1: The Wigner function (or quasi-distribution density) $ W_\psi : \R^2 \to \R $
2: corresponding to a state vector $ \psi $ has several equivalent (sometimes, up
3: to a coefficient) definitions. The `tomographic' definition (see also
4: \cite{BZ}, Sect.~6.2, Th.~6.1), stipulating that $ W_\psi $ returns correct
5: one-dimensional distributions, is based on the equality
6: \begin{equation}\label{tomo}
7: \iint f(aq + bp) \, W_\psi(q,p) \, \D q \D p = \langle \psi | f(aQ + bP) | \psi
8: \rangle
9: \end{equation}
10: for all $ a,b \in \R $ and all bounded measurable functions $ f : \R \to \R
11: $. However, there is a catch: $ W_\psi $ need not be integrable, that is, $
12: \iint | W_\psi(q,p) | \, \D q \D p = \infty $ for some $ \psi $ (for example,
13: $ \psi(q) = (q^2+1)^{-1/3} \E^{\I q^3} $ in the Schr\"odinger
14: representation). Thus, even the most well-known relation $ \iint W_\psi(q,p)
15: \, \D q \D p = 1 $ needs a careful interpretation!
16:
17: One may interprete the left-hand side of \eqref{tomo} as
18: \[
19: \lim_{\eps\to0+} \iint f(aq + bp) \, W_\psi(q,p) h((bq-ap)\eps) \, \D q \D p
20: \]for an appropriate weight function $ h $ such as
21: \[
22: h(x) = \E^{-x^2} \quad \text{or} \quad h(x) = \begin{cases}
23: 1 - |x| &\text{for $ |x| \le 1 $},\\
24: 0 &\text{otherwise}.
25: \end{cases}
26: \]
27: (It would be natural to put $ h(x) = 1 $ for $ |x| \le 1 $, otherwise $ 0
28: $. However, I do not know, whether this function fits or not.)
29: Treated this way, \eqref{tomo} holds for all $ \psi $.
30:
31: Alternatively, one may characterize the whole map $ \psi \mapsto W_\psi $ by
32: two conditions:
33: \begin{itemize}
34: \item[(a)] there exists a dense set of state vectors $ \psi $ such that $
35: W_\psi $ is integrable and \eqref{tomo} is satisfied;
36: \item[(b)] $ W_{\psi_n} (q,p) \to W_\psi(q,p) $ (as $ n\to\infty $) uniformly
37: in $ q,p $ whenever $ \psi_n \to \psi $.
38: \end{itemize}
39:
40: In the Schr\"odinger representation,
41: \[
42: W_\psi (q,p) = \frac1{2\pi} \int \psi \Big( q + \frac x 2 \Big) \overline{
43: \psi \Big( q - \frac x 2 \Big) } \E^{\I px} \, \D x
44: \]
45: for $ \psi \in L_2(\R) $, and of course,
46: \[
47: (Q\psi) (q) = q \psi(q) \, , \quad (P\psi) (q) = -\I \frac{\D}{\D q} \psi(q)
48: \]
49: for $ \psi $ good enough.
50:
51: We have
52: \[
53: \langle \psi | E(t) | \psi \rangle = \iint \theta( q\cos t + p\sin t ) W_\psi
54: (q,p) \, \D q \D p = \int_{t-\pi/2}^{t+\pi/2} W_\psi^\ang (\phi) \, \D\phi \, ,
55: \]
56: where
57: \[
58: W_\psi^\ang (\phi) = \int_0^\infty W_\psi ( r\cos\phi,r\sin\phi) \, r \D r
59: \]
60: may be called the angular Wigner function. Thus,
61: \begin{equation}\label{1.2}
62: \text{prob}(\psi) = \frac13 \langle \psi | E(0) + E(2\pi/3) +
63: E(4\pi/3) | \psi \rangle = \frac13 \int_{-\pi}^\pi f(\phi) W_\psi^\ang (\phi)
64: \, \D\phi \, ,
65: \end{equation}
66: where $ f(\cdot) $ takes on two values, $1$ and $2$, as follows:
67: \begin{equation}\label{1.3}
68: \begin{gathered}\includegraphics{pic1a.eps}\end{gathered}
69: \end{equation}
70: The inequality $ \frac13 \le \text{prob}(\psi) \le \frac23 $ holds if $
71: W_\psi^\ang (\phi) \ge 0 $ for all $ \phi $, but fails in general.
72:
73: Eigenvectors of the Hamiltonian are the number states $ |0\rangle, |1\rangle,
74: |2\rangle, \dots $;
75: \[
76: (Q^2+P^2) |n\rangle = (2n+1) |n\rangle \, .
77: \]
78: Generally,
79: \begin{gather*}
80: \psi = \sum_{n=0}^\infty c_n |n\rangle \, , \quad \sum_{n=0}^\infty |c_n|^2 =
81: 1 \, , \\
82: W_\psi (q,p) = \sum_{m,n} \bar c_m c_n w_{m,n} (q,p) \, ,
83: \end{gather*}
84: where functions $ w_{m,n} $ are defined by
85: \[
86: \iint f(aq + bp) w_{m,n} (q,p) \, \D q \D p = \langle m | f(aQ + bP) | n
87: \rangle \, .
88: \]
89: In fact,
90: \begin{equation}\label{*}
91: w_{m,n} (q,p) = \frac{(-1)^m}\pi \sqrt{ \frac{m!}{n!} } (2z)^{n-m}
92: \E^{-2|z|^2} L_m^{(n-m)} (4|z|^2)
93: \end{equation}
94: for $ m \le n $ (and $ w_{n,m} = \bar w_{m,n} $); here $ z = (q+\I p)/\sqrt2
95: $, and $ L_m^{(k)} $ is the associated Laguerre polynomial
96: \cite[Sect.~10.12]{BE},
97: \begin{gather*}
98: L_m^{(k)} (x) = (-1)^k \frac{\D^k}{\D x^k} L_{m+k}(x) \, , \\
99: L_n(x) = \frac1{n!} \E^x \frac{\D^n}{\D x^n} (x^n \E^{-x}) = \sum_{k=0}^n
100: (-1)^k \binom n k \frac{x^k}{k!} \, .
101: \end{gather*}
102: Equality \eqref{*} is a combination of \cite[(3.30)]{CG1} and
103: \cite[(4.36)]{CG2}.
104:
105: For the angular Wigner function,
106: \begin{gather*}
107: W_\psi^\ang (\phi) = \sum_{m,n} \bar c_m c_n w_{m,n}^\ang (\phi) \, , \\
108: w_{m,n}^\ang (\phi) = \int_0^\infty w_{m,n} ( r\cos\phi,r\sin\phi) \, r \D r
109: \, .
110: \end{gather*}
111: Using \eqref{*} we get, first,
112: \[
113: w_{m,n}^\ang (\phi) = \E^{\I(n-m)\phi} w_{m,n}^\ang (0)
114: \]
115: and second,
116: \begin{multline}\label{square}
117: w_{m,n}^\ang (0) = \int_0^\infty w_{m,n} (r,0) \, r \D r = \\
118: = \frac{(-1)^m}\pi \sqrt{ \frac{m!}{n!} } \int_0^\infty (\sqrt2 r)^{n-m}
119: \E^{-r^2} L_m^{(n-m)} (2r^2) \, r \D r = \\
120: = \frac{(-1)^{m+n}}\pi \sqrt{m!n!} \sum_{k=\max(m,n)}^{m+n} (-1)^k
121: 2^{k-\frac{m+n}2-1} \frac{ \Gamma(k-\frac{m+n}2+1) }{ (m+n-k)!(k-m)!(k-n)! }
122: \end{multline}
123: (the latter formula holds for all $ m,n $, but the formula with $ L_m^{(n-m)}
124: $ holds for $ m \le n $, of course). Using the generating function
125: \cite[Sect.~10.12, (17)]{BE}
126: \[
127: \sum_{m=0}^\infty L_m^{(k)}(x) t^m = (1-t)^{-k-1} \exp \Big( \! -
128: \frac{xt}{1-t} \Big)
129: \]
130: we get also (for $ m \le n $)
131: \begin{multline*}
132: w_{m,n}^\ang (0) = \\
133: \frac{(-1)^m}{\pi} \sqrt{ \frac{m!}{n!} } 2^{\frac{n-m}2-1}
134: \Gamma \Big( \frac{n-m}2 + 1 \Big) \frac1{m!} \frac{\D^m}{\D t^m} \bigg|_{t=0}
135: \frac1{ (1-t)^{(n-m)/2} (1+t)^{(n-m)/2+1} } \, ,
136: \end{multline*}
137: which leads (by contour integration) to a formula suitable for asymptotic
138: analysis,
139: \begin{multline*}
140: w_{m,n}^\ang (0) = \\
141: \frac{(-1)^m}{\pi} \sqrt{ \frac{m!}{n!} } 2^{\frac{n-m}2-1}
142: \Gamma \Big( \frac{n-m}2+1 \Big) \frac1{2\pi} \int_{-\pi}^\pi
143: \frac{1-r\E^{\I\phi}}{ (1-r^2\E^{2\I\phi})^{(n-m)/2+1} } \frac{\D\phi}{
144: (r\E^{\I\phi})^m } \, ;
145: \end{multline*}
146: it holds for every $ r \in (0,1) $ and is especially useful for $ r=\sqrt{m/n}
147: $. Here are three asymptotic results obtained this way.
148:
149: First, $ \lim_{m\to\infty} w_{m,m+k}^\ang (0) = 1/(2\pi) $ for every $
150: k=0,1,2,\dots $
151:
152: Second, let $ m,n \to \infty $, $ m/n \to c \in (0,1) $. If in addition $ m $
153: remains even, then $ w_{m,n}^\ang (0) \to (2\pi)^{-1} c^{-1/4} $. However, if
154: $ m $ remains odd, then $ w_{m,n}^\ang (0) \to (2\pi)^{-1} c^{1/4} $.
155:
156: Third, let $ m $ be fixed and $ n \to \infty $, then $ w_{m,n}^\ang (0) \sim
157: c_m n^{1/4} $ for some $ c_m $.
158:
159: Returning to \eqref{1.2} we have, first,
160: \begin{multline*}
161: \int f(\phi) W_\psi^\ang (\phi) \, \D\phi = \sum_{m,n} \bar c_m c_n \int
162: f(\phi) w_{m,n}^\ang (\phi) \, \D\phi = \\
163: = \sum_{m,n} \bar c_m c_n w_{m,n}^\ang (0) \int f(\phi) \E^{\I(n-m)\phi} \,
164: \D\phi
165: \end{multline*}
166: for any $ f \in L_2(0,2\pi) $, and second,
167: \begin{equation}\label{squaresquare}
168: \int f(\phi) \E^{\I k\phi} \, \D\phi = \begin{cases}
169: 3\pi & \text{if } k=0 , \\
170: 6/k & \text{if $ (k+3)/12 $ is an integer} , \\
171: -6/k & \text{if $ (k-3)/12 $ is an integer} , \\
172: 0 & \text{otherwise}
173: \end{cases}
174: \end{equation}
175: for the function $ f $ of \eqref{1.3}. The matrix elements of the
176: operator $ \frac13 \( E(0) + E(2\pi/3) + E(4\pi/3) \) $ in the basis of number
177: states are thus calculated,
178: \[
179: \frac13 \langle m | E(0) + E(2\pi/3) + E(4\pi/3) | n \rangle = \frac13
180: w_{m,n}^\ang (0) \int f(\phi) \E^{\I(n-m)\phi} \, \D\phi \, ,
181: \]
182: $ w_{m,n}^\ang (0) $ being given by \eqref{square} and the integral by
183: \eqref{squaresquare}. The spectrum of the corresponding infinite matrix
184: determines the possible values of $ \text{prob}(\psi) $, namely, $ \la_{\min}
185: \le \text{prob}(\psi) \le \la_{\max} $, where $ \la_{\min}, \la_{\max} $ are
186: the least and greatest elements of the spectrum.
187:
188: Restricting ourselves to a finite portion $ |0\rangle, |1\rangle, \dots,
189: |N-1\rangle $ of the basis of number states, we get a finite matrix, $ N
190: \times N $, and can compute its spectrum numerically. The results follow.
191: \[
192: \begin{matrix}
193: N & & 25 & 50 & 100 & 200 & 300 \\
194: \la_{\max}(N) & & 0.6961 & 0.6997 & 0.7025 & 0.7045 & 0.7054
195: \end{matrix}
196: \]
197: Probably, the number $ \la_{\max} = \lim_N \la_{\max}(N) $ is close to $ 0.71
198: $. But, who knows? It could happen that $ \la_{\max} = 1 $ and, say, $
199: 1/(1-\la_{\max}(N)) \approx 3 + 0.04 \ln N $ for large $ N $.
200:
201: In Section \ref{sect2} it is shown that $ \la_{\max} \ne 1 $.
202:
203: Here is the plot of the angular Wigner function $ W_\psi^\ang $, where $ \psi
204: = \psi_{\max}(300) $ is the \dimensional{300} eigenvector corresponding to $
205: \la_{\max}(300) $.
206: \[
207: \begin{gathered}\includegraphics{pic1b.eps}\end{gathered}
208: \]
209: Its (small) negative values are responsible for the (small) quantum violation
210: of the classical bound ($ 0.71 > 2/3 $).
211: