1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2: 

3: % This is a LaTex file, resubmitted on (26 April

4: % 2001)

5: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

6: 

7: %%%%%%%%%%%%%%%%%%%%%%%%%%%%

8: % authors: A. Carpio and L. L. Bonilla

9: %

10: %

11: %

12: % title: Wavefront depinning transition in

13: %   discrete one-dimensional reaction-diffusion

14: %   systems.

15: %

16: %

17: % Journal: PRL

18: %

19: %

20: % manuscript number:  LC8637 

21: %

22: % address: L. L. Bonilla

23: %       Escuela Politecnica Superior

24: %       Universidad Carlos III de Madrid

25: %       Avenida de la Universidad 30

26: %       E-28911 Leganes, Spain

27: %

28: % Internet: bonilla@ing.uc3m.es

29: %

30: % FAX:     34-91-624-9129

31: %

32: % Tel:     34-91-624-9445

33: %

34: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

35: \documentstyle[multicol,aps,epsfig]{revtex}

36: % for multicolumn

37: %\documentstyle[prl,aps,preprint,epsfig]{revtex}

38: % for camera-ready manuscript (RevTex 3.0)

39: \def\baselinestretch{2}

40: 

41: \tighten       %Gives single-space (RevTex 3.0)

42: 

43: 

44: \begin{document}

45: \draft %prints PACS numbers in

46: 

47: \title{Wavefront depinning transition in

48: discrete one-dimensional reaction-diffusion

49: systems}

50: \author{A. Carpio$^1$ and L. L. Bonilla$^2$ }

51: \address{$^1$Departamento de Matem\'{a}tica

52: Aplicada, Universidad Complutense de Madrid,\\

53: 28040 Madrid, Spain\\

54: $^2$Departamento de Matem\'{a}ticas, Escuela

55: Polit\'ecnica Superior,  Universidad Carlos III

56: de Madrid,\\ 

57: Avenida de la Universidad 30, 28911 Legan{\'e}s,

58: Spain }

59: \date{ \today  }

60: \maketitle

61: 

62: \begin{abstract}

63: Pinning and depinning of wavefronts are ubiquitous

64: features of spatially discrete systems describing

65: a host of phenomena in physics, biology, etc. A

66: large class of discrete systems is described by

67: overdamped chains of nonlinear oscillators with

68: nearest-neighbor coupling and controlled by

69: constant external forces. A theory of the

70: depinning transition for these systems, including

71: scaling laws and asymptotics of wavefronts, is

72: presented and confirmed by numerical calculations.

73: \end{abstract}

74: \pacs{82.40.-g; 05.45.-a}

75: 

76: \begin{multicols}{2}

77: \narrowtext

78: Spatially discrete systems describe physical

79: reality in many different fields: propagation of

80: nerve impulses along mielinated fibers

81: \cite{kee87,kee98}, pulse propagation through

82: cardiac cells \cite{kee98}, calcium release waves

83: in living cells \cite{bug97}, sliding of charge

84: density waves \cite{cdw}, superconductor Josephson

85: array junctions \cite{jj}, motion of dislocations

86: in crystals \cite{nab67}, atoms adsorbed on a

87: periodic substrate \cite{cha95}, arrays of

88: coupled diode resonators \cite{diode}, and weakly

89: coupled semiconductor superlattices

90: \cite{bon94,car00}. A distinctive feature of

91: discrete systems (not shared by continuous ones)

92: is the phenomenon of wavefront pinning: for

93: values of a control parameter in a certain

94: interval, wavefronts joining two different

95: constant states fail to propagate \cite{kee98}.

96: When the control parameter surpasses a threshold,

97: the wavefront depins and starts moving

98: \cite{kee87,cdw,nab67,car00}. The existence of

99: such thresholds is thought to be an intrinsecally

100: discrete fact, which is lost in continuum

101: aproximations. The characterization of propagation

102: failure and front depinning in discrete systems

103: is thus an important problem, which is still

104: poorly understood despite the numerous inroads

105: made in the literature

106: \cite{kee87,bug97,cdw,nab67,kla00,mit98,kin01}.

107: 

108: In this Letter, we study front depinning for

109: infinite one-dimensional nonlinear spatially

110: discrete reaction-diffusion (RD) systems. The

111: nature of the depinning transition depends on the

112: nonlinearity of the model, and is best understood

113: as propagation failure of the traveling front.

114: Usually, but not always, the wavefront profiles

115: become less smooth as a parameter $F$ (external

116: field) decreases. They become {\em discontinuous}

117: at a critical value $F_c$. Below $F_c$, the front

118: is pinned at discrete positions corresponding to a

119: stable steady state. Fig. \ref{fig1} shows

120: wavefront profiles near the critical field.

121: Individual points undergo abrupt jumps at

122: particular times, which gives the misleading

123: impression that the motion of the discrete fronts

124: proceeds by successive jumps. Wavefront velocity

125: scales with the field as $|F-F_c|^{{1\over 2}}$.

126: For exceptional nonlinearities, the wavefront

127: does not lose continuity as the field decreases.

128: In this case, there is a continuous transition

129: between wavefronts moving to the left for

130: $F>0$ and moving to the right for $F<0$: as for

131: continuous systems, front pinning occurs only at

132: a single field $F=0$. Wavefront velocity scales

133: then linearly with the field. We discuss the

134: characterization of the critical field (including

135: analytical formulas in the strongly discrete

136: limit), describe depinning anomalies (discrete

137: systems having zero critical field), and give a

138: precise characterization of stationary and moving

139: fronts near depinning (including front velocity)

140: by singular perturbation methods. Our

141: approximations show excellent agreement with

142: numerical calculations.

143: 

144: We consider chains of diffusively coupled

145: overdamped oscillators in a  potential $V$,

146: subject to a constant external force $F$:

147: \begin{equation}

148: {du_{n}\over dt} = u_{n+1}-2u_n + u_{n-1} + F -

149: A\,  g(u_n).   \label{Fd}

150: \end{equation}

151: Here $g(u)=V'(u)$ presents a `cubic' nonlinearity,

152: such that $A\, g(u)-F$ has three zeros, $U_1<U_2

153: <U_3$ in a certain force interval ($g'(U_i(F/A))

154: >0$ for $i=1,3$, $g'(U_2(F/A))<0$). Provided

155: $g(u)$ is odd with respect to $U_2(0)$, there is

156: a symmetric interval $|F|\leq F_c$ where the

157: wavefronts joining the stable zeros $U_1(F/A)$ and

158: $U_3(F/A)$ are pinned. For $|F|>F_c$, there are

159: {\em smooth traveling wavefronts}, $u_n(t)=

160: u(n-ct)$, with $u(-\infty)= U_1$ and $u(\infty)

161: =U_3$ \cite{zin92,car99}. The velocity $c(A,F)$

162: depends on $A$ and $F$ and it satisfies $cF<0$

163: and $c\to 0$ as $|F|\to F_c$ \cite{car99}.

164: Examples are the overdamped Frenkel-Kontorova

165: (FK) model ($g=\sin u$) \cite{FK} and the quartic

166: double well potential ($V=(u^2-1)^2 /4$). Less

167: symmetric nonlinearities yield a non-symmetric

168: pinning interval and our analysis applies to them

169: with trivial modifications.

170: 

171: {\em Critical field}. Stationary and traveling

172: wavefronts cannot coexist for the same value of

173: $F$ \cite{car99}. This follows from a comparison

174: principle for (\ref{Fd}) \cite{comparison}.

175: Pinning can be proved using stationary sub and

176: supersolutions, which can be constructed provided

177: the stationary solution is linearly stable. The

178: largest eigenvalue of the linearization of

179: (\ref{Fd}) about a stationary profile $u_n(A,F)$,

180: $u_n(t)= u_n(A,F) + v_n e^{\lambda t}$, is given

181: by

182: \begin{eqnarray}

183: -\lambda(A,F)=\mbox{Min}\, {\sum

184: (v_{n+1}-v_n)^2 +A g'(u_n(A,F)) v_n^2 \over \sum

185: v_n^2}\,, \label{var}

186: \end{eqnarray}

187: over a set of functions $v_n$ which decay

188: exponentially as $n\to\pm\infty$. The critical

189: field is uniquely characterized by $\lambda(A,F_c)

190: =0$ and $\lambda(A,F)<0$ for $|F| <F_c$. Thus two

191: facts distinguish the depinning transition: (i)

192: one eigenvalue becomes zero, and (ii) stationary

193: and moving wavefronts cannot coexist for the same

194: values of the field.

195: 

196: Equation (\ref{var}) shows that the critical field

197: is positive for large $A$ and typical

198: nonlinearities. In fact, consider the FK

199: potential. For $F=0$ there are two stationary

200: solutions which are symmetric with respect to

201: $U_2$, one taking on the value $U_2$ (unstable

202: dislocation), and the other one having $u_n \neq

203: U_2$ (stable dislocation) \cite{hob65}. For large

204: $A$, the stable dislocation has $g'(u_n)>0$ for

205: all $n$, and (\ref{var}) gives $\lambda(A,0)<0$.

206: Since $\lambda(A,F_c)=0$, this implies that the

207: critical field is nonzero. (A different proof can

208: be obtained using the comparison principle

209: \cite{kee87,car00}). As $A>0$ decreases, several

210: $u_n$ may enter the region of negative slope

211: $g'(u)$: the number of points with $g'(u_n)<0$

212: increases as $A$ decreases. It is then possible

213: to have $\lambda(A,0)=0$, i.e.\ $F_c=0$, for a

214: discrete system! Examples of this {\em pinning

215: anomaly} will be given below.

216: 

217: If $F>0$, the stable dislocation is no longer

218: symmetric with respect to $U_2$. If $F$ is not

219: too large, all $u_n(A,F)$ avoid the region of

220: negative slope $g'(u)<0$. For large $A$ and $F$

221: and the generic potentials above mentioned, we

222: have observed numerically that $g'<0$ for a single

223: point, labelled $u_0(A,F)$. This property

224: persists until $F_c$ is reached. How does $F_c$

225: depend on $A$? For $g=\sin u$, it is well known

226: that $F_c$ vanishes exponentially fast as $A$

227: goes to zero (the continuum limit). This was

228: conjectured by Indenbom \cite{ind58} on the basis

229: of a continuum approximation, and numerically

230: checked by Hobart \cite{hob65} in the context of

231: the Peierls stress and energy for dislocations.

232: More recently, Kladko et al

233: \cite{kla00} derived the formula $F_c =

234: C\,\exp(-\pi^2/\sqrt{A - A^2 /12})$ by means of a

235: variational argument. This argument can be used

236: for other potentials and it suggests that

237: $F_c\sim C\, e^{-\eta/\sqrt{A}}$ as

238: $A\to 0+$ (with positive $C$ and $\eta$

239: independent of $A$) holds for a large class of

240: nonlinearities \cite{kla00}. King and Chapman

241: have obtained an analogous result \cite{kin01}

242: using exponential asymptotics for the FK

243: potential, $F_c \sim C\, e^{-\pi^2/

244: \sqrt{A}}$ and the wavefront velocity after

245: depinning, $c\sim D\,\sqrt{(F^2- F_c^2)/A}$. This

246: later result agrees with the scaling law $c

247: \sim |F-F_c|^{{1\over 2}}$, found in a large class

248: of discrete RD equations \cite{kla00,mit98}.

249: 

250: {\em Anomalies of pinning}. Despite widespread

251: belief, it is not true that $F_c>0$ for all

252: discrete systems. Using the characterization

253: $\lambda(A,F_c)=0$, it is possible to see that

254: having a zero critical field is equivalent to

255: having a one-parameter family of continuous

256: increasing stationary profiles $u_n = u(n+\alpha)$

257: satisfying $u(x+1)+u(x-1)-2u(x) = A g(u(x))$,

258: with $u(-\infty)=U_1$, $u(\infty)=U_3$. In this

259: case, a standard perturbation argument yields a

260: wavefront speed having the same scaling as the

261: continuum approximation to the discrete system,

262: $c\sim C\, F$ as $F\to 0$. An example of

263: nonlinearity presenting this anomalous pinning

264: \cite{mcleod} can be obtained from $u(x)=\tanh

265: x$: it obeys the above equation with $A=1$, $U_1=

266: -1$, $U_3=1$ and $g(u)= - 2 \tanh^2(1)\, u

267: (1-u^2)/[1-\tanh^2(1)\, u^2]$. Furthermore the

268: wavefront velocity after depinning obeys the

269: relation, $c\sim - 3 F/2$ as $F\to 0$. Thus

270: nonlinearities presenting anomalous depinning

271: belong to a different universality class: the

272: wavefront velocity has a critical exponent 1 (and

273: $F_c=0$) instead of 1/2, which is the usual case

274: for discrete RD systems (having $F_c>0$).

275: 

276: {\em Asymptotic theory of wavefront depinning}.

277: We shall study the depinning transition in the

278: strongly discrete limit $A\gg 1$, in which the

279: structure of the wavefront is particularly

280: simple. Firstly, consider the symmetric

281: stationary profile with $u_n\neq U_2$ for $F=0$.

282: The front profile consists of two tails with

283: points very close to

284: $U_1$ and $U_3$, plus two symmetric points $u_0$,

285: $u_1$ in the gap region between $U_1$ and $U_3$.

286: As $F>0$ increases, this profile changes slightly:

287: the two tails are still very close to $U_1(F/A)$

288: and $U_3(F/A)$. As for the two middle points,

289: $u_1$ gets closer and closer to $U_3$ whereas

290: $u_0$ moves away from $U_1$. This structure is

291: preserved by the traveling fronts above the

292: critical field: there is only one active point

293: most of the time, which we can adopt as our

294: $u_0$. Then we can approximate $u_{-1}\sim U_1$,

295: $u_1\sim U_3$ in (\ref{Fd}), thereby obtaining

296: \begin{equation}

297: {du_{0}\over dt}\approx U_1\left({F\over A}

298: \right) + U_3\left({F\over A} \right)-2 u_0 -

299: A\, g(u_0) + F .\label{u0}

300: \end{equation}

301: This equation has three stationary solutions for

302: $F<F_c$, two stable and one unstable, and only one

303: stable stationary solution for $F>F_c$. The

304: critical field $F_c$ is such that the expansion

305: of the right hand side of (\ref{u0}) about the

306: two coalescing stationary solutions has zero

307: linear term, $2 +A g'(u_0)=0$, and

308: \begin{equation}

309: 2 u_0 + A \, g(u_0) \sim  U_1\left({F_c\over A}

310: \right) + U_3\left({F_c\over A} \right) + F_c .

311: \label{Fc1}

312: \end{equation}

313: These equations for $F_c$ and $u_0(A,F_c)$ have

314: been solved for the FK potential, for which $u_0 =

315: \cos^{-1}(-2/A)$ and $U_1+U_3 = 2\sin^{-1}

316: (F_c/A) + 2\pi$. The results are depicted in Fig.

317: \ref{fig2}, and show excellent agreement with

318: the numerical solution of (\ref{Fd})

319: for $A>10$. Our approximation performs less well

320: for smaller $A$, and it breaks down at $A=2$ with

321: the prediction $F_c=0$. Notice that $F_c(A)\sim

322: A$ as $A$ increases. In practice, only steady

323: solutions are observed for very large $A$.

324: 

325: Let us now construct the profile of the traveling

326: wavefronts after depinning, for $F$ sligthly above

327: $F_c$. Then $u_0(t)= u_0(A,F_c) + v_0(t)$ obeys

328: the following equation:

329: \begin{equation}

330: {dv_{0}\over dt}\approx (F-F_c) + A\, |g''(u_0)|\,

331: {v_{0}^{2}\over 2} ,\label{v0}

332: \end{equation}

333: where we have used $2+Ag'(u_0)=0$, (\ref{Fc1})

334: and ignored terms of order $(F-F_c)/A$ and higher.

335: This equation has the (outer) solution

336: \begin{equation}

337: v_0(t)\sim \sqrt{{2\, (F-F_{c})\over A\, |g''(

338: u_{0})|}}\, \tan\left(\sqrt{{A\, |g''(u_0)|\,

339: (F-F_{c})\over 2}}\, (t-t_0)\right)\,,

340: \label{outer}

341: \end{equation}

342: which is very small most of the time, but it

343: blows up when the argument of the tangent function

344: approaches $\pm \pi/2$. Thus the outer

345: approximation holds over a time interval

346: $(t-t_0)\sim \pi\sqrt{2}/\sqrt{A |g''(u_0)|\,

347: (F-F_{c})}$, which equals $\pi\sqrt{2} (A^2-4)^{

348: -{1\over 4}} (F-F_c)^{-{1\over 2}}$ for the FK

349: potential. The reciprocal of this time interval

350: yields an approximation for the wavefront

351: velocity,

352: \begin{equation}

353: |c(A,F)|\sim \sqrt{{A\, |g''(u_{0})|\,

354: (F-F_{c})\over 2\pi^{2}}}\,, \label{c}

355: \end{equation}

356: or $|c|\sim (A^2-4)^{{1\over 4}} (F-F_c)^{{1\over

357: 2}} /(\pi\sqrt{2})$ for a FK potential. In Fig.\

358: \ref{fig3} we compare this approximation with the

359: numerically computed velocity for $A=100$ and

360: $A=10$.

361: 

362: When the solution begins to blow up, the outer

363: solution (\ref{outer}) is no longer a good

364: approximation, for $u_0(t)$ departs from the

365: stationary value $u_0(A,F_c)$. We must go back

366: to (\ref{u0}) and obtain an inner approximation

367: to this equation. As $F$ is close to $F_c$ and

368: $u_0(t)-u_0(A,F_c)$ is of order 1, we solve

369: numerically (\ref{u0}) at $F=F_c$ with the

370: matching condition that $u_0(t)-u_0(A,F_c)\sim

371: 2/[\pi \sqrt{{1\over 2}\, A|g''(u_0)|/(F-F_c)} -A

372: |g''(u_0)|\, (t-t_0)]$, as $(t-t_0)\to -\infty$.

373: This inner solution describes the jump of $u_0$ to

374: values close to $U_3$. During this jump, the

375: motion of $u_0$ forces the other points to move.

376: Thus, $u_{-1}(t)$ can be calculated by using the

377: inner solution in (\ref{Fd}) for $u_0$, with

378: $F=F_c$ and $u_{-2}\approx U_1$. A composite

379: expansion \cite{bon87} constructed with these

380: inner and outer solutions is compared to the

381: numerical solution of (\ref{Fd}) in Fig.

382: \ref{fig4}.

383: 

384: Notice that (\ref{v0}) is the normal form

385: associated with a saddle-node bifurcation in a

386: one dimensional phase space. The wavefront

387: depinning transition is a {\em global} bifurcation

388: with generic features: each individual point

389: $u_n(t)$ spends a long time, which scales as

390: $|F-F_c|^{-{1\over 2}}$, near discrete values

391: $u_n(A,F_c)$, and then jumps to the next discrete

392: value on a time scale of order 1. The traveling

393: wave ceases to exist for

394: $F\leq F_c$. For these field values, discrete

395: stationary profiles $u_n(A,F)$ are found. The

396: above calculations give a normal form of the type

397: $d^2 v_0/dt^2 = \alpha\, (F-F_c) + \beta\, v_0^2$

398: instead of (\ref{v0}) for conservative discrete

399: systems (two time derivatives instead of one in

400: (\ref{Fd})). The solution of this equation blows

401: up in finite time as $(F-F_c)^{ -{1\over 4}}$,

402: which gives a critical exponent of 1/4 for the

403: wavefront velocity near the critical field.

404: 

405: The approximations to $F_c(A)$ and the wavefront

406: speed provided by the previous asymptotic theory

407: break down for small $A$. In particular, for the

408: FK potential and $A<2$, no double zeroes of $2x

409: +A\sin(x) - (F +U_{1} +U_3)$ are found for

410: $F=F_c$. What happens is that we need more than

411: one point to approximate wavefront motion.

412: Depinning is then described by a reduced system

413: of more than one degree of freedom corresponding

414: to active points. There is a saddle-node

415: bifurcation in this reduced system whose normal

416: form is of the same type as (\ref{v0}). The jump

417: of the active points after blow up is found by

418: solving the reduced system with a matching

419: condition \cite{else}. As we approach the

420: continuum limit, more and more points enter the

421: reduced system of equations and exponential

422: asymptotic methods become a viable alternative to

423: our methods.

424: 

425: 

426: In conclusion, we have studied depinning of

427: wavefronts in discrete RD equations. The normal

428: depinning transition can be viewed as a loss of

429: continuity of traveling front profiles as the

430: critical field is approached: below the critical

431: field, the fronts become pinned stationary

432: profiles with discontinuous jumps at discrete

433: values $u_n$. In the strongly discrete limit, the

434: critical field and these fronts can be

435: approximated by singular perturbation methods

436: which show excellent agreement with numerical

437: solutions. The leading order approximation to

438: the wavefront velocity is then correctly given

439: (scaling and prefactor) near the critical field.

440: Depinning transitions for discrete RD equations

441: apparently belong to two different universality

442: classes. In the normal class, the wavefront

443: velocity has a critical exponent 1/2. For certain

444: nonlinearities, the stationary fronts are

445: continuous functions of the discrete index at

446: zero field. Then the critical field is zero, the

447: depinning transition between stationary and

448: moving fronts is continuous, with a critical

449: exponent 1. This situation is the same as for

450: continuous RD equations and we have called it

451: anomalous pinning.

452: 

453: AC thanks S. Hastings and J.B. McLeod for fruifful

454: discussions. 

455: 

456: \begin{references}{}

457: 

458: \bibitem{kee87}

459: J. P. Keener, SIAM J. Appl. Math. {\bf 47}, 556

460: (1987).

461: 

462: \bibitem{kee98}

463: J.P. Keener and J. Sneyd, {\em Mathematical

464: Physiology} (Springer, New York, 1998). Chapter 9.

465: 

466: \bibitem{bug97}

467: A. E. Bugrim, A.M. Zhabotinsky and I.R. Epstein,

468: Biophys. J. {\bf 73}, 2897 (1997); J. Keizer,

469: G.D. Smith, S. Ponce Dawson and J.E. Pearson,

470: Biophys. J. {\bf 75}, 595 (1998).

471: 

472: \bibitem{cdw}

473: G. Gr\"uner, Rev. Mod. Phys. {\bf 60}, 1129

474: (1988); A.A. Middleton, Phys. Rev. Lett. {\bf

475: 68}, 670 (1992).

476: 

477: \bibitem{jj}

478: H. S. J. van der Zant, T. P. Orlando, S. Watanabe

479: and S. H. Strogatz, Phys. Rev. Lett. {\bf 75},

480: 174 (1995).

481: 

482: \bibitem{nab67}

483: F.R.N. Nabarro, {\em Theory of Crystal

484: Dislocations} (Oxford University Press, Oxford,

485: 1967).

486: 

487: \bibitem{cha95}

488: P. M. Chaikin and T. C. Lubensky, {\em Principles

489: of condensed matter physics} (Cambridge

490: University Press, Cambridge, 1995). Chapter 10.

491: 

492: \bibitem{diode}

493: M. L\"ocher, G.A. Johnson and E.R. Hunt, Phys.

494: Rev. Lett. {\bf 77}, 4698 (1996).

495: 

496: \bibitem{bon94}

497: L.L. Bonilla, J. Gal\'{a}n, J.A. Cuesta, F.C.

498: Mart\'{\i}nez and J. M. Molera, Phys. Rev. B {\bf

499: 50}, 8644 (1994); L. L. Bonilla, G. Platero and

500: D. S\'anchez, Phys. Rev. B {\bf 62}, 2786 (2000);

501: A. Wacker, in {\em Theory and transport

502: properties of semiconductor nanostructures},

503: edited by E.\ Sch\"oll (Chapman and Hall, New

504: York, 1998). Chapter 10.

505: 

506: \bibitem{car00}

507: A. Carpio, L. L. Bonilla, A. Wacker and E.

508: Sch\"oll, Phys. Rev. E {\bf 61}, 4866 (2000).

509: 

510: \bibitem{kla00}

511: K. Kladko, I. Mitkov and A.R. Bishop, Phys.\

512: Rev.\ Lett.\ {\bf 84}, 4505 (2000).

513: 

514: \bibitem{mit98}

515: I. Mitkov, K. Kladko and J.E. Pearson, Phys. Rev.

516: Lett. {\bf 81}, 5453 (1998).

517: 

518: \bibitem{kin01}

519: J.R. King and S.J. Chapman, Eur. J. Appl. Math.

520: (2001), to appear.

521: 

522: \bibitem{zin92}

523: B. Zinner, J. Diff. Eqs. {\bf 96}, 1 (1992);

524: A.-M.  Filip and S. Venakides, Comm. Pure Appl.

525: Math. {\bf 52}, 693 (1999).

526: 

527: \bibitem{car99}

528: A. Carpio, S.J. Chapman, S. Hastings, J.B.

529: McLeod, Eur.\ J.\ Appl.\ Math.\ {\bf 11}, 399

530: (2000).

531: 

532: \bibitem{FK}

533: J. Frenkel and T. Kontorova, Phys. Z. SowjUn.

534: {\bf 13}, 1 (1938).

535: 

536: \bibitem{comparison}

537: If we have $l_n(0)\leq u_n(0)$ such that

538: $\dot{u}_n \geq u_{n+1}-2u_n+u_{n-1} + F -A

539: g(u_n)$, and $\dot{l}_n \leq l_{n+1}-2l_n+l_{n-1}

540: + F -A g(l_n)$, then $l_n(t)\leq u_n(t)$ for all

541: later times. $l_n(t)$ and $u_n(t)$ are called sub

542: and supersolutions, respectively. See Ref.

543: \onlinecite{car99}.

544: 

545: \bibitem{hob65}

546: R. Hobart, J. Appl. Phys. {\bf 36}, 1948 (1965).

547: 

548: \bibitem{ind58}

549: V. L. Indenbom, Sov. Phys. -- Cryst. {\bf 3}, 193

550: (1958).

551: 

552: \bibitem{mcleod}

553: J. B. McLeod, private communication.

554: 

555: \bibitem{bon87}

556: L. L. Bonilla, J. Statist. Phys. {\bf 46}, 659

557: (1987).

558: 

559: \bibitem{else}

560: A. Carpio and L.L. Bonilla, unpublished.

561: 

562: \end{references}

563: 

564: \begin{figure}

565: \begin{center}

566: \includegraphics[width=8cm]{fig1.eps}

567: \caption{Traveling wavefront profiles near $F=

568: F_c$ for the FK potential and: (a) $A=2$, (b)

569: $A=100$.}

570: \label{fig1}

571: \end{center}

572: \end{figure}

573: 

574: 

575: \begin{figure}

576: \begin{center}

577: \includegraphics[width=8cm]{fig2.eps}

578: \caption{(a) $u_0(A,F_c)$; (b) Critical field as

579: a function of $A$; (c) Absolute and relative

580: errors in $F_c(A)$.}

581: \label{fig2}

582: \end{center}

583: \end{figure}

584: 

585: \begin{figure}

586: \begin{center}

587: \includegraphics[width=8cm]{fig3.eps}

588: \caption{Wavefront velocity versus $F$

589: near $F= F_c$ for the FK potential and: (a)

590: $A=100$, (b) $A=10$.}

591: \label{fig3}

592: \end{center}

593: \end{figure}

594: 

595: 

596: \begin{figure}

597: \begin{center}

598: \includegraphics[width=8cm]{fig4.eps}

599: \caption{Wavefront profiles near $F_c$ for the

600: FK potential and $A=100$. We show the three

601: largest jumps in Fig. \ref{fig1}(b). }

602: \label{fig4}

603: \end{center}

604: \end{figure}

605: 

606: \end{multicols}

607: \end{document}

608: