1: \errorcontextlines10

2: 

3: \documentclass[reqno]{amsart}

4: \usepackage{latexsym}

5: \usepackage{amstext}

6: \usepackage {amsmath}

7: \usepackage {amsfonts}

8: \usepackage {amssymb}

9: \usepackage {amsthm}

10: \usepackage {bbm}

11: \usepackage{enumerate}

12: \usepackage{array}

13: \usepackage{comment}

14: \usepackage{xspace}

15: \usepackage{graphicx}

16: \sloppy

17: %==================

18: % new commands

19: %==================

20: \def\cringle{\mathaccent"7017 }

21: \DeclareMathOperator{\dist}{dist}

22: \DeclareMathOperator{\graph}{graph}

23: \DeclareMathOperator{\sgn}{sgn}

24: \DeclareMathOperator{\spt}{supp}

25: \DeclareMathOperator\supp{supp}

26: \DeclareMathOperator{\aplim}{ap-lim}

27: \DeclareMathOperator{\dive}{div}

28: %\DeclareMathOperator{\loc}{loc}

29: \def\loc{{\mathrm{loc}}}

30: \DeclareMathOperator{\Lp}{L}

31: \DeclareMathOperator{\BV}{BV}

32: \DeclareMathOperator{\Lip}{Lip}

33: \DeclareMathOperator{\HS}{H}

34: \DeclareMathOperator{\p}{p}

35: \DeclareMathOperator{\diam}{diam}

36: \DeclareMathOperator{\modulo}{\, mod}

37: \DeclareMathOperator{\Int}{Int}

38: \DeclareMathOperator{\trace}{tr}

39: \DeclareMathOperator{\length}{length}

40: \DeclareMathOperator{\var}{var}

41: \DeclareMathOperator{\lin}{lin}

42: \newcommand{\HO}{\ensuremath{\cringle{\HS}\mbox{}}}

43: \newcommand{\eps}{\varepsilon}

44: \newcommand{\Chi}{\mathcal{X}}

45: \newcommand{\R}{\ensuremath{\mathbb{R}}}

46: \newcommand{\Rn}{\ensuremath{\mathbb{R}^n}}

47: \newcommand{\N}{\ensuremath{\mathbb{N}}\ }

48: \newcommand{\LL}{\ensuremath{\mathcal{L}}}

49: \newcommand{\K}{\ensuremath{\mathcal K}}

50: \newcommand{\sphere}{\ensuremath{\mathcal S}}

51: \newcommand{\Ha}{\ensuremath{\mathcal{H}}}

52: \newcommand{\eing}{\ensuremath{\lfloor}}

53: \newcommand{\schwto}{\ensuremath{\to}}

54: \newcommand{\schwstto}{\ensuremath{\overset{*}{\rightharpoonup}}}

55: \newcommand{\ddist}{\ensuremath{\mathsf{d}}}

56: \let\ds\displaystyle

57: \def\Mod#1{\left\|#1\right\|}

58: \def\mod#1{\left|#1\right|}

59: \def\vc#1{\ensuremath{\vcenter{\hbox{#1}}}}

60: \def\step#1{\\[2pt]\textsc{Step#1.}}

61: \newcommand{\schwsternto}{\ensuremath{\overset{*}{\rightharpoonup}}}

62: %\newcommand{}{}

63: %

64: % environments

65: %

66: \newcounter{counter-liste}

67: \newcounter{counter-liste2}

68: \newenvironment{liste2}{\ 

69: \begin{list}{{\arabic{counter-liste2}.}\hfill}{\usecounter{counter-liste2}

70: \setlength{\topsep}{0bp}

71: \setlength{\labelwidth}{10bp}

72: \setlength{\leftmargin}{12bp}%0.04\linewidth}

73: \setlength{\labelsep}{2bp}

74: \setlength{\itemindent}{0bp}%-0.2\leftmargin}

75: %\setlength{\itemsep}{3bp}%{3bp}

76: \setlength{\parsep}{0bp}}

77: }{\end{list}}

78: \newenvironment{liste}%

79:   {\ \begin{list}{{(\arabic{counter-liste})}\hfill}%

80:   {\topsep2mm\itemindent1ex\leftmargin0cm\usecounter{counter-liste}}

81:   }%

82:   {\end{list}}

83: \newenvironment{liste(a)}%

84:   {\ \begin{list}{{(\alph{counter-liste})}\hfill}%

85:   {\topsep2mm\itemindent1ex\leftmargin0cm\usecounter{counter-liste}}

86:   }%

87:   {\end{list}}

88: %

89: \def\R{\mathbb R}

90: \def\H{\mathcal H}

91: \def\W{\mathcal W}

92: \let\e\varepsilon

93: \def\mod#1{\left|#1\right|}

94: \def\BV{\mathrm{BV}}

95: \let\ds\displaystyle

96: \def\pref#1{(\ref{#1})}

97: \theoremstyle{plain}

98: \numberwithin{equation}{section}

99: \newtheorem{lemma}{Lemma}[section]

100: \newtheorem{theorem}[lemma]{Theorem}

101: \newtheorem{proposition}[lemma]{Proposition}

102: \newtheorem{definition}[lemma]{Definition}

103: \newtheorem{assumption}[lemma]{Assumption}

104: \newtheorem{corollary}[lemma]{Corollary}

105: \newtheorem{claim}[lemma]{Claim}

106: \theoremstyle{definition}

107: \newtheorem{remark}[lemma]{Remark}

108: %

109: 

110: %=====================================================================================================

111: % main

112: %=====================================================================================================

113: \begin{document}

114: 

115: \title[Travelling waves in a tissue-degradation model]{Travelling-wave analysis

116:   of a model describing tissue degradation by bacteria} 

117: \thanks{

118: The research of the authors was supported by

119: the European Community's

120: Human Potential Programme

121: under contract HPRN-CT-2002-00274,

122: FRONTS-SINGULARITIES.

123: }

124: 

125: \author[Hilhorst]{Danielle Hilhorst}

126: \address{Danielle Hilhorst, CNRS and Laboratoire de Math{\'e}matiques,

127:   Universit{\'e} de Paris-Sud, 91405 Orsay Cedex, France}

128: 

129: \author[King]{John R. King}

130: \address{John R. King, Centre for Mathematical Medicine, Theoretical Mechanics Section,

131: School of Mathematical \mbox{Sciences},

132: University of Nottingham,

133: Nottingham, NG7 2RD, UK}

134: 

135: \author[R\"{o}ger]{Matthias R\"{o}ger}

136: \address{Matthias R\"{o}ger, Max Planck Institute for Mathematics in the Sciences,

137: Inselstr. 22,

138: D-04103 Leipzig}

139: 

140: \subjclass[2000]{Primary 35K57; Secondary  92E20, 35B25, 80A22}

141: 

142: \keywords{Travelling waves, Reaction-Diffusion system, singular limit,

143:   Stefan problem, pulled fronts, pushed fronts}

144: 

145: \date{\today}

146: 

147: \begin{abstract}

148: We study travelling-wave solutions for a reaction-diffusion

149: system arising as a model for host-tissue degradation by bacteria.

150: This system consists of a parabolic equation coupled with an

151: ordinary differential equation.

152: For large values of the `degradation-rate parameter' solutions are

153: well approximated by solutions of a Stefan-like free boundary

154: problem, for which travelling-wave solutions can be found explicitly.

155: Our aim is to prove the existence of travelling waves

156: for all sufficiently large wave-speeds for the original reaction-diffusion system

157: and to determine the minimal speed.

158: We prove that for all sufficiently large degradation rates the minimal speed

159: is identical to

160: the minimal speed of the limit problem. In particular, in this parameter range,

161: {\itshape nonlinear} selection of the minimal speed occurs.

162: \end{abstract}

163: \maketitle

164: \tableofcontents

165: %=============================================

166: % introduction

167: %=============================================

168: \section{Introduction}

169: In this article we continue our mathematical analysis of a model

170: for the degradation of host tissue by extracellular bacteria.

171: This model was introduced in \cite{Ki} and consists of a

172: reaction-diffusion equation coupled with an ordinary differential equation.

173: In \cite{HKR} we proved the existence

174: of solutions to the time-dependent problem and the convergence to a

175: limit problem in the  

176: `large-degradation-rate' limit.

177: Here we turn to the question of existence

178: and behaviour of travelling-wave solutions.

179: 

180: There is an increasing interest in models which support the understanding of bacterial

181: infections, and we refer to \cite{Ki} and \cite{Wa} for further background and references.

182: This paper is in effect concerned with the specific issue of how rapidly

183: a bacterial infection in, for example, a burn wound may invade

184: the underlying

185: tissue (with dire potential consequences for the patient, notably mortality due

186: to septicemia). For the type of model with which we are concerned

187: here, the relevant invasion speed is expected to be governed by the corresponding

188: travelling-wave problem. Typically, the smallest possible wave-speed is

189: realized by a large class of solutions. Accordingly determining the

190: minimal speed of travelling waves becomes a central question

191: (with obvious implications

192: for the amount of time available for medical treatment, for instance).

193: 

194: In a dimensionless form, the model in \cite{Ki} is given by the equations

195: \begin{eqnarray}

196:         \partial_t {u} &=& \Delta{u} - {u} + {w} -\gamma k {u}(1-{w}),

197:                 \label{rd-u}\\

198:         \partial_t {w} &=& k {u}(1-{w}),\label{rd-w}

199: \end{eqnarray}

200: where $u$ describes the concentration of degradative enzymes, $(1-w)$

201: the volume fraction of healthy tissue and $\gamma,k$ are positive constants.

202: The key parameter here is the \emph{ degradation-rate} $k$, which is  very large

203: in practice.

204: Equations \eqref{rd-u}, \eqref{rd-w} are considered in a time-space cylinder,

205: with the upper half space of $\R^3$ as the spatial domain.

206: Finally, the system is complemented by initial conditions for $u$ and $w$,

207: a Neumann condition on the lateral boundary for $u$ and a decay condition

208: for $u$ and $w$ in the far field. In \cite{HKR} we gave a precise mathematical

209: formulation and proved the existence and uniqueness of solutions to a slightly more general

210: system, including the possibility of a diffusion term in \eqref{rd-w}.

211: One noteworthy aspect of \eqref{rd-u}, \eqref{rd-w} is the convergence of solutions

212: to the solution of a Stefan-like free boundary problem as the degradation rate $k$

213: tends to infinity. This \emph{large-degradation-rate limit} was identified

214: by a formal asymptotic analysis in \cite{Ki} and was proved in \cite{HKR}.

215: 

216: Reaction-diffusion systems of the general form

217: \begin{eqnarray}\label{eq-rd-gen}

218:         V_t &=& D\Delta V + F(V),

219: \end{eqnarray}

220: where $V:(0,T)\times\R^n\to \R^N$ is vector-valued, $F:\R^N\to\R^N$ is a given nonlinearity

221: and $D\in\R^{N\times N}$ is a diagonal positive-semi-definite matrix,

222: appear in a lot of different scientific areas.

223: One-dimensional travelling waves are solutions on $(0,\infty)\times\R$ of the special form

224: \begin{eqnarray*}

225:   \tilde{V}(t,x) &=& V(x-ct),

226: \end{eqnarray*}

227: where $c$ is called the speed and $V$ the profile of this travelling wave.

228: The question of existence and behaviour of travelling waves is of

229: enormous interest in many of the applications and pertinent results for

230: the vector case $N\geq 2$ remain restricted to rather specific systems.

231: 

232: The system \eqref{rd-u}, \eqref{rd-w} has, as we will see in Remark

233: \ref{rem:mono}, one stable equilibrium in $(u,w)=(1,1)$ and one

234: unstable equilibrium in $(u,w)=(0,0)$ and therefore belongs to the class of

235: monostable systems. Scalar monostable equations

236: \begin{eqnarray}\label{eq-Fi}

237:   v_t &=& \Delta v + f(v)

238: \end{eqnarray}

239: where

240: \begin{gather*}

241:   f(0)\,=\,f(1)\,=\,0,\quad f(v)>0\text{ for }0<v<1,\quad f^\prime(0)>0>f^\prime(1),

242: \end{gather*}

243: are well-studied, especially the famous Fisher equation, that is \eqref{eq-Fi} with

244: $f(v)=v(1-v)$, introduced in \cite{Fi}.

245: The rigorous analysis of equations of this type also started in the 1930s with the work of

246: Kolmogorov, Petrovskii and Piskunov \cite{KPP}. Under an extra assumption on $f$ they

247: proved the existence of travelling waves for all speeds $c\geq c_{\lin}$, where

248: $c_{\lin}$ can be found explicitly in terms of $f$ by a linearisation about $v=0$

249: (corresponding to a degenerate node in the travelling-wave phase plane).

250: Moreover, they proved that the solutions to \eqref{eq-Fi} with initial data decaying sufficiently fast

251: propagate with speed $c_{\lin}$.

252: For more general monostable $f$ the propagation speed was found to be either equal to

253: or larger than $c_{\lin}$ and therefore one distinguishes between a {\itshape linear}

254: or {\itshape nonlinear} selection of the propagation speed (the terminology

255: {\itshape pulled} and {\itshape pushed} fronts, respectively, having an equivalent

256: meaning).

257: Aronson and Weinberger \cite{AW} (see also Hadeler and Rothe \cite{HR} and \cite{Sto}

258: for other pioneering work on such matters and \cite{Saa-rev} for a recent review)

259: proved that for general monostable $f$, in both

260: the linear and the nonlinear selection case, the propagation speed for

261: solutions with initial data decaying sufficiently fast is given by

262: the minimal speed of travelling waves. They showed that monotonic travelling waves exist

263: for all speeds $c\geq c_{\min}$ and none for $c<c_{\min}$, where $c_{\min}\geq c_{\lin}$;

264: solutions of \eqref{eq-Fi} with sufficiently rapidly

265: decaying initial data propagate with speed $c_{\min}$.

266: For the nonlinear selection cases, $c_{\min}>c_{\lin}$, Rothe \cite{Rot} and Roquejoffre \cite{Roq}

267: proved that, if the initial data decays sufficiently rapidly, the large-time solutions to

268: \eqref{eq-Fi} not

269: only propagate with speed $c_{\min}$ but also approach the profile of a travelling wave

270: with minimal speed.

271: 

272: Whereas the connection between large-time behaviour and existence of travelling waves

273: for monostable equations is satisfactorily resolved, the distinction between nonlinear

274: or linear selection is still a challenging question, see for example

275: \cite{Saa}, \cite{ES}, \cite{BD}.

276: Only a few rigorous results for general monostable nonlinearities are available.

277: In \cite{LMN} a variational characterisation of travelling waves and

278: a concrete criterion for whether linear or nonlinear selection occurs for a given equation

279: was derived.

280: Even fewer general analytical results are available for monostable {\itshape systems}. An

281: existence theorem for travelling waves was proved in \cite{Vo} for monotone monostable systems,

282: which are systems of the form  \eqref{eq-rd-gen} in which the Jacobian matrix

283: $DF$ has only nonnegative off-diagonal elements.

284: Results on the existence of travelling waves and the long time behaviour of \eqref{eq-rd-gen}

285: for monostable gradient systems, that is for $n=N$ and nonlinearities $F=\nabla g$ with

286: $g:\R^n\to\R$, were given in \cite{Mur}.

287: To the best of our knowledge there are no more general results on the question of whether

288: linear or nonlinear selection will occur.

289: 

290: In this article we prove that for all $\gamma,k$ there exist

291: monotone travelling waves for the system \eqref{rd-u}, \eqref{rd-w}

292: for all speeds $c\geq c_{\min}$ and no speeds $c< c_{\min}$.

293: The minimal speed $c_{\min}$ in general depends on the

294: parameters $\gamma,k$.

295: 

296: We prove that for all $k>k_0$, where $k_0$ is explicitly given

297: in terms of $\gamma$, the minimal speed is larger than the value $c_{\lin}(k)$ obtained from

298: a linearisation at the unstable equilibrium. Surprisingly enough, for $k>k_0$ the minimal

299: speed of travelling waves is identical to the minimal speed of travelling waves for the

300: Stefan-like limit problem that was formulated in \cite{HKR}.

301: Our analysis is based on two main facts. One is the monotone structure of the

302: system, which makes possible the use of comparison principles for the parabolic problem.

303: The second is a remarkable reduction in order of the travelling-wave equations

304: that occurs when $c$ is given by the minimal speed $c_\infty$ of travelling waves of

305: the large-degradation limit.

306: We obtain the existence of travelling waves with speed $c_\infty$ for \eqref{rd-u}, \eqref{rd-w}.

307: Finally, a comparison argument allows us to prove that nonlinear selection occurs

308: for sufficiently large values of $k$, the minimal speed in this regime being identical to $c_\infty$.

309: %

310: 

311: This paper is organised as follows. In section \ref{sec-exist} we prove the existence of travelling

312: waves for the reaction-diffusion system \eqref{rd-u}, \eqref{rd-w}. In section \ref{sec-limit}

313: we recall the formulation of the large degradation limit problem

314: and consider travelling-wave solutions for this problem.

315: In section \ref{sec-speed} we return to the reaction-diffusion system and investigate

316: the selection of the minimal speed. Section \ref{sec-conv} deals with the convergence

317: of travelling waves for the reaction-diffusion system to travelling waves of the Stefan-like

318: free boundary problem as the reaction rate $k$ approaches infinity.

319: Finally, we give some conclusions and remarks on open problems in section \ref{sec-concl}.

320: %=====================================================

321: %       section existence

322: %=====================================================

323: \section{Existence of monotone travelling waves}

324: \label{sec-exist}

325: In this section we prove the existence

326: of monotone travelling waves. First we fix some notation and make some

327: remarks.

328: \begin{remark}\label{rem:mono}

329: A system of the general form

330: \eqref{eq-rd-gen} is called \emph{monotone} if the off-diagonal elements

331: of the Jacobian matrix $DF$ are

332: non-negative and \emph{strictly monotone} if they are positive (see \cite{Vo}).

333: A system of the form \eqref{eq-rd-gen} with two stationary points

334: is called \emph{monostable} if one of the stationary points is stable

335: and the other is unstable.\\

336: The system \eqref{rd-u}, \eqref{rd-w} is of the form \eqref{eq-rd-gen}

337: with

338: \begin{gather*}

339:   V\,=\,

340:   \begin{pmatrix}

341:     u\\

342:     w

343:   \end{pmatrix}\!,\qquad

344:   F(z_1,z_2)\,=\, 

345:   \begin{pmatrix}

346:     -z_1 +z_2 -\gamma kz_1(1-z_2)\\

347:     kz_1(1-z_2)

348:   \end{pmatrix}\!.

349: \end{gather*}

350: It follows that \eqref{rd-u}, \eqref{rd-w} is a monotone but not strictly monotone system.

351: Further, $z=(0,0)$ and $z=(1,1)$ are the only stationary points of \eqref{rd-u}, \eqref{rd-w}

352: and we obtain

353: \begin{gather*}

354:   DF(0,0)\,=\,

355:   \begin{pmatrix}

356:     -(1+\gamma k) & 1\\

357:     k & 0

358:   \end{pmatrix}

359: \end{gather*}

360: with one positive and one negative eigenvalue, and

361: \begin{gather*}

362:   DF(1,1)\,=\, 

363:   \begin{pmatrix}

364:     -1 & 1+\gamma k\\

365:     0 & -k

366:   \end{pmatrix}

367: \end{gather*}

368: with two negative eigenvalues. Therefore \eqref{rd-u}, \eqref{rd-w}

369: is a monostable monotone system.

370: \end{remark}

371: As remarked before, for a one-dimensional travelling wave $u,w$ of \eqref{rd-u},

372: \eqref{rd-w} with speed $c$, the functions

373: \begin{gather*}

374:         \tilde{u}(t,x) \,:=\, u(x-ct),\quad

375:         \tilde{w}(t,x) \,:=\, w(x-ct),

376: \end{gather*}

377: are solutions of \eqref{rd-u}, \eqref{rd-w} on $(0,\infty)\times\R$.

378: Therefore $u,w$ have to satisfy the \emph{travelling-wave equations}

379: \begin{eqnarray}

380:         0 &=& u^{\prime\prime} + cu^\prime -u + w -\gamma k u(1-w),\label{eq-tw-u}\\

381:         0 &=& cw^\prime +k u(1-w),\label{eq-tw-w}

382: \end{eqnarray}

383: We restrict our investigations to functions $u,w$ taking values only in $[0,1]$,

384: which is the range of

385: meaningful values in the tissue degradation model. % Observe that $(u,w)=(0,0)$ and $(u,w)=(1,1)$

386: %are the only stationary solutions of \eqref{eq-tw-u}, \eqref{eq-tw-w}.

387: We first summarise some basic properties of travelling-wave solutions.

388: \begin{lemma}\label{lem-tw-basic}

389: Assume that $u\in C^2(\R), w\in C^1(\R)$ with $0\leq u,w\leq 1$ satisfy

390: \eqref{eq-tw-u}, \eqref{eq-tw-w} with $c\in\R$.

391: Then the following properties hold.

392: \begin{enumerate}[(1)\hspace{-3pt}]

393: \item Invariance under space shifts. For any $x_0\in\R$ the functions defined by

394: $x\mapsto u(x+x_0)$, $x\mapsto w(x+x_0)$

395: solve \eqref{eq-tw-u}, \eqref{eq-tw-w} with the same $c\in\R$.

396: \item Invariance under inversion. The functions defined by $x\mapsto u(-x)$, $x\mapsto w(-x)$

397: solve \eqref{eq-tw-u}, \eqref{eq-tw-w} with $c$ replaced by $-c$.

398: \item Smoothness of travelling waves. The functions $u,w$ are infinitely differentiable.

399: \item Monotonicity of travelling waves. If $c>0$ and $u,w$ are not both constant

400: then $u,w$ are strictly monotone decreasing

401: and they approach unity as $x\to -\infty$ and zero

402: as $x\to\infty$.

403: \item Standing waves. For $c=0$ either $(u,w)=(0,0)$ or $(u,w)=(1,1)$.

404: \end{enumerate}

405: \end{lemma}

406: \begin{proof}

407: The statements (1), (2) are immediate from \eqref{eq-tw-u},

408: \eqref{eq-tw-w}. Statement (3) follows from a bootstrapping argument.\\

409: To prove (4) we observe from \eqref{eq-tw-w} that

410: $w^\prime\leq 0$ and $w^\prime <0$ if $u(1-w)>0$.

411: Next we differentiate \eqref{eq-tw-u} and obtain

412: \begin{eqnarray}

413:  0 &=& u^{\prime\prime\prime} + cu^{\prime\prime} -\big(1+\gamma k (1- w)\big)u^\prime

414:  +(1+\gamma ku) w^\prime.\label{eq:diff-u-eq}

415: \end{eqnarray}

416: From this equation we see that $u^\prime$ cannot have a positive local maximum.

417: Since $u$ is bounded,

418: $u^\prime$ cannot approach a positive supremum at $+\infty$ or $-\infty$

419: and we obtain that $u^\prime\leq 0$.

420: 

421: Assume now that $u^\prime(x_0)=0$. Therefore $u^\prime(x_0)$ is a local

422: maximum of $u^\prime$ and we deduce from \eqref{eq:diff-u-eq} that

423: $w^\prime(x_0)\geq 0$ and that, using $w^\prime\leq 0$ and

424: \eqref{eq-tw-w},

425: \begin{gather*}

426:   -\frac{k}{c} u(1-w)(x_0) \,=\, w^\prime(x_0)\,=\,0.

427: \end{gather*}

428: Let us consider the case  $u(x_0)=0$

429: (the case that $w$ takes the value one is analogous). Then, by $u^\prime\leq 0$ and

430: $u\geq 0$, we obtain

431: $u=0$ on $[x_0,\infty)$. From \eqref{eq-tw-u} it also follows that $w=0$ on $[x_0,\infty)$.

432: Since $u,w$ are smooth we find that $u,w$ solve \eqref{eq-tw-u}, \eqref{eq-tw-w}

433: with $u(x_0)=u^\prime(x_0)=w(x_0)=0$.

434: On the other hand solutions of \eqref{eq-tw-u}, \eqref{eq-tw-w} with these data prescribed

435: at $x_0$ are unique and thus identically equal to zero, in

436: contradiction to our assumption of non-constant solutions. Therefore $u(1-w)>0$ and $u^\prime,w^\prime<0$

437: holds.\\

438: Next, since $u(x),w(x)$ are monotone in $x$ and uniformly bounded, their limits as $x\to -\infty$ exist.

439: From \eqref{eq-tw-w} we deduce that $w^\prime(x)$ has a limit and, from $0\leq w\leq 1$, that

440: %\mbox{$w^\prime(x_j)\to 0$} for a subsequence $x_j\to-\infty$, which ensures

441: \begin{gather}\label{limit-u(1-w)}

442:         w^\prime(x) \,\to\, 0,\quad

443:         u(x)(1-w(x))\,\to\, 0\quad \text{ as }x\to-\infty.

444: \end{gather}

445: From \eqref{eq-tw-u} we similarly obtain that

446: \begin{eqnarray*}

447:         (u^\prime+cu)^\prime(x) &\to& 0\quad \text{ as }x\to-\infty,

448: \end{eqnarray*}

449: and thus

450: \begin{eqnarray*}

451:         -u(x)+w(x) &\to& 0 \quad\text{ as }x\to-\infty.

452: \end{eqnarray*}

453: By $0\leq u,w\leq 1$, \eqref{limit-u(1-w)}, and $u^\prime, w^\prime\leq 0$ this gives

454: \begin{gather*}

455:         u(x) \,\to\,1,\quad w(x) \,\to\, 1 \qquad\text{ as }x\to-\infty.

456: \end{gather*}

457: The proof that $u(x),w(x)$ approach zero as $x\to +\infty$ is similar.

458: %We now show that $u^\prime$ tends to zero as $x\to -\infty$. Observe that this function is bounded due to

459: %\eqref{bounds-u,w_x-infty} and assume the existence of two different limit points.

460: %Then we find a value $d<0$ in between and

461: %a subsequence $x_j\to -\infty $ such that $u^\prime(x_j)=d, u^{\prime\prime}(x_j)\leq 0$,

462: %in contradiction to \eqref{limit-u-prime+}. Therefore

463: %\begin{eqnarray*}

464: %       u^\prime(x) &\to& 0\quad\text{ as }x\to-\infty.

465: %\end{eqnarray*}

466: 

467: 

468: To prove claim (5) we first deduce from $c=0$ and \eqref{eq-tw-w} that

469: \begin{gather}

470:   u(1-w)\,=\, 0. \label{eq:u(1-w)}

471: \end{gather}

472: Let us assume that $u(x_0)>0$ for a point $x_0\in \R$. Then there exists

473: a maximal interval 

474: $(a,b)$ such that $x_0\in (a,b)$ and $u>0$ in $(a,b)$. From

475: \eqref{eq-tw-u}, \eqref{eq:u(1-w)}, and $c=0$ we then  obtain that 

476: \begin{gather}

477:   u\,>\, 0,\, w\,=\,1,\, u^{\prime\prime}\,<\, 0\qquad\text{ in }(a,b).

478:   \label{eq:prop-ab} 

479: \end{gather}

480: If $a>-\infty$ we deduce from the regularity assumptions on $u,w$ and

481: $u\geq 0$ that

482: \begin{gather*}

483:   u(a)\,=\,0,\, w(a)\,=\,1,\, u^\prime(a)=0

484: \end{gather*}

485: and finally $u^{\prime\prime}(a)\geq 0$, which yields a contradiction to

486: \eqref{eq-tw-u}. This shows that $a=-\infty$; by analogous arguments we

487: obtain that $b=\infty$ and $(a,b)=\R$. Therefore

488: \begin{gather*}

489:   0\,=\,u^{\prime\prime} -u+1\quad\text{ on }\R.

490: \end{gather*}

491: and from the boundedness of $u$ it follows that $u=1=w$.

492: 

493: Similarly we prove that $w=0=u$ if there exists $x_0\in\R$ such that

494: $w(x_0)<1$.

495: \end{proof}

496: Due to Lemma \ref{lem-tw-basic} we can restrict our investigations to

497: the following set of {\itshape admissible functions}.

498: \begin{definition}\label{def-admiss}

499: Let $\K$ be the set of functions $v\in C^\infty(\R)$, with

500: \begin{gather}

501:         0\,<\,v\,<\,1,\quad

502:         v^\prime\,<\,0,\label{K-1}\\

503:         v(x)\,\to\, 1,\quad \text{ as } x\to -\infty,\label{K-2}\\

504:         v(x)\,\to\, 0,\quad \text{ as } x\to +\infty.\label{K-3}

505: \end{gather}

506: We call $(c,u,w)\in \R\times\K\times\K$ satisfying (\ref{eq-tw-u}), (\ref{eq-tw-w})

507: a monotone (decreasing) travelling wave for \eqref{rd-u}, \eqref{rd-w}.\\

508: %The minimal speed of travelling waves is defined as

509: %\begin{eqnarray}\label{def-c0}

510: %       c_{\min} &:=& \inf\{c\in \R\,:\, \text{ there exists a monotone TW }(c,u,w)\}.

511: %\end{eqnarray}

512: \end{definition}

513: %The value of $c_{\min}=c_{\min}(k)$ in general depends on the parameter $k$ (and $\gamma$ as well).

514: 

515: We observe from \eqref{eq-tw-w} that the speed of a monotone (decreasing)

516: travelling

517: wave is always positive. 

518: \begin{theorem}[Existence of travelling waves]\label{the-exist}

519: For all $\gamma,k$ there is a positive number $c_{\min}<\infty$,

520: $c_{\min}=c_{\min}(\gamma,k)$, such 

521: that there 

522: exists a monotone travelling wave of \eqref{rd-u}, \eqref{rd-w} with speed $c$

523: for 

524: all $c\geq c_{\min}$ and such that there is no monotone travelling wave

525: with speed $c< c_{\min}$.

526: \end{theorem}

527: The value $c_{\min}$ gives thus the \emph{minimal speed} of travelling waves for

528: \eqref{rd-u}, \eqref{rd-w}.

529: 

530: In the remainder of this section we prove Theorem \ref{the-exist}.

531: 

532: For $0<\eps<(2\gamma)^{-1}$ we consider the

533: following strictly monotone and strictly parabolic approximation of

534: \eqref{rd-u}, \eqref{rd-w},

535: \begin{eqnarray}

536:   \partial_t \tilde{u}_\eps &=&

537:   \partial_{xx}\tilde{u}_\eps -\tilde{u}_\eps +\tilde{w}_\eps -\gamma k\tilde{u}_\eps(1-\tilde{w}_\eps),\label{rd-u-eps}\\

538:   \partial_t \tilde{w}_\eps &=&

539:   \eps\partial_{xx}\tilde{w}_\eps +\eps(\tilde{u}_\eps-\tilde{w}_\eps)+k\tilde{u}_\eps(1-\tilde{w}_\eps),\label{rd-w-eps}

540: \end{eqnarray}

541: and the corresponding travelling-wave equations

542: \begin{eqnarray}

543:   0 &=& u_\eps^{\prime\prime} +cu_\eps^\prime-u_\eps+w_\eps -\gamma k u_\eps(1-w_\eps)\label{tw-u-eps},\\

544:   0 &=& \eps w_\eps^{\prime\prime} +cw_\eps^\prime +\eps(u_\eps-w_\eps) +ku_\eps(1-w_\eps).\label{tw-w-eps}

545: \end{eqnarray}

546: Observe that the ODE system \eqref{tw-u-eps}, \eqref{tw-w-eps} has the same stationary points

547: as the original system \eqref{eq-tw-u}, \eqref{eq-tw-w}. For $\eps=0$ the systems \eqref{tw-u-eps}, \eqref{tw-w-eps}

548: and \eqref{eq-tw-u}, \eqref{eq-tw-w} coincide. The existence

549: of travelling waves for the auxiliary problem and a variational 

550: characterisation of the minimal speed follows from \cite{Vo}.

551: To state their result, we define functionals \mbox{$\Phi_1^\eps,\Phi_2^\eps: \K\times\K\to\R$},

552: \begin{eqnarray*}

553:   \Phi_1^\eps(\varrho,\sigma) &:=&\sup_{x\in\R}

554:   \frac{\varrho^{\prime\prime}(x)-\varrho(x) +\sigma(x)

555:         -\gamma k\varrho(x)(1-\sigma(x))}{-\varrho^\prime(x)},\\

556:   \Phi_2^\eps(\varrho,\sigma) &:=&

557:         \sup_{x\in\R}\,\frac{\eps\sigma^{\prime\prime}(x) +\eps(\varrho-\sigma)(x)+k\varrho(x)(1-\sigma(x))}

558:             {-\sigma^\prime(x)}.

559: \end{eqnarray*}

560: \begin{lemma}[\cite{Vo} Theorem I.4.2]\label{lem:volpert}

561: For all $c\geq c_\eps$, where $c_\eps\geq 0$ is defined by

562: \begin{eqnarray}

563:   c_\eps &:=& \inf_{\sigma,\varrho\in\K}

564:   \max\Big(\Phi_1^\eps(\varrho,\sigma),\Phi_2^\eps(\varrho,\sigma)\Big),

565:   \label{eq:def-c-eps} 

566: \end{eqnarray}

567: there exists a monotone travelling wave $(c,u_\eps,w_\eps)$ and $c_\eps$ is the minimal speed

568: of travelling waves for \eqref{rd-u-eps}, \eqref{rd-w-eps}. \hfill$\Box$

569: \end{lemma}

570: To prove the existence of travelling waves for \eqref{rd-u}, \eqref{rd-w} we pass in

571: \eqref{tw-u-eps}, \eqref{tw-w-eps} to the limit $\eps\to 0$.

572: First we derive bounds for $u_\eps,w_\eps$ which are uniform in $\eps>0$.

573: \begin{lemma}

574: Consider $\eps>0$ and let

575: $(c,u_\eps,w_\eps)$ be a monotone travelling wave for \eqref{rd-u-eps}, \eqref{rd-w-eps}.

576: Define $\mu,\nu>0$ to be

577: the positive solutions of

578: \begin{eqnarray}

579:   \mu^2+c\mu-1 &=& 0,\label{eq:def-mu}\\

580:   \nu^2-c\nu-(1+\gamma k) &=& 0.\label{eq:def-nu}

581: \end{eqnarray}

582: Then

583: \begin{eqnarray}

584:   \label{eq:bound-u-eps-prime}

585:   u_\eps^\prime &\geq& -\mu(1-u_\eps),\\

586:   u_\eps^\prime &\geq& -\nu u_\eps,

587:     \label{eq:bound-u-eps-prime+}\\

588:   |u_\eps^{\prime\prime}| &\leq& c\mu+1+\gamma k,\label{eq:bound-u-eps-prime2}\\

589:   \label{eq:bound-w-eps-prime}

590:   w_\eps^\prime &\geq& -\frac{\eps+k}{c}(1-w_\eps)

591: \end{eqnarray}

592: hold.

593: \end{lemma}

594: % In particular, for $c>\delta$ and any $l\in \N_0$ we obtain

595: %\begin{eqnarray}\label{eq:bounds-u,w-prime}

596: %  \|u_\eps\|_{C_l(\R)}, \|w_\eps\|_{C^l(\R)} &\leq& C(\gamma,k,\delta).

597: %\end{eqnarray}

598: \begin{proof}

599: We fix $x\in\R$ and define for $y\in (-\infty,x]$

600: \begin{eqnarray*}

601:   \hat{u}(y) &:=& 1-(1-u_\eps(x))e^{\mu (y-x)},\\

602:   \hat{w}(y) &:=& 1-(1-w_\eps(x))e^{L (y-x)},\quad\text{ where } L:=\frac{k+\eps}{c}.

603: \end{eqnarray*}

604: Then we observe that

605: \begin{gather*}

606:   \lim_{y\to -\infty}\hat{u}(y)\,=\,1,\quad \hat{u}(x)=u_\eps(x),\\

607:   \lim_{y\to -\infty}\hat{w}(y)\,=\,1,\quad \hat{w}(x)=w_\eps(x).

608: \end{gather*}

609: Moreover

610: \begin{gather*}

611:   \hat{u}^{\prime\prime} +c\hat{u}^\prime-\hat{u}+w_\eps-\gamma k\hat{u}(1-w_\eps)\\

612:   =\, -(1-u_\eps(x))\big(\mu^2+c\mu -1\big)e^{\mu (y-x)} -(1-w_\eps)-\gamma k\hat{u}(1-w_\eps)

613:   \,<\,0

614: \end{gather*}

615: and

616: \begin{gather*}

617:   \eps\hat{w}^{\prime\prime} +c\hat{w}^\prime-\eps\hat{w}+\eps u_\eps +ku_\eps(1-\hat{w})\\

618:   =\, -(1-w_\eps(x))\big(\eps L^2+cL-(\eps+k u_\eps)\big)e^{L (y-x)} -\eps(1-u_\eps)(x)\,<\, 0.

619: \end{gather*}

620: By the maximum principle, which we apply once for the scalar equation \eqref{tw-u-eps} and

621: once for the scalar equation \eqref{tw-w-eps}, we deduce that

622: \begin{gather*}

623:   u_\eps\,\leq\, \hat{u},\quad w_\eps\,\leq\, \hat{w}\qquad\text{ on }(-\infty,x],\label{eq:dom--u-eps}

624: \end{gather*}

625: thus

626: \begin{gather*}

627:   u_\eps^\prime(x)\,\geq \hat{u}^\prime(x)=-\mu\big(1-u_\eps(x)\big),\\

628:   w_\eps^\prime(x)\,\geq \hat{w}^\prime(x)=-L\big(1-w_\eps(x)\big),

629: \end{gather*}

630: which proves \eqref{eq:bound-u-eps-prime}, \eqref{eq:bound-w-eps-prime}.

631: The estimate \eqref{eq:bound-u-eps-prime2} follows from \eqref{tw-u-eps} and \eqref{eq:bound-u-eps-prime}.\\

632: In addition, by comparing $u_\eps$ on $[x,\infty)$ with

633: \begin{eqnarray*}

634:   \check{u}(y) &:=& u_\eps(x)e^{-\nu (y-x)}

635: \end{eqnarray*}

636: we obtain

637: \begin{eqnarray}

638:   \check{u}&\leq& u_\eps\quad\text{ on }[x,\infty),\label{eq:dom+-u-eps}\\

639:     \check{u}^\prime(x) &\leq& u_\eps^\prime(x),\notag

640: \end{eqnarray} 

641: which yields \eqref{eq:bound-u-eps-prime+}.

642: \end{proof}

643: %We now define%choose now a subsequence $\eps_i\to 0\,(i\to\infty)$ such that

644: %\begin{eqnarray}\label{eq:c-i}

645: %  c_0 &:=& \liminf_{\eps>0} c_\eps

646: %\end{eqnarray}

647: %and prove the existence of travelling waves for all speeds $c\geq c_0$.

648: By similar arguments one proves that corresponding properties

649: hold for travelling-wave solutions of \eqref{rd-u}, 

650: \eqref{rd-w}.

651: \begin{lemma}

652: Let $(c,u,w)$ be a monotone travelling wave for \eqref{rd-u}, \eqref{rd-w}.

653: Define $\mu,\nu>0$ to be

654: the positive solutions of

655: \begin{eqnarray}

656:   \mu^2+c\mu-1 &=& 0,\label{eq:def-mu-0}\\

657:   \nu^2-c\nu-(1+\gamma k) &=& 0.\label{eq:def-nu-0}

658: \end{eqnarray}

659: Then

660: \begin{eqnarray}

661:   \label{eq:bound-u-eps-prime-0}

662:   u^\prime &\geq& -\mu(1-u),\\

663:   u^\prime &\geq& -\nu u,

664:     \label{eq:bound-u-eps-prime+-0}\\

665:   |u^{\prime\prime}| &\leq& c\mu+1+\gamma

666:    k,\label{eq:bound-u-eps-prime2-0}\\ 

667:   \label{eq:bound-w-eps-prime-0}

668:   w^\prime &\geq& -\frac{k}{c}(1-w)

669: \end{eqnarray}

670: hold. \hfill$\Box$

671: \end{lemma}

672: We prove now the first statement in Theorem \ref{the-exist}.

673: \begin{proposition}\label{prop:ex-tw-c0}

674: For each $c\geq c_0$, where

675: \begin{eqnarray}\label{eq:c-i}

676:   c_0 &:= & \liminf_{\eps\to 0} c_\eps,

677: \end{eqnarray}

678: there exists a monotone travelling wave $(c,u,w)$ for \eqref{rd-u}, \eqref{rd-w}.

679: Moreover, the value $c_0$ is finite.

680: % Moreover,

681: %with $\mu,\nu$ as in \eqref{eq:def-mu}, \eqref{eq:def-nu}

682: %\begin{eqnarray}

683: % \label{eq:bound-u-prime}

684: %  u^\prime &\geq& -\mu(1-u),\\

685: %  u^\prime &\geq& -\nu u, \label{eq:bound-u-prime+}\\

686: %  |u^{\prime\prime}| &\leq& c\mu+1+\gamma k,\label{eq:bound-u-prime2}\\

687: %  \label{eq:bound-w-prime}

688: %  w^\prime &\geq& -\frac{k}{c}(1-w)

689: %\end{eqnarray}

690: %holds.

691: \end{proposition}

692: \begin{proof}

693: Assume first that $c_0<\infty$ and fix an arbitrary $c\geq c_0$ and a

694: subsequence $\eps_i\to 0\, (i\to\infty)$ such that 

695: \begin{eqnarray*}

696:   c_0 &=& \lim_{i\to\infty} c_{\eps_i}.

697: \end{eqnarray*}

698: By Lemma \ref{lem:volpert} there exists a sequence of monotone travelling waves

699: $(c_i,u_{\eps_i},w_{\eps_i})$ for \eqref{rd-u-eps}, 

700: \eqref{rd-w-eps}, with $c_i=c$ if $c>c_0$ and $c_i=c_{\eps_i}$ if

701: $c=c_0$ such that 

702: \begin{eqnarray}

703:   0 &=& u_{\eps_i}^{\prime\prime} +c_iu_{\eps_i}^\prime-u_{\eps_i}+w_{\eps_i}

704:   -\gamma k u_{\eps_i}(1-w_{\eps_i}),\label{eq:tw-u-i}\\

705:   0 &=& \eps_i w_{\eps_i}^{\prime\prime} +c_iw_{\eps_i}^\prime +\eps_i(u_{\eps_i}-w_{\eps_i})

706:   +ku_{\eps_i}(1-w_{\eps_i}).\label{eq:tw-w-i}

707: \end{eqnarray}

708: Since travelling waves are invariant under space shifts, we can assume

709: without loss of generality that

710: \begin{gather}

711:   u_{\eps_i}(0)=\frac{1}{2}\quad\text{ for all }i\in\N.\label{eq:u0=0.5}

712: \end{gather}

713: By $0< u_{\eps_i},w_{\eps_i}<1$, the monotonicity of $w_{\eps_i}$ and

714: \eqref{eq:bound-u-eps-prime}, \eqref{eq:bound-u-eps-prime2} 

715: there exists $u\in C^{1,1}(\R), w\in L^\infty(\R)$ such that for all

716: $0<\alpha<1$, $R>0$

717: \begin{eqnarray}

718:   u_{\eps_i} &\to& u\quad\text{ in }C^{1,\alpha}([-R,R]),\label{eq:conv-u-i}\\

719:   w_{\eps_i} &\to& w\quad\text{ pointwise almost everywhere in }\R\label{eq:conv-w-i}

720: \end{eqnarray}

721: hold for a subsequence $\eps_i\to 0 (i\to\infty)$.

722: 

723: Multiplying \eqref{eq:tw-u-i}, \eqref{eq:tw-w-i} by a function $\eta\in C^\infty_c(\R)$ and

724: integrating we deduce

725: \begin{eqnarray*}

726:   0 &=& \int_\R \Big(- \eta^\prime u_{\eps_i}^{\prime} +c_{i}\eta

727:   u_{\eps_i}^\prime-\eta(u_{\eps_i}-w_{\eps_i}) -\eta 

728:   \gamma k u_{\eps_i}(1-w_{\eps_i})\Big),\\

729:   0 &=& \int_\R\Big( \eps_i \eta^{\prime\prime}w_{\eps_i}

730:   -c_{i}\eta^\prime w_{\eps_i} +{\eps_i}\eta (u_{\eps_i}-w_{\eps_i}) 

731:   +k\eta u_{\eps_i}(1-w_{\eps_i})\Big).

732: \end{eqnarray*}

733: Due to \eqref{eq:conv-u-i}, \eqref{eq:conv-w-i} we can pass  to the

734: limit $\eps_i\to 0$ in these equations

735: and get

736: \begin{eqnarray*}

737:   0 &=& \int_\R \Big(- \eta^\prime u^{\prime} +c\eta u^\prime-\eta(u-w) -\eta

738:   \gamma k u(1-w)\Big),\\

739:   0 &=& \int_\R\big(  -c\eta^\prime w +k\eta u(1-w)\big).

740: \end{eqnarray*}

741: It follows that $(c,u,w)$ solve \eqref{eq-tw-u}, \eqref{eq-tw-w} and, by

742: a bootstrapping argument, that $u,w$ are smooth.

743: Moreover, \eqref{eq:u0=0.5} and \eqref{eq:conv-u-i} yield that $u(0)=1/2$

744: and by Lemma 

745: \ref{lem-tw-basic} we obtain that $c>0$ and that $(c,u,w)$ 

746: is a monotone travelling wave.

747: % By \eqref{eq:bound-u-eps-prime}, \eqref{eq:bound-u-eps-prime+}

748: %we deduce \eqref{eq:bound-u-prime}, \eqref{eq:bound-u-prime+}

749: %and from \eqref{eq:bound-w-eps-prime} we get \eqref{eq:bound-w-prime}.

750: Since $c\geq c_0$ was arbitrary, the first part of the Proposition is

751: proved. To prove that $c_0<\infty$, let $\varrho,\sigma$ be two smooth

752: strictly monotonically decreasing functions with

753: \begin{alignat*}{2}

754:   \varrho(x)\,&=\,\sigma(x)\,=\,e^{-x}\quad&&\text{ for all }x\geq 1,\\

755:   \varrho(x)\,&=\,\sigma(x)\,=\,1-e^{-x}\quad&&\text{ for all }x\leq -1.

756: \end{alignat*}

757: Then there exists a constant $C(k)<\infty$ such that

758: for all $0<\eps<1$ and all $|x|\geq 1$ 

759: \begin{align*}

760:   \frac{\varrho^{\prime\prime}(x)-\varrho(x) +\sigma(x)

761:     -\gamma k\varrho(x)(1-\sigma(x))}{-\varrho^\prime(x)}\,&\leq\, C(k),\\

762:   \frac{\eps\sigma^{\prime\prime}(x) +\eps(\varrho-\sigma)(x)+k\varrho(x)(1-\sigma(x))}

763:             {-\sigma^\prime(x)}\,&\leq\, C(k).

764: \end{align*}

765: We can estimate the same ratios for all $0<\eps<1$ and $|x|\leq 1$ by

766: a constant depending only on $\|\varrho\|_{C^2([-1,1])}$,

767: $\|\sigma\|_{C^2([-1,1])}$. By Lemma \ref{lem:volpert} and the

768: definition of $c_\eps$ in \eqref{eq:def-c-eps} it follows that

769: \begin{gather*}

770:   c_\eps\,\leq\, C(k)\quad\text{ for all }0<\eps<1.

771: \end{gather*}

772: In particular, $c_0\leq C(k)$ is finite.

773: \end{proof}

774: We now complete the proof of Theorem \ref{the-exist}.

775: \begin{proposition}

776: \label{prop:c0=cmin}

777: There is no monotone travelling wave for \eqref{rd-u}, \eqref{rd-w} with speed $c<c_0$.

778: \end{proposition}

779: \begin{proof}

780: Assume that $(c,u,w)$ with $c<c_0$, $u,w\in\K$ satisfies \eqref{eq-tw-u}, \eqref{eq-tw-w}.

781: We then obtain that

782: \begin{align}

783:   \Phi^\eps_1(u,w)\,&=\,c,\notag\\

784:   \Phi^\eps_2(u,w)\,&\leq\, c +

785:   \eps\sup_{x\in\R}\frac{w^{\prime\prime}(x) + u(x)-w(x)}{-w^\prime(x)}.\label{eq:est-phi2}

786: \end{align} 

787: From \eqref{eq-tw-w} and $u<1$ we deduce that

788: \begin{gather}

789:         \frac{u-w}{-w^\prime}\,\leq\, \frac{u(1-w)}{-w^\prime}\,=\,\frac{c}{k}.\label{eq:phi2-term1}

790: \end{gather}

791: A differentiation in \eqref{eq-tw-w} yields that

792: \begin{align*}

793:         -cw^{\prime\prime}\,&=\, ku^\prime(1-w) -kuw^\prime

794:                 \,=\, -cw^\prime\frac{u^\prime}{u} -kuw^\prime

795: \end{align*}

796: which gives, together with \eqref{eq:bound-u-eps-prime+-0}, that

797: \begin{gather}

798:  \frac{w^{\prime\prime}}{-w^\prime}\,=\, \frac{-u^\prime}{u} -\frac{k}{c}u\,\leq\ \nu

799:  \label{eq:phi2-term2}

800: \end{gather}

801: with $\nu$ as in \eqref{eq:def-nu-0}.

802: Using \eqref{eq:phi2-term1}, \eqref{eq:phi2-term2} in \eqref{eq:est-phi2} we obtain that

803: \begin{gather*}

804:  \Phi^\eps_2(u,w)\,\leq\ c+\eps C,

805: \end{gather*}

806: where $C$ is independent of $\eps>0$.

807: From Lemma \ref{lem:volpert} and the definition of $c_0$,

808: \eqref{eq:def-c-eps} and \eqref{eq:c-i} we deduce that

809: \begin{gather*}

810:  c_0\,\leq\, \liminf_{\eps\to 0} \max (\Phi^\eps_1(u,w),\Phi^\eps_2(u,w))\,=\, c,

811: \end{gather*}

812: which is a contradiction to our assumption $c<c_0$.

813: \end{proof} 

814: 

815: 

816: %===================================================

817: % section limit

818: %===================================================

819: \section{The Stefan-like limit problem}

820: \label{sec-limit}

821: The reaction-diffusion system \eqref{rd-u}, \eqref{rd-w} converges to a

822: Stefan-like free boundary problem as $k$ tends to infinity, see \cite{HKR}.  

823: For solutions ${U}_\infty,{W}_\infty$ of this limit problem

824: \begin{eqnarray}

825:   0 &=& U_\infty(1-W_\infty),\label{st-w0}

826: \end{eqnarray}

827: holds and the spatial

828: domain splits in a region where ${U}_\infty=0$, and a region where

829: ${U}_\infty>0$ and ${W}_\infty=1$. If we denote 

830: their common boundary at time $t$ by $\Gamma(t)$ then $U_\infty,

831: W_\infty$ satisfy

832: \begin{alignat}{2}

833:   \partial_t {U}_\infty \,&=\, \Delta {U}_\infty

834:   -{U}_\infty + 1&\quad&\text{ in }\{{W}_\infty=1\},\label{st-w=1}\\

835:   \gamma \partial_t {W}_\infty \,&=\, {W}_\infty&\quad&\text{ in

836:   }\{{U}_\infty=0\}\label{st-u=0}

837: \end{alignat}

838: and a continuity and jump condition on $\Gamma(t)$,

839: \begin{align}

840:   \mbox{}[U_\infty(t,.)]\,&=\, 0,\label{st-Gamma2}\\

841:    - [\nabla{U}_\infty(t,.)\cdot\nu(t,.)]\,&=\,\gamma

842:   [{W}_\infty(t,.)]\vec{v}(t,.)\cdot\nu(t,.)\label{st-Gamma}, 

843: \end{align}

844: where $\vec{v}(t,.)$ and $\nu(t,.)$ are the velocity and the unit normal of the free boundary

845: $\Gamma(t)$, pointing into

846: $\{U_\infty=0\}$, and $[.]$ denotes the jump across the free boundary from

847: the region $\{U_\infty>0\}$ to $\{U_\infty=0\}$.\\

848: As for the reaction-diffusion system, travelling-wave solutions

849: $U_\infty, W_\infty$ are

850: given by a speed $c\in\R$ and \emph{profile functions} 

851: $u_\infty,w_\infty$,

852: \begin{gather*}

853:   U_\infty(x,t)\,=\, u_\infty(x-ct),\quad W_\infty(x,t)\,=\,

854:  w_\infty(x-ct).

855: \end{gather*}

856: We are interested in monotone travelling waves which

857: connect unity and zero,

858: \begin{align}

859:   u_\infty(x),w_\infty(x) &\,\to\, 1\quad\text{ as }x\to -\infty,\label{eq:tw-lim-infty-}\\

860:   u_\infty(x),w_\infty(x) &\,\to\, 0\quad\text{ as }x\to \infty,\label{eq:tw-lim-infty}\\

861:   u_\infty,w_\infty \quad&\text{ are monotonically decreasing}.\label{eq:tw-lim-mono}

862: \end{align}

863: Due to the shift invariance of travelling waves, the condition

864: \eqref{st-w0} and the monotonicity of $u_\infty, w_\infty$ we can assume

865: that 

866: \begin{alignat}{2}

867:   u_\infty(x)\,&=\,0&\qquad&\text{ for }x\geq 0,\label{eq:tw-st0}\\

868:   u_\infty(x)\,&>\,0,\, w_\infty(x)\,=\,1&\quad&\text{ for }x<0.\label{eq:tw-st1}

869: \end{alignat}

870: From \eqref{st-w=1}-\eqref{st-Gamma} we then obtain

871: \begin{alignat}{2}

872:   0\,&=\, u_\infty^{\prime\prime}+cu_\infty^\prime -u_\infty +1&\quad&\text{

873:     for }x<0,\label{eq:tw-st2}\\

874:   0\,&=\, \gamma c w_\infty^\prime + w_\infty&\quad&\text{ for }x>0,

875:     \label{eq:tw-st3}

876: \end{alignat}

877: and the continuity and jump condition

878: \begin{align}

879:       u_\infty(0-)\,&=\,0,\label{eq:tw-st5}\\

880:   \gamma c (1-w_\infty(0+)) &\,=\,

881:     -u^\prime_\infty(0-)\label{eq:tw-st4}.

882: \end{align}

883: \begin{proposition}

884: \label{prop:ex-tw-lim}

885: For all $c\geq c_\infty$, where

886: \begin{eqnarray}\label{eq-c-infty}

887:         c_\infty &:=& \frac{1}{\sqrt{\gamma (1+\gamma)}},

888: \end{eqnarray}

889: there exists a unique solution $u_\infty,w_\infty$ of

890: \eqref{eq:tw-lim-infty-}-\eqref{eq:tw-st4}.

891: This solution is given by

892: \begin{eqnarray}

893:         u_\infty(x) &=&

894:         \begin{cases}

895:                 1- e^{\alpha x} &\text{ if } x\leq 0,\\

896:                 0 &\text{ if }x\geq 0,

897:         \end{cases}

898:         \label{tw-lim-u}\\

899:         w_\infty(x) &=&

900:         \begin{cases}

901:                 1 &\text{ if }x<0,\\

902:                 \beta e^{\textstyle{-\frac{x}{c\gamma}}} &\text{ if }x>0,

903:         \end{cases}\label{tw-lim-w}

904: \end{eqnarray}

905: where

906: \begin{gather}

907:         \alpha=\alpha(c) := -\frac{c}{2} + \sqrt{\frac{c^2}{4}+1},\label{eq:def-alpha-mu}\\

908:         \qquad \beta=\beta(c):=1-\frac{\alpha(c)}{\gamma c}.\label{eq:def-alpha}

909: \end{gather}

910: For $c < c_\infty$, there does not exist any solution.

911: \end{proposition}

912: \begin{proof}

913: We deduce that \eqref{eq:tw-st2} and \eqref{eq:tw-lim-infty-},

914: \eqref{eq:tw-st5} hold if and only if

915: \eqref{tw-lim-u} holds, with $\alpha$ as in \eqref{eq:def-alpha-mu}.

916: Similarly \eqref{eq:tw-st3} and \eqref{eq:tw-lim-infty} are satisfied if

917: and only if $w_\infty$ satisfies \eqref{tw-lim-w} with $\beta\in\R$.

918: Since

919: \begin{gather*}

920:   1- w_\infty(0+)\,=\, 1-\beta,\quad

921:   -u_\infty^\prime(0-)\,=\, \alpha

922: \end{gather*}

923: we obtain that the jump condition \eqref{eq:tw-st4} is satisfied if and

924: only if $\beta$ satisfies \eqref{eq:def-alpha}.

925: Finally, by \eqref{tw-lim-w} and \eqref{eq:def-alpha}, the solution

926: $w_\infty$ is nonnegative if 

927: and only if $c\geq c_\infty$.

928: \end{proof}

929: %===================================================

930: % section speed selection

931: %===================================================

932: \section{Selection of the minimal speed}

933: \label{sec-speed}

934: In this section we prove that for sufficiently large values $k>k_0$ a nonlinear selection principle determines

935: the minimal speed of travelling waves. The threshold $k_0$ is obtained explicitly in terms of

936: the given constant $\gamma$.

937: First, we have to analyse the behaviour of

938: travelling-wave solutions at infinity.

939: \subsection{Behaviour at infinity}

940: The linear selection principle for the minimal speed is based on the analysis

941: of the linearised system at the unstable stationary point $(u,w)=(0,0)$ of \eqref{eq-tw-u}, \eqref{eq-tw-w}.

942: In the next lemma we show that solutions have to decay exponentially to zero as $x$ tends to infinity.

943: \begin{lemma}\label{lem-decay-infty+}

944: Let $(c,u,w)$ be a monotone travelling wave. Then

945: \begin{gather}\label{eq-decay-infty+}

946:         \lim_{x\to\infty}\frac{1}{x}\log u(x)\,=\,\lim_{x\to\infty}\frac{1}{x}\log w(x)\,=\,\lambda,

947: \end{gather}

948: where $\lambda$ is a negative root of the cubic equation

949: \begin{eqnarray}\label{eq-lambda}

950:         0 &=& \lambda^3 + c\lambda^2 -(1+\gamma k)\lambda -\frac{k}{c}.

951: \end{eqnarray}

952: \end{lemma}

953: \begin{proof}

954: With the definitions

955: \begin{gather*}

956:   V_0(x)\,:=\,

957:   \begin{pmatrix}

958:     u^\prime(x)\\

959:     u(x)\\

960:     w(x)

961:   \end{pmatrix}\!,

962:   \quad

963:   A_0\,:=\,

964:   \begin{pmatrix}

965:     -c & 1+\gamma k & -1\\

966:     1 &0 & 0\\

967:     0 &-\frac{k}{c} &0

968:   \end{pmatrix}\!,

969:   \quad

970:   B_0\,:=\,

971:   \begin{pmatrix}

972:     -\gamma\\

973:     0\\

974:     \frac{1}{c}

975:   \end{pmatrix}

976: \end{gather*}

977: the system \eqref{eq-tw-u}, \eqref{eq-tw-w} is equivalent to

978: \begin{eqnarray}

979:   V_0^\prime(x) &=& A_0V_0(x) + ku(x)w(x)B_0

980: \end{eqnarray}

981: and the linearized system at $(0,0,0)$ is given by

982: \begin{eqnarray}\label{eq:xi-lin}

983:   \xi^\prime(x) &=& A_0\xi(x).

984: \end{eqnarray}

985: The eigenvalues of $A_0$ are the

986: solutions of \eqref{eq-lambda}. Since $s\mapsto s^3 + c s^2 -(1+\gamma k)s -\frac{k}{c}$ is negative

987: at $s=0$ and becomes positive as $s\to +\infty$ there exists a positive eigenvalue $\lambda_+>0$.

988: The other two solutions of \eqref{eq-lambda} satisfy the equation

989: \begin{eqnarray}\label{eq-lambda-+}

990:   0 &=& \lambda^2 +(c+\lambda_+)\lambda +\frac{k}{c\lambda_+} 

991: \end{eqnarray}

992: which has, depending on the values of $c,k,\lambda_+$, either two negative roots, one repeated negative root

993: or two complex-conjugate roots with negative real part.

994: Since $V_0(x)$ converges to zero as $x\to +\infty$, the curve

995: $x\mapsto V_0(x)$ is for sufficiently large values of $x$ contained in the stable manifold and

996: converges exponentially to zero; see for example \cite{Pe} Section 2.7.

997: By \cite{CL} Theorem XIII.4.5 there exists a solution $\xi=(\xi_1,\xi_2,\xi_3)$ of the linearized system

998: \eqref{eq:xi-lin} and $\delta>0$ with

999: \begin{eqnarray}\label{eq:exp-close}

1000:   V_0(x) &=& \xi(x) + O(e^{(\alpha-\delta)x}),\\

1001:   \xi(x) &=& O(e^{\alpha x})\notag

1002: \end{eqnarray}

1003: as $x\to +\infty$, where $\alpha<0$ is the real part of an eigenvalue $\lambda$ of $A_0$.

1004: One checks that $A_0$ has no eigenvector with a component equal to zero; therefore

1005: \begin{eqnarray}\label{eq:dec-comp}

1006:   \xi_i(x) &=& O(e^{\alpha x})\quad\text{ for } i=1,2,3

1007: \end{eqnarray}

1008: holds as $x\to +\infty$. Let us show that in fact $\lambda$ is real. Assume that

1009: $\lambda=\alpha+i\beta$ with $\beta\neq 0$. Then

1010: $\xi(x)$ describes, as $x\to +\infty$, a spiral around the origin contained

1011: in the plane $P$ spanned by the real and imaginary part of an eigenvector

1012: of $A_0$ with eigenvalue $\lambda$.

1013: But then, since the difference between $\xi$ and $V_0$ decays exponentially faster than $\xi$,

1014: $V_0$ has to take values outside the set $\{z=(z_1,z_2,z_3):z_1<0,z_2>0,z_3>0\}$,

1015: which is a contradiction to the assumption that $(c,u,w)$ is a

1016: monotone travelling wave. This shows that $\lambda=\alpha<0$ is real. Thus \eqref{eq-decay-infty+}

1017: follows from \eqref{eq:exp-close} and \eqref{eq:dec-comp}.

1018: \end{proof}

1019: The equation \eqref{eq-lambda} connects the speed and the decay

1020: rate at $+\infty$ of a travelling wave. We now further analyze

1021: this relation.

1022: \begin{lemma}\label{lem:curve-lambda-c}

1023: For all $k>0$ and all $\lambda<0$ there exists a unique value

1024: $\bar{c}_k(\lambda)>0$ such that $\lambda$ satisfies 

1025: \eqref{eq-lambda} for $c=\bar{c}_k(\lambda)$. The function

1026: $\lambda\mapsto \bar{c}_k(\lambda)$ 

1027: attains a positive minimum $c_{\lin}=c_{\lin}(k)$ at a unique value

1028: $\lambda_{\lin}=\lambda_{\lin}(k)$ and 

1029: $c_{\lin},\lambda_{\lin}$ are given by

1030: \begin{eqnarray}

1031:   \lambda_{\lin}&<& 0\notag\\

1032:   \label{eq-lambda-lin1}

1033:   3\lambda_{\lin}^2 &=& -(1+\gamma k)+2\sqrt{(1+\gamma k)^2+3k},\\

1034:   \lambda_{\lin} c_{\lin} &=& (1+\gamma k) -\sqrt{(1+\gamma k)^2+3k}.\label{eq-lambda-lin}

1035: \end{eqnarray}

1036: Moreover

1037: \begin{alignat}{2}

1038:   \bar{c}_k^\prime(\lambda)\,&<\, 0 \quad &&\text{ for

1039:  }\lambda<\lambda_{\lin},\label{eq:mon-ck-1}\\

1040:   \bar{c}_k^\prime(\lambda)\,&>\, 0 \quad &&\text{ for

1041:  }\lambda>\lambda_{\lin},\label{eq:mon-ck-2}

1042: \end{alignat}

1043: with

1044: \begin{eqnarray*}

1045:   \bar{c}_k(\lambda) &\to& \infty\quad\text{ as }\lambda\to -\infty,\\

1046:   \bar{c}_k(\lambda) &\to& \infty\quad\text{ as }\lambda\to 0.

1047: \end{eqnarray*}

1048: \end{lemma}

1049: \begin{proof}

1050: From \eqref{eq-lambda} we obtain that $\bar{c}_k$ is given by

1051: \begin{eqnarray*}

1052:   \bar{c}_k(\lambda) &=& -\frac{1}{2}\Big(\lambda-(1+\gamma k)\frac{1}{\lambda}\Big)

1053:   +\sqrt{\frac{1}{4}\Big(\lambda-(1+\gamma k)\frac{1}{\lambda}\Big)^2+\frac{k}{\lambda^2}},

1054: \end{eqnarray*}

1055: that $\bar{c}_k$ is strictly positive and that $\bar{c}_k$ tends to

1056: infinity as $\lambda\to -\infty$ or $\lambda\to 0$. 

1057: Therefore the positive minimum $c_{\lin}$ is attained at a value

1058: $\lambda_{\lin}$ and 

1059: $\bar{c}^\prime(\lambda_{\lin})=0$ holds. By \eqref{eq-lambda} this implies

1060: \begin{eqnarray}\label{eq:lam1}

1061:   0 &=& 3\lambda_{\lin}^2 + 2\lambda_{\lin} c_{\lin} - (1+\gamma k).

1062: \end{eqnarray}

1063: Moreover, by \eqref{eq-lambda},

1064: \begin{eqnarray}\label{eq:lam2}

1065:         0 &=& \lambda_{\lin}^3 + c_{\lin}\lambda_{\lin}^2 -(1+\gamma

1066:         k)\lambda_{\lin} -\frac{k}{c_{\lin}}. 

1067: \end{eqnarray}

1068: One checks that \eqref{eq-lambda-lin1}, \eqref{eq-lambda-lin} is equivalent

1069: to \eqref{eq:lam1}, \eqref{eq:lam2}.

1070: In particular, $\bar{c}_k^\prime$ has only one zero and we deduce that

1071: $\bar{c}_k^\prime(\lambda)<0$ for $\lambda<\lambda_{\lin}$

1072: and  $\bar{c}_k^\prime(\lambda)>0$ for $0>\lambda>\lambda_{\lin}$.

1073: \end{proof}

1074: \begin{corollary}\label{cor-c(lambda)}

1075: The minimal speed $c_{\min}(k)$ of travelling waves satisfies the estimate

1076: \begin{eqnarray}\label{eq-lower-bound}

1077:   c_{\min}(k) &\geq& c_{\lin}(k).

1078: \end{eqnarray}

1079: \end{corollary}

1080: \begin{proof}

1081: According to Lemma \ref{lem-decay-infty+} for a monotone travelling wave with speed $c$, a negative root

1082: $\lambda$ of \eqref{eq-lambda} exists.

1083: On the other hand $c_{\lin}(k)$ gives the minimal value of $c$ such that \eqref{eq-lambda} has a negative solution.

1084: \end{proof}

1085: We now state the result corresponding to Lemma \ref{lem-decay-infty+}

1086: if we consider $x$ approaching $-\infty$.

1087: \begin{lemma}\label{lem-decay-infty-}

1088: Let $(c,u,w)$ be a monotone travelling wave and let $k\geq 1$.

1089: Then

1090: \begin{eqnarray}

1091:         \lim_{x\to-\infty}\frac{1}{x}\log \big(1-u(x)\big) &=& \mu,\label{eq-decay-infty-u}\\

1092:         \lim_{x\to-\infty}\frac{1}{x}\log \big(1-w(x)\big) &=&\frac{k}{c},\label{eq-decay-infty-w}

1093: \end{eqnarray}

1094: where $\mu>0$ is given by

1095: \begin{eqnarray}\label{eq-lambda-}

1096:         \mu &=& -\frac{c}{2} +\sqrt{\frac{c^2}{4}+1}.

1097: \end{eqnarray}

1098: For $0<k<1$ the limits in \eqref{eq-decay-infty-u}, \eqref{eq-decay-infty-w}

1099: are either equal to $\mu$ or equal to $k/c$.

1100: \end{lemma}

1101: \begin{proof}

1102: With

1103: \begin{gather*}

1104:   V_1(x)\,:=\,

1105:   \begin{pmatrix}

1106:     -u^\prime(x)\\

1107:     1-u(x)\\

1108:     1-w(x)

1109:   \end{pmatrix}\!,

1110:   \quad

1111:   A_1\,:=\,

1112:   \begin{pmatrix}

1113:     -c & 1 & -(1+\gamma k)\\

1114:     1 &0 & 0\\

1115:     0 & 0 &\frac{k}{c}

1116:   \end{pmatrix}\!,

1117:   \quad

1118:   B_1\,:=\,

1119:   \begin{pmatrix}

1120:     \gamma\\

1121:     0\\

1122:     -\frac{1}{c}

1123:   \end{pmatrix}

1124: \end{gather*}

1125: the system \eqref{eq-tw-u}, \eqref{eq-tw-w} is equivalent to

1126: \begin{eqnarray}\label{eq:V-A-}

1127:   V_1^\prime(x) &=& A_1V_1(x) + k\big(1-u(x)\big)\big(1-w(x)\big)B_1

1128: \end{eqnarray}

1129: and the linearized system at $V_1=(0,0,0)$ is given by

1130: \begin{eqnarray}\label{eq:xi-lin-}

1131:   \zeta^\prime(x) &=& A_1\zeta(x).

1132: \end{eqnarray}

1133: The matrix $A_1$ has the positive eigenvalues

1134: \begin{gather*}

1135:   \mu\,=\, -\frac{c}{2} +\sqrt{\frac{c^2}{4}+1},\quad

1136:  \mu_1\,=\,\frac{k}{c}

1137: \end{gather*}

1138: and

1139: one negative eigenvalue 

1140: \begin{gather*}

1141:   \mu_2\, =\, -\frac{c}{2} - \sqrt{\frac{c^2}{4} +1}.

1142: \end{gather*}

1143: Since $V_1(x)\to 0$ as $x\to -\infty$ we deduce

1144: that $V_1(x)$ is for sufficiently small $x$ contained in the unstable manifold

1145: of \eqref{eq:V-A-} at $(0,0,0)$.

1146:  Using \cite{CL} Theorem XIII.4.5 we obtain the existence of a solution

1147: $\zeta$ of the linearized system

1148: \eqref{eq:xi-lin-} and a $\delta>0$ with

1149: \begin{eqnarray}

1150:   V_1(x) &=& \zeta(x) + O(e^{(\beta+\delta)x}),\label{eq:lin-close-}\\

1151:   \zeta(x) &=& O(e^{\beta x})\label{eq:lin-decay-}

1152: \end{eqnarray}

1153: as $x\to -\infty$, where $\beta=k/c$ or $\beta=\mu$.

1154: One checks that $k/c>\mu$ holds for $k\geq 1$.

1155: Thus, if $\beta=k/c$ then the trajectory of $V_1$ as $x\to -\infty$ has to be tangential

1156: to the eigenspace of $A_1$ corresponding to the eigenvalue $k/c$. On the other hand this

1157: eigenspace is spanned by a vector $(k/c,1,a)$, where $a<0$ for $k\geq 1$.

1158: Therefore $V_1$ leaves the region $\{z=(z_1,z_2,z_3): z_2,z_3>0\}$, which

1159: is a contradiction to $u(x),w(x)<1$.

1160: This proves that $\beta=\mu$ for $k\geq 1$ and implies that the trajectory of

1161: $\zeta(x)$ for $x\to -\infty$ is tangential to the eigenspace corresponding

1162: to the eigenvalue $\mu$, which is spanned by the vector $(\mu,1,0)$.

1163: Therefore

1164: \begin{eqnarray*}

1165:   \zeta_2(x) &=& O(e^{\mu x})

1166: \end{eqnarray*}

1167: and \eqref{eq-decay-infty-u} follows from \eqref{eq:lin-close-}. By

1168: \eqref{eq-tw-w} we deduce

1169: \begin{eqnarray*}

1170:   \frac{d}{dx}\log \big(1-w(x)\big) &=& \frac{k}{c} + O(e^{\mu x})\quad\text{ as }x\to -\infty,

1171: \end{eqnarray*}

1172: and thus \eqref{eq-decay-infty-w} holds.

1173: \end{proof}

1174: \begin{corollary}\label{cor:decay-}

1175: Let $k\geq 1$. If we consider two different monotone travelling waves

1176: for \eqref{rd-u}, \eqref{rd-w}, 

1177: the one with the lower speed converges faster to unity as $x$ approaches $-\infty$.

1178: \end{corollary}

1179: \begin{proof}

1180: For a monotone travelling wave $(c,u,w)$, by Lemma \ref{lem-decay-infty-} the convergence of $u,w$

1181: to unity as $x\to -\infty$

1182: is exponential with rate $\mu$ for $u$ and rate $k/c$ for $w$, where $\mu$ is given by \eqref{eq-lambda-}. Since

1183: \begin{gather*}

1184:   \frac{d}{dc}\Big( -\frac{c}{2} +\sqrt{\frac{c^2}{4}+1}\Big)\,=\,

1185:   \frac{1}{2}\Big(-1+\frac{c}{\sqrt{c^2+4}}\Big)\,<\, 0

1186: \end{gather*}

1187: we deduce that both convergence rates are decreasing with $c$. 

1188: \end{proof}

1189: %

1190: %

1191: %

1192: %

1193: %

1194: %

1195: \subsection{The reduced system for $c=c_\infty$}

1196: It was observed in \cite{Ki} that the travelling-wave equations \eqref{eq-tw-u}, \eqref{eq-tw-w}

1197: are for the speed $c=c_\infty$ defined in

1198: \eqref{eq-c-infty} remarkable in being equivalent to a system of two first-order equations.

1199: \begin{lemma}\label{lem:red}

1200: Let $u\in \K_0,w\in \K$. Then $(c_\infty,u,w)$ is a monotone travelling wave if and only if $u,w$ satisfy

1201: \begin{eqnarray}\label{eq-red-u}

1202:         u^\prime &=& \gamma c_\infty (u-w),\\

1203:         w^\prime &=& -k\gamma (1+\gamma) c_\infty u(1-w).\label{eq-red-w}

1204: \end{eqnarray}

1205: \end{lemma}

1206: \begin{proof}

1207: Multiplying \eqref{eq-red-w} by $c_\infty$ and using \eqref{eq-c-infty}

1208: we see that \eqref{eq-red-w} and \eqref{eq-tw-w} are equivalent for $c=c_\infty$.

1209: Next we obtain from \eqref{eq-c-infty} that

1210: \begin{eqnarray}

1211:   \big(u^\prime -\gamma c_\infty(u-w)\big)^\prime &=& u^{\prime\prime} +c_\infty u^\prime

1212:   -\frac{1}{\gamma c_\infty}u^\prime -\gamma k u(1-w)\label{eq:aux-red}

1213: \end{eqnarray}

1214: and we see that \eqref{eq-red-u} implies \eqref{eq-tw-u}.

1215: Conversely, from \eqref{eq-tw-u} and \eqref{eq:aux-red} we deduce that

1216: \begin{eqnarray*}

1217:   \big(u^\prime -\gamma c_\infty(u-w)\big)^\prime &=& -\frac{1}{\gamma c_\infty}\big(u^\prime-\gamma c_\infty(u-w)\big),

1218: \end{eqnarray*}

1219: whose solutions are

1220: \begin{eqnarray*}

1221:   \big(u^\prime -\gamma c_\infty(u-w)\big)(x) &=& a e^{-\frac{1}{\gamma

1222:   c_\infty}x}\quad\text{ for any } a\in\R. 

1223: \end{eqnarray*}

1224: The condition that $u(x),w(x)$ converge exponentially to unity as $x\to -\infty$

1225: implies that $a=0$ and \eqref{eq-red-u} therefore holds.

1226: \end{proof}

1227: This reduction allows us to prove the existence of a travelling wave with speed $c_\infty$

1228: for \eqref{rd-u}, \eqref{rd-w} by a phase-plane analysis for

1229: \eqref{eq-red-u}, \eqref{eq-red-w}.

1230: \begin{proposition}\label{prop-red}

1231: For all $\gamma,k>0$ there exists a monotone travelling wave  $(c_\infty,u_0,w_0)$

1232: for \eqref{rd-u}, \eqref{rd-w}. As $x$ tends to infinity, $u_0,w_0$ decay exponentially to

1233: zero with decay rate

1234: \begin{gather}\label{eq:dec-u-0}

1235:   \lambda_{\infty}\,=\,\lambda_\infty(k) \,:=\,

1236:   c_\infty\gamma\Big(\frac{1}{2}-\sqrt{\frac{1}{4}+k(1+\gamma)}\Big)\,<\,0. 

1237: \end{gather}

1238: \end{proposition}

1239: \begin{proof}

1240: The system \eqref{eq-red-u}, \eqref{eq-red-w}

1241: has the two stationary points $(0,0)$ and $(1,1)$.

1242: We define the set

1243: \begin{gather}\label{def:J}

1244:   J\,:=\,\{(x_1,x_2)^T\in\R^2 : 0\leq x_1<x_2\leq 1\}

1245: \end{gather}

1246: and observe that $\R^2\setminus J$ is an invariant region for

1247: \eqref{eq-red-u}, \eqref{eq-red-w} (see Figure \ref{fig:red}).\\

1248: Next we consider the linearisation at $(0,0)$ of \eqref{eq-red-u}, \eqref{eq-red-w}

1249: which is given by

1250: \begin{eqnarray*}

1251:   \zeta^\prime &=& c_\infty\gamma

1252:   \begin{pmatrix}

1253:     1 & -1\\

1254:     -k(1+\gamma) & 0

1255:   \end{pmatrix}

1256:   \zeta,

1257: \end{eqnarray*}

1258: for $\xi:\R\to\R^2$.

1259: The eigenvalues of this linear system are $\lambda_\infty$, as defined

1260: in \eqref{eq:dec-u-0}, and 

1261: \begin{gather*}

1262:   c_\infty\gamma\Big(\frac{1}{2}+\sqrt{\frac{1}{4}+k(1+\gamma)}\Big)>0.

1263: \end{gather*}

1264: An eigenvector with eigenvalue $\lambda_\infty$ is given by

1265: \begin{gather*}

1266:   v \,:=\,

1267:   \begin{pmatrix}

1268:     -\frac{1}{2} + \sqrt{\frac{1}{4}+k(1+\gamma)}\\

1269:     k(1+\gamma)

1270:   \end{pmatrix}

1271: \end{gather*}

1272: One checks that $v_1<v_2$ holds for the components of $v$ and deduces

1273: that the eigenspace corresponding to $\lambda_\infty$ intersects the

1274: set $J$ defined in \eqref{def:J}.

1275: \begin{figure}[t]

1276: \centering

1277: \includegraphics[height=65mm]{stable.eps}

1278: \caption{Schematic of the phase portrait of

1279:   \eqref{eq-red-u},\eqref{eq-red-w}. The dashed area represents the set

1280:   $J$, the incoming arc the stable manifold at $(0,0)$ and the arrows the

1281:   direction field.}

1282: \label{fig:red}

1283: \end{figure}

1284: 

1285: By the stable manifold theorem, see for example \cite{Pe}, Theorem 2.7, there exists a

1286: trajectory that, taking $t$ as its parameter variable, converges to $(0,0)$ as $t\to\infty$

1287: and starts at $t=0$ in $J$, since the stable manifold is tangential to the eigenspace corresponding

1288: to the eigenvalue $\lambda_\infty$.

1289: Following this trajectory back with decreasing $t$ we cannot leave $J$, since otherwise

1290: the trajectory would stay in $\R^2\setminus J$ as $t$ increases and thus

1291: could not reach any point in $J$ at $t=0$. In $J$, with decreasing $t$, the trajectory

1292: has to be monotone in both components and therefore has to approach the

1293: stationary point $(1,1)$. Thus the trajectory connects $(1,1)$ to $(0,0)$ and satisfies

1294: \eqref{eq-red-u}, \eqref{eq-red-w}. By Lemma \ref{lem:red}

1295: this shows that $(c_\infty,u_0,w_0)$ is a monotone travelling wave for

1296: \eqref{rd-u}, \eqref{rd-w}. Moreover $u_0(x),w_0(x)$ are, for sufficiently large $x$, in the stable

1297: manifold of \eqref{eq-red-u}, \eqref{eq-red-w}

1298: and we deduce that they converge exponentially fast to zero, with

1299: decay rate given by $\lambda_\infty$.   

1300: \end{proof}

1301: The existence of a travelling wave with speed $c_\infty$ implies immediately

1302: the following estimate.

1303: \begin{corollary}

1304: The minimal speed $c_{\min}(k)$ of travelling waves satisfies

1305: \begin{eqnarray}\label{eq-upper-bound}

1306:   c_{\min}(k) &\leq& c_\infty.

1307: \end{eqnarray}

1308: \end{corollary}

1309: As we will see, for sufficiently large values of $k$ the minimal speed

1310: is identical to the value $c_\infty$.

1311: \subsection{Analysis of the decay-rates at $+\infty$}

1312: In this section we further investigate the deacy of travelling

1313: waves to zero as $x$ tends to $+\infty$. With this aim we analyze the

1314: functions $\bar{c}_k$ defined in Lemma \ref{lem:curve-lambda-c}:

1315: $\bar{c}_k(\lambda)$ is the

1316: speed of a travelling wave with decay rate $\lambda$ at $+\infty$.

1317: \begin{remark}\label{rem-decay-red}

1318: Corresponding to the reduction \eqref{eq-red-u}, \eqref{eq-red-w} of the

1319: travelling-wave system \eqref{eq-tw-u}, \eqref{eq-tw-w}, we find that

1320: the equation \eqref{eq-lambda} for the possible decay rates factorises

1321: for the speed $c_\infty=(\gamma(1+\gamma))^{-1/2}$. The value

1322: \begin{gather}

1323:   \lambda_\infty^* \,:=\, -\frac{1}{\gamma c_\infty}\,=\,

1324:   -\sqrt{\frac{1+\gamma}{\gamma}} \label{eq:def-lambda-infty-*}

1325: \end{gather}

1326: is for all $k>0$ a negative root of \eqref{eq-lambda} with $c=c_\infty$.

1327: The decay rate of $u_0,w_0$ is given by the other negative root of this equation, which is

1328: the value $\lambda_\infty(k)$ defined in \eqref{eq:dec-u-0}.

1329: \end{remark}

1330: \begin{lemma}\label{lem:decay}

1331: For $k>0$ consider the values $\lambda_{\lin}(k), c_{\lin}(k)$ as

1332: defined in Lemma \ref{lem:curve-lambda-c} and the values

1333: $\lambda_{\infty}(k),\lambda_\infty^*$ as given in \eqref{eq:dec-u-0},

1334: \eqref{eq:def-lambda-infty-*}. Then there exists a unique $k_0>0$, which

1335: is is given explicitely by $k_0=(1+2\gamma)\gamma^{-2}$, such that

1336: \begin{gather}

1337:   \lambda_{\lin}(k_0)\,=\, \lambda_\infty^*.\label{eq:k0}

1338: \end{gather}

1339: Moreover, for $0<k_1<k_0<k_2$

1340: %\begin{gather}

1341: %  c_{\lin}(k_0)\,=\,c_\infty\label{eq:k0-c}

1342: %\end{gather}

1343: \begin{alignat}{3}

1344:   &\lambda_\infty(k_1) \,&&>\,\lambda_{\lin}(k_1)\,&&>\,\lambda_\infty^*

1345:   ,\label{eq:ineqs-k1}\\

1346:   &\lambda_\infty(k_0)

1347:   &&=\lambda_{\lin}(k_0)\,&&=\,\lambda_\infty^*,\label{eq:ineqs-k0}\\ 

1348:   &\lambda_\infty(k_2) \,&&<\,\lambda_{\lin}(k_2)\,&&<\,\lambda_\infty^*

1349:   \label{eq:ineqs-k2}

1350: \end{alignat}

1351: hold, see Figure \ref{fig:dec}.

1352: \end{lemma}

1353: \begin{proof}

1354: We have proved in Lemma \ref{lem:curve-lambda-c} that the functions

1355: $\bar{c}_k$ attain their minimum $c_{\lin}(k)$ at a unique 

1356: value $\lambda_{\lin}(k)$,

1357: \begin{align}

1358:   &c_{\lin}(k)\,=\, \bar{c}_k(\lambda_{\lin}(k)),\label{eq:ck-min}\\

1359:   &\bar{c}_k(\lambda)\,>\,c_{\lin}(k)\quad\text{ for all }\lambda<0,

1360:   \lambda\,\neq\,\lambda_{\lin}(k).\label{eq:ck-min-uni}

1361: \end{align}

1362: For convenience we recall that

1363: \begin{align}

1364:   3\lambda_{\lin}^2 \,&=\, -(1+\gamma k)+2\sqrt{(1+\gamma

1365:   k)^2+3k},\quad \lambda_{\lin}<0,\label{eq:recap-lambda-lin}\\ 

1366:   \lambda_{\lin} c_{\lin} \,&=\, (1+\gamma k) -\sqrt{(1+\gamma

1367:   k)^2+3k}.\label{eq:recap-c-lin} 

1368: \end{align}

1369: By Proposition \ref{prop-red}, Remark \ref{rem-decay-red} and the

1370: definition of $\bar{c}_k$ in Lemma \ref{lem:curve-lambda-c}

1371: \begin{gather}

1372:   \bar{c}_k(\lambda_\infty^*)\,=\,\bar{c}_k(\lambda_\infty(k)) \,=\,

1373:   c_\infty\quad\text{ for all }k>0 \label{eq:eqs-ck}

1374: \end{gather}

1375: holds.

1376: Next we see from

1377: \eqref{eq:recap-lambda-lin} that $\lambda_{\lin}(k)$ is strictly

1378: decreasing in $k$ 

1379: and that

1380: \begin{align*}

1381:   \lim_{k\to 0} \lambda_{\lin}(k)\,&=\, -\frac{1}{3}\sqrt{3}\,>\, -1,\qquad

1382:   \lim_{k\to\infty} \lambda_{\lin}(k) \,=\, -\infty.

1383: \end{align*}

1384: Since $\lambda_\infty^*<-1$ by \eqref{eq:def-lambda-infty-*} there is a

1385: unique value $k_0$ such that \eqref{eq:k0} holds.

1386: By \eqref{eq:k0}, \eqref{eq:ck-min} and \eqref{eq:eqs-ck} we deduce that

1387: \begin{gather}

1388:   c_{\lin}(k_0)\,=\, \bar{c}_{k_0}(\lambda_{\lin}(k_0))

1389:   \,=\, \bar{c}_{k_0}(\lambda^*_\infty)\,=\,

1390:   c_\infty,\label{eq:k0-c}

1391: \end{gather}

1392: and by \eqref{eq:eqs-ck}, \eqref{eq:k0-c} that

1393: \begin{gather}

1394:   c_{\lin}(k_0)\,=\,\bar{c}_{k_0}(\lambda_\infty(k_0)).

1395:   \label{eq:k0-lambda}

1396: \end{gather}

1397: By \eqref{eq:ck-min}, \eqref{eq:ck-min-uni} and \eqref{eq:k0} this yields

1398: \begin{gather*}

1399:   \lambda_\infty(k_0)\,=\, \lambda_\infty^*.

1400: \end{gather*}

1401: Finally one derives from \eqref{eq:k0}, \eqref{eq:recap-c-lin} and

1402: \eqref{eq:k0-c} that $k_0$ is given by

1403: \begin{gather*}

1404:   k_0\,=\, \frac{1+2\gamma}{\gamma^2}.

1405: \end{gather*}

1406: To prove the inequalities \eqref{eq:ineqs-k1}, \eqref{eq:ineqs-k2} we

1407: first observe that \eqref{eq:eqs-ck} implies that

1408: for $k\neq k_0$  there is a $\lambda$ between $\lambda_{\infty}^*$ and

1409: $\lambda_{\infty}(k)$ such that $\bar{c}_k^\prime(\lambda)=0$. By

1410: \eqref{eq:mon-ck-1}, \eqref{eq:mon-ck-2} we conclude that

1411: $\lambda=\lambda_{\lin}(k)$ and deduce that $\lambda_{\lin}(k)$

1412: lies between  $\lambda_{\infty}^*$ and

1413: $\lambda_{\infty}(k)$. Since $\lambda_{\lin}(k)$ is monotonically

1414: decreasing in $k$ we obtain from \eqref{eq:k0} that

1415: \begin{gather*}

1416:   \lambda_\infty(k_1) \,>\,\lambda_\infty^*,\qquad

1417:   \lambda_\infty(k_2) \,<\,\lambda_\infty^*,

1418: \end{gather*}

1419: which proves

1420: \eqref{eq:ineqs-k1}, \eqref{eq:ineqs-k2}.

1421: \end{proof}

1422: The conclusions of Lemma \ref{lem:decay} are illustrated in Figure

1423: \ref{fig:dec}. 

1424: \\

1425: \begin{figure}[h]

1426: \centering

1427: \includegraphics[height=75mm]{decay1.eps}

1428: \caption{Graph of the functions $\bar{c}_k$ for different values of $k$

1429:   and the decay rates $\lambda_\infty(k)$ of the travelling waves with

1430:   speed $c_\infty$ as found in Proposition \ref{prop-red}.}

1431: \label{fig:dec}

1432: \end{figure}

1433: 

1434: By the previous results we can now compare the decay of two different

1435: travelling waves as $x\to +\infty$, similarly as in Corollary

1436: \ref{cor:decay-} for the convergence to unity as $x\to -\infty$. 

1437: \begin{lemma}\label{lem:comp-decay}

1438: Let $k>k_0$ and assume that $(c,u,w)$ is a monotone travelling wave with speed

1439: $c<c_\infty$. Then, as $x\to +\infty$,

1440: $u(x),w(x)$ decay slower to zero

1441: than does the travelling wave $u_0(x),w_0(x)$ obtained in Proposition

1442: \ref{prop-red}. 

1443: \end{lemma}

1444: \begin{proof}

1445: By \eqref{eq:mon-ck-1} the function $\bar{c}_k$ is monotonically

1446: decreasing for $\lambda<\lambda_{\lin}(k)$. From \eqref{eq:ineqs-k2} and

1447: \eqref{eq:eqs-ck} we therefore deduce that

1448: \begin{gather*}

1449:   \bar{c}_k(\lambda)\,>\, \bar{c}_k(\lambda_\infty(k))\,=\,c_\infty\quad\text{ for all }

1450:   \lambda<\lambda_\infty(k).

1451: \end{gather*}

1452: Since we have assumed that $c<c_\infty$ this implies that the travelling

1453: wave $(c,u,w)$ decays with a rate $\lambda>\lambda_\infty(k)$. On the

1454: other hand, by Proposition \ref{prop-red}, the rate of the exponential

1455: deacy of $u_0,w_0$ as $x\to +\infty$ is given by

1456: $\lambda_\infty(k)$.

1457: \end{proof}

1458: %

1459: %

1460: %

1461: %

1462: \subsection{The nonlinear selection regime}

1463: In this section we prove that for $k>k_0$, where $k_0$ is given in Lemma

1464: \ref{lem:decay}, the minimal speed of 

1465: travelling waves for the reaction-diffusion system 

1466: \eqref{rd-u}, \eqref{rd-w} is identical to the minimal speed of travelling waves for the

1467: Stefan-like limit problem \eqref{st-w=1}-\eqref{st-Gamma}. In

1468: particular there is nonlinear selection of the minimal speed for $k>k_0$.

1469: This result follows from a comparison principle which is formulated in the 

1470: next Theorem. In general, invariant region arguments do not apply for elliptic systems,

1471: but here a shift parameter is chosen to play the role of

1472: the time parameter in the proof of comparison principles for parabolic systems.

1473: %A similar trick is used in the `moving plane technique', see \cite{Ber}. 

1474: \begin{theorem}\label{the-sel}

1475: Let $(c_1,u_1,w_1)$ and $(c_2,u_2,w_2)$ be two monotone travelling waves and assume that $c_1<c_2$.

1476: Let $\lambda_1,\lambda_2$ denote the decay rates at $+\infty$ of $u_1,w_1$ and $u_2,w_2$

1477: respectively. Then $\lambda_1\leq\lambda_2$ holds and, as $x$ tends to infinity, $u_1(x),w_1(x)$

1478: cannot converge exponentially slower to zero than $u_2(x),w_2(x)$ do.

1479: 

1480: In particular, a travelling wave has minimal speed if and only if its

1481: decay rate at $+\infty$ is the minimal one among all travelling waves.

1482: \end{theorem}

1483: \begin{proof}

1484: Let us assume that $\lambda_1>\lambda_2$ holds.

1485: Since $c_1<c_2$ the travelling wave $(u_1,w_1)$ converges by Corollary \ref{cor:decay-} 

1486: faster to $(1,1)$ as \mbox{$x\to -\infty$} than does $(u_2,w_2)$. In particular

1487: \begin{eqnarray*}

1488:   (u_1,w_1)(x) &>& (u_2,w_2)(x) \quad\text{ for sufficiently small }x<0.

1489: \end{eqnarray*}

1490: Since we have assumed that $\lambda_1>\lambda_2$, the decay of

1491: $(u_1,w_1)$ at $+\infty$ is slower than the decay of $(u_2,w_2)$ and we

1492: deduce from Lemma \ref{lem-decay-infty+}

1493: that

1494: \begin{eqnarray*}

1495:   (u_1,w_1)(x) &>& (u_2,w_2)(x)\quad\text{ for sufficiently large }x>0.

1496: \end{eqnarray*}

1497: This implies that there is a shift $x_0$, such that for $(\tilde{u}_1,\tilde{w}_1)=(u_1,w_1)(.+x_0)$ 

1498: \begin{gather*}

1499:         \tilde{u}_1 \,\geq\, u_2,\qquad \tilde{w}_1\,\geq\, w_2\quad\text{ on }\R

1500: \end{gather*}

1501: holds and such that there exists a $x_1\in\R$ with

1502: \begin{gather*}

1503:         \tilde{u}_1(x_1)\,=\,u_2(x_1)\quad \text{ or }\quad\tilde{w}_1(x_1)\,=\,w_2(x_1).

1504: \end{gather*}

1505: From equations \eqref{eq-tw-u}, \eqref{eq-tw-w} we obtain for $U:=u_2-\tilde{u}_1, W:=w_2-\tilde{w}_1$

1506: \begin{eqnarray}\label{eq-c,c0-U}

1507:         0 &=& U^{\prime\prime}+cU^\prime - (c_1-c_2)u_2^\prime-U+W -\gamma k (1-w_2)U +\gamma k \tilde{u}_1W,\\

1508:         0 &=& cW^\prime -(c_1-c_2)w_2^\prime +k(1-w_2)U-k\tilde{u}_1W.\label{eq-c,c0-W}

1509: \end{eqnarray}

1510: Assume that $\tilde{u}_1(x_1)=u_2(x_1)$, which gives

1511: \begin{gather*}

1512:         U(x_1) \,=\, \max_x U(x)\,=\,0

1513: \end{gather*}

1514: and $U^\prime(x_1)=0, U^{\prime\prime}(x_1)\leq 0$. Then \eqref{eq-c,c0-U} yields

1515: \begin{gather*}

1516:         0\,\leq\, -(c_1-c_2)u_2^\prime(x_1)\,<\,0,

1517: \end{gather*}

1518: which is a contradiction.\\

1519: If $\tilde{w}_1(x_1)=w_2(x_1)$ then

1520: \begin{gather*}

1521:         W(x_2) \,=\, \max_x W(x)\,=\,0

1522: \end{gather*}

1523: implies $W^\prime(x_2)=0$ and \eqref{eq-c,c0-W} yields

1524: \begin{gather*}

1525:         0\,\leq\, -(c_1-c_2)w_1^\prime(x_2)\,<\,0,

1526: \end{gather*}

1527: which is also a contradiction. Thus we deduce that

1528: $\lambda_1\leq\lambda_2$.

1529: 

1530: The final conclusion of the Theorem follows now  by a

1531: contradiction argument.

1532: \end{proof}

1533: The comparison principle Theorem \ref{the-sel} and Lemma \ref{lem:comp-decay} imply

1534: that for $k>k_0$ no monotone travelling wave  exists with lower speed than $c_\infty$.

1535: \begin{corollary}\label{cor:minspeed}

1536: For $k\geq k_0$ the minimal speed of travelling waves is given by

1537: \begin{eqnarray*}

1538:   c_{\min}(k) &=& c_\infty.

1539: \end{eqnarray*}

1540: \end{corollary}

1541: \begin{proof}

1542: Assume there is a monotone travelling wave $(c,u,w)$ with $c<c_\infty$ and let $(c_\infty,u_0,w_0)$ be the monotone travelling wave

1543: with speed $c_\infty$ which we have found in Proposition \ref{prop-red}.

1544: By Lemma \ref{lem:comp-decay} the functions $(u,w)$ decay  slower to zero at $+\infty$

1545: than $u_\infty,w_\infty$ do, which is a contradiction to Theorem \ref{the-sel}.

1546: Therefore $c_{\min}(k)\geq c_\infty$ holds and, recalling \eqref{eq-upper-bound}, the conclusion

1547: follows.

1548: \end{proof}

1549: %%===================================================

1550: % convergence of travelling waves

1551: %===================================================

1552: \section{Travelling waves in the fast-degradation-rate limit}

1553: \label{sec-conv}

1554: We complete our investigations by proving that the travelling waves of \eqref{rd-u},

1555: \eqref{rd-w} are, for large values of $k$, close to a travelling waves

1556: of the limit problem \eqref{st-w0}-\eqref{st-Gamma}.

1557: \begin{proposition}\label{prop:speed-lim-k}

1558: Let $(c,u_k,w_k)$, $k\in\N$, be a sequence of monotone travelling waves

1559: for \eqref{rd-u}, \eqref{rd-w} with speed $c\geq c_\infty$ and

1560: \begin{gather}

1561:   u_k(0)\,=\, \frac{1}{2}\quad\text{ for all }k>0. \label{eq:fix-u0}

1562: \end{gather}

1563: Then, as $k$ tends to infinity,

1564: \begin{alignat}{2}

1565:   u_k\,&\to\, U\quad&&\text{ uniformly, }\label{eq:conv-uk-U-1}\\

1566:   u_k\,&\to\, U\quad&&\text{ in }C^{0,\beta}_{\loc}(\R),\text{ for all

1567:   }0<\beta<1,\label{eq:conv-uk-U-2}\\ 

1568:   w_k\,&\to\, W\quad&&\text{ in }L^p_{\loc}(\R),\text{ for all }1\leq

1569:   p<\infty,\label{eq:conv-wk-W-1}

1570: \end{alignat}

1571: where $U,W$ is the unique travelling-wave solution of the limit

1572: problem with speed $c$ and $U(0)=1/2$.

1573: \end{proposition}

1574: \begin{proof}

1575: We recall that $u_k^\prime,w_k^\prime<0$, that $0<u_k,w_k<1$ and

1576: that by \eqref{eq:bound-u-eps-prime}, \eqref{eq:bound-w-eps-prime}

1577: \begin{gather}\label{est-tw-1}

1578:         -u_k^\prime \,<\,\mu

1579: \end{gather}

1580: holds uniformly in $k$, where $\mu$ was defined in \eqref{eq:def-mu}.

1581: This yields the existence of a subsequence $k_i\to\infty\ (i\to\infty)$

1582: and monotone decreasing functions $u,w$ with 

1583: $0\leq u,w\leq 1$ and $u\in H^{1,\infty}_{\loc}(\R)$ such that

1584: \begin{alignat}{3}

1585:         u_{k_i} &\to u \quad&&\text{ in } C^{0,\beta}_{\loc}(\R)&&\text{

1586:         for all }0<\beta<1,\label{conv-tw-u}\\ 

1587:         u_{k_i}^\prime &\schwto u^\prime \quad&&\text{ in

1588:         }L^p_{\loc}(\R)&&\text{ for all }1\leq p <\infty, 

1589:         \label{conv-tw-u-prime}\\

1590:         u_{k_i} &\to u \quad&&\text{ uniformly in }\R&&\text{ for all

1591:         }1\leq p <\infty, 

1592:         \label{conv-tw-u-uniform}\\

1593:         w_{k_i} &\to w \quad&&\text{ in }L^p_{\loc}(\R)&&\text{ for all

1594:         }1\leq p <\infty,\label{conv-tw-w},

1595: \end{alignat}

1596: Integrating equation \eqref{eq-tw-w} over $\R$, we obtain

1597: \begin{eqnarray}

1598:         \int_{-\infty}^\infty u_{k_i}(1-w_{k_i}) &=&

1599:         \frac{c}{k_i},\label{bound-L1-kterm} 

1600: \end{eqnarray}

1601: and by Fatou's Lemma we see that

1602: \begin{gather*}

1603:   \int_{-\infty}^\infty u(1-w)\,\leq\,

1604:   \liminf_{i\to\infty}\int_{-\infty}^\infty u_{k_i}(1-w_{k_i})\,=\,0, 

1605: \end{gather*}

1606: which implies that

1607: \begin{eqnarray}

1608:   u(1-w) &=& 0\quad\text{ almost everywhere in }\R.\label{eq:tw-lim-00}

1609: \end{eqnarray}

1610: By \eqref{eq:conv-uk-U-2}, \eqref{conv-tw-u} we obtain that $u(0)=1/2$

1611: and $u(x)\geq 1/2$ for $x<0$. From \eqref{eq:tw-lim-00} 

1612: we deduce that

1613: \begin{eqnarray*}

1614:   w(x) &=& 1 \quad\text{ for }x<0.

1615: \end{eqnarray*}

1616: The equations \eqref{eq-tw-u} and \eqref{est-tw-1},

1617: \eqref{bound-L1-kterm} yield the estimate 

1618: \begin{eqnarray}

1619:         \int_{-\infty}^0 (1-u_{k_i}) &=& -u_{k_i}^\prime(0) -

1620:         cu_{k_i}(0) +c +\int_{-\infty}^0 (1-w_{k_i}) 

1621:         +\int_{-\infty}^0\gamma ku_{k_i}(1-w_{k_i})

1622:         \notag\\

1623:         &\leq& C(c,\gamma)\label{est-tw-1-u}

1624: \end{eqnarray}

1625: and we obtain that $(1-u)\in L^1\big((-\infty,0)\big)$ and $u(x)\to 1$

1626: as $x\to -\infty$. 

1627: By \eqref{eq-tw-u} and \eqref{est-tw-1}, \eqref{bound-L1-kterm}

1628: \begin{gather}

1629:         \int_0^\infty (u_{k_i}-w_{k_i}) \,=\, -u_{k_i}^\prime(0)

1630:         -\frac{c}{2} -\gamma k_i \int_0^\infty u_{k_i}(1-w_{k_i}) 

1631:         \,\geq\, -C(\gamma,c)\label{est-tw-u}

1632: \end{gather}

1633: holds and we further deduce from \eqref{bound-L1-kterm} that

1634: \begin{gather}\label{est-tw-L1-sum}

1635:         \int_0^\infty (u_{k_i} -\frac{1}{2}w_{k_i}) \,\leq\,

1636:         \int_0^\infty u_{k_i}(1-w_{k_i})\,\leq\, \frac{c}{k_i}, 

1637: \end{gather}

1638: which gives, substracting \eqref{est-tw-u},

1639: \begin{eqnarray}\label{est-tw-L1-w}

1640:         \int_0^\infty w_{k_i} &\leq& C(\gamma,c)

1641: \end{eqnarray}

1642: for $k>1$. Further we find from \eqref{est-tw-L1-sum} that

1643: \begin{eqnarray}\label{est-tw-L1-u}

1644:         \int_0^\infty u_{k_i} &\leq& C(\gamma,c)

1645: \end{eqnarray}

1646: holds. By Fatou's Lemma this implies that $u,w\in L^1((0,\infty))$ and

1647: $u(x),w(x)\to 0$ as $x\to\infty$. By \eqref{conv-tw-u} and since the

1648: limits as $x\to\pm\infty$ of $u_{k_i}, U$ coincide, we deduce

1649: \eqref{eq:conv-uk-U-1} from \cite[Lemma 2.4]{odo}.\\

1650: The equations \eqref{eq-tw-u}, \eqref{eq-tw-w} yield that $u_{k_i},

1651: w_{k_i}$ satisfy 

1652: \begin{eqnarray*}

1653:         0 &=& \int_\R \big(u_{k_i}^\prime + c(u_{k_i}+\gamma

1654:         w_{k_i})\big)\eta^\prime + (u_{k_i}-w_{k_i})\eta 

1655: \end{eqnarray*}

1656: for all $\eta\in C^\infty_c(\R)$ and, according to

1657: \eqref{conv-tw-u}-\eqref{conv-tw-w}, we can pass in this 

1658: equation to the limit $i\to\infty$. This yields

1659: \begin{eqnarray}

1660:         0 &=& \int_\R \big(u^\prime + c(u+\gamma w)\big)\eta^\prime +

1661:         (u-w)\eta.\label{eq:tw-lim-1} 

1662: \end{eqnarray}

1663: Since $u,w$ are monotone decreasing from unity to zero and satisfy

1664: \eqref{eq:tw-lim-00}, we deduce that there is a $x_0\in\R$ such that

1665: \begin{alignat*}{2}

1666:   u(x)\,&=\,0&\qquad&\text{ for }x\geq x_0,\\

1667:   u(x)\,&>\,0,\, w(x)\,=\,1&\quad&\text{ for }x<x_0.

1668: \end{alignat*}

1669: Therefore \eqref{eq:tw-lim-1} yields that

1670: \begin{alignat*}{2}

1671:   0\,&=\, u^{\prime\prime}+cu^\prime -u +1&\quad&\text{

1672:     for }x<x_0,\\

1673:   0\,&=\, \gamma c w^\prime + w&\quad&\text{ for }x>x_0,

1674: \end{alignat*}

1675: and that the jump condition

1676: \begin{align*}

1677:   \gamma c (1-\lim_{x\downarrow x_0}w(x)) &\,=\,

1678:     -\lim_{x\uparrow x_0}u^\prime(x)

1679: \end{align*}

1680: has to be satisfied. This shows that $u,w$ is a travelling wave with

1681: speed $c$ of the limit system \eqref{st-w=1}, \eqref{st-u=0}.

1682: \end{proof}

1683: %=======================================================

1684: % conclusions

1685: %=======================================================

1686: \section{Conclusions}\label{sec-concl}

1687: We conclude our investigations with a brief summary and discussion of

1688: our results. 

1689: \subsection{Slow and fast decay at the threshold $c=c_\infty$}

1690: Our results on the existence of travelling waves for the system

1691: \eqref{rd-u}, \eqref{rd-w} and the selection mechanism of the minimal

1692: speed are summarized in Figure \ref{fig:dec2}. As in the preceding

1693: figure we have plotted the functions $\bar{c}_k$ which give for a

1694: travelling wave with speed $c$ the possible rates $\lambda$ of the

1695: exponential decay to zero at $x\to +\infty$. By the circles, squares and

1696: diamonds in

1697: Figure \ref{fig:dec2} we have indicated the decay rates which in fact

1698: are realized by a travelling wave: presuming that there is linear

1699: selection for $k<k_0$ travelling waves exists

1700: for all speeds $c\geq c_{\lin}$.

1701: By Theorem \ref{the-sel} the decay rates of these

1702: travelling waves corresponds to values on the increasing branch of

1703: $(\lambda,c_k(\lambda))$.

1704: This behaviour changes for $k> k_0$: the decay rate of $u_0,w_0$ is

1705: on the decreasing branch of the solution curve. By Corollary \ref{cor:minspeed}

1706: $c_\infty$ is the minimal speed and the decay rates of travelling waves with larger speeds

1707: are on the increasing branch.

1708: \\

1709: \begin{figure}[h]

1710: \centering

1711: \includegraphics[height=75mm]{decay2.eps}

1712: \caption{Graph of the functions $\bar{c}_k$ for different values of $k$ and the

1713:   points on the graph that are realized by a travelling wave.}

1714: \label{fig:dec2}

1715: \end{figure}

1716: 

1717: To give an explanation of what happens in the nonlinear selection regime

1718: $k>k_0$ if the speed $c$ falls below the minimal speed

1719: $c_{\min}(k)=c_\infty$ we consider the linearization of \eqref{rd-u},

1720: \eqref{rd-w} at $(u^\prime,u,w)=(0,0,0)$, which was given in \eqref{eq:xi-lin}. For all speeds

1721: $c$ in a neighbourhood of $c_\infty$ the linearized system has two

1722: negative eigenvalues; the stable manifold for \eqref{rd-u},

1723: \eqref{rd-w} at $(u^\prime,u,w)=(0,0,0)$ is two-dimensional. One checks that the

1724: eigenspace corresponding to a negative eigenvalue intersects with the

1725: set $\{z=(z_1,z_2,z_3) : z_1<0, z_2>0, z_3>0 \}$. A monotone travelling

1726: wave exists for $c>c_0$ and approaches $(0,0,0)$ tangentially to the

1727: eigenspace corresponding to the larger negative eigenvalue (`slow

1728: decay'), see Figure \ref{fig:dec2}. We expect also for $c<c_\infty$ an orbit

1729: $(u^\prime,u,w)$ connecting $(0,1,1)$ with $(0,0,0)$. Such an orbit will

1730: also converge to $(0,0,0)$ tangentially to the eigenspace corresponding

1731: to the larger negative eigenvalue but comes from the `wrong' side, 

1732: taking negative values for $u,w$. For the threshold $c=c_\infty$ there

1733: still exists a monotone travelling wave; this travelling wave

1734: approaches $(0,0,0)$ tangentially to the eigenspace of the smaller

1735: negative eigenvalue (`fast decay'). 

1736: 

1737: \subsection{Discussion}

1738: Travelling waves often determine the long-time behaviour of solutions for

1739: the initial-value problem for \eqref{rd-u}, \eqref{rd-w}. Typically,

1740: for solutions with sufficiently fast decaying initial data, the

1741: propagation speed of pertubations from the unstable equilibrium 

1742: is given by the minimal speed of travelling waves. The proof of

1743: such a result, as well as the uniqueness of travelling waves,

1744: for the system \eqref{rd-u}, \eqref{rd-w} is not in the

1745: scope of the present article. Nevertheless our result that the

1746: travelling wave with 

1747: minimal speed has the fastest decay supports that conjecture.

1748: 

1749: For the reaction-diffusion system \eqref{rd-u}, \eqref{rd-w} in

1750: arbitrary space-dimension $n$ our analysis yields a family of

1751: supersolutions: Consider for  $c\geq

1752: c_\infty$ a travelling 

1753: wave $u,w$ with speed $c$ and an arbitrary real number $r\in \R$. The

1754: functions $\bar{u},\bar{w}$ defined by 

1755: \begin{gather*}

1756:   \bar{u}(x,t)\,:=\, u(|x|-ct-r),\quad \bar{w}(x,t)\,:=\, w(|x|-ct-r)

1757: \end{gather*}

1758: satisfy

1759: \begin{align}

1760:   \partial_t {u} - \Delta{u} + {u} - {w} +\gamma k {u}(1-{w})\,&=\,

1761:   -\frac{n-1}{|x|} \bar{u}^\prime,\label{eq:n-dim-rd}\\

1762:   \partial_t {w} - k {u}(1-{w})\,&=\, 0.\notag

1763: \end{align}

1764: Since $\bar{u}^\prime$ is negative $(\bar{u},\bar{w})$ is

1765: a supersolution.

1766: In order to construct a subsolution one has to control the

1767:  dimension-depending correction term in \eqref{eq:n-dim-rd}.

1768: 

1769: In view of the applications, the robustness of the wave-speed to changes

1770: in the parameter values is a valuable feature: see the explicit formula

1771: \eqref{eq-c-infty} for the minimal speed and Proposition

1772: \ref{prop:speed-lim-k}. For the mathematical

1773: analysis of reaction-diffusion  \emph{systems} and the selection of the

1774: minimal speed the model that we have derived is a good paradigm. We

1775: prove that nonlinear selection occurs and determine explicitly the

1776: minimal speed. One crucial ingredient

1777: is the exact first integral obtained in Lemma

1778: \ref{lem:red}, which is a special property of the system \eqref{rd-u},

1779: \eqref{rd-w}. Other results, in particular the comparison principle

1780: Theorem \ref{the-sel} and the observation that the fastest decay is

1781: realized by a travelling wave with minimal speed, can be extended to general

1782: monotone systems. 

1783: %%%%%%%%%%%%%%%%%%%%%%%%

1784: % bibliography

1785: %%%%%%%%%%%%%%%%%%%%%%%%

1786: \newcommand{\etalchar}[1]{$^{#1}$}

1787: \begin{thebibliography}{WKK{\etalchar{+}}04}

1788: 

1789: \bibitem[AW75]{AW}

1790: D.~G. Aronson and H.~F. Weinberger.

1791: \newblock Nonlinear diffusion in population genetics, combustion, and nerve

1792:   pulse propagation.

1793: \newblock In {\em Partial differential equations and related topics (Program,

1794:   Tulane Univ., New Orleans, La., 1974)}, pages 5--49. Lecture Notes in Math.,

1795:   Vol. 446. Springer, Berlin, 1975.

1796: 

1797: \bibitem[BD94]{BD}

1798: R.~D. Benguria and M.~C. Depassier.

1799: \newblock Validity of the linear speed selection mechanism for fronts of the

1800:   nonlinear diffusion equation.

1801: \newblock {\em Phys. Rev. Lett.}, 73:2272--2274, 1994.

1802: 

1803: \bibitem[CL55]{CL}

1804: E.~A. Coddington and N.~Levinson.

1805: \newblock {\em Theory of ordinary differential equations}.

1806: \newblock McGraw-Hill Book Company, New York-Toronto-London, 1955.

1807: 

1808: \bibitem[Die77]{odo}

1809: O.~Diekmann.

1810: \newblock Limiting behaviour in an epidemic model.

1811: \newblock {\em Nonlinear Anal.}, 1:459--470, 1976/77.

1812: 

1813: \bibitem[EvS00]{ES}

1814: U.~Ebert and W.~van Saarloos.

1815: \newblock Front propagation into unstable states: universal algebraic

1816:   convergence towards uniformly translating pulled fronts.

1817: \newblock {\em Phys. D}, 146:1--99, 2000.

1818: 

1819: \bibitem[Fis37]{Fi}

1820: R.~A. Fisher.

1821: \newblock The wave of advance of advantageous genes.

1822: \newblock {\em Ann. Eugenics}, 7:355--369, 1937.

1823: 

1824: \bibitem[HKR07]{HKR}

1825: D.~Hilhorst, J.~R. King, and M.~R{\"o}ger.

1826: \newblock Mathematical analysis of a model describing the invasion of bacteria

1827:   in burn wounds.

1828: \newblock {\em Nonlinear Anal. (TMA)}, 66:1118--1140, 2007.

1829: 

1830: \bibitem[HR75]{HR}

1831: K.~P. Hadeler and F.~Rothe.

1832: \newblock Travelling fronts in nonlinear diffusion equations.

1833: \newblock {\em J. Math. Biol.}, 2:251--263, 1975.

1834: 

1835: \bibitem[KKC{\etalchar{+}}03]{Ki}

1836: J.~R. King, A.~J. Koerber, J.~M. Croft, J.~P. Ward, R.~E. Sockett, and

1837:   P.~Williams.

1838: \newblock {Modelling host tissue degradation by extracellular bacterial

1839:   pathogens}.

1840: \newblock {\em Math. Med. Biol.}, 20:227--260, 2003.

1841: 

1842: \bibitem[KPP37]{KPP}

1843: A.~N. Kolmogorov, I.~G. Petrovsky, and N.~S. Piskunov.

1844: \newblock Etude de l'e\'quation de la diffusion avec croissance de la

1845:   quantit\'e de mati\`ere et son application \`a un probl\`eme biologique.

1846: \newblock {\em Moskow Univ. Math. Bull.}, 1:1--25, 1937.

1847: 

1848: \bibitem[LMN04]{LMN}

1849: M.~Lucia, C.~B. Muratov, and M.~Novaga.

1850: \newblock Linear vs.\ nonlinear selection for the propagation speed of the

1851:   solutions of scalar reaction-diffusion equations invading an unstable

1852:   equilibrium.

1853: \newblock {\em Comm. Pure Appl. Math.}, 57:616--636, 2004.

1854: 

1855: \bibitem[Mur04]{Mur}

1856: C.~B. Muratov.

1857: \newblock A global variational structure and propagation of disturbances in

1858:   reaction-diffusion systems of gradient type.

1859: \newblock {\em Discrete Contin. Dyn. Syst. Ser. B}, 4:867--892, 2004.

1860: 

1861: \bibitem[Per01]{Pe}

1862: L.~Perko.

1863: \newblock {\em Differential equations and dynamical systems}, volume~7 of {\em

1864:   Texts in Applied Mathematics}.

1865: \newblock Springer-Verlag, New York, 2001.

1866: 

1867: \bibitem[Roq97]{Roq}

1868: J.-M. Roquejoffre.

1869: \newblock Eventual monotonicity and convergence to travelling fronts for the

1870:   solutions of parabolic equations in cylinders.

1871: \newblock {\em Ann. Inst. H. Poincar\'e Anal. Non Lin\'eaire}, 14:499--552,

1872:   1997.

1873: 

1874: \bibitem[Rot81]{Rot}

1875: F.~Rothe.

1876: \newblock Convergence to pushed fronts.

1877: \newblock {\em Rocky Mountain J. Math.}, 11:617--633, 1981.

1878: 

1879: \bibitem[Sto76]{Sto}

1880: A.~N. Stokes.

1881: \newblock On two types of moving front in quasilinear diffusion.

1882: \newblock {\em Math. Biosci.}, 31:307--315, 1976.

1883: 

1884: \bibitem[vS89]{Saa}

1885: W.~van Saarloos.

1886: \newblock Front propagation into unstable states. {II}. {L}inear versus

1887:   nonlinear marginal stability and rate of convergence.

1888: \newblock {\em Phys. Rev. A}, 39:6367--6390, 1989.

1889: 

1890: \bibitem[vS03]{Saa-rev}

1891: W.~van Saarloos.

1892: \newblock Front propagation into unstable states.

1893: \newblock {\em Physics Reports}, 386:29--222, 2003.

1894: 

1895: \bibitem[VVV94]{Vo}

1896: A.~I. Volpert, V.~A. Volpert, and V.~A. Volpert.

1897: \newblock {\em Traveling wave solutions of parabolic systems}, volume 140 of

1898:   {\em Translations of Mathematical Monographs}.

1899: \newblock American Mathematical Society, Providence, RI, 1994.

1900: 

1901: \bibitem[WKK{\etalchar{+}}04]{Wa}

1902: J.~P. Ward, J.~R. King, A.~J. Koerber, J.~M. Croft, R.~E. Sockett, and

1903:   P.~Williams.

1904: \newblock {Cell-signalling repression in bacterial quorum sensing}.

1905: \newblock {\em Math. Med. Biol.}, 21:169--204, 2004.

1906: 

1907: \end{thebibliography}

1908: 

1909: \end{document}

1910: