Take
Its variational chemical potential is
Conservation gives . Assuming an isotropic constant mobility , local linear irreversible thermodynamics gives . Hence the noiseless conserved order-parameter dynamics is
It decreases the free energy because under closed or no-flux boundary conditions.
For and , planar coexistence occurs at with . Put . A common small shift gives equal chemical potentials to first order,
The thermodynamic pressure is . Its difference between the positive interior and negative exterior is therefore
For a sphere the Young–Laplace equation gives . Consequently the Gibbs--Thomson relation is
The curved interface requires exterior composition , but the far field supplies only . Material therefore diffuses away from the droplet, which evaporates. Write outside. Linearization gives , and the quasistatic condition gives
Spherical symmetry yields . Taking the normal outward from the droplet,
The interface converts positive-phase material into negative-phase material. Integrating the continuity equation through the moving interface gives the Stefan condition
where , and hence
For , the linearized mean curvature is
The boundary value in the upper phase is therefore
The decaying harmonic extension into is
With the normal directed into the upper half-space,
Using only this upper-phase flux in gives
Positive surface tension raises the chemical potential at a crest, drives material away from it, and therefore smooths the interface, which fixes the negative sign of the decay rate.
The dynamics is nonlocal because the conserved composition must diffuse through the bulk. Solving Laplace's equation maps a boundary Fourier amplitude to its normal derivative by multiplication by . Combining this Dirichlet-to-Neumann factor with the curvature factor produces the nonanalytic rate described by diffusion-limited relaxation of an interface.
Both coexisting phases transport the conserved order parameter. The calculation above included only the harmonic field and normal current in the upper half-space. The lower half-space contributes an equal flux into the interface for equal mobilities and symmetric phases. The total interface speed is therefore twice the one-sided result, so the correct coefficient is

Articles by others on the same topic (0)

There are currently no matching articles.