International Journal of Engineering Science 41 (2003) 1685–1698 www.elsevier.com/locate/ijengsci
Explicit solutions to the one-phase Stefan problem with temperature-dependent thermal conductivity and a convective term
Mar ııa F. Natale, Domingo A. Tarzia * Depto. Matema tica, FCE, Universidad Austral, CONICET, Paraguay 1950, S2000FZF Rosario, Argentina Received 28 October 2002; accepted 18 December 2002
Abstract
A one-phase Stefan problem for a semi-infinite material with temperature-dependent thermalpffiffi conduc- tivity and convective term with a constant temperature or a heat flux condition of the type q0= t ðq0 > 0Þ at the fixed face x ¼ 0 is studied. For both cases a parametric representation of the solution of the similarity type is also obtained. Ó 2003 Elsevier Science Ltd. All rights reserved.
AMS: 35R35; 80A22; 35C05 Keywords: Stefan problem; Free boundary problem; Movingboundary problem; Phase-changeprocess; Nonlinear thermal conductivity; Fusion; Solidification; Similarity solution
1. Introduction
We consider a Stefan problem for a semi-infinite region x > 0 with phase-change temperature hf [18]. It is required to determine the evolution of the movingphase separation x ¼ sðtÞ and the temperature distribution hðx; tÞ. The modelingof this kind of systems is a problem with a great mathematical and industrial significance. Phase-change problems appear frequently in industrial processes an other problems of technological interest [1,2,9–11,13–16,19,30]. A large bibliography on the subject was given in [29].
* Correspondingauthor. Tel.: +54-3414-814990; fax: +54-3414-810505. E-mail addresses: [email protected] (M.F. Natale), [email protected] (D.A. Tarzia).
0020-7225/03/$ - see front matter Ó 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0020-7225(03)00067-3 1686 M.F. Natale, D.A. Tarzia / International Journal of Engineering Science 41 (2003) 1685–1698
Here, we consider a one phase-change process (one-phase Stefan problem) for a nonlinear heat conduction equation with a convective term. Owingto [26] we consider the followingfree boundary (fusion process) problem: oh o oh oh qc ¼ kðh; xÞ vðhÞ ; 0 < x < sðtÞ; t > 0 ð1Þ ot ox ox ox
oh q kðhð0; tÞ; 0Þ ð0; tÞ¼ p0ffiffi ; q > 0; t > 0 ð2Þ ox t 0
oh kðhðsðtÞ; tÞ; sðtÞÞ ðsðtÞ; tÞ¼ qlsðtÞ; t > 0 ð3Þ ox
hðsðtÞ; tÞ¼hf ; t > 0 ð4Þ
sð0Þ¼0 ð5Þ where the thermal conductivity kðh; xÞ and the velocity vðhÞ are given by
d vðhÞ¼qc ð6Þ 2ða þ bhÞ2
1 þ dx kðh; xÞ¼qc ð7Þ ða þ bhÞ2 and c, q and l are the specific heat, the density and the latent heat of fusion of the medium re- spectively, all of them are assumed to be constant with positive parameters a; b; d and a þ bhf > 0. The last condition guaranteed that v and k are well defined by maximal principle. This kind of nonlinear thermal conductivity or diffusion coefficients was considered in numerous papers, e.g. [4,7,8,17,21,23,27]. The nonlinear transport equation (1) arises in connection with unsaturated flow in heterogeneous porous media. If we set d ¼ 0 and b ¼ 0 in the free boundary problem (1)– (7) then we retrieve the classical one-phase Lam ee–Clapeyron–Stefan problem.pffiffi The first explicit solution for the one-phase Stefan problem was given in [18]. Here q0= t denotes the prescribed flux on the boundary x ¼ 0 which is of the type imposed in [28], where it was proven that the heat flux condition (2) on the fixed face x ¼ 0 is equivalent to the constant temperature boundary condition (43) for the two-phase Stefan problem for a semi-infinite material with constant thermal coefficients in both phases. This kind of heat flux at the fixed boundary x ¼ 0 was also considered in several applied problems, e.g. [3,12,22]. The goal of this paper is to determine which conditions on the parameters of the problem (in particular q0) must be satisfied in order to have an instantaneous phase-change process. In Section 2 we follow and improve [26] in the sense that the existence of an explicit solution of the problem (1)–(5) with nonlinear coefficients (6) and (7) is obtained for all data q0; q; c; l; hf ; a; b; d with the à restriction a þ bhf > 0. Moreover, this solution is given as a function of a parameter c which is M.F. Natale, D.A. Tarzia / International Journal of Engineering Science 41 (2003) 1685–1698 1687 given the unique solution of the transcendental equation (35). Although, if we replace the flux condition (2) for a new temperature boundary condition (43) on the fixed face x ¼ 0 we also obtain in Section 3 for all data the existence of solution which is given as a function of a parameter b which we prove it is the unique solution of the transcendental equation (71). This section is new with respect to [26]. In both cases, the explicit solution is given by a parametric representation of the similarity type. Other problems with nonlinear thermal coefficients in this subject are also given in [5,6,20,24,25].
2. Solution of the free boundary problem with heat flux condition on the fixed face
We consider the problem (1)–(5). Takinginto account (6) and (7) we can put our problem as ! oh o 1 þ dx oh d ¼ þ ; 0 < x < sðtÞ; t > 0 ð8Þ ot ox ða þ bhÞ2 ox 2bða þ bhÞ
1 oh qà ð0; tÞ¼ p0ffiffi ; t > 0 ð9Þ ða þ bhð0; tÞÞ2 ox t
1 þ dsðtÞ oh ðsðtÞ; tÞ¼ asðtÞ; t > 0 ð10Þ ða þ bhðsðtÞ; tÞÞ2 ox
hðsðtÞ; tÞ¼hf ; t > 0 ð11Þ sð0Þ¼0 ð12Þ
à where a ¼ l=c and q0 ¼ q0=qc: We define the followingtransformations in the same way as in [26]: 8 hi 1 > y ¼ 2 ð1 þ dxÞ2 1 < d hi 1 > SðtÞ¼2 ð1 þ dsðtÞÞ2 1 ð13Þ :> d hðy; tÞ¼hðx; tÞ and
8 R > yà ¼ yÃðy; tÞ¼ y ða þ bhðr; tÞÞdr þðab þ aÞSðtÞ > SðtÞ < tà ¼ t hà ¼ 1 ð14Þ > aþb h :> à à S ¼ y y¼S ¼ðab þ aÞSðtÞ 1688 M.F. Natale, D.A. Tarzia / International Journal of Engineering Science 41 (2003) 1685–1698
In order to obtain an alternative expression for yà we compute Z oyà y oh o ¼ ða þ bhðSðtÞ; tÞÞSðtÞþ b o ðr; tÞdr þðab þ aÞSðtÞ t SðtÞ t Z ! y o 1 oh ¼ abSðtÞþ b dr o 2 o S t r a bh r; t r ð Þ ð þ ð ÞÞ ! 1 oh 1 oh ¼ abSðtÞþb ðy; tÞ ÀÁðSðtÞ; tÞ 2 o 2 o ða þ bhðy; tÞÞ y a þ bhðSðtÞ; tÞ y b oh ¼ ðy; tÞ ða þ bhðy; tÞÞ2 oy Z ! y o b oh b oh ¼ ðr; tÞ dr þ ð0; tÞð15Þ o 2 o 2 o 0 r ða þ bhðr; tÞÞ r ða þ bhð0; tÞÞ y Z y oh bqà ffiffi0 ¼ b o ðr; tÞdr þ p ð16Þ 0 t t then Z Z Z t y o bqà y yÃðy; tÞ¼ ða þ bhðr; tÞÞdr þ pffiffiffi0 ds þ ða þ bhðr; 0ÞÞdr or s Z0 0 0 y pffiffi à ¼ ða þ bhðr; tÞÞdr þ 2bq0 t ð17Þ 0 Now, applying(13) and (14) the problem (8)–(12) is transformed in a classical Stefan problem given by
ohà o2hà pffiffiffiffi ¼ ; 2bqà tà < yà < SÃðtÃÞ; tà > 0 ð18Þ otà oyÃ2 0
ohà pffiffiffiffi qÃb pffiffiffiffi ð2bqà tÃ; tÃÞ¼p0ffiffiffiffi hÃð2bqà tÃ; tÃÞ; tà > 0 ð19Þ oyà 0 tà 0
ohà ðSÃðtÃÞ; tÃÞ¼aÃSÃðtÃÞ; tà > 0 ð20Þ oyÃ
à à à à à à h ðS ðt Þ; t Þ¼hf ; t > 0 ð21Þ SÃð0Þ¼0 ð22Þ where
à ab à 1 a ¼ ; hf ¼ ð23Þ ðab þ aÞða þ bhf Þ a þ bhf M.F. Natale, D.A. Tarzia / International Journal of Engineering Science 41 (2003) 1685–1698 1689
Then, if we introduce the similarity variable: yà nà ¼ pffiffiffiffiffiffiffiffiffiffi ð24Þ 2cÃtà where cà is a dimensionless positive constant to be determined, and the solution is sought of the type hÃðyÃ; tÃÞ¼HÃðnÃÞ pffiffiffiffiffiffiffiffiffiffi ð25Þ SÃðtÃÞ¼ 2cÃtà then, we get that (18)–(22) yields sffiffiffiffi d2Hà dHà 2 þ cÃnà ¼ 0; bqà < nà < 1 ð26Þ dnÃ2 dnà 0 cà sffiffiffiffi! sffiffiffiffi! dHà 2 pffiffiffiffiffiffiffi 2 bqà ¼ 2cÃqÃbHà bqà ð27Þ dnà 0 cà 0 0 cÃ
dHà ð1Þ¼aÃcà ð28Þ dnÃ
à à H ð1Þ¼hf ð29Þ The solution of the differential equation (26) is given by rffiffiffiffi ! cà HÃðnÃÞ¼Aerf nà þ B ð30Þ 2 where A and B are two unknown coefficients. From (27) and (29) we get