Multi-Component Phase Stability
Composition geometry · convex hulls · finite temperature
Nico Unglert
22.04.2026
Use arrow keys to navigate · Press F for fullscreen
Outline
From composition geometry to finite-T stability
1. Composition geometry
- Binary — Fe–O
- Ternary — triangle
- Quaternary — LiMnOF
- Quinary — (Li,Mn,Ti)(O,F)
2. Stability at T = 0
- Convex-hull LP (theory)
- Binary hull — Fe–O
- Ternary hull — Li–Mn–O
3. Stability at T > 0
- Free-energy hull (theory)
- Entropy softens the hull
- Sampling F(x, T) with MC
Appendix — Constrained Gibbs triangle (Li,Mn,Ti)(O,F)
Thermodynamical stability at T = 0
Convex hull of formation energies
Formation energy (per config \(\sigma\))
$$\Delta E_f(\sigma) = E(\sigma) - \sum_i x_i(\sigma)\, E_i^{\text{ref}}$$
Convex hull (function of \(\mathbf{x}\))
$$E_{\text{hull}}(\mathbf{x}) = \min_{\substack{\lambda_i \,\ge\, 0 \\[2pt] \sum_i \lambda_i\, \mathbf{x}(\sigma_i) \,=\, \mathbf{x} \\[2pt] \sum_i \lambda_i \,=\, 1}} \; \sum_i \lambda_i\, \Delta E_f(\sigma_i)$$
A linear program (LP) in \(\{\lambda_i\}\): optimum sits on a facet with \(\le d{+}1\) non-zero \(\lambda_i\) — the decomposition products.
Energy above hull
$$E_{\text{above}}(\sigma) = \Delta E_f(\sigma) - E_{\text{hull}}\!\big(\mathbf{x}(\sigma)\big) \;\ge\; 0$$
Stable ⇔ \(E_{\text{above}}(\sigma)=0\) — \(\sigma\) sits on a hull vertex.
Workflow
- Enumerate configurations \(\sigma\) at each composition
- Predict \(E(\sigma)\) with fitted CE
- Construct formation-energy hull \(E_{\text{hull}}(\mathbf{x})\)
- Hull vertices = thermodynamically stable phases
Thermodynamical stability at T > 0
Convex hull of free energies — rebuilt at each temperature
Helmholtz free energy (function of \(\mathbf{x}, T\))
$$F(\mathbf{x}, T) = U(\mathbf{x}, T) - T\, S_{\text{conf}}(\mathbf{x}, T)$$
Free-energy hull (at fixed \(T\))
$$F_{\text{hull}}(\mathbf{x}, T) = \min_{\substack{\lambda_i \,\ge\, 0 \\[2pt] \sum_i \lambda_i\, \mathbf{x}_i \,=\, \mathbf{x} \\[2pt] \sum_i \lambda_i \,=\, 1}} \; \sum_i \lambda_i\, F_i(T)$$
Same linear program as \(T = 0\) with \(F\) in place of \(\Delta E_f\); supporting facet = common-tangent line (binary) / plane (ternary).
Coexistence condition
$$\mu_i(\mathbf{x}^{(\alpha)}, T) = \mu_i(\mathbf{x}^{(\beta)}, T) \quad \forall i$$
Chemical potentials \(\mu_i = \partial F / \partial x_i\) equal across coexisting phases ⇒ common tangent.
Workflow
- Sample \(U(\mathbf{x},T)\) with canonical MC at many \(T\)
- Thermodynamic integration \(\to S_{\text{conf}} \to F(\mathbf{x},T)\)
- Build convex hull of \(F(\cdot,T)\) at each \(T\)
- Entropy can stabilise phases above the \(T=0\) hull
Summary
Composition geometry
Each constraint removes one DOF. Sum-to-1 → simplex · sublattice → product of sub-simplices · charge balance → flat facet.
Stability at T = 0
\(E_{\text{hull}}(\mathbf{x}) = \min_{\boldsymbol{\lambda}} \sum_i \lambda_i\, \Delta E_f^{(i)}\) · stable ⇔ hull vertex; decomposition = optimal \(\lambda_i\).
Stability at T > 0
\(F(\mathbf{x}, T) = U - T\, S_{\text{conf}}\) · same LP with \(F_i(T)\) · entropy softens the hull.
Common tangent
≡ hull facet. Supporting hyperplane touching \(F\) at coexisting compositions; \(\mu_i = \partial F/\partial x_i\) equal across phases.
In practice
MC (canonical / semi-grand canonical) + thermodynamic integration deliver \(F(\mathbf{x}, T)\); the hull is re-built at each temperature.
Appendix — Constrained Gibbs Triangle: (Li,Mn,Ti)(O,F)
5 elements − 3 constraints = 2 DOF → Gibbs triangle with adjustable basis
Basis:
Same grid, different reference corners