PHYSICAL REVIEW A 74, 053624 2006

Hyperspherical description of the degenerate Fermi gas: s-wave interactions

Department of Physics and JILA, University of Colorado, Boulder, Colorado 80309-0440, USA Department of Physics and Astronomy, University of Kentucky, Lexington, Kentucky 40506-0055, USA Received 12 July 2006; published 28 November 2006
1

Seth T. Rittenhouse,1 M. J. Cavagnero,2 Javier von Stecher,1 and Chris H. Greene1

We present a unique theoretical description of the physics of the spherically trapped N-atom degenerate Fermi gas at zero temperature based on an ordinary Schrdinger equation with a microscopic, two-body interaction potential. With a careful choice of coordinates and a variational wave function, the many-body Schrdinger equation can be accurately described by a linear, one-dimensional effective Schrdinger equation in a single collective coordinate, the rms radius of the gas. Comparisons of the energy, rms radius, and peak density of the ground-state energy are made to those predicted by Hartree-Fock HF theory. Also the lowest radial excitation frequency the breathing mode frequency agrees with a sum rule calculation, but deviates from a HF prediction. DOI: 10.1103/PhysRevA.74.053624 PACS number s : 03.75.Ss, 31.15.Ja

I. INTRODUCTION

The realization of a degenerate Fermi gas DFG in a dilute gas of fermionic atoms has triggered widespread interest in the nature of these systems. This achievement combined with the use of a Feshbach resonance allows for a quantum laboratory in which many quantum phenomena can be explored over a wide range of interaction strengths. This leads to a large array of complex behaviors including the discovery of highly correlated BCS-like pairing for effectively attractive interactions 15 . While these pairing phenomena are extremely interesting, we believe that the physics of a pure degenerate Fermi gas, in which the majority of atoms successively ll single-particle states, is worthy of further study even in the absence of pairing. This topic is addressed in the present paper, which follows a less complete archived study of the topic 6 . The starting point for this study is that of a hyperspherical treatment of the problem in which the gas is described by a set of 3N 1 angular coordinates on the surface of a 3N-dimensional hypersphere of radius R. Here N is the number of atoms in the system. This formulation is inspired by a similar study of the Bose-Einstein condensate BEC 79 . Furthermore, these same coordinates have been applied to nite nuclei 10 . The formulation is a rigorous variational treatment of a many-body Hamiltonian, aside from the limitations of the assumption of pairwise, zero-range interactions. In this paper, we consider only lled energy shells of atoms. This is done for analytic and calculational simplicity, but the treatment could apply to any number of atoms in open shells with modest extensions that are discussed briey. The main goal of this study is to describe the motion of the gas in a single collective coordinate R, which describes the overall extent of the gas. The benet of this strategy is that the behavior of the gas is reduced to a single onedimensional 1D linear Schrdinger equation with an effective hyperradial potential. The use of a real potential then lends itself to the intuitive understanding of normal Schrdinger quantum mechanics. This method also allows for the calculation of physical quantities such as the energy and rms radius of the ground state; these observables agree
1050-2947/2006/74 5 /053624 14

quantitatively with those computed using Hartree-Fock methods. The method also yields a visceral understanding of a low-energy collective oscillation of the gasi.e., a breathing mode. Beyond the intuitive benets of reducing the problem to an effective one-dimensional Schrdinger equation, another motivation for developing this hyperspherical viewpoint is that it has proven effective in other contexts for describing processes involving fragmentation or collisions in few-body and many-body systems 11,12 . Such processes would be challenging to formulate using eld theory or the random phase approximation RPA or conguration interaction viewpoints, but they emerge naturally and intuitively, once the techniques for computing the hyperspherical potential curves for such systems are adequately developed. To this end we view it as a rst essential step, in the development of a more comprehensive theory, to calculate the ground-state and low-lying excited-state properties within this framework. Then we can ascertain whether the hyperspherical formulation is capable of reproducing the key results of other, more conventional descriptions, which start instead from a meaneld theory perspective. The paper is organized as follows: Section II develops the formulation that yields a hyperradial 1D effective Hamiltonian. In Sec. III we apply this formalism to a zero-range s-wave interaction and nd the effective Hamiltonian in both the nite-N and large-N limits. In Secs. III A and III B we examine the nature of the resulting effective potential for a wide range of interaction strengths and give comparisons with other known methods, mainly the Hartree-Fock HF method. Section III C is a brief discussion of the simplications that can be made in the limit where N . Finally in Sec. IV we summarize the results and discuss future avenues of study.
II. FORMULATION

The formalism is similar to that of Ref. 7 , but we will reiterate it for clarity and to make this article self-contained. Consider a collection of N identical fermionic atoms of mass m in a spherically symmetric trap with oscillator frequency
2006 The American Physical Society

053624-1

RITTENHOUSE et al.

PHYSICAL REVIEW A 74, 053624 2006

, distributed equally between two internal spin substates. The governing Hamiltonian is H= 2 2m i=1
N 2 i

=
i j

2 ij,

ij

= xi

xj

xj

xi

2.6

1 + m 2

N 2 i=1

r2 + i
i j

Uint rij ,

2.1

for all Cartesian components xi of the 3N-dimensional space. The isotropic spherical oscillator potential becomes simply 1 m 2
2

where Uint r is an arbitrary two-body interaction potential and rij = ri r j. We ignore interaction terms involving three or more bodies. In general, the Schrdinger equation that comes from this Hamiltonian is very difcult to solve. Our goal is to simplify the system by describing its behavior in terms of a single collective coordinate. To achieve this aim we transform this Hamiltonian into a set of collective hyperspherical coordinates; the hyperradius R of this set is given by the root-mean-square distance of the atoms from the center of the trap: R 1 r2 N i=1 i
N 1/2

1 r2 = M i 2 i=1

2 2

R .

2.7

The sum of Eqs. 2.5 and 2.7 gives a time-independent Schrdinger equation H R , = E R , of the form 0= + 2 2M
2

R2

2 3N 1 3N 3 1 2 + M 2 2 2R R

2 2

Uint ri r j E R 3N1 /2
i j

R,

2.8

2.2

So far, we only have one coordinate for the system, but we need to account for all 3N spatial coordinates and also the spin degrees of freedom. This leaves 3N 1 angular coordinates left to dene. We have 2N of the angles as the independent-particle spherical polar coordinate angles 1 , 1 , 2 , 2 , . . . , N , N . The remaining N 1 hyperangles are chosen by the convention used in Ref. 10 , which describes correlated motions in the radial distances of the atoms from the trap center,
i

where R, has been multiplied by R 3N1 /2 to remove rst-derivative terms in the hyperradius. These mathematical transformations have not yet accomplished much. We started with a 3N-dimensional Schrdinger equation to solve, and we still have a 3N-dimensional Schrdinger equation. To simplify, we assume that is approximately separable into an eigenfunction of the operator 2, a hyperspherical harmonic, multiplied by an unknown hyperradial function. Hyperspherical harmonics HH; see Ref. 13 for more details are generally expressed as products of Jacobi polynomials for any number of dimensions. Their eigenvalue equation is
2

r2 j tan
i

j=1

+ 3N 1

2.9

ri+1

, 2.3

i = 1,2,3, . . . ,N 1. Alternatively we may write this as

N1

where = 0 , 1 , 2 , . . . and omitted below for brevity stands for the 3N 2 other quantum numbers that are needed to distinguish between the usually quite large degeneracies for a given . The separability ansatz implies that R, =F R . 2.10

rn = NR cos

n1 j=n

sin

j,

j = 1,2, . . . ,N 1,

2.4

where we dene cos 0 1 and N1 sin j 1. For the purj=N poses of this study, the set of 3N 1 hyperangles will be referred to collectively as . The particular denition of the hyperangles will not play a signicant role in the actual formalism, but we give the denitions here for completeness. After carrying out this coordinate transformation on the sum of Laplacians, the kinetic energy becomes 13
2 2 1 R3N1 2 . 2M R3N1 R R R

2.5

Here M = Nm and is the grand angular momentum operator which is similar to the conventional angular momentum operator and is dened by

is the HH that corresponds to Here we assume that the lowest value of the hyperangular momentum that is allowed for the given symmetry of the problemi.e., that is antisymmetric with respect to interchange of indistinguishable fermions. This choice of trial wave function indicates that we expect the overall energetics of the gas to be described by its size; as such, xing the hyperangular behavior is equivalent to xing the conguration of the atoms in the gas. In nuclear physics this is known as the K harmonic method. Alternatively, this may be viewed as choosing a trial wave function whose hyperradial behavior will be variationally optimized, but whose hyperangular behavior is that of a noninteracting-trap-dominated gas of fermions. To utilize this as a trial wave function, we evaluate the H , where the integration is taken expectation value over all hyperangles at a xed hyperradius. This approach gives a new effective linear 1D Schrdinger equation Hef f R 3N1 /2F R = ER 3N1 /2F R in terms of an effective Hamiltonian Hef f given by

053624-2

HYPERSPHERICAL DESCRIPTION OF THE

PHYSICAL REVIEW A 74, 053624 2006

2 d2 K K + 1 2M dR2 R2

1 + M 2

2 2

R +
i j

Uint rij

. 2.11

R 3N1 /2G R = A exp R2/2L2

R L

+3N/21/2

; 2.15

Here K = + 3 N 1 / 2. We now must force our trial wave function to obey the antisymmetry condition of fermionic atoms. To antisymmenote that Eq. 2.2 trize the total wave function F R indicates that R is completely symmetric under all particle coordinate exchange; thus the antisymmetrization of the . Finding a completely wave function must only affect for any given is generally quite difantisymmetric cult and is often done using recursive techniques like coefcient of fractional parentage expansions see Ref. 10,14,15 for more details or using a basis of Slater determinants of independent-particle wave functions 1619 . We use a simplied version of the second method combined with the following theorem, proved in Ref. 15 and developed in Appendix B. Theorem 1. The ground state of any noninteracting set of N particles in an isotropic oscillator is an eigenfunction of 2 with minimal eigenvalue + 3N 2 where is given by the total number of oscillator quanta in the noninteracting system: = ENI 3N , 2 2.12

here, A is a normalization constant and L = / M = l / N. : Combining Eq. 2.14 with Eq. 2.15 now gives us
N

1 pP ,
1, 2,

... ,

i=1

R ni i r i y GR

imi

m si , 2.16

where we must make the variable substitutions in the numerator using Eqs. 2.2 and 2.3 . In the following, for brevity, the spin coordinates 1 , . . . , N will be suppressed = , 1 , 2 , . . . , N . Interestingly, this must i.e., be a function only of the hyperangles, and thus all of the hyperradial dependence must cancel out on the right-hand side of Eq. 2.16 . We are now ready to start calculating the Uint . interaction matrix element
III. ZERO-RANGE s-WAVE INTERACTION

Here we specify Uint r as a zero-range two-body interaction with an interaction strength given by a constant parameter g: Uint r = g
3

where ENI is the total ground-state energy of the noninteracting N-body system. in terms of With theorem 1 in hand we may nd the independent-particle coordinates. The noninteracting ground state is given by a Slater determinant of singleparticle solutions:
NI

r .

3.1

r1,r2, . . . ,rN,
N

1,

2,

... ,

For a small two-body, s-wave scattering length a we know 2 20 . For stronger interactionsi.e., that g = 4 m a k f a 1this approximation no longer holds and g must be renormalized. Now we must calculate the effective interaction matrix element given by Uef f R = g
i j 3

=
P

1 pP
i=1

R ni i r i y

imi

m si .

2.13

rij

3.2

Here i is the spin coordinate for the ith atom and Rni i ri is the radial solution to the independent-particle harmonic oscillator for the ith particle given by rRn r = Nn exp r2 / 2l2 r / l +1Ln+1/2 r / l 2 where Ln r is an associated is an ordinary Laguerre polynomial with l = / m . y imi 3D spherical harmonic with i as the spatial solid angle for the ith particle, and msi is a spin ket that will allow for two spin species of atoms, and . The sum in Eq. 2.13 runs over all possible permutations P of the N spatial and spin coordinates in the product wave function. We now apply theorem 1 which directly leads to a NI that is separable into a hyperangular piece and a hyperradial piece:
NI

Degenerate ground states cause some complications for this formulation which we avoid by restricting ourselves to the nondegenerate ground states that correspond to lled energy shells of the oscillatori.e., magic numbers of particles. With moderate extensions the degeneracies can be taken into account by creating an interaction matrix, but the magic number restriction should still give a good description of the general behavior of systems with large numbers of atoms. The total number of atoms and the hyperangular momentum quantum number are most conveniently expressed in terms of the number n of single-particle orbital energies lled: N= n n+1 n+2 , 3 3.3a

r1,r2, . . . ,rN, ,

1, 1,

2,

... ,

N N

=G R

2, . . . ,

2.14

Here G R is the nodeless hyperradial wave function describing the ground state of N noninteracting atoms in an isotropic oscillator trap. A derivation of G R is given in Appendix A:
053624-3

n1 n n+1 n+2 , 4

3.3b

RITTENHOUSE et al.

PHYSICAL REVIEW A 74, 053624 2006

kf =

2m

n+

1 , 2

3.3c

TABLE I. N, , and CN / N7/2 for several lled shells. We can see that CN / N7/2 quickly converges to the Thomas-Fermi limiting value of 32 2 / 3 / 35 3 0.02408 to several digits. n 1 2 3 4 5 15 30 100 N 2 8 20 40 70 1360 9920 343400 0 6 30 90 210 14280 215760 25497450
1

where k f is the peak noninteracting Fermi wave number. In the limit where N 1, we write and k f in terms of the total number of particles, N, 3N 4 kf 2m
4/3

CN / N7/2
8
2

3.4a

3N

1/3

3.4b

Next we combine Eq. 3.4a with from Eq. 2.16 and calculate the interaction matrix element. is antisymmetric under particle exchange, Since we may do a coordinate transposition in the sum ri r2 and r j r1. Each transposition pulls out a negative sign from , and we are left with Uef f R = g
i j 3

0.0127 0.0637 0.0251 0.0244 0.02435 0.0241 0.02409 0.02408

r21

=g

N N1 2

r21

Appendix C details the calculation of the matrix element 3 r21 . The result is Uef f R = g N R CN 3/2 3 , 3.5

where CN is a constant that is dependent only on the number of atoms in the system. While CN has some complex behavior for smaller numbers of particles, we have seen that for N 100, CN quickly converges to CN 0.049 0.277 2 32N7/2 4/3 + 3 1+ 3 35 N2/3 N 0.049 0.277 4/3 + , N2/3 N 3.6

pected, for N = 1, with C1 = 0, Eq. 3.8 reduces to a single particle in a trap with K as the angular momentum quantum number . Note that the form of Vef f is very similar to the effective potential found for bosons by the authors of Ref. 7 . What may be surprising is the extra term of 3 N 1 / 2 contained in K. The kinetic energy term in Vef f is controlled by the hyperangular momentum, which in turn reects the total nodal structure of the N-fermion wave function. This added piece of hyperangular momentum summarizes the energy cost of conning N fermions in the trap. This repulsive barrier stabilizes the gas against collapse for attractive interactionsi.e., g 0. A nal transformation simplies the radial Schrdinger equationnamely, setting E = ENIE and R = R2 NIR , where ENI = + 3N , 2 + 3 2 3.9a

= 0.02408N7/2 1 +

R2

NI

= l2

3.9b

CN/(0.0241N )

where the higher-order terms in 1 / N were found by tting a curve to numerically calculated data. Values of CN / N7/2 are shown in Table I for several lled shells. Figure 1 shows both the calculated values of CN and the values from the t in Eq. 3.6 versus 1 / N. In both the table and the plot we can clearly see the convergence to the large N value of CN 0.024 08N7/2. We now may write the effective one-dimensional, hyperradial Schrdinger equation Hef f R 3N1 /2F R 3N1 /2 F R , where Hef f is given by = ER Hef f = d + Vef f R , 2M dR2
2 2

are the noninteracting expectation values of the energy and squared hyperradius in the ground state. In the following any
1.006 1.005
7/2

1.004 1.003 1.002 1.001 1 0.999 0 0.0005 0.001 0.0015 0.002 0.0025 1/N

3.7

where Vef f R is an effective hyperradial potential given by Vef f R = K K+1 1 + M 2 2 R 2M

2 2 2

R +g

CN 3/2 3 . N R

3.8

We reinforce the idea that this effective Hamiltonian describes the fully correlated motion of all of the atoms in the trap albeit within the aforementioned approximations. As ex-

FIG. 1. Values of CN divided by the large-N limit CN 0.02408N7/2 vs 1 / N are shown. The circles are the calculated value while the curve is the t stated in Eq. 3.6 .

053624-4

HYPERSPHERICAL DESCRIPTION OF THE

PHYSICAL REVIEW A 74, 053624 2006

-4 -4 -4 -4

6 5 Veff R ENI 4
E/ENI 1.15

410 310 210 110 0 0 5 kf 10 15

3 2 1 0

1.1

1.05

0.5

1 R

1.5 R2

2
NI

2.5

(EK-EHF)/ENI

kf

10

15

FIG. 2. The dimensionless rescaled effective potential Vef f / ENI as a function of the rescaled hyperradius for k f = 0 solid line , k f = 15 dashed line , and k f = 5 dot-dashed line .

hyperradius with a prime, R , denotes the hyperradius in units of R2 NI. Under this transformation, with the use of Eqs. 3.4a and 3.4b , and in the limit where N , the hyperradial Schrdinger equation becomes 1 d2 Vef f R * 2 + 2m dR ENI E R
3N1 /2

FIG. 3. The ground-state energy in units of the noninteracting energy vs k f for 240 atoms calculated using the K harmonic method curve and using the Hartree-Fock method circles . Inset: the difference in the ground-state energies predicted by the K harmonic EK and Hartree-Fock EHF methods. Clearly the K harmonic energies are slightly higher than the Hartree-Fock energies.

body, zero-range potential 21 . For k f 0 the positive 1 / R 3 serves to strengthen the repulsive barrier and pushes the gas further out.
A. Repulsive interactions g 0

F R = 0, 3.10

where m* = + 3N / 2 2. The effective potential takes on the simple form Vef f R ENI 1 1 2 kf , 2 + R + 2 2R R3 3.11

where = 1024/ 2835 3 and = g / l2. We note that the only parameter that remains in this potential is the dimensionless quantity k f and that in oscillator units = g. In the limit where N the effective mass m* becomes m* 1 3N 16
8/3

3.12

For large numbers of particles, the second derivative terms in Eq. 3.10 becomes negligible, a fact used later in Sec. III C. The behavior of Vef f R versus R is illustrated in Fig. 2 for various values of k f . For k f = 0 solid curve , the noninteracting limit, the curve is exact and the ground-state solution is given by Eq. 2.15 . For nonzero values of g, Vef f acquires an attractive k f 0 or repulsive k f 0 1/R 3 contribution as indicated by the dot-dashed and dashed lines, respectively. For k f 0 the DFG is metastable in a region which has a repulsive barrier which it may tunnel through and emerges in the region of small R where the interaction term is dominant. It should be noted, though, that small R means the overall size of the gas is small. Thus the region of collapse corresponds to a very high density in the gas. In this region several of the assumptions made can fall apart, most notably the assumption dealing with the validity of the two-

For positive values of the interaction parameter g we expect the predicted energy for this K harmonic method to deviate from experimental values, since the trial wave function does not allow any fermions to combine into molecular pairs as has been seen in experiments 1,2,4,5 . Our method can only describe the normal degenerate Fermi gas. The strong repulsive barrier for repulsive interactions shown in Fig. 2 arises as the gas pushes against itself which increases the energy and rms radius of the ground state. Figures 3 and 4 compare the ground-state energy and average radius squared, respectively, of 240 trapped atoms, plotted as a function of k f with a Hartree-Fock HF calculation. The inset in Fig. 3 shows that the K harmonic energies are slightly above the HF energies; since both methods are variational upper bounds, we can conclude that the Hartree-Fock solution is a slightly better representation of the true solution to the full Schrdinger equation with -function interactions. An added benet of the K harmonic method is that we now have an intuitively simple way to understand the energy of the lowest radial excitation of the gasi.e., the breathing mode frequency. Figure 2 shows that as k f increases, the repulsion increases the curvature at the local minimum, whereby stronger repulsion causes the breathing-mode frequency to increase. Figure 5 compares the breathing-mode frequency calculated using the K harmonic method to the sum rule prediction 22 based on HF orbitals and also the lowest eight radial excitation frequencies predicted by the Hartree-Fock method. As anticipated, the K harmonic method and the sum rule method agree that the breathingmode frequency will increase with added repulsion. Interest-

053624-5

RITTENHOUSE et al.

PHYSICAL REVIEW A 74, 053624 2006

1.2 <R >/<R >NI 1.15 1.1 1.05 1 0 5 kf 10 15

peak density (oscillator units)

0.75 0.7 0.65 0.6 0 2 4 6 8 kf 10 12 14

FIG. 4. The ground-state average squared radius of the gas atoms in units of the noninteracting rms squared radius is plotted vs k f . The calculations considered 240 atoms in both the K harmonic method curve and Hartree-Fock method squares .

FIG. 6. The peak density in units of / m 3/2 vs k f predicted by the K harmonic solid line and HF circles methods. Both sets of calculations were done for a lled shell of 240 atoms.

ingly both the K harmonic and sum rule methods disagree qualitatively with all eight of the lowest HF excitations. This difference is attributed to the fact that the Hartree-Fock method on its own can only describe single-particle excitations while both the sum rule and the K harmonic methods describe collective excitations in which the entire gas oscillates coherently. As another test of the K harmonic method we calculate the peak density of the gas. To do this we rst dene the density
N N

tion takes the form the antisymmetry of 0 =N

=F R gives
2

. Use of Eq. 3.13 and

dRR3N1 F R

rN

. 3.14

The hyperangular integration is carried out in Appendix D. The result is 0 = NI 0 where = N3/2
l3

dR

R3N1 F R R3

3.15

r =
j=1

d 3r j
i=1

ri r

3.13

NI 0 +3N/2

is the noninteracting peak density and . For the nth lled energy shell the noninterk3 1 2m f 2 = 6 6 2
3/2

It can be seen that integration over r using this denition gives r d3r = N. We recall that our separable approxima-

( +3 N1 /2)

acting peak density is given by

NI

n + 1/2

2 0/

1.9

1.8 0 2 4 6 kf 8 10

Note that the peak density is not given by R3N1 F R 2 evaluated at R = 0: this describes the probability of all of the particles being at the center at once. Figure 6 compares the K harmonic and HF peak densities. Clearly the density does decrease from the noninteracting value. The two methods are in good qualitative agreement, but the Hartree-Fock method seems to predict a slightly lower density, presumably a manifestation of the slightly in. ferior K harmonic wave function F R
B. Attractive interactions g

FIG. 5. The lowest breathing-mode excitation 0 in units of the trap frequency is plotted vs k f for the K harmonic method solid curve and for the sum rule circles . Also shown as dashed curves are the lowest eight radial excitation frequencies predicted in the Hartree-Fock approximation.

In this section we examine the behavior of the gas under 0 . For the inuence of attractive s-wave interactions k f attractive interactions the gas resides in a metastable region and can tunnel through the barrier shown in Fig. 2. Figure 7 shows the behavior of Vef f for several values of k f . The location of the local minimum gets pulled down with stronger attraction as the gas pulls in on itself and deeper into the

053624-6

HYPERSPHERICAL DESCRIPTION OF THE

5 4 Veff R ENI 3 2 1 0 0.5 R 1 R2 1.5
NI

PHYSICAL REVIEW A 74, 053624 2006

-3 -3 -3 -4

210 110

0.95
E/ENI

0.9 0.85 0.8

-12 -10 -8

-6 kf

-4

-2

510 0

FIG. 7. The dimensionless rescaled potential curve Vef f / ENI is shown as a function of the rescaled hyperradius R for several values of k f . k f = k f c dashed line and from top to bottom k f = 5.3, 8.3, 11.3, 19.3 all solid lines

-14

-12

-10

-8 -6 kf

(EK-EHF)/ENI

210

-4

-2

center of the trap. Further, as the strength of the interaction increases, the height of the barrier decreases. In fact beyond a critical interaction strength c the interaction becomes so strong that it always dominates over the repulsive kinetic term. At this critical point the local extrema disappear entirely and the gas is free to fall into the inner collapse region. The value of c can be calculated approximately by nding the point where Vef f loses its local minimum and becomes entirely attractive. This is not exact as the gas will have some small zero-point energy that will allow it to tunnel through or spill over the barrier before the minimum entirely disappears. This critical interaction strength is given by kf
c

FIG. 8. The ground-state energy in units of the noninteracting energy vs k f for 240 atoms calculated using the K harmonic curve and Hartree-Fock circles methods. Inset: the difference in the ground-state energies predicted by the K harmonic EK and Hartree-Fock EHF methods. Clearly the K harmonic prediction is slightly higher.

189 3 1 256 51/4

15.31.

Just before the minimum disappears, its location is given by Rmin = 51/4, with an energy of Vef f Rmin = 5ENI / 3 0.75ENI. This means that if the gas is mechanically stable for all values of the two-body scattering lengthi.e., a in this approximation, there must be a renormalization cutoff in the strength of the function such that k f 15.31 for all a. With this in mind, we begin to examine the behavior of the DFG for the allowed values of k f . Figures 8 and 9 show a comparison of the ground-state energy and rms radius of the gas versus k f down to the k f c as calculated in the K harmonic and Hartree-Fock methods. Again the Hartree-Fock method does just slightly better in energy, which we interpret as the Hartree-Fock method giving a slightly better representation of the actual ground-state wave function. The energy difference becomes largest as the interaction strength approaches the critical value. This increase is due to the fact that the Hartree-Fock method predicts that collapse occurs slightly earlier with k f c 14.1. As the interaction strength increases, the energy and rms radius of the gas decrease. As k f approaches k f c the overall size of the gas decreases sharply and from Eq. 3.15 we expect to see a very sharp rise in the peak density. This behavior is apparent in Fig. 10, which displays the peak density versus k f for both the K harmonic and Hartree-Fock methods. While the local minimum present in Vef f only supports metastable states, it is still informative to examine the behav-

ior of the energy spectrum versus k f , beginning with the breathing-mode frequency. As the interaction strength becomes more negative, Fig. 7 shows that the curvature about the local minimum in Vef f decreases. This softening of the hyperradial potential leads to a decrease in the breathingmode frequency in the outer well. Figure 11 shows the breathing mode versus k f predicted by the K harmonic curve method and also using the sum rule with HartreeFock orbitals circles . Also shown in Fig. 11 are the lowest eight Hartree-Fock excitation frequencies for a lled shell of 240 atoms. Again, the K harmonic method agrees quite well with the sum rule, while both differ qualitatively from the HF prediction. The sharp decrease in the breathing mode frequency that occurs as k f k f c is a result of the excited mode falling over the barrier into the collapse region as the barrier is pulled down by the interaction.

1 0.9 <R >/<R >NI 0.8 0.7 0.6 0.5 -15 -10 -5 0
2 2

kf

FIG. 9. The average squared radius of the Fermi gas ground state in units of the noninteracting value is plotted vs k f . The calculations are for 240 atoms in both the K harmonic method curve and Hartree-Fock method squares .

053624-7

RITTENHOUSE et al.

PHYSICAL REVIEW A 74, 053624 2006

peak density (oscillator units)

2 1.8 1.6 1.4 1.2 1 0.8 -14 -12 -10 -8 -6 kf -4 -2 0

E ENI

0.95 0.9 0.85 0.8 0.75 0.7 17 16 15 kf

FIG. 12. Color online A portion of the energy spectrum vs k f close to the critical point k f = k f c. Levels in the metastable region blue decrease slowly while levels in the collapse region red decrease very quickly. Energy levels above the barrier in Vs f purple ef reside in both the collapse region and the metastable region.

14

13

12

FIG. 10. The peak density in units of / m vs k f predicted by the K harmonic solid line and HF circles methods. Both calculations were carried out for a lled shell of 240 atoms.

3/2

Figure 12 displays some energy levels in the metastable region as functions of k f near k f c. Because of the singular nature of the 1 / R3 behavior in the inner region, we have added an inner repulsive 1 / R12 barrier to truncate the innitely many nodes of the wave function in the inner region. The behavior of the wave function is not correct within this region anyway because recombination would become dominant, and in any case the zero-range interaction is suspect beyond k f a 1 and it must be renormalized. Figure 12 shows three distinct types of energy level. Levels that are contained in the local minimum shown in blue are decreasing, but not as quickly as the others; levels that are in the collapse region shown in red have a very steep slope as they are drawn further into toward R = 0; and energy levels that are above the barrier in Vef f shown in purple have wave functions in both the collapse region and the local minimum. As k f decreases, the higher-energy levels fall

over the barrier into the collapse region earlier, until nally just before k f c is reached the lowest metastable level falls below the ground state. This corresponds to the breathingmode behavior seen in Fig. 11. Of course all of this applies only if there is no further hyperradial dependence of g. If the interaction becomes density dependent, as is the case in some types of renormalization, a change in the hyperradius will change the density and thus the interaction coupling parameteri.e., g g R .
C. Large-N limit

2.5

0/

In Secs. III A and III B the ground-state energy and expectation values discussed were found by solving the hyperradial effective Schrdinger equation 3.7 for a nite number of particles. Here we discuss the behavior of the N-fermion system in the limit where N is large. To do this we exploit the fact that the 2 / R2 term in the effective Hamiltonian, the hyperradial kinetic energy, becomes negligible. In this limit we see that the total energy of the system is merely given by E = Vef f Rmin , as is the case in dimensional perturbation theory 23 . To nd the ground-state energy we must merely nd the minimum local minimum for g 0 dV value of Vef f . Accordingly, we nd the roots of dRef f = 0. Using Eq. 3.11 we simplify this to kf = 1 R R 4 1 , 3 min min 3.16

1.5

-15

-10

kf

-5

FIG. 11. The frequency of the lowest-energy radial transition in units of the trap frequency vs k f predicted by the K harmonic method solid line and by the sum rule method circles . Also shown are the lowest eight radial transitions predicted by the Hartree-Fock method.

where Rmin is the hyperradial value that minimizes Vef f . The solutions to Eq. 3.16 are illustrated graphically in Fig. 13; for any given k f we need only look for the value of Rmin that gives that value. The exact solution of Eq. 3.16 cannot be determined analytically for all values of k f , but Fig. 13 shows that for k f 0 there is always only one positive, real Rmin that satises Eq. 3.16 . This corresponds to the global 0 minimum discussed for repulsive interactions. For k f things are a bit more complicated. Figure 13 shows that for kf c kf 0 there are two solutions to Eq. 3.16 . The inner

053624-8

HYPERSPHERICAL DESCRIPTION OF THE

PHYSICAL REVIEW A 74, 053624 2006

30 20 10 0
Energy difference (%)

-0.5

kf

-1

-1.5 -15

-10

-5

10 20 0.2 0.4 0.6 R' min 0.8 1

0 kf

10

15

FIG. 14. The percentage difference between the energy found by minimizing Vef f and the energy found by explicitly solving the hyperradial Schrdinger equation for 240 atoms.

FIG. 13. Plot of the interaction strength in the dimensionless combination k f vs the rescaled location s of the potential curve extrema, Rmin = Rmin / R2 NI, solutions of Eq. 3.16 . Examination of this plot tells us the behavior of Rmin for all allowed values of k f including the existence of the critical point k f c located at the minimum of the plot where the maximum and minimum coincide.

solution is a local maximum and corresponds to the peak of the barrier seen in Fig. 7; the outer solution corresponds to the local minimum where the DFG resides. The local minimum is the state that we are concerned with here as this will give the energy and hyperradial expectation values of the metastable Fermi gas. The value of k f where these two branches merge is the place where the local maximum merges with the local minimum: namely, the critical value k f c. Any value of k f less than k f c has no solution to Eq. 3.16 and thus we cannot say that there is a region of stability. For k f = 0, the noninteracting limit, we see that there are two solutions Rmin = 0 and Rmin = 1. The solution Rmin = 0 must be discounted as there is a singularity in Vef f at R = 0. Thus in the noninteracting limit Rmin 1 and Vef f Rmin ENI, as expected. Substitution of Eq. 3.16 into Eq. 3.11 gives the energy of the ground state, in the large-N limit, as a function of the size of the gas:
4 1 + 5Rmin Vef f Rmin = . 2 ENI 6Rmin

Another result from 3.10 is the fact that in the large-N limit the commutator Hef f , R 0. Thus for any operator that is solely a function of the hyperradius O R the groundstate expectation value in the large-N limit is given by the operator evaluated at Rmin: i.e., O R = O Rmin . This tells us that the large-N-limit wave function is given by R 3N1 /2G R 2 = R Rmin . We can perturb this slightly and say that the ground-state hyperradial wave function can be approximated by a very narrow Gaussian centered at Rmin. To nd the width of this Gaussian we approximate Vef f about Rmin as a harmonic oscillator with mass m* and frequency 0. We may nd 0 by comparing the oscillator potential with the second-order Taylor series about Rmin in Vef f i.e.,
0

1 1 m* ENI

Vef f R2

.
R =Rmin

3.17

The effective hyperradial oscillator length of R 3N1 /2G R is then given by l0 = / m* 0. The breathing-mode frequency is now simply found to be 0. The frequency in Eq. 3.17 is in units of the noninteracting energy, so to get back to conventional units we multiply by ENI / . From Eq. 3.10 we know that m* = mENIN R2 NI / 2. Noting that N R2 NI = l2ENI / this leads to
B 0

1 ENI

Vef f R2

3.18
R =Rmin

These solutions to Eq. 3.16 immediately give the groundstate energy of the gas versus k f . Figure 14 shows the percentage difference of the ground-state energy found by this minimization procedure and that of 240 particles found by diagonalizing Hef f in Eq. 3.7 . We see in Fig. 14 that for kf 10 the energy difference is less than 0.5%. As k f gets larger the difference increases, but in the range shown the difference is always less than 1.5%.

for B, the breathing-mode frequency in units of the trap 0 frequency. Using Eq. 3.11 and substituting into Eq. 3.16 to evaluate at the minimum gives that
B 0

1
4 Rmin

We note that this is now dependent only on the value of k f ; i.e., for a xed k f the predicted breathing mode is indepen-

053624-9

RITTENHOUSE et al.

PHYSICAL REVIEW A 74, 053624 2006

2 1.5
B 0

1 0.5 0

-10

kf

10

FIG. 15. The breathing mode in units of the oscillator frequency in the large-N limit vs k f . Note that as k f k f c the frequency drops to zero as the local minimum disappears.

dent of the number of atoms in the system in the large-N limit. This prediction can be compared with that predicted by the sum rule using the formula found by the authors of Ref. 22 :
B 0

4+

3 Eint . 2 Eho

3.19

Here Eint and Eho are the expectation values of the interaction potential and the oscillator potential in the ground state, respectively. If we insert the expectation values predicted here by the K harmonic method, we nd
B 0

4+3

kf 5 . Rmin

ingly, this approximation gives good results for the groundstate energy, rms radius, and peak number density of the gas. These are all in quantitative agreement with the HF method, while the breathing-mode frequency compares well with that found using the sum rule method, but gives a qualitative improvement over the lowest calculated HF radial excitation frequencies. The work presented here has been limited to the case of lled energy shellsi.e., magic numbers of atomsand it should be readily generalizable, with minor extensions, to open shells as well. The present results are limited to a spherically symmetric trap. Generalization to an cylindrically symmetric cigar-shaped trap should be possible with a judicious choice of coordinates and will be presented elsewhere. A more comprehensive description of this system would include the use of higher-order hyperspherical harmonics to get several adiabatic potential curves. The next-order term would involve expansions of all Slater determinants for singleparticle excitations 1619 and is beyond the scope of the present study. Since the -function Hamiltonian is too singular to have well-behaved solutions, in the absence of renormalization, we defer this discussion to future research. For strongly attractive interactions, this picture predicts an instability in the DFG in a manner qualitatively similar to the physics of an explosive collapse of a BEC or Bosenova 7 . For the gas to remain stable across the BEC-BCS crossover regime the strength of interaction potential must be 15.31. Preliminary results from bounded from below, k f another study 24 indicate that renormalization of the singular -function interaction accomplishes precisely that and apparently prevents collapse. The full interrelation between this picture and that of pairing in the BEC-BCS crossover region is beyond the scope of this study and is a subject that will be relegated to future publications.
ACKNOWLEDGMENTS

Substituting Eq. 3.16 for k f gives

B 0

1
4 Rmin

which agrees exactly with the frequency predicted in Eq. 3.19 . Figure 15 shows the breathing mode frequency predicted by Eq. 3.19 versus k f . We see the same behavior in this plot as was seen in Fig. 11 where the breathing-mode frequency dives to zero as k f k f c.
IV. SUMMARY AND PROSPECTS

This work was supported by funding from the National Science Foundation. We thank N. Barnea for introducing us to Refs. 18,19 . Discussions with L. Radzihovsky and V. Gurarie have also been informative.
APPENDIX A: HYPERRADIAL SOLUTION TO THE NONINTERACTING OSCILLATOR

We have demonstrated an alternative approach to describing the physics of a trapped degenerate Fermi gas from the point of view of an ordinary linear Schrdinger equation with two-body, microscopic interactions. The use of a hyperspherical variational trial wave function whose hyperangular behavior is frozen to be that of the K harmonic yields a one-dimensional effective potential 3.8 in a collective coordinate, the hyperradius R. This approach yields an intuitive understanding of the energy and size of the DFG in terms of familiar Schrdinger quantum mechanics. Perhaps surpris-

We wish to derive Eq. 2.15 , the nodeless hyperradial solution to the Schrdinger equation for N noninteracting particles in an isotropic oscillator. The Schrdinger equation for this system is given by
N i=1

2 2m

2 i

1 + m 2

2 2 ri

= 0,

A1

where r1 , r2 , . . . , rN are the Cartesian coordinates for each atom from the trap center. To begin we examine the radial Schrdinger equation for a single particle in an isotropic trap:

053624-10

HYPERSPHERICAL DESCRIPTION OF THE

PHYSICAL REVIEW A 74, 053624 2006

N

2 d2 2m dr2

+1 r2

1 + m 2

2 2

r En rf n r = 0. A2

2 + =
i=1

2ni +

A7

The solution to this is well known and is given by rf n r = An exp r2/2l r l

+1

Ln+1/2

r2 , l2

A3

We should note that this does not mean that the independentparticle solution and the hyperspherical solution are the same, only that the hyperspherical solution F R must be a linear combination of independent-particle solutions of the same energy.
APPENDIX B: PROOF OF THEOREM 1

with energy En = 2n + + 3 / 2 where l = / m and n is the number of radial nodes in the wave function. Transformation of Eq. A1 into hyperspherical coordinates using Eqs. 2.2 , 2.3 , 2.5 , and 2.7 yields a Schrdinger equation that separates into hyperradial and hyperangular pieces. The hyperangular solution is a hyperspherical harmonic that diagonalizes 2. The resulting hyperradial Schrdinger equation is given by 2 d2 K K + 1 2M dR2 R2 1 + M 2
2 2

R E R 3N1 /2F R = 0, A4

where K = + 3 N 1 / 2. Comparing this with Eq. A2 we see that if we make the substitutions K, n , m M, r R, and rf n r R 3N1 /2F K R , the single-particle radial Schrdinger equation becomes the N-particle hyperradial Schrdinger equation. With these replacements the solution is evidently R 3N1 /2G
K

Here we will use the results from Appendix A to prove theorem 1. We proceed by assuming that there exists a fully antisymmetric, ground-state, hyperspherical solution to the Schrdinger equation for N noninteracting fermions in an 0. If isotropic trap with energy given by Eq. A5b with we can show that this leads to a contradiction, then we have, from Eq. A7 , that there is only one for all of the groundstate congurations. From Appendix A we know that this hyperspherical solution must be a linear combination of antisymmetric, degenerate, ground-state solutions in independent-particle coordinates: F
K

R =

1,

2,

... ,

N 1, 2,

D r1,r2, . . . ,rN,

... ,

R =A

exp R2/2L

R L

K+1

LK+1/2

R2 , L2 A5a

where L = l / N and is the number of hyperradial nodes in the N-body system. For = 0 this is the same hyperradial wave function written in Eq. 2.15 . The total energy is given by E
K

where each D is of the form of Eq. 2.13 , is used to distinguish between degenerate states of the same energy, and F K R is given by Eq. A5a . Again, for brevity, we suppress the spin coordinates 1 , 2 , . . . , N in . We now note that the hyperradius R is completely symmetric under all transpositions of particle coordinates; thus all of the trans. position symmetry must be contained in the function we construct a new completely antisymmeFrom which has entrized hyperspherical solution F0K R ergy E= + 3N . 2

2 +K+

3 = 2

2 + +

3N . 2

A5b

Now that we have a hyperspherical solution we can compare it with the solution to Eq. A1 written in terms of independent-particle coordinates r1 , r2 , . . . , rN . This equation is clearly separable in each coordinate ri, and its solution is a product of N single-particle wave functions
N

Use of Eqs. A5b and A7 gives = ENI 3N 2 , 2 B1

=
i=1

f ni i r i y

imi

N where ENI = 3N / 2 + i=1 2ni + i is the ground-state enN ergy as dened by any of the functions D ri i=1 . Thus our has energy new function F0K R

where f ni i r is given by Eq. A3 , i are the spherical polar is a norangular coordinates of the ith particle, and y m mal 3D spherical harmonic. The energy for this independentparticle solution is given by E= 3N + 2
N

E = ENI 2 , which would lie below the ground-state energy, a contradiction unless = 0. In the above analysis we assumed a degenerate set of solutions at the ground-state energy. For nondegenerate solutions the proof becomes trivial, as any nondegenerate solution must be the same no matter what coordinate system it is expressed in. From this the rest of the proof follows in the same way, and the nal, unique for this system is given by Eq. B1 ,

2ni +
i=1

A6

Now that we have seen that Eq. A1 is separable in both hyperspherical and independent-particle coordinates, we may compare Eqs. A5b and A6 to nd that

053624-11

RITTENHOUSE et al.

PHYSICAL REVIEW A 74, 053624 2006

ENI

3N . 2

B2

1 = 2 i,j=1
ms ms
i j

d 3r 1d 3r 2
* i

* i

r1

* j

r2 Uint r21
i

r1

r2 C4

r2

* j

r1 Uint r21

r1

r2 ,

APPENDIX C: CALCULATION OF THE s-WAVE INTERACTION MATRIX ELEMENT

To calculate CN it is useful to start with a more general interaction. We assume that the interaction term in the total N-body Hamiltonian is such that at a xed hyperradius Uint rij is separable into a hyperradial function times a hyperangular integrali.e., U r12
vRv

where i r is the ith single-particle spatial wave function that appears in the product in Eq. 2.13 i.e., r
i

r = Nni i exp r2/2l2 r/l

i+1

Lni+1/2 r/l
i

imi

. C5

C1

From properties of the function and Eq. 2.4 it is easy to see that Uint r = g 3 r ts this criterion. While v might have some very complex form, it will be seen shortly that only the form of V R and Uint rij will matter: Uef f R =
i j

With this result we can now specify to the s-wave interaction. Following Eq. C1 we may identify Uint r21 g 3 r21 and V21R R 1 / R3. With this and Eq. 2.15 , the integral in Eq. C3b is found to be + = L3 3 N1 2 3N + 2

C6

Uint rij

. to exchange ri

We may again use the antisymmetry of r2 and r j r1 and arrive at Uef f R = N N1 2 Uint r21

Evaluating Eq. C4 begins by integrating the function over r2. This is simple, and we can clearly see that the two terms in the integral in Eq. C4 have a common factor of 2 2 i r1 j r1 . Factoring this out gives = g 1 2 i,j=1
N ms ms
i

From Eq. C1 we can write that Uef f R = VR N N 1 /2, C2

d 3r 1

r1

r1

C7

is the analog of gCN / N3/2 in Eq. where = v 3.5 . To nd we may substitute in the denition from Eq. 2.16 , multiplying by R3N1G R 2 on both sides, , integrating over R, and using Eq. C1 to replace V R v which gives a simple equation that may be solved for : = , R3N1G R 2V R dR,
N

This sum runs over all spatial and spin quantum numbers in a lled energy shell; we may break this up into two factors, a spin sum and a space sum: g = 1 2 ms ,ms =1/2
i j

1/2

ms ms
i

d 3r 1

r1

C3a =g d 3r 1

r1

C3b

r1

r1

C8

N N1 2

d 3r j D * r i
j=1

N i=1

Uint r21 D ri

N i=1

, C3c

where the Greek letters and stand for the set of spatial quantum numbers n , , m . The spin sum is trivial and is 1/2 given by m ,m =1/2 1 ms ms = 2. We may now calculate CN for any given lled energy shell using Eq. C8 and plugging into the relationship CN = N3/2 . g C9
si sj
i j

N where D ri i=1 is the Slater determinant dened in Eq. 2.13 and G R is dened as in Eq. 2.15 . We have also N

used the fact that R3N1dRd =

j=1

d3r j 13 .

can now be

found as a ratio of two integrals. The integral on the left-hand side LHS of Eq. C3a can be calculated directly. The integral on the right-hand side of Eq. C3a is now a diagonal, determinantal matrix may be drastically simplied by using the orelement. thogonality of the single-particle basis functions for details see Ref. 25 , Sec. 6-1 :

This has been done for the rst 100 lled shells, the results of which are summarized by Fig. 1. CN for a selection of shells is given in Table I. To extract CN in the limit as N we may examine the expression for in Eq. C8 . We may note that the sum r = r 2 is the denition of the density of a spin polarized degenerate Fermi gas of noninteracting particles in an isotropic oscillator. In the limit where N the trap energy of the noninteracting gas dominates the total energy. This

053624-12

HYPERSPHERICAL DESCRIPTION OF THE

PHYSICAL REVIEW A 74, 053624 2006

means that the Thomas-Fermi approximation becomes exact in this limit. Thus we may write that r = 1 6
2

2m
2

3/2

m 2r 2 2

3/2

C10a

N ms =

d 3r r .

C10b

may be found by the condition The chemical potential given in Eq. C10b where Nms is the number of particles with the same spin projection ms. The system we are considering is an equal-spin mixture so that N = N = N / 2. Thus we nd that = 3N 1/3. Plugging this into Eq. C8 gives =g r
2 3

describing the center-of-mass motion, G R is a nodeless hyperradial function, and is the lowest hyperspherical harmonic in the 3N 4 angular coordinates. We are then HR, where the intelooking for the matrix element gral is taken over all hyperangles at xed R. Theorem 1 still applies with the added idea that is given by the lowest s-wave state of the center of mass in an oscillator. From here the analysis presented in this section for trap centered coordinates still holds with the added change that the dimension of the hyperradial integral in Eq. C3b is three dimensions smaller. This leads to a factor in the interaction matrix element given by
2 3 N1 2 3N 3 N2 + + 2 2

d r=g

2 256 N3/2 . 3 315 3l3

C11

CN

CN .

We insert this into Eq. C9 with Eq. C6 to nd that 2 256 3/2 N 3 315 3 3N 2 3 N1 + 2 +

CN =

C12

For smaller N this factor changes the interaction considerably, but for larger N it quickly goes to 1. Thus, in the largeN limit, besides extracting the center-of-mass sloshing modes, there is very little difference from the trap center coordinate systems in this Jacobi coordinate formalism.
APPENDIX D: CALCULATING 0

Using Eq. 3.4a for the large-N behavior of , in the limit N , CN 2 32 7/2 N , 3 35 3 C13

We start from Eq. 3.14 : 0 =N dRR3N1 F R

2

which is the quoted large-N behavior in Eq. 3.6 . It should be said that the same formalism that is presented in this paper can be applied to this system in center-of-mass coordinates. This is done by rst choosing an appropriate set of Jacobi coordinates: xi = i / i + 1 ij=1r j / i ri+1 for i = 1 , . . . , N 1 with the center-of-mass vector dened as xcm = N r j / N. Hyperangular coordinates are now used to dej=1 scribe the 3N 3 degrees of freedom in the Jacobi coordi2 nates where R2 = N1x2 / N = N r2 xcm / N. The 3N 4 hyj=1 j j=1 j perangles needed are now dened with respect to the lengths of the Jacobi vectors in the same way as in Eqs. 2.3 and 2.4 . Under this coordinate transformation we nd the Hamiltonian to be given by H = HCM + HR, , where HCM is the Hamiltonian for the center-of-mass coordinate xcm and HR, is an operator entirely dened by hyperspherical coordinatesi.e., HCM = 2 2m
2 cm +

rN

2 , then we have the soluIf we can nd d 3 rN tion. From Eq. 2.4 and properties of the function we may say that

rN

R3

We multiply this on both sides by the noninteracting hyperradial function R3N1 G R 2 and integrate over R: N =
3

rN GR R3
2

GR

2 3N1

dRd D1

R3N1dR.

1 m 2

2 2 xcm ,

Inserting G R from Eq. 2.15 into the right-hand side this gives
2 2

HR, =

2 2 1 1 R3N4 2 + M 3N4 2 2M R R R R

G R 2 3N1 R dR = N3/2R3

3 N 1 /2 . L 3N/2
3

D2a

+
i j

Uint rij .

With this Hamiltonian we make the ansatz that for H = E , = xcm G R , where is a wave function

From the denition of in Eq. 2.16 we see that the lefthand side of Eq. D1 is a determinantal matrix element of a single-particle operator integrated over all independentparticle coordinates:

053624-13

RITTENHOUSE et al.

PHYSICAL REVIEW A 74, 053624 2006

rN
N

GR

2 3N1

dRd

NI

l3 N
3/2

3N/2 . 3 N 1 /2

=N
j=1

d 3r j D r i

N 2 3 i=1

rN ,

D2b

The peak density is then given by 0 = dR R3N1 F R R3

2

N where D ri i=1 is the Slater determinant wave function dened in Eq. 2.13 . Referring to the denition of the density in Eq. 3.13 we see that this is merely the peak density of the non-interacting system NI 0 . Thus

which is what we were seeking to show.

1 M. Bartenstein, A. Altmeyer, S. Riedl, S. Jochim, C. Chin, J. H. Denschlag, and R. Grimm, Phys. Rev. Lett. 92, 120401 2004 . 2 T. Bourdel, L. Khaykovich, J. Cubizolles, J. Zhang, F. Chevy, M. Teichmann, L. Tarruell, S. J. J. M. F. Kokkelmans, and C. Salomon, Phys. Rev. Lett. 93, 50401 2004 . 3 J. Kinast, S. L. Hemmer, M. E. Gehm, A. Turlapov, and J. E. Thomas, Phys. Rev. Lett. 92, 150402 2004 . 4 C. A. Regal, M. Greiner, and D. S. Jin, Phys. Rev. Lett. 92, 40403 2004 . 5 M. W. Zwierlein, C. A. Stan, C. H. Schunck, S. M. F. Raupach, A. J. Kerman, and W. Ketterle, Phys. Rev. Lett. 92, 120403 2004 . 6 S. T. Rittenhouse, M. J. Cavagnero, J. von Stecher, and C. H. Greene, e-print cond-mat/0510454. 7 J. L. Bohn, B. D. Esry, and C. H. Greene, Phys. Rev. A 58, 584 1998 . 8 Y. E. Kim and A. Zubarev, J. Phys. B 33, 55 2000 . 9 D. Kushibe, M. Mutou, T. Morishita, S. Watanabe, and M. Matsuzawa, Phys. Rev. A 70, 63617 2004 . 10 Y. F. Smirnov and K. V. Shitikova, Sov. J. Part. Nucl. 8, 44 1977 .

11 D. Blume and C. H. Greene, J. Chem. Phys. 112, 8053 2000 . 12 B. D. Esry, C. H. Greene, and J. P. Burke, Phys. Rev. Lett. 83, 1751 1999 . 13 J. Avery, Hyperspherical Harmonics: Applications in Quantum Theory Kluwer Academic, Dordrecht, 1989 . 14 N. Barnea, J. Math. Phys. 40, 1011 1999 . 15 M. J. Cavagnero, Phys. Rev. A 33, 2877 1986 . 16 M. Fabre de la Ripelle and J. Navarro, Ann. Phys. N.Y. 123, 185 1978 . 17 M. Fabre de la Ripelle, S. A. Soanos, and R. M. Adam, Ann. Phys. N.Y. 316, 107 2005 . 18 N. K. Timofeyuk, Phys. Rev. C 65, 064306 2002 . 19 N. K. Timofeyuk, Phys. Rev. C 69, 034336 2004 . 20 E. Fermi, Nuovo Cimento 11, 157 1934 . 21 B. D. Esry and C. H. Greene, Phys. Rev. A 60, 1451 1999 . 22 L. Vichi and S. Stringari, Phys. Rev. A 60, 4734 1999 . 23 T. C. Germann, D. R. Herschbach, M. Dunn, and D. K. Watson, Phys. Rev. Lett. 74, 658 1995 . 24 B. M. Fregoso and G. Baym, Phys. Rev. A 73, 043614 2006 . 25 R. D. Cowan, The Theory of Atomic Structure and Spectra University of California Press, Berkeley, 1981 .

053624-14