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\input{../../../../../../../../Network/Servers/lorien/Volumes/u1/residents/slater/Husimi.tex}
25:
26: %\documentclass[preprint,aps]{revtex4}
27: %\documentclass[preprint,aps,draft]{revtex4}
28: %\documentclass[prb]{revtex4}% Physical Review B
29:
30: \usepackage{graphicx}% Include figure files
31: \usepackage{dcolumn}% Align table columns on decimal point
32: \usepackage{bm}% bold math
33:
34:
35: %\nofiles
36:
37: \begin{document}
38:
39: \preprint{}
40:
41: \title{Analytic Fits to Separable Volumes and Probabilities for Qubit-Qubit
42: and Qubit-Qutrit Systems}
43:
44:
45: \author{Paul B. Slater}%
46: \email{slater@kitp.ucsb.edu}
47: \affiliation{%
48: ISBER, University of California, Santa Barbara, CA 93106\\
49: }%
50: \date{\today}% It is always \today, today,
51: % but any date may be explicitly specified
52:
53: \begin{abstract}
54: We investigate the possibility of deriving analytical formulas for
55: the $15$-dimensional {\it separable} volumes, in terms of any of a
56: number of metrics of interest (Hilbert-Schmidt [HS], Bures,...),
57: of the two-qubit (four-level) systems. This would appear to
58: require $15$-fold symbolic integrations over a complicated
59: convex body (defined by {\it both} separability and feasibility
60: constraints). The associated 15-dimensional
61: integrands --- in terms of the Tilma-Byrd-Sudarshan Euler-angle-based
62: parameterization of the $4 \times 4$ density matrices $\rho$ ({\it J. Phys. A}
63: 35 [2002], 10445) --- would be
64: the {\it products} of 12-dimensional Haar measure $\mu_{Haar}$
65: ({\it common} to each metric) and
66: 3-dimensional measures $\mu_{metric}$ ({\it specific}
67: to each metric) over the $3d$-simplex formed by the four eigenvalues
68: of $\rho$.
69: We attempt here to estimate/determine
70: the 3-dimensional integrands (the products of the various [known]
71: $\mu_{metric}$'s and an
72: {\it unknown} symmetric
73: weighting function $W$)
74: remaining after the (putative) 12-fold integration of $\mu_{Haar}$ over
75: the twelve Euler angles. We do this by {\it first} fitting $W$, so
76: that the conjectured
77: HS
78: separable volumes and hyperareas
79: ({\it Phys. Rev. A} 71 [2005], 052319; cf. quant-ph/0609006) are reproduced.
80: We further evaluate a number of possible such
81: choices of $W$ by seeing how well they {\it also} yield
82: the conjectured separable volumes
83: for the Bures, Kubo-Mori, Wigner-Yanase and (arithmetic)
84: average monotone metrics and the conjectured
85: separable Bures {\it hyperarea} ({\it J. Geom. Phys.} 53 [2005], 74, Table VI).
86: We, in fact, find two such exact (rather similar)
87: choices for $W$ that give these
88: five conjectured ({\it non}-HS)
89: values all within 5\%. In addition to the above-mentioned Euler angle
90: parameterization of $\rho$,
91: we make extensive use of the Bloore parameterization
92: ({\it J. Phys. A} 9 [1976], 2059) in a companion set of two-qubit
93: separability analyses.
94:
95: \end{abstract}
96:
97: \pacs{Valid PACS 03.67.-a, 2.40.Dr, 02.40.Ft, 02.30.Uu}% PACS, the Physics and Astronomy
98: % Classification Scheme.
99: %\keywords{Suggested keywords}%Use showkeys class option if keyword
100: %display desired
101: \maketitle
102: \section{Introduction}
103: In a pair of major, skillful papers, making use of the theory of
104: random matrices \cite{random}, Sommers and \.Zyczkowski were able to
105: derive explicit formulas for the volumes occupied by the
106: $d= (n^2-1)$-dimensional convex set of $n \times n$ (complex)
107: density matrices (as well as the $d=\frac{(n-1)(n+2)}{2}$-dimensional
108: convex set of real $n \times n$ density matrices),
109: {\it both} in terms of the Hilbert-Schmidt (HS) metric \cite{szHS} --- inducing the flat, Euclidean geometry --- and
110: the Bures metric \cite{szBures} (cf. \cite{szMore}).
111: (These results are also more lately discussed in the highly comprehensive
112: new text of Bengtsson and \.Zyczkowski \cite[chap. 14]{ingemar}.)
113: Of course, it would be of obvious
114: considerable quantum-information-theoretic
115: interest in the cases that $n$ is a {\it composite}
116: number, to also obtain HS and Bures volume
117: formulas {\it restricted} to those states that
118: are {\it separable} --- the sum of {\it
119: product} states --- in terms of some factorization of $n$ \cite{ZHSL}.
120: Then, by taking ratios --- employing these Sommers-\.Zyczkowski
121: results --- one would obtain corresponding
122: separability {\it probabilities}.
123:
124: In particular, again for the 15-dimensional complex
125: case, $n=4 = 2 \times 2$, {\it numerical}
126: evidence has been adduced
127: that the Bures volume of separable states is
128: (quite elegantly) $2^{-15} (\frac{\sqrt{2}-1}{3}) \approx
129: 4.2136 \cdot 10^{-6}$ \cite[Table VI]{slaterJGP}
130: and the HS volume
131: $(5 \sqrt{3})^{-7}
132: \approx 2.73707 \cdot 10^{-7}$ \cite[eq. (41)]{slaterPRA}.
133: Then, taking ratios (using the Sommers-\.Zyczkowski results
134: \cite{szHS,szBures}),
135: we have the {\it derived} conjectures that the Bures {\it separability}
136: probability
137: is $\frac{1680 (\sqrt{2}-1)}{\pi^8} \approx 0.0733389$
138: and the HS one,
139: considerably larger, $\frac{2^2 \cdot 3 \cdot 7^2 \cdot 11 \cdot 13 \sqrt{3}}{5^4 \pi^{6}} \approx 0.242379$ \cite[eq. (43), but misprinted as $5^3$ not
140: $5^4$ there]{slaterPRA}.
141: (Szarek, Bengtsson and \.Zyczkowski --- motivated by the numerical
142: findings of \cite{slaterPRA,slaterChinese} --- have recently
143: formally demonstrated
144: ``that the probability to find a random state to be separable equals 2 times the probability to find a random boundary state to be separable, provided the random states are generated uniformly with respect to the Hilbert-Schmidt (Euclidean) distance. An analogous property holds for the set of positive-partial-transpose states for an arbitrary bipartite system'' \cite{sbz}
145: (cf. \cite{innami}).
146: (``Since our reasoning hinges directly on the
147: Euclidean geometry,
148: it does not allow to predict any values of analogous ratios
149: computed with respect to Bures measure, nor other measures''
150: \cite[p. L125]{sbz}.)
151: These three authors
152: also noted \cite[p. L125]{sbz} that ``one could try to obtain similar
153: results for a general class of multipartite systems''. In this
154: latter vein,
155: numerical analyses of ours give some [but certainly not yet conclusive]
156: indication that for the {\it three}-qubit
157: {\it tri}separable states, there is an analogous probability ratio of
158: 6 --- rather than 2.)
159:
160: However, the analytical derivation of (conjecturally) {\it exact}
161: formulas for these HS and Bures (as well as other, such as the Kubo-Mori
162: \cite{petz1994} and Wigner-Yanase
163: \cite{wigneryanase,slaterPRA}) {\it separable} volumes still appears to be quite remote (cf. \cite{mathscinet}) --- the only such
164: progress to report so far being
165: certain exact formulas
166: when the number of dimensions of the 15-dimensional space of $4 \times 4$
167: density matrices has been severely
168: curtailed (nullifying or holding
169: constant {\it most} of the 15 parameters) to $d \leq 3$
170: \cite{pbsJak,pbsCanosa} (cf. \cite{slaterC}).
171: Most notably, in this research direction,
172: in \cite[Fig. 11]{pbsCanosa}, we were able to find
173: a highly interesting/intricate (one-dimensional) continuum
174: ($-\infty < \beta <\infty$) of two-dimensional
175: (the associated
176: parameters being $b_{1}$, the {\it mean}, and $\sigma_{q}^2$,
177: the {\it variance} of the Bell-CHSH observable)
178: HS separability
179: probabilities, in which the {\it golden ratio} \cite{livio} was
180: featured, among other items. (The associated
181: HS volume element --- $\frac{1}{32 \beta (1+\beta)}
182: d \beta d b_{q} d \sigma^2_{q}$ --- is
183: {\it independent} of $b_{1}$ and $\sigma_{q}^2$ in this
184: three-dimensional scenario. Extensions to higher-dimensional scenarios
185: $d>3$ appear problematical, though.)
186: Further, in \cite{pbsJak}, building upon work of
187: Jak\'obczyk and Siennicki \cite{jak}, we obtained a
188: remarkably wide-ranging variety of exact HS
189: separability ($n=4, 6$) and PPT (positive partial transpose)
190: ($n=8, 9, 10$) probabilities based on
191: {\it two}-dimensional {\it sections} of sets of
192: (generalized) Bloch vectors corresponding to $n \times n$
193: density matrices.
194:
195: In this paper we are able to report some
196: additional progress in these directions.
197: We obtain exact formulas for certain $d=4, n =4$ scenarios and
198: {\it upper} bounds for $d=7$ and $d=9$ instances.
199: (Nevertheless,
200: the full $d=9$ and/or $d =15$, $n=4$ real and complex scenarios
201: still appear quite daunting --- due to the numerous
202: separability constraints at work, some being active [binding]
203: in certain regions and
204: in complementary regions, inactive [nonbinding].
205: ``The geometry of the $15$-dimensional set of separable states of two
206: qubits is not easy to describe'' \cite[p. L125]{sbz}.)
207:
208: To proceed initially
209: (secs.~\ref{sc1} \ref{sc2} \ref{sc3}),
210: we employ the (quite simple) form of parameterization of the density matrices
211: put forth by Bloore \cite{bloore} some thirty years ago.
212: (Of course, there are several
213: other possible parametrizations
214: \cite{kk,byrd,sudarshan,vanik,fano,scutaru,stan}, a number of
215: which we have also utilized in various studies \cite{slaterA,slaterqip}
216: to estimate volumes of
217: separable states. Our greatest progress at this stage,
218: in terms of increasing dimensionality, has been achieved
219: with the Bloore parameterization --- due to a certain
220: computationally attractive feature of it, allowing us to
221: {\it decouple} diagonal and non-diagonal parameters --- as detailed
222: shortly below.)
223:
224: In our final
225: (quite differently structured) series of analyses (sec.~\ref{weighting functionsec}), though, we employ not the Bloore
226: parameterization, but the Euler-angle-based one
227: of Tilma, Byrd and Sudarshan \cite{sudarshan}.
228: Our motivation here is to bypass/circumvent the necessity of
229: the putatively achieveable, but computationally daunting
230: first twelve steps (over the twelve Euler angles) of a 15-fold
231: integration.
232: (``we would like to derive a subset of the ranges of the Euler angle
233: parameters...dividing the 15-parameter space into entangled and
234: separable subsets. Unfortunately, due to the complicated nature of the parameterization,
235: both numerical and symbolic calculations of the eigenvalues of the
236: partial transpose...have become computationally intractable using
237: standard mathematical software'' \cite[p. 10453]{sudarshan}.)
238: We seek to find the {\it three}-dimensional (weighting) function
239: $W(\lambda_{1},\lambda_{2},\lambda_{3},\lambda_{4})$
240: of the {\it four}
241: eigenvalues ($\lambda_{4}=1-\lambda_{1}-\lambda_{2}-\lambda_{3}$)
242: which would yield our various {\it conjectured} 15-dimensional
243: separable volumes and 14-dimensional separable hyperareas
244: for the two-qubit systems.
245: \cite{slaterJGP,slaterPRA}.
246: In sec.~\ref{n6}, we also apply this approach
247: exploratorily to the $n=6, d=35$ case
248: of qubit-{\it qutrit} pairs, seeking to find a suitable
249: 5-dimensional weighting function
250: $W_{n=6}(\lambda_{1},\lambda_{2},\lambda_{3},\lambda_{4},\lambda_{5},\lambda_{6})$.
251:
252: Let us also point out to the reader, our even more recent
253: companion-type study
254: \cite{slaterHSPRA}, in which solely for the
255: (relatively simple) Hilbert-Schmidt (non-monotone)
256: metric, are we able (making use of the Bloore parameterization)
257: to reduce the two-qubit (both real and complex)
258: separable volume determination
259: problems to those of finding {\it one}-dimensional weighting functions
260: (which turned out to be well-approximated by
261: certain incomplete beta functions --- functions of the ratio
262: $\frac{\rho_{11} \rho_{44}}{\rho_{22} \rho_{33}}$ of diagonal entries
263: of the $4 \times 4$ density matrices $\rho$).
264: \section{Bloore parameterization of the density matrices} \label{sc1}
265: The main presentation of Bloore \cite{bloore}
266: was made in terms of the $3 \times 3$ ($n=3$)
267: density matrices. It is clearly easily extendible to cases
268: $n > 3$.
269: The fundamental idea is to {\it scale} the {\it
270: off}-diagonal elements $(\rho_{ij},
271: i \neq j)$
272: of the density matrix in terms of the {\it square roots} of the diagonal
273: entries ($\rho_{ii}$). That is, we set (introducing the new [Bloore]
274: variables $z_{ij}$),
275: \begin{equation}
276: \rho_{ij} = \sqrt{\rho_{ii}
277: \rho_{jj}} z_{ij}.
278: \end{equation}
279: This allows the {\it determinant} of $\rho$ (and analogously
280: all its
281: {\it principal minors}) to be expressible as the {\it product}
282: ($|\rho| = A B$) of
283: two factors, one ($A$) of which is itself simply the product of
284: (positive)
285: diagonal entries ($\rho_{ii}$)
286: and the other --- in the $n=4$ case under investigation \newline
287: here (easily extendible from the case of real density matrices
288: to complex ones) ---
289: \begin{equation} \label{B}
290: B= \left(z_{34}^2-1\right) z_{12}^2+2 \left(z_{14}
291: \left(z_{24}-z_{23} z_{34}\right)+z_{13}
292: \left(z_{23}-z_{24} z_{34}\right)\right)
293: z_{12}-z_{23}^2-z_{24}^2-z_{34}^2+
294: \end{equation}
295: \begin{displaymath}
296: z_{14}^2
297: \left(z_{23}^2-1\right)+ z_{13}^2
298: \left(z_{24}^2-1\right)+2 z_{23} z_{24} z_{34}+2 z_{13}
299: z_{14} \left(z_{34}-z_{23} z_{24}\right)+1,
300: \end{displaymath}
301: involving ({\it only}) the $z_{ij}$'s ($i \neq j$)
302: \cite[eqs. (15), (17)]{bloore}.
303: Since, clearly, the factor $A$ is positive in all nondegenerate cases
304: ($\rho_{ii} \geq 0$),
305: one can --- by {\it only} analyzing $B$ --- essentially
306: ignore the diagonal entries (and thus {\it reduce} by ($n-1$)) the
307: dimensionality of the problem of finding nonnegativity
308: conditions to impose on $\rho$.
309: This is the feature we will seek to maximally
310: exploit here.
311:
312: It is, of course, necessary and sufficient for $\rho$ to serve
313: as a density matrix (that is, an Hermitian, nonnegative definite, trace
314: one matrix) that {\it all} its {\it principal} minors be nonnegative
315: \cite{horn}.
316: The condition --- quite natural in the Bloore
317: parameterization --- that all the principal $2 \times 2$ minors be
318: nonnegative requires simply that $-1 \leq z_{ij} \leq 1, i \neq j$. The
319: {\it joint} conditions that
320: {\it all} the principal minors be nonnegative are not as
321: readily apparent. But for the 9-dimensional {\it real}
322: case $n=4$ --- that is, $\Im(\rho_{ij})=0$ --- we have been able to obtain
323: one such set,
324: using the Mathematica implementation of the {\it cylindrical
325: algorithm decomposition} \cite{cylindrical}.
326: (The set of solutions of any system of real algebraic equations
327: and inequalities can be decomposed into a finite number of
328: ``cylindrical'' parts \cite{strzebonski}.)
329: Applying it, we were able to express the
330: conditions that
331: an arbitrary 9-dimensional $4 \times 4$ real density matrix
332: $\rho$ must fulfill.
333: These took the form, $z_{12}, z_{13}, z_{14} \in [-1,1]$ and
334: \begin{equation} \label{limits}
335: z_{23} \in [Z^-_{23},Z^+_{23}], \hspace{.1in}
336: z_{24} \in [Z^-_{24},Z^+_{24}], \hspace{.1in}
337: z_{34} \in [Z^-_{34},Z^+_{34}],
338: \end{equation}
339: where
340: \begin{equation}
341: Z^{\pm}_{23} =z_{12} z_{13} \pm \sqrt{1-z_{12}^2} \sqrt{1-z_{13}^2} ,
342: \hspace{.1in}
343: Z^{\pm}_{24} =z_{12} z_{14} \pm \sqrt{1-z_{12}^2} \sqrt{1-z_{14}^2} ,
344: \end{equation}
345: \begin{displaymath}
346: Z^{\pm}_{34} = \frac{z_{13} z_{14} -z_{12} z_{14} z_{23} -z_{12} z_{13} z_{24} +z_{23}
347: z_{24} \pm s}{1-z_{12}^2},
348: \end{displaymath}
349: and
350: \begin{equation}
351: s = \sqrt{-1 +z_{12}^2 +z_{13}^2 -2 z_{12} z_{13} z_{23} +z_{23}^2}
352: \sqrt{-1 +z_{12}^2 +z_{14}^2 -2 z_{12} z_{14} z_{24} +z_{24}^2}.
353: \end{equation}
354: Making use of these results, we were able to confirm {\it via} exact
355: {\it symbolic} integrations,
356: the (formally demonstrated)
357: result of \.Zyczkowski and Sommers
358: \cite{szHS} that the HS volume of the {\it real}
359: two-qubit ($n=4$) states is
360: $\frac{\pi^4}{60480} \approx 0.0016106$.
361: (This result was also achievable through a somewhat different
362: Mathematica computation, using the {\it implicit} integration feature
363: first
364: introduced in version 5.1. That is, the only integration limits employed were
365: that $z_{ij} \in [-1,1], i \neq j$ --- {\it broader} than those in
366: (\ref{limits}) --- while the Boolean constraints were imposed that
367: the determinant of $\rho$ and {\it one} [all that is needed to ensure
368: nonnegativity] of its principal $3 \times 3$ minors be nonnegative.)
369:
370: However, when we tried to combine these integration limits (\ref{limits})
371: with
372: the (Peres-Horodecki \cite{asher,michal,bruss} $n=4$)
373: {\it separability} constraint that the determinant ($C =|\rho_{PT}|$)
374: of the partial
375: transpose of $\rho$ be nonnegative \cite[Thm. 5]{ver},
376: we {\it exceeded} the memory availabilities of our workstations.
377: In general, the term $C$ --- unlike the earlier term $B$ --- unavoidably
378: involves
379: the diagonal entries ($\rho_{ii}$), so the
380: dimension of the accompanying integration problems must {\it increase} ---
381: in the $9$-dimensional real $n=4$ case from 6 to 9.
382: \subsection{Restricting diagonal entries}
383: Nevertheless, we found that by imposing the condition that the four diagonal
384: entries ($\rho_{ii}, i =1,\ldots,4$) fall into two equal pairs
385: (say, $\rho_{11} =\rho_{22}$ and $\rho_{33}=\rho_{44}$, so that
386: $\rho_{33} =\frac{(1- 2 \rho_{11})}{2}$), the determinant ($C$) of
387: the partial transpose could now be expressed as the {\it product}
388: \begin{equation}
389: C = |\rho_{PT}| = \frac{1}{4} (1-2 \rho_{11})^2 \rho_{11}^2 D,
390: \hspace{.5in} 0\leq \rho_{11} \leq \frac{1}{2},
391: \end{equation}
392: where
393: \begin{equation} \label{D}
394: D=\left(z_{24}^2-1\right) z_{13}^2+2 z_{23}
395: \left(z_{34}-z_{14} z_{24}\right)
396: z_{13}-z_{24}^2-z_{34}^2+
397: \end{equation}
398: \begin{displaymath}
399: \left(z_{14}-1\right)
400: \left(z_{14}+1\right) \left(z_{23}-1\right)
401: \left(z_{23}+1\right)+2 z_{14} z_{24} z_{34}+
402: \end{displaymath}
403: \begin{displaymath}
404: \left(z_{34}^2-1\right) z_{12}^2+2 \left(z_{13}
405: z_{14}+z_{23} z_{24}-\left(z_{14} z_{23}+z_{13}
406: z_{24}\right) z_{34}\right) z_{12}.
407: \end{displaymath}
408: Thus, the term $D$ (like $B$ in (\ref{B}))
409: is itself {\it independent} of the diagonal entries
410: of $\rho$ (in particular, the specific value of $\rho_{11}$) --- allowing us to proceed, as indicated,
411: with integrations in a {\it lower}-dimensional ($d \leq 7$) setting.
412: (This same form of factorizability takes place, as well, for all the [four]
413: $3 \times 3$ and [six] $2 \times 2$ minors of $\rho_{PT}$.)
414: If we can further guarantee the nonnegativity of $D$ --- in addition
415: to that of $B$ --- we can ensure
416: separability of $\rho$. Let us also note that
417: \begin{equation} \label{factors}
418: B-D= 2 (z_{14} -z_{23}) (z_{13} -z_{24}) (z_{12} -z_{34}).
419: \end{equation}
420: (So, if any of the three factors in (\ref{factors})
421: are zero, the associated
422: state must be separable.)
423:
424: We will now proceed
425: to some specific analyses within this more restrictive framework
426: (7-dimensional in nature, since we started with the 9-dimensional real
427: setting and have essentially only one free diagonal parameter [$\rho_{11}$]
428: left).
429: \section{7-Dimensional Real Setting
430: ($\rho_{11}=\rho_{22},\rho_{33}=\rho_{44}$)} \label{sc2}
431: \subsection{7-dimensional analysis}
432: The associated $7$-dimensional HS volume of all these states
433: (separable and nonseparable) is $\frac{\pi^2}{15120} \approx
434: 0.000652752$. The six principal $2 \times 2$ minors of the partial transpose
435: simply yield the (Bloore)
436: conditions that $z_{ij} \in [-1,1]$, so nothing can be
437: gained --- in terms of obtaining upper bounds on the separable volume --- by
438: using the nonnegativity of these six minors as further constraints in
439: our integrations. However,
440: if we require that {\it one} of the $3 \times 3$ principal minors of the
441: partial transpose be nonnegative,
442: we do succeed in obtaining a nontrivial upper bound of
443: $\frac{\pi^4}{172032} \approx 0.000566227$ on the
444: $7$-dimensional volume of
445: separable states. So, we have a derived {\it upper bound}
446: (probably rather weak, we surmise)
447: on the HS separability
448: probability for our $7$-dimensional real set of $4 \times 4$
449: density matrices of $\frac{45 \pi^2}{512} \approx 0.867446$.
450: \subsection{4-dimensional analyses}
451: \subsubsection{$z_{12}=z_{23}=z_{24}=0$}
452: Here, we set the three indicated Bloore parameters to
453: zero. (So, the {\it four} free parameters of the initial {\it seven}
454: are $\rho_{11}, z_{13}, z_{14}$
455: and $z_{34}$.)
456: Then, using the implicit integration feature of Mathematica
457: (rather than the limits (\ref{limits})), we
458: were able to obtain for the total HS volume
459: $\frac{\pi^2}{384} \approx 0.0257021$ and (further
460: {\it adding} the separability constraint that
461: $D$, given by (\ref{D}), be greater than 0) the separable volume of
462: $\frac{4 + \pi^2}{1536} \approx 0.00902969$.
463: Taking the appropriate quotient, we find for the HS separability
464: probability for this scenario (quite elegantly),
465: $\frac{4 + \pi^2}{4 \pi^2} \approx 0.351321$.
466: \subsubsection{$z_{23}=0,z_{24}=0,z_{34}=0$}
467: Now, $HS_{vol}^{tot} = \frac{\pi}{144} \approx 0.0218166$ and
468: $HS_{vol}^{sep}= \frac{4 + \pi^2}{1536} \approx
469: 0.00902969$,
470: so \newline $HS_{sep prob} \equiv \frac{HS_{vol}^{sep}}{HS_{vol}^{tot}} =
471: \frac{3 (4 + \pi^2)}{32 \pi} \approx 0.413891$.
472:
473: There are twenty possible $4$-dimensional scenarios. Of these four each
474: have one of these two nontrivial HS separability probabilities
475: ($\frac{4 + \pi^2}{4 \pi^2}$ or
476: $\frac{3 (4 + \pi^2)}{32 \pi}$). For
477: the remaining twelve cases,
478: the HS separability probabilities are simply 1.
479: \subsection{3-dimensional analyses}
480: We also observe that for the fifteen possible $3$-dimensional
481: scenarios, all the HS separability probabilities are
482: (trivially) 1. (Eight of these
483: scenarios have HS total volume equal to $\frac{\pi^2}{128}$, four,
484: $\frac{\pi}{48}$ and three $\frac{1}{12}$.)
485: \subsection{$5$-dimensional analyses}
486: For all twelve possible scenarios, setting two $z_{ij}$'s to zero,
487: the total HS volume of states is
488: $\frac{\pi^2}{1440} \approx 0.00685389$.
489: \subsubsection{$z_{23}=z_{24}=0$}
490: To compute the separable HS volume we had to resort to
491: numerical means, obtaining 0.00532303, for a separability probability
492: of 0.776643. (This is also the probability for
493: the 5-dimensional scenario $z_{12}=z_{13}=0$ --- and [at least] five others.)
494: \subsubsection{$z_{14}=0=z_{23}=0$}
495: Here, as indicated,
496: the total HS volume is $\frac{\pi^2}{1440} \approx 0.00685389$.
497: This is also the separable volume --- since $B=D$ in this case.
498: (This is also the situtation with the scenarios $z_{13}=z_{24}=0$ and
499: $z_{12}=z_{34}=0$.)
500: \section{$9$-dimensional real case} \label{sc3}
501: As previously mentioned, we know from the Sommers-\.Zyczkowski
502: analyses \cite{szHS} that the
503: HS volume of the $9$-dimensional convex set of {\it real}
504: $4 \times 4$ density matrices is $\frac{\pi^4}{60480} \approx
505: 0.0016106$. (In this section --- to fully accord with
506: their results \cite{szHS} --- we
507: have to adjust by an overall scaling factor of $2^4 =16$ the results of
508: our usual integration procedure employed previously
509: in this study, in which we simply employed 1 as our integrand, rather
510: than some other constant. This scaling,
511: of course, does not have any impact on
512: separability {\it probabilities}.)
513: We can also computationally verify the Sommers-\.Zyczkowski HS
514: volume formula in either of two manners:
515: (a) employing the integration limits (\ref{limits})
516: obtained by application of
517: the cylindrical decomposition algorithm or (b) implementing the
518: implicit integration feature (requiring here that the determinant of
519: $\rho$ and one of its $3 \times 3$ principal minors be nonnegative)
520: of Mathematica, using for the integration limits
521: simply that $-1 \leq z_{ij} \leq 1$ for all $i \neq j$.
522: We investigated
523: how far we could proceed in the full $9$-dimensional $n=4$ real case,
524: by imposing increasingly greater requirements (corresponding to the
525: Peres-Horodecki criterion \cite{asher,michal}) that would need to
526: be fulfilled for $\rho_{PT}$ to be nonnegative definite.
527: Now, of the six $2 \times 2$ principal minors of $\rho_{PT}$, only two are
528: distinct from the six such minors of $\rho$ itself.
529: If we demand that, in addition, to the
530: feasibility constraints on $\rho$ that {\it
531: one} of
532: these two minors be nonnegative we
533: obtain
534: (to high precision)
535: 0.0014242052589 yielding an {\it upper bound} on the HS separability
536: probability of the $9$-dimensional real two-qubit states of
537: 0.88426997055. Further (tighter) numerical results --- using
538: additional principal minors of $\rho_{PT}$ (even the
539: remaining $2 \times 2$ nonredundant one)--- proved difficult to
540: achieve, however.
541: \section{$3$-Dimensional weight for the
542: {\it Full} $15$-Dimensional Problem} \label{weighting functionsec}
543: The direct straightforward (``brute force'') computations
544: by {\it symbolic} means of the $15$-dimensional
545: {\it separable}
546: volumes (for any of a wide variety
547: of metrics) of the (complex) two-qubit states
548: seem to far exceed present workstation
549: capabilities. Nevertheless, we will here try to gain
550: some {\it analytical} insight into these formidable problems.
551:
552: Let us first note that the computation of the separable volumes
553: could be seen to require the
554: evaluation (in which the Peres-Horodecki separability
555: and feasibility criteria are {\it both}
556: enforced)
557: of a $15$-fold integral. Following the innovative Euler-angle-based
558: parameterization of $4 \times 4$ density matrices of
559: Tilma, Byrd and Sudarshan \cite{sudarshan},
560: the first twelve variables over which to integrate can be taken
561: to be the twelve Euler angles ($\alpha_{i}, i = 1,...,12$) and the last
562: three to be the eigenvalues $\lambda_{1}, \lambda_{2}, \lambda_{3}$ of
563: the density matrix $\rho$.
564: (Of course, $\Sigma_{i=1}^{4} \lambda_{i} =1$.)
565: Now for any of the metrics of interest, the
566: associated $15$-dimensional integrand
567: can be represented as the {\it product} of
568: a 12-dimensional Haar measure ($\mu_{Haar}$) \cite[eq. (34)]{sudarshan}
569: ({\it common} to {\it all} the metrics
570: of interest) over the twelve Euler angles
571: \begin{equation}
572: \mu_{Haar} =\cos(\alpha_{4})^3 \cos (\alpha_{6})
573: \cos(\alpha_{10})
574: \sin(2 \alpha_{2}) \sin(\alpha_{4}) \sin(\alpha_{6})^5 \sin(2 \alpha_{8})
575: \sin(\alpha_{10})^3 \sin(2 \alpha_{12})
576: d \alpha_{12}...d \alpha_{1}
577: \end{equation}
578: ($0 \leq \alpha_{even} \leq \frac{\pi}{2}, 0 \leq \alpha_{odd} \leq \pi$
579: \cite[eq. (47)]{sudarshan})
580: and a 3-dimensional {\it metric-specific} measure ($\mu_{metric}$)
581: over the eigenvalues (cf. \cite{hall}). (So, if the Peres-Horodecki criterion
582: is {\it not} enforced, we simply obtain the [known in the Bures and
583: HS cases] volumes of the separable
584: {\it and} nonseparable states, the volumes
585: decomposing into the {\it products} of the
586: results of 3-fold and 12-fold integrations
587: \cite[eq. (3.7)]{szHS},\cite[eq. (3.17)]{szBures}.)
588:
589: Consequently, after the first twelve steps of the (presumptively
590: theoretically achieveable, but apparently
591: totally impractical) integrations for the separable
592: volumes, we can imagine obtaining {\it three}-dimensional
593: integration problems (over the three-dimensional simplex of
594: eigenvalues). The integrands of the problem would now be the {\it products}
595: of $\mu_{metric}$ and a {\it common}
596: three-dimensional {\it weighting function} $W$,
597: acquired during the course of the 12-fold integration.
598: Certainly, $W$ should
599: be a {\it symmetric} function \cite{ig} of the four eigenvalues.
600: \subsection{First analysis} \label{firstanalysis}
601: We will try to fit $W$ to our various conjectures
602: for the separable volumes \cite{slaterJGP,slaterPRA},
603: previously obtained by {\it numerical}
604: methods.
605: In particular, we have found (after some limited trial and error) that
606: the (symmetric) choice (being neither a {\it convex} nor a {\it concave}
607: function, we observed --- nor even approximately so),
608: \begin{equation} \label{weighting function}
609: W(\lambda_{1},\lambda_{2},\lambda_{3},\lambda_{4})
610: = 6086. (\lambda_{1} \lambda_{2} \lambda_{3} +
611: \lambda_{1} \lambda_{2} \lambda_{4} +\lambda_{1} \lambda_{3} \lambda_{4} +\lambda_{2} \lambda_{3} \lambda_{4})^{\frac{53}{20}},
612: \end{equation}
613: reproduces the conjectures for {\it both}
614: the {\it Hilbert-Schmidt} volumes and hyperareas
615: of the separable two-qubit states to good accuracy (0.01643\%).
616: (These conjectured volume
617: and hyperarea are $(5 \sqrt{3})^{-7} \approx
618: 2.73707 \cdot 10^{-7}$ and
619: $(3^2 5^6)^{-1}
620: \approx 7.11111 \cdot 10^{-6}$, respectively
621: \cite[eqs. (41), (42)]{slaterPRA}.)
622: For the computation of the 14-dimensional separable hyperarea, the weighting function
623: (\ref{weighting function}) reduces
624: (since we can take $\lambda_{4} = 0$) to
625: \begin{equation} \label{weighting function2}
626: W(\lambda_{1},\lambda_{2},\lambda_{3})
627: = 6086 (\lambda_{1} \lambda_{2} \lambda_{3} )^{\frac{53}{20}}.
628: \end{equation}
629: Now, the {\it acid} test of the legitimacy/validity of our choice of
630: weighting function --- and the {\it raison d'\^etre} of our exercise --- is
631: to see how well (in {\it addition} to the separable
632: Hilbert-Schmidt volumes and hyperareas (which we constructed to
633: satisfy the Szarek-Bengtsson-\.Zyczkowski {\it two}-fold
634: ratio \cite{sbz}) it reproduces (our presumptively correct)
635: conjectures for metrics {\it other} than the Hilbert-Schmidt one.
636:
637: For the Bures (minimal monotone \cite{hansen})
638: metric, we found that the use of the weighting function (\ref{weighting function}) --- coupled with
639: $\mu_{Bures}$ --- predicted a
640: separable volume 0.938275 times the magnitude of the
641: conjectured value of $2^{-15} (\frac{\sqrt{2}-1}{3}) \approx 4.2136
642: \cdot 10^{-6}$.
643: For the Kubo-Mori monotone metric, we obtained an estimate that is 0.910768
644: times the magnitude of the conjectured value of $2^{-15}(10 (\sqrt{2}-1))
645: \approx 0.000126408$, for the (arithmetic) {\it average} monotone
646: metric, 0.903281 times the magnitude of the conjectured value of
647: $2^{-15} (\frac{29}{9} (\sqrt{2}-1)) \approx 0.0000407314$, and for
648: the Wigner-Yanase monotone metric, 0.919585 times the conjectured value
649: of $2^{-15} (\frac{7}{4} (\sqrt{2}-1)) \approx 0.000221214$.
650: So, our choice of $W$ works rather well, at least for a first
651: simply heuristic effort. (In \cite{slaterJGP}, we also had additional
652: volume
653: conjectures for the GKS (Grosse-Krattenthaler-Slater) (``quasi-Bures'')
654: monotone metric (cf. \cite{chernoff}). However, we encountered numerical difficulties in trying
655: to analyze it here, in the fashion of the other metrics.)
656:
657: If we try, as well,
658: to predict the 14-dimensional separable {\it hyperarea} for the
659: Bures metric using (\ref{weighting function2}), we obtain an estimate
660: of 0.0000262122, which is 0.940364 times as large as the conjectured value
661: of $2^{-14} (\frac{43 (\sqrt{2}-1)}{39}) \approx 0.0000278746$
662: \cite[Table VI]{slaterJGP}.
663: (For the Kubo-Mori metric, we also obtain an estimate of the separable hyperarea
664: of 0.0000399861 and of the separable probability of a state on the
665: 14-dimensional boundary
666: of 0.0214689 --- but there were no prior conjectures for these
667: quantities in \cite[Table VI]{slaterJGP}. For the arithmetic average metric,
668: our estimate of the separable hyperarea is 0.738784, while our
669: conjecture in \cite[Table VI]{slaterJGP} [to be corrected by a factor of
670: 8 as noted in \cite[p. 1-11]{slaterPRA}] amounts to
671: $2^{-14} (\frac{255 (\sqrt{2}-1)}{128}) \approx 0.825191$.)
672:
673: \subsection{Further analyses --- exact weighting functions}
674: We, then, altered our analytical strategy somewhat and succeeded (twice)
675: in reproducing our five indicators ---- the Bures, Kubo-Mori, Wigner-Yanase
676: and (arithmetic) average separable volumes and the Bures hyperarea ---
677: all now within 5\% of their conjectured values. We accomplished this by
678: {\it exactly}
679: fitting (by finding the values for $a$ and $b$)
680: the conjectured HS separable volume and separable hyperarea to
681: weighting functions of the form (and their $\lambda_{4}=0$ reductions),
682: \begin{equation} \label{exactexact}
683: W(\lambda_{1},\lambda_{2},\lambda_{3},\lambda_{4}) = a (\Sigma^{4}_{i<j} \lambda_{i} \lambda_{j})^{m_{1}}
684: + b (\Sigma^{4}_{i<j<k} \lambda_{i} \lambda_{j} \lambda_{k})^{m_{2}}.
685: \end{equation}
686: We conducted separate analyses for pairs of
687: low integral values of
688: the exponents ($1 \leq m_{1}, m_{2} \leq 4$).
689: For $m_{1} =3, m_{2}=3$ we had the results
690: \begin{equation} \label{EXACT1}
691: a= \frac{325909584 \sqrt{3}}{464375 \pi ^6} \approx 1.26422,
692: b= \frac{5070990172248 \sqrt{3}}{464375 \pi ^6} \approx 19673.7
693: \end{equation}
694: $(\frac{b}{a} =\frac{31119}{2})$,
695: and for $m_{1} =4, m_{2}=3$,
696: \begin{equation} \label{EXACT2}
697: a= \frac{8834477652 \sqrt{3}}{3109375 \pi ^6} \approx 5.11881,
698: b = \frac{33503284082268 \sqrt{3}}{3109375 \pi ^6} \approx 19412.2
699: \end{equation}
700: $(\frac{b}{a}=\frac{11377}{3})$.
701: (Numerical tests showed that neither of these two functions was
702: Schur-convex nor Schur-concave \cite{nielsen}, nor even approximately so.)
703: As indicated, for both these settings of (\ref{exactexact}), our five
704: ({\it non}-HS)
705: indicators all lay within {\it at most} 5\% of their conjectured values
706: \cite{slaterJGP}.
707: (Additionally, the estimated arithmetic average separable hyperarea
708: lay within 5\% of the conjectured value for the $m_{1}=4,m_{2}=3$ case,
709: and 10\% for the other. Since the two functions {\it both} fit the conjectured separable HS volume and hyperarea, any linear combination of them
710: will also. We have found that by weighting the function associated
711: with (\ref{EXACT1}) by 0.570347, and the other function by
712: 0.429653, there is no deviation in the
713: associated non-HS indicators by more than
714: 3.61\%.) The closeness of our estimates to their conjectured
715: values, certainly it would seem, should lend some further support
716: (beyond the original numerical evidence \cite{slaterJGP,slaterPRA}) for the
717: reasonableness of the associated conjectures.
718:
719:
720: Of course, it behooves us and possibly other interested
721: researchers, at this stage, to explore the properties
722: (and desirably derive {\it guiding} principles) of
723: additional candidates
724: for the presumptive three-dimensional weighting function $W$.
725: (We are compelled to note, however, that contrary to our
726: construction of $W$ so far, that it is clear that $W$ has to be
727: simple {\it flat}
728: in some finite neighborhood
729: of the fully mixed state $\lambda_{1}=\lambda_{2}=\lambda_{3}
730: =\lambda_{4}= \frac{1}{4}$ \cite{sepsize1} \cite[sec. 15.5]{ingemar}.
731: So, it seems that $W$ needs to be defined in a {\it piecemeal} manner over
732: differing domains (cf. \cite{pbsCanosa}).)
733:
734:
735: \section{5-dimensional qubit-{\it qutrit} weighting function} \label{n6}
736: Proceeding analogously to our (first)
737: analysis in sec.~\ref{firstanalysis}
738: for the two-qubit
739: case ($n=4$), we sought to obtain a 5-dimensional weighting function $W_{n=6}$
740: for the qubit-qutrit case ($n=6$).
741: Using the conjectures --- based on extensive numerical results --- stated in
742: \cite[sec. VI.D.2]{slaterPRA},
743: in particular, those for the {\it Hilbert-Schmidt} separable
744: volume and separable hyperarea of the
745: $35$-dimensional convex set of $6 \times 6$ density matrices
746: (conjectured to be $(2^{45} \cdot 3 \cdot 5^{13} \cdot 7 \sqrt{30})^{-1}
747: \approx 2.02423 \cdot 10^{-25}$ and
748: $(2^{46} \cdot 3 \cdot 5^{12})^{-1}
749: \approx 1.94026 \cdot 10^{-23}$, respectively), we fitted
750: the weighting function (reproducing the HS separable volume to very high accuracy
751: and the hyperarea to an
752: accuracy of .7\%),
753: \begin{equation}
754: W_{n=6}(\lambda_{1},\lambda_{2},\lambda_{3},\lambda_{4},\lambda_{5},\lambda_{6}) =
755: 986304.
756: (\Sigma_{i<j<k<l<m}^{6} \lambda_{i} \lambda_{j} \lambda_{k} \lambda_{l} \lambda_{m})^{\frac{9}{5}}.
757: \end{equation}
758:
759: Now, in applying this to the {\it Bures} metric, we derived an estimate
760: of the 35-dimensional Bures volume that was
761: 1.82587 times as large as the conjectured value of $2^{-77}
762: \cdot 3 \sqrt{8642986 \pi} \approx 1.03447 \cdot 10^{-19}$
763: \cite[eq. (32)]{slaterJGP}. Our estimate of the 34-dimensional Bures
764: hyperarea was 1.91223 times as large as the conjectured value
765: of $2^{43} \cdot 3 \cdot 5 \sqrt{8462986 \pi} \approx
766: 1.45449 \cdot 10^{-18}$ \cite[eq. (33)]{slaterPRA}.
767: Though, somewhat disappointingly large, these two early
768: results are certainly of the same order of magnitude as the conjectures
769: (and the underlying supporting numerical evidence), and suggest
770: additional research.
771:
772: \section{Remarks}
773: We have sought to determine a certain 3-dimensional weighting function
774: by fitting the conjectured values of the Hilbert-Schmidt
775: separable 15-dimensional
776: volume and 14-dimensional hyperarea \cite[eqs. (41), (42)]{slaterPRA}.
777: It would be of interest to attempt
778: to fit {\it additional} conjectured values as well (such as those for the
779: Bures [minimal monotone] metric). (No conjecture is presently available
780: for the HS separable
781: hyperarea of the 11-dimensional space spanned by the rank-2 density
782: matrices. Otherwise it could be incorporated into our
783: further analyses too --- if
784: the weighting function did not degenerate with two zero eigenvalues
785: present, and if the corresponding 11-dimensional
786: separable hyperarea is not actually zero [cf. \cite{lockhart} \cite[sec. VI.C.4]{slaterPRA}].)
787:
788: Let us --- as was done in \cite{sbz} --- bring to the reader's
789: attention some other studies, such as \cite{sz1,sz2,gb1,gb2,hildebrand}
790: pertaining to volumes of sets of
791: separable and/or positive-partial-transpose states, as well as our
792: more recent analysis \cite{slaterHSPRA}, concerning the Hilbert-Schmidt
793: metric.
794: (It becomes quite clear in this last study, that the separability constraint
795: on two-qubit systems is, in general, {\it quartic} in nature
796: (cf. \cite{wang,ulrich}), thus,
797: to some extent, explaining the associated difficulties in enforcing it --- as wwell as raising certain interesting topological questions
798: (cf. \cite{fortuna}).)
799:
800: In conclusion,
801: let us also make reference to a certain capsule review \cite{mathscinet}
802: in the database MathSciNet of our
803: previous paper \cite{slaterJGP} in this journal.
804: In particular, we add
805: emphasis to the final
806: sentence of the review, devising a response to which
807: comment has been the main
808: motivation of this paper, as well as that of \cite{slaterHSPRA}.
809:
810: ``The paper concerns properties of the convex set of
811: separable two-qubit states. Although the positive partial
812: transpose criterion gives in this very case a concrete answer to the
813: question of whether a given mixed state is separable,
814: the geometry of the 15-dimensional set ${S}$ of separable states
815: is still not well understood.
816:
817: The author analyzes numerically the volume of the set ${S}$
818: with respect to measures induced by several monotone metrics.
819: In particular, he studies the one-parameter family of metrics
820: interpolating between the maximal and the minimal (Bures) metrics.
821:
822: Working with the Bures measure he conjectures that the relative volume of
823: the set of separable states is equal to the silver mean, $\sigma=\sqrt{2}-1$.
824: In a similar way the volume of the $14$-dimensional hyperarea of
825: ${S}$ is estimated with respect to various measures, and
826: the ratios area/volume are analyzed.
827:
828: {\it The conjectures of Slater, based on numerical integration, still
829: await analytical confirmation.}''
830:
831:
832: \begin{acknowledgments}
833: I wish to express gratitude to the Kavli Institute for Theoretical
834: Physics (KITP)
835: for computational support in this research.
836:
837: \end{acknowledgments}
838:
839:
840: \bibliography{FJ5}% Produces the bibliography via BibTeX.
841:
842: \end{document}
843: %
844: % ****** End of file Test2.tex ******