7.

2 Bose–Einstein condensation in ultracold atomic gases 191

see equations (7) and (27). Eliminating T , we obtain

Pv5/3 = const. (47)

as the equation for an adiabat of the ideal Bose gas.

Incidentally, the foregoing results are exactly the same as for an ideal classical gas. There
is, however, a significant difference between the two cases; that is, while the exponent 35 in
formula (47) is identically equal to the ratio of the specific heats CP and CV in the case of
the ideal classical gas, it is not so in the case of the ideal Bose gas. For the latter, this ratio
is given by

CP 4 CV g1/2 (z)
γ≡ = 1+ (48a)
CV 9 Nk g3/2 (z)
5 g5/2 (z)g1/2 (z)
= ; (48b)
3 {g3/2 (z)}2

see Problems 7.4 and 7.5. It is only for T  Tc that γ ' 53 . At any finite temperature, γ > 5
3
and as T → Tc , γ → ∞. Equation (47), on the other hand, holds for all T .
In the mixed-phase region (T < Tc ), the entropy of the gas may be written as
 
5
5 ζ 2
S = Ne · k   ∝ Ne ; (49)
2 ζ 3
2

see equations (20) and (44b). As expected, the N0 particles that constitute the “condensate”
do not contribute toward the entropy of the system, while the Ne particles that constitute
the normal part contribute an amount of 25 kζ ( 25 )/ζ ( 32 ) per particle.

7.2 Bose–Einstein condensation in ultracold

atomic gases
The first demonstration of Bose–Einstein condensation in ultracold atomic gases came in
1995. Cornell and Wieman Bose-condensed 87 Rb (Anderson, Ensher, Matthews, Wieman,
and Cornell (1995)) and Ketterle Bose-condensed 23 Na (Davis, Mewes, Andrews, van
Druten, Durfee, Kurn, and Ketterle (1995)) using magneto-optical traps (MOTs) and
magnetic traps to cool vapors of tens of thousands of atoms to temperatures of a few
nanokelvin.9 A survey of the theory and experiments can be found in Pitaevskii and
Stringari (2003), Leggett (2006), and Pethick and Smith (2008).
The first step of the cooling of the atomic vapor uses three sets of counter-propagating
laser beams oriented along cartesian axes that are tuned just below the resonant frequency
9
Since 1995, many isotopes have been Bose-condensed including 7 Li, 23 Na, 41 K, 52 Cr, 84 Sr, 85 Rb, 87 Rb, 133 Cs, and
174 Yb. The first molecular Bose–Einstein condensates were created in 2003 by the research groups of Rudolf Grimm at the

University of Innsbruck, Deborah S. Jin at the University of Colorado at Boulder, and Wolfgang Ketterle at Massachusetts
Institute of Technology.
192 Chapter 7 . Ideal Bose Systems

of the atoms in the trap. Atoms that are stationary are just off resonance and so rarely
absorb a photon. Moving atoms are Doppler shifted on resonance to the laser beam that is
propagating opposite to the velocity vector of the atom. Those atoms preferentially absorb
photons from that direction and then reemit in random directions, resulting in a net
momentum kick opposite to the direction of motion. This results in an “optical molasses”
that slows the atoms. This cooling method is constrained by the “recoil limit” in which the
atoms have a minimum momentum of the order of the momentum of the photons used
to cool the gas. This gives a limiting temperature of (hf )2 /2mc2 k ≈ 1 µK, where f is the
frequency of the spectral line used for cooling and m is the mass of an atom.
In the next step of the cooling process, the lasers are turned off and a spatially vary-
ing magnetic field creates an attractive anisotropic harmonic oscillator potential near the
center of the magnetic trap

1  2 2 
V (r) = m ω1 x + ω22 y 2 + ω32 z2 . (1)
2

The frequencies of the trap ωα are controlled by the applied magnetic field. One can then
lower the trap barrier using a resonant transition to remove the highest energy atoms in
the trap. If the atoms in the vapor are sufficiently coupled to one other, then the remaining
atoms in the trap are cooled by evaporation.
If the interactions between the atoms in the gas can be neglected, the energy of each
atom in the harmonic oscillator potential is

1
εl1 ,l2 ,l3 = ~ω1 l1 + ~ω2 l2 + ~ω3 l3 + ~(ω1 + ω2 + ω3 ) , (2)
2

where lα (= 0, 1, 2, . . . ∞) are the quantum numbers of the harmonic oscillator. If the

three frequencies are all the same, then the quantum degeneracy of a level with energy
ε = ~ω(l + 3/2) is (l + 1)(l + 2)/2; see Problem 3.26.
For the general anisotropic case, the smoothed density of states as a function of energy
(suppressing the zero point energy and assuming ε  ~ωα ) is given by

Z∞Z∞Z∞
ε2
δ ε − ~ω1 l1 − ~ω2 l2 − ~ω3 l3 dl1 dl2 dl3 =


a(ε) = , (3)
2 (~ω0 )3
0 0 0

where ω0 = (ω1 ω2 ω3 )1/3 ; this assumes a single spin state per atom. The thermodynamic
potential 5, see Appendix H, for bosons in the trap is then given by

4 Z∞
4
kT 2

−x βµ
 kT
5(µ, T ) = − x ln 1 − e e dx = g4 (z), (4)
2 (~ω0 )3 (~ω0 )3
0
 7.2 Bose–Einstein condensation in ultracold atomic gases 193

where z = exp(βµ) is the fugacity and gν (z) is defined in Appendix D. Volume is not a
parameter in the thermodynamic potential since the atoms are confined by the har-
monic trap. The average number of atoms in the excited states in the trap is
3
∂5
  
kT
N(µ, T ) = = g3 (z) . (5)
∂µ T ~ω0

For fixed N, the chemical potential monotonically increases as temperature is lowered

until Bose–Einstein condensation occurs when µ = 0 (z = 1). The critical temperature for
N trapped atoms is then given by
 1/3
kTc N
= , (6)
~ω0 ζ (3)

where ζ (3) = g3 (1) ' 1.202. While the spacing of the energy levels is of order ~ω0 , the crit-
ical temperature for condensation is much larger than the energy spacing of the lowest
levels for N  1. A typical magnetic trap oscillation frequency f ≈ 100Hz. For N = 2 × 104 ,
as in Cornell and Wieman’s original experiment, kTc /~ω0 ≈ 25.5. The observed critical
temperature was about 170nK (Anderson et al. (1995)).
For T < Tc , the number of atoms in the excited states is
3 3
ζ (3)
 
Nexcited kT T
= = , (7)
N N ~ω0 Tc

so the fraction of atoms that condense into the ground state of the harmonic oscillator is
 3
N0 T
= 1− ; (8)
N Tc

see de Groot, Hooyman, and ten Seldam (1950), and Bagnato, Pritchard, and Kleppner
(1987). In the thermodynamic limit, a nonzero fraction of the atoms occupy the ground
state for T < Tc . By contrast, the occupancy of the first excited state is only of order N 1/3 , so
in the thermodynamic limit the occupancy fraction in each excited state is zero. A compar-
ison of the experimentally measured Bose-condensed fraction with equation (8) is shown
in Figure 7.7.

7.2.A Detection of the Bose–Einstein condensate

The linear size of the ground state wavefunction in cartesian direction α is
s
~
aα = , (9)
mωα
194 Chapter 7 . Ideal Bose Systems

1.0
12

N(104)
0.8
8

4
0.6

N0/N
0.0 0.5 1.0 1.5
0.4

0.2

0.0
0.0 0.5 1.0 1.5
T/Tc(N )

FIGURE 7.7 Experimental measurement of the Bose-condensed fraction vs. temperature, as compared to
equation (8). The scaled temperature on the horizontal axis is the temperature divided by the N-dependent critical
temperature given in equation (6). The inset shows the total number of atoms in the trap after the evaporative
cooling. From Ensher et al. (1996). Reprinted with permission; copyright © 1996, American Physical
Society.

while the linear size of the thermal distribution of the noncondensed atoms in that
direction is

s s
kT kT
athermal = = aα . (10)
mωα2 ~ωα

At trap frequency f = 100 Hz and temperature T = 100 nK, these sizes are about 1 µm
and 5 µm, respectively. Instead of measuring the atoms directly in the trapping potential,
experimenters usually measure the momentum distribution of the ultracold gas by a time-
of-flight experiment. At time t = 0, the magnetic field is turned off suddenly, eliminating
the trapping potential. The atomic cloud then expands according to the momentum dis-
tribution the atoms had in the harmonic trap. The cloud is allowed to expand for about 100
milliseconds. The speed of the atoms at this temperature is a few millimeters per second,
so the cloud expands to a few hundred microns in this period of time. The cloud is then
illuminated with a laser pulse on resonance with the atoms, leaving a shadow on a CCD
in the image plane of the optics. The size and shape of the light intensity pattern directly
measures the momentum distribution the atoms had in the trap at t = 0. The expanding
cloud can be divided into two components, the N0 atoms that had been Bose-condensed
into the ground state and the remaining N − N0 atoms that were in the excited states of the
harmonic oscillator potential. The Bose-condensed atoms have smaller momenta than the
atoms that were in the excited states. After time t, the quantum evolution of the ground
 7.2 Bose–Einstein condensation in ultracold atomic gases 195

state has a spatial number density

 !
3 2
N 0
Y 1 −r α
n0 (r, t) = N0 |ψ0 (r, t)|2 = 3/2  q exp
; (11)
π aα2 1 + ωα2 t 2
α=1 aα 1 + ωα
2 t2

see Pitaevskii and Stringari (2003), Pethick and Smith (2008), and Problem 7.15.
The atoms that are not condensed into the ground state can be treated semiclassi-
cally, that is, the position-momentum distribution function is treated classically while the
density follows the Bose–Einstein distribution function:

1
f (r, p, 0) = . (12)
βp2 βm
 
ω12 x2 + ω22 y 2 + ω32 z2 − βµ − 1


exp 2m + 2

After the potential is turned off at t = 0, the distribution evolves ballistically:

pt
 
f (r, p, t) = f r + , p, 0 . (13)
m

The spatial number density of atoms in the excited states is

pt
 
1
Z
nexcited (r, t) = f r + , p, t dp , (14)
h3 m

which can be integrated to give

  !
3
1 X eβµj  Y 
∞
1 −βjmωα2 rα2 
nexcited (r, t) = 3 exp , (15)
λ j3/2  2 1 + ωα2 t 2
q

1 + ω2 t 2
α=1

j=1 α

√
where λ = h/ 2πmkT is the thermal deBroglie wavelength; see Pethick and Smith (2008),
and Problem 7.16. The integrals over the condensed state and the excited states correctly
count all the atoms:
Z
N0 = n0 (r, t)dr, (16a)
Z
N − N0 = nexcited (r, t)dr = Nexcited ; (16b)

see Problem 7.18.

Note that at early times (ωα t  1) both the condensed and the excited distributions are
anisotropic due to the anisotropic trapping potential. However, at late times (ωα t  1), the
atoms from the excited states form a spherically symmetric cloud because of the isotropic
momentum dependence of the t = 0 distribution function. By contrast, the atoms that
were condensed into the ground state expand anisotropically due to the different spa-
tial extents of the ground state wavefunction at t = 0. The direction that has the largest
196 Chapter 7 . Ideal Bose Systems

FIGURE 7.8 The two-dimensional time-of-flight number density equations (11) and (15) at late times (ω0 t  1) for
T /Tc √
= 0.98 using the experimental parameters in Anderson et al. (1995): N = 2 × 104 atoms in the trap and
ω2 = 8ω1 . The plot shows the full density and, underneath, the broader isotropic density just due to the excited
states. The z-dimension has been integrated out. The Bose-condensed peak is anisotropic: the y-direction spread is
81/4 = 1.68 times larger than in the x-direction while the broad peak caused by the excited states is isotropic. The
√
distance scale v0 t = t ~ω1 /m determines the width of the distribution that results from the Bose-condensed peak
in the x-direction at late times; compare to Figure 7.9.

ωα is quantum mechanically squeezed the most at t = 0; so, according to the uncertainty

principle, it expands the fastest. This is an important feature of the experimental data that
confirms the onset of Bose–Einstein condensation;10 see Figures 7.8 and 7.9.

7.2.B Thermodynamic properties of the Bose–Einstein condensate

The temperature, condensate fraction, and internal energy can all be observed using time-
of-flight measurements. The internal energy can also be written in terms of the function
gν (z):

Z∞
4
ε3 1 kT
U(µ, T ) = dε = 3 g4 (z). (17)
2(~ω0 )3 eβ(ε−µ) − 1 (~ω0 )3
0

10
Repulsive interactions between atoms create additional forces that modify the time-of-flight expansion. This is
especially important in condensates with a very large numbers of atoms, as many as 107 or more in some experiments;
see Section 11.2.A.
 7.2 Bose–Einstein condensation in ultracold atomic gases 197

FIGURE 7.9 Time-of-flight images from the first observation of Bose–Einstein condensation in a dilute vapor of 87 Rb
by Anderson et al. (1995) at temperatures just above and below the phase transition temperature. The anisotropic
pattern of the Bose-condensed fraction is evident; compare to Figure 7.8. Courtesy of NIST/JILA/University of
Colorado.

The heat capacity at constant number can be written as

∂U ∂U ∂U ∂µ
    
  
CN (T ) = = +
∂T N ∂T µ ∂µ T ∂T N
   
∂U ∂N
∂U ∂T µ ∂T µ
 
= − . (18)
∂T
 
µ ∂N
∂µ T

Equations (5) and (6) can be used to determine the fugacity z numerically, as shown in
Figure 7.10(a). The fugacity can then be used in equation (17) to obtain the scaled internal
energy
  4
T ζ (4)
3 for T ≤ Tc ,


 Tc ζ (3)


U
= (19)
NkTc   4

 T g4 (z)
3
 for T ≥ Tc ;
Tc ζ (3)
198 Chapter 7 . Ideal Bose Systems

see Figures 7.10(b) and 7.12. The scaled specific heat is given by
 3
 12ζ (4) T

for T < Tc ,

CN  ζ (3) Tc


= (20)
Nk 
!
T 3 9g32 (z)
 1  
 12g4 (z) − for T > Tc ,
ζ (3) Tc g2 (z)



and is shown in Figure 7.11. Unlike the case of Bose–Einstein condensation of free parti-
cles in a box (Figure 7.4), the specific heat of a condensate in a harmonic trap displays a

1.5 6

1.0 4
z ⫽ exp(␤␮)

NkTc
U
0.5 2

0.0 0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
T/Tc T/Tc
(a) (b)

FIGURE 7.10 Fugacity (a) and scaled internal energy (b) vs. scaled temperature (T /Tc ) for a Bose–Einstein
condensate in a harmonic trap.

12

10

8
CN
Nk

0
0.0 0.5 1.0 1.5 2.0
T/Tc

FIGURE 7.11 Scaled specific heat of a Bose–Einstein condensate in a harmonic trap as a function of the scaled
temperature (T /Tc ); compare with Figure 7.4 for a free-particle Bose gas.
 7.2 Bose–Einstein condensation in ultracold atomic gases 199

discontinuity at the critical temperature:

 12ζ (4)
 ' 10.805 as T → Tc− ,
 ζ (3)


CN
→ (21)
Nk  12ζ (4) 9ζ (3)
→ Tc+ .

 − ' 4.228 as T
ζ (3) ζ (2)


Figure 7.12 shows experimental data for the internal energy of a Bose–Einstein con-
densate of 87 Rb. The break in slope is an indication of the discontinuous specific heat.
Naturally, in a system with a finite number of particles, all nonanalyticities associated with
the phase transition are removed. When N is finite, the condensate fraction approaches
zero smoothly and the discontinuity in the heat capacity is rounded off. Pathria (1998) has
derived N-dependent temperature markers that indicate the onset of Bose–Einstein con-
densation in terms of the condensate fraction and the specific heat; see also Kirsten and
Toms (1996) and Haugerud, Haugest, and Ravndal (1997).

1.5
U/NkTc

0.9 1 1.1
1
1.5

1.4

0.5 1.3

1.2

1.1
0
0.5 1 1.5
T/Tc

FIGURE 7.12 Comparison of the experimental measurements of Ensher et al. (1996) (diamonds) with the
noninteracting internal energy result — see equation (19) and Figure 7.10(b) — (dotted curve), the zero-order
solution including interactions (full curve), first-order perturbative treatment (dashed curve), and numerical
solution (circles). The straight line is the classical Maxwell–Boltzmann result. The inset is an enlargement of
the region around the critical temperature. The break in slope is an indication of the discontinuity in the
thermodynamic limit specific heat shown in Figure 7.11; from Minguzzi, Conti, and Tosi (1997). Reprinted
with permission; copyright © 1997, American Institute of Physics.