1: %% ****** Start of file template.aps ****** %
2: %%
3: %%
4: %% This file is part of the APS files in the REVTeX 4 distribution.
5: %% Version 4.0 of REVTeX, August 2001
6: %%
7: %%
8: %% Copyright (c) 2001 The American Physical Society.
9: %%
10: %% See the REVTeX 4 README file for restrictions and more information.
11: %%
12: %
13: % This is a template for producing manuscripts for use with REVTEX 4.0
14: % Copy this file to another name and then work on that file.
15: % That way, you always have this original template file to use.
16: %
17: % Group addresses by affiliation; use superscriqptaddress for long
18: % author lists, or if there are many overlapping affiliations.
19: % For Phys. Rev. appearance, change preprint to twocolumn.
20: % Choose pra, prb, prc, prd, pre, prl, prstab, or rmp for journal
21: % Add 'draft' option to mark overfull boxes with black boxes
22: % Add 'showpacs' option to make PACS codes appear
23: % Add 'showkeys' option to make keywords appear
24: \documentclass[aps,pre,preprint,groupedaddress,showpacs]{revtex4}
25: %\documentclass[aps,prl,preprint,superscriptaddress]{revtex4}
26: %\documentclass[aps,prl,twocolumn,groupedaddress]{revtex4}
27: \usepackage{graphicx}
28: % You should use BibTeX and apsrev.bst for references
29: % Choosing a journal automatically selects the correct APS
30: % BibTeX style file (bst file), so only uncomment the line
31: % below if necessary.
32: \bibliographystyle{apsrev}
33:
34: \begin{document}
35: \newcommand{\ud}{{\mathrm d}}
36: \newcommand{\umod}{\mathrm{mod}}
37: \newcommand{\sech}{\mathrm{sech}}
38: % Use the \preprint command to place your local institutional report
39: % number in the upper righthand corner of the title page in preprint mode.
40: % Multiple \preprint commands are allowed.
41: % Use the 'preprintnumbers' class option to override journal defaults
42: % to display numbers if necessary
43: %\preprint{}
44:
45: %Title of paper
46: \title{Subthreshold Stochastic Resonance: Rectangular signals can cause anomalous large gains }
47:
48: % repeat the \author .. \affiliation etc. as needed
49: % \email, \thanks, \homepage, \altaffiliation all apply to the current
50: % author. Explanatory text should go in the []'s, actual e-mail
51: % address or url should go in the {}'s for \email and \homepage.
52: % Please use the appropriate macro foreach each type of information
53:
54: % \affiliation command applies to all authors since the last
55: % \affiliation command. The \affiliation command should follow the
56: % other information
57: % \affiliation can be followed by \email, \homepage, \thanks as well.
58: \author{Jes\'us Casado-Pascual}
59: \email[]{jcasado@us.es}
60: \homepage[]{http://numerix.us.es}
61: \author{Jos\'e G\'omez-Ord\'o\~nez}
62: \author{ Manuel Morillo}
63: \affiliation{F\'{\i}sica Te\'orica,
64: Universidad de Sevilla, Apartado de Correos 1065, Sevilla 41080, Spain}
65:
66: %\thanks{}
67: %\altaffiliation{}
68:
69: \author{Peter H\"anggi}
70: \affiliation{Institut f\"ur Physik,
71: Universit\"at Augsburg, Universit\"atsstra\ss e 1, D-86135 Augsburg,
72: Germany}
73:
74: %Collaboration name if desired (requires use of superscriptaddress
75: %option in \documentclass). \noaffiliation is required (may also be
76: %used with the \author command).
77: %\collaboration can be followed by \email, \homepage, \thanks as well.
78: %\collaboration{}
79: %\noaffiliation
80:
81: \date{\today}
82:
83: \begin{abstract}
84: The main objective of this work is to explore aspects of stochastic
85: resonance (SR) in noisy bistable, symmetric systems driven by
86: subthreshold periodic rectangular external signals possessing a {\em
87: large} duty cycle of unity. Using a precise numerical solution of the
88: Langevin equation, we carry out a detailed analysis of the behavior of
89: the first two cumulant averages, the correlation function and its
90: coherent and incoherent parts. We also depict the non-monotonic behavior
91: versus the noise strength of several SR quantifiers such as the average
92: output amplitude, i.e. the spectral amplification (SPA), the
93: signal-to-noise ratio (SNR) and the SR-gain. In particular, we find
94: that with {\em subthreshold} amplitudes and for an appropriate duration
95: of the pulses of the driving force the phenomenon of stochastic
96: resonance (SR), is accompanied by SR-gains exceeding unity. This analysis
97: thus sheds new light onto the interplay between nonlinearity and the
98: nonlinear response which in turn yields nontrivial, unexpected SR-gains
99: above unity.
100: \end{abstract}
101:
102:
103:
104: % insert suggested PACS numbers in braces on next line
105: \pacs{05.40.-a, 05.10.Gg, 02.50.-r}
106: % insert suggested keywords - APS authors don't need to do this
107: %\keywords{}
108:
109: %\maketitle must follow title, authors, abstract, \pacs, and \keywords
110: \maketitle
111: \section{Introduction}
112: Most of the studies on the phenomenon of Stochastic Resonance (SR) in
113: dynamical systems have been devoted to systems driven by sinusoidal
114: terms (see \cite{PT96,RMP,Wiesenfeld98,Anish99,Chemphyschem02} for
115: reviews). Several analytical approximations have been put forward to
116: explain SR. In the approach of McNamara and Wiesenfeld \cite{McNWie89},
117: the Langevin dynamics is replaced by a reduced two-state model that
118: neglects the intra-well dynamics. The general ideas of Linear Response
119: Theory (LRT) have been applied to situations where the input amplitude
120: is small \cite{Hanggi78,PR82,RMP,EPL89,GM89,PRA91}. In
121: \cite{EPL89,PRA91} the Floquet theory has been applied to the
122: corresponding Fokker-Planck description. For very low input
123: frequencies, an adiabatic ansatz has been invoked \cite{PRA91}. Even
124: though these alternative analytical approaches provide an explanation of
125: SR for different regions of parameter values, their precise limits of
126: validity remain to be determined. In recent work,
127: \cite{EPLcasado,FNLcasado} we have explored the validity of LRT for
128: sinusoidal and multifrequency input signals with low frequency. Our
129: results indicate a breakdown of the LRT description of the average
130: behavior for low frequency, subthreshold amplitude inputs.
131:
132: Several quantifiers have been used to characterize SR in noisy,
133: continuous, systems. The average output amplitude, or the spectral
134: amplification (SPA), has been studied in Refs.~\cite{EPL89,PRA91} and
135: the phase of the output average in Refs. \cite{dykman92,junhan93},
136: respectively. Those parameters as well as the signal-to-noise ratio
137: (SNR) \cite{McNWie89}, exhibit a non-monotonic behavior with the noise
138: strength which is representative of SR. An important quantity is the
139: SR-gain, defined as the ratio of the SNR of the output over the input
140: SNR. It has been repeatedly pointed out that the SR-gain can not exceed
141: unity as long as the system operates in a regime described by LRT
142: \cite{dewbia95,casgom03}. Beyond LRT there exists no physical reason
143: that prevents the SR-gain to be larger than $1$, as it has been
144: demonstrated in \cite{haninc00} for super-threshold sinusoidal input
145: signals, and in analog experiments in \cite{ginmak00,ginmak01} for
146: subthreshold input signals with many Fourier components and a small
147: duty cycle \cite{casgom03}.
148:
149: In this work, we will make use of numerical solutions of the Langevin
150: equation following the methodology in \cite{casgom03}, to analyze SR in
151: noisy bistable systems, driven by periodic piecewise constant signal
152: with two amplitude values of opposite signs (rectangular signal) (see
153: Fig.~ \ref{fuerza}). There
154: are several relevant time scales in the dynamics of these systems: i)
155: $t_{asym}$, the time interval within each half period of the driving force,
156: during which the diffusing particle sees an asymmetric constant two-well
157: potential; ii) $t_{inter}$, the time scale associated with the inter-well
158: transitions in both directions; and iii) $t_{intra}$, the time scale
159: associated to intra-well dynamics. The inter-well and intra-well time
160: scales depend basically on the noise strength $D$ and the amplitude of
161: the driving term. The dependence of these two time scales with those
162: parameters is certainly very different, being more pronounced for
163: $t_{inter}$. Typically, for the range of parameter values associated
164: with SR, the intra-well time scale is shorter than the inter-well one.
165:
166: We will evaluate the long-time average behavior of the output and the
167: second cumulant. These two quantities were studied some time ago by two
168: of us \cite{Morillo95} for periodic rectangular driving signals. Here,
169: we will further extend our work to the analysis of the correlation
170: function and its coherent and incoherent parts. The knowledge of all
171: these quantities provides a very useful information for the explanation
172: of SR as indicated by the non-monotonic behavior with the noise strength
173: of the output amplitude and the SNR. In particular, the knowledge
174: of the incoherent part of the correlation function is of outmost
175: importance for a correct determination of the SNR. Furthermore, for a
176: given subthreshold amplitude, we will demonstrate that, if there exist a
177: range of noise values such that $t_{inter}$ is shorter than
178: $t_{asym}$, then it is possible to observe stochastic amplification and,
179: simultaneously, SR-gains larger than unity. This is strictly
180: forbidden by linear response theory, as we have recently
181: shown \cite{casgom03}. Thus, a simultaneous appearance of stochastic
182: amplification and SR-gains above $1$ implies a strong violation of linear
183: response theory.
184:
185:
186: \section{Model system and SR quantifiers}
187: Let us consider a system characterized by a single degree of freedom,
188: $x$, subject to the action of a zero average Gaussian white noise with
189: $\langle \xi(t)\xi(s)\rangle = 2D\delta(t-s)$ and driven by an
190: external periodic signal $F(t)$ with period $T$. In the Langevin
191: description, its dynamics is generated by the equation
192: \begin{equation}
193: \label{langev} \dot{x}(t)=-U'\left[ x(t) \right]+F(t)+\xi(t).
194: \end{equation}
195: The corresponding linear Fokker-Planck equation (FPE) for the
196: probability density $P(x,t)$ reads
197: \begin{equation}
198: \label{FP}
199: \frac{\partial}{\partial t}P(x,t)={\hat {\mathcal L}}(t) P(x,t),
200: \end{equation}
201: where
202: \begin{equation}
203: {\hat {\mathcal L}}(t)=\frac{\partial}{\partial x}\left[ U'(x)-F(t)+ D
204: \frac{\partial}{\partial x}\right].
205: \end{equation}
206: In the expressions above, $U'(x)$ represents the derivative of the
207: potential $U(x)$. In this work, we will consider a bistable potential
208: $U(x)=-x^2/2+x^4/4$. The periodicity of the external driving $F(t)$
209: allows its Fourier series expansion in the harmonics of the
210: fundamental frequency $\Omega=2\pi/T$, i.e.,
211: \begin{equation}
212: \label{fourf}
213: F(t)=\sum_{n=1}^\infty \left [ f_n \cos (n \Omega t) + g_n \sin
214: (n \Omega t) \right ],
215: \end{equation}
216: with the Fourier coefficients, $f_n$ and $g_n$, given by
217: \begin{eqnarray}
218: f_n&=&\frac 2T \, \int_0^T \ud t\, F(t) \cos (n \Omega t),
219: \nonumber \\
220: g_n&=&\frac 2T \, \int_0^T \ud t\, F(t) \sin (n \Omega t).
221: \end{eqnarray}
222: Here, we are assuming that the cycle average of the external driving
223: over its period equals zero. In this work, we will focus our attention
224: to multi-frequency input forces with ``rectangular'' shape given by
225: \begin{equation}
226: \label{pulse}
227: F(t)= \left \{ \begin{array}
228: {r@{\quad:\quad}l}
229: A & 0\le t < \frac T2 \\
230: -A& \frac T2 \le t < T
231: \end{array}
232: \right .
233: \end{equation}
234: as sketched in Fig.~\ref{fuerza}. The external force remains constant
235: at a value $A$ during each half-period and changes sign for the second
236: half of the period. The duty cycle of the signal is defined by the time
237: span the signal is nonzero over the total period of the signal; thus,
238: the rectangular signal in Eq.\ (\ref{pulse}) consequently possesses a
239: duty cycle of unity.
240:
241: The two-time correlation function $\langle x(t+\tau)
242: x(t)\rangle_{\infty}$ in the limit $t \rightarrow \infty$ is given
243: by
244: \begin{equation}
245: \langle x(t+\tau) x(t)\rangle_{\infty} =\int_{-\infty}^{\infty} \ud x'
246: \,x' P_{\infty}(x',t) \int_{-\infty}^{\infty}
247: \ud x \,x P_{1|1}(x,t+\tau|x',t),
248: \end{equation}
249: where $P_{\infty}(x,t)$ is the time-periodic, asymptotic long time
250: solution of the FPE and the quantity $P_{1|1}(x,t+\tau|x',t)$
251: denotes the two-time conditional probability density that the
252: stochastic variable will have a value near $x$ at time $t+\tau$ if
253: its value at time $t$ was exactly $x'$. It can been shown
254: \cite{RMP,PRA91} that, in the limit $t \rightarrow \infty$, the
255: two-time correlation function $\langle x(t+\tau)
256: x(t)\rangle_\infty$ becomes a periodic function of $t$ with the
257: period of the external driving. Then, we define the one-time
258: correlation function, $C(\tau)$, as the average of the two-time
259: correlation function over a period of the external driving, i.e.,
260: \begin{equation}
261: \label{ctau}
262: C(\tau)= \frac{1}{T} \int_{0}^{T} \ud t \, \langle x(t+\tau)
263: x(t)\rangle_{\infty}.
264: \end{equation}
265: The correlation function $C(\tau)$ can be written exactly as the
266: sum of two contributions: a coherent part, $C_{coh}(\tau)$, which
267: is periodic in $\tau$ with period $T$, and an incoherent part
268: which decays to $0$ for large values of $\tau$. The coherent part
269: $C_{coh}(\tau)$ is given by \cite{RMP,PRA91}
270: \begin{equation}
271: \label{chtau}
272: C_{coh}(\tau) =\frac{1}{T} \int_{0}^{T} \ud t \,\langle x(t+\tau)
273: \rangle_{\infty} \langle x(t) \rangle_{\infty},
274: \end{equation}
275: where $\langle x(t) \rangle_{\infty}$ is the average value evaluated
276: with the asymptotic form of the probability density, $P_{\infty}(x,t)$.
277:
278: According to McNamara and Wiesenfeld \cite{McNWie89}, the output SNR is
279: defined in terms of the Fourier transform of the coherent and
280: incoherent parts of $C(\tau)$. As the correlation function is even
281: in time and we evaluate its time dependence for $\tau \ge 0$, it
282: is convenient to use its Fourier cosine transform, defined as
283: \begin{equation}
284: \label{fourier}
285: \tilde{C}(\omega)=\frac 2\pi \int_0^\infty \ud\tau\,C(\tau) \cos (\omega
286: \tau);\; C(\tau)=\int_0^\infty \ud\omega\, \tilde{C}(\omega) \cos (\omega
287: \tau).
288: \end{equation}
289: The value of the output SNR is then obtained from:
290: \begin{equation}
291: \label{snr}
292: R_{out} =\frac {\lim_{\epsilon \rightarrow 0^+}
293: \int_{\Omega-\epsilon}^{\Omega+\epsilon} \ud\omega\;
294: \tilde{C}(\omega)}{\tilde{C}_{incoh}(\Omega)}.
295: \end{equation}
296: Note that this definition of the SNR differs by a factor $2$, stemming
297: from the same contribution at $\omega = - \Omega$, from the
298: definitions used in earlier works \cite {RMP,PRA91}. The periodicity of
299: the coherent part gives rise to delta peaks in the spectrum. Thus, the
300: only contribution to the numerator in Eq.\ (\ref{snr}) stems from the
301: coherent part of the correlation function. The evaluation of the SNR
302: requires the knowledge of the Fourier components of $C_{coh}(\tau)$ and
303: $C_{incoh}(\tau)$ at the fundamental frequency of the driving
304: force. Thus, rather than knowledge of the entire Fourier spectrum, only
305: two well defined numerical quadratures are needed. These are:
306: \begin{equation}
307: \label{snr1}
308: R_{out}=\frac{Q_u}{Q_l},
309: \end{equation}
310: where
311: \begin{equation}
312: \label{num}
313: Q_u= {\frac 2T} \int_0^T \ud \tau \,C_{coh}(\tau) \cos (\Omega
314: \tau),
315: \end{equation}
316: and
317: \begin{equation}
318: \label{den}
319: Q_l=\frac 2\pi \int_0^\infty \ud \tau \,C_{incoh}(\tau) \cos (\Omega
320: \tau ).
321: \end{equation}
322: The SNR for an input signal $F(t)+\xi(t)$ is given by
323: \begin{equation}
324: \label{snrinp}
325: R_{inp}=\frac{ \pi(f_1^2+g_1^2)}{4D}.
326: \end{equation}
327: The SR-gain $G$ is consequently defined as the ratio of the SNR of the
328: output over the SNR of the input, namely,
329: \begin{equation}
330: \label{gain}
331: G=\frac {R_{out}}{R_{inp}}.
332: \end{equation}
333: %
334: \section{Numerical solution}
335: Stochastic trajectories, $x^{(j)}(t)$, are generated by numerically
336: integrating the Langevin equation [Eq.\ (\ref{langev})] for every
337: realization $j$ of the white noise $\xi(t)$, starting from a given
338: initial condition $x_0$. The numerical solution is based on the
339: algorithm developed by Greenside and Helfand \cite{hel79,grehel81} (consult
340: also the Appendix in Ref.~\cite{casgom03}). After allowing for a relaxation
341: transient stage, we start recording the time evolution of each random
342: trajectory for many different trajectories. Then, we construct the long
343: time average value,
344: \begin{equation}
345: \label{avgnum}
346: \langle x(t) \rangle_\infty = \frac 1N \sum_{j=1}^N x^{(j)}(t),
347: \end{equation}
348: and the second cumulant,
349: \begin{equation}
350: \label{cumul}
351: \langle x^2(t)\rangle_\infty-\langle x(t)\rangle^2_\infty =\frac 1N
352: \sum_{j=1}^N \left [x^{(j)}(t)\right ]^2 - \left [ \frac 1N \sum_{j=1}^N
353: x^{(j)}(t) \right ] ^2,
354: \end{equation}
355: where $N$ is the number of stochastic trajectories considered.
356: We also evaluate the two-time ($t$ and $\tau$) correlation function,
357: i.e.,
358: \begin{equation}
359: \langle x(t+\tau) x(t) \rangle_\infty = \frac 1N \sum_{j=1}^N
360: x^{(j)}(t+\tau) x^{(j)}(t),
361: \end{equation}
362: as well as the product of the averages
363: \begin{equation}
364: \langle x(t+\tau)\rangle_\infty \langle x(t) \rangle_\infty = \left [\frac 1N
365: \sum_{j=1}^N x^{(j)}(t+\tau) \right ] \left [\frac 1N
366: \sum_{j=1}^N x^{(j)}(t) \right ].
367: \end{equation}
368: The correlation function $C(\tau)$ and its coherent part
369: $C_{coh}(\tau)$ are then obtained using their definitions in
370: Eqs.~(\ref{ctau}) and (\ref{chtau}), performing the cycle average over
371: one period of $t$. The difference between the values of $C(\tau)$ and
372: $C_{coh}(\tau)$ allows us to obtain the values for $C_{incoh}(\tau)$. It
373: is then straightforward to evaluate the Fourier component of
374: $C_{coh}(\tau)$ and the Fourier transform of $C_{incoh}(\tau)$ at the
375: driving frequency by numerical quadrature. With that information, the
376: numerator and the denominator for the output SNR [cf. Eqs.\
377: (\ref{snr1}), (\ref{num}) and (\ref{den})], as well as the SR-gain
378: [cf. Eq.\ (\ref{gain})], are obtained.
379: %
380: \section{Results}
381: \subsection{Response to a rectangular driving force with fundamental
382: frequency $\Omega=0.01$}
383: Consider an external driving of the type sketched in Fig.~\ref{fuerza}
384: with parameter values $\Omega=0.01$,
385: $A=0.25$. This amplitude is well below its threshold value defined,
386: for each driving frequency, as the minimum amplitude that can induce
387: repeated transitions between the minima of $U(x)$ in the absence of
388: noise. For the input considered here, the threshold amplitude is $A_T
389: \approx 0.37$. Note that this threshold value for the amplitude increases
390: with increasing driving frequency.
391:
392: In Fig.~\ref{wp1} we depict with several panels the behavior of the
393: first two cumulants, $ \langle x(t) \rangle_\infty$ and $\langle
394: x^2(t)\rangle_\infty -\langle x(t)\rangle^2_\infty$, for several
395: representative values of $D$ [from top to bottom $D=0.02$ (panel a),
396: $D=0.04$ (panel b), $D=0.06$ (panel c), $D=0.1$ (panel d) and $D=0.2$
397: (panel e)]. Notice that due to the transients, the time at which we
398: start recording data, $t=0$ in the graphs, does not necessarily coincide
399: with the start of an external cycle. The average is periodic with the
400: period of the driving force, while the second cumulant, due to the
401: reflection symmetry of the potential \cite{EPLcasado}, is periodic with
402: a period half of the period of the forcing term.
403:
404: Next we consider the case of small noise intensity $D$ (say, $D=0.02$ as
405: in panel a). The noise induces jumps between the wells. In each random
406: trajectory, a jump between the wells has a very short duration, but the
407: instants of time at which they take place for the different stochastic
408: trajectories are randomly distributed during a half-cycle. At this small
409: noise strength, the jumps are basically towards the lowest
410: minimum. Thus, because of this statistical effect, the average behavior
411: depicts the smooth evolution depicted in panel a of Fig.~\ref{wp1},
412: without sudden transitions between the wells. The evolution of the
413: second cumulant adds relevant information. The fact that it is rather
414: large during most of a period indicates that the probability density
415: $P_\infty(x,t)$ is basically bimodal during most of the external
416: cycle. It is only during very short time intervals around each
417: half-period that the probability distribution becomes monomodal around
418: one of the minima, and, consequently, the second cumulant is small. The
419: bimodal character arises from the fact that the noise is so small in
420: comparison with the barrier heights that jumps over the barrier are
421: rather infrequent during each half a period.
422:
423: As the noise strength increases, the time evolution of $ \langle
424: x(t)\rangle_\infty$ follows closely the shape of the external force,
425: (cf. see panels b, c, d in Fig.~\ref{wp1} for $0.04 \le D \le 0.1$). This
426: behavior indicates that, for these parameter values, the jumps in the
427: different random trajectories are concentrated within short time
428: intervals around the instants of time at which the driving force
429: switches sign. The second cumulant remains very small during most of a
430: period, except for short time intervals around the switching times of
431: the external driver. Thus, for these intermediate values of $D$, the
432: probability distribution, $P_\infty(x,t)$ is basically monomodal, except
433: for small time intervals around the switching instants of time of the
434: periodic driver. Finally, as the noise strength is further increased ($D
435: > 0.1$), the probability distribution remains very broad most of the
436: time. Even though a large majority of random trajectories will still
437: jump over the barrier in synchrony with the switching times of the
438: driver, the noise is so large in comparison with the barrier heights,
439: that the probability of crossings over the barrier in both directions
440: can not be neglected at any time during each half cycle. The probability
441: distribution remains bimodal during a whole period, but asymmetric: the
442: larger fraction of the probability accumulates around the corresponding
443: lower minima of the potential during each cycle. Therefore, the average
444: output amplitude decreases, while the second cumulant depicts plateaus at
445: higher values than for smaller noise strengths (compare panels e and c
446: in Fig.~\ref{wp1}).
447:
448: In Fig.~\ref{wp2}, we plot the coherent (left panels), $C_{coh}(\tau)$,
449: and incoherent (right panels), $C_{incoh}(\tau)$, components of the
450: correlation function $C(\tau)$ for the same parameter values as in
451: Fig.~\ref{wp1}. The coherent part shows oscillations with a period equal
452: to that of the driving force. Its shape changes with $D$. The amplitude
453: of the coherent part does not grow monotonically with $D$. Rather, it
454: maximizes at $D\approx 0.06$, which is consistent with the observed
455: behavior of $ \langle x(t)\rangle_\infty$ in Fig.~\ref{wp1}. This is
456: expected as the evaluation of $C_{coh}(\tau)$ involves only the time
457: behavior of $ \langle x(t)\rangle_\infty$ at two different instants of
458: time.
459:
460: Two features of the behavior of $C_{incoh}(\tau)$ are relevant: its
461: initial value and its decay time. The initial value of the incoherent
462: contribution, $C_{incoh}(0)$, is given by the cycle average of the
463: second cumulant. It has a non-monotonic behavior with $D$. For $D=0.02$,
464: $C_{incoh}(0)$ is large, consistent with the fact that the second
465: cumulant at this noise strength is appreciably different from $0$ during
466: a substantial part of a period. As the value of $D$ increases, ($D
467: \le 0.1$), the value of $C_{incoh}(0)$ decreases. This is expected as
468: the second cumulant is large just during those small time intervals
469: where most of the forward transitions take place every half-period. For
470: still larger values of $D$ there are frequent forward and backward jumps
471: that keep the stationary probability bimodal, and therefore, the initial
472: value $C_{incoh}(0)$ increases. For $D=0.02$, the decay of
473: $C_{incoh}(\tau)$ is very slow, although the decay time is still shorter
474: than the duration of half a period of the driving force. As $D$
475: increases, the decay time of the incoherent part becomes shorter. It is
476: worth to point out that the intra-well noisy dynamics manifests itself
477: in the behavior of $C_{incoh}(\tau)$. This is most clearly confirmed
478: noticing the fast initial decay observed in panel d. For smaller values of $D$,
479: this feature is masked by the long total relaxation time scale, while
480: for $D=0.2$, the noise strength is so large that there is not a
481: clear-cut separation between inter and intra-well time scales.
482:
483: The above considerations allow us to rationalize the behavior of the
484: several quantifiers used to characterize SR. Their behaviors with $D$
485: for $A=0.25$, $\Omega=0.01$ are depicted in Fig.~\ref{wp3}. It should
486: be noticed that the lowest value of the noise strength used in the
487: numerical solution of the Langevin equation is $D=0.01$. For this noise
488: strength, the values of $Q_u$ and $R_{out}$ are very small, although not
489: zero. For even lower noise strengths the task becomes computationally
490: very demanding and expensive, due to the extremely slow decay of the
491: correlations. For $D$ sufficiently small, however, one does expect $Q_u$
492: to be larger than $Q_l$, and, consequently, an increase of the numerical
493: $R_{out}$ as $D$ is lowered.
494:
495: The quantity $Q_u$ defined in Eq.\ (\ref{num}) depicts a non-monotonic
496: dependence on $D$ typical of the SR phenomenon. Its behavior is
497: expected from the dependence of the amplitude of $C_{coh}(\tau)$ with
498: $D$ in Fig.~\ref{wp2}.
499:
500: A non-monotonic behavior with $D$ for the numerically evaluated $Q_l$ is also
501: observed. The initial value, $C_{incoh}(0)$ and the decay time of
502: $C_{incoh}(\tau)$ are important in the evaluation of $Q_l$ [see
503: Eq. (\ref{den})]. For $D=0.01$, the decay time of the incoherent part of
504: the correlation function is longer than half a period of the driving
505: force, while for $D=0.02$, it is somewhat shorter than
506: $T/2$. Consistently with Eq.~(\ref{den}), the value of the integral for
507: $D=0.01$ is smaller than for $D=0.02$. As $D$ is further increased, the
508: influence of the cosine factor in Eq.~(\ref{den}) becomes less important
509: as the decay time is much shorter than $T/2$. The drastic fall in the
510: $Q_l$ values observed for $0.02 < D < 0.1$ is due to the decrease of
511: $C_{incoh}(0)$ with $D$ (see panels a-c in Fig.~\ref{wp2}) and the
512: shortening of the decay time. As $D$ is increased further,
513: $C_{incoh}(0)$ increases and, consequently, $Q_l$ also increases
514: slightly.
515:
516: Taking into account the definition of the SNR, [cf. Eq.~(\ref{snr1})],
517: its behavior with $D$ is not surprising. The numerically obtained SNR
518: peaks at $D=0.08$, a slightly different value of $D$ from the
519: one at which $Q_u$ peaks.
520:
521: \subsection{Anomalous SR-gain behavior for subthreshold driving}
522: The SR-gain is defined in Eq.\ (\ref{gain}). The numerically determined SR-gain
523: shows a most interesting feature: we observe a non-monotonic behavior
524: versus $D$, with values for the gain exceeding unity (!) for a whole
525: range of noise strengths. This is strictly forbidden within LRT
526: \cite{casgom03}; therefore, the fact that the SR-gain can assume values
527: larger than unity reflects a manifestation of the inadequacy of LRT to
528: describe the system dynamics for the parameter values considered.
529:
530: To rationalize this anomalous SR-gain behavior, we notice that the role of the
531: noise in the dynamics is twofold. On the one hand, it controls the decay
532: time of $C_{incoh}(\tau)$. On the other hand, the noise value is
533: relevant to ellucidate whether the one-time probability distribution is
534: basically monomodal or bimodal during most of the cycle and,
535: consequently, it controls the initial value $C_{incoh}(0)$. As
536: discussed above, if $D$ is small, the decay time is very large
537: compared to $t_{asym}$, and the one-time probability distribution is
538: essentially bimodal. For large values of $D$, the decay of
539: $C_{incoh}(\tau)$ is fast enough, and the distribution is also bimodal.
540: The large SR-gain obtained here requires the existence of a range of
541: intermediate noise values such that: i) $C_{incoh}(\tau)$ decays in a
542: much shorter time scale than $t_{asym}$ and, ii) the one-time probability
543: distribution remains monomodal during most of the external cycle.
544: \subsection{Response to a rectangular input driver with fundamental
545: frequency $\Omega=0.1$}
546:
547: As mentioned before, there are several time scales which are important
548: for the phenomenon of stochastic amplification and gain. In the previous
549: subsection we have considered an input frequency small enough so that
550: the inequality $t_{asym} > t_{inter}$ holds for a range of noise values.
551: Next, we shall analyze the system response to a driving force with
552: fundamental frequency $\Omega=0.1$, ten times larger than in the
553: previous case. We will take the same input amplitude as in the previous
554: case, $A=0.25$, which is still subthreshold. For this input frequency,
555: the threshold value for the amplitude strength is determined by
556: numerically solving the deterministic equation to yield $ A_T \approx
557: 0.39$.
558:
559: The behavior of the first two cumulants for several values of the noise
560: strength is depicted in Fig.~\ref{wp11} [from top to bottom $D=0.02$
561: (panel a), $D=0.04$ (panel b), $D=0.06$ (panel c), $D=0.1$ (panel d) and
562: $D=0.2$ (panel e)]. For all values of $D$, the second cumulant remains
563: large for most of each half-period. By contrast with the
564: lower frequency case, we detect no values of $D$ for which the
565: probability distribution is monomodal for a significant fraction of each
566: half period.
567:
568: In Fig.~\ref{wp22} the behavior of the coherent (left panels) and
569: incoherent (right panels) parts of the correlation function are
570: presented for (from top to bottom) $D=0.02$ (panel a), $D=0.04$ (panel
571: b), $D=0.06$ (panel c), $D=0.1$ (panel d) and $D=0.2$ (panel e). The
572: amplitude of the coherent oscillations shows a nonmonotonic behavior
573: with $D$. The incoherent part has initial values which remain very large
574: in comparison with the corresponding ones for $\Omega=0.01$ (compare
575: with Fig.~\ref{wp2}), consistently with the large value of the second
576: cumulant. The decay times are roughly the same for both frequencies.
577:
578: In Fig.~\ref{wp33} we show the behavior of the several SR quantifiers as
579: a function of $D$. The comparison of Figs.~\ref{wp3} and \ref{wp33}
580: indicates that $Q_u$ , $Q_l$ and $R_{out}$ have the same qualitative
581: behavior for both frequencies. The non-monotonic dependence on $D$ of
582: $Q_u$ and $R_{out}$ are indicative of the existence of SR (for both
583: frequencies) for the subthreshold input amplitude and in the ranges of
584: $D$ values considered. The most relevant quantitative difference is that
585: for $\Omega=0.1$ the SR-gain remains less than unity.
586:
587: \section{Conclusions}
588: With this work, we have analyzed the phenomenon of SR within the context
589: of a noisy, bistable symmetric system driven by time periodic,
590: rectangular forcing possessing a duty cycle of unity. The numerical
591: solution of the Langevin equation allows us to analyze the long-time
592: behavior of the average, the second cumulant and the coherent and
593: incoherent parts of the correlation function. For subthreshold input
594: signals we determined the SNR, together with its numerator and
595: denominator evaluated separately, for a wide range of noise strengths
596: $D$.
597:
598: As a main result we find the simultaneous existence of a typical
599: non-monotonic behavior versus the noise strength $D$ of several
600: quantifiers associated to SR; in particular SR-gains larger than unity
601: are possible for a subthreshold rectangular forcing possessing a duty
602: cycle of unity. This finding is at variance with the recent claim in
603: Refs.~\cite{ginmak01,makging02} that pulse-like signals with small duty
604: cycles are needed to obtain SR-gains larger than unity. This unexpected
605: result occurs indeed whenever the inequality $t_{asym} > t_{inter}$
606: holds. This is most easily achieved with low frequency inputs. As the
607: input frequency increases, that inequality is not satisfied for
608: sufficiently small values of $D$: even though SR then still exists, it
609: is not accompanied by SR-gains exceeding unity. Furthermore, in
610: Refs.~\cite{ginmak01,makging02}, SR-gains larger than unity are only
611: obtained with input amplitudes larger than $0.8\,A_T$. By contrast, in
612: this work, we have shown that such a large value for the input amplitude
613: is not needed (we have used $A\approx 0.68\,A_T$).
614:
615: The simultaneous occurrance of SR and SR-gains larger than unity is
616: associated to the fact that, for some range of noise values, the decay
617: time of the incoherent part of the correlation function is much shorter
618: than $t_{asym}$ and also the probability distribution is basically monomodal
619: during most of the cycle of the driving force.
620:
621:
622: \begin{acknowledgments}
623: We acknowledge the support of the Direcci\'on General de
624: Ense\~nanza Superior of Spain (BFM2002-03822), the Junta de
625: Andaluc\'{\i}a, the DAAD program "Acciones Integradas" (P.H., M.
626: M.) and the Sonderforschungsbereich 486 (project A10) of the Deutsche
627: Forschungsgemeinschaft.
628: \end{acknowledgments}
629:
630: \begin{thebibliography}{100}
631:
632: \bibitem{PT96}
633: A. R. Bulsara and L. Gammaitoni, Physics Today {\bf 49}, No. 3, 39
634: (1996).
635: \bibitem{RMP}
636: L. Gammaitoni, P. H\"anggi, P. Jung, and F. Marchesoni, Rev. Mod.
637: Phys. {\bf 70}, 223 (1998).
638: \bibitem{Wiesenfeld98}
639: K. Wiesenfeld and F. Jaramillo, Chaos {\bf 8}, 539 (1998).
640: \bibitem{Anish99}
641: V. S. Anishchenko, A. B. Neiman, F. Moss, and L. Schimansky-Geier,
642: Usp. Fiz. Nauk {\bf 169}, 7 (1999).
643: \bibitem{Chemphyschem02}
644: P. H\"anggi, CHEMPHYSCHEM {\bf 3}, 285 (2002).
645: \bibitem{McNWie89} B. McNamara and K. Wiesenfeld, Phys. Rev. A {\bf
646: 39}, 4854 (1989).
647: \bibitem{Hanggi78}
648: P. H\"anggi, Helv. Phys. Acta, {\bf 51}, 202 (1978).
649: \bibitem{PR82} P. H\"anggi, and H. Thomas, Phys. Rep. {\bf 88}, 207 (1982).
650: \bibitem{EPL89} P. Jung and P. H\"anggi, Europhys. Lett. {\bf 8}, 505 (1989).
651: \bibitem {GM89} L. Gammaitoni, E. Menichella-Saetta, S. Santucci,
652: F. Marchesoni, and C. Presilla, Phys. Rev. A {\bf 40}, 2114 (1989).
653: \bibitem{PRA91}
654: P. Jung and P. H\"anggi, Phys. Rev. A {\bf 44}, 8032 (1991).
655: \bibitem{EPLcasado}
656: J. Casado-Pascual, J. G\'omez-Ord\'o\~nez, M. Morillo, and P. H\"anggi,
657: Europhys. Lett. {\bf 58}, 342 (2002).
658: \bibitem{FNLcasado}
659: J. Casado-Pascual, J. G\'omez-Ord\'o\~nez, M. Morillo, and P. H\"anggi,
660: Fluct. Noise Lett. {\bf 2}, L127 (2002).
661: \bibitem{dykman92} M. I. Dykman,
662: R. Mannella, P. V. E. McClintock, and N. G. Stocks,
663: Phys. Rev. Lett. {\bf 68}, 2985 (1992).
664: \bibitem{junhan93} P. Jung and P. H\"anggi, Z. Physik B {\bf
665: 90}, 255 (1993).
666: \bibitem{dewbia95} M. DeWeese and W. Bialek, Il Nuovo Cimento
667: {\bf17D}, 733 (1995).
668: \bibitem{casgom03} Jes\'us Casado-Pascual, Claus Denk, Jos\'e
669: G\'omez-Ord\'o\~nez, Manuel Morillo, and Peter H\"anggi, Phys. Rev. E
670: {\bf 67}, 036109 (2003).
671: \bibitem{haninc00} P. H\"anggi, M. Inchiosa, D. Fogliatti and
672: A. R. Bulsara, Phys. Rev. E {\bf 62}, 6155 (2000).
673: \bibitem{ginmak00} Z. Gingl, R. Vajtai, and P. Makra, in ``Noise
674: in Physical Systems and 1/f Fluctuations '', ICNF 2001, G.
675: Bosman, editor (World Scientific, 2002), pp. 545-548.
676: \bibitem{ginmak01} Z. Gingl, P. Makra, and R. Vajtai, Fluct.
677: Noise Lett. {\bf 1}, L181, (2001).
678: \bibitem{Morillo95}
679: M. Morillo and J. G\'omez-Ord\'o\~nez, Phys. Rev. E {\bf 51}, 999
680: (1995).
681: \bibitem{hel79} E. Helfand, Bell Sci. Tech. J., {\bf 58}, 2289 (1979).
682: \bibitem{grehel81} H. S. Greenside and E. Helfand, Bell Sci. Tech. J.,
683: {\bf 60}, 1927 (1981).
684: \bibitem{makging02} P. Makra, Z. Gingl and L. B. Kish, Fluct.
685: Noise Lett. {\bf 1}, L147, (2002).
686: \end{thebibliography}
687:
688: \newpage
689: \begin{figure}
690: \includegraphics[width=10cm]{Fig1.ps}
691: \caption{\label{fuerza}Sketch of a rectangular periodic signal with
692: duty cycle 1, amplitude $A$ and period $T$}
693: \end{figure}
694: %
695: \begin{figure}
696: \includegraphics[width=10cm]{Fig2.ps}
697: \caption{\label{wp1} Time behavior of the average $\langle x(t)\rangle
698: _\infty$ (solid lines) and the second cumulant $\langle x^2(t)
699: \rangle_\infty - \langle x(t)\rangle _\infty^2$ (dashed lines) for a rectangular
700: driving force with duty cycle 1, fundamental frequency $\Omega=0.01$ and
701: subthreshold amplitude $A=0.25$ for several values of the noise
702: strength: $D=0.02$ (panel a), $D=0.04$ (panel b), $D=0.06$ (panel c),
703: $D=0.1$ (panel d), $D=0.2$ (panel e). Notice that
704: due to the transients, $t=0$ in the graphs, does not necessarily
705: coincide with the start of an external cycle.}
706: \end{figure}
707: %
708: \begin{figure}
709: \includegraphics[width=10cm]{Fig3.ps}
710: \caption{\label{wp2} Time behavior of $C_{coh}(\tau)$ (left panels) and
711: $C_{incoh}(\tau)$ (right panels) for a rectangular driving force with duty
712: cycle 1, fundamental frequency $\Omega=0.01$ and subthreshold
713: amplitude $A=0.25$ for several values of the noise strength: $D=0.02$
714: (panel a), $D=0.04$ (panel b), $D=0.06$ (panel c), $D=0.1$ (panel d),
715: $D=0.2$ (panel e).}
716: \end{figure}
717: %
718: \begin{figure}
719: \includegraphics[width=10cm]{Fig4.ps}
720: \caption{\label{wp3} Dependence with $D$ of several SR quantifiers: the
721: numerator of the SNR ($Q_u$), its denominator ($Q_l$), the output SNR
722: ($R_{out}$) and the SR-gain ($G$) for a rectangular driving force with
723: duty cycle 1,
724: fundamental frequency $\Omega=0.01$ and subthreshold amplitude
725: $A=0.25$.}
726: \end{figure}
727: %
728: \begin{figure}
729: \includegraphics[width=10cm]{Fig5.ps}
730: \caption{\label{wp11} Same as in Fig.~\ref{wp1} but for $\Omega=0.1$}
731: \end{figure}
732: %
733: \begin{figure}
734: \includegraphics[width=10cm]{Fig6.ps}
735: \caption{\label{wp22} Same as in Fig.~\ref{wp2} but for $\Omega=0.1$.}
736: \end{figure}
737: %
738: \begin{figure}
739: \includegraphics[width=10cm]{Fig7.ps}
740: \caption{\label{wp33} Same as in Fig.~\ref{wp3} but for $\Omega=0.1$.}
741: \end{figure}
742: \end{document}
743:
744: