1: \documentstyle[prb,epsfig,aps,preprint]{revtex}
2: \input{epsf}
3:
4: \begin{document}
5: \draft
6: \title{Observation of progressive motion of ac-driven solitons}
7: \author{Alexey V. Ustinov$^{a}$ and Boris A. Malomed$^{b}$}
8: \address{$^{a}$Physikalisches Institut III, Universit\"at\\
9: Erlangen-N\"urnberg, D-91058, Erlangen, Germany\\
10: $^{b}$Department of Interdisciplinary Studies, Faculty of\\
11: Engineering, Tel Aviv University, Tel Aviv, Israel}
12: \date{\today}
13: \maketitle
14:
15: \begin{abstract}
16: We report the first experimental observation of ac-driven phase-locked
17: motion of a topological soliton at a nonzero average velocity in a
18: periodically modulated lossy medium. The velocity is related by a resonant
19: condition to the driving frequency. The observation is made in terms of the
20: current-voltage, $I(V)$, characteristics for a fluxon trapped in an annular
21: Josephson junction placed into dc magnetic field. Large zero-crossing
22: constant-voltage steps, exactly corresponding to the resonantly locked
23: soliton motion at different orders of the resonance, are found on the
24: experimental $I(V)$ curves. A measured dependence of the size of the steps
25: vs. the external magnetic field is in good agreement with predictions of an
26: analytical model based on the balance equation for the fluxon's energy. The
27: effect has a potential application as a low-frequency voltage standard.
28: \end{abstract}
29:
30: \pacs{05.45.-a, 74.50.+r, 85.25.-j}
31:
32: %\wideabs{
33:
34: %}
35:
36: %end of WideAbs
37: %\newpage
38:
39: The important role played by nonlinear excitations in the form of solitons
40: in various physical systems is commonly known. However, experimental
41: observation of dynamical effects produced by solitons is often difficult
42: because real systems may be far from their idealized mathematical models
43: which give rise to soliton solutions. Among perturbations that destroy
44: soliton effects, dissipation is the most important one. To compensate
45: dissipative losses and thus make the soliton dynamics visible, one must
46: apply an external force that supports (in particular, {\em drives}) a
47: soliton \cite{oldreview}.
48:
49: Solitons of the simplest type are topological {\it kinks}, a well-known
50: example being a magnetic flux quantum (fluxon) in long Josephson junctions
51: (LJJs) \cite{Barone,Ust-rev98}. A fluxon in LJJ can be easily driven by bias
52: current applied to the junction. The motion of a fluxon gives rise to dc
53: voltage $V$ across the junction, which is proportional to the fluxon's mean
54: velocity. Varying the dc bias current $I$, one can produce a dependence $%
55: V(I) $, which is the main dynamical characteristic of LJJ. An experimentally
56: obtained $I(V)$ curve easily allows one to identify the presence of one or
57: more fluxons trapped in LJJ \cite{Barone}.
58:
59: Microwave field irradiating LJJ gives rise to an ac drive acting on the
60: fluxon. In a spatially homogeneous lossy system, an ac drive may only
61: support an oscillatory motion of a kink, which is hard to observe in LJJ due
62: to the absence of dc voltage. However, it was predicted \cite{Bob1} that an
63: ac drive can support motion of a kink with a {\em nonzero} average velocity $%
64: u$ in a system with a periodic spatial modulation. Indeed, a moving kink
65: passes the modulation length (period) $L$ during the time $L/u$. If this
66: time is commensurate with the period $2\pi /\omega $ of the ac drive, i.e., $%
67: m(L/u)=2\pi /\omega $ with an integer $m$, or
68: \begin{equation}
69: u=m(L\omega /2\pi ), \label{spectrum}
70: \end{equation}
71: one may expect a resonance (of order $m$) between the two periodicities. In
72: other words, a moving kink may be phase-locked to the ac drive. This
73: provides for permanent transfer of energy from the drive to the kink, making
74: it possible to compensate dissipative losses. The energy balance gives rise
75: to a minimum ({\it threshold}) value $\Gamma _{{\rm thr}}$ of the ac drive's
76: amplitude $\Gamma $, which can compensate the losses and support the motion
77: of the ac-driven kink.
78:
79: A more general case, when the system is simultaneously driven by the ac
80: field and dc bias current $I$, was considered in Ref.~\onlinecite{Bob2}. It was
81: predicted that the corresponding $V(I)$ characteristic has {\it steps}
82: (constant-voltage segments) at the resonant velocities (\ref{spectrum}). The
83: motion of the fluxon under the action of pure ac drive then corresponds to
84: {\it zero-crossings}, when the steps cross the axis $I=0$. In fact, the most
85: straightforward way to observe the ac-driven motion of the soliton is
86: through zero-crossings on the $V(I)$ characteristic.
87:
88: A formally similar feature is known in{\em \ small} Josephson junctions as
89: Shapiro steps \cite{Barone}: ac drive applied to the junction gives rise to
90: dc voltage across it (an inverse ac Josephson effect). However, a drastic
91: difference of the effect sought for in this work from the Shapiro steps is
92: that they are only possible at high frequencies exceeding the junctions's
93: plasma frequency, while the ac-driven motion of the fluxon can be supported
94: by the ac drive with an arbitrarily low frequency. This circumstance also
95: opens way for application to the design of voltage standards using easily
96: accessible sources of low-frequency radiation, which are not usable with the
97: usual voltage standards based on small Josephson junctions. Moreover,
98: because the single-valued dc voltage in LJJ is controlled not only by
99: the frequency and amplitude of the ac drive, but also by locally applied magnetic
100: field (see below), the system studied here may be useful for designing an ac voltage
101: standard.
102:
103: The inverse Josephson effect was also studied in detail in ac-driven
104: finite-length linear junctions with reflecting edges, corresponding to zero
105: external magnetic field \cite{Salerno}. A fluxon turns into an antifluxon
106: while bouncing from the edge. In that case, the system of constant-voltage
107: steps is complicated, on the contrary to a very simple set of two symmetric
108: steps in the circular junction, see below. This is explained by the fact
109: that the {\em shuttle motion} of the polarity-reversing fluxon/antifluxon in
110: the linear junction is different from the progressive motion of the
111: fluxon, without any polarity reversal, in the circular LJJ.
112:
113: An objective of this paper is to report direct experimental observation of
114: the ac-driven fluxon motion in periodically modulated LJJs. Frequently, it
115: is assumed that the necessary periodic spatial modulation along the junction
116: can be induced by periodically changing the thickness of the dielectric
117: barrier separating two superconductors. In the presence of the losses and
118: drive, the modulated LJJ is described by the perturbed sine-Gordon (sG)
119: equation,
120: \begin{equation}
121: \phi _{tt}-\phi _{xx}+\left( 1+\varepsilon \sin \frac{2\pi x}{L}\right) \sin
122: \phi =-\alpha \phi _{t}-\gamma -\Gamma \sin \omega t, \label{sG1}
123: \end{equation}
124: where $x$ and $t$ are the length along the junction and time, measured,
125: respectively, in units of the Josephson length and inverse plasma frequency,
126: $\varepsilon $ is the normalized modulation amplitude, while $\gamma $ and $%
127: \Gamma $ are the dc and ac bias current densities, both normalized to the
128: junction's critical current density.
129:
130: The harmonic modulation of the local magnitude of the maximum Josephson
131: current, assumed by this model, is very hard to realize in an experiment due
132: to the exponential dependence of the critical current on the thickness of
133: the dielectric barrier. A much easier and fully controllable way to induce a
134: harmonic periodic modulation is to use an {\em annular} (ring- shaped) LJJ,
135: to which uniform dc magnetic field is applied in its plane \cite{magnetic}.
136: As it was demonstrated experimentally, fluxons can be readily trapped in
137: annular LJJs \cite{vernik}. In this case, the sG model takes the form
138:
139: \begin{equation}
140: \phi _{tt}-\phi _{xx}+\sin \phi +h\sin \frac{x}{R}=-\alpha \phi _{t}-\gamma
141: -\Gamma \sin \omega t, \label{sG2}
142: \end{equation}
143: where $h$ is (renormalized) magnetic field, and $R$ is the radius of the
144: ring. In the case of the annular junction, solutions are subject to periodic
145: boundary conditions, $\phi _{x}(x+2\pi R)\equiv \phi _{x}(x)$ and $\phi
146: (x+2\pi R)\equiv \phi (x)+2\pi N$, $N$ being the number of the trapped
147: fluxons (in this work, $N=1$). Comparison with the experiment shows that,
148: unlike the model (\ref{sG1}), the one (\ref{sG2}) is, virtually, exact \cite
149: {UstMal}.
150:
151: We assume the spatial size of the fluxon, which is $\sim 1$ in the present
152: notation, to be much smaller than the circumference $L\equiv 2\pi R$ of the
153: annular junction. Large $L$ imposes an upper limit on the driving frequency $%
154: \omega $ which can support the ac-driven motion: as the fluxon's velocity
155: cannot exceed the maximum (Swihart) group velocity of electromagnetic waves
156: in LJJ, which is $1$ in our notation, Eq. (\ref{spectrum}) implies that $%
157: \omega \,\,_{\sim }^{<}\,\,1/L$.
158:
159: A different type of ac drive for fluxons in circular LJJs was proposed in
160: Ref.~\onlinecite {magnetic}, viz., ac magnetic field. In this case, the
161: terms $h\sin \left( x/R\right) $ and $\Gamma \sin (\omega t)$ in Eq. (\ref
162: {sG2}) are replaced by a single one, $h\sin \left( x/R\right) \sin (\omega
163: t) $, which may be decomposed into two waves traveling in opposite
164: directions, $(1/2)[\cos (x/R-\omega t)-\cos (x/R+\omega t)]$. As it was
165: shown \cite{magnetic}, either traveling wave may capture a fluxon, dragging
166: it at the wave's phase velocity $\pm \omega R$. A similar model was proposed
167: in Ref. \cite{Denmark}, in which the fluxon is dragged by {\em rotating}
168: magnetic field. A difference of our model (which corresponds to the real
169: experiment reported below) is the separation between the fields that induce
170: the spatial modulation and ac force. The separation makes it possible to
171: control the dynamics in a more flexible way.
172:
173: It is straightforward to derive an equation of motion for the fluxon in the
174: adiabatic approximation, following the lines of Refs.~\onlinecite{Bob1},%
175: \onlinecite{Bob2} ($\stackrel{.}{\xi }\equiv d\xi /dt$):
176: \begin{eqnarray}
177: \frac{d}{dt}\left( \frac{\stackrel{.}{\xi }}{\sqrt{1-\stackrel{.}{\xi}^{2}}}
178: \right) =\frac{\pi h}{4\sqrt{1-\stackrel{.}{\xi }^{2}}}\cos \frac{\xi }{R} -%
179: \frac{\alpha \stackrel{.}{\xi }}{\sqrt{1-\stackrel{.}{\xi}^{2}}} \nonumber
180: \\
181: +\frac{\pi }{4}\left[ \gamma +\Gamma \sin (\omega t)\right] \,.
182: \label{eqnmotion}
183: \end{eqnarray}
184:
185: For further analysis, one may assume, following Refs.~\onlinecite{magnetic}
186: and \onlinecite{Denmark}, that, in the lowest approximation, the fluxon is
187: moving at a constant velocity $\stackrel{.}{\xi _{0}}\equiv u$ belonging to
188: the resonant spectrum (\ref{spectrum}), so that $\xi (t)=ut+R\delta $, where
189: $\delta $ is a phase-locking constant. Then, the first correction to the
190: instantaneous fluxon's velocity, generated by the spatial modulation, can be
191: easily found from Eq. (\ref{eqnmotion}),
192:
193: \begin{equation}
194: \stackrel{.}{\xi }_{1}=\left( \pi Rh/4u\right) \left( 1-u^{2}\right) \sin %
195: \left[ (u/R)t+\delta \right] \,. \label{correction}
196: \end{equation}
197: The approximation applies provided that the correction (\ref{correction}) is
198: much smaller than the unperturbed velocity $u$, which amounts to $Rh\ll
199: u^{2}/\left( 1-u^{2}\right) $.
200:
201: A key ingredient of the dynamical analysis is the {\it energy-balance}
202: equation \cite{magnetic}. In the model (3) it is based on the correction (%
203: \ref{correction}) to the velocity \cite{Bob1,Bob2} (while in the
204: above-mentioned models with ac magnetic fields \cite{magnetic,Denmark} the
205: approximation $\stackrel{.}{\xi }=u$ was sufficient). In the case of the
206: fundamental resonance, with $m=1$ in Eq. (\ref{spectrum}), i.e., $u=R\omega $%
207: , the energy balance yields, after a straightforward algebra,
208: \begin{equation}
209: \gamma =\frac{4\alpha R\omega }{\pi \sqrt{1-(R\omega )^{2}}}\,-\frac{\pi
210: h\Gamma }{8R\omega ^{2}}\left[ 1-(R\omega )^{2}\right] \cos \delta .
211: \label{balance}
212: \end{equation}
213:
214: Setting $|\cos \delta |=1$ and $\gamma =0$ in Eq. (\ref{balance}) gives a
215: minimum ({\it threshold}) amplitude of the ac drive which can support the
216: fluxon's motion in the absence of the dc bias current, $\Gamma _{{\rm thr}%
217: }=\left( 32/\pi ^{2}h\right) \alpha R^{2}\omega ^{3}\left[ 1-(R\omega )^{2}%
218: \right] ^{-3/2}$. For the comparison with experimental results, the most
219: important consequence of Eq. (\ref{balance}) is an interval of dc bias
220: current density $\gamma $ in which the phase-locked ac-driven motion of the
221: fluxon is expected. It is produced by varying $\cos \delta $ in Eq. (\ref
222: {balance}) between $-1$ and $+1$:
223: \begin{equation}
224: \gamma _{1}^{-}<\gamma <\gamma _{1}^{+};\,\gamma _{1}^{\pm }\equiv \frac{%
225: 4\alpha R\omega }{\pi \sqrt{1-(R\omega )^{2}}}\,\pm \frac{\pi R\Gamma \left[
226: 1-(R\omega )^{2}\right] }{8(R\omega )^{2}}h. \label{interval}
227: \end{equation}
228: Note that the size of the interval strongly depends on the driving
229: frequency, while in the model with the ac magnetic field \cite{magnetic}\ it
230: does not depend on $\omega $ at all, provided that $2\pi R\gg 1$.
231:
232: Experiments have been performed with Nb/Al-AlO$_{x}$/Nb Josephson annular
233: junction with the mean diameter $2R=95\,\mu $m and the annulus width $5\,\mu
234: $m, applying the bias current $I$ and measuring the dc voltage $V$ across
235: the junction. The distribution of the bias current was uniform, which was
236: concluded from measurement of the critical current $I_{{\rm c}}$ in the
237: state without trapped fluxons at $H=0$. $I_{{\rm c}}$ was found to be about $%
238: 0.9$ of its value for the small junction. The annular LJJ had the Josephson
239: length $\lambda _{J}\approx 30\,\mu $m and plasma frequency $50\,$GHz. Note
240: that these parameters imply the ratio $\sim 10$ of the junction's length $%
241: 2\pi R$ to the fluxon's size, which is $\sim \lambda _{J}$, i.e., the
242: junction may indeed be regarded as a long one. The measurements were done at
243: the temperature $4.2\,$K, using a shielded low- noise measurement setup. The
244: ac driving current with the frequency $f=\omega /2\pi $ between $5$ and $26$
245: GHz was supplied by means of a coaxial cable ending with a small antenna
246: inductively coupled to the junction. The antenna was oriented coaxial to the
247: dc bias current supplied through the electrodes, therefore the ac magnetic
248: field was perpendicular to the dc magnetic field. Thus, the magnetic
249: component of the ac signal induced the driving force of the same type as the
250: bias current. The ac power levels mentioned below pertain to the input at
251: the top of the cryostat.
252:
253: Following Ref. \onlinecite{UstMal}, trapping of a fluxon in the junction was
254: achieved by cooling the sample below the critical temperature ${T_{c}}
255: \approx 9.2\,$K for the transition of Nb into the superconductive state,
256: with a small dc bias current applied to the junction. At $H=0$, the fluxon
257: depinning current $I_{{\rm dep}}$ was found to be very small, less than $1$
258: of the Josephson critical current $I_{{\rm c}}$ measured without the trapped
259: fluxon. As a fluxon can only be trapped by junction's local inhomogeneities
260: in the absence of the magnetic field, this indicates at fairly high
261: uniformity of the junction. At low values of the field $H$, linear increase
262: of $I_{{\rm dep}}$ with $H$ was observed, which is well described by the
263: theoretical model based on Eqs.~(\ref{sG2}) and (\ref{eqnmotion}): the
264: zero-voltage state exists as long as the maximum fluxon's pinning force
265: exerted by the field-induced potential remains larger than the driving force
266: induced by dc bias current, which is satisfied at $|\gamma |<$ $h\,$. Note
267: that the fluxon depinning and re-trapping in weak external magnetic fields
268: were studied experimentally and analytically in Ref.~\onlinecite{UstMal}.
269:
270: An evidence for the progressive ac-driven motion of the fluxon is presented
271: in Fig.~\ref{IVC1}a. This $I(V)$ characteristic was measured at $H=0.35\,$Oe
272: and $f=18.1\,$GHz. Its salient feature is two large symmetric
273: constant-voltage steps. The points where they intersect the zero-current
274: axis correspond to the fluxon moving around the junction with a nonzero
275: average velocity at zero dc driving force. Another remarkable feature is the
276: absence of any step at the zero voltage, i.e., in the present case the
277: fluxon cannot be trapped by the effective potential, even when the dc bias
278: current is small. For comparison, in Fig. \ref{IVC1}b we show the $I(V)$
279: curve measured at the same power and frequency of the drive, but with $H=0$.
280: In this case, a substantial zero-voltage step is seen, extending up to the
281: current $I_{0}\approx \pm 0.1$ mA. In the absence of the ac drive, the
282: critical current $I_{0}$ is much smaller, less than $20\,\mu $A (this
283: residual $I_{0}$ may be explained by small inhomogeneities of LJJ, see
284: above).
285:
286: The conspicuous zero-voltage step in Fig. \ref{IVC1}b may be explained by
287: the fact that the magnetic component of the ac drive creates its own
288: modulated potential. This argument also helps to explain two symmetric
289: constant-voltage steps at $V\approx 37\,\mu $V in Fig. \ref{IVC1}b as
290: resonant steps supported by the ac-drive-induced modulation. Note, however,
291: that the latter steps do not feature zero-crossing. All the data collected
292: in the experiments show that the zero crossing is possible {\em solely} on
293: the resonant steps that occur in the presence of dc magnetic field. In other
294: words, the ac-driven motion of the fluxons is not possible without a
295: stationary spatially periodic potential. This inference is in no
296: contradiction with numerical results of Ref.~\onlinecite{magnetic}, where the
297: drive itself was spatially modulated.
298:
299: Coming back to the resonant steps induced by the dc field $H$, which is the
300: main subject of the work, we have also measured their size vs. $H$, see Fig.~%
301: \ref{Istep-H}. The result is that both edge values $I_{1}^{+}$ and $%
302: I_{1}^{-} $ indicated in Fig.~\ref{IVC1} vary nearly linearly with $H$, up
303: to $H\approx 0.37\,$Oe. At still larger fields, the phase- locked ac-driven
304: motion of the fluxon gets interrupted in some current range (the
305: perturbation theory does not apply to so strong fields). These findings are
306: in reasonable agreement with the theoretical prediction (\ref{interval}), a
307: fit to which is shown by the dashed lines. This pertains to both the upward
308: shift of the lines $I_{1}^{\pm }(H)$ (recall that $I$ and $H$ are
309: proportional to $\gamma $ and $h$, respectively) and their linear change
310: with the magnetic field. The residual nonzero value of $I_{1}^{+}-I_{1}^{-}$
311: at $H=0$ matches the small non-zero-crossing step in Fig. \ref{IVC1}b. It is
312: noteworthy too that the current range of the zero-voltage state, $%
313: I_{0}^{-}<I<I_{0}^{+}$ (see Fig. \ref{IVC1}b), decreases nearly linearly
314: with $H$, disappearing at $H\approx 0.09\,$Oe.
315:
316: Equation (\ref{interval}) also predicts a linear dependence of the step's
317: size on the ac-drive's amplitude $\Gamma $. Comparison with experimental
318: data shows an agreement in a broad power range; we do not display detailed
319: results of the comparison, as they display no interesting features, the
320: linear dependence of the range of existence of the inverse Josephson effect
321: vs. the amplitude of the ac signal being a common feature of all the
322: manifestations of this effect, including the Shapiro steps in small
323: junctions \cite{Barone}.
324:
325: As for the dependence of the step's size on the ac-drive's frequency at a
326: fixed value of its power, it is hard to measure it, as variation of the
327: frequency inevitably entails a change in the ac power coupled to the
328: junction. Nevertheless, basic features reported in this work, i.e., the
329: zero-crossing steps at finite voltages and disappearance of the zero-voltage
330: state, have been observed in a broad range of the ac frequencies, starting
331: from about $5$ GHz and up. On the other hand, as it was mentioned above, the
332: condition that the moving fluxon cannot exceed the Swihart velocity $\bar{c}$
333: sets an upper cutoff for the frequency that can support the phase-locked
334: motion of fluxons. In our system, $\bar{c}$ corresponds to the dc voltage $V=%
335: \bar{c}\Phi _{0}/(2\pi R)\approx 80\,\mu $V, which translates, via Eq.~(\ref
336: {spectrum}), into the cutoff frequency $\sim 40\,$GHz for the case of the
337: fundamental resonance.
338:
339: All the above results pertained to the fundamental resonance, $m=1$ in Eq.~(%
340: \ref{spectrum}). It is also easy to observe zero-crossings corresponding to
341: higher-orders resonances. This is illustrated in Fig.~\ref{IVC2}, showing $%
342: V(I)$ curves with the resonant steps generated by both the fundamental and
343: second-order (corresponding to $m=2$) resonances.
344:
345: In conclusion, we have reported the first observation of ac-driven motion of
346: a topological soliton in a periodically modulated lossy medium. The
347: observation was made in an annular uniform Josephson junction placed into
348: constant magnetic field. Experimentally measured data, such as the size of
349: the constant-voltage step, are in good agreement with predictions of the
350: analytical model. The effect may take place in a broad class of nonlinear
351: systems and, in terms of the Josephson junctions, it may find a potential
352: application as a low-frequency voltage standard.
353:
354: This work was partly supported by a Grant G0464-247.07/95 from the
355: German-Israeli Foundation. We appreciate discussions with G. Filatrella, M.
356: Fistul, E. Goldobin, N. Gr{\o }nbech-Jensen, and J. Niemeyer.
357:
358: \begin{references}
359: \bibitem{oldreview} Y.S. Kivshar and B.A. Malomed, Rev. Mod. Phys. {\bf 61}%
360: , 763 ( 1989).
361:
362: \bibitem{Barone} A. Barone and G. Patern\'{o}, {\it Physics and
363: Applications of the Josephson Effect} (Wiley: New York, 1982).
364:
365: \bibitem{Ust-rev98} A.V. Ustinov, Physica D {\bf 123}, 315 (1998).
366:
367: \bibitem{Bob1} G. Filatrella, B.A. Malomed, and R.D. Parmentier, Phys.
368: Lett. A {\bf 198}, 43 (1995).
369:
370: %\bibitem{earlier} B.A. Malomed, Phys. Rev. A {\bf 45}, 4097 (1992).
371:
372: \bibitem{Bob2} G. Filatrella {\it et al}., Phys. Lett. A {\bf 228}, 250
373: (1997).
374:
375: \bibitem{Salerno} G. Costabile et al., Phys. Rev. B {\bf 42}, 2651 (1990);
376: B. A. Malomed, Phys. Rev. B {\bf 41}, 2037 (1990);
377: M. Salerno et al., Phys. Rev. B {\bf 41}, 6641
378: (1990); G. Filatrella et al., Phys. Lett. A {\bf 173}, 127 (1993).
379:
380: \bibitem{magnetic} N. Gr{\o }nbech-Jensen, P.S.\ Lomdahl, and M.R.\
381: Samuelsen, Phys.\ Lett.\ A\ {\bf 154}, 14 (1991).
382:
383: %\bibitem{ringexperiment} A.\ Davidson, B.\ Dueholm,
384: %B.\ Kryger, and N.\ F.\
385: %Pedersen, Phys.\ Rev.\ Lett.\ {\bf 55}, 2059 (1985).
386:
387: \bibitem{vernik} I.V. Vernik {\it et al}., J. Appl. Phys. {\bf 81}, 1335
388: (1997).
389:
390: \bibitem{UstMal} A.V. Ustinov, B.A. Malomed, and N. Thyssen, Phys. Lett.
391: {\bf A 233}, 239 (1997).
392:
393: \bibitem{Denmark} N. Gr{\o }nbech-Jensen, B.A. Malomed, and M.R. Samuelsen,
394: Phys. Rev. B {\bf 45}, 9688 (1992).
395: \end{references}
396:
397: \begin{figure}[tbp]
398: \caption{Current-voltage characteristics for a fluxon trapped in the annular
399: Josephson junction irradiated by the ac signal with the frequency $18.1$ GHz
400: and power $P_{{\rm ac}}=-8$dBm. The dc magnetic field is (a) $H=0.35\,$Oe
401: and (b) $H=0$.}
402: \label{IVC1}
403: \end{figure}
404:
405: \begin{figure}[tbp]
406: \caption{The critical values $I_{0}^{\pm }$ and $I_{1}^{\pm }$ of the dc
407: bias current, marked in Fig.~\ref{IVC1}, vs. the external dc magnetic field,
408: the dashed lines showing a fit to Eq.~\ref{interval}.}
409: \label{Istep-H}
410: \end{figure}
411:
412: \begin{figure}[tbp]
413: \caption{Current-voltage characteristics for a fluxon in the annular
414: Josephson junction at $H=0.40\,$Oe, irradiated by the ac signal at the
415: frequency $10.0\,$GHz. The signal's power $P_{{\rm ac}}$ is $-3.4\,$dBm
416: (solid line) and $-12.4\,$dBm (dashed line). The constant-voltage steps on
417: the two lines correspond, respectively, to the second-order and fundamental
418: resonance in Eq. (\ref{spectrum}).}
419: \label{IVC2}
420: \end{figure}
421:
422: \end{document}
423:
424: