1: % Revetex4 file for the paper p7.
2: % submitted to JKPS (and also ArXiv)
3: \documentclass[aps,pre,showpacs,showkeys,twocolumn]{revtex4}
4:
5: \usepackage{graphicx}
6: \usepackage{latexsym}
7:
8: \begin{document}
9:
10: %Title of paper
11: \title{Numerical studies on chaoticity of a classical hard-wall billiard
12: with openings}
13:
14: % Authors information
15: \author{Suhan Ree}
16: \affiliation{Department of Industrial Information,
17: Kongju National University, Yesan-Up, Yesan-Gun,
18: Chungnam, 340-802, South Korea}
19: \email{suhan@kongju.ac.kr}
20:
21: \date{\today}
22:
23: \begin{abstract}
24: Using fractal analysis, we investigate how the size of openings
25: affects the chaotic behavior of
26: a classical closed billiard when two
27: openings are made on the boundary of the billiard.
28: This kind of open billiards retains chaotic properties of
29: original closed billiards when openings are small compared to the size
30: of the billiard.
31: We calculate the fractal dimension using an
32: one-dimensional subset of all possible
33: initial conditions that will produce
34: trajectories of a particle injected from an opening
35: for the billiard,
36: and then observe how the opening size
37: can change the classical chaotic properties of this open billiard.
38: \end{abstract}
39:
40: % insert suggested PACS numbers in braces on next line
41: \pacs{05.45.Df, 05.45.Pq, 73.23.Ad}
42: % insert suggested keywords - APS authors don't need to do this
43: \keywords{Chaos, Billiard, Classical scattering, Fractal}
44:
45: %\maketitle must follow title, authors, abstract, \pacs, and \keywords
46: \maketitle
47:
48: % Body of the paper starts here
49: Two-dimensional (2D) hard-wall billiard systems have been a popular subject
50: for studying the dynamics of chaotic systems\cite{reichl}.
51: Theoretically, 2D billiard systems can be used to study manifestations of classical
52: chaos in semiclassical and quantum mechanics.
53: The classical dynamics of a billiard system shows three distinct types of behavior:
54: the system is either integrable (\emph{regular} behavior) or non-integrable
55: (either \emph{soft chaos}, characterized by mixed phase spaces that have
56: both regular and chaotic regions, or \emph{hard chaos}, characterized by
57: ergodicity and mixing)\cite{gutzwiller}.
58: In experiments, on the other hand, the billiard systems can be used as models
59: to explain fluctuating behavior of magnetoconductance through
60: 2D semiconductor heterostructures in the ballistic regime\cite{markus,chang}.
61: In such an experiment, the structure have the shape of a 2D billiard, of which the dynamic
62: properties of the billiard are already known (for example, the circle as
63: a regular system, or the stadium as a system with hard chaos).
64: To measure the conductance through the structure, leads
65: (from here on, we will call them \emph{openings}) are attached to the structure;
66: hence the whole structure becomes an open billiard.
67: One usually studies transmission properties through these open billiards,
68: and links results to the classical chaotic properties of original
69: closed billiards, because the properties of the closed billiard
70: are still observed indirectly when the size of the openings is small
71: compared to the size of the billiard
72: (see Refs.~\cite{baranger,ree2,fuchss}).
73: Therefore, to better relate the open billiard to the original closed billiard,
74: quantitative analysis on effects of openings in classical dynamics
75: is necessary.
76:
77: In this paper, we numerically calculate a quantity that represents the chaoticity
78: of classical open billiards, and see how the size of the openings affects the overall
79: classical dynamics.
80: In Ref.~\cite{bleher}, the fractal dimension was obtained by looking at
81: a two-dimensional set of initial conditions that will produce
82: trajectories injected from an opening of the Sinai billiard with two openings.
83: [A set of initial conditions in four-dimensional phase space can
84: be reduced to a two-dimensional set, because (1) energy of the particle
85: does not affect the trajectory of a particle, and (2) initial
86: locations of the particle can be restricted to an one-dimensional (1D) curve.]
87: Here, however, we will use
88: an 1D subset of all possible initial
89: conditions for calculations of the fractal dimension to make
90: numerical calculations simple.
91: The incident angle of the particle injected from the center of
92: one opening represents an 1D subset of
93: all possible initial conditions,
94: which give rise to some possible trajectories of injected particles.
95: In Ref.~\cite{ree3}, graphs of the ``exit opening'' (an opening from which
96: the particle exits) vs the incident angle from the center of an opening
97: were used.
98: There are ``geometrical channels''\cite{roukes,luna} inside a range of
99: incident angles.
100: When there are two openings, we can define
101: \emph{transmission windows} that represent initial angles for transmitting particles.
102: The boundaries of these windows will form a fractal\cite{bleher,ree3}.
103: In this paper, instead of calculating the fractal dimensions of boundaries of
104: these transmission windows,
105: we use the graph of the number of collisions vs the incident angle, and calculate the
106: fractal dimension of the set of singularities, at which the number
107: of collisions changes abruptly,
108: inside the set of all possible incident angles (see Ref.~\cite{eckhardt}).
109: The fractal dimensions from these two graphs will be eventually the same.
110: %But the latter method is more general, because it does not depend on
111: %the number of openings.
112:
113: The billiard used here for numerical calculations
114: is the circular billiard with a straight cut (\emph{the cut-circle
115: billiard}) with two openings
116: (see Fig.\ \ref{geometry}).
117: %==== figure <geometry>
118: \begin{figure}
119: \includegraphics[scale=0.5]{fig1.eps}
120: \caption{\label{geometry}
121: The geometry of the cut-circle billiard. The
122: billiard has two openings, I and II.
123: The size of the cut is given by the width $W$ (we define $w$ as $W/R$
124: to represent the relative size of the cut).
125: The size of two openings is $\Delta$, and the position of
126: the cut and the opening II, $\Omega$ and $\gamma$,
127: are measured from the position of the opening I.
128: A particle is injected
129: from the center of the opening I with an angle $\phi$.}
130: \end{figure}
131: %====
132: There are five parameters: (1) the width $W$ measured in the
133: direction perpendicular to the cut, (2) the radius $R$, (3) the angular
134: width $\Delta$ of the openings, which represents the opening size,
135: (4) the orientation angle
136: ${\Omega}$ of the cut relative
137: to the first opening, and (5) the position of the second
138: opening relative to the
139: first opening as measured by the angle $\gamma$.
140: We scale the width $W$ by $R$ so
141: $w\equiv W/R$, and thereby reduce the number of independent
142: parameters to four:
143: $w$, $\Delta$, $\Omega$, and $\gamma$. For all subsequent
144: discussions, we set
145: $\Omega=135^\circ$ and $\gamma=270^\circ$.
146: For these values, $w$ cannot be less than $[1-\cos(45^\circ)]\simeq0.293$, and
147: $\Delta$ has an upper limit $\Delta_{max}$,
148: \begin{equation}
149: \label{deltamax}
150: \Delta_{max} =\left\{
151: \begin{array}{ll}
152: 2\;[\cos^{-1}(1-w)-\pi/4]~ &\rm{when}~0.293<w<1\\
153: \pi/2 & \rm{when}~1\le w \le 2
154: \end{array}
155: \right. ,
156: \end{equation}
157: in radian.
158: When the billiard is closed ($\Delta=0$), it has been proven
159: that the phase
160: space is mixed (soft chaos) when $0<w<1$, and that the phase space is fully chaotic
161: (hard chaos) when $1<w<2$\cite{bunimovich}.
162: The system is integrable when $w=1$ and 2.
163: This billiard has a property of showing all three types of dynamic behaviors
164: just by changing a parameter $w$, and this is the reason why the cut-circle billiard
165: is a good example for analysis like this.
166:
167: The particle is injected with an incident angle $\phi$ ($-\pi/2<\phi<\pi/2$).
168: One can easily calculate several quantities by numerically following these
169: trajectories, but we focus only on
170: the number of collisions with the wall before the exit with respect to $\phi$.
171: In other words, we find a mapping from the incident
172: angle in opening I to the number of collisions before the exit.
173: In Fig.\ \ref{ncols}, we show how the number of collisions changes with $\phi$
174: for five different $w$ values ($w=0.5$,
175: $w=0.75$, $w=1.04$, $w=1.5$, and $w=1.71$) when $\Delta=30^\circ$.
176: %==== figure <ncols>
177: \begin{figure}
178: \includegraphics[scale=0.45]{fig2.eps}
179: \caption{\label{ncols}
180: Graphs of the number of collisions vs the incident angle $\phi$.
181: Here the number of collisions is scaled for each $w$ to fit in
182: graphs with the same sizes, because
183: we are only interested in positions of singularities.
184: Five cases with different cut sizes are shown : $w=0.5$,
185: $w=0.75$, $w=1.04$, $w=1.5$, and $w=1.71$.
186: }
187: \end{figure}
188: %====
189: In these calculations, there are finite number of singularities when $w=0.5$,
190: but, in other cases, there are regions where singular points are closely packed together
191: showing infinitely fine structures.
192: We observe that there tends to be more singularities as $w$ increases.
193: That is because (1) when $w$ is bigger than one, the
194: original closed billiard shows hard chaos, which means global chaos, and
195: (2) the opening size
196: is fixed here for all $w$'s even though the overall size of the billiard
197: increases as $w$ increases, as a result,
198: making the relative size of the openings smaller.
199:
200: We calculate the fractal dimension $d_f$ for
201: sets of singularities for various cases,
202: using a simplified box-counting algorithm.
203: For $N_P$ uniformly distributed incident angles, we numerically find
204: the number of collisions for each incident angle representing an initial condition.
205: For several $N_P$ values, we find the number of singularities $N_S$
206: (assuming that there only exists a singularity when the number of collisions changes
207: for two successive incident angles).
208: Then the fractal dimension $d_f$ is defined by
209: \begin{equation}
210: \label{df}
211: d_f\equiv\lim_{N_P\rightarrow\infty}\frac{\log_{10}{N_S}}{\log_{10}{N_P}},
212: \end{equation}
213: which is the slope of the graph of $\log_{10}{N_S}$ vs $\log_{10}{N_P}$.
214: We use the ordinary least-square fit
215: to find the slope using points in the graph with limited $N_P$.
216: Because the numerical accuracy for numbers is limited, $N_P$
217: have an upper limit.
218: Moreover the numerically given error of an initial condition will amplify as
219: the particle travels through the billiard and hits the boundary while traveling.
220: Here we use the double-precision numbers with 15 significant digits,
221: and then $N_P$ should be less than $10^{15}$, especially when the number
222: of collisions is big.
223: In most cases in our example, $N_P$ up to $10^7$ were used
224: to obtain $d_f$'s with two significant digits.
225: In some cases (for example,
226: when $w=1.71$ and $\Delta>59^\circ$), however, $N_P$'s up to $10^8$ were
227: used to find $d_f$'s,
228: because the slopes didn't converge sufficiently
229: when $N_P$ is less than $10^7$.
230:
231:
232: In Fig.\ \ref{angles}, we fix the width $w$ at five different values used in Fig.\
233: \ref{ncols}, and vary the opening size $\Delta$ up to $\Delta_{max}$ for each $w$.
234: %==== figure <angles>
235: \begin{figure}
236: \includegraphics[scale=0.45]{fig3.eps}
237: \caption{\label{angles}
238: Graphs of the fractal dimension $d_f$ vs the opening size $\Delta$.
239: For each $w$ ($w=0.5$, $w=0.75$, $w=1.04$, $w=1.5$, and $w=1.71$),
240: $\Delta$ is varied up to a maximum value $\Delta_{max}$.
241: We can observe that $d_f$ does not always decreases monotonically
242: as the opening size $\Delta$ increases.
243: }
244: \end{figure}
245: %====
246: An interesting feature of these curves is that they are not monotonically decreasing
247: as $\Delta$ increases. Small peaks appear even after $d_f$ reaches zero
248: at a certain $\Delta$ value. For example, the $d_f$-curve for $w=1.04$ reaches
249: zero near $\Delta\sim35^\circ$, but there is a small peak near $\Delta\sim79^\circ$.
250: The reason for this phenomenon is that there are more possible trajectories
251: when the opening is bigger.
252: In our method for the circular billiard,
253: the launching point of the particle
254: gets closer to the center of the billiard as the opening
255: size increases.
256:
257: In Fig.\ \ref{traj}, two trajectories that doesn't exist
258: when openings are smaller are shown.
259: %==== figure <traj>
260: \begin{figure}
261: \includegraphics[scale=0.45]{fig4.eps}
262: \caption{\label{traj}
263: Two trajectories showing why the function of $d_f$ with respect to $\Delta$
264: does not always decreases monotonically. As the opening size $\Delta$ increases, new
265: initial conditions for trajectories, which can be very sensitive
266: to the incident angle $\phi$, may appear. For example,
267: (a) $w=1.04$, $\Delta=79^\circ$, and $\phi=84.44^\circ$. (b) $w=1.71$,
268: $\Delta=59^\circ$, and $\phi=88.40^\circ$.
269: }
270: \end{figure}
271: %====
272: Since these trajectories have the higher numbers of collisions and there
273: is sensitive dependence on initial conditions near them, $d_f$ becomes non-zero.
274: In Fig.\ \ref{traj}(a), the billiard ($w=1.04$, $\Delta=79^\circ$) lost most
275: of the original shape of the closed billiard, but trajectories near
276: the one shown ($\phi=84.44^\circ$) still make up an infinitely fine structure.
277: In Fig.\ \ref{traj}(b), the billiard ($w=1.71$, $\Delta=59^\circ$) has
278: a trajectory ($\phi=88.40^\circ$)
279: that gets close to the 3-bounce closed orbit in the circular
280: billiard. (The closed orbits in the circular billiard are stable, but only
281: marginally\cite{berry,ree}.)
282: Even though only few bounces are shown, this trajectory actually bounces
283: more than two thousand times as the triangular shape rotates clockwise little by little
284: for each rotation, before the particle finally exits through either one opening.
285: A high peak near $\phi=59^\circ$ for $w=1.71$ in Fig.\ \ref{angles} is due to
286: trajectories near this one.
287:
288: In Fig.\ \ref{widths}, we calculate $d_f$'s as the width $w$ varies from 0.5
289: to 2 with the step size 0.01,
290: for several opening sizes ($\Delta=0.5^\circ$,
291: $\Delta=1^\circ$, $\Delta=5^\circ$, $\Delta=10^\circ$,
292: $\Delta=20^\circ$, $\Delta=30^\circ$, $\Delta=40^\circ$, and
293: $\Delta=50^\circ$).
294: %==== figure <widths>
295: \begin{figure}
296: \includegraphics[scale=0.45]{fig5.eps}
297: \caption{\label{widths}
298: Graphs of the fractal dimension $d_f$ vs $w$ for 8 different
299: opening sizes, $\Delta=0.5^\circ$ ($\circ$), $\Delta=1^\circ$ ($\Box$),
300: $\Delta=5^\circ$ ($\Diamond$),
301: $\Delta=10^\circ$ ($\bigtriangleup$), $\Delta=20^\circ$ ($\bigtriangledown$),
302: $\Delta=30^\circ$ ($+$),
303: $\Delta=40^\circ$ ($\circ$), and $\Delta=50^\circ$ ($\Box$).
304: When $\Delta=5^\circ$ and $\Delta=10^\circ$, the behavior of the graph
305: is clearly distinct for two regions:
306: hard chaos ($1<w<2$) and soft chaos ($0<w<1$). But when
307: $\Delta>20^\circ$, the distinction
308: between two regions disappears.
309: }
310: \end{figure}
311: %====
312: We can compare graphs in two different regions (found in
313: the closed cut-circle billiard): $0<w<1$ (soft chaos)
314: and $1<w<2$ (hard chaos).
315: When $w=1$ and $w=2$ (integrable cases), $d_f$ is zero in our calculation.
316: As the opening size $\Delta$ approaches zero (see graphs for
317: $\Delta=0.5^\circ$ and $\Delta=1^\circ$), we observe that the fractal
318: dimension $d_f$ approaches one in both regions in our calculation.
319: When the opening size $\Delta$ is not big (see graphs for
320: $\Delta=5^\circ$, $\Delta=10^\circ$, and $\Delta=20^\circ$),
321: we observe that behavior in two regions is clearly distinct.
322: In the region $0<w<1$, there are fluctuations, which comes from
323: the mixed phase space structures
324: of the billiard,
325: and in the region $1<w<2$,
326: graphs are smooth because the phase spaces of the billiard
327: have no structure due to ergodicity.
328: On the other hand, when the opening size gets bigger
329: (see graphs for $\Delta=30^\circ$,
330: $\Delta=40^\circ$, and $\Delta=50^\circ$), the distinction between
331: two regions, observed in cases with smaller openings, starts to disappear;
332: there are fluctuations in both regions.
333: Ergodicity in closed cut-circle billiards with hard chaos is no longer
334: an important factor in the dynamics of open billiards when openings are big,
335: as expected.
336:
337: We have so far observed numerical results for the cut-circle billiard with
338: two openings.
339: A method that calculates
340: the fractal dimension using an 1D subset of
341: all possible initial conditions has been used to determine
342: effects of openings on chaotic closed billiard quantitatively.
343: From numerical results, we have found that the fractal dimension
344: as a function of the opening size is not a monotonically
345: decreasing function,
346: because there are more trajectories, which can be sensitive to
347: initial conditions, for bigger openings.
348: These new trajectories can make up infinitely fine
349: structures, and subsequently can cause the fractal dimensions to increase
350: with $\Delta$.
351: We also have found that, as the opening size gets bigger, the distinction
352: caused by soft chaos and hard chaos of the original closed billiard
353: starts to disappear,
354: because the open billiard
355: that are constructed by introducing openings to the closed billiard
356: starts to lose the original shape of the closed billiard
357: as the size of openings gets bigger.
358:
359: There are several points worth mentioning.
360: First, if we use the 2D full set of initial conditions to calculate
361: the fractal dimensions as was done in Ref.~\cite{bleher},
362: $d_f$ will be in the range of $1\le d_f \le 2$, but the overall
363: behavior of $d_f$ is expected to be similar to that of the results obtained here.
364: Second, we can use a mapping from initial conditions to
365: other quantities like dwell times (or travel distances)
366: instead of the exit openings and
367: the number of collisions. Even though the dwell time
368: will be more expensive to calculate numerically, the set of
369: singularities from the dwell time will be almost the same as those
370: from the number of collisions.
371: Finally, it has been found to be important and necessary to understand
372: relations between classical dynamics and quantum (or
373: semiclassical) dynamics of billiards\cite{blumel,ott2,jung},
374: because the ballistic scattering is usually realized in microscopic scales.
375: Then it will be interesting to ask how this kind of
376: fractal dimensions can be related to the quantum and semiclassical
377: dynamics of the same kinds of open billiards.
378: % End of the body
379:
380: \begin{acknowledgments}
381: This work was supported by Kongju National University.
382: \end{acknowledgments}
383:
384: % Create the reference section using BibTeX:
385: %\bibliography{p7}
386:
387: \begin{thebibliography}{}
388: \bibitem{reichl}{
389: L.\ E.\ Reichl, \emph{The Transition to Chaos in Conservative Classical Systems: Quantum Manifestations} (Springer-Verlag, New York, 1992).
390: }
391:
392: \bibitem{gutzwiller}{
393: M.\ C.\ Gutzwiller, \emph{Chaos in Classical and Quantum Mechanics} (Springer-Verlag, New York, 1990).
394: }
395:
396: \bibitem{markus}{
397: C.\ M.\ Markus, R.\ M.\ Westerfelt, P.\ F.\ Hopkins, and A.\ C.\ Gossard, Chaos \textbf{3}, 643 (1993).
398: }
399:
400: \bibitem{chang}{
401: A.\ Chang, H.\ U.\ Baranger, L.\ N.\ Pfeifer, and K.\ W.\ West, Phys.\ Rev.\ Lett. \textbf{73}, 2111 (1994).
402: }
403:
404: \bibitem{baranger}{
405: H.\ U.\ Baranger, R.\ A.\ Jalabert, and A.\ D.\ Stone, Chaos \textbf{3}, 665 (1993).
406: }
407:
408: \bibitem{ree2}{
409: S.\ Ree and L.\ E.\ Reichl, Phys.\ Rev.\ B \textbf{59}, 8163 (1999).
410: }
411:
412: \bibitem{fuchss}{
413: K.\ Fuchss, S.\ Ree, and L.\ E.\ Reichl, Phys.\ Rev.\ E \textbf{63}, 016214 (2000).
414: }
415:
416: \bibitem{bleher}{
417: S.\ Bleher, C.\ Grebogi, E.\ Ott, and R.\ Brown, Phys.\ Rev.\ A \textbf{38}, 930 (1988).
418: }
419:
420: \bibitem{ree3}{
421: S.\ Ree and L.\ E.\ Reichl, Phys.\ Rev.\ E \textbf{65}, 055205(R) (2002).
422: }
423:
424: \bibitem{roukes}{
425: M.\ L.\ Roukes and O.\ L.\ Alerhand, Phys.\ Rev.\ Lett. \textbf{65}, 1651 (1990).
426: }
427:
428: \bibitem{luna}{
429: G.\ A.\ Luna-Acosta, A.\ A.\ Krokhin, M.\ A.\ Rodr\'{\i}guez, and P.\ H.\ Hern\'{a}ndez-Tejeda, Phys.\ Rev.\ B \textbf{54}, 11410 (1996).
430: }
431:
432: %\bibitem{farmer}{
433: % J.\ D.\ Farmer, Phys.\ Rev.\ Lett. \textbf{55}, 351 (1985).
434: %}
435:
436: %\bibitem{grebogi}{
437: % C.\ Grebogi, S.\ W.\ McDonald, E\. Ott, and J.\ A.\ Yorke, Phys.\ Lett. \textbf{110A}, 1 (1985).
438: %}
439:
440: %\bibitem{ott}{
441: % E.\ Ott, \emph{Chaos in Dynamical Systems} (Cambridge, New York, 1993).
442: %}
443:
444: \bibitem{eckhardt}{
445: B.\ Eckhardt, J.\ Phys.\ A \textbf{20}, 5971 (1987).
446: }
447:
448: \bibitem{bunimovich}{
449: L.\ A.\ Bunimovich, Commun.\ Math.\ Phys. \textbf{65}, 295 (1979).
450: }
451:
452: \bibitem{berry}{
453: M.\ V.\ Berry, Eur.\ J.\ Phys. \textbf{2}, 91 (1981).
454: }
455:
456: \bibitem{ree}{
457: S.\ Ree and L.\ E.\ Reichl, Phys.\ Rev.\ E \textbf{60}, 1607 (1999).
458: }
459:
460: \bibitem{blumel}{
461: R.\ Bl\"{u}mel and U.\ Smilansky, Phys.\ Rev.\ Lett. \textbf{60}, 447 (1988).
462: }
463:
464: \bibitem{ott2}{
465: E.\ Ott and T.\ T\'{e}l, Chaos \textbf{3}, 447 (1993).
466: }
467:
468: \bibitem{jung}{
469: C.\ Jung and T.\ Seligman, Phys.\ Rep. \textbf{285}, 77 (1997).
470: }
471:
472: \end{thebibliography}
473: \end{document}
474: