1: \documentclass[twocolumn,superscriptaddress,showpacs]{revtex4}
2: %\documentclass[twocolumn,showpacs]{revtex4}
3: %\documentclass[showpacs]{revtex4}
4: \usepackage{times} %,xspace}
5: \usepackage{amsbsy,amssymb,amsmath,bm}
6: \usepackage{graphicx,color,epsfig,rotate}
7: \usepackage{fancyhdr}
8: \def\one{{\mathchoice {\rm 1\mskip-4mu l} {\rm 1\mskip-4mu l} {\rm
9: 1\mskip-4.5mu l} {\rm 1\mskip-5mu l}}}
10: \def\bbbc{{\mathchoice {\setbox0=\hbox{$\displaystyle\rm C$}\hbox{\hbox
11: to0pt{\kern0.4\wd0\vrule height0.9\ht0\hss}\box0}}
12: {\setbox0=\hbox{$\textstyle\rm C$}\hbox{\hbox
13: to0pt{\kern0.4\wd0\vrule height0.9\ht0\hss}\box0}}
14: {\setbox0=\hbox{$\scriptstyle\rm C$}\hbox{\hbox
15: to0pt{\kern0.4\wd0\vrule height0.9\ht0\hss}\box0}}
16: {\setbox0=\hbox{$\scriptscriptstyle\rm C$}\hbox{\hbox
17: to0pt{\kern0.4\wd0\vrule height0.9\ht0\hss}\box0}}}}
18: \newcommand{\Naturals}{\ensuremath{\mathbb{N}}\xspace}
19: \newcommand{\Integers}{\ensuremath{\mathbb{Z}}\xspace}
20: \newcommand{\Rationals}{\ensuremath{\mathbb{Q}}\xspace}
21: \newcommand{\Reals}{\ensuremath{\mathbb{R}}\xspace}
22: \newcommand{\Complex}{\ensuremath{\mathbb{C}}\xspace}
23: \newcommand{\Zero}{\ensuremath{\mathbb{O}}\xspace}
24: \newcommand{\Punitary}{\ensuremath{\mathbb{I}}\xspace}
25: \newcommand{\UPunitary}{\ensuremath{\mathbb{U}}\xspace}
26: \newcommand{\VPunitary}{\ensuremath{\mathbb{V}}\xspace}
27: \newcommand{\ket}[1]{|{#1}\rangle}
28: \newcommand{\bra}[1]{\langle{#1}|}
29: \newcommand{\kets}[2]{|{#1}\rangle_{{}_{\!\!{#2}}}}
30: \newcommand{\bras}[2]{{}_{{}_{{#2}\!\!}}\langle{#1}|}
31: \newcommand{\slb}[2]{{{#1}^{({#2})}}}
32: \newcommand{\trace}{\mbox{tr}}
33: \renewcommand{\tensor}{\otimes}
34: \newcommand{\vspan}{\mbox{span}}
35: \newcommand{\lowertr}{\mbox{lt}}
36: \newcommand{\diag}{\mbox{dg}}
37: \newcommand{\id}{I}
38:
39: \newcommand{\ignore}[1]{}
40: \newcommand{\mComment}[1]{}
41: \newcommand{\gComment}[1]{}
42: \newcommand{\jComment}[1]{}
43: \newcommand{\rComment}[1]{}
44: \newcommand{\lComment}[1]{}
45:
46: \renewcommand{\mComment}[1]{\textcolor{blue}{Manny: #1}}
47: \renewcommand{\gComment}[1]{\textcolor{red}{Gerardo: #1}}
48: \renewcommand{\jComment}[1]{\textcolor{green}{Jim: #1}}
49: \renewcommand{\rComment}[1]{\textcolor{magenta}{Ray: #1}}
50: \renewcommand{\lComment}[1]{\textcolor{purple}{Rolando: #1}}
51:
52: %\pagestyle{fancy}
53: %\pagestyle{fancyplain}
54: %\footrulewidth 0.4pt
55: %\plainheadrulewidth 0.4pt
56: %\plainfootrulewidth 0.4pt
57: %\lhead{\large LA-UR-04-xxxx}
58: %\chead{ \today}
59: %\rhead{\sl submitted to Physical Review A}
60: %\cfoot{\sc\thepage}
61: %\lfoot{}
62: %\rfoot{}
63:
64: \begin{document}
65:
66: \title{The BCS - BEC Crossover In Arbitrary Dimensions}
67:
68: \author{Zohar Nussinov}
69: \email{zohar@viking.lanl.gov}
70: \affiliation{Department of Physics, Washington University, St. Louis,
71: MO 63160-4899, USA}
72: \affiliation{Theoretical Division,
73: Los Alamos National Laboratory, Los Alamos, NM 87545, USA}
74:
75: \author{Shmuel Nussinov}
76: \email{nussinov@post.tau.ac.il}
77: \affiliation{School of Physics
78: and Astronomy, Tel Aviv University,
79: Ramat-Aviv, Tel Aviv 69978, Israel}
80:
81: \date{Received \today }
82:
83:
84: \begin{abstract}
85: Cold atom traps and certain neutron star layers may contain fermions
86: with separation much larger than the range of pair-wise potentials yet
87: much shorter than the scattering length.
88: Such systems can display {\em universal} characteristics independent of
89: the details of the short range interactions.
90: In particular, the energy per particle is a fraction $\xi$ of the
91: Fermi energy of the free Fermion system.
92: Our main result is that for space dimensions D smaller than two and
93: larger than four a specific extension of this problem readily yields
94: $\xi=1$ for all $D \le 2$ whereas $\xi$ is rigorously
95: non-positive (and potentially
96: vanishing) for all $ D \ge 4$. We discuss the
97: $D=3$ case. A particular unjustified recipe suggests $\xi=1/2$ in
98: $D=3$.
99: \end{abstract}
100:
101: \maketitle
102: %Sec. I
103:
104:
105: \section{Introduction}
106:
107:
108: Bose-Einstein condensates (BEC) in dilute atomic gases were achieved
109: in 1995 for rubidium, sodium, and lithium
110: \cite{gas95}. Recently, with the production of cold atomic gases
111: close to the Feshbach resonance \cite{RGJ}, these systems
112: provided a vehicle for the study of the BCS to BEC
113: crossover. The (BCS) superconducting and
114: Fermi superfluid are mixed systems harboring
115: quasi-free and pair correlated fermions. In the dilute strong
116: coupling molecular BEC, all fermions are relatively tightly
117: bound into pairs forming a unique macroscopic state.
118: Several theoretical studies have been
119: carried, e.g. \cite{DH}. Many of these
120: works report on very interesting universal aspects of the
121: BCS to BEC transition. In such dilute systems with an
122: inter-particle separation much greater than
123: the range of the pair-wise potential yet far shorter than the scattering
124: length, the value of the energy density at this crossover
125: is independent of specific details. This {\bf universal}
126: energy density is only dimension dependent.
127: Direct results on the
128: three dimensional system are hard to obtain
129: and numerical works have been extremely
130: valuable. As is well known
131: from a multitude of other arenas, a dimensional
132: generalization of original problems posed in three
133: dimensions to arbitrary dimensions
134: often allows us an analytical access
135: to original three dimensional problems (e.g.
136: the well known $\epsilon = 4-D$ expansion,
137: with $D$ the spatial dimension,
138: which has been extremely fruitful in statistical
139: and quantum field theories, e.g. \cite{epsilonexpansion}).
140: Here, we follow suite and couch the BCS to BEC crossover problem
141: in arbitrary dimensions ($D$). We report new results
142: on the energy per particle
143: at the onset of this crossover in such an arbitrary dimension
144: (suitably defined by the appearance of zero-energy two particle bound states).
145: This extension enables us to examine crossovers
146: in low dimensions where no condensed phases occur
147: due to the enhanced role of low energy fluctuations.
148: We will illustrate, both by exact variational bounds and
149: normalization considerations, that
150: in all dimensions $D \ge 4$ the energy per particle
151: at the onset of the transition
152: is rigorously bounded from above by zero.
153: By contrast, due to localization
154: tendencies in low dimensions ($ D \le 2$),
155: each particle carries, on average, the mean
156: energy of a free Fermi system
157: at the appearance of the first two
158: particle bound states. The physically
159: pertinent well known question \cite{ChallProb}
160: concerns the energy per particle
161: in actual transitions
162: occurring in $D=3$ dimensions. The determination
163: of the energy per particle here is far more difficult.
164: In $D=3$, we depart from the more rigorous
165: results in higher and lower dimensions and
166: present a heuristic argument for
167: recently reported numerical results \cite{CCPS}.
168: These independent heuristic arguments bolster
169: the the result attained by dimensional interpolation.
170: The average of the exact bounds on the
171: energies in $D=2$ and $D=4$,
172: which is half the free fermion energy per particle,
173: is not far removed from the
174: numerical results reported in three
175: dimensions \cite{CCPS}.
176: In \cite{CCPS} the fraction of a half, derived here by
177: (i) dimensional interpolation and (ii)
178: an independent heuristic argument, is
179: replaced by $\xi \approx 0.44$.
180:
181: \section{Outline}
182:
183:
184: We begin, in section(\ref{form}), by formulating
185: the problem and briefly reviewing numerical results.
186: Next, in section(\ref{Dim}), we
187: turn to an exact variational bound concerning this
188: transition in high dimensions and an easier
189: result relating to bound states in
190: low dimensions.
191: Finally, in section(\ref{threeD}),
192: we present heuristic arguments for $D=3$.
193: These further correlate the
194: average energy per particle
195: with the observed differences
196: seen in finite size system
197: with an even or odd particle
198: number. In a brief appendix,
199: we provide the specifics of the
200: exact solution of the zero energy
201: bound state problem in a
202: D-dimensional spherical potential
203: well. Throughout the text,
204: in order to make the physics
205: and scaling very transparent, we will often
206: use simpler forms. Nevertheless, at the end
207: of all calculations, we will demonstrate that our
208: results go unchanged with the insertion
209: of the exact zero-energy bound state
210: solution.
211:
212:
213: \section{Formulating the Problem: Introductory and General Comments}
214: \label{form}
215:
216: We start by writing down the general Hamiltonian.
217: describing $N=2n$ spin-1/2 Fermions in a box of size $V=L^3$,
218: \begin{eqnarray}
219: H(g) & = & \sum_{I} \frac{{\bf{p}}_I^2}{2m} - g \sum_{I>J} V(|{\bf{r}}_I-{\bf{r}}_J|) \nonumber
220: \\ & = & \sum_{i} \frac{{\bf{p}}_i^2}{2m} + \sum_{j} \frac{({\bf{p}}'_j)^2}{2m} \nonumber
221: \\
222: & \;\;\;\;\;\;\;\; - & g \sum_{ij} V(|{\bf{r}}_i-{\bf{r}}'_j|).
223: \label{Hamiltonian} % { Eq. 1 }
224: \end{eqnarray}
225: Here, $ I,J $ (or $ i,j $) run from 1 to $N$ (or $n$) respectively.
226: The phase space coordinates ${\bf r}_i,{\bf p}_i$ and ${\bf r}'_j,{\bf p}'_j$
227: are the positions and momenta of the ``spin up" and ``spin down" atoms
228: respectively and $[-g V({\bf r}_i,{\bf r}'_j)]$ is the attractive short
229: range potential in mixed pairs. These conventions for the
230: coordinates [using upper case
231: characters to describe the total system (both spin up and spin
232: down) quantities and the use of lower case characters
233: (either ``primed'' or ``un-primed'') for spin-down and spin-up
234: coordinates] will be consistently employed throughout this work.
235:
236: Atoms of identical spin polarization cannot be
237: in a relative S wave state where the short range
238: interactions operate and thus, in Eq.(\ref{Hamiltonian}),
239: we set $V({\bf r}_i,{\bf r}_j)
240: = V({\bf r}'_k,{\bf r}'_l)= 0$. The coupling
241: $g$ is tuned to $g^*$ where the first (zero energy, S wave)
242: two-body
243: bound state appears and the scattering length $a$ diverges.
244: An illustration of the standard two-particle zero energy
245: bound state in three dimensions
246: is provided in Fig.(\ref{Fig1bound}).
247: With this identification of $g^{*}$ at hand \cite{ChallProb},
248: {\em the BCS to BEC crossover problem is now formally
249: defined in arbitrary dimension $D$},
250: although, due to the enhanced role of
251: low energy fluctuations,
252: actual condensates may form only in $D>2$.
253: Our physical interest is in $D=3$.
254: To make $g$ dimensionless in Eq.(\ref{Hamiltonian},
255: we scale V by $(m r_0^2)^{-1}$ with $r_0$ the range
256: of the potential and $m$ the particle mass.
257: We assume a dilute system with inter-particle separation $d=1/k_F$,
258: much larger than $r_0$. In practical terms,
259: this implies that
260: we will always consider the limit $r_{0} \to 0$.
261: To avoid the well known formal ``collapse problem'' in which
262: all particles may sit within the attractive potential
263: well [as the attractive $V({\bf r}_{i}, {\bf r'}_{j}) <0$ for
264: $|{\bf r}_{i} - {\bf r'}_{j}| \le r_{0}$, we may
265: envision a state in which all particles sit
266: within a sphere of radius $r_{0}$],
267: leading to a divergent negative energy density,
268: the problem is formulated with the provision that
269: the limit $r_{0} \to 0$ is taken before the thermodynamic
270: limit is considered \cite{ChallProb, Baker}. With these
271: definitions in tow, the problem emulates what transpires
272: in many dilute fermionic systems (cold atom traps,
273: neutron star layers) with separations
274: far larger than the range of the pair potentials
275: yet far smaller than the scattering length.
276:
277:
278:
279:
280: \begin{figure}
281: \centerline{\psfig{figure=bound.eps,height=
282: 5cm,width=6cm,angle=0}}
283: \caption{A schematic (thick lines) of the binding spherically symmetric
284: potential $V({\bf{r}}_{i} - {\bf{r'}}_{j}) \equiv V({\bf{r}}_{ij})$
285: (in three dimensions) which is non-zero only inside a sphere of
286: radius $r_{0}$ about the origin, $V({\bf{r}}) = - V_{0}
287: \Theta (r_{0} - |{\bf{r}}|)$.
288: A cartoon of a scaled very weakly bound s-wave state
289: ($\chi({\bf{r}}_{ij}) = |{\bf{r}}_{ij}| \phi({\bf{r}}_{ij})$)
290: is shown. To attain exactly one zero energy bound state we
291: need to fit exactly a quarter period for
292: $\chi(|{\bf{r}}|) \sim \sin \kappa r$
293: and set $ \kappa r_{0} = \frac{\pi}{2}$ with $\kappa
294: = \sqrt{2 \mu V_{0}}/\hbar$.
295: Here, $\mu = m/2$ is the reduced mass of a fermionic pair.
296: The zero-energy bound state wavefunction in general
297: dimension $D$ is derived in the appendix.}
298: \label{Fig1bound}
299: \end{figure}
300:
301:
302:
303:
304:
305:
306:
307:
308:
309: In three dimensions,
310: scaling arguments then imply a universal form for the energy per
311: particle
312: \begin{equation}
313: \frac{E}{N} = (\frac{3}{5} \, \varepsilon_{F}) \xi,
314: \label{TotEnergyNbody} %{Eq.2 }
315: \end{equation}
316: with $\varepsilon_F$ the Fermi energy of the free fermi system. \cite{scale}
317: The fraction $\xi$ is independent of the specific potential
318: $V({\bf r})$ chosen and is the same for
319: all short range potentials which have a zero energy S-wave bound state.
320: Cold dilute Fermionic atoms display many fascinating universal (and not
321: universal) features; see, e.g., \cite{DH}. Here we will focus just
322: on the above $\xi$ parameter.
323:
324:
325: In actual traps, an excited zero
326: energy state $\psi^e_0({\bf r})$ plays a key role. The continuum
327: states are effected by the extended zero energy state
328: and have negligible overlap with lower,
329: tightly bound, states of size $\sim r_0$ which can be
330: ignored unless we consider
331: very long time scales. As any zero
332: energy S wave bound state has the same form
333: $\psi_0({\bf r})= \psi^e_0({\bf r})= A/|{\bf r}|$
334: outside the range of the potential
335: for $|{\bf r}| > r_0$, we assume just {\it one} zero energy state. For $g=0$,
336: there are two free species with $n=N/2$ fermions in each, and
337: relative to Eq.(\ref{TotEnergyNbody}), with the standard results
338: \begin{equation}
339: \frac{E_{free-fermion}}{N}= \frac{3}{5} \varepsilon_F, \;\;\;
340: n= \frac{L^3
341: (4\pi/3)p_F^3}{(2\pi)^3},
342: \label{freespecies} %{Eq. 3 }
343: \end{equation}
344: where $p_F$ and $ \varepsilon_F= \frac{p_F^2}{2m}$
345: are the free particle Fermi momentum and Fermi energy respectively.
346: Throughout, we set $\hbar$=1. Thus, by definition,
347: \begin{equation}
348: \xi=1 \;\;\; {\rm for} \;\; {\rm g=0}.
349: \label{xi1} %{Eq. 4 }
350: \end{equation}
351: For small $g>0$ (weak attractive potentials), the BCS wave function is
352: adequate. By contrast, for strong coupling
353: $g|V(|{\bf r}| < r_{0})| >> \varepsilon_{F}$,
354: up-down spin pairs tightly bind into dimers
355: of sizes of order, $r_0$, the range of the potential V.
356: The $n=N/2$ dimers in the deepest bound state then behave
357: as point-like bosons and undergo BEC (Bose-Einstein
358: Condensation) to the traps' ground state with $p_{dimer}\sim 0$ so that
359: \begin{equation}
360: \frac{E}{N} = - \frac{|B.E.|}{2} \;\;\; {\rm if} \;\; g \rightarrow \infty,
361: \label{dimers} %{Eq. 5 }
362: \end{equation}
363: with $B.E.$ the binding energy.
364:
365: The focus of the current work is on
366: the crossover between the BCS and BEC regimes
367: wherein the initial Cooper pairs become tighter dimers
368: \cite{LeggetEngelbrecht}.
369: The existence of a bound state renders
370: a weak-coupling perturbation series inappropriate.
371: Similarly, strong coupling/tight binding schemes also fail.
372: As $g$ decreases so does the binding energy. Once $|B.E.| <\varepsilon_F$, the size of the bound state becomes larger
373: than the average inter-particle distance: $O(m \varepsilon_F)^{-1/2} \sim d$ and antisymmetrization
374: of spin up (and of spin down) atoms---which is negligible for small dimers---raises
375: $E/N$ above $\frac{|B.E.|}{2} =0$.
376:
377: The difficulty and interest of the problem motivated several calculations.
378: A semi-analytic approach utilizing the Pad\'{e} approximation has
379: been discussed at length in Ref. \cite{Baker}. Numerical Fixed
380: Nodal Planes-Green Function-Monte Carlo (FNPGFMC or MC)
381: calculations were used
382: to estimate $E(N)$\cite{CCPS}. The ground state wave function of $N$ fermions in a
383: periodic box of size $ L $ was estimated by evolving
384: (in imaginary time $\tau = - it$) an initial trial
385: function $\Psi({\bf r}_i,{\bf r'}_j;\tau=0)$ with
386: $\exp (-H \tau)$ . The short range potential
387: $V({\bf r}_i-{\bf r}'_j)$ was chosen to have precisely one
388: zero energy bound state.
389: The lowest energy obtained to date via these methods corresponds
390: to $\xi = 0.44$ in Eq.(\ref{TotEnergyNbody}) which is
391: consistent with present imprecise experimental values.
392: The $L/d \sim n^{1/3}=80^{1/3} \approx 3.5$ used
393: there implies a box size $L$ smaller
394: than the
395: (infinite) scattering length. This still allows
396: for an extraction of reliable data.
397:
398: In general, the
399: ground state energy $E/N$ decreases continuously with increasing
400: $g$. The instability against the formation of Cooper pairs for any
401: attractive potential adds a singular $\sim \exp (-g_{c}/g)$
402: (with $g_{c}$ a constant) term to E/N with the continuity in $g $ maintained.
403:
404: %Section II.
405: \section{Extension to arbitrary spatial dimensionality D}
406: \label{Dim}
407:
408: While the original problem is posed in $D=3$ dimensions, it is
409: instructive to address it in any ({\em continuous}) D with the simple
410: extension of Eqs.(\ref{TotEnergyNbody},\ref{freespecies}),
411: \begin{eqnarray}
412: \frac{E}{N} =\xi \frac{D}{(D+2)} \varepsilon_F.
413: \label{trivDim}
414: \end{eqnarray}
415:
416: To set the ground for future notation, we mention that the simple
417: $D$ dimensional extension of the latter part of
418: Eq.(\ref{freespecies}) has the total fermion
419: number (per spin flavor) within a Fermi sphere of radius $p_{F}$ given by
420: $n= \Omega_{D}L^{D} p_{F}^{D}/(2 \pi)^{D} $. Here,
421: $\Omega_{D} = \pi^{D/2}/\Gamma(\frac{D}{2} +1)$
422: is the volume of the D dimensional unit sphere.
423:
424:
425: In what follows, we
426: (i) establish by localization tendencies in dimensions $D \le 2$
427: that $\xi(D \le 2) ={\xi}(g=0)= 1$. We then (ii) illustrate
428: that $\xi( D \ge 4) \le 0$ by relying on divergence, at short length
429: scales, of the wave function normalization. This simple scaling
430: consideration is then made rigorous for all continuous dimensions
431: $D >4$ by employing a variational bound with the
432: use of an ``orbital'' Slater determinant wavefunction.
433: (Regretfully, we have not been
434: able to devise a wavefunction which will lead to new
435: stringent variational bounds on the three-dimensional
436: problem.)
437:
438:
439:
440: Putting all of the pieces together, we find by
441: extending the problem to any dimension $D$
442: while insisting on having precisely one zero energy bound state
443: that $\xi$ is non-trivial only
444: within the interval
445: $2 <D < 4$. From this perspective, the only interesting integer
446: dimensionality is, in fact, the original D=3!
447:
448:
449: There may be other extensions of the problem which might be non-trivial
450: in all dimensions. Many systems simplify at $D \rightarrow \infty$ and
451: a mean field approach may apply there also to Fermionic problems \cite{MV}.
452: It has been argued that a certain class of diagrams
453: dominates in this limit the perturbative series in an appropriate
454: effective field theory and can be analytically summed in certain fashions
455: to yield ${\xi}(D={\infty}) = \frac{4}{9}$ \cite{Steel} or $\frac{1}{2}$
456: \cite{SKC}.
457:
458: In what briefly follows, we employ
459: the above continuation with a primary focus on energetics (i.e., the
460: existence of a zero energy S wave bound state) rather than
461: on the infinite scattering length. We find the resulting simple
462: values $\xi = 1$ for all (continuous) dimensions smaller than
463: $D=2$ and $\xi \le 0$ for all $D \ge 4$ interesting on their own right.
464:
465:
466:
467: \subsection{Low Dimensions ($D \le 2$)}
468:
469:
470: Localization tendencies present in low dimensions afford us with
471: a direct result. Purely attractive
472: potentials $V({\bf r})$ have zero energy bound states in one {\it and} two dimensions
473: \cite{LL}, for any strength and $g$. It follows that
474: a bound state appears immediately as
475: the two body attraction is introduced, i.e.
476: $g^* = 0^{+}$. As the energy fraction ${\xi}(g)$ is
477: continuous in any dimension D, we find that
478: \begin{equation}
479: \xi ={\xi}(g=0)= 1 \; , \;\; D \le 2.
480: \label{D2} %{Eq 6}
481: \end{equation}
482: Similar incarnations of localization tendencies
483: in low dimensions have proven very useful in
484: other arenas (e.g. \cite{gang4} in which
485: a scaling theory of localization suggested $D=2$
486: as the marginal dimension for the appearance of
487: metallic states).
488:
489:
490: \subsection{High Dimensions ($D \ge 4$)}
491:
492: \subsubsection{Energy attained by Normalization conditions}
493: \label{norm}
494:
495: Next, let us assume that the potential V vanishes outside the range $r_0$.
496: Volume normalization factors aside, the spin-up spin-down, zero energy,
497: pair wave function for relative separation
498: $|{\bf r} |> r_0 $ is $ \phi({\bf r})
499: = \frac{A_{>}}{|{\bf r}|^{(D-2)}}$, with $A_{>}$
500: a dimension dependent numerical constant.
501: The following is
502: a very interesting observation: the integral representing the normalization of this
503: zero energy wave function, $ [\int d^D {\bf r} |\phi({\bf r})|^2]$ diverges at small
504: ${\bf r}$
505: for $D \ge 4$. We consequently find that in
506: $D \geq 4$ dimensions most of the normalization
507: of the zero
508: energy bound state wave function is concentrated near
509: $|{\bf r}| = r_{0}^{+}$ and the dimers formed
510: are actually compact---just like for large coupling in D=3.
511: Hence, we can again neglect the overlap of the different dimers and attendant kinetic
512: energy due to antisymmetrization in the spin up coordinates
513: $\{{\bf r}_i\}$ and separately in the spin down positions $\{{\bf r}'_j\}$.
514: The strong coupling result Eq. (\ref{dimers}) applies and we have:
515: $ E/N = - \frac{|B \cdot E|}{2} \le 0 $. Thus,
516: we find that $\xi$ may vanish (or become negative)
517: for all dimensions greater than four,
518: \begin{equation}
519: \xi \le 0 \; , \;\; D \ge 4.
520: \label{4D} % {Eq.7}
521: \end{equation}
522: In the marginal dimension $D=4$,
523: the size of the dimers is just barely
524: $r_{0}$ (where the attractive potentials
525: operate) and the upper bound may be more easily saturated
526: with a potentially vanishing $\xi(D=4)$.
527:
528:
529: \subsubsection{A Variational Upper Bound}
530:
531: To substantiate the above arguments, and directly derive
532: Eq.(\ref{4D}), we next employ a 2n-body ansatz wave
533: function $\Psi({\bf r}_1,{\bf r}_2,{\bf r}_i, \ldots {\bf r}_n;
534: {\bf r}'_1, \ldots ,{\bf r}'_j, \ldots {\bf r}'_n)$ and show
535: that in all continuous dimensions $D > 4$, we have $H|\Psi \rangle=0$.
536: The ansatz $\Psi$ has the form of a
537: Slater ``orbital" determinant. It represents the state of n
538: dimers, say, $\overline{\phi}({\bf r}_1-{\bf r}'_1) \ldots \overline{\phi}({\bf
539: r}_n-{\bf r}'_n)$ with $\overline{\phi}(|{\bf r}_i-{\bf r}'_j|)$ the zero
540: energy bound state, antisymmetrized over all the n! permutations:
541: $ P=i \rightarrow (p(i)), \; i=1, \ldots n$ of the n spin-up
542: atoms: \setcounter{equation}{7}
543: \begin{equation}
544: \Psi=\sum_{P} (-1)^P \Phi(P)\, ; \;\;\;\; \Phi(P) =
545: \prod_{i=1...n} \overline{\phi}({\bf r}_i-{\bf r}'_{p(i)})
546: \label{PsiSigmaEq8}
547: \end{equation}
548: with $(-1)^P$ the parity of the permutation $P$.
549: Here, $\overline{\phi}({\bf r}_i-{\bf r}'_{p(i)})$
550: represent normalized, zero energy, pair wavefunctions.
551:
552: The further required antisymmetrization over the n spin-down atoms
553: simply scales $\Psi$: ~ $\Psi \to [(n!) \Psi]$ and is redundant.
554: We note that $\Psi$ can be expressed as
555: a determinant,
556: \begin{eqnarray}
557: \Psi(\bf{r}_{1}, ...., \bf{r}_{n}; \bf{r'}_{1}, ..., \bf{r'}_{n})
558: = \nonumber
559: \\ \det \left( \begin{array}{cccc}
560: \overline{\phi}(\bf{r}_{1} - \bf{r'}_{1}) &
561: \overline{\phi}(\bf{r}_{1}- \bf{r'}_{2}) & ... &
562: \overline{\phi}(\bf{r}_{1} - \bf{r'}_{n}) \\
563: \overline{\phi}(\bf{r}_{2}- \bf{r'}_{1})
564: & \overline{\phi}(\bf{r}_{2} - \bf{r'}_{2}) & ... &
565: \overline{\phi}(\bf{r}_{2} - \bf{r'}_{n}) \\
566: . & . & ... & . \\
567: \overline{\phi}(\bf{r}_{n} - \bf{r'}_{1}) &
568: \overline{\phi}(\bf{r}_{n} - \bf{r'}_{2}) & ... &
569: \overline{\phi}(\bf{r'}_{n} - \bf{r'}_{n})
570: \end{array} \right).
571: \label{DetMijEq9}
572: \end{eqnarray}
573: We can use the variational ansatz
574: $\Psi$ in any dimension D to obtain a variational upper
575: bound on the N=2n-body ground state energy and on the corresponding
576: $\xi$:
577: \begin{equation}
578: E(2n) = \Big[ \frac{2n D}{D+2} \varepsilon_F \xi \ \Big] \le \,
579: \frac{\langle \Psi|H|\Psi \rangle}{\langle \Psi|\Psi \rangle}.
580: \label{E2n2nEq10}
581: \end{equation}
582: Clearly,
583: \begin{equation}
584: \langle \Psi|H|\Psi \rangle = \sum_{P,P'} \langle \Phi(P)|
585: H|\Phi(P') \rangle,
586: \label{PsiHPsiEq11}
587: \end{equation}
588: with a similar expression for $\langle \Psi|\Psi \rangle$.
589:
590: The local potential limit, say, $r_0 \rightarrow 0$ in the
591: ``square well'' spherical potential
592: \begin{equation}
593: V({\bf r})= - V_0 \Theta(r_{0}-|{\bf r}|)
594: \label{VrV0Eq12}
595: \end{equation}
596: implies for all dimensions $D>2$ the following ``selection rule"
597: for non-vanishing $\langle \Phi(P')|V_{i,j}|\Phi(P) \rangle$
598: matrix elements appearing when the 2n-body Hamiltonian (1) is
599: substituted in Eq. (\ref{PsiHPsiEq11}) above.
600:
601: Before going into the details of the variational
602: calculations, let us quickly give the reader a
603: glimpse of the final results derived
604: below and their implication.
605: The calculation detailed in this
606: subsection illustrates that
607: for the value of $V_{0}$ in Eq.(\ref{VrV0Eq12})
608: (the analog of $g$ in Eq.(\ref{Hamiltonian}))
609: which secures a single zero-energy bound,
610: state we have a variational state (i.e. $| \Psi \rangle$
611: of Eq.(\ref{DetMijEq9})) for which the corresponding
612: energy $E_{var}=0$.
613: This result then implies (by Eqs.(\ref{trivDim},\ref{E2n2nEq10}))
614: that $\xi \le 0$.
615:
616: For any $V_{i,j}$ among the $n^2$ spin-up--spin-down
617: potentials we have, for all $D>2$ with a local $V({\bf r})$ producing
618: one zero energy bound state, the following relation
619: \begin{equation}
620: \langle \Phi(P')|V_{i,j}|\Phi(P) \rangle =0 {\rm \; if \; no \;} p(i)=j
621: {\rm \; \underline{and} \; no\;} p'(i)=j.
622: \label{PhiP'VijEq13}
623: \end{equation}
624: To prove this claim, we perform first (among the 2n $\; d^D \, {\bf r}_k
625: d^D \, {\bf r}'_l $ integrations involved in evaluating the above
626: matrix element) the integrals $\int d^D \, {\bf r}_i \int d^D
627: \, {\bf r}'_j$ over the coordinates ${\bf r}_i$ and ${\bf r}'_j$
628: appearing in the particular potential term
629: $V({\bf r}_i-{\bf r}'_j)$ considered.
630:
631: Both ${\bf r}_i$ and ${\bf r}'_j$ appear in the product $\Phi(P)$
632: of Eq.(\ref{PsiSigmaEq8}) but, by assumption, in different
633: $\phi$ factors. The same holds for $\Phi(P')$. Thus, the
634: integration over ${\bf r}_i$ and ${\bf r}'_j$ is of the form,
635: \begin{eqnarray}
636: \int d^3{\bf r}_i \int d^3{\bf r}'_j
637: [\overline{\phi}({\bf r}_i-{\bf r'}_{p'(i)})
638: \overline{\phi}({\bf r}_{p'^{-1}(j)}-{\bf r}'_j) \nonumber
639: \\ \times V({\bf r}_i-{\bf
640: r}'_j)
641: \overline{\phi}({\bf r}_i-{\bf r'}_{p(i)})
642: \overline{\phi}({\bf
643: r}_{p^{-1}(j)}-{\bf r'}_j)], \label{integralsEq14}
644: \end{eqnarray}
645: with $p^{-1}$ the permutation inverse to p.
646: The condition for having one zero energy S wave bound state in the
647: square well potential of (\ref{VrV0Eq12}) in the D=3 case is
648: well known (see Fig.(\ref{Fig1bound})),
649: \begin{equation}
650: V_0= \hbar^2 \frac{(\pi^2/4)}{mr_0^2} \;. \label{VOEq15}
651: \end{equation}
652: By dimensional considerations, the same condition for one zero
653: energy bound state holds in all dimensions $D>2$ wherein the constant
654: $\pi^2/4$ is replaced by some other, dimension dependent,
655: numerical factor $c_D$,
656: \begin{equation}
657: V_0 = \hbar^2 \frac{c_D}{mr_0^2}. \label{VOcd}
658: \end{equation}
659: In the general $D$ dimensional problem,
660: $c_{D}$ of Eq.(\ref{VOcd}) is fixed by normalization of
661: the zero-energy wavefunction $\phi({\bf r})$.
662: In the appendix, we briefly present the
663: solution to the a zero-energy bound state
664: in a $D$-dimensional spherical potential
665: well.
666:
667:
668:
669: By changing variables $({\bf r}_i,{\bf r}'_j) \rightarrow ({\bf r}_i,
670: {\bf r}_{i,j})$ with ${\bf r}_{i,j} = ({\bf r}_i-{\bf r}'_j)$,
671: the local $V({\bf r}_{i,j})$ implies that ${\bf r}_i={\bf r}'_j$
672: in the arguments of all the four $\phi$ factors appearing above
673: which generically keep finite arguments and bound $\phi$ values.
674: Neglecting small variations of $\phi({\bf r})$ away from ${\bf r}
675: \sim 0$, the integral $\int d^3{\bf r} \, V({\bf r})$ over the D
676: dimensional sphere of radius $r_0$ where the square well
677: potential is non-vanishing yields the factor
678: $V_0 r_0^D \Omega_{D}$. This multiplicative factor
679: is, by virtue of Eq.(\ref{VOcd}), proportional to $[c_D r_0^{(D-2)}]$.
680: In dimensions $D>2$, this factor vanishes as $r_{0} \to 0$
681: and, as claimed earlier, so does the complete matrix
682: element. This is {\it not} the case in dimensions $D<2$ for which
683: $\xi=1$.
684:
685: Next, we consider dimensions $D>4$. In this case, the zero energy
686: bound states
687: \begin{eqnarray}
688: \overline{\phi}({\bf r}_{i}, {\bf r'}_{p(i)})
689: = {L^{-D/2}} \phi({\bf r}_{i} - {\bf r'}_{p(i)}),
690: \end{eqnarray} with
691: (see the Appendix for details),
692: \begin{eqnarray}
693: \phi({\bf r}_i-{\bf r}'_{p(i)}) = A_{>}|{\bf r}_i-{\bf
694: r}'_{p(i)}|^{-(D-2)} \nonumber
695: \\ {\mbox{for}} \; |{\bf{r}}| \equiv
696: |{\bf r_{i}- r`_{p(i)}}| >
697: r_0 ;\nonumber
698: \\ {\mbox{and}} \;\; \phi({\bf r}_i-{\bf r}'_{p(i)})
699: \sim A_{>} (r_0^{-(D-2)}) \nonumber
700: \\ {\rm for} \;
701: 0<|{\bf r}| < r_0,
702: \label{D>4EQ16}
703: \end{eqnarray}
704: where $A_{>} \approx (r_0^{(D/2)-2})$. Exact forms
705: are provided by Eqs.(\ref{bigr},\ref{smallr},
706: \ref{BesselV}, \ref{amp_ratios}). The comparison
707: between Eq.(\ref{D>4EQ16}) and the exact forms
708: is provided in Eq.(\ref{fz}).
709: The scaling of the results with $r_{0}$ (which we will
710: shortly obtain) becomes more transparent
711: with the use of Eq.(\ref{D>4EQ16}). With the
712: incorporation of Eq.(\ref{fz}) and its ensuing discussion,
713: the results which we will obtain using Eq.(\ref{D>4EQ16})
714: will further enable
715: as a rigorous upper bound on the matrix value elements to be discussed.
716: As seen from Eq.(\ref{D>4EQ16}), the wavefunctions $\phi$ are
717: strongly localized at $|{\bf r}| \sim r_0$ implying stronger
718: selection rules for non-zero matrix elements:
719: \begin{equation}
720: \langle \Phi(P')|\Phi(P) \rangle = 0 \;\; {\rm unless} \; P=P',
721: \label{PhiPhi17a}
722: \end{equation}
723: \vspace*{-.15in}
724: and
725: \begin{equation}
726: \langle \Phi(P')|V(r_i-r'_j)|\Phi(P) \rangle = 0 \;\; {\rm unless} \;
727: j=p(i)=p'(i). \label{PhiVPhi17b}
728: \end{equation}
729:
730: Eq.(\ref{PhiVPhi17b}) implies that also $\langle \Psi|H|\Psi\rangle$,
731: the numerator of Eq.(\ref{E2n2nEq10}), has only diagonal
732: contributions. Namely,
733: \begin{equation}
734: \langle \Phi(P')|H|\Phi(P) \rangle = 0 \;\; {\rm unless} \; P=P' \, .
735: \label{P=P'}
736: \end{equation}
737:
738: To illustrate Eq.(\ref{PhiPhi17a}), let us assume that $P$ and $P'$ differ
739: minimally: $p(i)=p'(i)$ for all $i>2$, but $p(1)=1$, $p(2)=2$
740: and $p'(1)=2$, $p'(2)=1$. The overlap in Eq.(\ref{PhiPhi17a}) is
741: then
742: \begin{eqnarray}
743: L^{-2D} \int d^D{\bf r}_1 d^D{\bf r}'_1 d^D{\bf r}_2
744: d^D{\bf r}'_2
745: \phi({\bf r}_1-{\bf r}'_1) \phi({\bf r}_2-{\bf r}'_2) \nonumber
746: \\ \times
747: \phi({\bf r}_1-{\bf r}'_2) \phi({\bf r}_2-{\bf r}'_1) \; \nonumber
748: \\ L^{-nD} \prod_{i,j>2} \int d^{D}{\bf r}_{i} d^{D}{\bf r'}_{j}
749: \langle \phi(i,p(i))|\phi(i,p(i)) \rangle. \label{n-2ProdEq.18}
750: \end{eqnarray}
751:
752:
753: The last line of Eq.(\ref{n-2ProdEq.18}) denotes a normalized
754: integral (equal to one). In the first integral over the 4 variables
755: $\{{\bf r}_{1,2}, {\bf r'}_{1,2}\}$, the number of integration
756: variables is equal to the number of wavefunctions
757: $\{\phi({\bf r}_{i} - {\bf r'}_{j})\}$ appearing in the
758: integrand. For each of the two particle
759: wavefunctions $\phi({\bf r}_{i} - {\bf r'}_{j})$, we insert
760: Eq.(\ref{D>4EQ16}) (or the exact Eqs.(\ref{bigr},\ref{smallr})
761: derived within the appendix) with $A_{>} \approx r_{0}^{(D/2)-2}$.
762: Insofar as scaling with $r_{0}$ is concerned,
763: all integrals over products of wavefunction forms
764: of Eq.(\ref{D>4EQ16}) multiplying $A_{>}$ amount to
765: product of either integrals of the type $I_{>} = \int_{|{\bf r}|>r_{0}}
766: d^{D} {\bf r} |r|^{2-D}$ or of the form $I_{<} = r_{0}^{2-D}
767: \int_{|{\bf r}| <r_{0}} d^{D} {\bf r} $.
768: Constant factors aside, both $I_{>}$ and $I_{<}$ scale as $r_{0}^{2}$.
769: The four normalization prefactors of $A_{>}$
770: (with $A_{>} \approx r_{0}^{(D/2)-2}$),
771: originating from the four factors of $\phi$ in the top two lines of
772: Eq.(\ref{n-2ProdEq.18}),
773: lead to an additional factor of $r_{0}^{2(D-4)}$ in tow. Thus,
774: in the final analysis, when the integrals are segregated into
775: all possible terms for $|{\bf r}_{i} - {\bf r'}_{j}|
776: > r_{0}$ and for $|{\bf r}_{i} - {\bf r'}_{j}| < r_{0}$ and
777: Eq.(\ref{D>4EQ16}) or Eqs.(\ref{bigr}, \ref{smallr})) are
778: inserted, we find that all terms scale as ${\cal A} \approx r_{0}^{2D}
779: \to 0$ as $r_{0} \to 0$.
780: This signals that the factor in the first two lines
781: of Eq.(\ref{n-2ProdEq.18}) vanishes as $r_{0} \to 0$
782: and thus so does the entire overlap of Eq.(\ref{PhiPhi17a}).
783: We can readily verify that for larger mismatches between $P$
784: and $P'$ the overlap vanishes as a higher power of $r_0$.
785: Now, in the diagonal case, with $p'(1) =1$ and $p'(2)=2$
786: the product of the four normalization
787: factor cancels and $\langle \Phi(P)| \Phi(P) \rangle =1$.
788: \cite{explainnormal}
789:
790: Turning to Eq.(\ref{PhiVPhi17b}), we assume that $p(i)=j$ but $p'(i)
791: \neq j$ and prove that the matrix element $\langle P'|V_{i,j}|P \rangle$
792: vanishes for $D>4$. Recall that when also $p(i) \neq j$ we showed
793: that this matrix element vanishes in {\it all} $D >2$. Unlike
794: that previous case, we have here only three (rather than four)
795: $\phi$ functions in which ${\bf r}_i$ and/or ${\bf r}'_j$ appear.
796: Factors of $L$ aside, the
797: relevant two integrations on ${\bf r}_i$ and ${\bf r}'_j$ in
798: the analog of Eq.(\ref{integralsEq14}) are now:
799: \begin{widetext}
800: \begin{eqnarray}
801: & & \int d^D{\bf r}_i \int d^D{\bf r}'_j ~ V({\bf r}_i-{\bf r}'_j)
802: \phi({\bf r}_i-{\bf r}'_{p(i)})\phi({\bf r}_{p'^{-1}(j)}-{\bf r}'_j)
803: \phi({\bf r}_i-{\bf r}'_j) \nonumber \\
804: & & = A_{<} V_0 \int d^D{\bf r_i} \int d^D{\bf r}
805: ~ \Theta(r_0-|{\bf r}|)
806: [(\kappa r)^{1-\frac{D}{2}} J_{\frac{D}{2}-1}(\kappa r)]
807: \phi({\bf r}_i-{\bf r}'_{p(i)}) \phi({\bf r}_{p'^{-1}(j)} - {\bf r}'_j),
808: \label{rel2integrationsEq.19}
809: \end{eqnarray}
810: \end{widetext}
811: where we invoke Eq.(\ref{smallr}). In particular, as
812: seen from the appendix, the wavenumber
813: $\kappa = c_{D}^{1/2}/r_{0}$.
814: In Eq.(\ref{rel2integrationsEq.19}), we employ, once again,
815: $p^{-1}$ for the inverse permutation, ${\bf r}
816: = {\bf r}_i-{\bf r}'_j $ is the argument of the square well
817: potential, and $\phi({\bf r}_i-{\bf r}'_j)=\phi({\bf r})$.
818: In what briefly follows, to flesh out more crisply the simple scaling
819: form of this term, we may replace the exact Bessel function form
820: of Eq.(\ref{smallr}) and instead invoke Eq.(\ref{D>4EQ16})
821: to replace the last factor $(\phi({\bf r}))$ for $r<r_{0}$
822: in the upper line of Eq.(\ref{rel2integrationsEq.19}).
823: The result which we will obtain in
824: this fashion will go unchanged if we employ the exact Bessel function
825: form of Eq.(\ref{rel2integrationsEq.19}), see \cite{explainproof}.
826: With the insertion of Eq.(\ref{D>4EQ16}), the $d^D{\bf r}$
827: integration of Eq.(\ref{rel2integrationsEq.19})
828: yields
829: \begin{eqnarray}
830: V_0 \int_0^{r_0} d^D{\bf r} ~ [ r_0^{\frac{D}{2}-2}
831: r_{0}^{(2-D)}] = \Omega_{D} V_0 \, r_0^2
832: r_0^{\frac{D}{2}-2}\nonumber
833: \\ = \frac{\hbar^{2}}{m} \Omega_{D} c_D \, r_0^{\frac{D-4}{2}} \to 0 \;\;\;\;
834: (\mbox{for $D>4$}) \; , \label{VoIntEq.20}
835: \end{eqnarray}
836: in the limit $r_{0} \to 0$.
837: Here, the definition of $c_{D}$ was
838: invoked from Eq.(\ref{VOcd}). The peculiar form given in
839: the square braces of Eq.(\ref{VoIntEq.20})
840: follows from Eq.(\ref{D>4EQ16}) with the approximate
841: $A_{>}\approx (r_0^{(D/2)-2})$. The incorporation of the exact Bessel
842: function form of Eq.(\ref{smallr}) in Eq.(\ref{rel2integrationsEq.19})
843: whose derivation is detailed in the appendix does not alter this simple
844: result \cite{explainproof}. This concludes the proof of Eq.(\ref{PhiVPhi17b}).
845:
846:
847: With all shown thus far, vanishing in Eq.(\ref{rel2integrationsEq.19})
848: may be avoided only when $p'(i)=j=p(i)$. In this case, the
849: ${\bf r}_i$ and ${\bf r}'_j$ integrations involve only two identical
850: $\phi$ functions and become (invoking Eq.(\ref{smallr}) once again),
851: \begin{eqnarray}
852: \int d^D{\bf r}_i d^D{\bf r}~ \phi^{2}({\bf r}) \, V_0
853: \cdot \Theta({\bf r}_0-|{\bf r}|) \nonumber
854: \\= V_0 A_{<}^{2} \int d^D{\bf r}_i \int_{0<|{\bf r}|
855: <r_0} d^D{\bf r}~~
856: [(\kappa r)^{1-\frac{D}{2}} J_{\frac{D}{2}-1}(\kappa r)]^{2} \nonumber
857: \\ \sim \int d^D{\bf r}_i \, V_0. \label{identicalphiEq.21}
858: \end{eqnarray}
859:
860: By virtue of
861: the zero-energy Schr\"odinger equation that $\phi({\bf r})$
862: satisfy, the contribution of Eq.(\ref{identicalphiEq.21})
863: (albeit divergent)
864: identically cancels against the kinetic terms
865: $T_i+T'_j = - (\hbar^2/2m) ({\bf \nabla}^2_i +
866: {\bf \nabla}'^2_j)$. The coordinates
867: ${\bf r}_i$ and ${\bf r}'_j$ now appear
868: in both $\Phi(P)$ and $\Phi'(P)$ only in the above two
869: identical $\phi$ functions on the left and on the right, $T_i +
870: T'_{j}$ can be paired with the $V_{i,j}$ considered here into
871: $H_{i,j}=T_i+T'_j+V_{i,j}$. In the above shorthand notation,
872: the matrix elements of $H_{i,j}$
873: are
874: \begin{widetext}
875: \begin{equation}
876: \langle \Phi(P')|H_{i,j}|\Phi(P) \rangle = \langle \phi(1,p'(1)) \ldots
877: \phi(n,p'(n))\;|\;V_{i,j}-\frac{\hbar^2}{2m}(\nabla^2_i
878: + \nabla'^2_j)\;|\;
879: \phi(1,p(1)) \ldots \phi(n,p(n)) \rangle.
880: \label{HijshorthandEq22}
881: \end{equation}
882: \end{widetext}
883: The Schr\"odinger equation for $\phi$ with the assumed small
884: binding energy (eventually $\epsilon \to 0$) which defines our problem is:
885: \begin{equation}
886: [\frac{-\hbar^2({\bf \nabla}^2_i + {\bf \nabla}'^2_j)}{2m}
887: + V({\bf r}_i-{\bf r}'_i)]
888: \phi(|{\bf r}_i - {\bf r}'_j|) = -\epsilon\phi(|{\bf r}_i - {\bf r}'_j|),
889: \label{h2DeltaEq23}
890: \end{equation}
891: or $H_{i,j}\phi_{i,j} =-\epsilon \phi_{i,j}$. Consequently, the
892: matrix element of $H_{i,j}$ between the two $\phi_{i,j}$
893: in $\Phi(P)$ and in $\Phi(P')$ is just a number ($\epsilon$).
894: Omitting, then, $\phi_{i,j}=\phi(i,p(i)) = \phi(i,p'(i))$ from
895: the product of n functions $\phi$ in $\Phi(P)$ and $\Phi(P')$
896: leaves an overlap of
897: two monomials each composed of a product of the remaining $(n-1)$
898: orbital wavefunctions $\phi$ in both the bra and in the ket, i.e.
899: $\langle \Phi(P'_{n-1})|\Phi(P_{n-1}) \rangle$ with $P_{n-1}$
900: and $P'_{n-1}$, the
901: permutations of the remaining $(n-1)$ elements obtained by removing
902: $i \rightarrow p(i)=j$ and $i \rightarrow p'(i)=j $. Eq.(\ref{PhiPhi17a})
903: implies that this overlap vanishes unless these
904: remaining permutations are identical. Inserting
905: $i \rightarrow p(i)=j$ and $i \rightarrow p'(i)=j$ implies that
906: $P=P'$. It follows that in the numerator of
907: Eq.(\ref{E2n2nEq10}) we have,
908: similar to the denominator, only diagonal terms.
909:
910: By definition, the permutations $(i,p(i))$ reproduce the $n$
911: distinct $(i,j)$ terms for given $i$. We may, therefore,
912: operate on $|\Phi(P) \rangle$ on the right with the $n$
913: potential terms $V(i,j)$ and the
914: $n$ kinetic energy operators
915: $T_{i,j}$ with matching $\phi(i,p(i))$ terms in $\Phi(P)$. In
916: the process we employ each of the Laplacian terms
917: ($\{\nabla^2_i\}$ and $\{\nabla'^2_j\}$)
918: exactly once. All further unmatched $V(i,j)$ have vanishing matrix
919: elements.
920:
921: Hence,
922: \begin{equation}
923: H|\Phi(P) \rangle = -n\epsilon |\Phi(P)\rangle.
924: \label{PHPhiPEq24}
925: \end{equation}
926: This holds for each $\Phi(P)$ and, hence, for the full Slater
927: orbital ansatz of Eq.(\ref{DetMijEq9}) $\Psi$ in $D>4$ dimensions,
928: \begin{equation}
929: H|\Psi \rangle = E(2n) = -n\epsilon|\Psi\rangle \, .
930: \label{HPsiEq25}
931: \end{equation}
932:
933: Thus, the total energy of the system is just that of
934: $n$ independent Bose-Einstein condensed
935: dimers. When the binding energy
936: $\epsilon \rightarrow 0$ we have $E(2n) \rightarrow 0
937: $ and for the variational orbital Slater determinant
938: of Eq.(\ref{DetMijEq9}), we find $\xi_{var}(D>4) = 0$ as claimed.
939: By the standard variational theorem, this proves that that for continuous
940: $D$,
941: \begin{eqnarray}
942: \xi(D>4) \le 0.
943: \end{eqnarray}
944: The exact variational
945: result fortifies the conclusion arrived
946: at earlier (subsection(\ref{norm}), with Eq.(\ref{4D}) in particular)
947: via normalization considerations. In the final analysis,
948: normalization was present in our slightly elaborate
949: variational bound and ultimately
950: dictated the scaling form of our expressions
951: with $r_{0}$ (see, e.g.,
952: Eq.(\ref{n-2ProdEq.18}), \cite{explainnormal})
953: as $r_{0} \to 0$.
954:
955:
956:
957: \section{Heuristic considerations for \boldmath{$\xi(D=3)$}} %Sec III
958: \label{threeD}
959:
960:
961: The extension of the problem to arbitrary continuous
962: dimension $D$ allowed us to firmly establish
963: that $\xi =1$ in $D \le 2$ and $\xi =0$ in $D \ge 4$ (with the
964: particular point $D=4$ accessed by general normalization conditions;
965: our rigorous variational bound $\xi(D>4) \le 0$ held for all continuous $D>4$).
966: These results suggest that in a lowest
967: order $\epsilon$ expansion interpolating between these
968: two limits, $\xi(D=3) \approx 1/2$. [Or, if the upper bound
969: in the variational inequality for all continuous large $D$,
970: $\xi(D>4) \le 0$ is not saturated for the
971: marginal $D=4$,
972: that, more generally, $\xi(D=3) \le 1/2$.] This is, remarkably,
973: not far from the results obtained
974: by very intensive numerical work \cite{CCPS}
975: (which led to an upper bound of $\xi(D=3) \approx 0.44$).
976: Unfortunately, we have not been able to
977: attain direct potent bounds on the three dimensional
978: problem which are close to the numerically
979: attained values. In the current section, we
980: attack the harder $D=3$
981: problem by a heuristic approach.
982: This approach, although imprecise, may shed light on
983: some of the observed physics in three dimensions.
984: Similar to the simple minded dimensional interpolation
985: result it, too, suggests that $\xi(D=3) \approx 1/2$. As an
986: additional bonus, this
987: allows for a heuristic argument for
988: even-odd oscillations numerically observed.
989:
990: Unlike the normalization of the zero energy bound state, which for
991: $|{\bf r}| > r_0$ is uniform
992: in $|{\bf r}|=|{\bf r}_i-{\bf r}'_j|$, the kinetic energy,
993: \begin{eqnarray}
994: \int d^{3}{\bf r}(\nabla \phi({\bf r}))^2 =
995: \int d^{3}{\bf r} \frac{A_{>}^2}{|{\bf r}|^4},
996: \label{3Dint+}
997: \end{eqnarray}
998: is mostly concentrated at small $|{\bf r}| \sim r_0 <<d$
999: values. In the last equality of Eq.(\ref{3Dint+}),
1000: we inserted the well known ($D=3$) zero-energy S-wave wavefunction
1001: form outside a potential well, $\phi = \frac{A_{>}}{|{\bf r}|}$.
1002: As a consequence of this concentration of kinetic energy,
1003: a zero energy bound state of a mixed $(i,j')$ pair,
1004: can manifest over any smooth background function $B$
1005: of all other variables by a $\frac{1}{|{\bf r}_i-{\bf r}'_j|}$
1006: enhancement at distances $|{\bf r}_i-{\bf r}'_j|$ smaller
1007: than typical particle spacing.
1008: (Due to antisymmetry, the i$^{th}$ spin up
1009: atom is equally likely to bind with any one of the
1010: n down spin atoms at ${\bf r}'_j$, with $j =1, ..., n$.)
1011:
1012: This is indeed the case for the wave functions
1013: serving as the staring point of the above MC
1014: where correlations in mixed pairs are introduced via a Jastrow product $J$,
1015: \begin{equation}
1016: \Psi = B \times J \equiv B \prod_{j=1,..n} \, \phi(|{\bf{r}}_{i}-{\bf{r}}'_j|),
1017: \label {J} % {Eq 8}
1018: \end{equation}
1019: where the background B of free Fermion Slater determinants or BCS
1020: form accounts for the antisymmetrization.
1021:
1022: The other $(n-1)$ factors of the form
1023: $ \; $ $ \frac{1}{|{\bf r}_i-{\bf r}'_k|}$
1024: with $k$ different from $j$
1025: still allow for
1026: $\Psi \sim \frac{C}{|{\bf r}_i-{\bf r}'_j|}$
1027: for $|{\bf r}_i-{\bf r}'_j| <<d$. Thus, let the remaining ($n-1$)
1028: down spin atoms
1029: form a cubic lattice of spacing $d$, the triply periodic product,
1030: $\prod_{j=1...n-1} \, \phi(|{\bf r}'_j-{\bf r}_i|)$, is then invariant under the
1031: point symmetry group of the simple cubic lattice. Its finite
1032: extrema are at the centers of the n cubes (each of volume
1033: $d^3$) and C varies
1034: slowly for $|{\bf r}|<<d$.
1035:
1036: This suggests (yet does {\it not} justify) the following ad-hoc
1037: approach which yields $\xi=1/2$. The
1038: potentials $V(|{\bf r}_i-{\bf r}'_j|)$ involve atom pairs yet the kinetic energy is a sum over
1039: single atoms:
1040: \begin{equation}
1041: \sum_{I=1,..N=2n}\, \frac{\bf{p}^{2}_I}{2m} \,=\,
1042: \sum_{i=1,...n}\, \frac{\bf{p}_i^2}{2m}\,
1043: + \,\sum_{j=1,...n} \,\frac{(\bf{p}'_j)^2}{2m}.
1044: \label{KE} % {Eq.9}
1045: \end{equation}
1046: We formally associate kinetic energies with pairs by using the identity :
1047: \begin{equation}
1048: \sum \, \frac{{\bf{p}}_I^2}{2m} \, = \, \frac{1}{2m (N+1)} \, \sum_{I<J=1,2...N}
1049: \, ({\bf{p}}_I- {\bf{p}}_J)^2,
1050: \label{KEvarpairs} %{Eq.10}
1051: \end{equation}
1052: where ${\bf{P}}_{tot}^2 =(\sum {\bf{p}}_I)^2=0$ was subtracted cancelling
1053: all $I<J$ terms $({\bf p}_I \cdot {\bf p}_J)$.
1054: Next, we separate the contribution to the Hamiltonian of same (up-up, down-down) and
1055: mixed spin pairs with the potentials included in the second term:
1056: \begin{eqnarray}
1057: H & = & H_{same} + \, H_{mixed} \nonumber
1058: \\ & = & \frac{1}{2m (N+1)} \sum_{i<l=1,...n}
1059: ({\bf{p}}_i-{\bf{p}}_l)^2 \nonumber
1060: \\ & \;\;\;\;\;\;\;\;+ & \frac{1}{2m (N+1)}
1061: \sum_{j<k=1,...n} ({\bf{p}}'_j-{\bf{p}}'_k)^2 \nonumber \\
1062: & \;\;\;\;\;\;\;\; + & \sum_{i,j=1,...n} \, [\frac{({\bf{p}}_i-{\bf{p}}'_j)^2}{2m(N+1)} -
1063: gV({\bf r}_i,{\bf r}'_j)].
1064: \label{updownmixpairs} %{ Eq.11}
1065: \end{eqnarray}
1066:
1067: For $g=0$ the kinetic terms in $H_{same}$ and $H_{mixed}$ are equal up to 1/N corrections.
1068:
1069: Next, we note that if the following assumptions are made:
1070:
1071: (i) Just as for isolated up down pairs also in the N-body ground state
1072: $\Psi$,
1073: the attractions between different atoms cancel the kinetic energy
1074: of the relative motion, i.e., $\langle \Psi |H_{mixed}|\Psi
1075: \rangle = 0$, and that
1076:
1077: (ii) turning on the coupling $g$ does not change the expectation value of
1078: $H_{same}$, then:
1079:
1080: \begin{eqnarray}
1081: (E/N)|_{N=even} = \frac{(n-1)}{(2n-1)} \; \frac{3}{5} \; \varepsilon_F \nonumber
1082: \\ \simeq \frac{1}{2}[1 + \frac{1}{N}]\; \frac{3}{5} \; \varepsilon_F.
1083: \label{evenN} %(Eq. 12)
1084: \end{eqnarray}
1085: For odd $N$, say $n+1$ spin up and $n$ spin down atoms, only $n=N/2-1/2$
1086: bound pairs can form
1087: The decrease of potential attraction energy by $(N-1)/N$ produces a ``gap"
1088: \begin{equation}
1089: \frac{E}{N}|_{N=2n+1}-\frac{E}{N}|_{N=2n} \approx
1090: \varepsilon_F,
1091: \label{oddN} %{Eq 13}
1092: \end{equation}
1093: consistent with the finding of the MC calculations. Furthermore,
1094: \begin{equation}
1095: \xi \to 1/2 \; \;\; (\mbox{as~~} N\rightarrow{\infty}),
1096: \end{equation} %(Eq.14)
1097: which is barely consistent with the variational bound $\xi< 0.43-0.45$ from
1098: the above calculations.\cite{Cohen}
1099:
1100:
1101: \section{Conclusions}
1102: In conclusion, we examined an extension
1103: of the BCS to BEC crossover problem to arbitrary
1104: dimension $D$. We report new results on the ground
1105: state energy per particle at the crossover point by employing
1106: direct variational bounds in high dimensions
1107: and by considering the consequences
1108: of localization in low dimensions. In particular, we
1109: find that the ground
1110: state energy per particle at the onset of
1111: the BCS to BEC crossover is zero (or negative)
1112: in all dimensions $D \ge 4$ while it is the energy of a
1113: free Fermi system in
1114: all dimensions $D \le 2$. The interpolation of these
1115: bounds to the physical
1116: three dimensional problem leads to an
1117: energy per particle which is half that of
1118: the free fermion energy and is close to current
1119: numerical results. We outlined
1120: a simple heuristic argument
1121: for the same result which further mandates
1122: even-odd variations which are indeed
1123: seen numerically.
1124:
1125:
1126:
1127: {\bf{Acknowledgments}}
1128:
1129: We benefited from discussions with G. A. Baker,
1130: A. Casher, R. Furnstahl, A. Schwenk and would like to thank J.
1131: Carlson, G. Ortiz and E. M. Timmermans, the organizers of the 2004
1132: Los Alamos/Santa Fe workshop on cold atoms.
1133:
1134:
1135: \appendix
1136:
1137: \section{The Zero Energy Bound State Problem in a D-dimensional
1138: spherical potential well}
1139: \label{ap1}
1140:
1141:
1142: Here, we provide the solution of the zero energy
1143: bound state problem in $D$ dimensions for the spherical
1144: potential well of Eq.(\ref{VrV0Eq12}) wherein the
1145: depth of the well ($V_{0}$) is tuned to get a zero-energy bound
1146: state. The solution is a straightforward extension of the
1147: standard $D=1,2,3$ dimensional spherical potential well problems.
1148: In these,
1149: without any angular dependence, in the
1150: the ``S-wave'' representation (i.e. the scalar (``$l=0$'') representation
1151: of the $SO(D)$ rotation group), the D-dimensional Laplacian is
1152: $\nabla^{2} = [ \frac{d^{2}}{dr^{2}} + \frac{D-1}{r} \frac{d}{dr}]$.
1153: This appendix explicitly illustrates how the numerical constant
1154: $c_{D}$ of Eq.(\ref{VOcd}) may be determined (from the implicit
1155: Eq.(\ref{BesselV})). Its results further allow us to compare the approximate
1156: form of Eq.(\ref{D>4EQ16}) introduced within the text to allow a clear
1157: understanding of the scaling form of the overlap integrals with
1158: the exact form of $\phi(r)$ (Eqs.(\ref{bigr},\ref{smallr})).
1159: In the aftermath, it will be shown that the incorporation
1160: of the exact functional from does not change the scaling results
1161: derived in the main text (following Eq.(\ref{D>4EQ16})).
1162:
1163: We proceed with the solution of the translationally
1164: invariant problem specified
1165: by the spherical potential well
1166: of Eq.(\ref{VrV0Eq12}). By translational invariance
1167: of the center of mass, the two body wavefunction
1168: \begin{eqnarray}
1169: \overline{\phi}({\bf x}, {\bf x}')
1170: = L^{-D/2} \phi({\bf r})
1171: \end{eqnarray}
1172: with ${\bf r} \equiv {\bf x}' - {\bf x}$. In the
1173: potential-free region $(r=|{\bf r}|>r_{0})$, the wavefunction
1174: $\phi$ satisfies a single particle Schr\"odinger equation
1175: with a reduced mass $\mu = m/2$,
1176: \begin{equation}
1177: \frac{d^{2}}{dr^{2}} \phi(r) +
1178: \frac{D-1}{r} \frac{d}{dr} \phi(r) + \kappa^{2} \phi(r) = 0,
1179: \label{schr}
1180: \end{equation}
1181: with $\kappa=0$. Within the spherical potential
1182: well ($r < r_{0}$), we have Eq.(\ref{schr}) with
1183: $\kappa^{2} = (m V_{0})/\hbar^{2}$. We immediately find
1184: that
1185: \begin{eqnarray}
1186: \phi(r>r_{0}) = A_{>} r^{2-D},
1187: \label{bigr}
1188: \end{eqnarray}
1189: while within the potential well
1190: \begin{eqnarray}
1191: \phi(r<r_{0}) = A_{<}[(\kappa r)^{1-\frac{D}{2}}
1192: J_{\frac{D}{2}-1}(\kappa r)],
1193: \label{smallr}
1194: \end{eqnarray}
1195: with $\kappa$ and the constant prefactors $A_{>}$ and $A_{<}$
1196: determined by continuity at
1197: $r=r_{0}$ and global normalization. Here, $J_{\frac{D}{2}-1}$ is
1198: a Bessel function of order $(\frac{D}{2}-1)$. (Inserting
1199: $J_{\frac{1}{2}}(\kappa r) = [(\kappa r)^{-1/2} \sin \kappa r]$,
1200: the pertinent
1201: three dimensional result, illustrated in Fig.(\ref{Fig1bound}),
1202: is recovered.) Continuity at $r=r_{0}$
1203: restrains
1204: \begin{eqnarray}
1205: \phi(r<r_{0}) \approx A_{>} r_{0}^{2-D}.
1206: \label{contA}
1207: \end{eqnarray}
1208: For small $r$ (i.e. $r \to 0$), employing the asymptotic form
1209: of the Bessel functions,
1210: we find that
1211: \begin{eqnarray}
1212: \phi(r \to 0) = \frac{A_{<}}{\Gamma(\frac{D}{2})2^{\frac{D}{2}-1}}.
1213: \end{eqnarray}
1214: Anywhere within the potential well ($r <r_{0}$), the wave function
1215: \begin{eqnarray}
1216: \phi(r<r_{0}) = f(\frac{r}{r_{0}}) A_{>} r_{0}^{2-D},
1217: \label{fz}
1218: \end{eqnarray}
1219: with
1220: $f(0\le z \le 1)$ a bounded function satisfying $f(1) =1$ and
1221: $f(0) = \frac{A_{<}}{A_{>}} \frac{r_{0}^{D-2}}{\Gamma(\frac{D}{2})
1222: 2^{\frac{D}{2}-1}}$. Within the text following
1223: Eq.(\ref{D>4EQ16}), we set $f$ to be one
1224: and we attained vanishing overlaps
1225: in Eq.(\ref{VoIntEq.20}) in the
1226: limit $r_{0} \to 0$. The incorporation
1227: of the more explicit, yet {\em bounded}, $f(z)$
1228: given by Eq.(\ref{smallr}) does
1229: not alter this result. Formally, our
1230: results derived in the text can be made
1231: rigorous by deriving an upper bound which
1232: differs from the results given in Eq.(\ref{VoIntEq.20})
1233: by a factor of $|f|_{\max}$, the maximal value of $|f|$.
1234: As Eq.(\ref{VoIntEq.20}) vanishes by simple scaling,
1235: the incorporation of a finite factor leads to a vanishing
1236: upper bound, proving that Eq.(\ref{rel2integrationsEq.19}) vanishes
1237: in the limit of small $r_{0}$. Although inconsequential for this
1238: bound (all that matters is its bounded norm),
1239: the prefactor $f(r/r_{0})$ is everywhere positive
1240: as the zero energy ground state is nodeless
1241: and of uniform sign.
1242:
1243:
1244: For completeness, we now outline the general solution.
1245: Matching the two functional forms of $\phi(r)$ and their
1246: derivatives at $r=r_{0}$ (and simplifying with the
1247: aid of the standard Bessel function
1248: recursion relations) leads to the implicit equation
1249: \begin{eqnarray}
1250: (1- \frac{D}{2}) J_{\frac{D}{2}-1}(\kappa r_{0}) \nonumber
1251: \\ = \frac{1}{2} (\kappa r_{0})
1252: [J_{\frac{D}{2}}(\kappa r_{0}) + J_{\frac{D}{2}- 2}(\kappa r_{0})].
1253: \label{BesselV}
1254: \end{eqnarray}
1255: The solution of Eq.(\ref{BesselV}) enables a determination
1256: of $\kappa$ and thus of the spherical potential depth $V_{0}$ and
1257: the constant $c_{D}$ in Eq.(\ref{VOcd})
1258: which ensures a zero-energy ``S-wave''
1259: state in D dimensions.
1260: The amplitudes $A_{>}$ and $A_{<}$ are then determined
1261: by
1262: \begin{eqnarray}
1263: \frac{A_{<}}{A_{>}} = \frac{r_{0}^{2-D}}{(\kappa r_{0})^{1 - \frac{D}{2}}
1264: J_{\frac{D}{2}-1}(\kappa r_{0})},
1265: \label{amp_ratios}
1266: \end{eqnarray}
1267: in conjunction with the global normalization of $\phi({\bf r})$.
1268: Normalization demands that, up to dimension dependent
1269: numerical constants, $A_{>} \approx r_{0}^{D/2-2}$
1270: and consequently $A_{<} \approx r_{0}^{-D/2}$.
1271:
1272:
1273:
1274: \begin{thebibliography}{99}
1275:
1276: \bibitem{gas95} M. H. Andersen, J. R. Ensher, M. R. Matthews, C. E. Wieman,
1277: and E. A. Cornell, Science {\bf 269}, 198 (1995); K. B. Davis,M.-O. Mewes,
1278: M. R. Andrews, N. J. van Druten, D. S. Durfee, D. M. Kurn,
1279: and W. Ketterle, Phys. Rev. Lett. {\bf 75}, 3969 (1995);
1280: C. C. Bradley, C. A. Sackett, J. J. Tollett and R. G. Hulet, Phys. Rev. Lett.
1281: {\bf 75}, 1687 (1995); C. C. Bradley,
1282: C. A. Sackett, and R. G. Hulet, Phys. Rev. Lett.
1283: {\bf 78}, 985 (1997)
1284:
1285:
1286: \bibitem{RGJ} C. A. Regal, M. Greiner and D. S. Jin, Phys. Rev. Lett. {\bf
1287: 92}, 040403 (2004);
1288: M. Zwierlein, et.al., Phys. Rev. Lett. {\bf 92}, 120403 (2004).
1289:
1290: \bibitem{DH} R. B. Diener and T-L. Ho,
1291: Phys. Rev. Lett. {\bf 94}, 090402 (2005)
1292:
1293: \bibitem{epsilonexpansion} K. G. Wilson, Phys. Rev. Lett. {\bf 28}, 548
1294: (1972); K. G. Wilson and M. E. Fisher,
1295: Phys. Rev. Lett. {\bf 28}, 240
1296: (1972); K. G. Wilson and J. Kogut, Phys. Rep. {\bf 12}, 75
1297: (1974); J. Zinn-Justin, {\em Quantum Field Theory and Critical Phenomena},
1298: Oxford University Press (1989)
1299:
1300: \bibitem{ChallProb} G. F. Bertsch, ``Challenge Problems in Many-Body
1301: Physics",
1302: http//:www.phys.washington.edu/$\sim$mbx/george.html (1998).
1303:
1304: \bibitem{CCPS} J. Carlson, S.-Y. Chang, V. R. Pandharipande, and K.
1305: E. Schmidt, Phys. Rev. Lett. {\bf 91}, 050401 (2003).
1306:
1307: \bibitem{Baker} G. A. Baker, Jr., Phys. Rev. C {\bf 60}, 054311
1308: (1999).
1309:
1310: \bibitem{scale} The hierarchy of scales, $a>>d>>r_0$, which is
1311: implicit here may not suffice to
1312: guarantee the same universal expression as E/N could also depend on $d^2/(a
1313: r_0)$.
1314:
1315: \bibitem{LeggetEngelbrecht} A. J. Legget, in {\it Modern Trends in the
1316: Theory
1317: of Condensed Matter}, edited by A. Pekalski and R.Przystawa
1318: (Springer-Verlag, Berlin, 1980); J. R. Engelbrecht,
1319: M. Randeria, and C. sa de Melo, Phys. Rev. B {\bf 55}, 15 153 (1997).
1320:
1321:
1322:
1323: \bibitem{LL} Landau, Lifshitz, {\it Quantum Mechanics}
1324: (Pergamon Press), third edition (1977)
1325:
1326: \bibitem{LdeLAndRDS} E. H. Lieb and M. de Liano, J.
1327: Math. Phys.{\bf 19},860
1328: (1978); M. Randeria, J.-M. Duan, and L.-Y. Shieh,
1329: Phys. Rev. Lett.
1330: {\bf 62}, 981 (1989).
1331:
1332: \bibitem{MV} W. Metzner and D. Voldhardt, Phys. Rev.
1333: Lett. {\bf 62}, 324
1334: (1989).
1335:
1336: \bibitem{Steel} J.W.Steel, arXiv:nucl- th/0010066
1337: (2000).
1338:
1339: \bibitem{SKC}T. Schaefer, C-W Kao, and S. R. Cotanch, arXiv:nucl-th/0504088
1340: (2005)
1341:
1342:
1343: \bibitem{gang4} E. Abrahams, P. W. Anderson, D. C. Licciardo,
1344: and T. V. Ramakrishnan, Phys. Rev. Lett. {\bf 42}, 673 (1979)
1345:
1346: \bibitem{explainnormal} In this case, we recover (by definition)
1347: the normalization integral. For a single up/down spin
1348: pair (${\bf x}, {\bf x}')$
1349: this (with the aid of Eqs.(\ref{bigr}, \ref{smallr}))
1350: is of the forms
1351: \begin{eqnarray}
1352: {\cal I}_{>} = A_{>}^{2} L^{-D} \int d^{D}{\bf R} \int_{r>r_{0}} \frac{d^{D}
1353: {\bf r}}{r^{2(D-2)}} \nonumber
1354: \\ \mbox{ ~or~~~~}
1355: {\cal I}_{<} = A_{<}^{2} L^{-D} \int d^{D} {\bf R} \int_{r<r_{0}} d^{D}{\bf r}
1356: \frac{[J_{\frac{D}{2} -1}(\kappa r)]^{2}}{ (\kappa r)^{D-2}},
1357: \label{calI}
1358: \end{eqnarray}
1359: with ${\bf R} = ({\bf x}+ {\bf x}')/2$
1360: the center of mass, whose integration
1361: cancels identically against the volume factors
1362: of $L^{D}$. Due to the higher power
1363: of $r$ in the denominators (vis a vis
1364: the integrals in Eq.(\ref{n-2ProdEq.18})),
1365: a lower power of $r_{0}$ is attained in ${\cal I}_{>},{\cal I}_{<}$
1366: with respect to their permuted counterparts (see below) .
1367: In ${\cal I}_{>}$, the scaling is that of $A_{>}^{2} r_{0}^{4-D}$
1368: which mandates the relation $A_{>} \approx r_{0}^{D/2-2}$ appearing
1369: in the text (Eq.(\ref{D>4EQ16})). The lower power of $r_{0}$ appearing
1370: here in the normalization integrals when no permutations are
1371: performed relative to the power of $r_{0}$ resulting from
1372: any integral appearing with a permutation (i.e. the scaling of of the integrals
1373: ${\cal I}_{>}$ of Eq.(\ref{calI}) relative to that of $I_{>}$ and $I_{<}$
1374: defined following Eq.(\ref{n-2ProdEq.18})))
1375: ensures that all integrals of the type of Eq.(\ref{n-2ProdEq.18}))
1376: vanish in the $r_{0} \to 0$ limit.
1377:
1378: \bibitem{explainproof}
1379: Indeed, this may be rigorously proven by replacing the positive function
1380: $f(z)$ defined via Eq.(\ref{fz}), (with the scaled radial coordinate
1381: $z \equiv r/r_{0}\sim \kappa r$), by its maximal value
1382: (denoted henceforth by $f_{\max}$)
1383: in the interval $[0,1]$ to attain an upper bound on
1384: the integral of Eq.(\ref{rel2integrationsEq.19}). This upper
1385: bound differs from the result shown in Eq.(\ref{VoIntEq.20})
1386: by a factor $f_{\max}$. As $f_{max}$ is finite,
1387: the vanishing of Eq.(\ref{rel2integrationsEq.19}) cannot be avoided
1388: in a spherical well of small radial extent $(r_{0} \to 0)$.
1389:
1390: \bibitem{Cohen} This feature motivated the extension to
1391: unequal numbers of spin-up and spin-down atoms leading to a new
1392: exact and interesting result for the asymmetry energy, T. D. Cohen,
1393: Phys. Rev. Lett. {\bf 95}, 120403 (2005)
1394: (arXiv:cond-mat/0505080)
1395:
1396: \end{thebibliography}
1397:
1398: \end{document}
1399:
1400:
1401: