Page 102 -
P. 102

6 Free Boundary Problems and Phase Transitions
         96

                              seminal paper [7], investigating the ice layer formation in a water-ice phase tran-
                              sition.Interestingly enough,Stefancomparedhis data,obtainedbymathemati-
                              cal modeling, to measurements taken in the quest of the search of a north-west
                              passage [11] through the northern polar sea. In his paper [7] Stefan investigated
                              the non-stationary transport of heat in the ice and formulated a free boundary
                              problem, which is now known as the classical Stefan problem and which has
                              given rise to the modern research area of phase transition modeling by free
                              boundary problems. As a basic reference we refer to [10].
                                 Some of the photographs associated to this chapter show icebergs in lakes
                              of Patagonia. The evolution of their water-ice phase transition free boundary is
                              modeled by the 3-dimensional Stefan problem formulated below.
                                                                   d
                                 Therefore, consider a domain G ∈ R (of course d = 1, 2or3for physical
                              reasons but there is no mathematical reason to exclude larger dimensions here),
                              in which the ice-water ensemble is contained. At time t> 0 assume that the
                              domainGisdividedinto2subdomains,G 1 (t)containingthesolidphase(ice)and
                              G 2 (t) containing the liquid phase (water). These subdomains shall be separated
                              by asmoothsurface Γ(t), where the phase transition occurs. Γ(t)isthe free
                              boundary, an unknown of the Stefan problem. Heat transport is modeled by the
                              linear heat equation:
                                          θ t = Δθ + f ,  x ∈ G 1 (t)and x ∈ G 2 (t), t> 0 ,    (6.13)

                              where f is a given function describing external heat sources/sinks. Here we
                              assumed that the local mass density, the heat conductivity and the heat capacity
                              at constant volume are equal and constant 1 in both phases. More realistically,
                              piecewise constants can be used for modeling purposes. The parabolic PDE
                              (6.13) has to be supplemented by an initial condition

                                                        θ(t = 0) = θ 0  in G                    (6.14)

                              and appropriate boundary conditions at the fixed boundary ∂G. Usually, the
                              temperatureisfixedthere


                                                          θ = θ 1  on ∂G                        (6.15)
                              or the heat flux through the boundary is given:

                                                   gradθ .ν = f 1  on ∂G , t> 0 .               (6.16)
                              Here ν denotes the exterior unit normal to ∂G. Also, mixed Neumann–Dirichlet
                              boundary conditions can be prescribed, corresponding to different types of
                              boundary segments.
                                 Disregarding the phase transition the problem (6.13), (6.14), (6.15) or (6.16)
                              is well-posed. Thus, additional conditions are needed to determine the free
                              boundary. The physically intuitive condition says that the temperature at the free
                              boundary is the constant melting temperature θ m of the solid phase. Obviously
   97   98   99   100   101   102   103   104   105   106   107