Curvature-induced melting-temperature depression Created 2026-10-07 Updated 2026-10-20
Curvature depresses the equilibrium melting temperature of solid convex into liquid. Here is the sum of principal curvatures, equal to twice a common convention for mean curvature, and is absolute temperature. The coefficient uses solid-liquid surface energy, mass density and latent heat. This stabilizes fine corrugations of a solidification front.
Take upward from the unperturbed interface in the apparatus frame, with liquid at and pulling velocity . The primary ice contains no salt. We use the dilute, equal-density approximation and neglect fluid motion relative to the mould. The initial salt inventory per unit horizontal area, with a common density factor suppressed, is . Its conservation is essential: the far liquid has zero concentration, so an imposed nonzero far-field composition would describe a different experiment.
The steady advection-diffusion equation is . Decay into the pure water and interfacial salt rejection give
Since , the finite-inventory directional salt rejection solution is
The decay length is . Local liquidus equilibrium gives
Thus the thermal apparatus selects the interface position at this isotherm. The stated speed inequality is precisely , so the interface is warmer than the eutectic temperature, . These are steady values after the initial finite saline layer has rearranged; the final exponential distribution is not confined to the original depth .
For , a steady pure-ice front would require an interfacial composition beyond the eutectic composition. That branch is unavailable. During the transient, rejected brine concentrates until it reaches ; secondary salt-bearing crystals can then form and the residual brine freezes as an eutectic system. Salt is incorporated into the product rather than remaining indefinitely in a super-eutectic exponential layer. The detailed pattern can include brine pockets and a nonplanar freezing region; it depends on the thermal history and the morphological stability. The speed inequality alone does not decide that latter stability.
For the stability calculation use an interface displacement and a liquid perturbation . Here is the wavenumber, is the temporal rate and gives spatial decay. The frozen thermal field is . Define , where is solid-liquid surface energy, is absolute pure-solvent melting temperature, is mass density and is specific latent heat. For a solid protrusion into liquid the curvature is , and the Gibbs-Thomson effect gives local equilibrium . We assume isotropic surface energy and instantaneous interface kinetics, so no kinetic undercooling term is present.
Linearizing the equilibrium condition at the displaced interface gives
The perturbation advection-diffusion equation gives
The local solute balance at a moving graph, to first order, is
Its perturbation is , because the terms and cancel. Consequently the full frozen-temperature solutal interface dispersion is
No quasistatic approximation to solute diffusion was made here. For a growing mode the unique admissible liquid root is
Writing and eliminating gives . The root continuously connected to the growing interface branch is therefore
In particular, unstable wavelengths exist exactly when , and their band is
This is the constitutional supercooling criterion: immediately above the plane, the local liquidus rises at rate , so it overtakes the imposed thermal gradient when that rate exceeds . The Gibbs-Thomson effect cuts off short waves. At long wavelength for positive . The exactly uniform inventory-changing displacement is not an instability of a fixed-inventory experiment. On stable branches the full diffusion problem also admits relaxation transients; the root selection above is sufficient to identify every exponentially growing interface mode.
If one additionally neglects the perturbation time derivative in the liquid diffusion equation, the alternative reduced rate is
It has the same onset and cutoff, but need not have the same rate away from the slow-growth regime. The frozen-temperature approximation by itself does not justify this extra reduction.
After onset, protrusions reach more weakly saline liquid and advance faster, while grooves accumulate rejected brine. The planar interface develops cells and, as growth becomes sufficiently strong, crystal dendrites with concentrated liquid between them. Nonlinear competition changes the selected spacing; linear rates alone cannot fix the final geometry. The interdendritic liquid forms a mushy layer, concentrates toward the eutectic composition, and ultimately freezes into a salt-bearing eutectic product when it reaches . This replaces the pure planar salt-rejection picture with trapping and freezing of brine between crystals.
Use the one-sided thermal conduction model encoded by the requested dispersion relation: neglect solid-side perturbation heat flux, take equal phase densities, and let the normal point from solid into liquid. Write the isotropic linear kinetic undercooling law using the solidification attachment mobility :
Here is solid-liquid surface energy, is the sum of principal curvatures, and is absolute temperature. Positive curvature convex into liquid lowers the equilibrium freezing temperature through the Gibbs--Thomson relation. The planar base state has .
Set and define
The paper's is the latent-to-sensible heat ratio, the reciprocal of the common Stefan number convention. Below, velocities, lengths and growth rates are dimensionless in these scales. In the travelling coordinate , the liquid heat equation gives , while the bounded solid steady field is constant. Matching a decaying liquid profile to the Stefan condition gives
The kinetic undercooling condition is . Thus
The speed is constant for , determined by and the attachment coefficient in the specified scales. Thermal diffusivity sets the exponential decay length , not the selected speed. Including solid-side perturbation conduction would change the stability equation below.
For linear stability analysis, corrugate the interface to and write the liquid perturbation as . The liquid heat equation requires
For this graph the linearized curvature is . Evaluate the temperature on the displaced interface: its perturbation is . The kinetic undercooling law therefore gives
Likewise the derivative evaluated at the displaced surface is , with . The Stefan condition at perturbation order reads
Elimination proves the one-sided kinetic solidification dispersion:
The destabilizing term comes from a protrusion encountering colder liquid; capillarity and kinetic relaxation oppose it.
At , and the base liquid has no thermal contrast generated by latent heat. The interface obeys : positive surface energy damps all nonzero modes, while is neutral at this order. Here the thermal perturbation amplitude vanishes, so the auxiliary decay exponent is immaterial to that pure kinetic relaxation. At , and the formal limit gives
For it has a negative real branch and no growing branch; at the interface branch is neutral. Strictly, is a singular endpoint of the travelling profile: no finite-width stationary exponential connects to the far-field value zero. These stability statements describe the limit of the travelling-front model.
For , assume at small wavenumber. Expand
Substituting into the dispersion relation cancels on both sides and gives
At the capillary stability threshold for a kinetic freezing front , continuing the same expansion gives : the nonzero long waves are damped even at threshold.
For large wavenumber, multiply the dispersion relation by and eliminate using the thermal equation. The exact algebraic form is
It prevents the uncontrolled replacement when itself is of order . In the usual small-capillarity range the leading results are
For the second result, set in the cubic; its quadratic-order balance gives . Thus positive capillarity stabilizes sufficiently short waves. If also and , the useful reduced approximation is
its finite terms should not be treated as a uniform expansion for arbitrary fixed .
For completeness, large dimensionless capillarity has different balances. At , and . For , , so . At , the real branch reaches : for its growth rate is , and for that endpoint is exact. For at this boundary, or for , the selected real- normal mode ceases to exist at sufficiently large wavenumber; the full diffusion problem includes a thermal continuous spectrum and cannot be assigned a real isolated interface growth rate by this ansatz there. Indeed, writing the bulk thermal perturbation as reduces its heat equation to , whose half-line diffusion modes have decay rates for . The thermal branch endpoint is . Loss of a damped isolated interface mode is not a morphological instability.
The approximate instability condition is already clear from the long-wave coefficient: , equivalently . In fact it is exact for growing modes of the present one-sided model. At , let , so . The right side of the dispersion relation becomes
For real , the residual is strictly increasing, since
Therefore a unique positive root exists precisely when . For , the unstable band is
There are no unstable oscillatory roots hidden by this real-root test. Indeed, if and , the thermal equation implies . For , the imaginary parts of the two equations would require , contradicting . Thus growing modes have real and real .
Figure 1.
Growth-rate curves for a one-sided kinetic freezing front
. Details.
The sketch shows zero capillarity approaching the positive plateau , weak capillarity giving a positive band followed by short-wave damping, and threshold/stronger capillarity giving damping at every nonzero wavenumber. The zero-wavenumber mode is a translation. As with fixed positive , both the unstable bandwidth and the positive growth rates shrink to zero; the small-wavenumber expansion is not uniform in that limit.
Let increase into the liquid, with the planar solidification front at . Denote the common thermal conductivity by . A bounded travelling temperature profile satisfies . In the solid, boundedness as forces ; in the liquid,
The Stefan condition then gives
Thus the travelling solidification front in a pure supercooled melt releases just enough latent heat to warm the initially supercooled liquid to . There is no separate imposed heat flux at infinity in this bounded travelling-wave problem.
For the interface perturbation write . Use the Gibbs--Thomson relation for curvature-induced melting-temperature depression
Here is the solid-liquid surface energy, in this expression is absolute temperature in kelvin, and is the sum of principal curvatures, positive for solid convex into liquid. For this graph, . If mean curvature is defined as half this sum, ; this fixes the capillary factor unambiguously.
Under the small-wavelength thermal approximation for a solidification front, take harmonic perturbation temperatures that decay away from the interface:
This approximates their temperature equation by Laplace's equation, neglecting perturbation time dependence and the moving-frame advection. Expanding the interfacial temperature at the displaced surface gives
The second expression includes displacement through the negative base liquid gradient. Linearizing the Stefan condition also requires evaluation of the base gradient at the displaced surface:
Using gives the requested reduced dispersion relation:
or
The destabilizing term describes a protrusion entering colder liquid. The cubic term expresses stabilization by the curvature-induced temperature decrease.
There is an important accuracy qualification. The harmonic approximation gives the leading short-wave thermal exponents, but the retained constant term is not a uniformly controlled next-order correction to the full moving-boundary problem. If perturbation diffusion is retained, its exponents are
Substitution into exactly the same linearized boundary balances gives the full thermal stability of a travelling solidification front:
For example, with capillarity removed, solves this relation exactly, whereas the harmonic reduction gives . The difference is smaller than the leading term when , but demonstrates why the reduced formula should not be asserted as an exact dispersion law or a precise onset calculation. To neglect perturbation time dependence one also needs ; extremely large capillary damping is outside that approximation.
To analyse the displayed cubic, define
Then . It is negative at zero and at sufficiently large . Its positive stationary point is the unique maximum:
Thus the capillary threshold of a reduced solidification dispersion is
with equality giving a neutral maximum and strict inequality giving a band of unstable positive wavenumbers. At equality, however, , not . The number is therefore the threshold of the stated reduced cubic, not a controlled threshold of the full thermal problem under the stated short-wave ordering. Far above this threshold its fastest-growing scale does satisfy , making the short-wave mechanism meaningful.
The analogy for a snowflake is dendritic growth of a snowflake. A protrusion has better access to the driving supersaturation or undercooling and grows faster than a sheltered region; this amplifies tips and promotes branching into crystal dendrites. Interfacial capillarity suppresses arbitrarily fine tips. In an actual snowflake, the main growth is deposition from water vapour, so the diffusion field and its driving variable differ from this pure-melt thermal example. The hexagonal ice crystal anisotropy supplies six equivalent preferred arm directions; an isotropic planar stability calculation alone cannot predict six arms. Nonlinear tip selection, vapour transport and side branching are needed for the later detailed morphology. The analysis explains the competition between diffusion-driven protrusion growth and capillary smoothing, with this physically necessary qualification.
The attachment mobility is the coefficient converting interfacial undercooling into a solidification speed. In the linear isotropic law, , and capillarity makes . Thus . This kinetic coefficient refers to a moving phase boundary, rather than an order-parameter relaxation coefficient. The Gibbs--Thomson relation determines for isotropic solid-liquid surface energy .