cond-mat0502551/2iv.tex
1: \documentclass[a4,twoside,12pt]{article}
2: %\documentclass[a4,twoside,12pt]{article}
3: 
4: %\documentstyle[a4,twoside,12pt]{article}
5: %\documentstyle[aps,preprint,tighten]{revtex}
6: 
7: \usepackage{color}
8: \usepackage{epsfig}
9: 
10: \renewcommand{\evensidemargin}{-0.3cm}
11: \renewcommand{\oddsidemargin}{-0.3cm}
12: \renewcommand{\topmargin}{-0.5cm}
13: \renewcommand{\textwidth}{16.5cm}
14: %\renewcommand{\textheight}{55\baselineskip}
15: \renewcommand{\baselinestretch}{1.5}
16: 
17: % \newcommand{\myred}[1]{{\color{red}#1}}
18: % \newcommand{\myblue}[1]{{\color{blue}#1}}
19: % \newcommand{\mygreen}[1]{{\color{green}#1}}
20: 
21: \baselineskip18pt
22: 
23: \begin{document}
24: 
25: \title{
26: Ivantsov parabolic solution for two combined moving interfaces}
27: 
28: \medskip
29: 
30: \author{D. Temkin \\[0.5cm]
31: Institut f\"ur Festk\"orperforschung, Forschungszentrum J\"ulich, \\
32: D-52425 J\"ulich, Germany \\
33: Contact e-mail: d.temkin@gmx.de}
34: 
35: \maketitle
36: 
37: \begin{abstract}
38: We demonstrate that for a migration of a
39: liquid layer between the melting and the solidification front
40: an exact steady-state solution with two parabolic fronts 
41: can be found. A necessary condition hereby is that the temperature
42: of the solidification front exceeds the temperature of the 
43: melting front (both temperatures are supposed to be constant). 
44: It is shown that in pure materials and alloys there exist
45: two types of solutions with two convex and with two concave
46: parabolas respectively. 
47: While a steady-state process with two planar interfaces is only
48: possible for a single point, the processes with two parabolas
49: are possible inside a region of control parameters. 
50: The relations between the Peclet numbers and the control
51: parameters are obtained.
52: \end{abstract}
53: 
54: \newpage
55: 
56: \section{Introduction} 
57: Liquid film migration (LFM) is well-known phenomenon 
58: which has been observed in many alloy systems during 
59: sintering in the presence of liquid phase \cite{yoon1,yoon2} 
60: and in Cu-In solid solutions during melting started at grain 
61: boundaries \cite{musch1}. In LFM, one of crystals is melted 
62: and the other one is solidified. The both solid-liquid interfaces 
63: are moving together with the same velocity. In the investigated 
64: alloys systems the migration velocity is of the order of $10^{-8}-10^{-7}$ 
65: $m\cdot s^{-1}$ and it is controlled  by the solute diffusion through 
66: a thin liquid layer between the two interfaces \cite{4}. The migration 
67: velocity is much smaller than the characteristic velocity of 
68: atomic kinetics at the interfaces. Therefore the both solids at the 
69: interfaces must be at the local thermodynamic equilibrium with 
70: liquid. The theory \cite{4} answers the question about 
71: the different equilibrium states at the melting and solidification fronts. 
72: In a steady-state regime, the difference in equilibrium relates to coherency 
73: stresses appearing only at the melting front due to the sharp profile of 
74: composition ahead the moving melting front. Thus, the liquid composition
75: at the melting front which depends on the coherency strains 
76: and a curvature of the front differs from the liquid composition 
77: at unstressed and curved solidification front. The migration velocity is 
78: proportional to the difference of these compositions divided by the 
79: film thickness \cite{4}. But what controls the thickness? 
80: 
81: Consequently, a new problem of two combined moving solid-liquid 
82: interfaces with a liquid between them appears. In the present article, 
83: this problem is considered under simplified boundary conditions: 
84: the temperature and chemical composition along each interface 
85: are constant. These constants are different for the melting and 
86: solidification fronts and differ from those far from the migrating 
87: liquid film. It means that any capillary, kinetic and crystallographic 
88: effects at the interfaces are neglected. It is found that under these simplified 
89: boundary conditions two co-focal parabolic fronts can move together 
90: with the same velocity. The situation is rather similar to a steady-state  motion 
91: of one parabolic solidification front into a supercooled melt \cite{1,2} 
92: or one parabolic melting front into a superheated solid. 
93: 
94: \section{Solutions for one parabolic front}
95: Needle-like stationary solutions were first obtained by Ivantsov 
96: for the crystallization of a pure material from a supercooled
97: melt \cite{1} and were extended to binary alloys \cite{2}. 
98: This solutions describes a parabolic interface of a solid phase
99: at constant temperature $T_m$ which extends into a 
100: supercooled melt. Inside the melt the temperature drops 
101: and reaches far from the interface its asymptotic value $T_{0} < T_m$. 
102: 
103: For the two-dimensional case the Ivantsov relation is
104: \begin{equation}
105:   \Delta = F(P) \equiv \sqrt{\pi P} e^P {\rm erfc} (\sqrt{P}) 
106:   \quad , \quad 0 \le \Delta < 1.
107:   \label{ivantsov1}
108: \end{equation}
109: It connects the Peclet number $P = VR/2D$ and the supercooling
110: $\Delta = (T_m-T_{0}) c/q$. Here $V$ and $R$ are the front-velocity
111: and the tip-radius of the parabola respectively; $D$ and $c$ are
112: the thermal diffusivity and the specific heat of the melt, respectively, $q$ is
113: the latent heat. 
114: 
115: Eq. (\ref{ivantsov1}) can easily be obtained in 2D-parabolic
116: coordinates $(\eta,\xi)$ (see, for example, Ref. \cite{3}) 
117: \begin{eqnarray}
118:   \eta &=& \sqrt{x^2+z'^2} + z' \quad , \quad \xi =
119:   \sqrt{x^2+z'^2} - z' \quad, \quad ß0 \leq (\eta, \xi) 
120:   < \infty, \nonumber \\ x &=& \pm \sqrt{\eta \xi} \quad , 
121:   \quad z' = \frac{1}{2} \ (\eta - \xi) \quad , 
122:   \quad - \infty < (x,z') < \infty,
123: \end{eqnarray}
124: where $(x,z'=z-Vt,t)$ are the Cartesian coordinates and the $t$ is the time.
125: In this description the parabolic solidification front
126: ($z'\!=\!\frac{R}{2}\,-\,\frac{x^{2}}{2R}$), 
127: moving with velocity $V$ in the positive $z$-direction,
128: has the coordinate $\eta = R$. The regions $0 \leq \eta < R$ and 
129: $R < \eta < \infty$ correspond to the solid and the liquid phase
130: respectively.
131: 
132: In this set of coordinates, the temperature field $T$ in both phases
133: depends only on $\eta$. The thermal diffusion equation 
134: \begin{equation}
135:   \frac{\partial}{\partial \eta} \ \left(  \sqrt{\eta} \ \frac{\partial T}
136:   {\partial \eta}\right) + \frac{V}{2D} \left( \sqrt{\eta} \ 
137:   \frac{\partial T}{\partial \eta} \right) \ = \ 0
138:   \label{diffusion1}
139: \end{equation}
140: is easily solved by the ansatz
141: \begin{equation}
142: T (\eta) = A+B \int \eta^{-1/2} e^{-V\eta/2D} d\eta
143: \end{equation}
144: with different constants $A$ and $B$ for the solid and the
145: liquid phase. These constants and Eq. (\ref{ivantsov1}) can
146: be obtained by using the appropriate boundary conditions,
147: namely the asymptotic boundary condition
148: $T (\eta \to \infty) = T_{0}$ and the interfacial conditions
149: of temperature continuity $T(R-0)=T(R+0)=T_m$, and 
150: heat balance at the interface
151: \begin{equation}
152:   V q/c = - 2D T' (R+0) + 2DT'(R-0),
153: \end{equation}
154: where the prime denotes the derivative with respect to 
155: $\eta$. 
156: For reasons of simplicity we assumed the same values for
157: $D$ and $c$ in both phases.
158: 
159: Apart from the solution with convex parabolic front there
160: exists as well a solution with a concave front 
161: ($z'\!=\!-\frac{R}{2}+\frac{x^2}{2R}$).
162: In the parabolic coordinates, the interface is defined by the  
163: relation $\xi = R$ while solid and liquid phase are at 
164: $\infty < \xi < R$ and  $R > \xi \ge 0 $, respectively.
165: In contrast to the former case the temperature 
166: field $T$ depends only on $\xi$. Instead of 
167: Eq. (\ref{diffusion1}) the diffusion equation therefore
168: reduces to
169: \begin{equation}
170:   \frac{\partial}{\partial\xi} \ \left( \sqrt{\xi} \ 
171:   \frac{\partial T}{\partial \xi} \right) \ - \
172:   \frac{V}{2D} \left( \sqrt{\xi} \ \frac{\partial T}
173:   {\partial \xi} \right) \ = \ 0 \qquad ,
174:   \label{diffusion2} 
175: \end{equation}
176: and is solved by the ansatz
177: \begin{equation}
178:   T(\xi) \ = \ A + B \, \int \xi^{-1/2} e^{V\xi/2D} d\xi \qquad.
179: \end{equation}
180: While the conditions of continuity and heat balance at the interface
181: stay the same as for the convex parabola,  
182: the asymptotic condition changes to $T (\xi \to 0) = T_{0}$. 
183: Thus for the concave front instead of Eq. (\ref{ivantsov1})
184: we obtain the relation
185: \begin{equation}
186:   \Delta = \Phi (P) \equiv 2 \sqrt P \ e^{-P} \ 
187:   \int^{\sqrt P}_0 e^{x^2} dx \qquad \quad 0 \leq \Delta \leq 1.284... \qquad , 
188: \label{ivantsov2} 
189: \end{equation}
190: between the supercooling $\Delta$ and the Peclet number $P$.
191: The function $\Phi (P)$, in contrast to the function $F (P)$ 
192: in Eq. (\ref{ivantsov1}), is not monotonous: it has a maximum 
193: % $F_{max}\approx 
194: $1.284...$ at $P = 2.25...$ and
195: approaches 1 from above for $P\to\infty$.
196: 
197: Naturally, Eq. (\ref{ivantsov1}) and Eq. (\ref{ivantsov2}) describe 
198: as well a parabolic melting front: the superheated solid
199: phase with $T_{0} > T_m$ is outside the convex parabola 
200: ($\eta = R$) at $R<\eta<\infty$ or inside the concave one
201: ($\xi=R$) at $R>\xi\geq 0$. The melting heat is $-q$. 
202: 
203: \section{Two parabolic fronts in a pure material}
204: It turns out, that an exact steady-state solution with a combined
205: motion of the two co-focal parabolic fronts 
206: (melting and solidification front with a liquid layer in between)
207: can be found. For this the solidification front temperature 
208: $T_m$ has to exceed the temperature $\tilde T_m$ on the melting front. 
209: %
210: This may be the case for pure materials where the melted solid
211: phase (S1) has with respect to the solidified solid (S2) an 
212: additional contribution $\epsilon$ to the energy-density 
213: (for example due to preliminary mechanical treatment). 
214: In this case the melting heat equal to
215: $- (q - \epsilon)$ and the melting temperature 
216: $\tilde T_m = T_m (1-\epsilon/q)$  differ from the values 
217: $q$ and $T_m$ for the solidification front. 
218: The additional energy $\epsilon$ is the driving force for a
219: recrystallization process, which takes place in the 
220: solid either due to a migration of a boundary between 
221: a S1- and a S2-grain or due to a possible recrystallization 
222: through a liquid layer between S1 and S2. 
223: 
224: It should be pointed out, that a combined stationary motion
225: of two planar interfaces (with equilibrium temperatures $T_m$ and
226: $\tilde T_m$) is only possible at the temperature $T_{0,0}$
227: at which the temperature increase $(T_m - T_{0,0})$ exactly
228: compensates the additional energy, $c(T_m - T_{0,0})= \epsilon$. 
229: All temperatures deviating from $T_{0,0}$ lead to an increase 
230: (at $T_{0} > T_{0,0}$) or to a decrease and disappearance 
231: (at $T_{0} < T_{0,0}$) of the liquid layer between the planar
232: interfaces. Here only two co-focal parabolic interfaces can lead to a
233: steady-state motion. 
234: 
235: Similar to the considerations described for one interface
236: an analysis of the temperature field in parabolic coordinates
237: leads to the following relations:\\
238: %
239: For two convex parabolas with $P_1 > P_2$ (Fig. 1):
240: %
241: % 
242: \begin{eqnarray}
243:   \Delta_m &=& 2 \sqrt{P_2} e^{P_2}
244:   \ \int^{\sqrt{P_1}}_{\sqrt{P_2}} e^{-x^2} \ dx \label{ivantsovneu1a}\\
245:   \Delta~~ &=& \left[ 1-\alpha \Delta_m - \sqrt{\frac{P_2}{P_1}} \ 
246:   e^{P_2 - P_1} \right] \ F(P_1). \label{ivantsovneu1b}
247: \end{eqnarray}
248: For two concave parabolas with $P_1 < P_2$ (Fig. 1):
249: \begin{eqnarray}
250:   \Delta_m &=& 2 \sqrt{P_2} e^{-P_2} \ 
251:   \int^{\sqrt{P_2}}_{\sqrt{P_1}} e^{x^2} \ dx, \label{ivantsovneu2a}\\
252:   \Delta~~ &=& \left[ 1-\alpha \Delta_m - \sqrt{\frac{P_2}{P_1}} \ 
253:     e^{P_1 - P_2} \right] \ \Phi (P_1), \label{ivantsovneu2b}
254: \end{eqnarray}
255: with $P_i = V R_i/2D$, $\Delta_m = (T_m - \tilde T_m) c/q$,
256: $\Delta = (T_{0} - \tilde T_m) c/q$ and $F(P)$ and $\Phi (P)$
257: as in Eq. (\ref{ivantsov1}) and Eq.(\ref{ivantsov2}). 
258: $\alpha = q/c T_m$ is a material parameter (e.g., for Ni one has 
259: $\alpha \cong 0,25$).
260: % 
261: The term $(1-\alpha \Delta_m)$ arises due to the fact that the 
262: melting heat of a solid phase S1 with an additional 
263: energy density $\epsilon$ differs from the equilibrium 
264: melting heat, $-q$ 
265: (note that $-(q-\epsilon) = -q\,(1-\alpha \Delta_m)$). 
266: 
267: In the corresponding limiting cases Eqs. 
268: (\ref{ivantsovneu1a})-(\ref{ivantsovneu2b}) reduce to 
269: Eqs. (\ref{ivantsov1}) and (\ref{ivantsov2}):
270: Eq. (\ref{ivantsovneu1a}) for $P_1\to\infty$ and  Eq. (\ref{ivantsovneu2a}) 
271: for  $P_1 \to 0$
272:  lead to the solidification relations  Eq. (\ref{ivantsov1}) and
273: Eq. (\ref{ivantsov2}), respectively, with supercooling $\Delta_m$ instead of $\Delta$.
274: Eq. (\ref{ivantsovneu1b}) for $P_2\to 0$ and 
275: Eq. (\ref{ivantsovneu2b}) for $P_2\to\infty$
276: lead to the melting relations  Eq. (\ref{ivantsov1}) and
277: Eq. (\ref{ivantsov2})
278: with the normalized superheating 
279:  $\Delta/(1-\alpha \Delta_m)$ instead of $\Delta$.
280: 
281: \section{Analysis of convex and concave solutions}
282: Even if Eqs. (\ref{ivantsovneu1a}) - (\ref{ivantsovneu2b}) 
283: are valid for arbitrary values of the normalized 
284: driving force $\Delta_m > 0$, we consider in the
285: following analysis the case of small
286: driving forces $\Delta_m \ll 1$. With this assumption 
287: Eqs. (\ref{ivantsovneu1a}) and (\ref{ivantsovneu2a}), respectively, simplify to 
288: \begin{eqnarray}
289:   \sqrt{P_2} &\cong& \frac{1}{2} \ \left[ \sqrt{P_1} \pm 
290:   \sqrt{P_1 - 2\Delta_m} \right] \quad \quad (P_1 > P_2)~
291:   \label{ivantsov3}\\
292:   \sqrt{P_2} &\cong& \frac{1}{2} \ \left[ \sqrt{P_1} + 
293:   \sqrt{P_1 + 2\Delta_m} \right] \quad  \quad (P_1 < P_2).
294:   \label{ivantsov4}
295: \end{eqnarray}
296: Eq. (\ref{ivantsov3}) together with Eq. (\ref{ivantsovneu1b}) and 
297: Eq. (\ref{ivantsov4}) together with Eq. (\ref{ivantsovneu2b})
298: define for the case of small driving forces $\Delta_m \ll 1$
299: the dependency of $\Delta (P_1)$ for both types of solutions 
300: (concave and convex parabolas) (see Fig. 2).
301: %
302: % 
303: 
304: For  $P_1 \gg \Delta_m$ 
305:  the upper branch of Eq. (\ref{ivantsov3}) gives  
306: \begin{equation}
307:   P_2 \cong P_1 - \Delta_m \qquad   (P_1 > P_2)
308: \end{equation}
309: and from Eq. (\ref{ivantsov4})
310: \begin{equation}
311:   P_2 \cong P_1 + \Delta_m \qquad (P_1 < P_2).  
312: \end{equation}
313: Using this simplification and taking the limit $\Delta_m \to 0$ we 
314: obtain from Eqs. (\ref{ivantsovneu1b}) and  (\ref{ivantsovneu2b})
315: the simplified expressions
316: \begin{eqnarray}
317:   \frac{\Delta_{~~}}{\Delta_m} &=& \left( 1-\alpha \, + \, 
318:   \frac{1}{2P_1} \right) \ F(P_1) \quad , \quad 
319:   (P_1 > P_2) \qquad , \label{exequation20} \\
320:   \frac{\Delta_{~~}}{\Delta_m} &=& \left( 1-\alpha \, - \, 
321:   \frac{1}{2P_1} \right) \ \Phi (P_1) \quad , \quad
322:   (P_1 < P_2) \qquad . \label{exequation21}
323: \end{eqnarray}
324: 
325: A stationary solution with two convex parabolas $(P_1 > P_2)$ exists for
326: \begin{equation}
327:   \Delta_{0} \leq \Delta \leq \Delta^*  \label{regiondef1}
328: \end{equation}
329: with 
330: \begin{equation}
331:   \Delta_{0} \equiv (1-\alpha) \, \Delta_m \qquad\mbox{and}\qquad
332:   \Delta^* \equiv 1 - \alpha \Delta_m  .
333: \end{equation}
334: %
335: The point $\Delta = \Delta_{0}$ corresponds to the 
336: steady-state solution of two moving planar interfaces mentioned above.
337: 
338: If $\Delta_{0} \leq \Delta \ll \Delta_1$, the dependence of 
339: $\Delta$ on $P_1$ is defined by Eq. (\ref{exequation20}) 
340: and $P_2 \cong P_1 - \Delta_m$. At the point 
341: $\Delta = \Delta_1 \cong \sqrt{\pi \Delta_{m}/2}$ 
342: one has $P_1 \cong 2 \Delta_m$ and $P_2 \cong \Delta_m/2$. 
343: In the vicinity of the $\Delta_1$-point one can obtain 
344: \begin{equation}
345:   \frac{\Delta-\Delta_1}{\Delta_1} \ \cong \ \mp
346:   \sqrt{\frac{P_1}{2\Delta_m} - 1}. 
347: \end{equation}
348: 
349: From Eqs. (\ref{ivantsovneu1a}) and (\ref{ivantsovneu1b}), one can 
350: obtain that close to the limiting point $\Delta^*$ one gets 
351: \begin{equation}
352:   P_1 \ \cong \ \frac{\Delta^*}{2(\Delta^*-\Delta)} \ \gg \ 1 \quad 
353:   and \quad P_2 \ \cong \ \frac{\Delta_{m}^2}{\pi} \ll 1.
354: \end{equation}
355: %
356: These relations describe the independent motion of both parabolic 
357: fronts propagating with the same velocity.
358: 
359: Solutions of Eqs. (\ref{ivantsovneu2b}) - (\ref{ivantsov4})
360:  for a pair of concave parabolas exist in the region 
361: \begin{equation}
362:   -\mid\Delta_4 \mid\ \leq\ \Delta\ \leq\ \Delta_2 \label{regiondef2}
363: \end{equation}
364: between two extreme points on the $\Delta (P_1)$-dependence
365: (Fig. 2): minimum point $\mid \Delta_4 \mid \ \cong \ \Delta_m$ 
366: at $P_1 \cong \sqrt{\Delta_m/4 (1-\alpha)}$ and maximum point
367:  $\Delta_2$ which is close to $\Delta_{0}$. In the limit $P_1 \to
368:  \infty$ the value of $\Delta$ tends to the same point $\Delta_{0}$
369:  as for the convex parabolas, $\Delta \cong \Delta_{0} -\alpha/2P_1$.
370: 
371: \section{Parabolic solutions for binary alloys}
372: Extending our analysis to a two-component alloy, we have to take into
373: account that in contrast to a pure material the temperatures 
374: $T_1$ and $T_2$ of the melting and solidification fronts are 
375: unknown (in a pure material $T_1 = \tilde T_m$, $T_2 = T_m$). 
376: In order to define $T_1$ and $T_2$ (and  consequently 
377: $\Delta_m = (T_2 - T_1) c/q$, and $\Delta = (T_{0} - T_1) c/q$ 
378: in Eqs. (\ref{ivantsovneu1a})-(\ref{ivantsovneu2b})) we have
379: to obtain two additional equations.
380: 
381: In our description, we denote the molar fraction of the
382: second component with $C$ and the diffusion 
383: coefficients in the solid and the liquid phase with 
384: $D_S$ and $D_L$ respectively. While for convex parabolas
385: the concentration fields, $C(\eta,\xi)$, only depend on $\eta$,
386: they depend only on $\xi$ in the concave case. 
387: In the first case, the field $C(\eta)$ satisfies the
388: equilibrium boundary conditions at both interfaces, 
389: namely the continuity condition
390: \begin{equation}
391:   C(R_{2}-0) = C_S (T_2) \ , \ 
392:   C(R_{2}+0) = C_L (T_2) \ , \
393:   C(R_{1}-0) = \tilde C_L (T_1) \ , \
394:   C(R_{1}+0) = \tilde C_S (T_1) \quad , \label{alloycond1}
395: \end{equation}
396: the far field condition $C(\eta\!\to\!\infty) = C_{0}$
397: and the conservation conditions at the interfaces
398: \begin{eqnarray}
399:   V \left[ C_L(T_2) - C_S(T_2)\right] &=& -2D_L C' (R_2+0) +
400:   2D_S C' (R_2-0) \quad ,  \label{alloycond2}\\
401:   V \left[ \tilde C_L(T_1) - \tilde C_S(T_1)\right] &=&
402:   -2D_L C' (R_1-0) + 2D_S C' (R_1+0) \quad . \label{alloycond3}
403: \end{eqnarray}
404: 
405: Here $C_S (T_2)$ and $C_L (T_2)$ are the solidus and liquidus
406: compositions which are defined by the equilibrium phase 
407: diagram of the alloy, while $\tilde C_S(T_1)$,
408: $\tilde C_L (T_1)$ are the corresponding values defined 
409: by a disturbed phase diagram. The additional
410: energy density $\epsilon$, changing the melting point of the melted
411: solid phase S1, will change the equilibrium compositions at 
412: the melting interface as well. In this case, the composition 
413: differences $\tilde C_S (T_1) - C_S (T_1)$, $\tilde C_L (T_1)
414: - C_L (T_1)$, are proportional to $\epsilon/q$. Another reason
415: for the distortion of the phase diagram are coherency stresses
416: due to compositional inhomogeneities in front of the melting 
417: interface \cite{4}.
418: 
419: By solving the diffusion equation for $C(\eta)$ and applying the
420: appropriate boundary conditions (\ref{alloycond1})-(\ref{alloycond3}),
421: we obtain for the case of convex parabolas the solution:
422: \begin{eqnarray}
423:   \frac{C_L (T_2) - \tilde C_L (T_1)}{C_L (T_2) - C_S (T_2)} \!&\!=\!&\!
424:   2\sqrt{P_{2L}} e^{P_{2L}} \ \int^{\sqrt{P_{1L}}}_{\sqrt{P_{2L}}}
425:   e^{-x^2} dx \quad , \quad P_{1L} > P_{2L} \label{alloysol1a}\\
426:   \frac{C_{0}-\tilde C_{s}(T_1)}{C_L(T_2)-C_{s}(T_{2})}\!&\!=\!&\!
427:   \left[
428:      \frac{\tilde C_L (T_1) - \tilde C_s (T_1)}{C_L (T_2) - C_s (T_2)}
429:      -\sqrt{\frac{P_{2L}}{P_{1L}}}\,e^{P_{2L} - P_{1L}} 
430:   \right] \ F(P_{1S}) \label{alloysol1b}
431: \end{eqnarray}
432: ($P_{iL} = DP_i / D_L$, $P_{1S} = D_L P_{1L} / D_S$; $F(P)$ 
433: is the same function as in Eq. (\ref{ivantsov1})). A similar
434: consideration for the case of concave parabolas with 
435: concentration field $C(\xi)$ leads to:
436: \begin{eqnarray}
437:   \frac{ C_L (T_2) - \tilde C_L (T_1)}{C_L (T_2) - C_s (T_2)} \!&\!=\!&\!
438:   2 \sqrt{P_{2L}} \ e^{-P_{2L}} \ \int^{\sqrt{P_{2L}}}_{\sqrt{P_{1L}}} 
439:   e^{x^2} dx \quad , \quad P_{1L} < P_{2L} \label{alloysol2a}\\
440:   \frac{C_{0}-\tilde C_{s}(T_1)}{C_L(T_2)-C_{s}(T_{2})}\!&\!=\!&\!
441:   \left[ \frac{\tilde C_L (T_1) - \tilde C_s (T_1)}{C_L (T_2) - C_s (T_2)}
442:      -\sqrt{\frac{P_{2L}}{P_{1L}}}\,e^{P_{1L} - P_{2L}} 
443:   \right] \ \Phi(P_{1S}), \label{alloysol2b}
444: \end{eqnarray}
445: where $\Phi(P)$ is defined by the same manner as 
446: in Eq. (\ref{ivantsov2}). 
447: 
448: % Usually the diffusion coefficients fulfill the relation 
449: % $D_s \gg D_L \gg D$, while the diffusion in the solid phase
450: % \myred {can be neglected completely}. As main difference 
451: % it becomes possible for a small but non-vanishing value of
452: % $D_s$ to replace analytically the composition \myred{concentration ??} 
453: % $\tilde C_S (T_1)$ in Eq. (\ref{alloycond3}) with the initial
454: % composition $C_{0}$ (see left hand side of Eqs. (\ref{alloysol1b})
455: % and (\ref{alloysol2b})). 
456: % An additional effect of $D_S$ (present in the right hand
457: % side of Eqs. (\ref{alloysol1b}) and (\ref{alloysol2b})) is small
458: % \myred{vanishes ??} if $P_{1L} \ll D_S/D_L$ and terms with
459: % $[1/F(P_{1S}) - 1]$ and $[-1/\Phi (P_{1S})-1 ]$, which are
460: % proportional to $1/P_{1S}$, can be omitted.
461: 
462: 
463: For the convex configuration, Eqs. 
464: (\ref{ivantsovneu1a})-(\ref{ivantsovneu1b}) together with Eqs. 
465: (\ref{alloysol1a})-(\ref{alloysol1b}) describe the dependency
466: of the quantities $T_1$, $T_2$, $P_1$, $P_2$ on the initial
467: control parameters $T_{0}$, $C_{0}$, and $\epsilon$.
468: %($\Delta_m\!\equiv\!(T_2-T_1)c/a$ and $\Delta\!\equiv\!(T_{0}- T_1)/c/q$).
469: The corresponding relations for the concave configuration are given by 
470: Eqs. (\ref{ivantsovneu2a})-(\ref{ivantsovneu2b}) and 
471: Eqs. (\ref{alloysol2a})-(\ref{alloysol2b}).
472: In systems with vanishing $\epsilon$ the coherency strain effect
473: can support the combined motion of two interfaces as, for example,
474: in liquid film migration \cite{4}.
475: 
476: \section{Discussion} 
477: 
478: The considered two fronts process (TFP) is an alternative to an one front process (OFP).
479: TFP is possible at some conditions (for example, at $\tilde T_m < T_m$ for a pure 
480: material) and do exist as a steady-state process in a definite region 
481: of control parameters.
482:  In this region there is one or a few OFP. For  pure materials the TFP exists at 
483: $-|\Delta_4|\le \Delta\le \Delta^{\ast}$ (see Eqs. (19) and (23)), i.e. at 
484: $\tilde T_m-(T_m-\tilde T_m)\le T_0\le \tilde T_m+(q-\epsilon)/c$.
485: When the initial temperature is $T_0<\tilde T_m<T_m$, the OFP of the transition 
486: $S1\rightarrow S2$ proceeds at the grain boundary $S1/S2$. 
487: When $\tilde T_m<T_0<T_m$, the OFP of the melting of $S1$ at the interface $S1/L$ 
488: is also possible. At $T_0>T_m$ the grain $S2$ can also be melted at the interface 
489: $S2/L$. The TFP can proceed faster than the corresponding OFP due to a heat transfer 
490: between melting and solidification fronts through a thin liquid layer. 
491: In order to initiate the TFP (especially at 
492: $ \tilde T_m-(T_m-\tilde T_m)\le T_0\le \tilde T_m$) a liquid phase must be created 
493: inside the system. Then the grain boundary $S1/S2$ splits into two interfaces, 
494: $S1/L$ and $S2/L$, and the TFP proceeds as a self-sustained process.
495: 
496: It should be noted that the TFP can take place in a pure material in which 
497: there are several polymorphic modifications. Then a low temperature modification plays 
498: a role of the melted grain $S1$ in Fig. 1 and its melting temperature $\tilde T_m$ 
499: is lower than the melting temperature $T_m$ for a high temperature modification, $S2$.
500: In such a case the solutions given by Eqs. (\ref{ivantsovneu1a}) - 
501: (\ref{ivantsovneu2b}) are valid for $\Delta_m=(T_m-\tilde T_m)c/q$ as a material 
502: parameter and $(1-\alpha\Delta_m)$  is replaced by another material parameter 
503: $\tilde q/q$, where $\tilde q$ and $q$ are the melting heats of the low and high 
504: temperature phases respectively. 
505: 
506: Generally, one can consider the TFP of the transition of a phase $S1$ into another 
507: phase $S2$ through an intermediate phase $L$ (which is not necessarily a liquid one) 
508: as an alternative to the OFP of the direct transition $S1\rightarrow S2$.      
509: 
510: The obtained Eqs. (\ref{ivantsovneu1a}) - (\ref{ivantsovneu2b}) 
511: together with Eqs. (\ref{alloysol1a}) - (\ref{alloysol2b}), 
512: which define relations between the Peclet numbers $P_1$, $P_2$ 
513: and control parameters $T_0$, $C_0$, give the continuous spectrum 
514: of solutions with the only free parameter (e.g., the velocity $V$). 
515: Therefore an additional equations, i.e. ``selection'' relationship, 
516: is needed in order to define the unique solution. This is similar 
517: to a well-known ``selection'' problem in dendritic growth \cite{5}. 
518: A search for the selection condition will be a subject of future 
519: investigation. The only point which might be stressed here 
520: is the following. 
521: 
522: The structure of the fronts is usually more complicated compared to pure parabolic 
523: ones due to possible cellular structures or due to finite size of the sample. 
524: In this case,
525: two parabolas describe only a part of the moving fronts. 
526: The complete shape of the fronts must be subjected to 
527: additional boundary conditions.  One can speculate on  
528:  possible cellular structures with two moving 
529: interfaces which are shown in Fig. 3. These structures should appear due to the 
530: diffusional interaction between different cells and capillary effects play also 
531: crucial role. 
532: The structures have 
533: central parabolic parts which are convex (Fig. 3(a,b)) 
534: or concave (Fig. 3(c)) and satisfy boundary conditions 
535: of zero heat- and mass-fluxes across boundaries of cells. 
536: The first structure (Fig. 3(a)) can be related to a melting processes
537: and the other two ones, for example, to sintering of two 
538: solid grains $S1$ and $S2$ in liquid phase $L$. 
539: The structure in Fig. 3(b) may correspond to a sintering process 
540: with a supersaturated solid $S1$, while
541: the structure in Fig. 3(c) with a concave central part 
542: may correspond to the case of undersaturated solid $S1$.
543: 
544: In the process in the channel (or cell) the fronts velocity may be controlled 
545: either by the above mentioned cell boundary conditions or by the ``selection''
546: which takes place mostly in the parabolic region. Such a problem 
547: arises also in the selection of the growth velocity of the 
548: classical dendrite in the channel (the structure with one front) \cite{5}. 
549: 
550: In this paper the solutions are obtained for two parabolic fronts 
551: in the two dimensions. Similar solutions can be easily found 
552: for two paraboloidal interfaces in the three dimensions. 
553:  
554: A few words about the solutions (\ref{ivantsov1}) and (\ref{ivantsov2}) 
555: for one convex and one concave parabolic interface might be outlined. 
556: The convex parabola and Eq. (\ref{ivantsov1}) are playing an important role 
557: in dendritic solidification (see, e.g., Ref. \cite{5}) and must be important 
558: in ``dendritic'' melting. It can be supposed that concave parabola 
559: and Eq. (\ref{ivantsov2}) are playing a role in such ``doublon'' 
560: structures with two convex parts and a concave part in between 
561: which are similar to the profile of $S2/L$-interface in Fig. 3(c). 
562: Whether or not those ``doublon'' structures exist in solidification 
563: or melting processes with one front? 
564: 
565: \section{Conclusions} 
566: A theoretical model for combined motion of two solid-liquid fronts with 
567: a liquid layer in between has been developed for pure materials 
568: and binary alloys. The temperature and chemical compositions 
569: at the interfaces were supposed to be constants but different for 
570: fronts of melting and solidification. 
571: 
572: An exact solution which describes the steady-state motion of two 
573: co-focal parabolic interfaces has been found. 
574: It is shown that there exist two types of solutions 
575: with two convex and with two concave parabolas. The relations 
576: between the Peclet numbers and the control parameters of the process 
577: for both types of solutions are obtained. 
578: 
579: As usual in theories of the steady-state growth, 
580: the continuous spectrum of solutions with one 
581: free parameter exists. 
582: The unresolved problem of ``selection'' of unique solution 
583: is briefly discussed. 
584: 
585: \bigskip
586: 
587: \section{Acknowledgments}
588: This work was supported in part by the Deutsche Forschungsgemeinschaft under project 
589: SPP 1120.
590: The author would like to thank L. Bagrova, E. Brener, P. Galenko, 
591:  and H. M\"uller-Krumbhaar for fruitful discussions.
592: 
593: \newpage
594: 
595: \begin{thebibliography}{99}
596: \bibitem{yoon1} Yoon DN, Hupmann WJ. Acta Metall 1979;27:973. 
597: \bibitem{yoon2} Yoon DN. Int Mater Rev 1995;40:149. 
598: \bibitem{musch1} Muschik T, Kaysser WA and Hehenkamp T. 
599: Acta Metall 1989;37:603. 
600: \bibitem{4} Yoon DN, Cahn JW,  Handwerker CA,  Blendell JE,
601:   Baik YJ. In: Interface Migration and Control of Microstrucutres.
602:   Am Soc. Metals. Park. Ohio (1985), pp. 19-31. 
603: \bibitem{1} Ivantsov GP. Dokl Akad Nauk SSSR 1947;58:567.
604: \bibitem{2} Ivantsov GP. Dokl Akad Nauk SSSR 1952;83:573.
605: \bibitem{3} Saito Y. Statistical Physics of Crystal Growth, World
606:   Scientific Publishing, Singapore, 1996.
607: \bibitem{5}  Brener EA,   Mel'nikov VI. Adv Phys 1991;40:53.
608: \end{thebibliography} 
609: 
610: 
611: 
612: \newpage
613: FIGURE CAPTIONS
614: 
615: Figure 1: Two types of combined moving parabolic fronts: convex (left) and 
616: concave (right).
617: 
618: Figure 2: The dependence of the reduced temperature $\Delta$ on the Peclet number
619: $P_1$ for convex and concave fronts. The insert represents the region of small
620: $\Delta$. Parameters are: $\Delta_m=0.01$ and $\alpha=0.25$.
621: 
622: Figure 3: Three possible cellular structures 
623: of two combined moving interfaces with convex [(a) and (b)] 
624: and concave (c) central parabolic part. $S1$ and $S2$ are 
625: melting solid and growing solid, and $L$ is the liquid phase. 
626: The direction of motion is defined by the arrow of $V$. 
627: Dashed regions in (a)-(c) show the central parabolic part of interfaces. 
628: 
629: 
630: 
631: \newpage
632: 
633: \begin{figure}
634: \begin{center}
635: \epsfig{file=fig1a.eps, width=8cm}
636: \epsfig{file=fig1b.eps, width=8cm}
637: \caption{	}
638: \end{center}
639: \end{figure}
640: 
641: 
642: \newpage
643: 
644: \begin{figure}
645: \begin{center}
646: \epsfig{file=fig2.eps, width=10cm}
647: \caption{	}
648: \end{center}
649: \end{figure}
650: 
651: \newpage
652: 
653: \begin{figure}
654: \begin{center}
655: \epsfig{file=fig3.eps, width=17cm}
656: \caption{	}
657: \end{center}
658: \end{figure}
659: 
660: 
661: 
662: 
663: \end{document}
664: