Ultra Cold Quantum GasesProblem Set 3

Elvis BejkoGroup A
LudwigMaximiliansUniversit
at, Fakult
at f
ur Physik
WS 2013-14

1 Rabi oscillations in the dressed state picture

We numerically solved the Schrodinger equation for the Hamiltonian

0 0
~
=
,
H
2
0 2
where L 21 and 0

(1)

e(x )E0
. Our solution is presented in Fig.1. We notice that as we increase the
~

detuning , our occupation probability oscillation increases its frequency.1 At resonance we see that our atom
undergoes, periodically, population inversion with probability 1. As we detour from resonance, the population
inversion occurs with ever decreasing probability.
1.0

=0

=0.5

=0.7

=1.5

0.8

|c1,2|2

0.6

0.4
0.2
0.0
1.0
0.8

|c1,2|2

0.6

0.4
0.2
0.0

Figure 1. The numerical solution to the Schrodinger equation with Hamiltonian given by Eq.(3). We solved it using Python and
the RungeKutta 4 integration scheme. The solid line corresponds to |c2 (t)|2 , while the dashed one to |c1 (t)|2 .

At resonance, the Hamiltonian assumes a simple form

0 0
~
=
.
H
2
0 0

Elvis.Bejko@campus.lmu.de

It is obvious if one looks at the various graphs of Fig.1 vertically.

(2)

Excited state occupation probability for various

0.8

|c2|2

0.6
0.4
0.2
0.0

Figure 2. The excited state occupation probability as a function of time. The effect of detuning is more obvious in this figure;
i.e. with increasing , the amplitude decreases while the frequency increases.

Finding the eigenvalues and the eigenvectors of the above Hamiltonian, is quite straightforward; we get E1,2 = 0 ,
with corresponding eigenvectors

1
1
|S1,2 i =
2
1
1
1
= |gi + |ei .
2
2

(3)

Since the dressed state vectors |S1,2 i form a complete set, we can write the initial state, |gi, as follows

|gi = |S1 i hS1 | gi + |S2 i hS2 | gi

1
1
= |S1 i |S2 i .
2
2

(4)

The Schrodinger equation

i~

d
|i ,
|i = H
dt

(5)

has the formal solution

it

|(t)i = e ~ H |(0)i .

(6)

It is obvious, from the formal solution, that if the initial state |(0)i is an eigenstate of the Hamiltonian,2 then the
operation of the exponential operator on that state is trivial, in the sense that

|(t)i = e

itEn
~

|(0)i .

(7)

In other words, the time evolution appends a phase in front of the eigenstate. If the initial state is a superposition
itEn
of eigenstates of the Hamiltonian, each eigenstate will get a different phase e ~ , depending on its energy.

|(0)i = |ni.

3
Having established the above, in the sudden approximation, the evolution of the ground state after we turn on
our light field, is given by


i0 t
1  i0 t
|0 (t)i |g(t)i = e 2 |S1 i e+ 2 |S2 i .
2

(8)

To calculate the population of the excited state at time t; |he| (t)i| , we need to express the excited state as a
function of |S1,2 i, as we did for the ground state in Eq.(4)

|ei = |S1 i hS1 | ei + |S2 i hS2 | ei

1
1
= |S1 i + |S2 i .
2
2

(9)

Now we just need to take the inner product of the above two Eqs (8-9). Taking into account that the crossterms
vanish, due to the fact that the |S1,2 i are orthogonal, we get

he| (t)i =


i0 t
1  i0 t
e 2 e+ 2 .
2

(10)

Multiplying with the complex conjugate, we arrive at the desired result

|c2 (t)|2 =

1
(1 cos0 t) ,
2

(11)

which is the solution that matches exactly Figs (1-2)!

Now that we have the analytic solution, it is obvious that a pulse, of time duration = /0 , will completely
invert the population of the atom from the ground state to the excited state, because

|c2 (/0 )|2 = 1.

(12)

By the same token, a /2pulse will send the atom to a superposition state where, the atom is equally probable to be
in the excited as it is to be in the ground state!

2 Magnetic trap: 2D quadrupole

The magnetic field of an infinitely long currentcarrying wire can be found from Amperes law

I
B dl = 0 I,

(13)

whereby, symmetry arguments lead to

B=

0
I ,
2r

(14)

where we denote by the unit vector, in polar coordinates, that circulates around the origin.
Let us calculate, explicitly, the components of the magnetic field (cartesian coordinates) induced by the wire in the
origin of our coordinate system. From elementary geometrical considerations, it is clear that

= sin
x + cos
y
y
x
= x
+ y.
r
r

(15)

Inserting into Eq.(14), we get

B1 =

0 I
1
(y
x + x
y) ,
2 x2 + y 2

(16)

where the subscript denotes that we are refering to the magnetic field of the currentcarrying wire at the origin of our
coordinate system. To create a magnetic quadrupole we place a currentcarrying wire, at distance d along the xaxis,

Quadrupole magnetic field lines

x
Figure 3. The quadrupole magnetic field lines. Plotted with the help of Python and Eqs (16 and 17).

with the current with the opposite direction. Then we place a currentcarrying wire, with the opposite direction, a
distance d along the y axis. Finally, with the current reversed again, we place a wire at a distance d along the y axis,
with respect to the first wire. The other magnetic fields are

0 I
1
(+y
x (x d)
y)
2 (x d)2 + y 2
1
0 I
((y d)
x + (x d)
y)
B3 =
2 (x d)2 + (y d)2
0 I
1
B4 =
(+(y d)
x x
y)
2 x2 + (y d)2

B2 =

(17)

The resulting magnetic field lines are shown in Fig.3.

If we calculate the total magnetic field at the center of the configuration (x = d/2, y = d/2), it will come out zero,
as can be seen from Fig.3 and Eqs (16, 17). We canhoweverTaylor expand our magnetic field at the center, to
calculate the magnetic field gradient. We do so for the B1

:
0
T
2
B1 (r) = 
B1(r
0 ) + J B1 (r0 ) (r r0 ) + O((r r0 ) ),

(18)

where r0 = (x = d/2, y = d/2) and

B1x B1x
x
y

J B1 =
,
B1y B1y
x
y

(19)

the socalled Jacobian matrix.

Some simple differentiations reveal that the offdiagonal terms of the Jacobian are zero, while the diagonal terms
are

B1x
0 I 1
B1y
0 I 1
=
, and
=
.
2
x
d
y
d2

(20)

Since the same is true for all four magnetic fields, we conclude that, for small deviations from the center of the
configuration

Bx =

40 I 1
40 I 1
x
, while By =
y,
2
d
d2

(21)

5
where x
and y measure the distance from the center of the configuration. Indeed, if one looks at the Fig.3, above, it
is obvious that for constanst x
, if we move toward positive y, the By component becomes negative, hence justifying
the minus sign of Eq.(21).
If we add a homogeneous magnetic field B = B0 ez to our configuration, then the magnitude of the magnetic field3
is

B=

r
Bx2 + By2 + Bz2 =

B02 +

1620 I 2 2
(
x + y2 ).
2 d4

(22)

Around the center of our configuration, the second term within the square root of the above equation, is small in
comparison with the first. That allows us to Taylor expand the above equation

s
B = B0

1+


= B0

1620 I 2 2
(
x + y2 )
2 d2 B02


1 1620 I 2 2
2
(
x + y ) .
1+
2 2 d2 B02

Now, since the potential created by the magnetic field is gF F B |B|, comparing with
the trap frequency is given by

s
=

1620 I 2 gF F B |B|
.
m 2 d2 B02

(23)

1
m 2 (x2 + y 2 ), we find that
2

(24)

The Zeeman energy for the hyperfine interaction is given by

EZ = gF F B |B|,

(25)

where

gF ' gJ

f (f + 1) + j(j + 1) i(i + 1)
,
2f (f + 1)

(26)

e
. For 87 Rb, gF = 1/2. Also, the minimum magnetic field intensity is at the center of our configuration,
2m
where |B| = B0 . With everything known, we calculate the Zeeman energy (mF = 2)
and B =

1
2 (5.788 105 eV T1 ) (2 104 T)
2
' 1.1 108 eV.

EZ =

(27)

Of course, in general, the Zeeman energy is spatially dependent

r
EZ (r) = gF F B

B02 +

1620 I 2 2
(
x + y2 ).
2 d4

(28)

The reason why we need an extra homogeneous field is so that we can prevent our magnetic field from having a zero
minimum. Why is that bad? When the atom traverses the zerofield region, sufficiently fast, its magnetic moment
cannot adiabatically follow the rapidly changing magnetic field direction. The change in mutual orientation of the
atoms magnetic moment and the magnetic field, as a result, causes the atoms to escape the trap, phenomenon
known also as Majorana losses.4

3
4

For not too large deviations from the center of our configuration.
The rapid change of the magnetic field, may cause the sign of the B to change, thus allowing the atom to escape the trap, as it is no more
in low energy state.

3 Phase space density

We know, from classical statistical mechanics, that the probability to find a molecule5 at position r with momentum

p equals
d3 rd3 p
e
h30

P (r, p)d3 rd3 p

p2
2m +U (r)

(29)

The velocity distribution is given by integrating the above equation with respect to d3 r. Now the point is that the
integral over all positions is extremely hard to carry out, since the interaction can be extremely complicated. However,
in an ideal gas, we assume that the gas is dilute enough that we can actually neglect the interactions between the
molecules. In that approximation the velocity distribution is independent of the potential shape

P (p)d3 p = Ce

p2
2m

d3 p,

(30)

where the proportionality constant, C , comes from the spatial integration and is equal to V N . It is obvious that the
above approximation collapses in the case of a BEC, since there we have the extreme case of a nondilute gas,
where most of the atoms condensate in the ground state. There the velocity distribution is strongly correlated with
the potential shape.
To calculate the density of states for various physical systems, it is useful
N (), as the number of states
 to define

with energy less than or equal to . For a 1D harmonic oscillator,  = ~

n+

1
. So the number of states between
2

with energy less than or equal to  is

N () =


1
.
~ 2

The density of energy states, by definition equals ()

harmonic oscillator, () =

1
; i.e. independent of energy.
~

(31)

dN ()
, so from the above equation, we get, for the 1D
d

For the 3D harmonic oscillator, the situation is slightly more complicated, nevertheless the physics is the same.
We have

3
x + y + z
 = ~x nx + ~y ny + ~z nz +
, with
=
2
3

(32)

Now the number of states with energy less than or equal to  = x + y + z will be given by an integral of the form

N () =

1
~x ~y ~z

Z
dx

x

Z
dy

3
=
.
6~x y z

x y

dz
0

(33)

The density of states, of the 3D harmonic oscillator potential, is simply given by differentiating the above expression
with respect to energy

() =

2
.
2~x y z

(34)

The calculation for the 3D box potential, runs on the same path. If we suppose that the dimensions of the box
potential are all equal to L, then we have the constraint

k=

In the canonical ensemble.

n
, with n = 1, 2, . . .
L

(35)

7
where k is the wavenumber of the wavefunction. Let us calculate the number of states that have wavenumber6 less
than or equal (in magnitude) to k

N (k) =

V 3
1 43 k 3
k ,
3 =g
8 (L)
6 2

(36)

where the factor 1/8 comes because we take into account only possitive values of k , hence we integrate the first
octant of a sphere in k space, and the factor g is the degeneracy, which we put in by hand. As usual, the density
of states is equal to

(k) =

N (k)
V
= g 2 k2 .
dk
2

(37)

We want the density of energy states and not the density of k states. Since, for a particle in a box, it holds that

=

~2 k 2
,
2m

(38)

and since (k)dk = ()d, we get

V
() = g 2
2

2m
~2

3/2

.

(39)

4 BoseEinstein Condensation
As we explained in the Problem 2; the main issue that needed to be solved, at the time the paper was written, was
the Majorana losses due to the nonadiabatic spin flips of the atoms, as they pass near the center, where the field
changes rapidly direction. Avoiding lower confinement (which was the issue with TOP trap processes), the group
effectively plugged the magnetic field hole by tightly focusing an intense bluedetuned laser, which generated a
repulsive potential for the atoms. Far detuning minimized the heating due to photon scattering.
The main observable of the paper, as always, is demonstrated in the abstract. The group observed a BoseEinstein
condensation of sodium atoms. In order to do so, they employed magnetic, optical forces,7 and evaporative cooling.
How did they understand that they created a BEC? The main experimental signature of the BEC was the bimodal8
velocity distribution below the critical temperature; indicating that the atoms are condensing to the ground state.
There exists; however, a second experimental evidence of the formation of BEC, that is the sudden change of the area
of the cloud, indicating that a phase transition has occured.

6
7
8

Which is directly related with the energy; however, the calculation is easier in the kphase space.
Optical force is the plugging of the region where the magnetic field was zero.
Meaning a superposition of two distributions; a broad one due to the normal gas and a narrow one due to the condensate.