Formation and propagation of a shock wave due to evaporation processes at imperfect interfaces Yoshimoto Onishi, OOSHIDA Takeshi and Katsuhiko Tsubata Department of Applied Mathematics and Physics, Tottori University, Tottori 680-8552, Japan Abstract. Transient motions of a vapor due to evaporation processes from its plane condensed phase at imperfect conditions have been considered numerically based not only on the Boltzmann equation of BGK type but also on the formulation of fluid dynamic level, i.e., the Navier-Stokes equations subject to the boundary conditions appropriate for evaporation and condensation derived earlier from the kinetic theory analysis. The imperfect ness of the interface of the condensed phase has been taken into account in terms of an adjustable parameter ac first introduced by Wort berg and his co-workes. The parameter, variously called the condensation coefficient or condensation parameter or evaporation coefficient, may presumably be associated with some kind of imperfectness of the interface but has nothing to do with the condensation coefficient commonly defined at the level of kinetic theory. The results based on both of these systems of equations agree quite well, describing even the process of establishment of the flow field as well as its established state. Some of the results obtained are compared with the experimental results available for Helium II by Furukawa and Murakami. The comparison between the present results and the experimental ones shows that the incorporation of the adjustable parameter seems to work fairly well in some cases but not in other cases. No decisive conclusion can yet be drawn. However, this parameter is simple and may serve to incorporate to some extent ambiguous nature associated with the imperfectness of the interface. INTRODUCTION Flow problems involving phase-change processes are the ones to which the ordinary continuum-based fluid dynamics is not directly applicable because of the existence of a nonequilibrium region called the Knudsen layer in the vicinity of the condensed phase. The analysis for such problems, therefore, must necessarily be based on kinetic equations because it is this nonequilibrium region that is responsible for the phase-change processes to occur and its existence can never be neglected in any problems even in the continuum limit. However, if this nonequilibrium region is small in its thickness, the flow field may well be described by the Navier-Stokes equations. The problem arising then is that what kind of boundary conditions for fluid dynamic quantities are to be specified at the condensed phase. The answer is to be given from the analysis of kinetic theory, and the conditions for fluid dynamic quantities to satisfy at the condensed phase have already been given in general terms (see e.g., [1]), although, of course, there are some restrictions on the values of the parameters involved and hence on the applicability of the conditions. From a number of studies having so far been done in our laboratory, these conditions are good enough when applied to the system of Navier-Stokes equations, giving good agreement with the results based on the Boltzmann equation of BGK type [2]. Therefore, the system of Navier-Stokes equations subject to these conditions at the condensed phases may be called the fluid dynamic formulation [3], which is valid for problems with phase change processes involved. However, the experimental verification of the numerical results on both of these systems of equations has not yet been done because of lack in the experimental results appropriate for the comparison. A recent publication by Furukawa fe Murakami [4] on experiments of evaporation processes of Helium II from its plane condensed phase has now made it possible to make comparison between the two results, the experimental and the numerical. Here we consider a simple flow field of a vapor due to evaporation and condensation processes from its plane condensed phase based not only on the Boltzmann equation of BGK type [2] but also on the fluid dynamic formulation just mentioned above. A new feature in the present study is that some imperfectness of the interface of the condensed phase is taken into account in terms of a parameter ac variously called the CP585, Rarefied Gas Dynamics: 22nd International Symposium, edited by T. J. Bartel and M. A. Gallis © 2001 American Institute of Physics 0-7354-0025-3/01/$18.00 591 condensation coefficient or condensation parameter or evaporation coefficient, which probably has first been used by Wortberg and his co-workes [5] in their experimental and approximate analytical studies on this kind of problems. So far the commonly used boundary condition at the interface for the velocity distribution function of molecules leaving the condensed phase has been a Maxwellian distribution function corresponding to a saturated state of the vapor at a temperature of its condensed phase, i.e., a distribution function characterized by the temperature and velocity of the condensed phase, its number density being the saturated vapor number density at that temperature. The introduction of ac, on the other hand, represents that a cetrain fraction 1 — ac of molecules leaving the condensed phase may be subject to a velocity distribution of diffusive type holding at a solid wall, namely, a Maxwellian distribution function characterized by the temperature and velocity of the condensed phase but with the number density being determined by the no net mass flow condition across the boundary surface. This parameter ac, which of course has nothing to do with the condensation coefficient commonly defined at the level of kinetic theory, is assumed here to be associated with the imperfectness of the interface of the condensed phase. Actually, it may happen that otc gives a certain measure of the imperfectness of the interface or the imperfect accommodation of molecules at the interface. As to the boundary conditions at the level of the fiuid dynamic formulation, on the other hand, these conditions have to be modified also to incorporate ac, because they have been derived from an analysis of the Boltzmann equation of BGK type subject to the commonly used equilibrium distribution function as the boundary condition at the interface (see [1]). The method of inclusion of ac into the conditions has been proposed by Sone and Sugimoto [6] and we have used the modified boundary conditions for the analysis to follow. FORMULATION OF THE PROBLEM Let a plane condensed phase be located at x = 0 and a half-space ( x > 0 ) be occupied by the stationary vapor. Initially, the condensed phase and the vapor phase are in complete equilibrium at a temperature TO . The pressure, density and number density of the vapor at this state are PQ, p$ and N$ , respectively. Suppose that, at time t = 0, the temperature of the condensed phase is suddenly changed, i.e., TO —> TW - This leads to the onset of phase change processes at the condensed phase, giving then rise to transient motions of the vapor accompanied by shock waves and the contact regions (sometimes expansion waves involved). We investigate the transient motions based not only on the Boltzmann equation of BGK type [2] but also on the Navier-Stokes equations. The boundary conditions at the condensed phase for these two systems of equations will be specified in accordance with the statement given in the Introduction. Kinetic formulation The Boltzmann equation of BGK type [2], which is adopted here as the governing system in kinetic theory for the description of the motions of the vapor in the present one-dimensional problem, may be written as = ^e-/), (1) N 1 £>X (3) (4) where t is the time; x is the the molecular velocity vector; distribution characterized by number density, the velocity, coordinate measured from the original position of the membrane; (£x, £ y , £z ) is / is the molecular velocity distribution function, Fe being the local Maxwellian the local fluid dynamic quantities; N, w, T, P and p are, respectively, the the temperature, the pressure and the density of the gas; m is the molecular 592 mass; k is the Boltzmann constant and R = k/m the gas constant per unit mass of gas. vc is a constant associated with the collision frequency ( Nvc is the local collision frequency) and, hence, can be calculated either from the viscosity // or from the thermal conductivity A of the gas at a certain reference state, say, at the initial equilibrium state, by the following relation NM = ^ = \R^, ^0 * (5) ^0 the suffix 0 being understood to indicate the quantities at the initial state. It may be noted here that, in the BGK model equation, the relation A = (5/2)1? // holds and, hence, the Prandtl number is unity for this model equation. The initial condition for the distribution function / for the present problem is, at t = 0 ^n ^ f x>0 for 6 - The boundary condition for / at infinity (x —> oc) is exactly of the same form as the initial condition (6) for all times and for £x < 0. The boundary condition at the condensed phase (x = 0), however, is , e for ^ n at . <•>« w *= with Nw = acNw + (l- ac)NB (0 < ac < 1), (8) where NW is the number density for molecules emitted from the condensed phase. NW is the saturated vapor number density at the temperature TW of the condensed phase to be determined by the Clapeyron-Clausius relation as w > =exp "" p i P$ { rr T\lw"" 1 i j rr= ^* 9/ O x The other quantity NB, on the other hand, is the number density for molecules reflected from the condensed phase determined by the condition of no net mass flow across the phase boundary. Ng, therefore, is obtained from the distribution function for incident molecules to the condensed phase (^ < 0) as From this, it may be noted that NB is yet an unknown constant to be determined as part of the solution. Fluid dynamic formulation The governing system at the fluid dynamic level for the description of the motions of the gas in the present problem is the set of the Navier-Stokes equations, which may be written as 8 d = 0, (11) with P = pRT, (12) T, (13) 7-1 4 du n .£ T 1 = 1. 593 (14) w TABLE 1. Characteristic length and time scales, L and TO, corresponding to the experimental conditions [4]. Case A B C T 0 (°K) 1.74 1.94 2.04 Po(Pa) 1315 2610 3512 c 0 (m/s) 75.91 79.40 81.01 L (/urn) 2.820293xlO~ 2 1.666244 x!0~2 1.331662 x!0~2 /uo(Pas) 4.46xlO~ 7 5.00xlO~ 7 5.27xlO~ 7 TO (/usec) 3.3916xlO~4 1.9157xlO~4 1.5006xlO~ 4 where e and h are the specific internal energy and enthalpy, respectively. cp is the specific heat at constant pressure and 7 is the specific heat ratio (7 = 5/3 assumed). rxx is the viscous stress and q the heat flux. ju and A are here functions of T, which is in consistence with the temperature-dependence of these transport coefficients derived from the Boltzmann equation of BGK type. The initial conditions for this set of equations are, at t = 0 u = 0, P = PQ T = T0, for x > 0 . (16) As to the boundary conditions, the conditions at infinity is exactly the same as those in (16). The conditions at the condensed phase (x = 0), on the other hand, may be specified as P - Pw _ Pw TW u 4 1 2 ' ( } - where C| = —2.132039, d\ = —0.446749. These are the macroscopic conditions holding at the phase boundary having been given earlier by the kinetic theory analysis [1]. For the corresponding expressions with the imperfectness of the condensed phase ac incorporated, however, the constant C| above should be replaced by C% = C% — 2-y/TT (1 — a c )/a c , whereas d\ is left unchanged (see [6]). This modification in the constant C| will be made where and when it becomes necessary. RESULTS AND DISCUSSION For the present problem, we introduce the length scale L and the time scale TO defined as (2JRT0)1/2 P0 where CQ and /o are the sound speed and the molecular mean free path, respectively, at the initial state. They are defined by CQ = (7 RT$)1^ and /o = (jUo/Po)(8^To/7r) 1 / 2 . It is mentioned that, since the length scale L is of the order of the molecular mean free path JQ, the time scale TO here represents the mean collision time of gas molecules at the initial state. In Table 1, the actual values of the length and time scales, L and TO, corresponding to the experimental conditions done by Furukawa & Murakami [4] are listed just for reference. With these length and time scales together with the fluid dynamic quantities at the initial state, both of these systems of the governing equations, the Boltzmann equation of BGK type and the Navier-Stokes equations, are appropriately nondimensionalized. The Prandtl number Pr defined by AO 7 — 1 AO (20) which appear in the nondirnensional Navier-Stokes equations, has been set equal to unity here for simplicity (cf. the Prandtl number is always unity for the Boltzmann equation of BGK type). Incidentally, it may be noted that the Reynolds number Re defined by Re = p^c^L/jJL^ simply becomes Re = (27) 1 / 2 in this case. 594 1.01 1.000 100 X/L2QQ 0 100 X/L2QQ 100 X/L 1.000200 FIGURE 1. A transient flow of a vapor due to evaporation from its plane condensed phase with its different imperfectness, ac = 1.0 and ac = 0.72. TW/T0 = 1.00755, Pw/Po = 1.05083 and Pr = 1. Solid lines: Boltzmann. Dashed lines: Navier-Stokes. • : experimental data measured ata;/L = 5 x ! 0 4 from the interface, the initial temperature and pressure being TO = 1.74°K and PO = 1315Pa, respectively (Furukawa and Murakami [4]). The numbers in the graphs indicate the values of time £/TO. The initial shock wave with its Mach number Ms = 1.00667 is from left to right. The pressure and the flow velocity behind the shock wave are P/Po = 1.0168 and U/CQ =0.0100, respectively. In the pressure distribution of the Boltzmann's solution, the Knudsen layer is clearly visible in the close vicinity of the interface. Therefore, the parameters which characterize the present flow fields are, in addition to the Prandtl number having already been set equal to unity here, ac Tw_ To' Pw (21) where PW is the saturated vapor pressure at TW and is known for He II. Numerical calculations have been carried out for various sets of the parameters in (21), based on both governing systems of equations. First of all, we have confirmed that, for several sets of the parameters, the 595 results based on the Boltzmann equation agree quite well with those based on the fluid dynamic formulation or simply the Navier-Stokes system. With this fact in mind, most of the calculations for various other sets of the parameters are carried out based on the fluid dynamic formulation. Figure 1 shows one of the results for the transient behavior of the relevant fluid dynamic quantities for Case A in Table 1 with different values of ac, i.e., ac = 1.0 and ac = 0.72. Good agreement can be seen between the results of the Boltzmann and those of the fluid dynamic formulation. Incidentally, it is also recongnized that a wave or a region with steep gradients is produced associated with the evaporation process occurring at the interface and the wave develops into a shock wave and a contact region as time goes on. The time required for a shock wave to be fully established is about t/TQ = 50, which corresponds approximately to an actual time t = O.OlT^usec for the He II experiment done by Furukawa & Murakami [4]. Unfortunately, their experiments correspond only to the established states of the unsteady flow fields and, therefore, no experimental data is available for the establishing process of the flow fields, i.e., for the process of formation of waves at very early stages to the eventual shock waves and contact regions. The contact region formed behind the shock wave develops gradually into a slightly wider region as time goes on due to the diffusion effects involved within the region, for example, the region expands by about po)^ ~ 0.14//m in both directions for a period of time £/TQ = 50 for the present experimental situation (or in terms of the diffusion velocity, y/ PQ / (pat) ~ 0.109co, where CQ ~ 76m/s at TO = 1.74°K). Consequently, the contact region with its left edge reached to the interface (because of the relatively large diffusion velocity compared with the flow velocity) seems to be developing from the interface. Within this contact region, the temperature and density have fairly large slopes as indicated in the figure. Furthermore, the velocity has a hump (very small in this case), which was pointed out by Onishi et. al [7] as a new feature within the contact region. At about this stage, the Knudsen layer has already established itself and its thickness is about four times the length of the molecular mean free path at the initial state (see e.g., the pressure distribution). The thickness of the established shock wave is about 50I/ = 1.41 jum and its Mach number is Ms = 1.00667, where Ms = US/CQ, Us being the speed of the shock. As to the transient processes of the flow fields, there is almost no qualitative difference between the results for ac = 0.72 and for ac = 1. The experimental data by Furukawa & Murakami [4], which are reported to have been measured at a distance of x/L = 5 x 104 from the interface, are also indicated by a symbol • in the figure. The agreement between the numerical and the experimental results happens to be excellent in this case with ac = 0.72 but this is not always the case. Actually, there are a number of cases in which the numerical and the experimental results are not in good agreement, whatever values of ac may be. One of such cases is shown in Fig. 2, which corresponds to Case C of Table 1. The experimental values are far larger than the numerical ones. No definite reason can be given at the moment for this difference but one reason may be attributed to the errors involved in the experiments in this case and in other cases which are not in good agreement with the numerical results. Of course, this may also be interpreted to indicate that a single adjustable parameter such as ac introduced here is not sufficient to account for the imperfectness or the ambiguous nature of the interface. However, a further study should be continued of the roles and the effects of this parameter ac because the parameter is simple and, as indicated in Fig. 1, it might serve in certain situations to account for the ambiguous and complex nature of the interface to some extent. Finally, in order to see further the effects of ac on various fluid dynamic quantities at the calculation al level, we have shown in Fig. 3 some of the numerical results for various ac, which have been calculated based on both the kinetic and the fluid dynamic formulations. It is clearly seen that as the parameter ac becomes small, the evaporation process weakens and hence a weaker shock wave is produced. This is clear from the nature and the role of the parameter ac (see Eq. (8)). From the calculational results of this kind, it may become possible to estimate, in turn, the imperfectness of the interface by comparison with the experiments. The authors are grateful to Prof. Murakami and Dr. Furukawa at Tsukuba University for providing their experimental data. This work was partially supported by the Grant-in-Aid (No. 10650175) from the Ministry of Education in Japan. REFERENCES 1. Onishi Y., and Sone Y., /. Phys. Soc. Japan 47, 1676-1685 (1979). 2. Bhatnager P.L., Gross E.P., and Krook M., Phys. Rev. 94, 511-525 (1954). 3. Tanaka T., Onishi Y. and Miura H., Fluids Engineering Conference'99 held at Tot tori, Japan on October, 1999, pp. 113-114 (in Japanese). 4. Furukawa T., and Murakami M., Rarefied Gas Dynamics, edited by R. Brun, R. Camp argue, R. Gatignol, J.C. Lengrand, Cepadues-Editions, Toulouse, France, 1999, pp. 519-526. 596 1.02 1.01- 1.000 100 x/L 200 0 100 x/L 200 100 x/L 200 1.005 1.000FIGURE 2. A transient flow of a vapor due to evaporation from its plane condensed phase with its different imperfectness, ac = 1.0 and ac = 0.72. TW/T0 = 1.00658, Pw/Po = 1.03889 and Pr = 1. Solid lines: Boltzmann. Dashed lines: Navier-Stokes. • : experimental data measured ata;/L = 5 x l 0 4 from the interface, the initial temperature and pressure being TO = 2.04°K and PQ = 3512Pa, respectively (Furukawa and Murakami [4]). The numbers in the graphs indicate the values of time £/TO. The initial shock wave with its Mach number Ms = 1.00699 is from left to right. The pressure and the flow velocity behind the shock wave are P/Po = 1.0176 and U/CQ =0.0105, respectively. In the pressure distribution of the Boltzmann's solution, the Knudsen layer is clearly visible in the close vicinity of the interface. 5. Mager R., Adomeit G., and Wortberg G., Rarefied Gas Dynamics: Physical Phenomena (Progress in Astronautics and Aeronautics, Vol. 117), edited by E.P. Muntz, D.P. Weaver and D.H. Campbell, Washington B.C.: AIAA, 1989, pp. 460-469. 6. Sone Y., and Sugimoto H., Adiabatic Waves in Liquid-Vapor Systems, edited by G.E.A. Meier and P.A. Thompson, Berlin: Springer-Verlag, 1990, pp. 293-304. 7. Onishi Y., OOSHIDA T. and Umemura N., "On a small structure in velocity field within a contact region" to be published in Phys. Fluids. 597 p_ Po 1.02 1.01 1.01 1.00 1.000 100 x/L 200 0 100 x/L 100 x/L 200 0 100 JC/L 200 200 1.000FIGURE 3. Effects of the parameter ac on a transient flow of a vapor at time I/TQ = 220 due to evaporation from its plane condensed phase. TV/To = 1.00755, Pw/Po = 1.05083 and Pr = 1. Solid lines: Boltzmann. Dashed lines: Navier- Stokes. • : experimental data measured at x/L = 5 x l 0 4 from the interface, the initial temperature and pressure being TO = 1.74°K and PO = 1315Pa, respectively (Furukawa and Murakami [4]). In the pressure distribution of the Boltzmann's solution, the Knudsen layer is clearly visible in the close vicinity of the interface. 598