nlin0505055/3rd.tex
1: \documentclass[12pt]{article}
2: \usepackage{epsf}  
3: \newcommand{\beq }{\begin{eqnarray}}
4: \newcommand{\eeq }{\end{eqnarray}}
5: \newcommand{\lr }{\left(}
6: \newcommand{\rl }{\right)}
7: \newcommand{\lan }{\langle}
8: \newcommand{\ran }{\rangle}
9: \newcommand{\SRa}{ 
10: \put(0,6) { \vector(1,0){10}  }
11: \put(10,0){ \vector(-1,0){10} } 
12: \qquad}
13: \newcommand{\SRb}{ 
14: \put(0,10) { \vector(1,-1){10}  }
15: \put(10,10){ \vector(-1,-1){10} } 
16: \qquad}
17: 
18: \begin{document}
19: 
20: \title{Spectral Form Factor for Chaotic Dynamics \\ 
21: in a Weak Magnetic Field}
22: 
23: \author{Keiji Saito {\footnote {\tt email: 
24: saitoh@spin.phys.s.u-tokyo.ac.jp}}~~and 
25: Taro Nagao {\footnote {\tt email: 
26: nagao@math.nagoya-u.ac.jp}}  
27: }
28: \date{}
29: \maketitle
30: 
31: \begin{center}
32: \it
33: $^*$ Department of Physics, Graduate School of Science,
34: University of Tokyo, Hongo 7-3-1, Bunkyo-ku, Tokyo 113-0033, Japan
35: \end{center}
36: \begin{center}
37: \it
38: $^\dag$ Graduate School of Mathematics,
39: Nagoya University, Chikusa-ku, \\ Nagoya 464-8602, Japan
40: \end{center}
41: 
42: \begin{abstract}
43: Using semiclassical periodic orbit theory for a chaotic system in a
44: weak magnetic field, we obtain the form factor predicted 
45: by Pandey and Mehta's two matrix model up to the third order. The third
46: order contribution has a peculiar term which exists only in the 
47: intermediate crossover domain between the GOE (Gaussian Orthogonal 
48: Ensemble) and the GUE (Gaussian Unitary Ensemble) universality classes. 
49: The exact expression is obtained by taking account of the contribution 
50: from encounter regions where orbit loops are connected.  
51: \end{abstract}
52: 
53: PACS: 05.45.+b; 05.40.+j
54: 
55: \medskip
56: 
57: KEYWORDS: quantum chaos; periodic orbit theory; random matrices
58: 
59: \newpage
60: 
61: \section{Introduction}
62: 
63: Universal level statistics of classically chaotic systems 
64: in agreement with the RMT (Random Matrix Theory) prediction 
65: has been intensively studied since it was conjectured two 
66: decades ago\cite{BGS84}.
67:  
68: Berry\cite{B85} first evaluated the first order term of the 
69: spectral form factor using the semiclassical periodic orbit 
70: theory\cite{GW}. His method is called the diagonal approximation. 
71: It is applicable to both of the GOE (Gaussian Orthogonal Ensemble) 
72: and GUE (Gaussian Unitary Ensemble) universality classes.
73: In the absence of time reversal invariance, the first order 
74: term is evaluated as a sum over the periodic orbits and agrees 
75: with the prediction derived from the GUE of random matrices. 
76: If the system is time reversal invariant, we additionally need 
77: to sum up over the pairs of the periodic orbits mutually 
78: time reverse. The result is also in agreement with the RMT 
79: prediction derived from the GOE.
80: 
81: Recently Sieber and Richter (SR) showed that a family of orbit pairs 
82: with an encounter gave the first correction to the diagonal approximation
83: \cite{SR01}. Their idea clarifies what corresponds to the {\em order} 
84: of each term in the context of the periodic orbit theory. Heusler et al. 
85: \cite{HMBH04} further generalized SR's idea by using a combinatorial 
86: analysis and M\"uller et al.\cite{MHBHA04,MHBHA05} obtained the 
87: full expansion which yields the RMT form factor.
88: 
89: In this paper, we investigate the form factor of chaotic 
90: systems in a weak magnetic field, extending the method 
91: of Ref.\cite{HMBH04}. Our aim is to reproduce 
92: the universal level statistics in the crossover domain 
93: between the GOE and GUE universality classes. The corresponding 
94: RMT formula is derived from Pandey and Mehta's two matrix 
95: model\cite{PM83}. This formula interpolates the form 
96: factors of GOE and GUE, which correspond 
97: to chaotic systems without and with a magnetic field, respectively.
98: Using a quantum graph, the authors previously obtained the second order 
99: term by employing Sieber and Richter's periodic orbit pairs and 
100: showed that it agreed with the RMT prediction\cite{NS03}. 
101: Here we consider the third order term of the 
102: form factor and clarify a mechanism to yield the RMT
103: expression. We show that the RMT expression is obtained 
104: by taking account of the interference of the gauge 
105: potential in the encounters of periodic orbits. 
106: 
107: Let us explicitly present the RMT prediction of the form factor. 
108: It is derived from Pandey and Mehta's two matrix 
109: model, as explained in\cite{NS03}. For small $\tau 
110: (0\le \tau \le {1\over 2})$, the RMT form factor $K_{\rm RM}(\tau)$ 
111: with a parameter $\mu$ is written as
112: \beq
113: K_{\rm RM} (\tau ) &=& \tau + {1\over 2} 
114: \int_{1-2\tau }^{1}  dk \, {k \over k + 2\tau }
115: e^{-\mu(k + \tau)}  
116: \nonumber \\
117: &=&  \tau + \tau e^{-\mu} - 2\tau^2 e^{-\mu} 
118: + \tau^3 \left( 
119: {\mu^2 \over 6} e^{-\mu} + 
120: 2 e^{-\mu} \right) + \cdots . \label{pd}
121: \eeq
122: 
123: The GOE and GUE limits correspond to $\mu\to0$ and 
124: $\mu\to\infty$, respectively. Up to the second order of 
125: $\tau$, each term is a monotonic function of the parameter 
126: $\mu$. However, the third order term in (\ref{pd}) 
127: includes a peculiar nonmonotonic term $\tau^3 
128: {\mu^2 \over 6} e^{-\mu}$, which vanishes in both of the GOE 
129: and GUE limits. Such terms generally appear in higher order 
130: contributions. We can thus expect that a semiclassical 
131: analysis of the third order term will offer a basis to 
132: reproduce the full expansion of $K_{\rm RM}(\tau)$. 
133: 
134: \section{Chaotic System in a Magnetic Field and Periodic Orbit Theory}
135: 
136: We consider a quantum system in a magnetic field and 
137: assume that the corresponding classical dynamics is chaotic. 
138: In the semiclassical limit $\hbar \rightarrow 0$, the form factor 
139: of such a system can be 
140: formally written in terms of the classical periodic 
141: orbits\cite{GW};
142: \beq
143: K_{\rm po}(\tau ) =
144: \left\langle
145: \sum_{\gamma , \gamma'} A_{\gamma}A_{\gamma'}^{\ast} e^{i( S_{\gamma}  
146: + \theta_{\gamma})/\hbar }
147: e^{ -i( S_{\gamma'} + \theta_{\gamma'} )/\hbar }
148: \delta \left( \tau - {T_{\gamma} + T_{\gamma'} \over 2T_{\rm H}}\right)
149: \right\rangle ,
150: \label{fmr}
151: \eeq
152: where $S_{\gamma}$, $T_{\gamma}$ and $A_{\gamma}$ are 
153: the action without gauge potential 
154: (including the Maslov phase), period and 
155: dimensionless stability amplitude of the 
156: periodic orbit $\gamma$. The phase $\theta_{\gamma}$ is 
157: an effect of the gauge potential in the action along 
158: the orbit $\gamma$, i.e., 
159: $\theta_{\gamma}=
160:  \int_{\gamma} d{\bf x}_{\gamma}    \cdot {\bf a}({\bf x}_{\gamma}) 
161: = \int_{\gamma} dt \, d{\bf x}_{\gamma}/dt \cdot {\bf a}({\bf x}_{\gamma})$, 
162: using the gauge potential ${\bf a}({\bf x}_{\gamma})$ at the 
163: position ${\bf x}_{\gamma}$.
164: The angular bracket stands for an average over 
165: the mean energy $(E+E')/2$ and over a time interval much smaller than 
166: the Heisenberg time $T_{\rm H}$. Since the classical dynamics 
167: is chaotic, each periodic orbit shows a diffusive 
168: behavior\cite{TR03,R_text}; successive changes of the velocity 
169: can be regarded as independent events. Hence, if the time $T$ 
170: elapsed on an orbit $\gamma$ is sufficiently large, we assume 
171: that the term $g_{\gamma} (t) \equiv d{\bf x}_{\gamma}/dt 
172: \cdot {\bf a}({\bf x}_{\gamma}) $ can be replaced by the 
173: Gaussian white noise satisfying the correlation 
174: $\lan \lan g_{\gamma} (t) g_{\gamma }(t') \ran \ran 
175: = 2D \delta (t-t')$. Here a Gaussian average $\lan\lan 
176: \cdots\ran\ran$ is computed as a functional integral 
177: \beq 
178: \label{gaussian}
179: \lan \lan \ F[g_{\gamma}] \ \ran \ran = \frac{\displaystyle 
180: \int {\cal D}g_{\gamma} \ {\rm exp}\left[ 
181: -\frac{1}{4 D} \int_0^T dt (g_{\gamma}(t))^2 
182: \right] F[g_{\gamma}]} 
183: {\displaystyle \int {\cal D}g_{\gamma} \  
184: {\rm exp}\left[ -\frac{1}{4 D} \int_0^T dt (g_{\gamma}(t))^2 
185: \right]}.
186: \eeq
187: Incorporating such a Gaussian average, we obtain a formula 
188: \beq
189: K_{\rm po} (\tau ) = 
190: \left\langle
191: \sum_{\gamma , \gamma'} A_{\gamma}A_{\gamma'}^{\ast}
192: e^{i( S_{\gamma}  -S_{\gamma'} )/\hbar }
193: \lan\lan e^{ i( \theta_{\gamma} - \theta_{\gamma'} )/\hbar} \ran\ran
194: \delta \left( \tau - {T_{\gamma} + T_{\gamma'} \over 2T_{\rm H}}\right)
195: \right\rangle . \label{pobs}
196: \eeq
197: We start from Berry's diagonal approximation, which yields 
198: the first order term\cite{B85}. This term is obtained from 
199: two types of periodic orbit pairs 
200: $({\gamma,\gamma})$ and $({\gamma, \overline{\gamma}})$. Here 
201: a periodic orbit $\overline{\gamma}$ is the time reverse of $\gamma$. 
202: For the pairs of identical orbits $({\gamma,\gamma})$, the dependence 
203: on the gauge potential is canceled, so that the result does not 
204: depend on the magnetic field. The estimate for the form 
205: factor is given by the Hannay and Ozorio de Almeida (HOdA)'s 
206: sum rule\cite{HODA84};
207: \beq
208: \sum_{\gamma}\left| A_{\gamma}\right|^2
209: \delta \lr \tau - {T_{\gamma}\over T_{H}}\rl 
210: = \tau.
211: \eeq 
212: For the pairs of mutually time reverse orbits $({\gamma,\overline{\gamma}})$, 
213: on the other hand, the result depends on the magnetic field. 
214: It is estimated as 
215: \beq
216: \sum_{\gamma}\left| A_{\gamma}\right|^2
217: \delta \lr \tau - {T_{\gamma}\over T_{H}}\rl 
218: \lan\lan e^{ i( \theta_{\gamma} - \theta_{\overline{\gamma}})/\hbar} \ran\ran
219: =\tau \lan\lan e^{2i \theta_{\gamma} /\hbar } \ran\ran 
220: = \tau e^{-bT},
221: \eeq
222: where the last equality results from the 
223: Gaussian average (\ref{gaussian}). 
224: Here $T$ is the period $T(=\tau T_{H})$ and $b$ 
225: is defined in terms of the diffusion constant $D$ 
226: as $b = 4D/\hbar^2$, which depends on the magnetic field. 
227: Putting the above results together, we reproduce the 
228: first order term of the form factor 
229: \cite{BGOS95} as
230: \beq
231: K_{\rm po}^{(1)} &=& \tau + \tau e^{-bT},
232: \eeq
233: which reproduces the first two terms of the 
234: RMT prediction (\ref{pd}). The RMT parameter 
235: $\mu$ is identified with $b T$. 
236: 
237: We next consider the second order term. 
238: As shown by Sieber and Richter (SR) \cite{SR01}, a family of orbit pairs 
239: with one encounter plays a crucial role\cite{NS03,TR03}.
240: We can symbolically write SR pairs as $\gamma=E L_1 \overline{E} L_2$ and 
241: $\gamma'=E \overline{L}_1 \overline{E} L_2$ (see Figure 1). Here $E$ and its almost 
242: time reverse $\overline{E}$ are the orbit segments in the encounter where two 
243: loops are connected. The loops are denoted as $L_1$ and $L_2$, 
244: respectively, and $\overline{L}_1$ is the time reverse of $L_1$. 
245: Let us see the behavior of the periodic orbits $\gamma$ and 
246: $\gamma'$ in the encounter region. The directions of encounter 
247: orbits are depicted as $\SRa$ for 
248: $\gamma$ and $\SRb$ for $\gamma'$. For a system with two 
249: degrees of freedom, it is convenient to use the 
250: Poincar\'e section ${\cal P}$ in the phase space 
251: \cite{HMBH04}. The orbit $\gamma$ pierces through 
252: $\cal P$ at two phase space points denoted by vectors 
253: $x_a$ and $x_b$. Let us express 
254: ${\cal T}x_b - x_a$ (${\cal T}$ is the time reversal 
255: operator) in terms of the pairwise normalized 
256: vectors $\hat{e}_{s}$ and $\hat{e}_{u}$ as $
257: {\cal T}x_b - x_a = u \hat{e}_{u} + s \hat{e}_{s}$.
258: To demand that the orbits in the encounter are mutually 
259: close, we introduce a bound $c$ and assume that $|u|,|s| 
260: < c$. Then the partner orbit $\gamma'$ (or the 
261: time reverse of $\gamma'$) pierces at $x_a^p = 
262: x_a + u \hat{e}_{u}$ and at $x_b^p= {\cal T}(x_a 
263: + s\hat{e}_{s})$. We can consequently estimate 
264: the duration $t_{\rm enc}$ of the encounter 
265: and the action difference $\Delta S$ for the 
266: two orbits as
267: \beq
268: t_{\rm enc} = 
269: {1\over \lambda} \ln {c^{2}\over |us|} 
270: \qquad {\rm and}\qquad 
271: \Delta S = us, 
272: \eeq
273: where $\lambda$ is the Liapunov exponent. 
274: 
275: We need to estimate the number of encounters 
276: in one periodic orbit of a period $T$. 
277: This can be computed as
278: \beq
279: \int_{-c}^c du ds 
280: \int_{0}^{T -2t_{\rm enc}} d T_1
281: {T \over {\cal N} t_{\rm enc} \Omega},
282: \label{enb} 
283: \eeq
284: where $\Omega^{-1} (={1\over 2\pi\hbar T_H})$ 
285: is the ergodic return probability per a unit action. 
286: The factor $T \int_{0}^{T - 2 t_{\rm enc}} {\rm d}T_1$ 
287: indicates that one of the two piercings occurs in the 
288: time interval $[0,T]$ and that the time $T_1$ elapsed 
289: on the loop $L_1$ or $\overline{L}_1$ lies in $[0,T-2 t_{\rm enc}]$.
290: The combinatorial factor ${\cal N} = 2$ is 
291: necessary to avoid a double counting due to 
292: the equivalence of $E$ and $\overline{E}$ in 
293: $\gamma$. 
294: 
295: Let us now evaluate the contribution of the gauge potential 
296: to the second order term. The Gaussian average (\ref{gaussian}) 
297: yields a factor $e^{-b T_1}$ as a contribution from the 
298: the pair of the loops $L_1$ and $\overline{L}_1$. In the region of the encounter, 
299: the gauge potential destructively interferes with itself 
300: and yields no contribution. That is, since mutually 
301: almost time reversed orbits are combined in the encounter, 
302: a product of the phase factors $e^{i \theta_{\gamma}}$ 
303: and $e^{-i \theta_{\gamma^{\prime}}}$ 
304: gives a factor $1$. Incorporating the factor $2$ due to the 
305: time reversal of the orbit $\gamma^{\prime}$, we 
306: obtain the second order term as
307: \beq
308: K_{\rm po}^{(2)}(\tau) &=& 2 \sum_{\gamma}\left| A_{\gamma}\right|^2
309: \delta \lr \tau - {T_{\gamma}\over T_{H}}\rl
310: \int_{-c}^{c} du ds
311: \int_{0}^{T -2t_{\rm enc}} d T_1  
312: {T \over {\cal N} t_{\rm enc} \Omega} 
313: e^{-b T_{1}} 
314: e^{i us/\hbar} 
315: \nonumber \\
316: &\rightarrow&-2\tau^2 e^{-b T}, 
317: \eeq
318: again in agreement with (\ref{pd}). Here we calculated only the 
319: term independent of $t_{\rm enc}$, as the other terms vanish 
320: due to appearances of extra factors $\hbar$ or rapid 
321: oscillations\cite{MHBHA05}.
322:  
323: \section{Third Order Term}
324: We now calculate the third order contribution, following  
325: the recipe explained in previous section. We will find 
326: that the calculation is somewhat different from the second 
327: order term because the interference of the gauge 
328: potential in the encounter regions gives a 
329: significant contribution. For the 
330: periodic orbit pairs which contribute 
331: to the third order term, Heusler 
332: et al.\cite{HMBH04} introduced several 
333: notations (aas, api, ppi, ac and pc). 
334: In order to take account of the time 
335: reversal of the orbit $\gamma^{\prime}$, 
336: we modify their notations by putting 
337: suffixes (-a and -b) as drawn in 
338: Figures 2 and 3. 
339: 
340: Let us first consider a general orbit pair 
341: $\alpha$ with $L$ loops and $V$ encounters. 
342: Extending the result (\ref{enb}) for 
343: the second order term, we can compute 
344: the number of encounters in one periodic 
345: orbit of a period $T$ as
346: \beq 
347: \int d{\bf u}d{\bf s} 
348: \int_0^{T-t_{\alpha}} dT_1  
349: \int_0^{T-t_{\alpha}-T_1} dT_2 \cdots 
350: \int_0^{T-t_{\alpha}-T_1-T_2-\cdots-T_{L-2}} dT_{L-1} 
351: \ Q_{\alpha}, 
352: \eeq
353: where
354: \beq
355: Q_{\alpha} = \frac{  T}{{\cal N}_{\alpha} \  
356: \prod_{v=1}^V t_{{\rm enc},v} \ \Omega^{L-V}}. 
357: \eeq
358: The integration measures are $d{\bf u} = \prod_{j=1}^{L-V} 
359: du_j$ and $d{\bf s} = \prod_{j=1}^{L-V} ds_j$ in terms of 
360: appropriate phase space coordinates $(u_j,s_j)$. The times elapsed 
361: on the $L$ loops are depicted as $T_1,T_2,\cdots,T_L$. 
362: The duration of the $v$-th encounter is written as $t_{{\rm enc},v}$. 
363: The total duration of the encounters in one periodic orbit 
364: is $t_{\alpha} = \sum_{v=1}^V l_v t_{{\rm enc},v}$ ($l_v$ 
365: is the number of the orbit segments contained in the $v$-th 
366: encounter). A combinatorial factor ${\cal N}_{\alpha}$ 
367: depends on the structure of the orbit pair. 
368: \par 
369: We then consider the contribution of the gauge potential. 
370: The Gaussian average (\ref{gaussian}) on the loops yields 
371: a factor $e^{-b T_1} e^{-b T_2} \cdots e^{-b T_K}$, 
372: where $K$ is the number of the pairs of mutually 
373: time reversed loops. In Figures, the encounter orbits 
374: are schematically drawn along with the passing 
375: directions through the encounter regions. For the $v$-th 
376: encounter, we introduce the number $n_{{\rm enc},v}$ of 
377: the orbit segment pairs which cause the 
378: significant gauge potential contribution. 
379: Let us fix an arbitrary direction $(+)$ 
380: of the orbit passing and call the opposite 
381: direction $(-)$. Then the number 
382: $n_{{\rm enc},v}$ can be computed as 
383: \beq
384: n_{{\rm enc},v} = \frac{1}{2}
385: \left| \left\{ \#^{(+)}(\gamma)-\#^{(-)}(\gamma) \right\} - 
386:  \left\{ \#^{(+)}(\gamma') - \#^{(-)}(\gamma') \right\} \right|.
387: \eeq
388: Here $\#^{(+)}(\gamma)$ and $\#^{(-)}(\gamma)$ mean the 
389: numbers of passings of $\gamma$ through the encounter 
390: in $(+)$ and $(-)$ directions, respectively. We again compute 
391: the Gaussian average (\ref{gaussian}) to evaluate the effect 
392: of the gauge potential. The contribution from an encounter 
393: is then estimated as $\lan \lan  e^{2 i n_{{\rm enc},v} 
394: \theta_{\gamma}/\hbar} \ran \ran 
395: = e^{- b n_{{\rm enc},v}^2 t_{{\rm enc},v}}$. 
396:    
397: Putting the above results together, we find the 
398: contribution to the form factor from the orbit 
399: pair $\alpha$ as 
400: \beq
401: & & K_{\rm po,\alpha}(\tau) 
402: = \tau \int d{\bf u}d{\bf s} 
403: \nonumber \\ & \times  & 
404: \int_0^{T-t_{\alpha}} dT_1  
405: \int_0^{T-t_{\alpha}-T_1} dT_2 \cdots 
406: \int_0^{T-t_{\alpha}-T_1-T_2-\cdots-T_{L-2}} dT_{L-1} 
407: Q_{\alpha} R_{\alpha} e^{i \Delta S/\hbar} \nonumber \\
408: \label{kpo} 
409: \eeq
410: with
411: \beq
412: R_{\alpha} = e^{-b(T_1 + T_2 + \cdots + T_K)} 
413: e^{-b(n_{{\rm enc},1}^2 t_{{\rm enc},1}  
414:  + n_{{\rm enc},2}^2 t_{{\rm enc},2} + \cdots +   
415: n_{{\rm enc},V}^2 t_{{\rm enc},V})}.
416: \eeq
417: This contributes to the terms of order $L-V+1$. 
418: Here $\tau$ appears due to the HOdA sum rule 
419: and the action difference is estimated as 
420: $\Delta S = \sum_{j=1}^{L-V} u_j s_j$. 
421: 
422: Let us go back to the calculation of the third order 
423: term with $L-V+1=3$. From Figures 2 and 3, 
424: we can readily see the followings. For $\alpha=$ aas,api 
425: and ppi, $L=4$, $V=2$ and $t_{\alpha}=2\lr 
426: t_{\rm enc, 1} + t_{\rm enc, 2}\rl$. On the other hand, 
427: for $\alpha=$ ac and pc, $L=3$, $V=1$ and 
428: $t_{\alpha}=3 t_{\rm enc}$. Thus we can 
429: calculate $R_{\alpha}$ as
430: \beq
431: R_{\rm aas-a}
432: &=& R_{\rm aas-b} = e^{-b\lr T_1 + T_2 \rl} , \\
433: R_{\rm api-a} &=& R_{\rm api-b} = e^{-b\lr T_1 + T_2  
434: + t_{\rm enc, 1}+ t_{\rm enc , 2} \rl} , \\
435: R_{\rm ppi-a} &=& 1 , \\
436: R_{\rm ppi-b} &=& 
437: e^{-b\lr T -t_{\rm ppi}\rl 
438: -4b\lr t_{\rm enc , 1}+ t_{\rm enc ,2}\rl } , \\
439: R_{\rm ac-a} &=& e^{-b\lr T_1 + T_2 \rl} , \\
440: R_{\rm ac-b} &=& e^{-b\lr T_1 + t_{\rm enc} \rl} , \\
441: R_{\rm pc-a} &=& 1 , \\
442: R_{\rm pc-b} &=& e^{-b\lr T -t_{pc} \rl - 9 b t_{\rm enc}} . 
443: \eeq
444: Moreover the combinatorial factor ${\cal N}_{\alpha}$ 
445: is evaluated in \cite{HMBH04} as
446: \beq 
447: ({\cal N}_{\rm aas}, {\cal N}_{\rm api}, {\cal N}_{\rm ppi},
448: {\cal N}_{\rm ac}, {\cal N}_{\rm pc})=(2,2,4,1,3).
449: \eeq
450: 
451: We substitute these formulas into (\ref{kpo}) and 
452: again extract the terms independent of the encounter 
453: durations $t_{{\rm enc} ,1}$, $t_{{\rm enc} ,2}$ and 
454: $t_{\rm enc}$. After a straightforward calculation, 
455: we obtain contributions from the orbit pairs
456: \beq
457: K_{\rm po,aas-a}(\tau ) &=& K_{\rm po, aas-b}(\tau ) = 2 \tau^3 e^{-bT}  , \\
458: K_{\rm po,api-a}(\tau ) &=& K_{\rm po, api-b} (\tau ) = 
459: \tau^3 \left[ 
460: -{e^{-bT} -1 \over bT} + {1 + e^{-bT} \over 2}
461: \right]  , \\
462: K_{\rm po,ppi-a}(\tau ) &=&  \tau^3 ,\\
463: K_{\rm po,ppi-b}(\tau ) &=& \tau^3 e^{-bT} \left[ 
464: {(bT)^2 \over 6} + bT + 1 \right] ,\\
465: K_{\rm po,ac-a}(\tau ) &=& -3\tau^3 e^{-bT} ,\\
466: K_{\rm po,ac-b}(\tau ) &=&  \tau^3 \left[ 
467: -1 + {2\lr e^{-bT} - 1 \rl \over bT}
468: \right] ,\\
469: K_{\rm po,pc-a}(\tau ) &=& -\tau^3 , \\
470: K_{\rm po,pc-b}(\tau ) &=& -\tau^3 e^{-bT} \lr  1 + bT\rl .
471: \eeq
472: Summing up these values, we arrive at the third order term of 
473: the form factor
474: \beq
475: K_{po}^{(3)}(\tau ) = \sum_{\alpha}  K_{\rm po, \alpha}(\tau ) =
476:  \tau^3 \left( 
477: {(bT)^2 \over 6} e^{-bT} + 
478: 2 e^{-bT} \right),
479: \eeq
480: which agrees with the RMT prediction (\ref{pd}). One might think that 
481: $K_{\rm po,api}$ and $K_{\rm po,ac}$ should disappear in the 
482: limit $b\to\infty$, because a pair of mutually time reversed 
483: loops exists. However, as these contributions come from the 
484: limiting situations that the times elapsed on those loops are 
485: short, there is no inconsistency. 
486: 
487: In summary, we investigate the level statistics of 
488: a classically chaotic system in the crossover domain 
489: between GUE and GOE universality classes. Summing up 
490: the contributions from the relevant periodic orbit pairs 
491: of a chaotic system, we find an agreement between the 
492: semiclassical form factor and the RMT result up 
493: to the third order. We expect that this agreement 
494: holds up to any arbitrary order. The periodic orbit pairs 
495: consist of loops and encounters. The durations of 
496: the encounters are logarithmically divergent 
497: in the limit $\hbar \rightarrow 0$ but 
498: infinitesimally small compared to the orbit 
499: periods, which are of the order of the 
500: Heisenberg time $T_H$ ($\sim O(1/\hbar)$ 
501: for a system with two degrees 
502: of freedom)\cite{HMBH04}. Nevertheless, 
503: the contribution from the encounter regions 
504: is crucial. This mechanism should also be 
505: a key factor in the derivation of higher 
506: order terms. It might moreover shed light 
507: on a similar counting problem\cite{BSW03} 
508: for a quantum graph.  
509: 
510: \section*{Acknowledgement}
511: 
512: One of the authors (T.N.) is grateful to Dr. Gregory Berkolaiko, 
513: Dr. Petr Braun, Dr. Sebastian M\"uller and Dr. Holger Schanz 
514: for valuable discussions.
515: 
516: \begin{thebibliography}{notitle}
517: 
518: \bibitem{BGS84}
519: O. Bohigas, M.J. Giannoni and C. Schmit, Phys. Rev. Lett. {\bf 52} (1984) 1. 
520: \bibitem{B85}
521: M.V. Berry, Proc. R. Soc. London {\bf A400} (1985) 229.
522: \bibitem{GW}
523: M. Gutzwiller, {\em Chaos in Classical and Quantum Mechanics}
524: (Springer, 1990).
525: \bibitem{SR01}
526: M. Sieber and K. Richter, Physica Scripta {\bf T90} (2001) 128.
527: \bibitem{HMBH04}
528: S. Heusler, S. M\"{u}ller, P. Braun and F. Haake, 
529: J. Phys. {\bf A37} (2004) L31. 
530: \bibitem{MHBHA04}
531: S. M\"{u}ller, S. Heusler, P. Braun, F. Haake and A. Altland, 
532: Phys. Rev. Lett. {\bf 93} (2004) 014103-1. 
533: \bibitem{MHBHA05}
534: S. M\"{u}ller, S. Heusler, P. Braun, F. Haake and A. Altland, 
535: nlin.CD/0503052.
536: \bibitem{PM83}
537: A. Pandey and M.L. Mehta, Commun. Math. Phys. {\bf 87} (1983) 449.
538: \bibitem{NS03}
539: T. Nagao and K. Saito, Phys. Lett. {\bf A311} (2003) 353.
540: \bibitem{TR03}
541: M. Turek and K. Richter, J. Phys. {\bf A36} (2003) L455.
542: \bibitem{R_text}
543: K. Richter, {\em Semiclassical Theory of Mesoscopic Quantum Systems}
544: (Springer, 2000).
545: \bibitem{HODA84}
546: J.H. Hannay and A.M. Ozorio de Almeida, J. Phys. {\bf A17} 
547: (1984) 3429.
548: \bibitem{BGOS95}
549: O. Bohigas, M.-J. Giannoni, A.M. Ozorio de Almeida and C. Schmit, 
550: Nonlinearity {\bf 8} (1995) 203. 
551: \bibitem{BSW03}
552: G. Berkolaiko, H. Schanz and R.S. Whitney, J. Phys. {\bf A36} 
553: (2003) 8373.
554: 
555: \end{thebibliography}
556: 
557: \newpage
558: 
559: \begin{figure}[!t]
560: \epsfxsize=14cm
561: \centerline{\epsfbox{sr.eps}}
562: \caption{The Sieber-Richter (SR) pair of periodic orbits}
563: \end{figure}
564: 
565: \begin{figure}[!t]
566: \epsfxsize=14cm
567: \centerline{\epsfbox{two.eps}}
568: \caption{The periodic orbit pairs with two encounters.}
569: \end{figure}
570: 
571: \begin{figure}[!t]
572: \epsfxsize=14cm
573: \centerline{\epsfbox{one.eps}}
574: \caption{The periodic orbit pairs with one encounter.}
575: \end{figure}
576: 
577: 
578: \end{document}
579: 
580: 
581: