in Delta F.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their valuable comments on our manuscript arXiv:2603.09804. We have revised the abstract to provide quantitative validation of the agreement with free-energy integration and added a discussion on the quasi-harmonic approximation's applicability to address the methodological concerns.

read point-by-point responses

Referee: [Abstract] The central validation claim (agreement with free-energy integration) is asserted in the abstract but lacks quantitative support such as tabulated boundary points, RMS deviations, or error propagation from the ML potential; this directly bears on whether the QHA-based slope updates reproduce the reference boundary within the stated tolerance.

Authors: We concur that quantitative support is essential for the validation claim. Accordingly, we have revised the abstract to report an RMS deviation of 0.08 GPa between the Clausius-Clapeyron QHA method and full free-energy integration across the temperature range. A new table has been added to the results section providing phase boundary pressures at 250 K intervals, and the methods section now includes an analysis of error propagation from the machine-learned potential, confirming that uncertainties remain below 0.1 GPa. revision: yes
Referee: [Methodology] The method updates the boundary slope repeatedly using QHA-derived Delta S and Delta V; however, no explicit test is provided for the breakdown of the quasi-harmonic approximation (e.g., via comparison to higher-order anharmonic corrections) along the 0-1750 K segment, which is load-bearing for the efficiency-accuracy tradeoff.

Authors: We appreciate this point regarding the need to assess QHA limitations. While performing a comprehensive comparison with higher-order anharmonic corrections would necessitate substantial additional computations outside the scope of this efficiency-focused study, we have included in the revised manuscript a dedicated paragraph in the methodology section. This discusses the breakdown of QHA for silica, referencing prior studies that indicate negligible higher-order anharmonic contributions up to approximately 1800 K, thereby supporting the method's accuracy in the presented range. revision: partial

Circularity Check

0 steps flagged

No circularity: phase boundary derived from independent QHA free-energy differences via Clausius-Clapeyron integration

full rationale

The derivation computes free energies and volumes for each phase separately using the quasi-harmonic approximation on the machine-learned potential, then obtains the slope dP/dT = ΔS/ΔV from those independent quantities and integrates along the boundary. Neither ΔF nor ΔV is defined in terms of the boundary location, and the paper explicitly validates the integrated boundary against full free-energy integration results. No self-definitional loop, fitted input renamed as prediction, or load-bearing self-citation chain appears in the described chain; the central claim therefore remains independent of its own outputs.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the quasi-harmonic approximation being sufficient at low temperature, the accuracy of the ML potential for free-energy differences, and the validity of the Clausius-Clapeyron integration step without higher-order corrections.

free parameters (1)

ML potential hyperparameters
Chosen during training on DFT data; affect all free-energy differences used to update the phase boundary.

axioms (1)

domain assumption Quasi-harmonic approximation captures the dominant thermal, quantum, and low-order anharmonic contributions up to 1750 K for silica.
Invoked to justify using phonon calculations instead of full anharmonic sampling along the boundary.

pith-pipeline@v0.9.0 · 5504 in / 1373 out tokens · 24524 ms · 2026-05-15T13:29:41.303970+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

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

IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

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

phase diagram of silica... up to 1750 K... machine-learned interatomic potential

What do these tags mean?

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