q-bio0402040/weak.tex
1: \documentclass[aps,pre,preprint,groupedaddress,showpacs]{revtex4}
2: %\documentclass[twocolumn,secnumarabic,amssymb, nobibnotes, aps, prd,
3: %showpacs]{revtex4}
4: \usepackage{epsfig}
5: \begin{document}
6: 
7: \preprint{}
8: 
9: \title{Turbulence near cyclic fold bifurcations in birhythmic media}
10: 
11: \author{Dorjsuren Battogtokh} 
12: \email{dbattogt@vt.edu}
13: \author{John J. Tyson}
14: \affiliation{
15: Department of Biology, Virginia Polytechnic and State University
16: Blacksburg, VA 24061-0106
17: } 
18: 
19: \date{\today}
20: 
21: \begin{abstract}
22: We show that at the onset of a cyclic fold bifurcation,
23: a birhythmic  medium composed of glycolytic oscillators 
24: displays turbulent dynamics. By computing the largest Lyapunov 
25: exponent, the spatial correlation function,  and the average 
26: transient lifetime, we classify it as a weak turbulence 
27: with transient nature. Virtual heterogeneities
28: generating  unstable fast oscillations 
29: are the  mechanism of the transient turbulence.  
30: In the presence of  wavenumber instability, unstable oscillations 
31: can be reinjected leading to 
32: stationary turbulence. We also find similar turbulence
33: in a cell cycle model. These findings  
34: suggest that  weak turbulence may be universal 
35: in  biochemical birhythmic media exhibiting cyclic 
36: fold bifurcations.
37: \end{abstract}
38: \pacs{ 05.45.Jn, 05.45.Pq, 82.39.Fk, 82.40.Bj, 89.75.Fb}
39: 
40: \maketitle
41: \section{Introduction}
42: In studies of chemical turbulence
43:  in reaction diffusion systems near 
44: a Hopf bifurcation, a reduction of the model to the complex Ginzburg Landau 
45: equation(CGLE) is very useful \cite{kura1,cross}. First, it allows to 
46: define a parameter set in the model leading to turbulence  without 
47: carrying out simulations 
48: \cite{kbn}. Secondly, the
49: detailed knowledge of the CGLE's dynamics can be very  helpful \cite{
50: nozaki,saka,shraiman,kramer1}, because  mathematical models from different 
51: disciplines displaying dynamics near a Hopf bifurcation
52: obey the same qualitative dynamics of the CGLE \cite{kura2}. 
53: 
54: However, the CGLE alone is insufficient 
55: for a qualitative description of realistic models in a neighborhood of 
56: a Hopf bifurcation, when other bifurcations occur nearby
57: \cite{kramer2,show1}. For example, near a 
58: supercritical Hopf bifurcation point, another stable limit cycle may exist, 
59: so that, depending on initial conditions, oscillations with two different 
60: frequencies and amplitudes are possible. Such a situation called birhythmicity
61: is a characteristic feature of a number of well known models in biochemical 
62: oscillations \cite{gold1,borisuk}. For these systems, the CGLE cannot be used 
63: without appropriate modifications. Often, the best way to approach 
64: these problems is simulations of the original models \cite{show1,yama}.
65: 
66: To the best of our knowledge, little is known about turbulence in birhythmic 
67: media. Intuitively, in a regime of a strong wavenumber instability, 
68: birhythmicity should not be a factor. Therefore, turbulence in 
69: homogeneous birhythmic media and in coupled limit cycle oscillators
70: should have similar characteristics. In the absence of wavenumber instability,
71: high frequency oscillations are supposed to suppress slow 
72: oscillations and restore uniform oscillations. 
73: But at the onset of a cyclic fold (CF) bifurcation in birhythmic media of 
74: a biochemical origin, high frequency oscillations may be  unstable. 
75: Thus, a complete suppression of slow oscillations may not be achieved in
76: these systems. On the contrary, if unstable oscillations emerge persistently, 
77: complex spatiotemporal motions are possible.
78: 
79: The goal of this work is to  show that near cyclic fold bifurcations
80: in birhythmic media, virtual heterogeneities creating unstable oscillations 
81: can lead to a peculiar turbulence, 
82: intermittency of small and large
83: amplitude oscillations.
84: We will first compute complex
85: spatiotemporal behavior in a birhythmic 
86: medium composed of glycolytic oscillators. By calculating 
87: the maximal Lyapunov exponent, the spatial correlation function, and the 
88: average transient lifetime, we will provide evidence 
89: that this behavior is weak transient turbulence. In the presence of 
90: wavenumber instability 
91: transient turbulence may become stationary. Mathematically, the instability of the
92: faster oscillations is a result of  a CF 
93: bifurcation driven by the terms representing enzymatic regulations,
94: suggesting that weak turbulence may be common 
95: in biochemical birhythmic media exhibiting CF bifurcations.
96: As further evidence, we demonstrate weak turbulence 
97: in a cell cycle model. 
98: A biological system where weak turbulence might possibly  be found 
99: is presented in the closing section.
100: 
101: \section{A birhythmic medium of glycolytic oscillators}
102: Let us introduce a birythmic medium composed of glycolytic oscillators,
103: 
104: \begin{eqnarray}
105: \frac{d\alpha}{dt}=\nu+\frac{\sigma_i \gamma^n}{K^n+\gamma^n}-\sigma \phi
106: +D_{\alpha}\Delta\alpha,   {\hspace{0.62cm}}\\
107: \frac{d\gamma}{dt}=Q \sigma \phi -k_s \gamma 
108: -\frac{Q \sigma_i \gamma^n}{K^n+\gamma^n}+D_{\gamma}\Delta\gamma,\\
109: \phi=\frac{\alpha (1+\alpha) (1+\gamma)^2}{L+ (1+\alpha)^2  (1+\gamma)^2}.
110: \hspace{2.15cm}\nonumber
111: \end{eqnarray}
112: 
113: In Eqn. (1-2), $\alpha$ and $\gamma$ represent dimensionless substrate
114: and product concentrations of glycolytic reactions, $K$,
115: $n$, $\nu$, $\sigma_i$, $\sigma$, $k_s$,
116: $L$ and $Q$ are parameters. For convenience, we assume $Q\equiv 1$ throughout
117: this paper. $D_{\alpha}$, $D_{\gamma}$ are  
118: diffusion constants for the substrate and product. 
119: Our units of time and space are {\rm sec} and {\rm cm}, respectfully. 
120: When $D_{\alpha}=0$, $D_{\gamma}=0$ and $\sigma_i=0$, 
121: Eqn. (1-2) is called the glycolytic oscillator. The term
122: $\frac{\sigma_i \gamma^n}{K^n+\gamma^n}$ represents
123: substrate recycling that drives  birhythmicity. 
124: Recently,  in Ref. \cite{gold2}, Eqn. (1-2)  were shown to support 
125: multiple wave fronts.  Our concern in this paper is 
126: a different parameter region where irregular spatiotemporal  
127: motions develop. 
128: 
129: A phase plane analysis of Eqn. (1-2) shows that the mechanism of birhythmicity 
130: is two regions of negative slope in the product nullcline \cite{gold1}. 
131: A convenient way  to illustrate birhythmicity is a bifurcation diagram. 
132: We used a  well known software package, AUTO \cite{doedel}, for 
133: bifurcation analysis of the local model($D_{\alpha}=D_{\gamma}=0$
134: in Eqn. (1-2)).  Solid lines in Fig. 1 show stable steady states,
135: dashed lines show unstable steady states.  Stable limit
136: cycles are shown by filled symbols, unstable
137: limit cycles by open circles. Filled circles represent
138: small amplitude oscillations with high frequencies. 
139: Large amplitude oscillations with low frequencies are shown by 
140: filled diamonds in Fig. 1. A Hopf bifurcation point $HB$ is located
141: at $\sigma_{i,cr} \approx 1.282$.  There are two 
142: CF bifurcations in Fig. 1, where 
143: stable limit cycles are replaced
144: by unstable ones. Between these two CF points,
145: which occur  at 
146: $\sigma_{i,CF^1} \approx 1.77$ and $\sigma_{i,CF^2} \approx 1.83$,
147: two stable limit cycles coexist. Therefore, 
148: depending on initial conditions, one of the 
149: limit cycles will be selected in simulations of 
150: the glycolytic oscillator with substrate recycling. 
151: 
152: A general mechanism of turbulence in oscillatory
153: reaction diffusion systems is wavenumber instability, i.e.,
154: instability of uniform oscillations against phase-like 
155: fluctuations \cite{kura1}. 
156: In Eqn. (1-2), there are
157: two different uniform oscillations that might undergo 
158: wavenumber  instability. We want to provide evidence that
159: these oscillations are stable against phase-like fluctuations
160: for the parameters in Fig. 1.
161: For the fast, uniform oscillations
162: which originate from the Hopf bifurcation point 
163: shown by filled circles in Fig. 1,  the stability condition 
164: can be obtained by reducing Eqn. (1-2) to the CGLE,
165: \begin{equation}
166: \dot A=(1+ i c_0) A-(1+i c_2)|A|^2 A+(1+i c_1) \Delta A.
167: \end{equation}
168: In Eqn. (3) $A$ is the complex amplitude, and $\omega$, $\beta$, $\gamma$ are
169: real parameters. The CGLE has a uniform oscillatory 
170:  solution,  $A=exp(i (c_0-c_2) t)$, that is stable if the  condition
171: $ 1+ c_1 c_2 > 0$ holds.
172: In the appendix, we calculated $c_0$, $c_1$ and 
173: $c_2$ corresponding to Eqn. (1-2).
174: Our results show that the uniform  oscillations are stable for
175: the parameters used in Fig. 1, for any $D_{\alpha}>0$
176: and  $D_{\gamma}>0$. For $D_{\alpha}=D_{\gamma}$ we find that  $c_1=0$. 
177: Hence, the parameter region we are interested in is deep inside  
178: the Benjamin-Feir stability region  given by  $ 1+ c_1 c_2 > 0$. 
179: Although  the CGLE is valid only near the $HB$ point, 
180: it is likely that the uniform  oscillations will remain stable 
181: until the the next bifurcation in the system, i.e., 
182: $CF^1$ in Fig. 1 \cite{kura2}. Next,  consider the uniform 
183: oscillations with low frequencies.
184: Unlike the case of fast oscillations, 
185: no analytic approach is available in this case.
186: Note that oscillations shown by filled circles and diamonds in Fig. 1
187: occur at the same parameters. Therefore, it is rather unlikely 
188: that the slow oscillations undergo wavenumber instability, 
189: contrary to the fast ones. Thus, we can assume that uniform, 
190: slow oscillations are also stable.
191: 
192: It is known that strong perturbations can switch oscillations from one stable 
193: orbit to another in the glycolyctic model with substrate recycling \cite{gold1}. 
194: Therefore, even if both uniform oscillations in Eqn. (1-2)  are stable 
195: against wavenumber instability, strong perturbations can excite the system 
196: by switching the oscillations. This kind of excitability, however, will 
197: not lead to turbulence; in the parameter 
198: interval $[\sigma_{i,CF^1},\sigma_{i,CF^2}]$,
199: the fast oscillations will suppress 
200: the slow ones as time progresses. 
201: But for $|\sigma_i-\sigma_{i,CF^1}| \ll 1 $
202: where the fast  oscillations become unstable, it is apparent  that
203: a complete suppression of slow oscillations is impossible. 
204: Here, because of complex interactions between stable, slow and  unstable, 
205: fast oscillations,  interesting spatiotemporal dynamics might develop.
206: Therefore, we carried out a detailed numerical study in the 
207: neighborhood of $CF^1$.
208: 
209: \section{Weak turbulence 
210: in a birhythmic medium of glycolytic oscillators}
211: For numerical integrations of Eqn. (1-2) in 
212: one spatial dimension, we used the fourth order
213: Runge-Kutta method. Diffusion terms were approximated by the 
214: finite difference method. Numerical parameters are $\delta x=0.005{\rm cm}$, 
215: $\delta t=0.05{\rm s}$. The system size is defined as $l=N \delta x$, 
216: where $N$ is the number of spatial grid points. 
217: In this paper we present results for
218: periodic boundary conditions, but  we also tested the main results with 
219: no-flux boundary conditions. We also tested selected examples with smaller
220: values of $\delta x$ and $\delta t$ for fixed $l$.
221: Our simulations show that 
222: Eqn. (1-2) are sensitive to initial conditions. By choosing 
223: initial conditions as small perturbations of uniform, 
224: slow oscillations with large amplitudes, 
225: we found that these oscillations are stable
226: for $\sigma_i<\sigma_{i,CF^2}$. But, near and to the left of $CF^1$, 
227: uniform, fast  oscillations with small amplitudes
228: are found to be unstable. They spiral out from unstable  orbits towards
229: the  orbit of stable, large amplitude oscillations.  
230: For strong perturbations near the $CF^1$ bifurcation point,
231: we found spatiotemporal irregular motions in Eqn. (1-2). 
232: 
233: Fig. 2 shows a gray scale plot of spatiotemporal 
234: dynamics in Eqn. (1-2). Oscillations between the white and black colors show 
235: large amplitude oscillations displayed by $\gamma(x,t)$. There are also 
236: oscillations with  higher frequency and smaller amplitude in Fig. 2.
237: Because the latter ones  are unstable, they  
238: can not suppress large amplitude domains.
239: Although uniform, large amplitude oscillations are stable 
240: against small fluctuations,
241: phase slips created by strong initial perturbations cannot be 
242: eliminated as time progresses. As a result, spatially 
243: nonuniform distributions of concentrations are seen at 
244: given time moments, Fig. 3. On the phase plane, these 
245: nonuniform distributions generate motions attracted by unstable 
246: orbits around the inner cycle shown in Fig. 4. We found that such 
247: unstable orbits act as a weak, virtual heterogeneity emerging
248: randomly. They cannot entrain the bulk oscillations, 
249: but in their presence, phase slips cannot be eliminated. 
250: Instead, persistent spatiotemporal irregular motions develop.
251: 
252: To characterize the irregular motions in Fig. 2,
253: we calculated the maximum Lyapunov exponent $\lambda^{max}_{lyap}$
254: in $2N$ dimensional phase space \cite{wolf}. First we made a very 
255: long run of Eqn. (1-2) to confirm that the turbulence is stationary. 
256: Then, by using the same initial conditions, 
257: we simulated Eqn. (1-2) and its linear
258: system for  computation of $\lambda^{max}_{lyap}$
259: for $T_1=2 \cdot 10^5{\rm s}$. 
260: We found that the largest Lyaponov exponent 
261: is positive and small,
262: $\lambda^{max}_{lyap} \approx 2\cdot 10^{-3}$. We also calculated 
263: a two-point correlation function, $C(x)=<\gamma(x_0,t) \gamma(x_0+x,t)>$, where
264: $<..>$ stands for an average over space and time \cite{bataa}. 
265: Fig. 5 shows that 
266: $C(x) \approx const$ at small values of $x$, indicating 
267: strong local coupling 
268: and an absence of short waves. A power-law decay of the correlation function
269: at intermediate values of $x$ implies the  presence of chaotic  
270: motions. We found that the slope is $\kappa \approx -0.15$. 
271: We also found no significant variations of $\kappa$
272: and $\lambda^{max}_{lyap}$ 
273: with changes of $\sigma_i$ and $l$. The small values of $\kappa$
274: and $\lambda^{max}_{lyap}>0$ suggest that spatiotemporal 
275: irregular motions shown in Fig. 2-4 can be characterized 
276: as a weak turbulence. 
277: 
278: We found that in Eqn. (1-2), stationary irregular motions  
279: can develop only for certain initial conditions and 
280: system sizes. In simulations with 
281: different initial conditions and system sizes, we observed
282: sudden collapses  of turbulent dynamics. 
283: Collapse of turbulence in Eqn. (1-2) means a complete suppression
284: of small amplitude oscillations.
285: Thus, we defined the transient lifetime of turbulence $t_p$ as the
286: time interval from initial conditions to the moment 
287: when all oscillators come within a distance $d$ of the orbit of stable, slow, 
288: large amplitude oscillations. In our simulations we used $d=0.03$.
289: Following Ref.  \cite{show2}, we plot an average transient 
290: lifetime $t_p$ versus the system size $l$ in Fig. 6. 
291: Here, each filled circles is an  average of 20 simulations  
292: with different initial conditions. Fig. 6 shows that, 
293: as the system size increase, $t_p$ grows exponentially.
294: 
295: For some initial conditions, when $l$ is close to 
296: $2{ \rm cm}$,
297: the turbulent solution does not collapse. 
298: The inset in Fig. 6 shows the number of cases, among 20 different 
299: simulations, when a collapse of turbulence has not occurred by
300: $T=10^6{\rm s}$. (These cases were not included in calculations of 
301: the filled  circles in Fig. 6.) We continued two cases
302: in the inset (at $l=1.75{\rm cm}$) up to $T=10^8 {\rm s}$ and did
303: not observe a collapse of motions near the inner cycle in Fig. 4. 
304: 
305: Numerical experiments indicate that if  virtual heterogeneities 
306: reside sufficiently far from 
307: each other, a stationary pattern is possible
308: in the interval 
309: $\sigma_i \in [1.055,1.075]$.  Fig. 7 gives an example
310: of such a pattern. Here virtual heterogeneities are 
311: located from each other by distances between $ 0.5{\rm cm}$ and  
312: $1{\rm cm}$. Note that these quasi-periodic structures
313: are not related to a  Turing
314: instability, which emerges due to differences in diffusion
315: coefficients. Unstable oscillations at the onset of a CF  bifurcation
316: are the instabilities leading to these structures. 
317: The cellular structures in Fig. 7 are breathing
318: because of the unstable oscillations. Numerical results show 
319: that as  the system size increases, the cells breath coherently.
320: 
321: A collapse of turbulence can be prevented if there is a reinjection
322: mechanism for the unstable oscillations generated by the virtual 
323: heterogeneities. Naturally, wavenumber instability can be such
324: a  mechanism. Using our calculations in the appendix, 
325: we simulated Eqn. (1-2) for
326: parameters when the corresponding CGLE displays a wavenumber instability and 
327: found that small amplitude oscillations exhibit phase instability
328: near Hopf bifurcation. For $|\sigma-\sigma_{i,CF^1}| \ll 1$ we found 
329: stationary weak turbulence.
330: 
331: \section{weak turbulence in a cell cycle model}
332: In section III, we demonstrated that the $CF^1$ bifurcation point
333: is crucial  for turbulence in  Eqn. (1-2). 
334: Mathematically, the term representing  substrate recycling drives
335: CF bifurcations. In models of biochemical oscillations,
336: terms representing enzymatic activities naturally 
337: arise. As an enzyme can quickly switch from being 
338: active to inactive and back again, 
339: ideal conditions for CF bifurcations exist 
340: in these models. Therefore, other
341: biochemical reaction diffusion models may also display
342: the weak turbulence discussed in the previous section.
343: As an example, consider  a three variable model of the 
344: budding yeast cell cycle,
345: 
346: \begin{widetext}
347: \begin{eqnarray}
348: %%%Equation#1
349: \frac{d X}{dt} =m (k_1+k_2 T) - ( k_3+k_4 Y
350: +k_5 Z)X + D_X\Delta X, {\hspace{3cm}}\\
351: %%%Equation#2
352: \frac{d Y}{dt}=\frac{(k_6+k_7 Z)(1-Y)}{J_1+1-Y}-
353: \frac{(k_8 m+k_9 X) Y}{J_1+Y} +  D_Y\Delta Y,  
354: {\hspace{2.75cm}}\\
355: %%%Equation#3
356: \frac{d Z}{dt}=(k_{10}+k_{11} X) -k_{12} Z +  D_Z\Delta Z,
357: {\hspace{5.8cm}}\\
358: %%%Equation#4
359: T=G(X,P, J_2, J_2),{\hspace{8.7cm}}\\
360: G(a,b,c,d)= 
361: \frac{2 a d}{ b-a+b c +a d +\sqrt{(b-a+b c +a d)^2-4 a d (b-a)}},
362:  \hspace{0.45cm}
363: \end{eqnarray}
364: \end{widetext}
365: where the transcription factor $T$  for $X$ is given by the 
366: Goldbeter-Koshland function $G$ \cite{gold1}. $X$, $Y$, $Z$ 
367: are dimensionless variables 
368: and $m$ is a dimensionless parameter.  Our units of time and space are 
369: {\rm min} and {\rm cm}, respectfully. 
370: 
371: When $D_X=D_Y=D_Z=0$, Eqn. (4-8) are  the reduced version of 
372: a budding yeast cell cycle model \cite{kathy}.
373: Here, $X$ represents the  concentration of  cyclin-dependent 
374: protein kinase(CDK),
375: $Y$ and $Z$ are concentrations of 
376: two different anaphase promoting complexes(APC),
377: APC/Cdh1 and APC/Cdc20 respectively. 
378: In Eqn. (4-5), $m$ represents the cell's  mass, which will be used as a primary
379: bifurcation parameter.  Eqn. (4-8) display CF bifurcations 
380: as shown in Fig. 8.
381: For small $m$, Eqn. (4-8) also display saddle node bifurcations, 
382: a feature universal in cell cycle
383: models \cite{borisuk,kathy}. Here, our concern is 
384: the neighborhood of $CF^1$  in Fig. 8. 
385: 
386: There are no experimental measurements of  diffusion coefficients for 
387: CDK, and APC factors. But it is known that diffusion coefficients
388: of average sized proteins in cytoplasm are order of 
389: $10^{-4}\frac{{\rm cm^2}}{{\rm min}}$ or smaller \cite{woj,zubay}. 
390: As our goal is a demonstration of weak turbulence 
391: in a representative model of biochemical oscillations, 
392: we choose $D_X$ and $D_Z$ arbitrarily subject to this upper bond. For
393: simplicity, we assume  $D_Z=0$. 
394: 
395: For simulations of Eqn. (4-8) we used the same method as in the previous
396: section with $\delta t=0.05{\rm min}$, $\delta x=0.005{\rm cm}$.
397: We found numerically that for strong perturbations,
398: Eqn. (4-8) display a weak turbulence, Fig. 9.
399: Typically, for  $D_X \leq D_Y$, 
400: we found transient, weak turbulence. 
401: When  $D_X < < D_Y$, numerical experiments lead to
402: stationary turbulence. For instance, 
403: we simulated Eqn. (4-8)
404: up to  $T=10^7{\rm min}$ for $m=3$, 
405: $D_X=6\cdot10^{-7}\frac{ {\rm cm^2}}{{\rm min}}$, 
406: $D_Y=10^{-4}\frac{{\rm cm^2}}{{\rm min}}$, $D_Z=0$ 
407: and  $l=1.28{\rm cm}$ and found 
408: stationary turbulence for a number of different initial conditions. 
409: 
410: 
411: \section{Discussion}
412: We have shown in this paper that two representative mathematical models of
413: biochemical oscillations exhibiting  birhythmicity, glycolytic and cell 
414: cycle models, display weak turbulence, intermiitency
415: of large and small amplitude oscillations. We revealed that 
416: unstable oscillations near cyclic fold bifurcations
417: are the mechanism of transient turbulence in birhythmic media.
418: In the presence of wavenumber instability, weak turbulence is stationary.
419: 
420: Recently,  Stich et. al. \cite{alex1,alex2} proposed an amplitude model for 
421: birhythmic media. An interesting question is whether the weak 
422: turbulence we discussed in this paper 
423: can be found in their model? First, let us mention two important 
424: differences between our models and the amplitude model of birhythmic media. 
425: In our case, a cyclic fold bifurcation is crucial for turbulence,
426: but the amplitude model describes a pitch-fork bifurcation
427: of limit cycles. Secondly, both fast and slow 
428: oscillations in the amplitude equation are smooth, but in our case, 
429: slow oscillations are strongly anharmonic. 
430: Besides these differences, it is well known that 
431: if phase slips develop,
432: the CGLE equation generates defects \cite{chate}.
433: Thus, these facts indicate that
434: instead of intermittency of small and large
435: amplitude oscillations, defect
436:  turbulence
437: is likely in the amplitude model of birhythmic media. 
438: 
439: To date, there are no experimental evidences of weak turbulence in glycolysis
440: or in the cell cycle. Our results are pure theoretical predictions of mathematical 
441: models. The system sizes we simulated are much larger 
442: than the typical size of an yeast cell ($10^{-3}{\rm cm}$). Therefore, weak 
443: turbulence is not expected in yeasts. Interestingly, 
444: some slime molds grow as syncytial 
445: plasmodia (many nuclei in a common cytoplasmic pool) that are many 
446: times larger than a typical yeast cell;
447: cells $15{\rm cm}$ in diameter can be grown in the laboratory \cite{moh}.
448: Waves of nuclear division are observed in these multinucleate
449: plasmodia \cite{haskin,tysonnovak}, and, as we have shown it is possible that 
450: these waves exhibit weak turbulence.  
451: Note that weak turbulence in the cell cycle would
452: mean irregular oscillations of CDK. But for a normal cell cycle, 
453: large amplitude oscillations of CDK are essential; CDK activity must drop below 
454: a certain threshold for nuclei to exit mitosis and divide. Therefore, 
455: hypothetically, weak turbulence in syncytial plasmodia might lead 
456: to mitotic arrest of certain nuclei in the plasmodium.
457: 
458: A more quantitative characterization of wavenumber instability of unstable 
459: oscillations at the onset of a cyclic fold bifurcation,
460: as well as simulations in two spatial 
461: dimensions \cite{bat} are problems in the future.
462: 
463: \appendix
464: \section{Coefficients of the  CGLE for a glycolytic model with substrate inhibition}
465: In this appendix, 
466: following standard procedures  in Ref. \cite{kura1}, 
467: we will calculate
468: coefficients of  CGLE for the glycolytic model.
469: For a convenience we assume $Q\equiv 1$ in Eqn. (1-3).
470: First, let us find uniform steady state solutions $\alpha_0$ and $\gamma_0$,
471: \begin{eqnarray}
472: \gamma_0=\mu/k_s, \hspace{6cm}\\
473: \alpha_0=\frac{
474: K^4 (-2\mu+\sigma)+\gamma_0^4(-2(\mu+\sigma_i)+\sigma) }
475: {\tilde c
476: }-\nonumber\hspace{2cm}\\
477: -\frac{
478: \sqrt{-4 \tilde a^2
479: +4 L\sigma \tilde a \tilde b
480: +(1+\gamma_0)^2\sigma^2\tilde b^2}
481: }
482: {
483: (1+\gamma_0)\tilde c
484: }\nonumber\hspace{2cm}\\
485: \end{eqnarray}
486: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
487: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
488: where, 
489: $\tilde a=K^4\mu+\gamma_0^4(\mu+\sigma_i)$,
490: $\tilde b=(K^4+\gamma_0^4)$,
491: $\tilde c=2(K^4 (\mu-\sigma)+\gamma_0^4(\mu+\sigma_i-\sigma))$.
492: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
493: Next we perform a linear stability analysis of $(\alpha_0,\gamma_0)$ 
494: against small fluctuations $\delta \alpha, \delta \gamma \propto
495: exp(iq x+i \lambda t)$. At the critical wavenumber $q_{cr}=0$, we obtain
496: a characteristic equation,
497: \begin{equation}
498: \lambda^2+(a_1+a_2+k_s)\lambda+a_1k_s=0.
499: \end{equation} 
500: In Eqn. (A3), 
501: $a_1$ and $a_2$ are given by 
502: \begin{equation}
503: a_1=
504: \frac{\sigma(1+\gamma_0)^2(L+2 L\alpha_0+(1+\alpha_0)^2(1+\gamma_0)^2)}
505: {(L+(1+\alpha_0)^2(1+\gamma_0)^2)^2},
506: \end{equation}
507: \begin{equation}
508: a_2=\frac{4 K^4 \gamma_0^3 \sigma_i}{(K^4+\gamma_0^4)^2}-
509: \frac{2\sigma L\alpha_0(1+\alpha_0)(1+\gamma_0)}
510: {(L+(1+\alpha_0)^2(1+\gamma_0)^2)^2}.
511: \end{equation} 
512: Let us define  such a critical value  for the bifurcation 
513: parameter $\sigma_i=\sigma_{i,cr}$  that 
514: \begin{equation}
515: a_1+a_2+k_s\equiv0.
516: \end{equation}
517: Eqn. (A6) is the condition for a Hopf bifurcation;
518: the characteristic equation has 
519: pure imaginary solutions,
520: $\lambda_0=\pm i\sqrt{a_1 k_s}$. 
521: 
522: Let $\mu$ be defined by $\mu=\frac{\sigma_i-\sigma_{i,cr}}{\sigma_{i,cr}}$. 
523: We develop the Jacobian matrix $L$ of Eqn. (1-3) in powers of $\mu$,
524: \begin{equation}
525: L=L_0+\mu L_1+....
526: \end{equation}
527: At $\mu=0$ the Jacobian  is given by
528: \begin{eqnarray}
529: L_0=
530: \left|
531: \begin{array}{c}
532: \begin{array}{cc}
533: -a_1\\
534: a_1
535: \end{array}
536: \begin{array}{cc}
537: \ \ \ \ a_2\\
538: \ \ \ \ -k_s-a_2.
539: \end{array}
540: \end{array}\right|.
541: \end{eqnarray} 
542: We find the right $\bf{u}_0$  and 
543: left ${\bf{u}_0}^*$ eigenvectors of $L_0$ corresponding to
544: $\lambda_0$, 
545: \begin{eqnarray}
546:  \bf{u}_0=
547: \left(
548: \begin{array}{c}
549: -1+ i \sqrt{\frac{k_s}{a_1}}\\
550: 1
551: \end{array}\right),
552: \end{eqnarray} 
553: \begin{equation}
554: { \bf{u}_0}^*= \frac{1}{2} \left(-i 
555: \sqrt{\frac{a_1}{ k_s}}, 1 - i \sqrt{\frac{a_1}{ k_s}}\right).
556: \end{equation} 
557: We find further,
558: \begin{eqnarray}
559: L_1=
560: \frac{4 K^4 \gamma_0^3  \sigma_{i,cr}}{(K^4+\gamma_0^4)^2 }\left|
561: \begin{array}{c}
562: \begin{array}{cc}
563: 0\\
564: 0
565: \end{array}
566: \begin{array}{cc}
567: \ \ \ \ -1\\
568: \ \ \ \ 1
569: \end{array}
570: \end{array}\right|.
571: \end{eqnarray} 
572: Let us first find $c_0$ in the CGLE.
573: It  is given by $c_0=\frac{Im\lambda_1}{Re\lambda_1}$, where
574: \begin{equation}
575: \lambda_1={ \bf{u}_0}^* L_1{\bf{u}_0}= 
576: \frac{2 K^4 \gamma_0^3}{(K^4+\gamma_0^4)^2} \sigma_{i,cr}.
577: \end{equation} 
578: We see that   $\lambda_1$ is pure real, therefore, 
579: $c_0=0$. Now following again \cite{kura1}, 
580: we find $c_1$, 
581: \begin{eqnarray}
582: D=
583: \left|
584: \begin{array}{c}
585: \begin{array}{cc}
586: D_{\alpha}\\
587: 0
588: \end{array}
589: \begin{array}{cc}
590: \ \ \ \ 0\\
591: \ \ \ \ D_{\gamma}
592: \end{array}
593: \end{array}\right|,
594: \end{eqnarray} 
595: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
596: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
597: 
598: \begin{equation}
599: d'+id''={ \bf{u}_0}^* D \bf{u}_0,
600: \end{equation} 
601: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
602: \begin{equation}
603: c_1=d''/d'=\sqrt{\frac{a_1}{k_s}}
604: \left(\frac{D_{\alpha}-D_{\gamma}}{D_{\alpha}+D_{\gamma}}\right).
605: \end{equation}
606: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
607: Calculation of $c_2$ is a little more tedious. We need to find \cite{kura1}, 
608: \begin{eqnarray}
609: {\bf V}_+={\bf \bar{V}}_-=-{(L_0-2 \lambda_0)}^{-1}{\bf M}_0{\bf u}_0{\bf u}_0,\\
610: {\bf V}_0=-2 {L_0}^{-1} {\bf M}_0 {\bf u}_0 {\bar{\bf u}}_0,
611: \end{eqnarray}
612: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
613: \begin{eqnarray}
614: g=g'+i g''= -2 { \bf{u}_0}^*{\bf M}_0{\bf u}_0 {\bf V}_0-
615: 2  { \bf{u}_0}^*{\bf M}_0  {\bar{\bf u}}_0 {\bf V}\nonumber\\
616: -3 { \bf{u}_0}^* {\bf N}_0 { \bf{u}_0}^* { \bf{u}_0}^* {\bar{\bf u}}_0.
617: \hspace{1.5cm}
618: \end{eqnarray}
619: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
620: Parameter $c_2$ in the CGLE is given by a formula, $c_2=g''/g'$. 
621: We find that $c_2=\frac{\tilde g''} {\tilde g'}$ , where
622: \begin{eqnarray}
623: \tilde g'=-3 k_s [k_s m_{ \alpha^2} (2 m_{\alpha^2}-m_{\alpha\gamma})
624: +a_1((2 m_{\alpha^2}-m_{\alpha\gamma})
625: (m_{\alpha^2}-m_{\alpha,\gamma}+m_{\gamma^2})\nonumber
626: \ \ \ \ \ \ \ \ \ \ \\
627: -k_s ({n_{\alpha^2\gamma}}-3 {n_{\alpha^3}}))
628: +3 a_1^2( {n_{\alpha\gamma^2}}-{n_{\alpha^2\gamma}}+ {n_{\alpha^3}}-
629:  {n_{\gamma^3}})], \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \\
630: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
631: \tilde g''=\sqrt{\frac{k_s}{a_1}}[10 k_s^2 {m_{\alpha^2}}^2+ 
632: a_1 k_s (14 {m_{\alpha^2}}^2-14 {m_{\alpha^2}}m_{\alpha,\gamma}
633: +{m_{\alpha\gamma}}^2+10 {m_{\alpha^2}} {m_{\gamma^2}}
634: +9k_s {n_{\alpha^3}})+\nonumber\\
635: +a_1^2 {(4 ({m_{\alpha^2}}-
636: m_{\alpha,\gamma}+{m_{\gamma^2}})}^2+
637: 3 k_s ( {n_{\alpha\gamma^2}}-2 {n_{\alpha^2\gamma}}+3 {n_{\alpha^3}}))].
638: \ \ \ \ \ \ \ \ \ \ \ \ \ \ 
639: \end{eqnarray}
640: In the above expressions, 
641: $m_{\alpha^2}=(\frac{\partial^2 \phi(\alpha,\gamma)}
642: {\partial\alpha^2})_{\alpha_0,\gamma_0}$,
643: $m_{\alpha\gamma}=(\frac{\partial^2 \phi(\alpha,\gamma)}
644: {\partial\alpha\partial\gamma})_{\alpha_0,\gamma_0}$,
645: $m_{\gamma^2}=\frac{2 (3 K^8 \gamma_0^2-5 K^4 \gamma_0^6) \sigma_{i,cr}}
646: {(K^4+\gamma_0^4)^3}-
647: (\frac{\partial^2 \phi(\alpha,\gamma)}
648: {\partial\gamma^2})_{\alpha_0,\gamma_0}$, 
649: $ {n_{\alpha^3}}= 
650: (\frac{\partial^3 \phi(\alpha,\gamma)}
651: {\partial\alpha^3})_{\alpha_0,\gamma_0}$,
652: ${n_{\alpha^2\gamma}}=
653: (\frac{\partial^3 \phi(\alpha,\gamma)}
654: {\partial\alpha^2\partial\gamma})_{\alpha_0,\gamma_0}$,
655: ${n_{\alpha\gamma^2}}=
656: (\frac{\partial^3 \phi(\alpha,\gamma)}
657: {\partial\alpha\partial\gamma^2})_{\alpha_0,\gamma_0}$,
658: ${n_{\gamma^3}}=
659: (\frac{\partial^3 \phi(\alpha,\gamma)}
660: {\partial\gamma^3})_{\alpha_0,\gamma_0}$. To save space we
661: do not present here cumbersome expressions for
662: these derivatives.
663: 
664: For parameters in Fig. 1 we find that $\sigma_{i,cr}=1.282$ and
665: $c_2\approx 2.21$. From Eqn. (A15) we see that if  $D_{\alpha}=
666: D_{\gamma}$, $c_1=0$. Therefore, $1+c_1 c_2>0$ for parameters used in
667: this paper. If $D_{\alpha}=0$, Eqn. (A15) gives the minimal
668: value, $c_1=-0.47$. In this case $1+c_1 c_2 \approx -0.03$, therefore,
669: wavenumber instability is possible. However, turbulence must be weak
670: as the parameters are very close to the stability condition $1+c_1 c_2>0$
671: \cite{cross}.  A stronger wavenumber instability is possible, for example, for 
672: $\mu=0.28{\rm s^{-1}}$,  $K=12$, 
673: $D_{\alpha}=5 \cdot 10^{-7}\frac{{\rm cm^2}}{{\rm s}} $, 
674: $D_{\gamma}=1\cdot 10^{-5} \frac{{\rm cm^2}}{{\rm s}}$
675: and other
676: papameters are the same as in Fig. 1. In this case, we find that
677: $\sigma_{i,cr}\approx 1.095$ and $1+c_1 c_2=-0.416$.  
678: 
679:  
680: \begin{thebibliography}{}
681: %1
682: \bibitem{kura1} Y. Kuramoto, {\em Chemical Oscillations, Waves and 
683: Turbulence}, Springer-Verlag, Berlin, 1984.
684: %
685: \bibitem{cross} M. C. Cross and P. C. Hohenber, Rev. Mod. Phys. {\bf 65},
686: 851 (1993).
687: %
688: \bibitem{kbn} Y. Kuramoto, D. Battogtokh, and H. Nakao, 
689: Phys. Rev. Lett. {\bf 81}, 3543 (1998). 
690: %
691: \bibitem{nozaki} K. Nozaki and N. Bekki, Phys. Lett. A {\bf 110},
692: 133 (1985).
693: %
694: \bibitem{shraiman} B. I. Shraiman, A. Pumir, W. van. Saarlos, 
695: P. C. Hohenberg, H. Chate, and M. Holen, Physica D {\bf 57}, 241 (1992).
696: %
697: \bibitem{saka} H. Sakaguchi, Progr. Theor. Phys. {\bf 84}, 792 (1990).
698: 
699: \bibitem{kramer1} I. S. Aronson, L. Kramer, Rev. Mod. Phys. {\bf 74},
700: 99 ( 2002).
701: %
702: \bibitem{kura2} H. Mori and Y. Kuramoto, {\em Dissipative
703: Structures and Chaos}, Springer-Tokyo, 1998. 
704: %
705: \bibitem{kramer2} M. Ipsen, L. Kramer, P. G. Sorensen, Physics Reports
706: {\bf 337}, 193(2000).
707: %
708: \bibitem{show1} J. H. Merkin, V. Petrov, S. K. Scott and K. Showalter,
709: Phys. Rev. Lett. {\bf 76}, 546(1996).
710: %
711: \bibitem{yama} Y. Nishiura and D. Ueyama, Physica D {\bf 150}, 137(2001).
712: %
713: \bibitem{show2} R. Wackerbauer and K. Showalter, Phys. Rev. Lett.
714: {\bf 91}, 174103, 2003.
715: %
716: \bibitem{gold1} A. Goldbeter, {\em Biochemical Oscillations and Cellular 
717: Rhythms. The Molecular Bases of Periodic and Chaotic Behavior},
718: Cambridge University Press, UK, 1996.
719: %
720: \bibitem{borisuk} M. T. Borisuk and J. J. Tyson, Journ. Theor. Biol.
721: {\bf 195},  69 (2000).
722: %
723: \bibitem{gold2} C. Perez-Iratxeta, J. Halloy, F. Moran, J. L. Martel,
724: A. Goldbeter, Biophys. Chem. {\bf 74}, 197 (1998).
725: %
726: \bibitem{doedel} E. J. Doedel, A. R. Champneys, T. F. Fairgrieve, 
727: Y. A. Kuznetsov, B. Sandstede, X. Wang, 
728: {\em Auto97, Continuation and Bifurcation Software for
729: Ordinary Differential Equations(with HomCont)},  1997. 
730: %
731: \bibitem{wolf} A. Wolf, J. B. Swift, H. L. Swinney and J. Vastano, 
732: Physica D {\bf 16}, 285 (1985).
733: %
734: \bibitem{bataa} D. Battogtokh, Phys. Rev. E {\bf 66}, 066202 (2002).
735: %
736: \bibitem{kathy} K. Chen, A. Csikasz-Nagy, B. Georffy, J. Val, 
737: B. Novak, J. J. Tyson, Mol. Biol. Cell {\bf 11}, 369 (2000);
738: D. Battogtokh and J. J. Tyson, preprint (2004).
739: %
740: \bibitem{woj} J. W. Wojcieszyn, R. A. Schlegel, E. S. Wu,
741: K. A. Jacobson, PNAS USA {\bf 78}, 4407 (1981).
742: %
743: \bibitem{zubay} G. Zubay, {\em Biochemistry}, Addison Wesley, 
744: Reading MA, 1983.
745: %
746: \bibitem{alex1} M Stich, M. Ipsen and A. S. Mikhailov, Phys. Rev. Lett.,
747: {\bf 86}, 4406 (2001).
748: %
749: \bibitem{alex2} M. Stich, M. Ipsen, A. S. Mikhailov,
750: Physica D {\bf 171}, 19 (2002).
751: %
752: \bibitem{chate} H. Chate, Nonlinearity {\bf 7}, 185, (1994).
753: %
754: \bibitem{moh} J. Mohberg and H. P. Rusch, Journal of Bacteriology {\bf 97}, 
755: 1411 (1969).
756: %
757: \bibitem{haskin} E. F. Haskins, {\em Film E 2000} Goetingen. Inst. Wiss. Film
758:  1974.
759: %
760: \bibitem{tysonnovak} B. Novak and J. J. Tyson, Journ. Theor. Biol.
761: {\bf 165}, 101 (1993).
762: 
763: \bibitem{bat} Our preliminary results  show existence of the weak 
764: turbulence in dimension two.
765: \end{thebibliography}
766: 
767: \listoffigures
768: 
769: FIG. 1.  A bifurcation diagram of Eqn. (1-3). {\rm HB} marks 
770: a Hopf bifurcation point, ${\rm CF^{1,2}}$  mark cyclic fold
771: bifurcations. Parameters are:
772: $\nu =0.25{\rm s^{-1}}$, $n=4$, $K=11.5$, $\sigma=11{\rm s^{-1}}$, 
773: $k_s=0.05{\rm s^{-1}}
774: $, $L=3400000$.
775: 
776: FIG. 2.  Space-time pattern of $\gamma$ in 
777: a weak turbulent regime of Eqn. (1-2).
778: The space and time spans are $l=1.75{\rm cm}$ 
779: and $T=5\cdot10^3{\rm s}$. The pattern was obtained
780: by recording $\gamma(x)$ with a time interval $\tau=5{\rm s}$.
781: $D_{\alpha}=D_{\gamma}=1\cdot 10^{-5} \frac{{\rm cm^2}}{{\rm s}}$
782: and $\sigma_i=1.065{\rm s^{-1}}$.  
783: Other parameters are the same as in Fig. 1.
784: 
785: FIG. 3.  Snapshots of spatial distributions of $\alpha$ at two different
786: time moments. Parameters are the same as in Fig. 2.
787: 
788: FIG. 4.  A phase plane view. The outer cycle shows the 
789: orbit of stable uniform oscillations with a period $\tau=300{\rm s}$.
790: The inner cycle shows the orbit of small amplitude,
791: fast oscillations with a period  $\tau=290 {\rm s}$ at 
792: $\sigma_i=1.08{\rm s^{-1}}$.
793: With the decrease of $\sigma_i$, the inner cycle disappears,
794: but it still can  attract neighboring trajectories creating
795: a virtual, chaotic heterogeneity in Eqn. (1-2). The solid lines 
796: show oscillator distributions at two different time moments.
797: Parameters are the same as in Fig. 2.
798: 
799: FIG. 5.  A log-log plot of the 
800: spatial correlation function. Parameters are the same as in Fig. 2.
801: 
802: FIG. 6   Average transient lifetime versus the system size. The inset
803: shows the cases when a collapse of turbulence has not occurred by
804: $T=10^6{\rm s}$. Parameters are the same as in Fig. 2.
805: 
806: FIG. 7.  Breathing periodic structures. $l=3.5 {\rm cm}$,
807: other parameters, as well as the time 
808: and space spans are the same as in Fig. 2. 
809: 
810: FIG. 8.  Bifurcation diagram of a cell cycle model.
811: Rate constants  $k_i$ are in units ${\rm min^{-1}}$,
812: $k_1=0.002$, $k_2=0.053$,
813: $k_3=0.01$, $k_4=2$, 
814: $k_5=0.05$, $k_6=0.04$,
815: $k_7=1.5$, $k_8=0.19$, $k_9=0.64$, 
816: $k_{10}=0.005$, $k_{11}=0.07$, $k_{12}=0.08$.
817: Other parameters are 
818: $P=0.15$, $J_1=0.05$, and $J_2=0.01$, 
819: $l=1.28{\rm cm}$, 
820: $D_X=6\cdot10^{-7}\frac{{\rm cm^2}}{{\rm min}}$, 
821: $D_Y=10^{-4}\frac{{\rm cm^2}}{{\rm min}}$ and $D_Z=0$.
822: 
823: 
824: FIG. 9. Turbulence in a cell cycle model.
825: Space time plot of $Y$ field in Eqn. (4-8). 
826: The space and time spans are $L=1.28{\rm cm}$ 
827: and $T=2500{\rm min}$. The pattern was obtained
828: by recording $Y(x)$ with a time interval $\tau=5{\rm min}$.  
829: 
830: %\newpage
831: %\newpage
832: %\centerline{
833: %\epsfig{file=Figure1.eps,scale=1,angle=0}}
834: %\bigskip
835: %\centerline{{\LARGE Figure 1}}
836: 
837: %\newpage
838: %\bigskip
839: %\centerline{{\LARGE Figure 2}}
840: %\epsfig{file=Figure2.eps,scale=1,angle=0,
841: %bbllx=100,bblly=500,bburx=350,bbury=700}
842: 
843: 
844: %\newpage
845: %\centerline{
846: %\epsfig{file=Figure3.eps,scale=1,angle=0}}
847: %\bigskip
848: %\centerline{{\LARGE Figure 3}}
849: 
850: %\newpage
851: %\centerline{
852: %\epsfig{file=Figure4.eps,scale=1,angle=0}}
853: %\bigskip
854: %\centerline{{\LARGE Figure 4}}
855: 
856: %\newpage
857: %\centerline{
858: %\epsfig{file=Figure5.eps,scale=1,angle=0}}
859: %\bigskip
860: %\centerline{{\LARGE Figure 5}}
861: 
862: %\newpage
863: %\centerline{
864: %\epsfig{file=Figure6.eps,scale=1,angle=0}}
865: %\bigskip
866: %\centerline{{\LARGE Figure 6}}
867: 
868: %\newpage
869: %\bigskip
870: %\centerline{{\LARGE Figure 7}}
871: %\epsfig{file=Figure7.eps,scale=1,angle=0,
872: %bbllx=100,bblly=500,bburx=350,bbury=700}
873: 
874: %\newpage
875: %\centerline{
876: %\epsfig{file=Figure8.eps,scale=1,angle=0}}
877: %\bigskip
878: %\centerline{{\LARGE Figure 8}}
879: 
880: %\newpage
881: %\bigskip
882: %\centerline{{\LARGE Figure 9}}
883: %\epsfig{file=Figure9.eps,scale=1,angle=0,
884: %bbllx=100,bblly=500,bburx=350,bbury=700}
885: 
886: \end{document}