Fact-checked by Grok 2 weeks ago

Stefan problem

The Stefan problem is a class of moving boundary value problems in partial differential equations that model phase transitions in materials, such as the melting of a solid or the freezing of a liquid, where the interface between phases evolves over time due to heat transfer.^[1] It typically involves solving the heat equation in each phase, subject to initial and boundary conditions, along with the Stefan condition at the interface, which balances the heat fluxes from both sides with the latent heat released or absorbed during the phase change: k_s \frac{\partial T_s}{\partial n} - k_l \frac{\partial T_l}{\partial n} = \rho L v, where k denotes thermal conductivity, T temperature, \rho density, L latent heat, v the interface velocity, and n the normal direction.^[1] Named after the Slovenian-Austrian physicist Josef Stefan (1835–1893), who formulated it in 1889 while studying the growth of polar ice based on expedition measurements, the problem builds on earlier 19th-century work by mathematicians like Gabriel Lamé, Émile Clapeyron, and Franz Neumann, who addressed related heat conduction issues in phase changes.^[2] Stefan's original application involved comparing theoretical predictions of ice thickness with empirical data from Arctic explorations, marking it as one of the first free-boundary problems in mathematical physics.^[2] Over time, the framework has expanded to include one-phase (e.g., only liquid above melting temperature) and two-phase variants, with explicit similarity solutions like the Neumann solution providing benchmarks for more complex numerical methods.^[3] The Stefan problem is fundamental in applied mathematics and engineering, with applications in processes like metal casting, crystal growth, and latent heat thermal energy storage systems, where accurate prediction of interface motion is crucial for design and efficiency.^[1] Its study has grown extensively since the mid-20th century, driven by advances in computational techniques to handle nonlinearities from variable properties, convection, or mushy zones, influencing fields from materials science to geophysics.^[2]

History and Development

Historical Origins

The movement of phase boundaries during natural phenomena, such as the formation of ice on bodies of water and the solidification of molten metals in casting, was noted by 19th-century scientists as a process governed by heat transfer across the interface between phases.^[4] These observations underscored the need for mathematical models to describe the dynamic evolution of the boundary under varying thermal conditions.^[2] In 1831, Gabriel Lamé and Benoît Paul Émile Clapeyron published the first mathematical treatment of transient heat conduction in a solid with a phase interface, modeling the solidification of a liquid sphere cooling from its surface while assuming the interior liquid remained at the melting temperature.^[5] Their work, detailed in "Mémoire sur la solidification par refroidissement d’un globe solide" in the Annales de Chimie et de Physique, calculated the thickness of the growing solid shell as proportional to the square root of time, laying foundational concepts for later moving-boundary problems.^[6] Josef Stefan advanced this framework in a series of publications between 1889 and 1891, introducing the moving boundary condition for unsteady heat conduction in phase-change processes like melting and freezing.^[2] In his seminal 1889 paper "Über die Theorie der Eisbildung, insbesondere über die Eisbildung im Polarmeere," published in the Sitzungsberichte der Mathematisch-Naturwissenschaftlichen Classe der Kaiserlichen Akademie der Wissenschaften, Stefan applied the model to ice formation in polar seas, deriving the Stefan condition that equates the latent heat released or absorbed at the interface to the difference in heat fluxes across the boundary. He extended these ideas in subsequent works, including a 1891 publication in Annalen der Physik und Chemie, generalizing the approach to arbitrary geometries and initial conditions.^[4] Stefan validated his theoretical model through comparisons of predicted ice thickness with empirical data from polar expeditions, demonstrating the model's accuracy in capturing the square-root-of-time dependence of interface position observed in natural settings.^[6]^[2] As a precursor, Franz Neumann had derived a similarity solution in 1860 for a related transient problem akin to Lamé and Clapeyron's formulation.^[4]

Key Contributors and Milestones

In the mid-20th century, John Crank's seminal book The Mathematics of Diffusion (1956) formalized the Stefan problem as a free boundary partial differential equation, highlighting approximations for the one-phase case where only the phase change region is modeled dynamically.^[7] This work provided a foundational framework for analyzing diffusion-driven phase transitions, integrating latent heat release as a key physical premise.^[8] During the 1970s, Olga Oleinik advanced the theoretical understanding by proving existence and uniqueness theorems for weak solutions to multidimensional Stefan problems, building on earlier introductions of weak formulations and enabling broader analytical tractability.^[9] Concurrently, Boris Rubinsky and Eduardo G. Cravalho applied the Stefan model to biomedical contexts, such as cryosurgical freezing of biological tissues, demonstrating its utility in predicting thermal stresses and phase boundaries in vivo. The 1980s and 1990s saw significant progress in regularity theory, led by Luis A. Caffarelli, whose collaborations established optimal regularity results for free boundaries in the one-phase Stefan problem, including characterizations of non-smooth behaviors like cusps and corners at interfaces.^[10] These results quantified the smoothness of evolving interfaces under parabolic scaling, resolving longstanding questions about solution stability.^[11] A recent milestone came in 2024 with Alessio Figalli, Xavier Ros-Oton, and Joaquim Serra's proof that the singular set of the free boundary in the Stefan problem has parabolic Hausdorff dimension at most n - 1/2 in n dimensions, with constant expansion at singular points, sharpening prior bounds on interface irregularities.^[10] This work refines the geometric structure of singularities, impacting both theoretical and applied analyses. Post-2000, research shifted toward computational methods and inverse problems, with inverse Stefan formulations enabling parameter estimation—such as latent heat or diffusivity—from observed phase boundaries, as exemplified in numerical schemes using method of fundamental solutions.^[12] This evolution facilitated simulations in engineering and materials science, bridging analytical foundations with practical optimization.^[13]

Physical and Mathematical Foundations

Physical Premises of Phase Transitions

Phase transitions involve changes between distinct states of matter, primarily solid, liquid, and gas, where the molecular arrangement and thermal energy differ significantly. In the solid phase, molecules vibrate around fixed lattice positions with relatively low thermal energy, maintaining a rigid structure. The liquid phase features molecules that slide past one another with increased thermal motion, allowing flow while retaining short-range order. The gas phase, though less central to classical Stefan problems, consists of widely separated molecules moving freely with high kinetic energy. These phases are separated by interfaces, which are thin regions of abrupt changes in material properties such as density (typically 5-10% variation, up to 30% in some cases) and specific heat capacity (e.g., approximately doubling from ~2.09 kJ/kg·K for ice to 4.19 kJ/kg·K for water at 0°C).^[14]^[15] A key physical feature of phase transitions is the involvement of latent heat, the energy absorbed or released during the change without altering the temperature at the interface. During melting, latent heat is absorbed to overcome intermolecular forces, transitioning the material from solid to liquid at the constant melting temperature; conversely, freezing releases this heat, solidifying the material. This isothermal process at the interface ensures that the phase change occurs at a fixed temperature, such as 0°C for the ice-water transition, distinguishing it from sensible heat effects where temperature varies with energy input. The magnitude of latent heat reflects the energy required to reorganize molecular bonds, playing a crucial role in the dynamics of the moving boundary between phases.^[14] The movement of the phase interface is primarily driven by heat conduction, the transfer of thermal energy through the material without bulk motion, as described by Fourier's law, which states that heat flux is proportional to the negative temperature gradient. This conduction creates temperature imbalances that supply or remove the latent heat needed for the phase change, causing the interface to advance or recede. In classical models, this process assumes thermodynamic equilibrium at the interface, where the phases coexist stably at the melting temperature, and kinetic effects—such as undercooling or rapid attachment rates—are negligible, simplifying the description to diffusion-dominated transport. These assumptions hold for many natural and industrial scenarios where interface velocities remain slow relative to molecular scales.^[14]^[16] Representative examples illustrate these principles. In glaciers, the ice-water interface advances during freezing as heat conduction through the ice releases latent heat into surrounding water, forming thicker ice layers without temperature rise at the boundary; this process is evident in refreezing of meltwater within glacial structures. In contrast, rapid solidification in metallurgy, such as during continuous casting of alloys, involves high cooling rates where latent heat release at the solid-liquid interface influences microstructure formation, though classical assumptions may require adjustments for non-equilibrium conditions at very high speeds. These cases highlight how phase transitions govern material behavior in environmental and manufacturing contexts.^[17]^[18] Josef Stefan's 19th-century observations of freezing fronts in polar seas provided early empirical motivation for studying these physical processes.^[19]

Mathematical Prerequisites and Closure Conditions

The mathematical modeling of phase transitions in the Stefan problem relies on the heat equation as the governing partial differential equation for temperature evolution within each phase. In both the solid and liquid phases, the temperature u(x, t) satisfies the diffusion equation

\frac{\partial u}{\partial t} = \alpha \frac{\partial^2 u}{\partial x^2},

where \alpha = k / (\rho c) denotes the thermal diffusivity, with k the thermal conductivity, \rho the density, and c the specific heat capacity (potentially varying between phases). This equation, derived from Fourier's law of heat conduction, assumes isotropic and homogeneous material properties within each phase and neglects convective effects or other transport mechanisms. To specify the problem in a domain, boundary conditions are required at fixed outer boundaries, typically involving prescribed temperature (Dirichlet) or heat flux (Neumann) conditions, such as u = u_0 or \frac{\partial u}{\partial x} = q at the domain endpoints. Across the phase interface, continuity of temperature is enforced to ensure thermodynamic equilibrium, yielding u_s = u_l = u_m at x = s(t), where subscripts s and l denote the solid and liquid phases, respectively, and u_m is the phase-change temperature (often taken as the melting point). These conditions maintain the physical consistency of the temperature field while the interface evolves. The Stefan problem constitutes a free boundary problem due to the unknown a priori position of the interface s(t), which moves according to the phase transition dynamics. The evolution of s(t) necessitates a kinematic condition that couples the interface velocity to the underlying temperature distribution, transforming the problem into a system where both the solution and the domain boundary are to be determined simultaneously. Without such a closure, the mathematical system remains underdetermined. Closure is achieved through the Stefan condition, which enforces energy conservation by equating the rate of latent heat release or absorption to the jump in conductive heat flux across the interface:

\rho L \frac{ds}{dt} = k_s \frac{\partial u_s}{\partial x} - k_l \frac{\partial u_l}{\partial x} \bigg|_{x = s(t)},

where L is the latent heat per unit mass. This condition captures the discontinuity in heat flux arising from the phase change, with the sign depending on melting or solidification. Initial conditions for u(x, 0) and s(0) further complete the formulation, enabling the prediction of the transient interface motion. Despite this closure, the Stefan problem presents significant challenges to well-posedness, particularly in higher dimensions or without regularization terms like surface tension. Classical solutions may exhibit interface singularities, such as cusp formation or infinite curvature, which can lead to loss of regularity and breakdown of existence or uniqueness over finite time intervals. These issues arise from the nonlinear coupling between the parabolic equation and the free boundary, highlighting the need for careful analysis of solution smoothness and stability.^[20]

Classical Mathematical Formulation

One-Dimensional One-Phase Problem

The one-dimensional one-phase Stefan problem models the solidification of a liquid initially at its melting temperature in a semi-infinite domain. The liquid occupies the region x > 0 at t = 0, with temperature uniform at u_m. A cold boundary condition u(0, t) = u_0 < u_m is applied at the fixed wall x = 0, inducing solidification and forming a solid layer in 0 < x < s(t), where s(t) denotes the position of the moving phase-change interface. In this approximation, the liquid remains at uniform temperature u_m (no superheat or subcooling), and the temperature u(x, t) in the solid region 0 < x < s(t) obeys the heat equation

\frac{\partial u}{\partial t} = \alpha \frac{\partial^2 u}{\partial x^2}, \quad 0 < x < s(t), \ t > 0,

with thermal diffusivity α assumed constant. The interface is maintained at the melting temperature, u(s(t), t) = u_m, and the initial interface position is s(0) = 0. The Stefan condition at the interface x = s(t) couples the motion of the free boundary to the heat flux, accounting for the release of latent heat during solidification:

\rho L \frac{ds}{dt} = k \left. \frac{\partial u}{\partial x} \right|_{x = s(t)^-},

where ρ is the density, L is the latent heat of fusion, and k is the thermal conductivity (all assumed constant). In this one-phase formulation, heat conduction is considered only in the solid phase, with the liquid phase idealized as having no temperature gradient.^[3] This setup assumes a semi-infinite domain and constant material properties, simplifying the problem to capture the essential physics of phase change driven by a fixed cold boundary. The initial uniform temperature u_m in the liquid implies no superheat, so the temperature field in the solid 0 < x < s(t) develops a gradient due to heat extraction at the cold wall. The problem is well-posed under these conditions, with the interface advancing as latent heat is released and conducted through the growing solid layer.^[3] An exact similarity solution exists for this configuration. The interface position takes the self-similar form

s(t) = 2\lambda \sqrt{\alpha t},

where λ is a constant determined by the transcendental equation

\sqrt{\pi} \lambda e^{\lambda^2} \erf(\lambda) = \frac{c (u_m - u_0)}{L}.

Here, c is the specific heat capacity of the solid, u_0 is the boundary temperature, and the right-hand side is the Stefan number Ste, quantifying the relative importance of sensible to latent heat. The parameter λ is solved numerically for given material parameters and is unique for physically relevant Stefan numbers greater than zero. This solution assumes the similarity variable η = x / (2 √(α t)), reducing the partial differential equation to an ordinary differential equation in η. The resulting temperature profile in the solid is

u(x, t) = u_0 + (u_m - u_0) \frac{\erf(\eta)}{\erf(\lambda)},

ensuring compatibility with the boundary and interface conditions. This similarity solution provides insight into the square-root time dependence of the interface motion, characteristic of diffusion-driven phase changes in semi-infinite domains. However, it relies on several idealizations, including constant thermophysical properties (no temperature dependence of c, k, α, or L), a purely diffusive heat transfer mechanism (neglecting convection or radiation), and an infinite domain (no finite-size effects or opposite boundary). These assumptions limit applicability to early-time behavior or large-scale systems but break down for long times or when property variations are significant, necessitating numerical or approximate methods for more realistic scenarios.

One-Dimensional Two-Phase Problem

The one-dimensional two-phase Stefan problem describes phase transitions where both solid and liquid phases actively participate, with heat diffusion occurring in each domain separated by a moving interface s(t). In the classical semi-infinite formulation, the solid phase occupies the region x < s(t), while the liquid phase occupies x > s(t), over x \in (-\infty, \infty). Far-field conditions are imposed: as x \to -\infty, u_s \to T_s < u_m in the solid, and as x \to \infty, u_l \to T_l > u_m in the liquid. This setup captures asymmetric thermal influences from both phases, with the net motion depending on the Stefan numbers in each phase. Finite domain versions exist but are extensions of the classical case.^[21] In the solid domain, the temperature u_s(x, t) satisfies the heat equation

\frac{\partial u_s}{\partial t} = \alpha_s \frac{\partial^2 u_s}{\partial x^2}, \quad x < s(t), \quad t > 0,

where \alpha_s = k_s / (\rho_s c_s) is the thermal diffusivity, with k_s, \rho_s, and c_s denoting thermal conductivity, density, and specific heat of the solid, respectively. Similarly, in the liquid domain,

\frac{\partial u_l}{\partial t} = \alpha_l \frac{\partial^2 u_l}{\partial x^2}, \quad x > s(t), \quad t > 0,

with corresponding liquid properties \alpha_l = k_l / (\rho_l c_l). At the interface, temperature continuity holds:

u_s(s(t), t) = u_l(s(t), t) = u_m,

where u_m is the melting temperature. The Stefan condition enforces energy balance across the interface:

\rho_s L \frac{ds}{dt} = k_s \frac{\partial u_s}{\partial x}\bigg|_{x = s(t)^-} - k_l \frac{\partial u_l}{\partial x}\bigg|_{x = s(t)^+},

reflecting the latent heat L absorbed or released during phase change, with heat fluxes directed appropriately based on the interface normal.^[21] Initial conditions specify the starting interface position s(0) = 0 and uniform temperature distributions: u_s(x, 0) = T_s < u_m for x < 0 in the solid, and u_l(x, 0) = T_l > u_m for x > 0 in the liquid. These conditions ensure compatibility with the far-field temperatures and interface requirements. Unlike the one-phase problem, which simplifies by assuming one phase remains at u_m and focuses on a single temperature field, the two-phase version couples two distinct fields with potentially differing properties, leading to greater mathematical complexity in tracking the interface evolution. Solutions to this problem require additional regularity assumptions, such as Lipschitz continuity of the initial data or bounds on the interface velocity, to ensure well-posedness; without them, there is potential for non-uniqueness, particularly in weak formulations where multiple interface paths may satisfy the conditions.^[22]

Analytical Approaches and Solutions

Similarity and Exact Solutions

The similarity solutions for the Stefan problem provide closed-form expressions for the temperature fields and interface position under idealized conditions, such as semi-infinite domains and constant material properties. These solutions, first derived in the nineteenth century and elaborated in the mid-twentieth century, rely on self-similar variables that reduce the partial differential equations to ordinary differential equations solvable via error functions. In the one-dimensional case, the Neumann solution for the one-phase problem was extended to the two-phase formulation, capturing both solid and liquid phases.^[3] For the classical two-phase Stefan problem in one dimension, the interface position s(t) advances as s(t) = 2\lambda \sqrt{\alpha_s t}, where \lambda is a constant determined by the Stefan condition, and \alpha_s is the thermal diffusivity of the solid phase. The temperature profile in the solid phase is given by

u_s(x,t) = u_0 + (u_m - u_0) \frac{\erf\left( \frac{x}{2\sqrt{\alpha_s t}} \right)}{\erf(\lambda)},

while in the liquid phase, it takes the complementary form

u_l(x,t) = u_l + (u_m - u_l) \frac{\erfc\left( \frac{x}{2\sqrt{\alpha_l t}} \right)}{\erfc\left( \lambda \sqrt{\frac{\alpha_s}{\alpha_l}} \right)},

with u_m the melting temperature, u_l the initial liquid temperature, u_0 the fixed boundary temperature in the solid, and \erf and \erfc the error function and its complement. The parameter \lambda satisfies a transcendental equation derived from the energy balance at the interface, involving the latent heat and Stefan number. These expressions were systematically derived and popularized by John Crank in the 1940s and 1950s through his work on diffusion with moving boundaries.^[7]^[3] Exact solutions extend to radial and spherical geometries under cylindrical or spherical symmetry, particularly relevant for processes like droplet solidification. In the one-phase Stefan problem for inward solidification of a cylindrical or spherical domain, the similarity variable \eta = r / \sqrt{4\alpha t} transforms the heat equation into a form solvable by confluent hypergeometric functions or series expansions. For spherical symmetry, the interface radius evolves as R(t) = \beta \sqrt{t}, where \beta is found by solving an algebraic equation balancing diffusion and latent heat release; the temperature field involves integrals of the Green's function adapted to spherical coordinates. These solutions apply to scenarios such as the freezing of a supercooled liquid droplet, where the solid phase grows radially inward from the surface. Crank provided explicit derivations for these cases, highlighting their utility for validating numerical models in bounded but symmetric domains.^[7]^[23] In multi-dimensional settings, self-similar solutions describe the evolution of planar fronts in two or three dimensions, assuming translational invariance perpendicular to the front. The temperature and interface satisfy the same error-function profiles as in one dimension, but perturbations introduce curvature effects analyzed via linear stability theory. Small amplitude perturbations of the form s(x_\perp, t) = 2\lambda \sqrt{\alpha t} (1 + \epsilon \cos(k \cdot x_\perp) e^{\sigma t / \sqrt{t}}) reveal the Mullins-Sekerka instability, where the growth rate \sigma becomes positive for long-wavelength modes when the velocity exceeds a critical value, leading to dendritic patterns. This instability arises from competition between stabilizing surface tension and destabilizing solute rejection or thermal diffusion fields, first quantified in the seminal analysis for diffusive growth. Uniqueness of these self-similar solutions holds under suitable boundary conditions in cylindrical and spherical symmetries.^[24] These exact and similarity solutions are inherently limited to infinite or semi-infinite domains with constant thermal properties and no convection, precluding direct application to finite geometries or variable coefficients where no closed forms exist. They serve primarily as benchmarks for theoretical understanding and numerical verification rather than practical predictions.^[7]

Regularity and Well-Posedness Results

The existence of weak solutions to the one-dimensional Stefan problem was established in the 1960s using approximation methods and comparison principles, with Olga Oleinik providing criteria based on monotonicity to ensure convergence to a weak solution satisfying the integrated form of the equations.^[25] These results rely on constructing a sequence of penalized problems where the phase change is regularized, and monotonicity arguments guarantee the limit satisfies the weak formulation of the heat equation and the Stefan condition in the sense of distributions.^[9] In higher dimensions, regularity theory for the one-phase Stefan problem advanced significantly in the 1980s through the work of Luis Caffarelli and Avner Friedman, who proved the continuity of the temperature field across the interface away from potential singularities.^[26] Building on this, they established that the free boundary exhibits C^{1,\alpha} regularity for some \alpha > 0 at regular points, meaning the interface is locally a graph of a C^{1,\alpha} function over its tangent plane, excluding isolated singular points where the normal might not exist. This partial regularity implies that singularities, if present, form a set of measure zero in the parabolic space-time domain. Singularities in the Stefan problem include interface cusps, where the free boundary develops sharp corners, and waiting times, during which the interface remains stationary despite nonzero temperature gradients, particularly in the one-phase setting with undercooled initial data. Recent advances by Alessio Figalli, Xavier Ros-Oton, and Joaquim Serra in 2024 characterize the singular set more precisely, showing it has parabolic Hausdorff dimension at most n-1 in n spatial dimensions and finite (n-1)-Hausdorff measure for almost every time slice, thereby bounding the "size" of possible cusps and excluding dense singularities.^[27] Examples of ill-posedness arise in multi-dimensional settings without small data assumptions, as demonstrated by Emanuele DiBenedetto and Avner Friedman, who constructed non-unique weak solutions for the supercooled two-phase Stefan problem, where multiple free boundary evolutions satisfy the same initial and boundary conditions. This non-uniqueness stems from the possibility of instantaneous complete solidification or partial freezing, highlighting the need for additional regularity or entropy conditions to select physical solutions. Modern developments emphasize partial regularity in higher dimensions through monotonicity formulas, analogous to those introduced by Hans Wilhelm Alt, Luis Caffarelli, and Avner Friedman for elliptic free boundary problems but adapted to the parabolic Stefan setting. These formulas quantify the growth of energy functionals around blow-up points on the free boundary, enabling the classification of regular points where the interface is C^\infty smooth and the identification of the singular set as stratified with decreasing dimensions.^[27]

Numerical Methods

Finite Difference and Front-Tracking Methods

Finite difference methods, combined with front-tracking techniques, provide a foundational approach for numerically solving Stefan problems by discretizing the heat equation on a fixed or adaptive grid while explicitly or implicitly resolving the moving phase-change interface. These methods are particularly suited to one-dimensional formulations, where the position of the interface s(t) is updated at each time step using the Stefan condition, which relates the interface velocity to the heat flux jump across phases. Early implementations in the 1970s employed explicit finite difference schemes to approximate solutions to both one-phase and two-phase problems, demonstrating practical feasibility despite stability constraints.^[28] Explicit finite difference schemes advance the temperature field via forward Euler time-stepping for the heat equation in each phase, followed by an interface update derived from the Stefan condition \dot{s}(t) = \frac{1}{L} \left( k_s u_x(s^-, t) - k_l u_x(s^+, t) \right), where L is the latent heat, k the thermal conductivity, and subscripts denote solid (s) and liquid (l) phases. To handle the latent heat without resolving the interface separately, the enthalpy method embeds it by reformulating the problem as a single-domain partial differential equation:

\frac{\partial H(u)}{\partial t} = \frac{\partial}{\partial x} \left( k(u) \frac{\partial u}{\partial x} \right),

where the enthalpy H(u) is a nonlinear function incorporating the sensible heat c u and a jump of magnitude L at the melting temperature, effectively treating the phase change as a diffusive process across the domain. This approach avoids explicit grid deformation near the interface but requires careful approximation of the nonlinear H(u), often via piecewise linear functions, and has been shown to maintain conservation properties in one-phase settings.^[29] Implicit schemes, such as the Crank-Nicolson method, offer improved stability for larger time steps by averaging explicit and implicit discretizations of the heat equation, solving a tridiagonal system at each step while front-tracking explicitly resolves s(t) through an iterative or extrapolated update based on the Stefan condition. In the Crank-Nicolson framework, the temperature in each phase is approximated to second-order accuracy in space and time, with the interface position advanced using a predictor-corrector strategy to ensure consistency with the moving boundary. This combination enhances unconditional stability in one-phase problems and has been applied to track interface evolution without severe oscillations, though computational cost increases due to matrix inversions.^[30] Error analysis for these methods reveals first-order convergence in time and space for explicit front-tracking schemes in the one-phase Stefan problem, yielding overall rates of O(\Delta t + \Delta x), limited by the interface update's accuracy and potential smearing of the temperature profile. In two-phase problems, challenges arise from the Gibbs phenomenon, where sharp discontinuities in temperature or flux across the interface induce spurious oscillations, reducing convergence and requiring smoothing or higher-order corrections to achieve reliable results. Early numerical implementations, such as those developed in the late 1970s using explicit schemes on uniform grids, validated these methods against similarity solutions for benchmark one-phase melting scenarios, establishing their utility despite the noted limitations.^[31]^[32]^[33]

Finite Element and Level-Set Methods

The finite element method (FEM) provides a variational framework for discretizing the heat equation in the subdomains separated by the moving interface in the Stefan problem, enabling simulations on complex geometries. In this approach, the weak formulation of the parabolic partial differential equation governing temperature evolution is approximated using piecewise polynomial basis functions, typically linear or quadratic, over a triangulation of the domain. The interface motion, driven by the Stefan condition, is handled by remeshing or deforming the computational mesh to track the free boundary accurately. A key advancement involves the arbitrary Lagrangian-Eulerian (ALE) formulation, which allows the mesh to move with a prescribed velocity independent of the material flow, thereby adapting to interface deformation without excessive distortion. This method has been shown to maintain stability and conservation properties in two-phase configurations, as demonstrated in coupled Stefan-Navier-Stokes simulations where capillary effects are included.^[34] The level-set method represents the interface implicitly as the zero-level set of a smooth scalar function \phi(\mathbf{x}, t), defined over a fixed Eulerian grid, which facilitates handling topological changes and complex interface geometries without explicit tracking. The evolution of the level-set function follows the advection equation

\frac{\partial \phi}{\partial t} + v_n |\nabla \phi| = 0,

where v_n is the normal velocity of the interface, derived from the Stefan condition incorporating heat fluxes from both phases. This is coupled to the heat equation in each subdomain via jump conditions at the interface, enforced through extrapolated values or ghost fluid techniques to resolve discontinuities in temperature and flux. The method's simplicity and robustness for dendritic solidification problems were established in early implementations, achieving accurate interface capturing even for oscillatory or branching morphologies.^[35] Within numerical frameworks, phase-field approximations can be integrated to smooth the interface via a diffuse profile, replacing the sharp boundary with a transition zone governed by a double-well potential and gradient energy terms, which enhances stability by avoiding explicit jump enforcement. This diffuse interface approach, when discretized with FEM or combined with level sets, mitigates oscillations near the interface and improves convergence for multi-dimensional problems.^[36] Higher-order accuracy in these methods is achieved through adaptive mesh refinement, concentrating computational effort near the interface where singularities in gradients occur, leading to O(h^2) convergence rates for linear elements in space and first-order in time. Extended finite element techniques further enrich the approximation space across the interface without remeshing, preserving polynomial continuity while accommodating discontinuities.^[36] Modern developments, particularly from the 2000s, include the Dziuk-Elliott evolving surface finite element method, which parametrizes the interface as a evolving manifold and discretizes surface diffusion and normal velocity laws directly on the moving hypersurface, applicable to Stefan problems with curvature-dependent kinetics. This parametric approach ensures geometric conservation and has been extended to weak formulations handling undercooling and Gibbs-Thomson effects.^[37]^[38]

Recent Advances

Since the 2020s, numerical methods for Stefan problems have incorporated machine learning techniques, such as physics-informed neural networks (PINNs), which approximate solutions by minimizing residuals of the governing equations and boundary conditions, offering flexibility for high-dimensional and nonlinear problems without traditional meshing. These data-driven approaches have demonstrated accuracy comparable to classical methods for benchmark Stefan scenarios, with applications in rapid prototyping of phase-change simulations as of 2023. Additionally, cut finite element methods have emerged for handling complex interface geometries by embedding the domain in a background mesh and using Nitsche's method to enforce interface conditions weakly, improving efficiency for two-phase flows with free boundaries, as shown in implementations up to 2023.^[39]^[40]^[41]

Applications

Engineering and Materials Processing

In engineering and materials processing, the Stefan problem serves as a foundational model for predicting phase change dynamics during solidification, enabling optimization of processes involving heat transfer and interface evolution in controlled environments. This application is particularly vital in manufacturing where precise control of thermal gradients prevents defects like cracks or uneven microstructures. Seminal studies have demonstrated its utility in simplifying complex multiphysics interactions through one-phase approximations, focusing on the solidifying phase while neglecting the liquid for rapid cooling scenarios.^[42] In casting and welding of metals, the Stefan problem facilitates predictions of dendrite growth and porosity formation by modeling the moving solidification front. For instance, cellular automaton-finite difference methods coupled with Stefan formulations simulate solute redistribution and gas bubble entrapment in aluminum alloys, revealing how interdendritic flow influences microporosity levels during directional solidification. In welding processes, one-phase Stefan approximations capture rapid cooling effects, aiding in the forecast of heat-affected zones and residual stresses that contribute to porosity in high-strength steels. These models, validated against experimental microstructures, underscore the role of latent heat release in dictating dendrite arm spacing and overall casting quality.^[43]^[44]^[45] Additive manufacturing, such as laser powder bed fusion of alloys, employs Stefan-based interface tracking to simulate the evolution of melt pools and mushy zones, where partial solidification occurs. Phase-field extensions of the Stefan problem integrate mushy zone dynamics, accounting for dendritic growth and elemental segregation under rapid thermal cycles, which helps mitigate defects like keyhole pores in titanium alloys. Numerical simulations using these approaches have shown that incorporating mushy zone drag forces stabilizes the solidification front, improving part density and mechanical properties in 3D-printed components. Representative studies highlight how Stefan-derived models predict grain refinement in Al-Cu alloys, linking process parameters like scan speed to microstructure uniformity.^[46]^[47]^[48]^[49] In food processing, the Stefan problem models freezing of products like fruits or biological tissues, incorporating inverse formulations to estimate thermal properties from observed phase change data. Enthalpy-based methods solve these problems for irregular geometries, predicting ice front propagation and minimizing cellular damage in cryopreserved foods. For example, exact solutions to finite Stefan problems with variable boundary temperatures have been applied to thawing cycles, optimizing energy use while preserving texture in meat products. Inverse Stefan analyses further enable parameter identification, such as diffusivity, from experimental freezing curves, enhancing process efficiency in industrial freezers.^[50]^[51]^[52] Biomedical applications leverage the Stefan problem in cryosurgery for tumor ablation, where models account for perfusion effects in biological tissues to predict ice ball formation. One-dimensional inverse Stefan solutions incorporate blood flow and metabolic heat, estimating the ablation zone size during needle-based freezing procedures. These formulations, often using heat balance integrals, guide probe placement and cooling rates to achieve complete necrosis while sparing healthy tissue, as validated in prostate cancer treatments. Multiphase extensions briefly reference classical two-phase setups to handle tissue heterogeneity, improving predictive accuracy for perfused organs.^[53]^[54]^[55] A prominent case study is the continuous casting of steel, where Stefan solutions inform real-time process control by modeling shell thickness growth along the caster. Optimal control strategies for the single-phase Stefan problem adjust spray cooling to maintain metallurgical length, preventing breakouts and ensuring uniform solidification. Feedback algorithms using enthalpy methods track the free boundary, demonstrating in simulations that precise heat flux modulation reduces centerline segregation in billet casting. These approaches, rooted in well-posedness results for the Stefan system, have been implemented in industrial controllers to enhance steel quality and throughput.^[56]^[57]^[58]

Geophysical and Climate Modeling

In geophysical modeling, the Stefan problem describes the evolution of phase boundaries in natural ice systems, such as glaciers and sea ice, where heat diffusion drives melting or freezing influenced by environmental forcings. For glaciers, one-dimensional multi-phase Stefan formulations model the freezing of water in crevasses or basal melt beneath ice sheets, capturing the moving interface between solid, liquid, and potentially mushy phases. In sea ice dynamics, the Stefan condition governs thermodynamic growth and decay of ice thickness, balancing net surface fluxes against latent heat release at the ice-ocean interface, often coupled with Navier-Stokes equations to account for fluid flow effects on melt rates in polar regions. This coupling is essential for simulating advective influences on interface evolution, as seen in viscous-plastic rheology models integrated with thermodynamic solvers.^[59]^[60]^[61] Permafrost thawing under global warming is modeled as a two-phase Stefan problem, representing the transition from frozen soil-ice mixtures to thawed, water-saturated ground, with the moving boundary dictated by heat conduction and latent heat absorption. The Darcy-Stefan approach incorporates porous media flow, coupling thermal-hydraulic-mechanical processes to predict the radial expansion of the thawed zone around heat sources like wellbores, leading to ground subsidence as ice volume decreases. Numerical solutions using enthalpy-porosity methods simulate seasonal freeze-thaw cycles, reproducing observed thaw depths with errors below 8% and highlighting accelerated subsidence risks from rising temperatures. These models forecast subsidence in Arctic regions, depending on flow coupling.^[62]^[63] In climate applications, the Stefan problem is parameterized within global climate models (GCMs) to represent sea ice thermodynamics, contributing to Arctic amplification through enhanced polar warming from ice-albedo feedbacks. Sea ice schemes solve the Stefan equations using multi-layer discretizations (typically 1–4 levels) to compute growth and melt, though low resolution can overestimate surface temperature variability by factors of 2–3. Enthalpy-based methods improve efficiency by formulating the problem in a single domain without explicit interface tracking, incorporating salinity-dependent thermodynamics to handle brine pockets and phase changes accurately. This approach, as in the Community Ice CodE (CICE), enables robust simulations of ice thickness distribution under varying forcings, reducing computational costs while capturing essential feedbacks in GCMs like CCSM3.^[64]^[65]^[61] Volcanic lava cooling exhibits Stefan-like fronts in multi-component flows, where solidification propagates from the surface inward, releasing latent heat over a temperature range between liquidus and solidus phases. One-dimensional Stefan models, adapted for density contrasts and thermal contraction, simulate post-emplacement cooling of basaltic flows, predicting subsidence rates that decay exponentially from ~20 mm/year initially to ~2 mm/year after 15 years, as validated by InSAR observations at Hekla volcano. These formulations account for variable thermal properties in multi-phase mixtures, aiding hazard assessments for infrastructure near active flows.^[66] Recent studies in the 2020s emphasize the Stefan problem's role in sea-level rise projections, particularly through basal melt modeling of Antarctic ice sheets under variable climate forcings like ocean warming. Three-phase Stefan approaches quantify mass loss rates for coastal glaciers, linking interface dynamics to global projections of 0.3–1.0 meters rise by 2100, with implications for coastal inundation and permafrost-driven subsidence amplifying local effects. Enthalpy methods in these models enhance projections by efficiently handling irregular boundaries and forcings in ensemble simulations.^[59]^[65]

Extensions and Modern Developments

Multi-Phase and Variable Property Problems

In extensions of the classical Stefan problem, material properties such as thermal diffusivity \alpha(x,u) and conductivity k(u) are allowed to depend on position x and temperature u, leading to nonlinear heat equations of the form u_t = \nabla \cdot (k(u) \nabla u) in each phase and modified Stefan conditions at the interface that incorporate these variations, such as k_s \frac{\partial u_s}{\partial n} - k_l \frac{\partial u_l}{\partial n} = \rho [L](/page/L') v, where the latent heat L may also vary.^[67] These generalizations account for realistic scenarios where properties like specific heat and conductivity change with temperature, as seen in high-temperature melting processes, departing from the constant-property assumptions of the classical one-phase or two-phase models.^[67] Research since the 1970s has established existence and uniqueness for such problems under conditions like p > -1 for power-law dependencies k(u) = u^p.^[67] Multi-phase Stefan problems extend the framework to scenarios involving more than two phases, such as solid-liquid-gas transitions or alloy solidification with eutectic points, where multiple interfaces evolve and interact. In these models, the domain is partitioned into N phases with distinct heat equations in each, coupled by Stefan conditions at each interface, often analyzed in one dimension for self-similar solutions on half-spaces with Dirichlet or Neumann boundary data.^[68] Seminal work from 1977 proved existence, uniqueness, and stability of solutions for the multi-phase problem using variational inequalities and monotonicity arguments, establishing a foundation for handling arbitrary numbers of phases without assuming constant properties.^[69] For three-phase problems, such as melting with vaporization, analytical solutions reveal interface positions scaling as \sqrt{t}, with numerical validations confirming energy conservation across phases.^[70] Coupled flow extensions integrate the Stefan problem with Navier-Stokes equations to model convection-driven phase changes, as in crystal growth or alloy casting, where fluid velocity influences heat transport via advection terms in the energy equation \rho c (u_t + \mathbf{v} \cdot \nabla u) = \nabla \cdot (k \nabla u).^[71] The interface evolution then satisfies a modified Stefan condition incorporating normal velocity components from the flow. Arbitrary Lagrangian-Eulerian (ALE) finite element methods have been developed for such systems, enabling accurate tracking of free boundaries in two dimensions while handling mesh movement and incompressibility.^[72] These approaches demonstrate stability for high Rayleigh numbers, where buoyancy-driven convection significantly alters solidification patterns compared to pure conduction cases.^[73] At the nanoscale, Stefan problems incorporate size-dependent properties and non-local effects, such as kinetic undercooling where the interface temperature deviates from the bulk melting point due to curvature and finite atomic scales, modifying the classical Stefan condition to T_I = T_m (1 - \frac{\gamma \kappa}{L}) - \mu v, with \gamma as the Gibbs-Thomson coefficient and \mu the kinetic undercooling parameter.^[74] Non-local heat transport is modeled via a size-dependent effective thermal conductivity k^* = k (1 + \ell^*/L^*), where \ell^* is the phonon mean free path and L^* the characteristic length, leading to faster initial solidification rates that approach classical limits for large scales (Knudsen number \mathrm{Kn} \ll 1).^[75] Asymptotic analyses reveal three regimes: small-time finite growth rates, intermediate similarity solutions, and large-time linear interface advancement proportional to the Biot number over the Stefan number.^[75] These modifications are crucial for modeling nanoparticle melting, where surface effects dominate and latent heat release scales with particle radius.^[76] Solution approaches for these problems leverage perturbation methods for small property variations, expanding solutions around classical similarity forms to capture nonlinear effects up to first order.^[74] For larger variations, numerical adaptations include finite difference schemes on fixed grids with enthalpy formulations to handle nonlinearity, achieving second-order accuracy in time and space for coupled systems.^[74] Immersed boundary methods with smooth extensions ensure stability in flow-coupled cases, while one-phase reductions simplify computations when solid conductivity is negligible, preserving energy balance.^[73] These techniques prioritize conceptual fidelity over exhaustive benchmarks, with validations showing errors below 1% for Stefan numbers up to 10 in variable-coefficient benchmarks.^[67]

Phase-Field Models and Alternatives

The phase-field approach provides a diffuse-interface alternative to the sharp-interface Stefan problem by introducing an order parameter \phi that varies smoothly across an interfacial width \epsilon, representing the transition between phases such as solid and liquid. This parameter evolves according to the Allen-Cahn equation for non-conserved fields or the Cahn-Hilliard equation for conserved quantities, coupled with the heat conduction equation to account for temperature fields. The dynamics are derived from a free energy functional F[\phi, T] = \int \left( f(\phi, T) + \frac{\epsilon^2}{2} |\nabla \phi|^2 \right) d\mathbf{x}, where f is a double-well potential, leading to the evolution equation \frac{\partial \phi}{\partial t} = -M \frac{\delta F}{\delta \phi} with mobility M. In the limit as \epsilon \to 0, the phase-field model asymptotically recovers the classical Stefan problem, including the Stefan condition at the interface, through matched asymptotic expansions that demonstrate convergence to the sharp-interface dynamics. This equivalence ensures physical accuracy while avoiding explicit interface tracking, allowing the model to naturally handle complex phenomena like interface pinching, merging, or triple junctions without ad hoc geometric reconstructions. The approach originated in the context of phase transitions and was rigorously adapted to solidification problems, enabling simulations on fixed grids. Key advantages of phase-field models include their ability to simulate automatic topology changes in interfaces, which is particularly beneficial for evolving morphologies in multi-dimensional settings, as opposed to front-tracking methods that require remeshing. Applications encompass detailed simulations of dendritic growth during alloy solidification, where solute diffusion and thermal gradients drive branching patterns, and microstructure evolution in materials processing, capturing grain boundary motion and phase coarsening in polycrystalline alloys. For instance, quantitative predictions of dendrite tip velocities and selection mechanisms have been achieved through thin-interface asymptotics that minimize solute trapping effects. Compared to sharp-interface numerical methods, phase-field models incur higher computational costs due to the need to resolve the diffuse layer, but they excel in handling intricate geometries and multi-phase interactions without parametric restrictions on interface curvature. Recent advancements in the 2020s have integrated machine learning techniques, such as physics-informed graph neural networks, to accelerate simulations by orders of magnitude—up to 50 times faster—while preserving accuracy for large-scale microstructure predictions in additive manufacturing and alloy design.^[77] These hybrids leverage data-driven surrogates to approximate the stiff phase-field equations, addressing scalability for practical engineering applications.^[77]

References

[1]
Stefan Problem - an overview | ScienceDirect Topics
The Stefan problem is a moving boundary problem involving temperature distribution during solid-liquid phase changes, where the interface changes over time.
[2]
(PDF) Some historical notes about the Stefan problem - ResearchGate
Problems involving a moving freeze-thaw boundary are known as the Stefan problems (Vuik, 1993) . The traditional models of freeze-thaw processes in porous media ...<|control11|><|separator|>
[3]
[PDF] CHAPTER 2 PROBLEMS WITH EXPLICIT SOLUTIONS - UTK Math
These simplest possible, explicitly solvable Stefan problems form the backbone of our understanding of all phase-change models and serve as the only means of.
[4]
[PDF] Beyond the classical Stefan problem - CORE
In the 19th century, the physicist. Gabriel Lamé (1795-1850) and the mechanical engineer Emile Clapeyron (1799-1864) were the first to mathematically couple the ...
[5]
Stefan's work on solid-liquid phase changes - ScienceDirect.com
28. G. Lamé, E. Clapeyron. Mémoire sur la solidification par refroidissement d'un globe liquide. Ann. Chem. Phys., 47 (1831), pp. 250-256. Google Scholar. 29 ...
[6]
[PDF] Some historical notes on the Stefan problem. 1 Introduction.
Nowadays a large class of problems - containing a free or moving boundary - are called Stefan problems (after J. Stefan, 1835-1893).
[7]
[PDF] THE MATHEMATICS OF DIFFUSION - eng . lbl . gov
Page 1. THE MATHEMATICS. OF DIFFUSION. BY. J. CRANK. BRUNEL UNIVERSITY. UXBRIDGE. SECOND EDITION ... mathematics of diffusion in the next few years. Meanwhile, ...
[8]
A three-phase free boundary problem with melting ice and ...
Jan 10, 2014 · Crank, J. (1956) The Mathematics of Diffusion, Clarendon Press, New York, NY. ... (1990) Solutions for the two-phase Stefan problem with the Gibbs ...
[9]
CLASSICAL SOLUTIONS OF THE STEFAN PROBLEM - Project Euclid
Aug 22, 1979 · In [17], Kamenomostskaja introduced the notion of the weak solution of this problem, and proved its global exist-.
[10]
The singular set in the Stefan problem
Jul 3, 2023 · As mentioned above, the singular set could be (n−1)-dimensional in space for a fixed time t = t0. Hence our result gives the optimal bound for ...
[11]
The importance of Luis Caffarelli's work in the study of fluids. - arXiv
Mar 4, 2025 · The work of Caffarelli and collaborators on the regularity for the Stefan problem is the subject of Section 5. Section 6 deals with results ...
[12]
A method of fundamental solutions for the one-dimensional inverse ...
The numerical results presented show that an accurate approximation to the inverse Stefan problem can be obtained using the MFS with small computational cost.Missing: post- | Show results with:post-
[13]
[PDF] deep learning of free boundary and stefan problems - arXiv
Jun 4, 2020 · For example, in its classical form, the inverse Stefan problem is focused on either inferring unknown thermophysical parameters, or.Missing: post- | Show results with:post-
[14]
[PDF] MATHEMATICAL MODELING OF MELTING AND FREEZING ...
The growing use of Silicon brings us into intimate contact with processes involving freezing of supercooled liquid; alloy processing in space leads us to ...
[15]
[PDF] Diffusion and heat transfer∗
flux is linear in the gradient is known as Fourier's law of heat conduction q = −k. dT dx. Fourier's law of heat conduction. (1) in which q is the heat flux ...
[16]
[PDF] Glacial Hydrology
Some or all of the water entering the glacier refreezes, releasing latent heat and warming the surrounding ice.
[17]
[PDF] Selection of Solidification Pathway in Rapid Solidification Processes
Feb 21, 2023 · Consequently, the rapid release of latent heat increases the local temperature of the interface, resulting in coalescence. Due to different ...
[18]
[PDF] Modeling Ice Dynamics As A Thin-Film Stefan Problem
First posed by Josef Stefan as a model of ocean ice forming in arctic regions, the Stefan problem has since found applications in fields ranging from geology to ...
[19]
Well-posedness for the Classical Stefan Problem and the Zero ...
Nov 19, 2016 · We develop a framework for a unified treatment of well-posedness for the Stefan problem with or without surface tension. In the absence of ...Missing: challenges | Show results with:challenges
[20]
The One-Dimensional Heat Equation
The One-Dimensional Heat Equation. The One-Dimensional Heat Equation. The One ... Chapter 18 - The One-Phase Stefan Problem with Flux-Boundary Specification: Some ...
[21]
[PDF] Lecture notes on the Stefan problem
A fixed point of this transform corresponds to a solution of the complete problem. Remark 1.3. The meaning of conditions (1.1) and (1.5) in the context of two.
[22]
The classical solution of the one-dimensional two-phase stefan ...
The classical solution of the one-dimensional two-phase stefan problem with energy specification. Download PDF. John R. Cannon &; John van der Hoek. 366 ...Missing: formulation | Show results with:formulation
[23]
Exact and Approximate Solutions - ResearchGate
Nov 4, 2025 · There exist very few exact solutions to practical Stefan problems. The key solutions are presented in the first section for both one- and ...
[24]
Existence and Uniqueness for Similarity Solutions of One ... - jstor
solutions are among the few explicit solutions obtainable for Stefan problems ... cylindrical, and spherical symmetries which establishes uniqueness when ...<|control11|><|separator|>
[25]
O. A. Oleinik, “A method of solution of the general Stefan problem ...
O. A. Oleinik, G. Kallianpur, “On free boundary problems arising in probability theory (existence theorems)”, Russian Math. Surveys, 52:1 (1997), 222–223 ...
[26]
Continuity of the Temperature in the Stefan Problem - jstor
An argument given by Caffarelli and Friedman [5; proof of Theorem 9.1] can then be used to deduce that. (6.1) the set of singular points of the free boundary ...
[27]
AMS :: Journal of the American Mathematical Society
Abstract: In this paper we analyze the singular set in the Stefan problem and prove the following results: The singular set has parabolic Hausdorff dimension ...
[28]
[PDF] tr/69 december 1976 numerical solution of stefan problems
S. L. Kamenomostskaja, On the Stefan problem, Mat. Sb. 53 (95) ,. 489 - 514, (1961). 6. D. R. Atthey, A finite difference scheme for melting problems,. J ...
[29]
Accurate solutions of moving boundary problems using the enthalpy ...
Abstract. After highlighting the problems associated with the conventional numerical implementations of Stefan problems using the enthalpy formulation, a simple ...
[30]
(PDF) Unconditionally Stable Fully Explicit Finite Difference Solution ...
Aug 6, 2025 · An unconditionally stable fully explicit finite difference method for solution of conduction dominated phase-change problems is presented.<|control11|><|separator|>
[31]
Extrapolated Crank-Nicolson approximation for a linear Stefan ...
May 21, 2001 · In this paper, we apply finite element Galerkin method to a single-phase linear Stefan problem with a forcing term.
[32]
[PDF] A FINITE DIFFERENCE METHOD FOR A STEFAN PROBLEM by ...
The purpose of this paper is to approximate the solution of a. Stefan problem using an implicit finite difference analog for the heat equation and an explicit ...Missing: implementations Lynch O'Neill
[33]
Implicit and fully discrete approximation of the supercooled Stefan ...
Jun 29, 2022 · We consider two implicit approximation schemes of the one-dimensional supercooled Stefan problem and prove their convergence, even in the presence of finite ...
[34]
(PDF) On an Explicit Method for the Solution of a Stefan Problem
Aug 10, 2025 · Also, we shall show that the finite difference solution we obtain converges uniformly to the actual solution. Numerous numerical schemes for ...
[35]
An ALE finite element method for a coupled Stefan problem and ...
Aug 7, 2025 · An adaptive immersed finite element method with arbitrary lagrangian-eulerian scheme for parabolic equations in time variable domains.
[36]
A Simple Level Set Method for Solving Stefan Problems
A simple level set method for solving Stefan problems is presented. This method can be applied to problems involving dendritic solidification.Missing: instability | Show results with:instability
[37]
Stefan Problem through Extended Finite Elements: Review and ...
In its original formulation, the Stefan problem is a representation of (temperature-driven) first-order phase transitions in matter through a boundary value ...<|control11|><|separator|>
[38]
Finite element methods for surface PDEs* | Acta Numerica
Apr 2, 2013 · In this article we consider finite element methods for approximating the solution of partial differential equations on surfaces.
[39]
A Stefan problem on an evolving surface - PMC - NIH
In this paper, we present the theory of weak solutions associated with the so-called enthalpy approach [1] to the Stefan problem on an evolving curved ...
[40]
[PDF] Modeling of dendritic solidification and numerical analysis of ... - arXiv
Oct 26, 2022 · The theoretical foundations of dendritic solidification lie in the classical Stefan problem, a moving boundary problem, that describes the ...
[41]
Microporosity formation and dendrite growth during solidification of ...
The dendrite growth and solute transport are simulated using a CA-FDM approach. The nucleation, growth, and movement of gas bubbles, as well as the transport of ...
[42]
Interaction of local solidification and remelting during dendrite ...
Dec 19, 2017 · The simulations reveal how local solidification and melting stimulate each other through the complicated interactions between phase transformation, interface ...
[43]
[PDF] Multiscale modeling of solidification: from microstructures to castings
Jan 4, 2021 · In this chapter I will present the most salient developments: we explained the link between microstructure and macrosegregation in steel ingots ...
[44]
Prediction of microstructure in laser powder bed fusion process
Feb 1, 2017 · The classical mathematical formulation is so-called Stefan problem with a free moving boundary. This is the sharp-interface model approach [11] ...
[45]
Mechanism of keyhole pore formation in metal additive manufacturing
Jan 27, 2022 · Additionally, the drag force from the mushy zone is pertinent as it determines the keyhole fluctuation at the bottom of the melt pool.
[46]
[PDF] Multiscale modeling of metal additive manufacturing
The theoretical foundations of dendritic solidification lie in the classical Stefan problem, a moving boundary problem, that describes the evolution of a solid- ...
[47]
Modelling grain refinement under additive manufacturing ...
This study explores grain refinement in Al-Cu alloys with a process-microstructure linking cellular automata-finite difference (CAFD) approach.Modelling Grain Refinement... · 2. Methods · 2.3. Grain Structure...<|control11|><|separator|>
[48]
An exact solution for the finite Stefan problem with temperature ...
Crank. How to deal with moving boundaries in thermal problems. R.W. Lewis ... The Mathematics of Diffusion. Clarendon Press, Oxford (1979). Google Scholar.
[49]
(PDF) An enthalpy-based finite element method for solving two ...
This paper presents numerical solution of one-dimensional two-phase Stefan problem by using finite element method. The governing equations involved in Stefan ...Missing: seminal | Show results with:seminal
[50]
[PDF] Numerical and Experimental Prediction of Food Freezing
STEFAN PROBLEM WITH ENTHALPY TRANSFORMING METHOD. Stefan problems represented heat conduction problems with phase change. This problem is treated by the ...
[51]
[PDF] Dimensional Inverse-Stefan Problem in Nonideal Biological Tissues
A new analytic solution of the inverse-Stefan problem in biological tissues is presented. The solution, which is based on the enthalpy method, assumes that ...
[52]
Combined Solution of the Inverse Stefan Problem for Successive ...
The solution includes the thermal effects of blood perfusion and metabolic heat generation. The analysis combines a heat balance integral solution in the frozen ...Missing: food processing problems
[53]
[PDF] Inverse Stefan Problem for Successive Freezing/Thawing in ...
Moreover, the analyses presented by Rubinsky and Shitzer. (1976) and Rabin and Shitzer (1995) assume constant blood perfusion and metabolic heat generation ...
[54]
Optimal Control of Free Boundary of a Stefan Problem ... - IEEE Xplore
The process is described as a single-phase Stefan problem. The temperature and the shell growth are controlled by the steel surface heat flux generated by the ...
[55]
Control of the Stefan System and Applications: A Tutorial
May 3, 2022 · The so-called Stefan system describes the dynamical model of the liquid–solid phase change in materials ranging from water and ice in the ...<|control11|><|separator|>
[56]
[PDF] Enthalpy-based feedback control algorithms for the Stefan problem ...
After a brief introduction to a continuous casting process, this talk presents derivation of a control law for full-state feedback control of the single-phase ...
[57]
Modelling of Basal Melt of Antarctic Ice Sheet Based on ... - NASA ADS
... glacier's average mass loss rate being 29 Gt ... A numerical model based on the one-dimensional three-phase Stefan problem was developed.Missing: modeling | Show results with:modeling
[58]
Modeling Sea Ice - American Mathematical Society
For 𝐮 = 0, (10)–(11) can be used with appropriate boundary conditions to solve a Stefan problem for a diffu- sively growing ice layer with vertical variation of ...
[59]
[PDF] Numerical modeling of sea ice in the climate system
the ice known as a Stefan condition: q dh dt= ∑ F. (6) where h is the ice thickness and F is the net flux into the top and bottom surfaces, including the ...
[60]
Application of the Darcy-Stefan model to investigate the thawing ...
Jun 25, 2019 · In this paper, the Darcy-Stefan model is coupled with the porous elastic equation, and a THM coupling model for calculating the thawing ...
[61]
Modeling Heat Transfer through Permafrost Soil Subjected to ... - MDPI
The active layer's depth is calculated by solving the heat transfer problem in the soil with a phase change; this problem is called the Stefan problem. The ...
[62]
The Importance of Ice Vertical Resolution for Snowball Climate and ...
Sea ice schemes in GCMs solve the Stefan problem of sea ice growth and melting (Stefan 1891). These schemes are generally based on a simplification of ...
[63]
A one-dimensional enthalpy model of sea ice | Annals of Glaciology
Sep 14, 2017 · We present a 1-D enthalpy-based model of sea ice that allows for quantitative studies of sea ice and its interaction with radiative ...Missing: GCM | Show results with:GCM
[64]
Post‐emplacement cooling and contraction of lava flows: InSAR ...
Jan 26, 2017 · This mathematized problem of the development of a crust due to solid-liquid phase transition is commonly known as Stefan problem. The ...
[65]
An Analysis of the One-Phase Stefan Problem with Variable Thermal ...
May 19, 2023 · In classical Stefan problems, the substance's specific heat and thermal conductivity are both constants. Since the 1970s, researchers have ...
[66]
On self-similar solutions of a multi-phase Stefan problem in the half ...
Mar 16, 2024 · We study self-similar solutions of a multi-phase Stefan problem for a heat equation on the half-line x>0 with a constant initial data and with Dirichlet or ...
[67]
Some results on the multi-phase stefan problem
(1977). Some results on the multi-phase stefan problem. Communications in Partial Differential Equations: Vol. 2, No. 10, pp. 1017-1044.
[68]
Analytical and Numerical Solutions to the Three-Phase Stefan ...
The standard two-phase Stefan problem, in which only the heat equation is used, and density change-induced fluid flow is not involved, remains the gold standard ...
[69]
Stefan problem coupled with natural convection - AIP Publishing
Jun 1, 2023 · This paper investigates the process of dissolution of a solute in a solvent placed in an horizontal concentric cylinder.
[70]
An ALE finite element method for a coupled Stefan problem and ...
Aug 6, 2012 · A system of Navier-Stokes equations with free capillary surface and Stefan problem is used to model the partial melting of a workpiece of ...
[71]
A stable and accurate scheme for solving the Stefan problem coupled with natural convection using the Immersed Boundary Smooth Extension method
### Summary of Numerical Method for Stefan Problem Coupled with Fluid Flow
[72]
[PDF] The Stefan problem with variable thermophysical properties ... - arXiv
Apr 12, 2019 · 6 Numerical solution. There are numerous published methods for the numerical solution of Stefan problems, such as finite difference, finite ...
[73]
Non-local effects and size-dependent properties in Stefan problems ...
Non-local effects and size-dependent properties in Stefan problems ... Asymptotic analysis of the Guyer–Krumhansl–Stefan model for nanoscale solidification.
[74]
Nanoparticle melting as a stefan moving boundary problem - PubMed
The melting of spherical nanoparticles is considered from the perspective of heat flow in a pure material and as a moving boundary (Stefan) problem.