1: \documentclass[aps,pre,twocolumn,amsfonts,amssymb,amsmath,floatfix,showpacs]{revtex4}
2: \usepackage[dvips]{graphicx}
3: \usepackage{color}
4:
5: \addtolength{\topmargin}{10mm}
6:
7: % Keystroke saving macros
8: \newcommand{\beps}{\varepsilon}
9: \newcommand{\const}{\mathrm{const}}
10: \newcommand{\crit}{_{\mathrm{cr}}}
11: \newcommand{\dt}{h_t}
12: \newcommand{\dx}{h_x}
13: \newcommand{\Estim}{E_{\mathrm{stim}}}
14: \newcommand{\f}{f}
15: \newcommand{\Heav}{\Theta}
16: \newcommand{\INa}{I_{\mathrm{Na}}}
17: \newcommand{\ms}{\mathrm{ms}}
18: \newcommand{\mV}{\mathrm{mV}}
19: \renewcommand{\O}[1]{\mathcal{O}\left(#1\right)}
20: \newcommand{\ustim}{u_{\mathrm{stim}}}
21: \newcommand{\Vstim}{V_{\mathrm{stim}}}
22: \newcommand{\xf}{x_f}
23: \newcommand{\xm}{\Delta}
24: \newcommand{\xstim}{x_{\mathrm{stim}}}
25:
26: \newtheorem{conjecture}{Conjecture}
27:
28: \def\eq(#1){(\ref{eq:#1})}
29: \def\eqtwo(#1,#2){(\ref{eq:#1},\ref{eq:#2})}
30: \def\eqthree(#1,#2,#3){(\ref{eq:#1},\ref{eq:#2},\ref{eq:#3})}
31: \newcommand{\Fig}[1]{Fig.~\ref{fig:#1}}
32: \newcommand{\fig}[1]{Fig.~\ref{fig:#1}}
33: \newcommand{\figs}[1]{Figs.~\ref{fig:#1}}
34:
35: \newcommand{\NB}[1]{\textbf{\color{red}#1}}
36:
37: \begin{document}
38: \title{Critical fronts in initiation of excitation waves}
39: \author{I. Idris}
40: \author{V. N. Biktashev}
41: \affiliation{Department of Mathematical Sciences,
42: University of Liverpool, Liverpool L69 7ZL, UK }
43: \date{\today}
44: \begin{abstract}
45: We consider the problem of initiation of propagating wave
46: in a one-dimensional excitable fiber.
47: In the FitzHugh-Nagumo theory, the key role is played by
48: ``critical nucleus'' and ``critical pulse'' solutions whose (center-)stable
49: manifold is the threshold surface separating initial conditions
50: leading to propagation and those leading to decay.
51: We present evidence that in cardiac excitation models,
52: this role is played by ``critical front'' solutions.
53: \end{abstract}
54: \pacs{%
55: 87.19.Hh% Cardiac dynamics
56: , 87.19.La% Neuroscience
57: , 02.90.+p% Other topics in mathematical methods in physics
58: }
59: \maketitle
60:
61: %##############################
62: \section{Introduction}
63:
64: An excitable medium is a thermodynamically non-equilibrium system that
65: has a stable spatially uniform ``resting state'', but responds to an
66: above-threshold localized stimulus by a propagating non-decaying
67: ``excitation wave''. Excitation waves play key roles in living
68: organisms and are observed in chemical and physical systems,
69: e.g. nerves, heart muscle, catalytic redox reactions, large aspect
70: lasers and star formation in galaxies
71: \cite{Krinsky-Swinney-1991}. Understanding conditions of successful
72: initiation of excitation waves is particularly important for heart
73: where such waves trigger coordinated contraction of the muscle and
74: where a failure of initiation can cause or contribute to serious or
75: fatal medical conditions, or render inefficient the work of pacemakers
76: or defibrillators \cite{Zipes-Jalife-2000}.
77:
78: The theoretical understanding of excitability stems from FitzHugh's
79: simplified model of a nerve membrane~\cite{FitzHugh-1961}. One of his
80: key concepts is ``quasi-threshold'', which gets precise in the limit
81: of large time scale separation between the processes of excitation and
82: recovery. Then the fast subsystem has unstable ``threshold''
83: equilibria. Initial conditions below such an equilibrium lead to decay,
84: and those above it lead to excitation.
85:
86: In a spatially extended FitzHugh-Nagumo (FHN)
87: system~\cite{FitzHugh-1961,Nagumo-etal-1962}, the ability of a
88: stimulus to initiate a wave depends also on its spatial extent, the
89: aspect summarized by Rushton's~\cite{Rushton-1937} concept of
90: ``liminal length''. More generic is the concept of
91: ``critical curve'' in the stimulus strength-spatial extent plane
92: (see \fig{fhn}). A stimulus initiates a wave if its parameters are above this
93: curve, or leads to decay if below.
94:
95: Mathematically, the problem is about classification of initial
96: conditions that will or will not lead to a traveling wave
97: solution. The key question is the nature of the boundary between the
98: two classes. A detailed analysis of this boundary has been done for
99: the FitzHugh-Nagumo system and its variations. This has led to the
100: concept of a \emph{critical nucleus}, discussed below in more
101: detail. Roughly, this is a spatially extended analog of a threshold
102: equilibrium in the point system: critical nucleus is also a stationary
103: but unstable solution, and its small perturbations lead to either
104: initiation of an excitation wave, for perturbations of one direction,
105: or to decay, for perturbations of the opposite direction.
106:
107: We stress that although the role of FHN as a universal prototype of
108: excitable systems has been disputed, to our knowledge, there are still
109: no alternatives to the critical nucleus concept, as far as initiation
110: problem is concerned.
111:
112: In this paper, we present evidence that cardiac excitation provides an
113: example of alternative type of system, in which there is no place for
114: the critical nucleus. Two independent observations led to this
115: study. Firstly, numerical simulations of the cardiac excitation models
116: reveal significant qualitative differences in the way initiation
117: occurs in such models, compared to the FHN-style
118: systems~\cite{Starmer-etal-2003}. Secondly, asymptotic analysis of
119: detailed cardiac excitation models reveals that in the fast subsystem
120: there, there is \emph{no analog} of the unstable threshold
121: equilibrium of FHN systems~\cite{Biktashev-Suckley-2004}, and the
122: threshold there has a completely different mathematical nature.
123: Further, elementary arguments show that in cardiac equations
124: there are no nontrivial stationary solutions that could play the role
125: similar to the critical nucleus in FHN system.
126:
127: Thus we have a theoretical vacuum here. Obviously, one cannot even
128: begin to think about investigating initiation criteria without
129: understanding the nature of the critical solutions. This paper aims to
130: fill this vacuum, and clarify the nature of the critical solutions.
131: We analyze a simplified model of cardiac excitation, and use the knowledge
132: of its exact solutions to demonstrate
133: that for this model the concept of critical nucleus should be replaced
134: with a new concept of \emph{critical front}.
135: We also confirm numerically the relevance of this new concept on an example of
136: a detailed ionic cardiac excitation model.
137:
138:
139: %##############################
140: \section{FitzHugh-Nagumo system}
141:
142: First we recapitulate some known theoretical concepts related to
143: initiation of waves
144: (see e.g. \cite{Flores-1989,Flores-1991} and references therein).
145: We consider FitzHugh-Nagumo system in the form
146: %
147: \begin{align}
148: u_t &= u_{xx}+\f(u)-v , \qquad \f(u) = u(u-\theta)(1-u) \nonumber\\
149: v_t &= \beps (\alpha u-v) \label{eq:fhn1}
150: \end{align}
151: %
152: where $\beps>0$, $\alpha>0$, $\theta\in(0,1/2)$
153: (some works consider piecewise linear functions $\f$ of similar shape)
154: on a half-fiber, $(x,t)\in[0,\infty)\times[0,\infty)$ with a no-flux boundary
155: %
156: \begin{equation}
157: u_x(0,t)=0 , \label{eq:fhn1b}
158: \end{equation}
159: %
160: and a rectangular
161: initial perturbation of width $\xstim$ and amplitude $\ustim$,
162: %
163: \begin{equation}
164: u(x,0) = \ustim\Heav(\xstim-x), \qquad
165: v(x,0) = 0 \label{eq:fhn1i} \\
166: \end{equation}
167: where $\Heav(\cdot)$ is the Heaviside step function.
168:
169: %%%%%%%%%%%%%%%%%%
170: \begin{figure*}[ht]
171: \includegraphics{fig1.eps}
172: \caption[]{
173: Initiation of excitation in FitzHugh-Nagumo system.
174: %
175: (a,b) Full system \eqthree(fhn1,fhn1b,fhn1i) ``FHN'' for parameter values:
176: $\alpha=0.37$, $\theta=0.13$, $\beps=0.02$.
177: Stimulation parameters: $\xstim=2.10$ for both,
178: below-threshold $\ustim=0.43$, leading to decay, for (a) and above-threshold $\ustim=0.44$,
179: leading to initiation of excitation propagation, for (b).
180: %
181: (c,d) Fast subsystem \eqthree(zfk,zfkb,zfki) ``ZFK'':
182: same parameters as in (a,b) except $\beps=0$.
183: Stimulation: $\xstim=2.10$ for both, below-threshold
184: $\ustim=0.3304831$ for (c) and above-threshold
185: $\ustim=0.3304833$ for (d).
186: Bold black lines: initial conditions.
187: (e) The corresponding critical curves,
188: separating initiation initial conditions from decay initial conditions.
189: Simulation done on an interval $x\in[0,L]$, $L=120$ with Neuman
190: boundaries, central space differencing with $\dx=0.15$, and explicit
191: Euler timestepping with $\dt=0.01$.
192: }\label{fig:fhn}
193: \end{figure*}
194: %%%%%%%%%%%%%%%%%%
195:
196: \Fig{fhn}(a,b) shows two typical results of the initiation process: a
197: successful initiation, leading to generation of a propagating pulse,
198: and an unsuccessful, leading to decay of excitation in the whole
199: half-fiber into the resting state.
200:
201: If $\beps=0$, problem
202: \eqthree(fhn1,fhn1b,fhn1i) reduces to an initiation problem
203: for Zeldovich-Frank-Kamenetsky (ZFK) equation \cite{Zeldovich-FrankKamenetsky-1938}
204: also known as Nagumo equation \cite{McKean-1970},
205: %
206: \begin{eqnarray}
207: u_t &=& u_{xx}+\f(u), \label{eq:zfk}\\
208: u_x(0,t)&=& 0, \label{eq:zfkb}\\
209: u(x,0) &=& \ustim\Heav(\xstim-x). \label{eq:zfki}
210: \end{eqnarray}
211: %
212: \Fig{fhn}(c,d) illustrates the initiation and its failure in this
213: reduced problem. Instead of a propagating pulse, successful
214: initiation produces a propagating front. In the full model with small $\beps$, this front
215: is followed, in time scale $\O{\beps^{-1}}$, by a wave-back to
216: form a full excitation pulse.
217:
218: A key role in understanding initiation belongs to
219: an unstable nontrivial bounded time-independent solution of
220: \eqtwo(zfk,zfkb), sometimes called \emph{critical nucleus},
221: by analogy with phase transitions theory.
222: Such solution is unique;
223: for a cubical nonlinearity $\f$ as in \eq(fhn1),
224: this solution has the form
225: %
226: \(
227: u\crit(x)=3\theta\sqrt 2%
228: \left[(1+\theta)\sqrt{2}+\cosh(x\sqrt \theta)\sqrt{2-5\theta+\theta^2}\right]^{-1}
229: \).
230: %
231: Its linearization spectrum has exactly one unstable
232: eigenvalue, while all other eigenvalues are stable. So the stable
233: manifold of this stationary solution has codimension one, and divides
234: the phase space of \eqtwo(zfk,zfkb) to two open sets. One of these
235: sets corresponds to initial conditions leading to successful
236: initiation, and the other to decay. In particular, if the initial
237: condition satisfies $u(x,0)<u\crit(x)$, $x\in[0,\infty)$ then $u(x,t)$ decays as
238: $t\to\infty$, and if $u(x,0)>u\crit(x)$, $x\in[0,\infty)$ then
239: $u(x,t)$ approaches a stable propagating front solution.
240: Moreover, if a continuous one-parametric
241: family of initial conditions contains some that initiate a wave and
242: some that lead to decay, then there is always at least one that does
243: neither, but gives a solution that approaches the critical nucleus.
244: This critical nucleus is the same for all such families, e.g. does not
245: depend on the shape of the initial distribution $u(x,0)$, as long as
246: its amplitude is at the threshold corresponding to that shape. Initial conditions
247: very close to the threshold generate solutions which approach the
248: critical nucleus and then depart from it, either toward propagation
249: or toward decay. This transient stationary state can be seen in
250: \fig{fhn}(c,d) where the initial conditions are selected very close to
251: the threshold.
252:
253: %%%%%%%%%%%%%%%%%%%
254: \begin{figure*}[ht]
255: \includegraphics{fig2.eps}
256: \caption[]{
257: The critical pulse is a universal transient for any near-threshold
258: initial condition.
259: The solutions to \eq(fhn1) for slightly below-threshold (a,c)
260: and slightly above-threshold (b,d) amplitudes, for
261: smaller stimulus width $\xstim=2.10$ in (a,b) and larger
262: $\xstim=10.05$ in (c,d).
263: Parameter values:
264: $\beps=0.02$,
265: $\alpha =0.37$,
266: $\dt =0.01$,
267: $\dx =0.15$,
268: $L=120$.
269: Stimulus amplitudes:
270: $\ustim=0.431929399574766$ for (a),
271: $0.431929399574768$ for (b),
272: $0.191802079312694$ for (c) and
273: $0.191802079312696$ for (d).
274: In all cases we see a slow, low-amplitude unstable propagating pulse
275: which subsequently either decays or evolves into a fast,
276: high-amplitude stable propagating pulse.
277: }\label{fig:critpulse}
278: \end{figure*}
279: %%%%%%%%%%%%%%%%%%%
280:
281: For small $\beps>0$, system \eq(fhn1) does not have nontrivial
282: stationary solutions.
283: However, for $x\in(-\infty,\infty)$ equation \eq(fhn1)
284: has an unstable
285: propagating pulse solution $\tilde{u}\crit(x-ct)$,
286: $\tilde{v}\crit(x-ct)$ such that $\tilde{u}\crit(x)\to u\crit(x)$,
287: $\tilde{v}\crit(x)\to0$ and $c=\O{\beps^{1/2}}$ as
288: $\beps\searrow0$.
289: Due to translational symmetry, this solution (in the comoving frame of reference)
290: has a zero eigenvalue
291: corresponding to the eigenfunction $(\tilde{u}\crit',\tilde{v}\crit')$.
292: This solution also has a single unstable eigenvalue.
293: So its center-stable
294: manifold has codimension one and is the threshold hypersurface
295: dividing the phase space into the decay domain and the initiation
296: domain. So here we have a \emph{critical pulse} solution, which we define as
297: an unstable traveling wave that is asymptotic to the resting
298: state for both limits $x-ct\to\pm\infty$.
299: For small $\beps$, the critical pulse is
300: essentially a slowly traveling variant of the critical nucleus. Any
301: solution with the initial condition at the threshold hypersurface will
302: asymptotically approach this critical pulse (suitably shifted),
303: and any solutions starting close to the threshold will approach this
304: critical pulse as a transient~\endnote{
305: The symmetry $x \leftrightarrow -x$ means there are two stable pulse solutions,
306: one propagating to the right and one propagating to the left.
307: Likewise, there are two critical pulse solutions, and two
308: center-stable manifolds. The relationship between these two
309: critical hypersurfaces is complicated, since some families of
310: initial conditions can generate two oppositely traveling pulses
311: and some can generate only one.
312: }. This is illustrated in \fig{critpulse}.
313:
314: With this understanding, the excitation condition in terms of
315: $(\xstim,\ustim)$ reduces to computing the intersection of the
316: two-parametric manifold described by \eq(fhn1i) with the codimension 1
317: stable (center-stable) manifold of the critical nucleus (critical
318: pulse). This gives the curve on the $(\xstim,\ustim)$ plane separating
319: initial conditions leading to excitation propagation from those
320: leading to decay. This can be done numerically or,
321: with appropriate simplifications, analytically. An example of dealing
322: with this problem in the ZFK equation, using Galerkin style
323: approximations can be found in \cite{Neu-etal-1997}.
324: Here we concentrate on the principal question, about the nature of the
325: threshold hypersurface in the functional space, for cardiac excitation
326: equations. It appears that in cardiac equations, this nature is
327: different from the FitzHugh-Nagumo theory just considered.
328:
329: %##############################
330: \section{Simplified cardiac excitation model}
331:
332: Now we consider the simplified model of $\INa$-driven excitation
333: fronts in typical cardiac excitation models proposed in
334: \cite{Biktashev-2002}:
335: %
336: \begin{equation}
337: E_t = E_{xx}+\Heav(E-1)h, \quad
338: h_t = (\Heav(-E)-h)/\tau \label{eq:bik}
339: \end{equation}
340: with boundary condition
341: %
342: \begin{equation}
343: E_x(0,t) =0 \label{eq:bik1b}
344: \end{equation}
345: %
346: and initial condition
347: %
348: \begin{eqnarray}
349: E(x,0) = -\alpha+\Estim\Heav(\xstim-x), \quad
350: h(x,0) = 1, \label{eq:bik1a}
351: \end{eqnarray}
352: %
353: where the variable $E$ represents transmembrane potential of the
354: cardiac tissue, $h$ is the probability of the Na-gates being
355: open, $\tau$ is a dimensionless parameter and $\alpha>0$ represents
356: the pre-frontal voltage which we consider fixed in this paper. System
357: \eq(bik) can be obtained by simplifying right-hand sides of the
358: \emph{fast} subsystem in an appropriate asymptotic limit of a typical
359: cardiac excitation model \cite{Biktasheva-etal-2006}. In that sense,
360: system \eq(bik) plays the same role for a typical cardiac excitation model,
361: as ZFK equation \eq(zfk) plays for a classical activator-inhibitor
362: excitable system like \eq(fhn1).
363:
364: System \eq(bik) does not have nontrivial
365: bounded stationary solutions: if $E_t=h_t=0$ then any bounded solution
366: has the form $E=a$, $h=\Heav(-a)$, $a=\const$. So, \emph{there are no
367: critical nuclei} in this system.
368: Nevertheless, system \eq(bik) gives propagating front solutions for
369: initial conditions above a threshold and decay for those below it.
370: Hence there is a question, what happens when the
371: initial condition is exactly at the threshold.
372:
373: System \eq(bik) has a family of propagating front solutions
374: %
375: \begin{eqnarray}
376: E(z) &=& \left\{\begin{array}{ll}
377: \displaystyle \omega - \frac{\tau^2 c^2}{1+\tau c^2}\exp\left(\frac{z}{\tau c}\right) & (z\leq-\xm) , \\[1.5ex]
378: \displaystyle -\alpha + \alpha \exp(-cz) & (z\geq-\xm),
379: \end{array}\right. \nonumber \\ \nonumber\\
380: h(z) &=& \left\{\begin{array}{ll}
381: \displaystyle \exp\left( \frac{z}{\tau c} \right) & (z\leq 0) , \\
382: \displaystyle 1 & (z\geq 0) ,
383: \end{array}\right. \label{eq:Ehprofile}
384: \end{eqnarray}
385: %
386: where $z=x-ct$,
387: $\omega=1+\tau c^2(\alpha+1)$,
388: $\xm=\frac{1}{c}\ln\left(\frac{1+\alpha}{\alpha}\right)$ and
389: parameters $c$, $\alpha$ and $\tau$ related by
390: %
391: \begin{equation}
392: \tau c^2 \ln\left((1+\alpha)(c^2+\tau^{-1})\right) +
393: \ln\left(1+\alpha^{-1}\right)=0.
394: \label{eq:transalpha}
395: \end{equation}
396: %
397: For a fixed $\alpha$, there is a $\tau_*(\alpha)$ such that for
398: $\tau>\tau_*$, equation \eq(transalpha) has two solutions for $c$,
399: $c=c_{\pm}(\alpha,\tau)$, $c_+>c_-$
400: \cite{Biktashev-2002}. There is numerical and analytical evidence that solutions
401: with $c=c_+$ are stable and those with $c=c_-$ are unstable with one
402: positive eigenvalue \cite{Biktashev-2002,Hinch-2004}.
403:
404: Hence by analogy with the FHN system, we propose the following
405: %
406: \begin{conjecture}
407: The center-stable manifold of the unstable front solution
408: \eq(Ehprofile) with $c=c_-(\alpha,\tau)$
409: is the threshold hypersurface, separating the initial conditions leading to
410: initiation from the initial conditions leading to decay.\footnote{
411: Again, symmetry $x\leftrightarrow-x$
412: implies there are actually two hypersurfaces, partly connected with
413: each other.
414: }
415: \end{conjecture}
416:
417: That is, instead of a critical nucleus or a critical pulse solution,
418: the role of the threshold solution is played by a ``critical front'',
419: which we define as a traveling wave solution with different asymptotics
420: at $x-ct\to+\infty$ and $x-ct\to-\infty$: the pre-frontal state and the post-frontal state.
421:
422: An ``experimentally testable'' consequence of this conjecture is that
423: for any initial condition exactly at the threshold, the solution will
424: approach the unstable front as $t\to+\infty$. For any initial
425: condition near the threshold, the solution will come close to the
426: unstable front and stay in its vicinity for a long time: if the
427: positive eigenvalue is $\lambda$ and the initial condition is
428: $\delta$-close to the threshold, the transient front should be
429: observed for the time of the order of $\lambda^{-1}|\ln\delta|$. This
430: transient front solution \emph{will not depend on the initial
431: condition}, as long as the initial condition is at the threshold.
432:
433: %%%%%%%%%%%%%%%%
434: \begin{figure}[ht]
435: \includegraphics{fig3.eps}
436: \caption[]{
437: Evolution of two different near-threshold initial conditions toward the critical
438: front solution in system \eq(bik).
439: Initial stimuli:
440: $\xstim=0.3$, $\Estim=12.716330706144868$ (upper row)
441: and
442: $\xstim=1.5$, $\Estim=2.619968799545055$ (lower row).
443: Other parameters:
444: $\tau=8.2$, $\alpha=1$, $\dx=0.075$, $\dt=0.0025$, $L=50$.
445: }\label{fig:transient}
446: \end{figure}
447: %%%%%%%%%%%%%%%%
448:
449:
450: We have tested these predictions by numerical simulation of
451: \eqthree(bik,bik1b,bik1a). The results are shown in \figs{transient}
452: and~\ref{fig:critfront}.
453:
454: \Fig{transient} illustrates two solutions starting from initial conditions
455: with different $\xstim$ values. In both cases, $\Estim$ values have been
456: chosen close to the respective thresholds with high precision. In both
457: cases, the solutions evolve in the long run toward the same
458: propagating front.
459:
460: \Fig{critfront} presents an analysis of a pair of solutions, one with slightly
461: above-threshold and the other with slightly below-threshold initial
462: conditions. To separate the evolution of the front shape from its
463: movement, we employed the idea of symmetry group decomposition with
464: explicit representation of the orbit manifold (see
465: e.g. \cite{Biktashev-Holden-1998}). Practically, we
466: define the front point $\xf=\xf(t)$ via
467: \[
468: E(\xf(t),t)=E_*
469: \]
470: for some constant $E_*$ which is guaranteed to be represented exactly
471: once in the front at every instant of time (we have chosen $E_*=0$).
472: Then $E(x-\xf(t),t)$ gives the voltage profile ``in the standard position'', and
473: $\xf(t)$ describes the movement of this profile.
474:
475: %%%%%%%%%%%%%%%%
476: \begin{figure}[ht]
477: \includegraphics{fig4.eps}
478: \caption[]{
479: Transient critical fronts are close to the unstable front solution
480: of \eq(bik). Initial conditions: $\xstim=1.5$, with
481: $\Estim=2.619968799545055$ in the upper row and
482: $\Estim=2.619968799545054$ in the lower row,
483: other parameters the same as in \fig{transient}. Left column: evolution
484: of the $E$ profiles in the laboratory frame of reference. Middle
485: column: same evolution, in the frame of reference comoving with the
486: front. Right column: speed of the front. Blue/green dashed lines in the
487: middle and right columns correspond to the exact fast/slow front
488: solutions of \eq(bik).
489: }\label{fig:critfront}
490: \end{figure}
491: %%%%%%%%%%%%%%%%
492:
493: The predictions based on the Conjecture are that the voltage profile
494: should, after an initial transient depending on the initial condition,
495: approach the profile of the slow unstable front solution given by
496: \eq(Ehprofile) with $c=c_-(\tau,\alpha)$ and stay close to it for some time, before
497: either developing into the fast stable front
498: \eq(Ehprofile) with $c=c_+(\tau,\alpha)$
499: or decaying. Likewise, the speed of the front should, after an initial
500: transient, be close to the speed of the slow unstable front
501: $c_-(\tau,\alpha)$, before either switching the speed of the fast
502: stable front $c_+(\tau,\alpha)$ or dropping to zero. This is precisely
503: what is seen on \fig{critfront}, where we have taken advantage of
504: knowing the exact solutions $E(x-c_{\pm}t)$ and $c_\pm$ for both the fast and
505: the slow fronts.
506:
507: Initial conditions with different $\xstim$ and $\Estim$ close to the
508: corresponding threshold produce the same picture with the exception
509: of the initial transient. We have also checked that the length of the time
510: period during which the solution stays close to the unstable front is,
511: roughly, a linear function of the number of correct decimal figures in
512: $\Estim$, as it should be according to the Conjecture.
513:
514: %##############################
515: \section{Detailed cardiac excitation model}
516:
517: %%%%%%%%%%%%%%%%
518: \begin{figure}[ht]
519: \includegraphics{fig5.eps}
520: \caption[]{
521: Critical fronts in CRN model~\cite{Courtemanche-etal-1998}.
522: Shown are voltage profiles in every $10\,\ms$.
523: Parameter values:
524: $\dt =0.01\,\ms$,
525: $\dx =0.2$,
526: $L=40$, the length unit chosen so that voltage diffusion coefficient equals 1.
527: Stimulus witdh $\xstim=2$, stimulus amplitudes:
528: $\Vstim=29.31542299307152\,\mV$ (left panel) and
529: $\Vstim=29.31542299307153\,\mV$ (right panel).
530: The critical fronts are formed within first $10\,\ms$ and then
531: are seen for subsequent $80\,\ms$ on both panels before exploding into an excitation
532: wave of much bigger amplitude and speed on the right panel, and decaying on the left panel.
533: }\label{fig:CRN}
534: \end{figure}
535: %%%%%%%%%%%%%%%%
536:
537: The simplified model \eq(bik) is quantitatively very far from any
538: realistic ionic model of cardiac excitation, and has many peculiar
539: qualitative features, stemming from the non-standard asymptotic embedding
540: leading to it. Hence the newly described phenomenon of critical front
541: could be an artifact of the simplifications. To eliminate this
542: possibility, we have tested the relevance of the critical front
543: concept to a full ionic model of cardiac excitation. We have chosen
544: the model of human atrial tissue due to Courtemanche\ et\
545: al. (CRN)~\cite{Courtemanche-etal-1998}, which is less stiff than a
546: typical ventricular or Purkinje fiber model, is well formulated in a
547: mathematical sense and is popular among cardiac modellers. This
548: model operates with 21 dynamic variables including the transmembrane
549: voltage $V$. We have used the default parameter values as described in
550: \cite{Courtemanche-etal-1998} and supplemented the equation for $V$ in the ODE system with a
551: diffusion term $D\partial^2V/\partial{x}^2$. As the spatial scale is
552: not important for the question at hand, we assumed $D=1$. The
553: initial conditions for $V$ were taken in the form
554: \[
555: V(x,0) = V_r+\Vstim\Heav(\xstim-x)
556: \]
557: where $V_r=-81.18\,\mV$ is the standard resting potential, and for all
558: other 20 variables at their resting values as described in
559: \cite{Courtemanche-etal-1998}. \Fig{CRN} illustrates a pair of
560: solutions with initial conditions slightly above and slightly below
561: the threshold. The critical front solution is clearly seen there: it
562: has the upper voltage of about $-46\,\mV$ and during $80\,\ms$ of its
563: existence propagates with a speed approximately $0.06$ space units per
564: millisecond. Then for the above-critical case it develops into an
565: excitation front with maximal voltage about $+3\,\mV$ and speed 0.8
566: space units per millisecond, and decays for the below-critical case.
567:
568: Mathematically, the post-front voltage of about $-46\,\mV$ observed in
569: \fig{CRN} is not a true equilibrium of the full CRN model so the
570: critical front can only be an asymptotic concept in an appropriate
571: asymptotic embedding, say as ones described in
572: \cite{Biktasheva-etal-2006} or
573: \cite{Simitev-Biktashev-2006}, and the observed critical front may well be
574: the front of a critical pulse solution in the full model. However
575: \fig{CRN} demonstrates that the critical front is a practical and well
576: working concept even for the full model, unlike the critical pulse
577: which may be theoretically existing but practically unobservable:
578: notice the number of significant decimal digits in initial conditions
579: required to produce only the critical front observed for $80\,\ms$, and
580: recall that the number of decimals is roughly proportional to the
581: duration of the observation of an unstable solution.
582:
583:
584: %##############################
585: \section{Discussion}
586: We have presented numerical evidence that the
587: center-stable manifold of the unstable slow front solution of
588: the fast subsystem of a cardiac excitation model
589: serves as the threshold hypersurface separating initial conditions
590: leading to successful initiation and those leading to decay. This
591: means e.g. that a critical curve for a two-parametric family of
592: initial conditions, be that family \eq(bik1a) with parameters
593: $(\xstim,\Estim)$ or any other, can be found as an intersection of
594: this codimension-1 critical hyper-surface with the two-dimensional
595: manifold of those initial conditions. Finding this hypersurface can be
596: done numerically or analytically using suitable approximation, e.g. as
597: it was done in \cite{Neu-etal-1997,Moll-Rosencrans-1990} for the ZFK equation;
598: this is a subject for further investigation.
599:
600: Another problem for future study is to verify that the findings
601: remain qualitatively true for formal asymptotic embeddings
602: of various cardiac excitation models, and in particular,
603: which mathematical features of these embeddings are essential for the existence
604: of the critical fronts.
605: Of principal
606: importance is the conclusion that for cardiac equations, instead of the
607: ``critical nucleus'' or its slowly moving variant ``critical pulse'' known from
608: the FitzHugh-Nagumo theory, we now have a ``critical front solution''.
609: This, in particular, means physically that the make-or-break conditions of cardiac
610: excitation wave are restricted to a vicinity of its front.
611:
612: Finally, we believe that this study sets a useful example for
613: initiation problems in other types of excitable systems, alternative
614: to the existing critical nucleus theory, since not all, if any,
615: real-world excitable systems are well described by an asymptotic
616: structure as in \eq(fhn1).
617:
618: Authors are grateful to C.F.~Starmer and E.E.~Shnol for inspiring
619: discussion. The study has been supported by EPSRC grant GR/S75314/01
620: and MacArthur Foundation grant 71356-01.
621:
622: % \bibliography{cf}
623: \begin{thebibliography}{19}
624: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
625: \expandafter\ifx\csname bibnamefont\endcsname\relax
626: \def\bibnamefont#1{#1}\fi
627: \expandafter\ifx\csname bibfnamefont\endcsname\relax
628: \def\bibfnamefont#1{#1}\fi
629: \expandafter\ifx\csname citenamefont\endcsname\relax
630: \def\citenamefont#1{#1}\fi
631: \expandafter\ifx\csname url\endcsname\relax
632: \def\url#1{\texttt{#1}}\fi
633: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
634: \providecommand{\bibinfo}[2]{#2}
635: \providecommand{\eprint}[2][]{\url{#2}}
636:
637: \bibitem[{\citenamefont{Krinsky and Swinney}(1991)}]{Krinsky-Swinney-1991}
638: \bibinfo{editor}{\bibfnamefont{V.}~\bibnamefont{Krinsky}} \bibnamefont{and}
639: \bibinfo{editor}{\bibfnamefont{H.}~\bibnamefont{Swinney}}, eds.,
640: \emph{\bibinfo{title}{Wave and patterns in biological and chemical excitable
641: media}} (\bibinfo{publisher}{North-Holland}, \bibinfo{address}{Amsterdam},
642: \bibinfo{year}{1991}).
643:
644: \bibitem[{\citenamefont{Zipes and Jalife}(2000)}]{Zipes-Jalife-2000}
645: \bibinfo{editor}{\bibfnamefont{D.~P.} \bibnamefont{Zipes}} \bibnamefont{and}
646: \bibinfo{editor}{\bibfnamefont{J.}~\bibnamefont{Jalife}}, eds.,
647: \emph{\bibinfo{title}{Cardiac electrophysiology: From cell to bedside}}
648: (\bibinfo{publisher}{W B Saunders Co}, \bibinfo{year}{2000}).
649:
650: \bibitem[{\citenamefont{FitzHugh}(1961)}]{FitzHugh-1961}
651: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{FitzHugh}},
652: \bibinfo{journal}{Biophysical Journal} \textbf{\bibinfo{volume}{1}},
653: \bibinfo{pages}{445} (\bibinfo{year}{1961}).
654:
655: \bibitem[{\citenamefont{Nagumo et~al.}(1962)\citenamefont{Nagumo, Arimoto, and
656: Yoshizawa}}]{Nagumo-etal-1962}
657: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Nagumo}},
658: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Arimoto}}, \bibnamefont{and}
659: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Yoshizawa}},
660: \bibinfo{journal}{Proc. IRE} \textbf{\bibinfo{volume}{50}},
661: \bibinfo{pages}{2061} (\bibinfo{year}{1962}).
662:
663: \bibitem[{\citenamefont{Rushton}(1937)}]{Rushton-1937}
664: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Rushton}},
665: \bibinfo{journal}{Proc. Roy. Soc. Lond. ser. B}
666: \textbf{\bibinfo{volume}{124}}, \bibinfo{pages}{210} (\bibinfo{year}{1937}).
667:
668: \bibitem[{\citenamefont{Starmer et~al.}(2003)\citenamefont{Starmer, Grant, and
669: Colatsky}}]{Starmer-etal-2003}
670: \bibinfo{author}{\bibfnamefont{C.~F.} \bibnamefont{Starmer}},
671: \bibinfo{author}{\bibfnamefont{A.~O.} \bibnamefont{Grant}}, \bibnamefont{and}
672: \bibinfo{author}{\bibfnamefont{T.~J.} \bibnamefont{Colatsky}},
673: \bibinfo{journal}{Cardiovasc. Res.} \textbf{\bibinfo{volume}{57}},
674: \bibinfo{pages}{1062} (\bibinfo{year}{2003}).
675:
676: \bibitem[{\citenamefont{Biktashev and Suckley}(2004)}]{Biktashev-Suckley-2004}
677: \bibinfo{author}{\bibfnamefont{V.~N.} \bibnamefont{Biktashev}}
678: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Suckley}},
679: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{93}},
680: \bibinfo{pages}{168103} (\bibinfo{year}{2004}).
681:
682: \bibitem[{\citenamefont{Flores}(1989)}]{Flores-1989}
683: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Flores}}, \bibinfo{journal}{J.
684: Diff. Eq.} \textbf{\bibinfo{volume}{80}}, \bibinfo{pages}{306}
685: (\bibinfo{year}{1989}).
686:
687: \bibitem[{\citenamefont{Flores}(1991)}]{Flores-1991}
688: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Flores}},
689: \bibinfo{journal}{SIAM J. Math. Anal.} \textbf{\bibinfo{volume}{22}},
690: \bibinfo{pages}{392} (\bibinfo{year}{1991}).
691:
692: \bibitem[{\citenamefont{Zel'dovich and
693: Frank-Kamenetsky}(1938)}]{Zeldovich-FrankKamenetsky-1938}
694: \bibinfo{author}{\bibfnamefont{Y.~B.} \bibnamefont{Zel'dovich}}
695: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{D.~A.}
696: \bibnamefont{Frank-Kamenetsky}}, \bibinfo{journal}{Doklady AN SSSR}
697: \textbf{\bibinfo{volume}{19}}, \bibinfo{pages}{693} (\bibinfo{year}{1938}).
698:
699: \bibitem[{\citenamefont{McKean}(1970)}]{McKean-1970}
700: \bibinfo{author}{\bibfnamefont{H.~P.} \bibnamefont{McKean},
701: \bibfnamefont{Jr.}}, \bibinfo{journal}{Adv. Appl. Math.}
702: \textbf{\bibinfo{volume}{4}}, \bibinfo{pages}{209} (\bibinfo{year}{1970}).
703:
704: \bibitem[{\citenamefont{Neu et~al.}(1997)\citenamefont{Neu, Preissig, and
705: Krassowska}}]{Neu-etal-1997}
706: \bibinfo{author}{\bibfnamefont{J.~C.} \bibnamefont{Neu}},
707: \bibinfo{author}{\bibfnamefont{R.~S.} \bibnamefont{Preissig}},
708: \bibnamefont{and}
709: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Krassowska}},
710: \bibinfo{journal}{Physica D} \textbf{\bibinfo{volume}{102}},
711: \bibinfo{pages}{285} (\bibinfo{year}{1997}).
712:
713: \bibitem[{\citenamefont{Biktashev}(2002)}]{Biktashev-2002}
714: \bibinfo{author}{\bibfnamefont{V.~N.} \bibnamefont{Biktashev}},
715: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{89}},
716: \bibinfo{pages}{168102} (\bibinfo{year}{2002}).
717:
718: \bibitem[{\citenamefont{Biktasheva et~al.}(2006)\citenamefont{Biktasheva,
719: Simitev, Suckley, and Biktashev}}]{Biktasheva-etal-2006}
720: \bibinfo{author}{\bibfnamefont{I.~V.} \bibnamefont{Biktasheva}},
721: \bibinfo{author}{\bibfnamefont{R.~D.} \bibnamefont{Simitev}},
722: \bibinfo{author}{\bibfnamefont{R.~S.} \bibnamefont{Suckley}},
723: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{V.~N.}
724: \bibnamefont{Biktashev}}, \bibinfo{journal}{Phil. Trans. Roy. Soc. Lond. ser.
725: A} \textbf{\bibinfo{volume}{364}}, \bibinfo{pages}{1283}
726: (\bibinfo{year}{2006}).
727:
728: \bibitem[{\citenamefont{Hinch}(2004)}]{Hinch-2004}
729: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Hinch}},
730: \bibinfo{journal}{Bull. Math. Biol.} \textbf{\bibinfo{volume}{66}},
731: \bibinfo{pages}{1887} (\bibinfo{year}{2004}).
732:
733: \bibitem[{\citenamefont{Biktashev and Holden}(1998)}]{Biktashev-Holden-1998}
734: \bibinfo{author}{\bibfnamefont{V.~N.} \bibnamefont{Biktashev}}
735: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{A.~V.}
736: \bibnamefont{Holden}}, \bibinfo{journal}{Physica D}
737: \textbf{\bibinfo{volume}{116}}, \bibinfo{pages}{342} (\bibinfo{year}{1998}).
738:
739: \bibitem[{\citenamefont{Courtemanche et~al.}(1998)\citenamefont{Courtemanche,
740: Ramirez, and Nattel}}]{Courtemanche-etal-1998}
741: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Courtemanche}},
742: \bibinfo{author}{\bibfnamefont{R.~J.} \bibnamefont{Ramirez}},
743: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Nattel}},
744: \bibinfo{journal}{Am. J. Physiol.} \textbf{\bibinfo{volume}{275}},
745: \bibinfo{pages}{H301} (\bibinfo{year}{1998}).
746:
747: \bibitem[{\citenamefont{Simitev and Biktashev}(2006)}]{Simitev-Biktashev-2006}
748: \bibinfo{author}{\bibfnamefont{R.~D.} \bibnamefont{Simitev}} \bibnamefont{and}
749: \bibinfo{author}{\bibfnamefont{V.~N.} \bibnamefont{Biktashev}},
750: \bibinfo{journal}{Biophysical Journal} \textbf{\bibinfo{volume}{90}},
751: \bibinfo{pages}{2258} (\bibinfo{year}{2006}).
752:
753: \bibitem[{\citenamefont{Moll and Rosencrans}(1990)}]{Moll-Rosencrans-1990}
754: \bibinfo{author}{\bibfnamefont{V.}~\bibnamefont{Moll}} \bibnamefont{and}
755: \bibinfo{author}{\bibfnamefont{S.~I.} \bibnamefont{Rosencrans}},
756: \bibinfo{journal}{SIAM J. Appl. Math.} \textbf{\bibinfo{volume}{50}},
757: \bibinfo{pages}{1419} (\bibinfo{year}{1990}).
758:
759: \end{thebibliography}
760:
761: \end{document}