1: % Revetex4 file for the paper p8.
2: \documentclass[aps,prl,showpacs,showkeys,twocolumn]{revtex4}
3: %\documentclass[aps,prl,showpacs,showkeys,preprint]{revtex4}
4:
5: \usepackage{graphicx}
6: \usepackage{latexsym}
7:
8: \begin{document}
9:
10: %Title of paper
11: \title{Fractal analysis on a closed classical
12: hard-wall billiard using a simplified box-counting algorithm}
13:
14: % Authors information
15: \author{Suhan Ree}
16: \email{suhan@kongju.ac.kr}
17: \affiliation{Department of Industrial Information,
18: Kongju National University, Yesan-Up, Yesan-Gun,
19: Chungnam, 340-802, South Korea}
20:
21: \date{\today}
22:
23: \begin{abstract}
24: We perform fractal analysis on a closed classical
25: hard-wall billiard, the circular billiard with a straight cut, assuming
26: there are two openings on the boundary.
27: We use a two-dimensional set of initial conditions
28: that produce all possible trajectories of a particle injected from one opening,
29: and numerically compute the fractal dimension of singular points of a function
30: that maps an initial condition to
31: the number of collisions with the wall before the exit.
32: We introduce a simplified box-counting algorithm, which uses points from
33: a rectangular grid inside the two-dimensional set of the initial conditions, to
34: simplify the calculation,
35: and observe the classical chaotic properties
36: while varying the parameters of the billiard.
37: \end{abstract}
38:
39: % insert suggested PACS numbers in braces on next line
40: \pacs{05.45.Df, 05.45.Pq, 73.23.Ad}
41: % insert suggested keywords - APS authors don't need to do this
42: \keywords{Chaos, Billiard, Classical scattering, Fractal}
43:
44: %\maketitle must follow title, authors, abstract, \pacs, and \keywords
45: \maketitle
46:
47: % Body of the paper starts here
48: % <Introduction>===================================================
49: Chaotic systems have recently attracted many researchers from
50: various fields,
51: partly because the fast development of computer hardwares
52: has enabled us to solve equations
53: almost insolvable in the past\cite{ott,reichl}.
54: For physicists, the two-dimensional (2D) billiard system
55: has been a popular subject
56: for studying the dynamics of the Hamiltonian chaotic systems.
57: Classically, the dynamics of the 2D billiard system shows three
58: distinct types of behaviors.
59: The system is either integrable (\emph{regular} behavior) or
60: non-integrable
61: (either \emph{soft chaos}, characterized by mixed phase spaces
62: that have
63: both regular and chaotic regions, or \emph{hard chaos},
64: characterized by ergodicity and mixing)\cite{gutzwiller}.
65:
66: Here we use a closed 2D hard-wall billiard
67: with two openings on the boundary, and perform fractal analysis using the
68: classical dynamics
69: while varying the shape of the bliiard and the size of the openings.
70: To calculate the fractal dimension,
71: we use a set of initial conditions that will produce trajectories
72: of a particle injected from one opening,
73: and calculate certain values for each trajectory, such as
74: the \emph{exit opening} (an opening from which
75: the particle exits), the number of collisions with the wall
76: before the exit,
77: the dwell time, and so on.
78: When the system is chaotic, singular points of this kind of response
79: functions form a fractal set\cite{eckhardt,bleher,ott2}.
80: In this Letter, we will calculate the fractal dimension of
81: these singular points of the function that maps an initial condition
82: to the number of collisions,
83: by introducing a simplified box-counting algorithm.
84:
85: % <Method>===================================================
86: The billiard we have chosen is the circular billiard with a straight cut,
87: or the ``cut-circle'' billiard (see Fig.\ \ref{geometry}).
88: %==== figure <geometry>
89: \begin{figure}
90: \includegraphics[scale=0.7]{fig1.eps}
91: %\includegraphics[scale=1.0]{fig1.eps}
92: \caption{\label{geometry}
93: (a) Geometry of the cut-circle billiard. There are
94: two openings, I and II, and a straight cut.
95: The size of the cut is given by the width $W$ (here we use $w\equiv W/R$
96: as the relative size of the cut), and
97: the size of two openings is $\Delta$.
98: The positions of the cut and the opening II, $\Omega$ and $\gamma$,
99: are measured from the position of the opening I.
100: (b) When a particle is injected
101: from the central point of the opening I with an angle $\phi$, the set of possible
102: initial conditions is one dimensional $(-\pi/2 < \phi < \pi/2)$.
103: (c) To fully represent all significant initial conditions,
104: a particle is injected from a point on the circular boundary of the
105: opening I with the angular momentum $P_{\theta \rm in}$. This set, whose element is
106: represented by
107: a pair of values, $(\theta_{\rm in},P_{\theta \rm in})$, is
108: a closed two-dimensional subspace of the four-dimensional phase space.
109: }
110: \end{figure}
111: %====
112: There are five parameters: (1) the width $W$ measured in the
113: direction perpendicular to the cut, (2) the radius $R$, (3) the angular
114: width $\Delta$ of the openings, which represents the opening size,
115: (4) the orientation angle ${\Omega}$ of the cut relative
116: to the first opening, and (5) the position of the second
117: opening relative to the
118: first opening as measured by the angle $\gamma$.
119: We scale the width $W$ by $R$, so
120: $w\equiv W/R$, thereby reducing the number of independent
121: parameters to four:
122: $w$, $\Delta$, $\Omega$, and $\gamma$.
123: For all subsequent discussions, we set $\Omega=135^\circ$ and $\gamma=270^\circ$.
124: For the closed cut-circle billiard (when $\Delta=0$), it has been proved in Ref.\
125: \onlinecite{bunimovich} that the phase
126: space is mixed (soft chaos) when $0<w<1$, and that the phase space is fully chaotic
127: (hard chaos) when $1<w<2$.
128: And it is integrable when $w=1$ and 2.
129: Consequently this billiard has a single shape parameter $w$
130: that makes the billiard exhibit all three types of chaotic behaviors.
131:
132: In Ref.\ \onlinecite{ree7}, a one-dimensional (1D) set of initial conditions
133: of the cut-circle billiard
134: was used to calculate fractal dimensions while varying the opening size $\Delta$ and the
135: shape parameter $w$ [see Fig.\ \ref{geometry}(b)].
136: The particle is injected with an incident angle $\phi$ ($-\pi/2<\phi<\pi/2$).
137: In Ref.\ \onlinecite{legrand2}, authors used a 1D set of initial conditions
138: for the \emph{stadium},
139: and calculated the recurrence time for each trajectory to obtain the fractal dimension.
140: These 1D sets, however, are subsets of the set of all significant initial conditions
141: that represent all possible trajectories;
142: in our billiard, a 2D subspace in the four-dimensional (4D) phase space
143: is a minimal set that will produce all possible trajectories originating
144: from one opening [see Fig.\ \ref{geometry}(c)].
145: (A 2D set of initial conditions were also used in Ref.\ \onlinecite{bleher}
146: for the Sinai billiard.)
147: There are two reasons that make this reduction possible.
148: First, the value of the energy does not change the trajectory of the particle.
149: This is due to the flat
150: potential inside the billiard with infinitely hard boundaries.
151: Second, the distance from
152: the center of the billiard to the initial location of the particle can be
153: fixed at $R$.
154:
155: With these two kinds of sets, we numerically find a function that maps an initial
156: condition to the number of collisions before the exit.
157: (In numerical computations, all possible initial conditions cannot be used;
158: only a finite and discrete subset can be considered.)
159: In Fig.\ \ref{ncols}, we show how the number of collisions changes for these
160: sets of initial conditions when $w=1.5$ and $\Delta=60^\circ$.
161: %==== figure <ncols>
162: \begin{figure}
163: \includegraphics[scale=0.45]{fig2.eps}
164: %\includegraphics[scale=1.0]{fig2.eps}
165: \caption{\label{ncols}
166: Calculations of the number of collisions when $\Omega=135^\circ$,
167: $\gamma=270^\circ$, $w=1.5$, and $\Delta=60^\circ$.
168: (a) Graph of the incident angle $\phi$ vs
169: the number of collisions $N_C$
170: when the
171: particle is injected as in Fig.\ \ref{geometry}(b).
172: (b) When the particle is injected as in Fig.\ \ref{geometry}(c), all significant
173: initial conditions lie inside the closed two-dimensional space bounded
174: by the bold line in the graph. In the graph, only singular points are shown.
175: The dotted line represents the
176: one-dimensional subset of initial conditions used in (a).
177: Five geometrical channels with low $N_C$'s are also shown.
178: }
179: \end{figure}
180: %====
181: In Fig.\ \ref{ncols}(a), the graph of $\phi$ vs the number of collisions is
182: shown when the 1D subset of initial conditions as in Fig.\ \ref{geometry}(b)
183: is used (see Ref.\ \onlinecite{ree7} for more results).
184: In Fig.\ \ref{ncols}(b), the graph shows the full 2D set of initial conditions,
185: $(\theta_{\rm in},P_{\theta\rm in})$, bounded by two equations,
186: \begin{equation}
187: -\frac{\Delta}{2} \le \theta{\rm in} \le \frac{\Delta}{2},
188: \end{equation}
189: \begin{equation}
190: -\cos\left(\frac{\Delta}{4}+\frac{\theta_{\rm in}}{2}\right) \le
191: \frac{P_{\theta\rm in}}{\sqrt{2mE}R} \le \cos\left(\frac{\Delta}{4}-
192: \frac{\theta_{\rm in}}{2}\right),
193: \end{equation}
194: which are represented by the bold line.
195: Instead of the number of collisions,
196: the graph shows singular points that constitute singular boundaries
197: at which the number of collisions changes.
198: We can observe that singular points have infinitely fine structures.
199: The number of collisions are shown only for five
200: \emph{geometrical channels}\cite{roukes,luna} in this example.
201: The dotted line is the 1D subset used in Fig.\ \ref{ncols}(a).
202:
203: Now we calculate the fractal dimension $d_f$ of
204: a set of singular points in this 2D set using a simplified box-counting algorithm.
205: As a generalization of the box-counting algorithm used in Ref.\ \onlinecite{ree7}
206: for the 1D set,
207: we use a uniform 2D rectangular grid.
208: Then there are uniform rectangular boxes, and each of them is
209: represented by four points on the grid.
210: For all of these uniformly distributed grid points inside the 2D set,
211: we numerically find
212: the number of collisions before the exit,
213: and for each box, we compare the number of collisions at four corners.
214: With an assumption that a box does not contain a singular point
215: when four values are all equal, we can count the number of boxes
216: containing singularities out of all boxes inside the 2D set.
217: The number of all boxes inside the set is $N_P$, and
218: the number of boxes that contain any singular point, based on the
219: above assumption, is $N_S$.
220: Then the fractal dimension $d_f$ is defined by
221: \begin{equation}
222: \label{df}
223: d_f\equiv\lim_{N_P\rightarrow\infty}\frac{\log_{10}{N_S}}{\log_{10}{N_P}^{1/2}},
224: \end{equation}
225: which is the slope of the graph of $0.5\log_{10}{N_P}$ vs $\log_{10}{N_S}$.
226: In numerical calculations, we find $N_S$ for several $N_P$ values,
227: and then use the ordinary least-square fit
228: to find the slope using the points in the graph.
229:
230: % <Numerical Results>===================================================
231: In Fig.\ \ref{angles}(a), the log-log graph is shown for several $\Delta$ values
232: when $w=1.5$. (For all calculations, $N_P$ up to $10^8$ will be used.)
233: %==== figure <angles>
234: \begin{figure}
235: \includegraphics[scale=0.7]{fig3.eps}
236: %\includegraphics[scale=1.0]{fig3.eps}
237: \caption{\label{angles}
238: (a) When $w=1.5$,
239: the number of boxes containing the singular points, $N_S$, is counted
240: as the number of boxes, $N_P$, increases,
241: for five different $\Delta$ values.
242: Then $d_f$ can be found from the slopes of these log-log graphs.
243: (b) Graphs of $\bar{d_f}$ vs the opening size $\Delta$
244: for $w=0.75$, $1.04$, and $1.5$.
245: Here we compare the values of $\bar{d_f}$ for the 2D set ($\circ$) and
246: for the 1D set ($*$).
247: }
248: \end{figure}
249: %====
250: Unlike the calculations using the 1D set, $d_f$ will be in the range of
251: $1\le d_f \le 2$.
252: We can define $\bar{d_f}$, which better represents the fractality,
253: \begin{equation}
254: \bar{d_f}\equiv \left\{ \begin{array}{rl}
255: 2-d_f~&\mbox{(for the 2D set)}\\
256: 1-d_f~&\mbox{(for the 1D set)}
257: \end{array} \right..
258: \end{equation}
259: In Fig.\ \ref{angles}(b), $\bar{d_f}$'s both for the 2D set and for the 1D set
260: are compared for three different $w$ values while varying $\Delta$.
261: Like the curves for the 1D set,
262: not all curves monotonically increase
263: as $\Delta$ increases, even though it is not as significant.
264: The reason for this phenomenon is that there are more possible trajectories
265: when the opening gets bigger, and it also explains why
266: curves don't match very well for large $\Delta$.
267: We also observe that the cases with hard chaos ($w=1.04$, and 1.5) do not behave
268: predictably for large $\Delta$, which will be more clearly seen in the following
269: graph.
270: When the original shape is changed sufficiently by the presence of the
271: openings, the ergodicity of the billiard
272: is no longer an important factor; the relative locations of the
273: openings start to
274: have more effect on the fractal dimension as in the cases of soft chaos.
275:
276: In Fig.\ \ref{widths}, we calculate $\bar{d_f}$ as the width $w$ varies from 0.5
277: to 2 with a step size of 0.02
278: for seven opening sizes
279: ($\Delta=1^\circ$, $\Delta=5^\circ$, $\Delta=10^\circ$,
280: $\Delta=20^\circ$, $\Delta=30^\circ$, $\Delta=40^\circ$, and
281: $\Delta=50^\circ$), using the 2D set.
282: %==== figure <widths>
283: \begin{figure}
284: \includegraphics[scale=0.45]{fig4.eps}
285: %\includegraphics[scale=1.0]{fig4.eps}
286: \caption{\label{widths}
287: Graphs of $w$ vs $\bar{d_f}$ for seven different
288: opening sizes: $\Delta=1^\circ$ ($\circ$), $\Delta=5^\circ$ ($\Box$),
289: $\Delta=10^\circ$ ($\Diamond$),
290: $\Delta=20^\circ$ ($\bigtriangleup$), $\Delta=30^\circ$ ($\bigtriangledown$),
291: $\Delta=40^\circ$ ($+$), and
292: $\Delta=50^\circ$ ($*$), for the 2D set.
293: When $\Delta$ is small enough ($\Delta\le20^\circ$ in the graph),
294: the behavior of the graph
295: is clearly distinct for two regions:
296: hard chaos ($1<w<2$) and soft chaos ($0<w<1$). But for larger $\Delta$,
297: the distinction between two regions starts to disappear.
298: }
299: \end{figure}
300: %====
301: We can compare graphs in two different regions (found in
302: the closed cut-circle billiard): $0<w<1$ (soft chaos)
303: and $1<w<2$ (hard chaos).
304: We observed similar behaviors seen in results for the 1D set in Ref.\
305: \onlinecite{ree7}.
306: However $\bar{d_f}$ does not reach one
307: ($\bar{d_f}=1$ represents \emph{non-chaos}),
308: even when the billiard is integrable ($w=1$ or 2),
309: and this is due to the fine structure near trajectories
310: bouncing very close to the circular boundary.
311: When the opening size $\Delta$ is not big (see graphs for
312: $\Delta=5^\circ$, $\Delta=10^\circ$, and $\Delta=20^\circ$),
313: we observe that the behavior in two regions is clearly distinct.
314: In the region $0<w<1$, there are fluctuations, which comes from
315: the mixed phase space structures
316: of the billiard,
317: and in the region $1<w<2$,
318: graphs are smooth because the phase spaces of the billiard
319: have no structure due to ergodicity.
320: On the other hand, when the opening size gets bigger
321: (see graphs for $\Delta=30^\circ$,
322: $\Delta=40^\circ$, and $\Delta=50^\circ$), the distinction between
323: two regions, observed in cases with smaller openings, disappears.
324: There are fluctuations in both regions.
325:
326: % <Discussions>===================================================
327: In this Letter, we calculated the fractal dimension for the cut-circle
328: billiard with two openings by introducing the simplified box-counting algorithm
329: for the 2D set of initial conditions.
330: The simplified box-counting algorithm introduced here is possible
331: because the function used to find singularities is constant
332: except singular points (i.\ e., the function is constant with
333: different values in all regions surrounded by singular boundaries).
334: With this method, some boxes containing singular points will be missed,
335: but the number of missed boxes is negligible
336: when we calculate the fractal dimension with big enough $N_P$ values.
337: Results for the 2D set were close to those for the 1D set
338: only when $\Delta$ is small enough;
339: hence using the 1D set in fractal analysis does not
340: fully represent the chaoticity of the billiard
341: for most cases.
342: In conclusion, this kind of fractal analyses gives us one of the fundamental
343: measures of the chaoticity of classical billiard systems,
344: and finding the relations with the quantum and semiclassical dynamics
345: of the same kind of billiards will be one of the interesting future works.
346: % End of the body
347:
348: \begin{acknowledgments}
349: This work was supported by Kongju National University.
350: \end{acknowledgments}
351:
352: % Create the reference section using BibTeX:
353: \bibliography{p8}
354:
355: \end{document}
356:
357: