1: \documentstyle[12pt]{article}
2: \input epsf.tex
3:
4: \begin{document}
5:
6: \title
7: {
8: About periodic solutions of a planar system modelling neural activity%
9: \thanks{The work is supported by the Russian Foundation for Fundamental
10: Research (Grant No. 99-01-00574)}
11: }
12:
13: \author{S.~Treskov, E.~Volokitin}
14: \date{}
15: \maketitle
16:
17: \begin{abstract}
18: We derived explicit symbolic expressions for the first, second, and third
19: Lyapunov coefficients of the complex focus
20: of a planar system modelling activity of a neural network.
21: The analysis of these expressions allowed us to obtain new results about
22: the number and location of limit cycles in the model.
23: \end{abstract}
24:
25: \noindent {\bf Introduction}.
26: In this paper we study the mathematical model which describe processes
27: in a neural network consisting from two neurons.
28: The model was suggested in [1] and is defined as a planar
29: differential system
30: \refstepcounter{equation}
31: %\label{sys1}
32: $$ \begin{array} {ll}
33: \dot u_1=-u_1+q_{11} \varphi(u_1) - q_{12} u_2 + e_1,\\
34: \dot u_2=-u_2+ q_{21} \varphi(u_1) +e_2,
35: \end{array}
36: \eqno{(1)}
37: $$
38: where $q_{11}, q_{12}, q_{21} >0, e_1, e_2 \in {\bf R}$ are parameters, and
39: $$
40: \varphi(u_1)=\frac{1}{1+e^{-4u_1}}.
41: $$
42:
43: In [1] it was fulfilled a study of model (1) with
44: the help of the bifurcation theory and it was listed possible types
45: of dynamical behavior of the neural network;
46: with analytic and numerical techniques it was constructed the bifurcation
47: diagram of system (1) for some parameter values.
48:
49: In [2] it was proved by analytic methods the correctness of the
50: bifurcation diagram for some specific cases.
51:
52: In [2] it was demonstrated that the change of variables
53: $$
54: u=u_1, v=\frac{u_2-e_2}{q_{21}},
55: $$
56: forms system (1) to the system
57: \refstepcounter{equation}
58: %\label{sys2}
59: $$ \begin{array} {ll}
60: \dot u=-u+a \varphi(u) - b v + c,\\
61: \dot v=-v + \varphi(u),
62: \end{array}
63: \eqno{(2)}
64: $$
65: which involves three parameters only
66: $$
67: a=q_{11}, b=q_{12}q_{21}, c=e_1-q_{12}e_2, a,b>0, c \in {\bf R}.
68: $$
69:
70: We intend to study system (2).
71:
72: Using our results from [3], we derived analytic formulas for the
73: first, second, and third Lyapunov coefficients of the complex focus
74: of system (2). A sophisticated treatment of the formulas
75: allows us to assert that for some parameter values system (2)
76: has three concentric limit cycles around a unique steady state,
77: two of these cycles being stable. Such a phase portrait is absent
78: in the list of possible phase portraits of (2) from [1,2]
79: although on our mind it is interesting from a physical point of view.
80:
81: \quad
82:
83: \noindent {\bf 1}.
84: Let us to note the following property of system (2).
85: If $u(t), v(t)$ is the solution of the system then
86: $u_1(t)=-u(t), v_1(t)=1-v(t)$ is the solution of the system
87: $$ \begin{array} {ll}
88: \dot u=-u+a \varphi(u) - b v - c + b -a,\\
89: \dot v=-v + \varphi(u),
90: \end{array}
91: $$
92: This fact means that if we have a phase portrait of (2)
93: for parameter values $a=a_1, b=b_1, c=c_1$ then there exists
94: a symmetric with respect to the point $u=0, v=1/2$ phase portrait of the
95: system for parameter values $a=a_1, b=b_1, c=c_2=-a_1+b_1-c_1$.
96: In the parameter space points $(a_1,b_1, c_1)$ and $(a_1, b_1, c_2)$
97: lie on different sides from the hyperplane $a-b+2á=0$ (on equal distances).
98: So, it is sufficiently to study system (2) for parameter values
99: from one of a half-space, for example, from the half-space $a-b+2c \geq 0.$
100:
101: The mentioned property is obvious from the
102: bifurcation diagrams constructed with numerical methods in [2].
103:
104: The authors of [1,2] solved all problems of the number, location,
105: and types of steady states of system (2). The system
106: may have between one and three steady states in the finite part of a
107: phase plane. If the system has three steady states then one of them is always
108: a saddle. Two remaining steady states are nodes or foci (stable or
109: unstable). We distinguish between them a left and right steady state.
110:
111: The authors of [4] introduced the following method to describe
112: the phase portraits of system (2).
113: The symbol $s$ ($u$) denotes a stable (unstable) equilibrium;
114: the symbol $S$ ($U$), a stable (unstable) limit cycle.
115: The subscript $1$ refers to the symbol corresponding to the left
116: equilibrium or to a limit cycle around it;
117: the subscript $2$, to the symbol corresponding to the right
118: equilibrium or to a limit cycle around it.
119: A symbol has no subscripts when the steady state is unique.
120: The symbols $S$ and $U$ without subscripts refer also to limit cycles
121: around all three steady states.
122: If there are several cycles with the same subscript or without subscript,
123: then they are listed in order from inside.
124:
125: The authors of [1,2] detected following phase portraits of
126: system (2).%
127: \footnote
128: {The subscript 1 for the number of a phase portrait implies that
129: a corresponding portrait is symmetric to the portrait with the same
130: number without the subscript.
131: }
132:
133: \refstepcounter{equation}
134: %\label{sp}
135: \begin{center}
136: \begin{tabbing}
137: 1$_{\rm 1}$. \= $s_1U_1u_2S$ \quad \ \ \ \ \= 10$_{\rm 1}$.\= $s_1U_1u_2S_2US$ \quad \= 20$_{\rm 1}$. \= $s_1U_1u_2S_2US$ \= \hfill \kill
138: 1.\> $s_1s_2$ \> 5$_{\rm 1}$.\> $s_1u_2$ \> 9$_{\rm 1}$.\> $s_1U_1u_2S$ \> \\
139: 2.\> $s_1s_2US$ \> 6.\> $u_1s_2US$ \> 10.\> $u_1s_2S$ \> \\
140: 3.\> $s_1U_1s_2US$ \> 6$_{\rm 1}$.\> $s_1u_2US$ \> 10$_{\rm 1}$.\> $s_1u_2S$ \> \\
141: 3$_{\rm 1}$.\> $s_1s_2U_2US$ \> 7.\> $s_1U_1s_2S$ \> 11.\> $u_1u_2S$ \> \` (3) \\
142: 4.\> $s_1U_1s_2$ \> 7$_{\rm 1}$.\> $s_1s_2U_2S$ \> 12.\> $s$ \> \\
143: 4$_{\rm 1}$.\> $s_1s_2U_2$ \> 8.\> $s_1U_1s_2U_2S$ \> 13.\> $sUS$ \> \\
144: 5.\> $u_1s_2$ \> 9.\> $u_1s_2U_2S$ \> 14.\> $uS$ \> \\
145: \end{tabbing}
146: \end{center}
147:
148: As a result of our study, we detect one more phase portrait of
149: system (2) which is described by the sequence
150: \refstepcounter{equation}
151: %\label{sp1}
152: $$
153: {\rm 15}.\ uSUS
154: \eqno{(4)}
155: $$
156: and corresponds to the situation when in the system two
157: stable limit cycles coexist which are separated with an unstable cycle.
158:
159: The detection is due to a more comprehensive analysis of the Lyapunov
160: coefficients of the complex focus of system (2).
161: The next section is devoted to this analysis.
162:
163: \quad
164:
165: \noindent {\bf 2}.
166: In [3] we derived symbolic expressions for the first, second, and
167: third Lyapunov coefficients of the complex focus of the system
168: $$
169: \begin{array}{l}
170: \dot x=y,\\
171: \dot y=a_{10} x + a_{01} y + a_{20} x^2 + a_{11} x y + a_{02} y^2 + \ldots
172: \end{array}
173: $$
174: in terms of the coefficients $a_{ij}$.
175:
176: The expressions are rather unwieldy and in general case are not amenable
177: to theoretical study but in some special cases they may be simplified
178: noticeably. In particular, if we have a Li\'{e}nard system with the complex
179: focus in the origin
180: \refstepcounter{equation}
181: %\label{sysl}
182: $$
183: \begin{array} {ll}
184: \dot x=y,\\
185: \dot y=p(x)+y q(x) \equiv p_1 x + p_2 x^3 + \ldots + y(q_1 x + q_2 x^2 + \ldots),
186: \end{array}
187: \eqno{(5)}
188: $$
189: then the expressions for the first three Lyapunov coefficients
190: with accuracy to a positive factor coincide with ones
191: \refstepcounter{equation}
192: %\label{lvl}
193: $$
194: \begin{array}{ll}
195: l_1 = & p_2 q_1 - p_1 q_2,\\
196: l_2 = & 5(p_2 q_3 - p_3 q_2) + 3(p_4 q_1 - p_1 q_4),\\
197: l_3 = & 14 p_2 (p_2 q_4 - p_4 q_2) + 21 p_1 (p_3 q_4 - p_4 q_3) +
198: 35 p_1 (p_5 q_2 - p_2 q_5) + \\
199: \ & + 15 p_1 (p_1 q_6 - p_6 q_1).
200: \end{array}
201: \eqno{(6)}
202: $$
203:
204: Let $(u_0,v_0)$ be a steady state of system (2).
205: Then we have
206: $$
207: u_0-(a-b)\varphi (u_0) -c =0, v_0=\varphi (u_0).
208: $$
209:
210: Let us make the following change of variables in system (2)
211: $$
212: x=u-u_0, \ y=-u+a \varphi (u) - b v + c.
213: $$
214:
215: After it the steady state $(u_0, v_0)$ transfers to the origin.
216:
217: We obtain a Li\'{e}nard system which is equivalent to system (2)
218: \refstepcounter{equation}
219: %\label{sys3l}
220: $$
221: \begin{array}{ll}
222: \dot x=y,\\
223: \dot y=p(x) + y q(x),
224: \end{array}
225: \eqno{(7)}
226: $$
227: where
228: \refstepcounter{equation}
229: %\label{pq}
230: $$
231: \begin{array}{ll}
232: p(x)= c + (a-b)/(1 + \exp(-4u_0 - 4x)) - u_0 - x,\\
233: q(x)= -2 + (4 a \exp(-4 u_0 - 4 x)/(1 + \exp(-4 u_0 - 4 x))^2.
234: \end{array}
235: \eqno{(8)}
236: $$
237:
238: The origin will be the complex focus of system (7), (8)
239: if $p(0)=q(0)=0, p'(0)<0,$ that is under the fulfillment of the conditions
240: $$
241: \begin{array}{ll}
242: c + (a-b)/(1 + \exp(-4u_0)) - u_0=0,\\
243: -2 + (4 a \exp(-4 u_0)/(1 + \exp(-4 u_0))^2=0,\\
244: -1 + 4(a-b) \exp(4 u_0))/(1 + \exp(4 u_0))^2 <0,
245: \end{array}
246: $$
247: which may be written as
248: \refstepcounter{equation}
249: %\label{cond}
250: $$
251: \begin{array}{ll}
252: c =(b -a)/(1 + \exp(-4u_0)) + u_0,\\
253: a = (1 + \exp(4 u_0))^2/(2 \exp(4 u_0)),\\
254: -1 + 4(a-b) \exp(4 u_0))/(1 + \exp(4 u_0))^2 <0.
255: \end{array}
256: \eqno{(9)}
257: $$
258:
259: Under conditions (9) system (7), (8)
260: takes form (5).
261:
262: Evaluating needed coefficients $p_i, q_i$ and substituting them to (6),
263: we obtain the expressions for the Lyapunov coefficients of system
264: (7), (8).
265:
266: After some transformations we obtain that the Lyapunov coefficients
267: with accuracy to a positive factor are equal to the expressions
268:
269: \refstepcounter{equation}
270: %\label{lv12}
271: $$
272: \begin{array}{ll}
273: l_1 =& 1 - 2 \exp(4 u_0) - (6-8a+8b) \exp(8 u_0) - 2 \exp(12 u_0) +\\
274: \ & +\exp(16 u_0),\\
275: l_2 =& 3 - 72 \exp(4 u_0) + (45 + 8(a-b))\exp(8 u_0) +\\
276: \ & (240 - 368(a -b)) \exp(12 u_0) + (45 + 8(a-b))\exp(16 u_0) +\\
277: \ & - 72 \exp(20 u_0) + 3 \exp(24 u_0),\\
278: l_3=&1 -4(29 +a -b)\exp(4u_0) + (717+160(a-b))\exp(8u_0)+\\
279: &+16(102 -205(a-b) -6(a-b)^2 )\exp(12u_0) - \\
280: &-2(903- 1904(a-b) -384(a-b)^2)\exp(16u_0)+\\
281: &+8(651 + 1813(a-b) -1608( a-b)^2 )\exp(20u_0)-\\
282: &-2(903 -1904(a-b) -364(a-b)^2)\exp(24u_0)+\\
283: &+16(102 -205(a-b) -6(a-b)^2 )\exp(28u_0) + \\
284: &+(717+160(a-b))\exp(32u_0) -4(29 +a -b)\exp(36u_0) +\\
285: &+\exp(40u_0).
286:
287: \end{array}
288: \eqno{(10)}
289: $$
290:
291: Conditions (9) define a two-dimensoinal
292: bifurcation manifold in the parameter space. This manifold corresponds
293: to the codim 1 Andronov-Hopf bifurcation. The condition $l_1=0$
294: from (10) defines in this manifolds a one-dimensional bifurcation
295: codim 2 manifold corresponding to the degenerate Andronov-Hopf bifurcation
296: (the Bautin bifurcation, the Takens bifurcation, the vanishing of the first
297: Lyapunov coefficient).
298:
299: The last manifold is piecewise smooth and it may be defined as follows
300:
301: \refstepcounter{equation}
302: %\label{l1par}
303: $$
304: \begin{array}{ll}
305: a= (1 + \vartheta)^2/(2 \vartheta),\\
306: b=(1 + \vartheta)^2 (1 + \vartheta^2)/(8 \vartheta^2),\\
307: c=(1 - 3 \vartheta - 3 \vartheta^2 + \vartheta^3 + 2 \vartheta \ln \vartheta)/(8 \vartheta),\\
308: 0<\vartheta<+\infty, \vartheta \neq 1.
309: \end{array}
310: \eqno{(11)}
311: $$
312:
313: Coordinates $(u_0,v_0)$ of a corresponding unique equilibrium may be
314: obtained from equations $\exp (4 u_0)=\vartheta, v_0=\varphi(u_0).$
315:
316: If parameters $a, b, c$ are evaluated accordingly to (11)
317: then system (7), (8) (and then system (2) also)
318: has the complex focus of the multiplicity 2.
319: Stability of the focus is defined by the sign of the expression $\bar l_2$
320: which is the second Lyapunov coefficient $l_2$ from (10) evaluating
321: accordingly to the (11) and to the equality $\vartheta=\exp(4 u_0).$
322: $$
323: \bar l_2=2(1+\vartheta)^2 (1-14 \vartheta + 6 \vartheta^2 -14 \vartheta^3 + \vartheta^4).
324: $$
325:
326: The equation $\bar l_2=0$ has two positive roots $\vartheta_1, \vartheta_2$
327: $$
328: \begin{array}{ll}
329: \vartheta_1=(7+3\sqrt{5}+\sqrt{6}\sqrt{15+\sqrt{5}})/2 \approx 13.6349,\\
330: \vartheta_2=(7+3\sqrt{5}-\sqrt{6}\sqrt{15+\sqrt{5}})/2 \approx .0733414,\\
331: \end{array}
332: $$
333: to which correspond the following parameter values
334: $a_1 \approx 7.8541, b_1\approx 26.9164,$ $c_1\approx 18.4129$ and
335: $a_2=a_1, b_2=b_1, c_2 \approx 0.64937,$
336: $a_1-b_1+2á_1>0, a_2-b_2+2á_2<0.$
337:
338: The third Lyapunov coefficient is negativ for $(a_1, b_1, c_1)$
339: and $(a_2, b_2, c_2)$.
340:
341: For such parameters system (2) has a unique steady state
342: which is the complex focus of multiplicity 3.
343:
344: It is well known that in the parameter space to the point corresponding
345: to the vanishing of the second Lyapunov coefficient is adjacent a region
346: of parameter values for which the system has three concentric limit
347: cycles around the equilibrium.
348:
349: Thus, we proved that there are parameter values $a,b,c,$ for which
350: system (2) has three limit cycles surrouding a unique steady
351: state (an unstable focus). Two from these cycles are stable and they
352: are separated by an unstable cycle.
353:
354: \quad
355:
356: \noindent {\bf 3}.
357: Fig.1 gives a part of a complete bifurcation diagram of system (2)
358: for a fixed value of parameter $a>a_1.$ The rest of the diagram may be
359: easily restored from the about mentioned reasons of the symmetry.
360: The diagram is fairly typical. In particular, it has many common features
361: with the diagram given in [1,2] for $a<a_1.$ We use the notations
362: introduced in [1,2] and restrict ourselves to only brief remarks.
363: More detailed description of the mentioned bifurcations may be found
364: in [6,7], for example.
365:
366: The curve $sn_1$ corresponds to the availability in the system of a
367: a double steady state which by an appropriate parameter perturbation
368: splits into a saddle and a node which lies lefter then a saddle.
369: On the curve $sn_1$ there is a point $tb_1,$ which corresponds
370: to the Bogdanov-Takens bifurcation and divides the curve into two
371: segments with stable and unstable saddle-nodes.
372:
373: The line $h_1\ (h_2)$ corresponds to the Andronov-Hopf bifurcation
374: of the left (right) steady state if parameters are inside the region
375: of multiplicity of steady states or of a unique steady state.
376: As we noted above these lines are defined by formulas (9).
377: The point $dh_1$ on the line $h_1$ corresponds to the degenerate
378: Andronov-Hopf bifurcation and divides the line into two segments
379: corresponding to a soft or hard loss in stability. The second
380: Lyapunov coefficient is positive at the point $dh_1$. Let us note
381: that this coefficient is negative at the point $dh_1$ on the diagram
382: presented in [1,2] for $a<a_1$.
383:
384: The curves $sl_1$ and $sl_2$ correspond to separatrix loops around
385: left and right steady states. The curve $sl_4$ corresponds to a separatrix
386: loop around both steady states.
387:
388: The curve $snpo$ corresponds to a double cycle which is stable from
389: outside.
390: The curve $snpo_1$ corresponds to a double cycle which is stable from
391: inside. The curves $snpo$ and $snpo_1$ coalesce and form a cusp
392: at the point corresponding to a triple cycle of system (2).
393:
394: The listed curves divides the parameter space into regions within which
395: the system has distinct rough phase portraits describing
396: in (3), (4).
397:
398: As usual, we give only the scaleless relative positions of the
399: bifurcation sets. The drawing of these sets to real scale involves technical
400: dificulties.
401:
402: The local bifurcation sets $sn_i, h_i, tb_i, dh_i$
403: can be described by explicit formulas
404: (see (9), (11), for example).
405: Therefore, construction of these sets involves no difficulties.
406: As to sets of nonlocal bifurcations $sl_i, snpo, snpo_1$,
407: in constructing them we used the results of the bifurcation theory,
408: and the results of the qualitative theory of ordinary differential
409: equations, and the results of simulations.
410:
411: A considerable difference of our bifurcation diagram from the diagram
412: in [1,2] (for $a<a_1$) is the presence of region 15 and curve
413: $snpo_1$ whose existence we proved above.
414:
415: In Fig.2 is shown the phase portrait of system (2)
416: for parameter values $a=16., b=130., c=111.165$, corresponding
417: to region 15.%
418: \footnote
419: {To confirm the existence of three limit cycles in system (2)
420: for the specified parameter values we studied the proper Poincar\'{e}
421: map numerically and established that it has three fixed points.
422: }
423:
424:
425: \begin{center}
426: \quad \epsfbox{pic1.eps} \quad \\
427: Fig.1\\
428: \quad \\
429: \quad \epsfbox{pic2.eps} \quad \\
430: Fig.2
431: \end{center}
432:
433: Hence our analysis of formulas for the Lyapunov coefficients for the
434: system describing the dynamics of a neural network consisting from
435: two sells enables us to supplement results of other authors.
436: In particular, to earlier descrided phase portraits we add one more
437: portrait with three limit cycles around an unstable steady state.
438: Studing a complete bifurcation diagram, we specify the position
439: of this portrait in the parameter space.
440:
441: The results of [1,2] and our results give grounds to advocate
442: that all possible phase portraits of system (2) are included
443: in the join of lists (3), (4).
444:
445: \noindent {\bf 4}.
446: As we noted above, the bifurcation diagram from [1,2]
447: is fairly typical and occurs during a study of many mathematical
448: models, see [8,9] for references.
449: In particular, a simular diagram was consnructed by us in [10]
450: where we investigated a planar system modelling a nerve conduction
451: in the squid giant axon.
452:
453: {\bf Acknowledgmens.} We thank Professor Fotios Giannakopoulos for materials
454: which he placed at our disposal.
455:
456: \newpage
457:
458: \begin{thebibliography}{10}
459:
460: \bibitem{1}
461: F.~Giannakopoulos, O.~Oster.
462: {\it Bifurcation propeties of a planar system modelling neural activity.}
463: Planar Nonlinear Dynamical Systems (Delft, 1995), Differential and
464: Dynamical Systems. 1997. V. 5. 229--242.
465:
466: \bibitem{2}
467: R.E.Kooij, F.~Giannakopoulos.
468: {\it Periodic orbits in planar systems modelling neural activity.}
469: Quarterly of Applied Mathematics. 2000. V.63. No 3. 437--457.
470:
471: \bibitem{3}
472: Volokitin E.P, Treskov S.A.
473: {\it About the Lyapunov values of the complex focus of a planar
474: dynamical system.}
475: Transactions of Russian Academy of Natural Sciences.
476: 1997. V. 1. No 1. 59--72. (in Russian)
477:
478: \bibitem {4}
479: Vaganov, D.A., Samoilenko, N.G. and Abramov, V.G.
480: {\it A determination of regions of
481: periodic regimes of continuous stirred tank reactors.}
482: Transactions of Academy of Sciences of the USSR.
483: 1977. V. 234, 640--643. (in Russian)
484:
485: \bibitem{6}
486: Bazykin, A., Kuznetsov, Y. and Khibnik, A.
487: {\it Portraits of Bifurcations: Bifurcation Diagrams of Planar
488: Dynamical Systems}. 1989. Znanie, Moscow. (in Russian)
489:
490: \bibitem{7}
491: Y.A.Kuznetsov.
492: {\it Elements of Applied Bifurcation Theory.}
493: 1995. Springer-Verlag, New York.
494:
495: \bibitem{8}
496: A.Khibnik, B.Krauskopf, C.Rousseau.
497: {\it Global study of a family of cubic Lienard equations.}
498: Preprint. CRM-2471, April 1997.
499:
500: \bibitem{9}
501: A.Khibnik, B.Krauskopf, C. Rousseau.
502: {\it Global study of a family of cubic Lienard equations.}
503: Nonlinearity. 1998. V.11. No 6, 1505--1519.
504:
505: \bibitem{10}
506: Volokitin E.P, Treskov S.A.
507: {\it A parametric portrait of the Fitz-Hugh system.}
508: Mathematical modelling. 1994. No 6. 65--78. (in Russian)
509:
510: \end{thebibliography}
511:
512: \end{document}
513: