1: 

2: \documentclass[prl,twocolumn]{revtex4}

3: \usepackage{graphicx}

4: \usepackage{psfrag}

5: \usepackage{psfig}

6: 

7: \begin{document}

8: \title{Asymptotic behaviour of the Rayleigh--Taylor instability}

9: \author{Laurent Duchemin$^1$, Christophe Josserand$^2$ and Paul Clavin$^3$\\

10: \small $^1$ Department of Applied Mathematics and Theoretical Physics, University of Cambridge,

11: Cambridge CB3 0WA, United Kingdom\\

12: \small  $^2$ Laboratoire de Mod\'elisation en M\'ecanique\\

13: \small UPMC-CNRS UMR 7607, 4 place Jussieu, 75252 Paris C\'edex 05 France\\

14: \small $^3$ IRPHE, Universit\'es d'Aix-Marseille I \& II-CNRS, 49 rue Joliot-Curie, BP 146, 13384 Marseille Cedex France\\

15: }

16: 

17: \begin{abstract}

18: We investigate long time numerical simulations of the inviscid Rayleigh-Taylor instability at Atwood 

19: number one using a boundary integral method. We are able to attain the asymptotic behavior for the

20: spikes predicted by Clavin \& Williams\cite{clavin} for which we give a simplified demonstration.  In

21: particular we observe that the spike's curvature evolves like $t^3$ while the overshoot in acceleration

22: shows a good agreement with the suggested $1/t^5$ law. Moreover, we obtain consistent results

23: for the prefactor coefficients of the asymptotic laws. Eventually we exhibit the self-similar behavior of

24: the interface profile near the spike.

25: \end{abstract}

26: \maketitle

27: 

28: \section{Introduction}

29: The Rayleigh-Taylor (RT) instability appears when, under gravity, an heavy liquid is placed over a lighter one\cite{rayleigh}. 

30: This instability is crucial for our understanding of different phenomena in fluid mechanics: mixing, thermal 

31: convection (\cite{kada} and cited ref. herein) and also finger number selection in splashes\cite{gueyffier}. 

32: It is also important in inertial confinment fusion where the mass ablation provides a stabilizing effect to

33: the interface instability\cite{sanz}. Without ablation, after the exponential growth of the perturbations due

34: to the linear RT instability, nonlinear profiles develop through the formation of bubbles of lighter fluid 

35: rising into the heavier one and falling

36:  spikes of the heavier liquid penetrating the lighter one. In the 

37: general situations of viscous fluids which are immiscible and/or have Atwood number not equal to unity ($A_T=(\rho_h

38: -\rho_l)/(\rho_h+\rho_l)$ with $\rho_h$ and $\rho_l$ being the density of the heavier and lighter fluids

39: respectively), famous mushrooms-like structures grow for 

40: larger times\cite{kada,sohn1,sohn2}. The limit of an inviscid fluid above a vacuum ($A_T=1$) without surface tension plays 

41: a specific role since no stabilizing effects are present in the linear dynamics. Numerous theoretical

42: and numerical work have focused on this idealized limit in order to track insights into the instability

43: itself\cite{layzer,zhang,hazac,abar,mika,inog}. It has been

44: shown using a conformal mapping that a finite time singularity might appear in the conformal 

45: plane\cite{tanveer} and

46: it is also suspected that for some sufficiently irregular initial conditions finite time singularities should also 

47: be observed in the physical plane. However, starting with sufficiently smooth initial conditions, the 

48: asymptotic dynamics\cite{mika,zhang,inog}  presents a constant velocity rising bubble separated by free falling tiny spikes as 

49: displayed on figure \ref{profils}. Although the

50: rising bubble motion has been described using local properties of the flow\cite{gonch}, the asymptotic dynamics

51: of the spikes is far from being well understood. The single mode approach gives a fair 

52: description of the constant velocity of the rising bubble ($v_b=\sqrt{g/(3k)}$ where $g$ is the 

53: acceleration of the gravity and $k$ the wavenumber of the perturbation) but gives only partial results

54: for the spike\cite{zhang}. The fluid there obeys free fall dynamics to a good approximation and the 

55: pressure field of the flow leads to an overshoot in the acceleration. The accelerated

56: motion of the liquid stretches the spike geometry and one expects self-similar behaviour of the

57: tip of the spikes. 

58: \begin{figure}[h]

59: \begin{center}

60: \psfrag{vertical velocity}{\Large Vertical velocity}

61: \psfrag{time}{\Large time}

62: \psfrag{-pi}{$-\pi$}

63: \psfrag{0}{$0$}

64: \psfrag{-1}{$-1$}

65: \psfrag{-2}{$-2$}

66: \psfrag{1}{$1$}

67: \psfrag{2}{$2$}

68: \psfrag{3}{$3$}

69: \psfrag{4}{$4$}

70: \psfrag{5}{$5$}

71: \psfrag{6}{$6$}

72: \psfrag{7}{$7$}

73: \psfrag{8}{$8$}

74: \psfrag{5}{$5$}

75: \psfrag{10}{$10$}

76: \psfrag{pi}{$\pi$}

77: \psfrag{2p}{$2\pi$}

78: \psfrag{3p}{$3\pi$}

79: \psfrag{4p}{$4\pi$}

80: \psfrag{5p}{$5\pi$}

81: \psfrag{6p}{$6\pi$}

82: \psfrag{6.5}{$t=6.5$}

83: \psfrag{7.5}{$t=7.5$}

84: \psfrag{9.5}{$t=9.5$}

85: \centerline {\includegraphics[width=6cm]{all.eps}}

86: \end{center}

87: \caption{Snapshots of the interface subject to the Rayleigh-Taylor instability for time

88: ranging from $t=0$ to $10$, starting with a small amplitude sine mode (left). On the

89: right is shown the velocity of several points along the interface, non-dimensionalized with the stationnary 

90: bubble rising velocity $\sqrt{g/3k}$, as a function of time.\label{all}}

91: \label{profils}

92: \end{figure}

93: Recently, an asymptotic theory using a parallel flow description of the velocity field near the 

94: spikes has been constructed \cite{clavin}. The interface dynamics is nonlinear for large time and can 

95: be described using the theory of characteristics which gives rise to finite time singularity 

96: solutions. In the case of regular dynamics a self-similar 

97: description of the peak is obtained for large time: the maximal curvature of the interface at the

98: peak tip is found to behave like the cubic power of time $t^3$. Moreover, the spike position,

99: following the free fall $ \frac12 g t^2$ at leading order, is shown to converge to the constant 

100: acceleration $g$ with an overshoot in acceleration decreasing like $t^{-5}$. In this letter, we 

101: present a numerical study of the Rayleigh-Taylor instability which focuses on the large time dynamics 

102: of the spikes in order to investigate the self similar dynamics predicted in \cite{clavin}.

103: We consider  the dynamics for an inviscid liquid (heavy)  with an exterior fluid of zero density 

104: ($A_t=1$) and no surface tension. The numerics use a boundary integral method (BIM later on).

105: Due to strong numerical instabilities, a careful treatment of the

106: interface using conformal mapping is needed as explained below. The results are then 

107: shown and compared with the theory.

108: 

109: \section{Asymptotic analysis and numerical method}

110: 

111: We consider the two-dimensional motion of an inviscid fluid 

112: above a vacuum, subject to a negative acceleration $-g$. A periodic

113: sine perturbation of the interface of wave number $k$ is implemented as initial conditions.

114: Neglecting surface tension, the equations of motion have no 

115:  control parameter after rescaling the time, the position and the velocity potential $\varphi$ by 

116:  factors $\sqrt{gk}$, $k$ and $\sqrt{k^3/g}$ respectively.

117: The interface is described by $y=\alpha(x,t)$, where $y$ is the direction along the 

118: gravity and $x$ orthogonal to it (see figure \ref{map}). 

119: The velocity field ${\bf U}=(u,v)$ satisfies the dimensionless Euler equation

120: $$ \frac{d{\bf U}}{dt}=-{\bf \nabla}P +{\bf e_y} $$

121: where $P(x,y,t)$ is the pressure, ${\bf e_y}$ the non-dimensional acceleration due to gravity and 

122: the fluid density $\rho=1$.

123:  The kinetic equation for the interface reads~:

124: $$ \frac{\partial \alpha(x,t)}{\partial t}+u\frac{\partial \alpha(x,t)}{\partial x}=v$$

125: with the velocity field $(u,v)$ evaluated at the interface $(x,\alpha(x,t))$. 

126: Starting at time $t=0$ with a small sine amplitude interface, we observe for large time 

127: that the fluid particles located in the vicinity of the tiny spikes come from an almost free fall from the 

128: initial interface region. Therefore, following \cite{clavin}, 

129: we assume quasi-parallel steady flow for the velocity field which gives then in the tip region $ |u|\ll |v|$ and~:

130: $$ v \sim \sqrt{2y}$$ 

131: with $y \sim \frac12 t^2$ for large time.

132: Writing a perturbation expansion of the velocity field in the tip region $|x| \ll y$, we in fact consider:

133: $$ v=\sqrt{2(y+f(x,y,t))} $$

134: with $ f(x,y,t) \ll y$. Taking a Taylor expansion in $x$ of the perturbation $f$, we obtain by symmetry:

135: $$ v = \sqrt{2y}+\frac{f_0(y,t)}{\sqrt{2y}}+\frac{x^2}{2} \frac{f_2(y,t)}{\sqrt{2y}} + O(x^4) $$

136: We limit our expansion to the second order in $x$ for the velocity field later on.

137: Incompressibility gives~:

138: $$ u=-\left(\sqrt{\frac{1}{2y}}+\frac{\partial (f_0(y,t)/\sqrt{2y})}{\partial y}\right)x + O(x^3).$$

139: At the leading order (where we neglect even the perturbation $f(x,y,t)$) we obtain the following 

140: expression for the interface location:

141: $$  \frac{\partial \alpha(x,t)}{\partial t}-\frac{x}{\sqrt{2\alpha(x,t)}} \frac{\partial \alpha(x,t)}{\partial x}=\sqrt{2\alpha(x,t)} $$

142: which can be solved using the methods of charasteristics (see \cite{clavin}).

143: Writing $\alpha(x,t)=t(\frac{t}{2} - \gamma(x,t))$ and noting that $\gamma(x,t)\ll t/2$ in the spike region, 

144: we obtain, after linearisation~:

145: $$ \frac{\partial \gamma(x,t)}{\partial t}-\frac{x}{t}\frac{\partial \gamma(x,t)}{\partial x}=0 $$

146: which has self-similar solution of the form $ \gamma(x,t)=\theta(xt)$. A first conclusion can be drawn 

147: about the curvature of the interface at the tip,  $\kappa=-\partial^2\alpha/\partial x^2|_{x=0}$, which is 

148: thus found to increase as the cubic power of time~:

149: \begin{equation}

150:  \kappa=t^3 \theta''(0)

151: \label{cubic}

152: \end{equation}

153: 

154: The next order terms of the expansion allow the determination of the function $f_0(y,t)$ near the tip.

155: Using the constant value of the pressure at the interface we use the projection of the Euler

156: equation at the interface on its local tangent~:

157: $$ \frac{du}{dt} + \frac{\partial \alpha(x,t)}{\partial x} \frac{dv}{dt} = \frac{\partial \alpha(x,t)}{\partial x}. $$

158: Since on the interface $dP(x,\alpha(x,t),t)/dx=0$.

159: We develop this equation at first non-zero order (which will end up to be the first order in $x$) 

160: with the expansion $\theta(xt)=\theta(0)+x^2t^2 \theta''(0)/2+O(x^4)$. 

161: Remembering that $|f| \ll y$, we can neglect also the larsge scale terms $\partial^2 (f_0(y,t)/\sqrt{2y})/\partial t \partial y$ and $\sqrt{2y}\partial^2 (f_0(y,t)/\sqrt{2y})/\partial y^2$ with respect to the others.

162: We obtain finally for the tip position $y=y_s$~:

163: $$ \frac{\partial f_0(y_s,t)}{\partial t} +\sqrt{2y_s}\frac{\partial f_0(y_s,t)}{\partial y}=\frac{df_0(y_s,t)}{dt}=

164: \sqrt{\frac{2}{y_s}} \frac{1}{\kappa} $$

165: 

166: Recalling that:

167: $ \frac{dy_s}{dt}=\sqrt{2y_s}+\frac{f_0(y_s,t)}{\sqrt{2y_s}} $

168: we obtain for the tip acceleration at leading order:

169: \begin{equation}

170:  \frac{d^2y_s}{dt^2}=1+\frac{1}{\sqrt{2 y_s}} \frac{df_0(y_s,t)}{dt}=1+\frac{2}{t^5 \theta''(0)} 

171: \label{quint}

172: \end{equation}

173: which corresponds to an overshoot in the spike acceleration decreasing as the fifth

174: power of time.

175: 

176: The numerical method is elaborated using the incompressible and potential properties

177: of the flow. The velocity field can thus be evaluated everywhere when the velocity 

178: potential is known on the interface thanks to Cauchy's theorem, in the spirit of pionnering works\cite{vinje,cokelet,BMO,MZ}. 

179: The non-dimensional Bernoulli equation on the free surface reads~:

180: \begin{equation}

181: \frac{\partial \varphi}{\partial t} = -\frac{1}{2}(\nabla \varphi)^2 + y,

182: \end{equation}

183: where the velocity potential $\varphi$ is a harmonic function in the 

184: fluid domain $\Omega$~:

185: \begin{equation}

186: \Delta \varphi = 0

187: \label{lap}

188: \end{equation}

189: 

190: The kinematic condition on the free surface expresses the fact that 

191: fluid particles move with the same normal velocity than the free surface 

192: itself~:

193: \begin{equation}

194: \frac{d \bf x}{dt} \cdot {\bf n} = \nabla \varphi \cdot {\bf n}

195: \label{kin}

196: \end{equation}

197: 

198: Knowing $\varphi$ on the free surface at a given time-step, 

199: we search for the solution of equation (\ref{lap}) that satisfies 

200: this boundary condition (\ref{kin}). We use the complex potential $\beta(z) = \varphi + i \psi$ 

201: and the conformal map $f(z) = exp(-i z)$ (Cf. Figure \ref{map}), 

202: where $z=x+iy$ and $\psi$ is the stream function. The conformal map transforms the periodic 

203: domain $\Omega$ into the closed domain $M$. 

204: Since $\psi$ is harmonic inside $\Omega$, $\beta(z)$ is analytic 

205: inside $\Omega$ and therefore $\gamma(\zeta)=\beta(f(z))$ 

206: is analytic inside $M$. Using Cauchy's theorem, we obtain a

207: Fredholm equation of the second kind for the stream function $\psi$ which is solved using

208: discretization of the free surface ($\partial \Omega$ and thus $\partial M$).

209: This linear system of equations is solved using a $LU$ decomposition. 

210: Once we know $\psi$ on each point on $\partial M$, the complex velocity of each marker in the physical plane is given by~:

211: \begin{equation}

212: \frac{d \beta}{dz} = u - i \, v

213: \end{equation}

214: where $u$ and $v$ are the horizontal and vertical velocities 

215: respectively. 

216: This complex velocity is computed with a finite difference scheme 

217: using the values of the complex potential on the collocation points 

218: on $\partial \Omega$.

219: \begin{figure}[h]

220: \psfrag{omega}{$\Omega$}

221: \psfrag{zeta}{$\zeta=f(z)$}

222: \psfrag{x}{$y$}

223: \psfrag{y}{$x$}

224: \psfrag{function}{$f(z) = e^{-i z}$}

225: \psfrag{M}{$M$}

226: \centerline {\includegraphics[width=8cm]{map_new.eps}}

227: \caption{\label{map} Conformal map used to transform the physical periodic plane $\Omega$ into a closed domain $M$.}

228: \end{figure}

229: The position of the surface markers (kinematic condition) 

230: and the value of the velocity potential on each of these markers 

231: (Bernoulli equation) are then updated in time using a fourth 

232: order Runge-Kutta method. 

233: 

234: \section{Results and discussions}

235: We have performed numerical simulations of the Rayleigh-Taylor instability using the numerical

236: method described above. We start with a small amplitude sine-mode. The unavoidable

237: numerical noise cannot be damped by the numerics and the calculations always end up subject to 

238: numerical instabilities. 

239: Nevertheless, we emphasize that the numerical scheme used here is remarkably robust and 

240: can be accurately evolved to reach the large time where the scalings predicted by the theory \cite{clavin} are valid.

241:  Comparing our simulations with recent numerical works\cite{sohn1,sohn2,hazac}, we have been able to run the

242:  dynamics at least twice as far which corresponds roughly to an increase of a factor of $8$ in the tip's curvature.

243: 

244: The position of the spike is shown on figure \ref{spike} as function of time. We observe that the 

245: asymptotics dynamics are very well approximated by the relation $ y_s =\frac12 g(t-t_0)^2$ as shown in 

246: the inset to the figure with $t_0=3.74$. This remarkable behavior, in good agreement with the free fall 

247:  hypothesis,

248: suggests that $t_0$ is the time delay accounting for the initial exponential development of the instability. 

249: We will therefore present further data on the curvature dependance and the acceleration of the tip as functions

250: of this delayed time $t-t_0$ instead of $t$.

251: \begin{figure}[h]

252: \centerline {

253: \includegraphics[width=7cm]{spike_zoom2.eps}}

254: \caption{Position of the spike $y_s(t)$ as a function of time. The inset shows in a log-log plot of the spike

255: position (black curve) as function of time $t-t_0$ with $t_0=3.74$ obtained by a second order 

256: polynomial fit of $y_s$. the dashed line shows the expected behavior $\frac12 t^2$. }

257: \label{spike}

258: \end{figure}

259: The curvature $\kappa_s$ at the tip is then shown on figure \ref{courbure}. The large time asymptotic 

260: behavior is similarly found to follow the cubic law (see equation \ref{cubic}) with $\theta''(0)=1.5$.

261: \begin{figure}[h]

262: \centerline {

263: \includegraphics[width=7cm]{courspike.eps}}

264: \caption{Spike curvature $\kappa_s$ calculated at the tip $y=y_s$ as function of the delayed time $t-t_0$ 

265: in a log-log plot. The dashed line displays the cubic law (\ref{cubic}) with $\theta''(0)=1.5$.}

266: \label{courbure}

267: \end{figure}

268: In addition, the acceleration of the tip is computed by finite differences on the tip velocity and

269: the overshoot in the acceleration is presented on figure \ref{accel}. We observe that the results 

270: look noisier than the two previous ones. Two factors can explain such noise: firstly, we are

271: looking to a finite difference which decreases to zero so that the numerical errors are relatively more

272: important. However,

273: we note that the overshoot in acceleration shows a good agreement with the $1/t^5$ law, noting

274: that no adjustable parameter is used in this comparison.

275: \begin{figure}[h]

276: \centerline {

277: \includegraphics[width=7cm]{overshoot.eps}}

278: \caption{Overshoot in acceleration, defined as the difference between the tip acceleration and the

279: gravity. The plot is in log-log scale and with the delayed time $t-t_0$. The dashed line shows the 

280: theoretical prediction (\ref{quint}) using the value of $\theta''(0)$ obtained from figure \ref{courbure}}

281: \label{accel}

282: \end{figure}

283: Moreover, the self similar structure of the interface near the tip has been exhibited on figure \ref{similar}.

284: We observe after the proper rescaling on the left part of the figure that the interface 

285: profiles collapse onto a single curve near the spike.

286: \begin{figure}[h]

287: \centerline {

288: \includegraphics[width=6cm]{selfsim.eps}}

289: \caption{Self-similar structure of the tip: the interface profile around the spike have been superimposed 

290: on the right side of the figure for different time $t$ ranging from $4$ to $12$. The left side of the figure

291: shows the same curves rescaled by factor $1/(t-t_0)$ and $(t-t_0)$ for the $x$ and $y$ coordinates

292: respectively, following the scaling behavior predicted by the theory.}

293: \label{similar}

294: \end{figure}

295: 

296: 

297: We have thus exhibited large times numerical simulations of the Rayleigh-Taylor instability which 

298: present asymptotic scaling behavior in agreement with theoretical predictions using Taylor

299: expansions of the free fall velocity field at the spike\cite{clavin}. Although our numerics always stops

300: due to numerical instability, we have been able to reach large time enough to exhibit the cubic power in time dependance for the spike 

301: curvature and the inverse of the quintinc power of time decreasing of the overshoot in acceleration.

302: 

303: It is our pleasure to thank J. Ashmore for useful comments.

304: We acknowledge also the support of CEA through the contract CEA/DIF N° 4600051147/P6H29.

305: 

306: \begin{thebibliography}{99}

307: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi

308: \bibitem{rayleigh}  Lord Rayleigh, {\em Scientific Papers II (Cambridge University Press, Cambridge,

309: United Kingdom, 1900)}, p 200.

310: 

311: \bibitem{kada} B. Castaing, G. Gunaratne, F. Heslot, L. Kadanoff, A. Libchaber, S. Thomae, X. Wu, S. Zaleski and G. Zanetti, {\em J. Fluid Mech.} {\bf 204}, 1-30 (1989). 

312: 

313: \bibitem{gueyffier}  D. Gueyffier and S. Zaleski, {\em C. R. Acad. Sci. Paris IIb} {\bf 326}, 839-844 (1998).

314: 

315: \bibitem{sanz} J. Sanz, J. Ramirez, R. Ramis, R. Betti and R.P.J. Town, {\em Phys. Rev. Lett.} {\bf 89},

316: 195002 (2002).

317: 

318: \bibitem{sohn1} S.-I. Sohn, {\em Phys. Rev. E} {\bf 67}, 026301 (2003).

319: 

320: \bibitem{sohn2} S.-I. Sohn, {\em Phys. Rev. E} {\bf 69}, 036703 (2004).

321: 

322: \bibitem{layzer} D. Layzer, {\em Astrophys. J.} {\bf 122}, 1 (1955).

323: 

324: \bibitem{zhang}  Q. Zhang, {\em Phys. Rev. Lett.} {\bf 81}, 3391 (1998).

325: 

326: \bibitem{hazac} G. Hazak,{\em Phys. Rev. Lett.} {\bf 76}, 4167 (1996).

327: 

328: \bibitem{abar} S.I. Abarzhi,{\em Phys. Rev. Lett.} {\bf 81}, 337 (1998).

329: 

330: \bibitem{mika} K.O. Mikaelian,{\em Phys. Rev. Lett.} {\bf 80}, 508 (1998).

331: 

332: \bibitem{inog} N. Inogamov,{\em Astrophys. Space Phys.} {\bf 10}, 1-335 (1999). 

333: 

334: \bibitem{tanveer} S. Tanveer, {\em Proc. R. Soc. Lond. A} {\bf 441}, 501-525 (1993).

335: 

336: \bibitem{gonch} V.N. Goncharov, {\em Phys. Rev. Lett.} {\bf 88}, 134502 (2002).

337: 

338: \bibitem{clavin} P. Clavin \& F. Williams accepted for publication in J. Fluid Mech. (2005).

339: 

340: \bibitem{vinje} T. Vinje \& P. Brevig, {\em Adv. Water Resources} {\bf 4}, 77 (1981).

341: 

342: \bibitem{cokelet}  M.S. Longuet-Higgins \& E.D. Cokelet {\em Proc. R. Soc. Lond. A} {\bf 350}, 1--26 (1976).

343: 

344: \bibitem{BMO} G.R. Baker, D.I. Meiron and S.A. Orszag, {\em Phys. Fluids} {\bf 23}, 1485 (1980).

345: \bibitem{MZ} R. Menikoff and C. Zemach, {\em J. Comput. Phys.} {\bf 51}, 28 (1983).

346: 

347: %\bibitem{hecht}

348: 

349: \end{thebibliography}

350: 

351: \end{document} 

352: 

353: We can write Cauchy's theorem on $\partial M$~:

354: \begin{equation}

355: \int_\mathcal{\partial M} \frac{\gamma(\zeta)}{\zeta-\zeta_e}d\zeta = 0,

356: \label{cauchy}

357: \end{equation}

358: where $\zeta_e$ is a point outside $\partial M$.

359: It can be shown that the real part of equation \ref{cauchy} -- 

360: when $\zeta_e$ approaches $\partial M$ -- 

361: is a Fredholm equation of the second kind for the stream function $\psi$, 

362: which is known to have a unique solution. 

363: If we discretize the free surface in the physical plane with collocation 

364: points, the images of these points are on $\partial M$ and we can 

365: write a discrete version of equation \ref{cauchy}~:

366: \begin{equation}

367: \sum_{j=1}^N \left( Re(\Gamma_{k,j})\psi_j - Im(\Gamma_{k,j})\varphi_j \right) = 0 

368: \label{discrete}

369: \end{equation}

370: where the coefficients $\Gamma_{k,j}$ are known if we consider $\gamma$ 

371: to be linear between two collocation points on $\partial M$. 

372: Having $N$ collocation points $\zeta_k$ on $\partial M$, 

373: we can write $N$ equations like \ref{discrete} 

374: corresponding $\zeta_e \to \zeta_k$. 

375: 

376: :

377: 

378: $$ -t^3\theta"(0) (\frac{\partial F_0(x_s,t)}{\partial t} +\frac{\partial f_0(x_s,t)}{\partial x})

379: -\frac{\partial^2 F_0(x,t)}{\partial t \partial x}+f_2(x_s,t)+\frac{g}{2x_s}=0 $$

380: 

381: which simplifies into

382: