1: % ****** Start of file Schack.tex ******
2: %
3: % This file is part of the APS files in the REVTeX 4 distribution.
4: % Version 4.0 of REVTeX, August 2001
5: %
6: % Copyright (c) 2001 The American Physical Society.
7: %
8: % See the REVTeX 4 README file for restrictions and more information.
9: %
10: % TeX'ing this file requires that you have AMS-LaTeX 2.0 installed
11: % as well as the rest of the prerequisites for REVTeX 4.0
12: %
13: % See the REVTeX 4 README file
14: % It also requires running BibTeX. The commands are as follows:
15: %
16: % 1) latex Schack.tex
17: % 2) bibtex Schack
18: % 3) latex Schack.tex
19: % 4) latex Schack.tex
20: %
21: %\documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
22: \documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
23:
24: % Some other (several out of many) possibilities
25: %\documentclass[preprint,aps]{revtex4}
26: %\documentclass[preprint,aps,draft]{revtex4}
27: %\documentclass[prb]{revtex4}% Physical Review B
28:
29: \usepackage{graphicx}% Include figure files
30: \usepackage{dcolumn}% Align table columns on decimal point
31: \usepackage{bm}% bold math
32:
33: %\nofiles
34:
35: \begin{document}
36:
37: \preprint{}
38:
39: \title{Volumes and Hyperareas of the Spaces of
40: Separable and Nonseparable Qubit-{\it Qutrit} Systems: Initial Numerical Analyses}
41:
42: \author{Paul B. Slater}%
43: \email{slater@kitp.ucsb.edu}
44: \affiliation{%
45: ISBER, University of California, Santa Barbara, CA 93106\\
46: }%
47: \date{\today}% It is always \today, today,
48: % but any date may be explicitly specified
49:
50: \begin{abstract}
51: Paralleling our recent computationally-intensive work for the case $N=4$
52: (quant-ph/0308037),
53: we undertake the task for $N=6$ of computing to
54: high numerical accuracy, the formulas
55: of Sommers and \.Zyczkowski (quant-ph/0304041) for the $(N^2-1)$-dimensional volume and $(N^2-2)$-dimensional
56: hyperarea of the (separable and nonseparable)
57: $N \times N$ density
58: matrices, based on the Bures (minimal monotone) metric. At the same time,
59: we estimate the {\it unknown} volumes and hyperareas based on
60: a number of other monotone metrics of interest. Additionally, we
61: estimate --- but
62: with perhaps unavoidably diminished
63: accuracy --- all these volume and hyperarea quantities, when restricted
64: to the ``small'' subset of
65: $6 \times 6$ density matrices that are separable (classically correlated) in nature. The ratios of separable to separable plus nonseparable volumes, then,
66: yield corresponding estimates of the ``probabilities of separability''.
67: We are particularly interested in
68: exploring
69: the possibility that a number of the various 35-dimensional volumes and 34-dimensional hyperareas, possess {\it exact}
70: values --- which we had, in fact, conjectured to be the case for the qubit-qubit systems ($N=4$), with the ``silver mean'', $\sqrt{2}-1$, appearing to play
71: a fundamental role as regards the separable states.
72: \end{abstract}
73:
74: \pacs{Valid PACS 03.65.Ud,03.67.-a, 02.60.Jh, 02.40.Ky}% PACS, the Physics and Astronomy
75: % Classification Scheme.
76: %\keywords{Suggested keywords}%Use showkeys class option if keyword
77: %display desired
78: \maketitle
79:
80: \section{Introduction}
81: In a recent highly comprehensive
82: analysis \cite{hans1} (cf. \cite{hansrecent}), Sommers and \.Zyczkowski obtained
83: ``a fairly general expression for the Bures volume of the submanifold of the states of rank $N-n$ of the set of complex ($\beta=2$) or real ($\beta=1$)
84: $N \times N$ density matrices
85: \begin{equation} \label{HZ1}
86: S^{(\beta)}_{N,n}= 2^{-d_{n}} \frac{\pi^{(d_{n}+1)/2}}{\Gamma((d_n+1)/2)} \Pi^{N-n}_{j=1} \frac{\Gamma(j \beta/2) \Gamma[1+(2 n+j-1) \beta/2]}{\Gamma[(n+j) \beta/2] \Gamma[1+(n+j-1) \beta/2]},
87: \end{equation}
88: where $d_{n}= (N-n) [1+(N+n-1) \beta/2] -1$ represents the dimensionality of the
89: manifold \ldots for $n=0$ the last factor simply equals unity and (\ref{HZ1})
90: gives the Bures volume of the entire space of density matrices, equal to that
91: of a $d_{0}$-dimensional hyper-hemisphere with radius 1/2. In the case
92: $n=1$ we obtain the volume of the surface of this set, while for $n=N-1$ we get the volume of the set of pure states \ldots which for $\beta=1(2)$ gives correctly the volume of the real (complex) projective space of dimensions
93: $N-1$'' \cite{hans1}.
94:
95: The Bures metric on various spaces of density matrices has been widely
96: studied \cite{hubner1,hubner2,ditt1,ditt2}. In a broader context, it serves
97: as the {\it minimal} monotone metric \cite{petz1}.
98:
99: Let us apply (\ref{HZ1}) to the cases of specific interest in this study,
100: $N=6,n=0,\beta=2$ and $N=6,n=1,\beta=2$ --- that is, the Bures 35-dimensional volume and
101: 34-dimensional hyperarea of the complex $6 \times 6$ density matrices.
102: We then have that
103: \begin{equation} \label{m1}
104: S^{(2)}_{6,0}= \frac{{\pi }^{18}}{12221326970165372387328000} \approx
105: 7.27075 \cdot {10}^{-17}
106: \end{equation}
107: and
108: \begin{equation} \label{m2}
109: S^{(2)}_{6,1}= \frac{{\pi }^{17}}{138339065763438059520000} \approx 2.04457 \cdot {10}^{-15}.
110: \end{equation}
111: Here, we are able (somewhat paralleling our recent work for $N=4$
112: \cite{slatersilver}, but in a
113: rather
114: more systematic manner than there) through numerical (quasi-Monte Carlo/quasi-random) methods to reproduce
115: both of these values (\ref{m1}), (\ref{m2}), to a considerable accuracy.
116: At the same time, we compute numerical values --- it would seem reasonable
117: to assume,
118: with roughly the same level of accuracy --- of these two quantities, but for the replacement of the Bures metric by five other {\it monotone} metrics of
119: interest. These are the Kubo-Mori \cite{hasegawa,petz3,michor,streater},
120: (arithmetic) average \cite{slatersilver}, Wigner-Yanase \cite{gi,wy,luo,luo2},
121: Grosse-Krattenthaler-Slater (GKS) \cite{KS,gillmassar}
122: and (geometric) average monotone
123: metrics --- the two ``averages'' being formed from the minimal and {\it maximal} monotone metrics, following the suggested procedure in
124: \cite[eq. (20)]{petz2}. No proven
125: formulas, such as (\ref{HZ1}), are presently available for these various quantities,
126: although our research here and in \cite{slatersilver} strongly suggests that
127: the Kubo-Mori volume of the $N \times N$ density matrices is equal
128: to simply $2^{N(N-1)/2)} S^{(2)}_{N,0}$, which for our case of $N=6$ would be $32768 S^{(2)}_{6,0}$ (cf. Tables I and II).
129: (In light of the considerable attention recently devoted to the (Riemannian, but non-monotone) Hilbert-Schmidt metric \cite{hansrecent,hilb1,hilb2} --- it would
130: certainly seem appropriate to include it as
131: well in any further analyses along the lines here and in \cite{slatersilver}.)
132:
133: Further, we compute for all these six metrics the 35-dimensional volumes and
134: 34-dimensional hyperareas restricted to the {\it separable} $2 \times 3$ and
135: $3 \times 2$ systems. (Then, we can, obviously, by taking ratios obtain ``probabilities of separability'' --- a topic which was first investigated in \cite{ZHSL},
136: and studied further, using the Bures metric, in \cite{zycz2,slaterA,slaterC}).
137: For this purpose, we employ the convenient Peres-Horodecki necessary {\it and}
138: sufficient positive
139: partial
140: transposition criterion for separability \cite{asher,michal} --- asserting that a $4 \times 4$ or $6 \times 6$ density matrix is separable if and only if all the eigenvalues of its
141: partial transpose are positive. (But in the $6 \times 6$ case, we have the qualitative difference that partial transposes can be determined in two inequivalent ways, either by transposing in place, in the natural
142: manner,
143: the nine $2 \times 2$ submatrices or the four
144: $3 \times 3$ submatrices. We will throughout this study --- as in \cite{qubitqutrit} --- analyze results using {\it both} forms of partial transpose.
145: It is our anticipation --- although without a formal demonstration --- that in the limit of large sample size, the two sets of results will converge
146: to true {\it common} values.)
147:
148: Our main analysis takes the form of a quasi-Monte Carlo (Tezuka-Faure
149: \cite{tezuka})
150: numerical integration over the 35-dimensional hypercube ($[0,1]^{35}$) and a
151: 34-dimensional subhypercube of it. In doing so, we implement
152: a parameterization of the $6 \times 6$ density matrices in terms of thirty Euler angles (parameterizing $6 \times 6$ unitary matrices) and {\it five} hyperspherical angles
153: (parameterizing the {\it six} eigenvalues --- constrained to sum to 1
154: \cite{sudarshan,toddecg}). We hold a single one of the five hyperspherical angles fixed in the 34-dimensional analysis. (The parameters are linearly transformed so
155: that they all lie in the interval [0,1].)
156:
157: We have previously pursued a similar numerical analysis in investigating the separable and nonseparable volumes and hyperareas of the $4 \times 4$ density matrices \cite{slatersilver}.
158: Highly accurate results (as gauged in terms of {\it known}
159: Bures quantities \cite{hans1}) --- based on two {\it billion} points of a Tezuka-Faure (``low discrepancy'')
160: sequence lying in the 15-dimensional hypercube --- led us to advance several strikingly simple conjectures.
161: For example, it was indicated that the Kubo-Mori volume of separable and nonseparable states was exactly $64 =2^6$ times the Bures volume. (The exponent
162: 6 is expressible --- in terms of our general conjecture,
163: mentioned above,
164: relating the Bures and Kubo-Mori volumes --- as $N(N-1)/2$, $N=4$.)
165: Most prominently, though, it appeared that the statistical distinguishability
166: (SD) volume
167: was simply expressible as $\frac{\sigma_{Ag}}{3}$, where the ``silver mean''
168: \cite{christos,spinadel,gumbs,kappraff},
169: $\sigma_{Ag} = \sqrt{2}-1 \approx 0.414214$, and $10 \sigma_{Ag}$ in terms of
170: (four times) the Kubo-Mori
171: metric. (The SD metric is identically four times the Bures metric
172: \cite{caves}. Consequently, the SD 15-dimensional volume of the $4 \times 4$ complex density matrices
173: is $2^{15}$ times that of the Bures volume --- given by (\ref{HZ1}) for $N=4,n=0,\beta=2$ --- thus equalling the volume
174: of a 15-dimensional hyper-hemisphere with radius 1, rather than $\frac{1}{2}$
175: as in the Bures case itself.)
176: Unfortunately, there appears to be little in the way of ``clues'' in the literature, as to how one might {\it formally} prove or disprove these conjectures --- ``brute
177: force'' {\it symbolic} integration appearing to be well beyond present technical/conceptual capabilities --- although the author ``suspects'' that at least in the Bures/{\it minimal}
178: monotone case, a proof might conceiveably be based on the concept of ``minimal volume'' \cite{bayard,bowditch,bambah}. (Certainly, Sommers and \.Zyczkowski \cite{hans1} did not directly
179: employ symbolic integration methodologies in deriving the Bures volume, hyperarea...for $N$-level [separable {\it and} nonseparable] systems, but rather, principally,
180: used concepts of random matrix theory.)
181:
182: The monotone metrics (of which we study five, in addition to the
183: Bures) can all be expressed
184: in the form
185: \begin{equation}
186: g_{\rho}(X',X)
187: = \frac{1}{4} \Sigma_{\alpha,\beta} |\langle \alpha |X| \beta \rangle |^2 c_{monotone}(\lambda_{\alpha},\lambda_{\beta})
188: \end{equation}
189: (cf. \cite{hubner1,hubner2}).
190: Here $X,X'$ lie in the tangent space of all Hermitian $N \times N$ density
191: matrices $\rho$ and $|\alpha \rangle, \alpha =1, 2 \ldots$ are eigenvectors
192: of $\rho$ with eigenvalues $\lambda_{\alpha}$.
193: Now, $c_{monotone}(\lambda_{\alpha},\lambda_{\beta})$ represents the specific {\it Morozova-Chentsov} function for the monotone metric in question \cite{petz2}. This function takes the form
194: for the Bures metric,
195: \begin{equation} \label{Bures}
196: c_{Bures}(\lambda_{\alpha},\lambda_{\beta}) = \frac{2}{\lambda_{\alpha} +\lambda_{\beta}},
197: \end{equation}
198: for the Kubo-Mori metric (which, up up to a scale factor, is
199: the unique monotone
200: Riemannian metric with respect to which the exponential and mixture
201: connections are dual \cite{streater}),
202: \begin{equation} \label{KM}
203: c_{KM}(\lambda_{\alpha},\lambda_{\beta}) =\frac{\log{\lambda_{\alpha}}-\log{\lambda_{\beta}}}{\lambda_{\alpha}-\lambda_{\beta}},
204: \end{equation}
205: for the (arithmetic) average metric (first discussed in \cite{slatersilver}),
206: \begin{equation}
207: c_{arith}(\lambda_{\alpha},\lambda_{\beta}) = \frac{4 (\lambda_{\alpha}+\lambda_{\beta})}{\lambda_{\alpha}^2 + 6 \lambda_{\alpha} \lambda_{\beta} +\lambda_{\beta}^2},
208: \end{equation}
209: for the Wigner-Yanase metric (which corresponds
210: to a space of {\it constant curvature} \cite{gi}),
211: \begin{equation}
212: c_{WY}(\lambda_{\alpha},\lambda_{\beta}) =\frac{4}{(\sqrt{\lambda_{\alpha}} +\sqrt{\lambda_{\beta}})^2},
213: \end{equation}
214: for the GKS/quasi-Bures metric (which yields the asymptotic redundancy for
215: universal quantum data compression \cite{KS}),
216: \begin{equation}
217: c_{GKS}(\lambda_{\alpha},\lambda_{\beta})= \frac{{\frac{\lambda_{\alpha}}{\lambda_{\beta}}}^{\lambda_{\alpha}/(\lambda_{\beta}-\lambda_{\alpha})}}{\lambda_{\beta}} e
218: \end{equation}
219: and for the (geometric) average metric (apparently previously unanalyzed),
220: \begin{equation} \label{geom}
221: c_{geom}(\lambda_{\alpha},\lambda_{\beta}) =\frac{1}{2 \sqrt{ \lambda_{\alpha} \lambda_{\beta}}}.
222: \end{equation}
223: \section{Analyses}
224: Based on the first 600 million points of a Tezuka-Faure sequence, to which we are continuing to copiously add, we
225: obtained the results reported in Tables~\ref{tab:table1}-\ref{tab:table7}. (We followed the Bures
226: formulas in \cite[secs. III.C, III.D]{hans1}, substituting the Morozova-Chentsov functions
227: given above (\ref{KM})-(\ref{geom}),
228: in the appropriate manner, to obtain their counterparts for the various non-Bures monotone metrics.)
229:
230: In Table~\ref{tab:table1}, we scale the estimates of the volumes and hyperareas by the {\it known} values (\ref{m1}), (\ref{m2}) of
231: $S^{(2)}_{6,0}$ and $S^{(2)}_{6,1}$, while in Table~\ref{tab:table2} we scale these estimates by the {\it estimated} values ($7.21259 \cdot 10^{-17}$ and $2.04607 \cdot 10^{-15}$) of these two quantities.
232: (We use {\it both} approaches because we are uncertain as to which may be more revealing as to possible exact ratios --- the possibility of which is
233: suggested by our work in \cite{slatersilver}. It is interesting to observe that the convergence to the true values of $S^{(2)}_{6,0}$ and $S^{(2)}_{6,1}$ appears
234: to be more pronounced in the 34-dimensional case than in the 35-dimensional one, although the Tezuka-Faure sequence we employ is specifically designed
235: as a {\it 35}-dimensional one --- of which we take an essentially arbitrary
236: 34-dimensional {\it projection} (cf. \cite[sec. 7]{morokoff}).)
237:
238: \begin{table}
239: \caption{\label{tab:table1}Scaled estimates based on the Tezuka-Faure sequence of 600 million points of
240: the 35-dimensional volumes and 34-dimensional
241: hyperareas of the $6 \times 6$ density matrices, using several monotone metrics. The scaling factors are the {\it known}
242: values of the volume and
243: hyperarea for the Bures metric, given by (\ref{HZ1}), and more specifically
244: for the cases $N=6$, $n=0,1$, $\beta=2$ by
245: (\ref{m1}) and (\ref{m2}).}
246: \begin{ruledtabular}
247: \begin{tabular}{rrr}
248: metric & volume/$S_{6,0}^{(2)}$ & hyperarea/$S_{6,1}^{(2)}$ \\
249: \hline
250: Bures & 0.992001 & 1.00073\\
251: KM & 31046.8 & 45.5328 \\
252: arith & 614.789 & 31.3225 \\
253: WY & 130.323 & 9.78041 \\
254: GKS & 12.2984 & 3.56433\\
255: geom & $6.53456 \cdot 10^{38}$ & $2.1581 \cdot 10^{9}$ \\
256: \end{tabular}
257: \end{ruledtabular}
258: \end{table}
259: \begin{table}
260: \caption{\label{tab:table2}Scaled estimates based on the Tezuka-Faure sequence of 600 million points of the 35-dimensional volumes and 34-dimensional
261: hyperareas of the $6 \times 6$ density matrices, using several monotone metrics. The scaling factors ($\tilde{S}$) are the {\it estimated}
262: values ($7.21259 \cdot 10^{-17}$ and \newline
263: $2.04606 \cdot 10^{-15}$ ) of the volume and hyperarea for the Bures metric.}
264: \begin{ruledtabular}
265: \begin{tabular}{rrr}
266: metric & volume/$\tilde{S}^{(2)}_{6,0}$ &
267: hyperarea/$\tilde{S}^{(2)}_{6,0}$ \\
268: \hline
269: KM & 31297.2 & 45.4995\\
270: arith & 619.747 & 31.2995\\
271: WY & 131.374 & 9.77326\\
272: GKS & 12.3976 & 3.56172 \\
273: geom & $6.58726 \cdot 10^{38}$ & $2.15652 \cdot 10^{9}$ \\
274: \end{tabular}
275: \end{ruledtabular}
276: \end{table}
277: In Tables~\ref{tab:table3} and Tables~\ref{tab:table4}, we report our estimates (scaled by the values obtained for the Bures metric)
278: of the volumes and hyperareas of the $6 \times 6$ separable complex density matrices.
279: Let us note, however, that
280: to compute the hyperarea of the complete boundary of the separable states, one must also include those $6 \times 6$ density matrices of full rank,
281: the partial transposes of which have a zero eigenvalue, and all other eigenvalues nonnegative
282: \cite{shidu}. (We do not compute this contribution here, as it would slow considerably
283: the overall process in which we are engaged, since high-degree polynomials
284: would need to be
285: solved at each step.) In \cite{slatersilver}, we had been led to conjecture that
286: that part of the 14-dimensional boundary of separable $4 \times 4$ density matrices consisting
287: generically of rank-{\it four} density matrices had
288: SD hyperarea $\frac{55 \sigma_{Ag}}{39}$
289: and that part composed of rank-{\it three} density matrices, $\frac{43 \sigma_{Ag}}{39}$, for a
290: total 14-dimensional boundary
291: SD hyperarea of $\frac{98 \sigma_{Ag}}{39}$. We, then, sought to apply the
292: ``Levy-Gromov isoperimetric inequality'' to the relation between the
293: known and estimated SD
294: volumes and hyperareas of the separable and separable plus nonseparable
295: states \cite[sec. VII.C]{slatersilver}.
296:
297:
298: In Table~\ref{tab:table3} we compute the partial transposes of the $6 \times 6$ density matrices by transposing in place the four $3 \times 3$ submatrices, while
299: in Table~\ref{tab:table4} we transpose in place the nine
300: $2 \times 2$ submatrices.
301: \begin{table}
302: \caption{\label{tab:table3}Scaled estimates based on the Tezuka-Faure sequence of 600 million points of the 35-dimensional volumes and 34-dimensional
303: hyperareas of the {\it separable} $6 \times 6$ density matrices, using several monotone metrics. The scaling factors are the {\it estimated}
304: values ($1.58803 \cdot 10^{-19}$ and $1.86401 \cdot 10^{-18}$) --- the true values being unknown --- of the volume and hyperarea for the Bures metric. To implement the Peres-Horodecki positive partial transposition criterion,
305: we compute
306: the partial transposes
307: of the four $3 \times 3$ submatrices (blocks) of the density matrix.}
308: \begin{ruledtabular}
309: \begin{tabular}{rrr}
310: metric & Bures-scaled volume & Bures-scaled hyperarea \\
311: \hline
312: KM & 2491.77 & 9.83988\\
313: arith & 226.332 & 11.3251\\
314: WY & 49.7186 & 4.30152 \\
315: GKS & 8.05581 & 2.35047\\
316: geom & $3.66418 \cdot 10^{23}$ & 52086.9 \\
317: \end{tabular}
318: \end{ruledtabular}
319: \end{table}
320: \begin{table}
321: \caption{\label{tab:table4}Scaled estimates based on the Tezuka-Faure sequence of 600 million points of the 35-dimensional volumes and 34-dimensional
322: hyperareas of the {\it separable} $6 \times 6$ density matrices, using several monotone metrics. The scaling factors are the {\it estimated}
323: values ($1.03484 \cdot 10^{-19}$ and $1.46223 \cdot 10^{-18}$) --- the true values being unknown --- of the volume and hyperarea
324: for the Bures metric. To implement the Peres-Horodecki positive partial transposition criterion, we compute
325: the partial transposes of the nine $2 \times 2$ submatrices (blocks) of the density
326: matrix.}
327: \begin{ruledtabular}
328: \begin{tabular}{rrr}
329: metric & Bures-scaled volume & Bures-scaled hyperarea \\
330: \hline
331: KM & 2563.64 & 8.41556 \\
332: arith & 198.163 & 10.3172 \\
333: WY & 46.5238 & 3.98471 \\
334: GKS & 7.59133 & 2.25601 \\
335: geom & $7.61764 \cdot 10^{23}$ & 12275.1 \\
336: \end{tabular}
337: \end{ruledtabular}
338: \end{table}
339: In Table~\ref{tab:table5}, we only require the density matrix
340: in question to pass {\it either} of the two tests, while
341: in Table~\ref{tab:table6}, we require it
342: to pass {\it both} tests for separability.
343: (Of the 600 million points of the Tezuka-Faure 35-dimensional sequence so
344: far generated,
345: approximately 2.91 percent yielded density matrices passing the test for Table I, 2.84 percent for
346: Table II, 4 percent for Table III and 1.74 percent for Table IV.)
347: \begin{table}
348: \caption{\label{tab:table5}Scaled estimates based on the Tezuka-Faure sequence of 600 million points of the 35-dimensional volumes and 34-dimensional
349: hyperareas of the {\it separable} $6 \times 6$ density matrices, using several monotone metrics. The scaling factors are the {\it estimated}
350: values ($2.59687 \cdot 10^{-19}$ and $3.30035 \cdot 10^{-18}$)--- the true values being unknown --- for the Bures metric.
351: A density matrix is included here if it passes {\it either} form of the positive partial transposition test.}
352: \begin{ruledtabular}
353: \begin{tabular}{rrr}
354: metric & Bures-scaled volume & Bures-scaled hyperarea \\
355: \hline
356: KM & 2534.46 & 9.25073\\
357: arith & 216.379 & 10.9214 \\
358: WY & 48.6732 & 4.17378 \\
359: GKS & 7.89451 & 2.31228 \\
360: geom & $4.63187 \cdot 10^{23}$ & 34855.9 \\
361: \end{tabular}
362: \end{ruledtabular}
363: \end{table}
364: \begin{table}
365: \caption{\label{tab:table6}Scaled estimates based on the Tezuka-Faure sequence of 600 million points of the 35-dimensional volumes and 34-dimensional
366: hyperareas of the {\it separable} $6 \times 6$ density matrices, using several monotone metrics. The scaling factors are the {\it estimated}
367: values ($2.59945 \cdot 10^{-21}$ and $3.59687 \cdot 10^{-31}$) --- the true values being unknown --- for the Bures metric.
368: A density matrix is included here {\it only}
369: if it passes
370: {\it both} forms of the positive partial transposition test.}
371: \begin{ruledtabular}
372: \begin{tabular}{rrr}
373: metric & Bures-scaled volume & Bures-scaled hyperarea \\
374: \hline
375: KM & 1087.9 & 4.50082\\
376: arith & 98.6297 & 5.85995 \\
377: WY & 26.9706 & 2.69208 \\
378: GKS & 5.67912 & 1.80686 \\
379: geom & $6.43785 \cdot 10^{24}$ & 117.375 \\
380: \end{tabular}
381: \end{ruledtabular}
382: \end{table}
383:
384: In Table VII, we ``pool'' (average) the results for the separable volumes
385: and hyperareas reported in Tables III and IV, based on the two
386: distinct forms of
387: partial transposition, to obtain possibly superior estimates of these
388: quantities, which presumably are actually one and the same {\it independent}
389: of the particular form of partial transposition.
390: \begin{table}
391: \caption{\label{tab:table7}Scaled estimates
392: obtained by pooling the results from Tables III and
393: IV --- based on the two forms of partial transposition --- for
394: the separable volumes and hyperareas. The Bures scaling factors
395: (pooled volume and hyperarea) are $1.31143 \cdot 10^{-19}$
396: and $1.66312 \cdot 10^{-18}$}
397: \begin{ruledtabular}
398: \begin{tabular}{rrr}
399: metric & Bures-scaled volume & Bures-scaled hyperarea \\
400: \hline
401: Bures & 2520.12 & 9.21374 \\
402: arith & 215.212 & 10.882 \\
403: WY & 48.4581 & 4.16225 \\
404: GKS & 7.87255 & 2.30894 \\
405: geom & $5.224 \cdot 10^{23}$ & 34585.5 \\
406: \end{tabular}
407: \end{ruledtabular}
408: \end{table}
409: \section{Discussion}
410: Of course, by taking the {\it ratios} of estimates of the volumes/hyperareas
411: of separable states to the estimates of the volumes/hyperareas of
412: separable plus nonseparable states, one would, in turn, obtain estimates of
413: the probabilities of separability \cite{ZHSL}
414: for the various monotone metrics studied. (Obviously, scaling the estimated volumes
415: and hyperareas by the corresponding estimates for the
416: Bures metric, as we have done in the
417: tables above for numerical convenience and possible insightfulness,
418: would be inappropriate in such a process.) The Bures metric gives the
419: {\it largest} probability of separability.
420:
421: In \cite{qubitqutrit}, we attempted a somewhat similar quasi-Monte Carlo
422: qubit-qutrit
423: analysis (but restricted to the Bures metric) to that reported above, but based on many fewer points (70 million {\it vs.} the 600 million so far
424: used here) of a
425: (Halton) sequence.
426: At this stage, having made use of considerably increased computer power (and streamlined MATHEMATICA programming --- in particular
427: employing the Compile command, which enables the program to
428: proceed under the condition
429: that certain variables will enter a calculation only as
430: machine numbers, and
431: not as lists, algebraic objects or any other kind of expression),
432: we must regard this earlier study as superseded by the one here.
433: (Our pooled estimate of the Bures volume of the separable qubit-qutrit
434: systems here [Table VII]
435: is $1.31143 \cdot 10^{-19}$, while in \cite{qubitqutrit}, following our earlier work for $N=4$ \cite{firstqubitqubit},
436: we formulated a conjecture (\cite[eq. (5)]{qubitqutrit}) --- in which we can now have but
437: very little
438: confidence --- that would give [converting from the SD metric to the
439: Bures] a value of
440: $2.19053 \cdot 10^{-9} \cdot 2^{-35} \approx 6.37528 \cdot 10^{-20}$.)
441: We also anticipate revisiting the $N=4$ (qubit-qubit) case \cite{slatersilver}
442: with our newly accelerated programming methods.
443:
444: We continue to add to the 600 million points of the Tezuka-Faure sequence employed above, and hope to report considerably more accurate results in
445: the future (based on which, hopefully,
446: we can advance plausible hypotheses as to
447: the true underlying
448: values of the 35-dimensional volumes and
449: 34-dimensional hyperareas). In fact, at the time of submission of this paper, we have already generated an additional 110 million points, using several {\it
450: independent} processors. Since the additional points are not numbered 600,000,0001 to 710,000,000 in the
451: sequence --- with the gaps remaining to be filled in --- it seems inappropriate to report the results fully now. Let us only indicate that if we were to do so,
452: the first line of Table I (that is, 0.992001=1/1.00806 and 1.00073=1/0.999269) would be replaced
453: by 0.9985 and 0.999644, so certainly the prognosis for considerably greater accuracy, as we extend the length of the sequence,
454: is good.
455:
456: It would be interesting to conduct analogous investigations to those reported here and in \cite{slatersilver} for the case $N=4$, using
457: quasi-random sequences {\it other}
458: than Tezuka-Faure ones \cite{tezuka}, particularly those for which it is possible to do {\it statistical} testing on the results
459: (such as constructing confidence intervals) \cite{hong}. (It is, of course,
460: possible to conduct statistical testing using simple Monte Carlo methods, but
461: their convergence is much slower than that of the quasi-Monte Carlo
462: procedures. Since we are dealing with quite high-dimensional spaces, good
463: convergence has been our dominant consideration in the selection of
464: numerical integration methodologies to employ.)
465:
466:
467: \begin{acknowledgments}
468: I wish to express gratitude to the Kavli Institute for Theoretical
469: Physics for computational support in this research and to Giray \"Okten
470: for supplying the MATHEMATICA code for the Tezuka-Faure quasi-Monte Carlo
471: procedure.
472:
473: \end{acknowledgments}
474:
475:
476: \bibliography{QQ}% Produces the bibliography via BibTeX.
477:
478: \end{document}
479: %
480: % ****** End of file Test2.tex ******