1: \rcsInfo $Id: skew.tex,v 3.1 2001/06/26 11:37:38 guest Exp guest $
2: \section{Ergodic skew maps}
3: \label{sec:skew-map}
4: Another interesting class of parabolic maps on the two-torus~$\TT$ are
5: the so-called skew maps. We will consider skew maps of the form
6: \begin{equation}
7: \label{eq:skew-map}
8: \begin{array}{ccccc}
9: \Skew:&\TT&\longrightarrow&\TT& \\[1ex]
10: &
11: \begin{pmatrix}
12: p \\
13: q \\
14: \end{pmatrix}
15: &\longmapsto&
16: \begin{pmatrix}
17: p+\beta \\
18: q+2p \\
19: \end{pmatrix}
20: &\pmod 1 \ .
21: \end{array}
22: \end{equation}
23: For $\beta\in\Sone\smallsetminus\Q$ one has again a uniquely ergodic but
24: non-mixing map, for details see~\cite{Fur61,AbrRoh65,CouHam97}.
25: If~$\beta$ is rational the map is in general almost integrable. This
26: situation will be discussed in the following section, and is our main
27: motivation for also analysing skew maps in addition to the simplest
28: case of parabolic maps, namely translations of the torus. As in the
29: case of Kronecker maps symmetries play an important r\^ole. Skew maps
30: also have several symplectic symmetries. First of all there is a
31: translation in momentum
32: \begin{equation}
33: \label{eq:s-momentum-symmetry}
34: T_p:
35: \begin{pmatrix}
36: p \\
37: q \\
38: \end{pmatrix}
39: \longmapsto
40: \begin{pmatrix}
41: p+\frac{1}{2} \\
42: q \\
43: \end{pmatrix}
44: \ \pmod 1
45: \ ,
46: \end{equation}
47: and a family of translations in position
48: \begin{equation}
49: \label{eq:s-position-symmetry}
50: T_q:
51: \begin{pmatrix}
52: p \\
53: q \\
54: \end{pmatrix}
55: \longmapsto
56: \begin{pmatrix}
57: p \\
58: q+\gamma \\
59: \end{pmatrix}
60: \ \pmod 1
61: \ ,\quad \gamma\in\Sone
62: \ .
63: \end{equation}
64: Moreover, skew maps are invariant under two anti-symplectic
65: transformations,
66: \begin{equation}
67: \label{eq:s-anti-symplectic-symmetry}
68: \tau_{\phantom{p}}:
69: \begin{pmatrix}
70: p \\
71: q \\
72: \end{pmatrix}
73: \longmapsto
74: \begin{pmatrix}
75: -p+\delta \\
76: q \\
77: \end{pmatrix}
78: \ \pmod 1
79: \ ,\quad 2\delta\equiv2\beta \bmod 1
80: \ ,
81: \end{equation}
82: Thus, in contrast to Kronecker maps, we now have a (non-conventional)
83: time reversal symmetry.
84:
85: \begin{figure}[!t]
86: \subfigure[The level spacing distribution for a skew map is shown.
87: The distribution is obviously non generic.
88: The level spacing density is given by
89: $P(s)=\nicefrac{1}{3}(\,\delta(s)+\delta(s-1)+\delta(s-2)\,)$,
90: see~\cite{BacHaa99}.] {\includegraphics[width=0.49\textwidth]
91: {levelspacing-skew-ergodic}\label{fig:skew-ergodic-levelspacing}}
92: \subfigure[The form factor has peaks for
93: $\tau=\nicefrac{k}{D}=\nicefrac{k}{3}$, $k\in\N$. For multiples of
94: the Heisenberg time $\THB=N$ one gets peaks of the height $N$. The
95: average for $\tau>1$ gives a mean degeneracy for the eigenangles of
96: $\nicefrac{5}{3}$.] {\includegraphics[width=0.49\textwidth]
97: {formfactor-skew-ergodic}\label{fig:skew-ergodic-formfactor}}
98: \caption{Spectral statistics for a skew map with
99: $\beta=\nicefrac{(\sqrt{5}-1)}{2}$ and $N=3000$, implying
100: $D=\gcd(b,N)=3$ and $M=\nicefrac{N}{D}=1000$.}
101: \label{fig:skew-ergodic-statistics}
102: \end{figure}
103:
104: The first quantisation of skew maps, which fulfils an Egorov property
105: for all $N\in\N$, cf.~\eqref{eq:k-egorov}, was given by Marklof and
106: Rudnik~\cite{MarRud00}. This quantisation
107: is similar to the quantisation of the Kronecker map which we presented
108: in section~\ref{sec:quant-mech-kron}. First of all one has to
109: approximate the translation in momentum by a rational number
110: $\nicefrac{b}{N}$ with
111: \begin{equation}
112: \label{eq:s-approximation}
113: \abs{\beta-\fracN[b]}\leq\frac{1}{2N} \ , \quad b \in \Z
114: \ .
115: \end{equation}
116: The propagator $\op{U}(\Skew)$ can then be defined similarly
117: to~\eqref{eq:van-vleck} by using a generating function
118: $G(q_{n+1},p_{n})$. In mixed representation one obtains
119: \begin{equation}
120: \label{eq:s-propagator-mixed}
121: \braopket{Q}{\op{U}(\Skew)}{P}
122: =\frac{1}{\sqrt{N}}
123: \exp\left(\zpi N \, G_{\fracN[b]}
124: \left(\fracN[Q],\fracN[P]\right) \right)
125: =\frac{1}{\sqrt{N}} \,
126: \ue^{\fracN[\zpi]\left(-P^2+Q(P+b)\right)}
127: \ ,
128: \end{equation}
129: and in momentum representation $\op{U}(\Skew)$ is simply given by
130: \begin{equation}
131: \label{eq:s-propagator-momentum}
132: \braopket{P}{\op{U}(\Skew)}{P^\prime}
133: = \ue^{-\fracN[\zpi]P^{\prime 2}}\ \delta_{P,P'+b} \ .
134: \end{equation}
135: For this propagator the Egorov property was proven in~\cite{MarRud00}.
136: Again the quantisation does not preserve all classical symmetries.
137: For our purposes it is not necessary to give a complete list of all
138: quantum symmetries, instead we only remark that there are always
139: unitary quantum symmetries given by translations in position
140: \begin{equation}
141: \label{eq:s-qm-symmetrie-position}
142: \op{T}\left(\frac{k}{D},0\right)
143: \ ,\quad\text{with}\quad D\defin\gcd(b,N)
144: \ ,\quad k\in\ZN[D]
145: \ ,
146: \end{equation}
147: and an anti-unitary symmetry~$\greek{t}\defin\Op(\tau)$
148: involving a translation in
149: momentum
150: \begin{equation}
151: \label{eq:s-qm-symmetrie-momentum}
152: \left[\greek{t}\psi\right](P)=\overline{\psi\left(-P+\fracN[b]\right)}
153: \ .
154: \end{equation}
155:
156: \begin{figure}[!t]
157: \subfigure[The level spacing distribution agrees with the
158: CUE.]
159: {\includegraphics[width=0.49\textwidth]
160: {levelspacing-skew-spin-ergodic}
161: \label{fig:skew-spin-ergodic-levelspacing}}
162: \subfigure[Smoothed form factor, where we averaged over
163: $\Delta\tau=0.2$.] {\includegraphics[width=0.49\textwidth]
164: {formfactor-skew-spin-ergodic-mean}
165: \label{fig:skew-spin-ergodic-formfactor}}
166: \caption{Spectral statistics for a skew map with spin~$\nicefrac{1}{2}$.
167: The parameters are identical to those
168: chosen in figure~\ref{fig:skew-ergodic-statistics}, \ie
169: $N=3000$, $\beta=\nicefrac{(\sqrt{5}-1)}{2}$, $D=3$
170: and $M=1000$.}
171: \label{fig:skew-spin-ergodic-statistics}
172: \end{figure}
173:
174: As for Kronecker maps one can calculate the eigenvalues and
175: eigenvectors of the propagator~\eqref{eq:s-propagator-mixed}
176: explicitly~\cite{MarRud00}. Analysing the spectrum of skew maps one
177: observes non-generic spectral statistics determined by the value of
178: $D$ defined in~\eqref{eq:s-qm-symmetrie-position}. Furthermore, there is
179: even no limit distribution~\cite{BacHaa99}. The non-generic behaviour
180: of the spectral statistics of a quantum skew map is illustrated in
181: figure~\ref{fig:skew-ergodic-statistics}. For a semiclassical analysis
182: we again use a trace formula. Details of the derivation can be found
183: in appendix~\ref{app:skew-trace}. With~\eqref{eq:app-trace-skew-spin-7},
184: see also~\eqref{eq:app-trace-2}, we obtain for the spectral density
185: \begin{equation}
186: \label{eq:skew-trace-formula}
187: d(\vartheta)
188: =\frac{N}{2\pi}
189: +\frac{\sqrt{N}}{2\pi}\sum_{k\in\Z\smallsetminus\{0\}} \ \sum_{\nu\in\Z_{2k}}
190: \frac{M}{\sqrt{2\ui kM}}\,
191: \ue^{\frac{\zpi}{N}kM\vartheta} \,
192: \ue^{\zpi N \fancyS^{kM}_\nu}
193: \ ,
194: \end{equation}
195: with $M\defin\nicefrac{N}{D}$.
196: The periods of periodic orbits of the approximating map
197: are given by $kM$,
198: and $\nu$ labels families of periodic points sharing the same action.
199: These families form invariant manifolds
200: $\E_{\nicefrac{\nu}{2kM}}(\nicefrac{b}{N})$,
201: see~\eqref{eq:disjoint-circles} below, which in particular are
202: invariant under translations in position as is the classical map.
203: Again we have high degeneracies in both the quantum spectrum and in
204: the the trace formula which are the origin of the non-generic spectral
205: statistics of the quantum map (see
206: figure~\ref{fig:skew-ergodic-statistics}).
207:
208: As in the case of Kronecker maps we couple the quantised skew map
209: to a spin~$\nicefrac{1}{2}$ in order to break these symmetries. As we
210: pointed out above the skew map also has symmetries in both position
211: and momentum. Therefore, we use the same magnetic field~$\vecB$ as for
212: the Kronecker map~\eqref{eq:B-Felder}, depending on both~$q$ and~$p$.
213: The same calculation as in section~\ref{sec:maps-with-spin}
214: then yields the Egorov property for skew maps with spin.
215:
216: In appendix~\ref{app:skew-trace} we also derive a semiclassical
217: trace-formula for the skew map with spin.
218: With~\eqref{eq:app-trace-skew-spin-7} the spectral density reads,
219: \begin{equation}
220: \label{eq:skew-spin-trace-formula}
221: d(\vartheta)
222: \sim\frac{N}{\pi}
223: +\frac{\sqrt{N}}{2\pi}\sum_{k\in\Z\smallsetminus\{0\}} \
224: \sum_{\nu\in\Z_{2k}}
225: \frac{ \ue^{\frac{\zpi}{N}kM\vartheta} \,
226: \ue^{2\pi\ui N\fancyS^{kM}_\nu}}
227: {\sqrt{2\ui kM}} \sum_{j\in\Z_M}
228: \int_0^1
229: \tr d_{kM} \!\left(q,\tilde{p}^{(\nu)}+\frac{j}{M}\right) \intd q
230: \ ,
231: \end{equation}
232: where now in contrast to~\eqref{eq:skew-trace-formula} the different
233: periodic orbits of one family, labelled by~$\nu$, enter with different
234: spin weights~$\tr d$. Hence the
235: magnetic field again breaks the degeneracies in the periodic orbit
236: sum~\eqref{eq:skew-spin-trace-formula}, but does not change the
237: classical dynamics on the torus in the sense that the set of periodic
238: orbits remains the same and only their amplitudes are altered.
239:
240: In figure~\ref{fig:skew-spin-ergodic-levelspacing} we show the
241: level spacing distribution and the form factor for an
242: ergodic skew map with spin~$\nicefrac{1}{2}$. As in the case of the
243: Kronecker map we observe good agreement with the CUE. Hence, we have a
244: further example of a system which is only ergodic, but obeys the
245: BGS conjecture.
246:
247: By using the same arguments as in
248: section~\ref{sec:trace-formula-with-spin} and the semiclassical trace
249: formula~\eqref{eq:skew-spin-trace-formula} we conclude that for the
250: form factor of ergodic skew maps
251: \begin{equation}
252: \label{eq:s-form-factor-small-tau}
253: \lim_{\substack{N\to\infty\\\tau\to0}}K(\tau)=0
254: \end{equation}
255: holds. But again, as in the case of Kronecker maps, the behaviour
256: of $K(\tau)$ for small but non-zero~$\tau$ cannot simply be determined
257: by a semiclassical analysis.
258:
259: %%% Local Variables:
260: %%% mode: latex
261: %%% TeX-master: "main"
262: %%% End:
263: