Lecture 17: Optical Bloch equations

Optical Bloch equations, Bloch sphere

Time evolution of the Bloch vector
Optical Bloch equations: Having identied the role of the parameters and in the
eective equations of motion for the atomic ip operators, Eqs. (16.4) and (16.5), we will
attempt to solve these equations. First note that the evolution of the atomic ip operators
free (rA )
is determined by the time dependence of the free electric eld. Suppose now that E
represents a monochromatic eld with frequency that has been prepared in a coherent
state | . Recall from the denition of the coherent states that, when taking the expectation
(+) (rA )| d/ iei t where
value of the eld in this state, we have to replace 2 |E
free

2
A (rA ) d

(17.1)

is the Rabi frequency. Although it has the dimension of a frequency, it is in general a complex
quantity and measures the strength of the atom-light interaction.
We now absorb the fast (free) time dependence ei t by rewriting Eqs. (16.4) and (16.5)
we then obtain (we
in a frame co-rotating with frequency . By re-dening
= ei t
omit thesign again)

= i
+
2
2

z )

z = (1 +

(17.2)
(17.3)

where we dened the detuning = A + .

The eective equations of motion for the atomic transition operators are not very helpful
if one tries to understand the atomic evolution. However, they can be converted into cnumber dierential equations by taking the trace over the atomic density matrix. Recall
is given by O
= Tr(
Here the operator
that the expectation value of an operator O
O).

is in fact a time derivative of one of the

O
i . Hence, an expectation value of the type

i ] is being computed in the Heisenberg picture in which the operators carry the

i = Tr[

time dependence. In contrast, in the Schrodinger picture, it is the density operator, i.e. the
quantum state, that varies in time. However, expectation values can be computed in both
i ].
pictures and they give the same answer. We thus seek to compute
i = Tr[
66

The density operator of a two-level atom is, written in the {|g |e}-basis, a Hermitian
22-matrix. The space of Hermitian 22-matrices is spanned by the identity matrix in that
space and the three Pauli matrices (these matrices form a basis into which all matrices can
be expanded). This means that the density matrix of any two-level system can be written
in the form
=

I +u
2

(17.4)

is the vector of Pauli matrices.

where u = (u v w)T is a real vector with norm |u| 1 and
j ) = 2ij , the elements of the vector
Due to the trace properties of the Pauli matrices, Tr(
i
u are given by

i ) .
ui = Tr(

(17.5)

In particular, with our denitions of the Pauli matrices, u = Tr[

(
+
)], v = iTr[
(

)]
and w = Tr(

z ). The expression (17.4) converts the (Hermitian) density matrix into a real
vector inside a three-dimensional ball.
Obviously, u and v are just the real and imaginary parts, respectively, of the o-diagonal
atomic matrix elements. Then we can rewrite Eqs. (17.2) and (17.3) as

v =
.
0 I v
v
2

w
R I 0
w
2(1 + w)

(17.6)

This set of equations is called optical Bloch equations and it describes the motion of
the Bloch vector u = (u v w)T on the Bloch sphere. In order to gain some insight into
Eq. (17.6), we have a look at what particular values of the Bloch vector mean physically. By
its denition, the vector uT = (0 0 1), i.e. the south pole of the Bloch sphere, represents

the ground state |g of the atom, and the vector uT = (0 0 1) pointing towards the north
pole represents the excited state |e. A vector in the equatorial plane with w = 0 must
therefore be a superposition of |e and |g whose phase is determined by u and v. It is easy

+
|A = cos
to verify that an atomic state |A = (|e + ei |g)/ 2 yields u = A |
)|A = sin .
and v = A |i(

Time evolution of the Bloch vector: The time evolution of the Bloch vector is determined by the Rabi frequency and the detuning from the resonance. For zero detuning
and negligible spontaneous decay, the external driving eld simply causes the atom to op
67

between its ground and excited states with the Rabi frequency. Starting in the ground state
|g, i.e. with the initial conditions u(0) = 0 v(0) = 0 w(0) = 1, the solutions to the Bloch
equations read
u(t) = sin t

v(t) = 0

w(t) = cos t

( = = 0) .

(17.7)

This process is called Rabi opping and is depicted in Figure 22 (red curve).
We see that we can bring the atom into a coherent superposition state |e + ei |g if
the interaction time is chosen to be exactly t = /(2). Because the azimuthal angle of the
Bloch vector changes by /2 during this evolution, this process is called a /2-pulse. If the
interaction time is chosen to be t = /, one transfers the population completely into the
excited state |e (-pulse).

FIG. 22: Time evolution of the Bloch vector for dierent detunings and = 0 (left gure). The
values starting with the red curve are: = 0, 0.5, 1, 2, 3. The Rabi frequency was set to = 1.
Time evolution of the Bloch vector with = 0.2. The other parameters are = 0.5 and = 1.

For non-zero detuning, the Bloch vector precesses on a cone starting from the south pole
of the Bloch sphere with an opening angle that is determined by (see Figure 22),

sin 2 + 2 t
2 + 2

v(t) = 2
(1

cos
2 + 2 t)
+ 2

2 (1 cos 2 + 2 t)
w(t) = 1 +

2 + 2
u(t) =

68

(17.8)

with a precession frequency

2 + 2 > . We can see why if we look at the rst term on

the right of Eq. (17.6) which we can write as a vector product

u = u = R .

(17.9)

Dierential equations such as Eq. (17.9) appear already in classical physics as the equation
of motion of the gyroscope or spinning top. The Bloch vector u precesses, starting from its
initial state, around the vector . For zero detuning and real Rabi frequency, the vector
points in the v-direction and the Bloch vector describes a great circle in the (u w)-plane
(red curve in left panel of Fig. 22).
Note that in the above examples, because has been set to zero, the Bloch vector is of
unit length, |u| = 1 which, by Eq. (17.4), means that the atom is in a pure state for all
times. Obviously, because spontaneous decay is always present, the atomic density matrix
will in general be mixed and the Bloch vector will have a length |u| < 1. The right panel
in Fig. 22 shows a typical evolution of the Bloch vector starting from the ground state |g
until the steady-state value is reached.
Stationary solutions of the Bloch equations: Finally, we will have a look at the stationary solutions of Eq. (17.6). They are characterized by the condition u = 0. For simplicity, let us assume that the Rabi frequency is real. Then we nd that
ustat = 2
2

+ 2 + ( 2 )2

vstat = 2
2

2
2 +

( 2 )2

wstat = 1 +

2
2

2
2
2
+

.
( 2 )2
(17.10)

In the limit when the driving is weak compared to the spontaneous decay rate, ,
we obtain from Eq. (17.10) that wstat 1. This means that the atom resides closely to
its ground state |g. In the opposite limit, when the driving dominates the restoring force,
, and the frequency of the incoming radiation eld is on resonance with the atomic
transition, = 0, then wstat 0. In this case there is a roughly equal probability for nding
the atom in its ground state |g or in its excited state |e. However, for any values of the
parameters, wstat is strictly negative, wstat < 0, so that the probability to nd the atom in
its ground state is always larger than the excited-state probability. In other words, there is
no steady-state solution for a two-level atom in which the atom is excited.
69

Lecture 18: Ramsey interferometer

/2-pulses revisited
Stationary solutions to the Bloch equations
/2-pulses revisited: Let us return to the Bloch equations (17.6) and recall their solutions
in the absence of spontaneous decay, = 0, and vanishing detuning, = 0. Starting at the
south pole of the Bloch sphere, uT (0) = (0 0 1), the solution is simply [see Eq. (17.7)]
u(t) = sin t
w(t) = cos t
which corresponds to motion in the (u w)-plane. After a time t = /(2), the Bloch vector
performs a rotation around the v-axis by an angle /2, uT (/2) = (1 0 0). The result is
an equal superposition of ground and excited states,
1
|(/2) = (|e |g)
2

(18.1)

which lies on the equator of the Bloch sphere. Clearly, another /2-pulse will bring the
atom into its excited state |e.
All the states on the equator of the Bloch sphere are equal superpositions of |e and |g,
albeit with dierent relative phases between them. Consider now a situation in which an
external perturbation induces a motion along the equator. In our gyroscopic picture this
corresponds to a rotation around the w-axis which can only be generated by a perturbation
Hamiltonian of the form
pert = g
z
H

(18.2)

Upert (t) = eigtz = cos(gt) i sin(gt)

z .

(18.3)

that generates a time evolution

Applied to the state (18.1) gives

1
1
Upert (t)|(/2) = eigt |e eigt |g = [cos(gt) (|e |g) i sin(gt) (|e + |g)] .
2
2
(18.4)
This rotates the Bloch vector by an azimuthal angle = gt around the equator to the
position uT = ( cos(gt) sin(gt) 0).
70

If we now apply a second /2-pulse, we do not necessarily end up in the excited state
anymore. Instead, the nal state after the pulse will be
| = cos(gt)|e i sin(gt)|g .

(18.5)

A subsequent projective measurement in the {|g |e}-basis will show that the atom is in its

ground state with probability pg = sin2 (gt) and in the excited state with probability pe =

cos2 (gt). Such a measurement can be performed by irradiating the atom with radiation that
is capable of ionizing only the excited state and, in a second step, with radiation that ionizes
also a ground-state atom. This is the basic principle behind the Ramsey interferometer (see
Fig. 23) which nds a wide range of applications in quantum optics and atomic physics.

e
1
|e
cos gt|e
|g2 |e
2

z
igt
e
i sin gt|g
1 |g

2 |g
2
igt

/2

/2

FIG. 23: Schematic Ramsey interferometer. The separate paths are internal states, not external
degrees of freedom. See description in the text.

Let us comment on the sensitivity of the Ramsey interferometer. As before with the
MachZehnder interferometer, we will estimate the phase error that one makes in such a
measurement. The signal in the Ramsey interferometer is the expectation value of
z ,

z = cos2 (gt) sin2 (gt) = cos(2gt) .

(18.6)

The noise, i.e. the uncertainty in a single measurement, is then

(
z )2 =

1 cos2 (2gt) = | sin(2gt)| .

The phase sensitivity for a single measurement, Eq. (11.7), is thus

(
z )2
= 1.
=

(2gt)

(18.7)

(18.8)

If we perform N independent measurements, i.e. we let a stream of N independent atoms

pass through the interferometer, the signal strength increases by a factor N. The noise,
71

however, increases only by a factor

(i)

N which we can show as follows. Let

z be the

inversion operator associated with the ith atom. Then the total variance is

i=1

z(i)

i=1

z(i)

z(i)

z(i)

i=1

i=1

(i) (j)

(i)

i=j

z(i)
z(j)

i=1

z(i)

(i) 2 (i) (j)

z
z

z
i=1

= N |sin(2gt)| .

i=j

(18.9)

(j)

Here we have used that

z
z =
z
z because we assumed that the measurements
are all independent. Combining these results we nd that the phase noise of N independent
measurements is
1
(18.10)
=
N
which shows the same scaling as the MachZehnder interferometer. The dierence in the
interpretation is that in Ramsey interferometry N denotes the number of atoms, whereas in
MachZehnder interferometry n
is the average number of photons in the coherent state.
The MachZehnder interferometer for photons (see lecture 11) uses matter in the form of
beam splitters to create two separate paths for the photons. In the Ramsey interferometer
it is light in the form of /2-pulses that forces the atom into dierent paths. This latter
type of interferometer is peculiar in that there is no separate spatial pathways that the atom
takes. The objects that pick up the phases or phase dierences are the two internal states
|g and |e. However, we will see in a later lecture that the absorption and emission of a
photon is connected with a change of momentum which could then be used to indeed force
the atom into separate pathways.

72

Lecture 19: Resonance uorescence

Intensity of the scattered light
Intensity correlations, photon anti-bunching
squeezing in resonance uorescence
Intensity of the scattered light: If a two-level atom is driven by an external radiation
eld, we have seen that the atom periodically absorbs and emits photons. The emitted (or
scattered) light from the driven atom can be collected at a photodetector. The measurement of the intensity of the scattered light reveals details of the nature of the atom-light
interaction.
Let us recall the equation of motion (15.15) for the photonic amplitude operators,

+ A (rA ) d
a
= i a

and let us again integrate it with respect to time as in Eq. (16.1),

i t
dt (t t )ei (tt )
(t ) .
a + A (rA ) d
a
(t) = e

Inserting this solution into the expression for the electric-eld strength, we obtain for the
positive frequency components
(+) (r t) + i
(+) (r t) = E
E
free

[A (r) A (rA )] d

(t ) . (19.1)
dt (t t )ei (tt )

We can rewrite this expression as

(+) (r t) +
(+) (r t) = E
E
free

dt (t t )G(r t; rA t ) d
(t )

(19.2)

where the propagator function

G(r t; rA t ) = i

[A (r) A (rA )] ei (tt )

(19.3)

propagates an excitation from the space-time point (rA t ) to (r t).

At the moment, we are not interested in the details of this propagator function. However,
in view of the Markov approximation discussed earlier, we introduce the slowly-varying

73

 (t ) and take them at the retarded time t |r rA |/c

atomic operators
(t ) = eiA t

outside the time integral,

(+) (r t) + (t |r rA |/c)
(+) (r t) = E
E
free

dt (t t )eiA (trrA /ct ) G(r t; rA t ) d .

(19.4)

If we now let t under the integral, the resulting function g(r rA) is an eective
propagator and the electric-eld strength can nally be written as
(+) (r t) + g(r rA)
(+) (r t) = E
(t |r rA |/c) .
E
free

(19.5)

Since the photocurrent at a photodetector is proportional to the light intensity, we have to

(+) (r t) with the electric-eld strength given by Eq. (19.5).
() (r t) E
compute I(r t) = E
Because we are interested only in the contribution from the scattered light, we obtain
I(r t) = |g(r rA )|2
(t |r rA |/c)
(t |r rA |/c) .

(19.6)

But the expectation value of the product of atomic operators is nothing but the excited-state
probability (1 + w)/2 taken at the retarded time. Hence, the observed intensity reects the
dynamics of the optical Bloch equations.
In the limit of zero detuning, the excited-state probability can be computed to be

2
( 2 )2 2 t
3
sinh

. (19.7)
1 e3t/2 cosh ( )2 2 t +

ee (t) =
2
22 + 42
2 ( )2 2
2

In the limit of a weak driving eld, 2 2 , we obtain for the intensity

Iweak (r t + |r rA |/c) = |g(r rA)|2

2
t/2 2
1

e
2

(19.8)

which approaches the steady-state solution |g(r rA )|2 2 /2 that can be read o from

Eq. (17.10). In the opposite limit of strong driving, 2 2 , we obtain

1
Istrong (r t + |r rA |/c) = |g(r rA)|2 1 e3t/4 cos t
2

(19.9)

which oscillates with the Rabi frequency . These oscillations are damped on a time scale
given by the spontaneous decay rate (see Fig. 24).
Intensity correlations, photon anti-bunching: In order to describe the statistics of the
scattered light, we need to look at the intensity correlation function. As we saw, the function
()
() (r t + )E (+) (r t + )E (+) (r t)
G(2) (r t + r t) = Ei (r t)E
j
j
i

74

(19.10)

0.25
0.6
0.2
/=5

0.5

/=2

ee

0.4
(t)

Squeezing<:(E) :>/|g|

0.15

0.3

0.2

0.05
0
0.05

/=0.6
0.1

0.1

0.1

0.5

1.5

2.5
t

3.5

4.5

0.15

0.1

0.2

0.3

0.4

0.5
/

0.6

0.7

0.8

0.9

FIG. 24: Time evolution of the excited-state probability ee (t) for dierent ratios / (left gure).
Squeezing in resonance uorescence (right gure).

can be used to infer whether a given quantum state of light shows nonclassical behaviour
such as photon anti-bunching. With the expression (19.5) for the electric-eld strength, we
nd that [r = |r rA |]
(t)
(t + )
(t + )
(t)
G(2) (r t + r/c + r t + r/c) = |g(r rA)|4
(t)
ee (t + )
(t) .
= |g(r rA)|4

(19.11)

In the coincidence limit, = 0, we nd that G(2) (r t + r/c r t + r/c) = 0. This means that
the joint probability of registering two photons at the same time in the detectors vanishes
and indicates perfect photon anti-bunching. The steady-state intensity correlation function
G(2) ( ) = limt G(2) (r t + r/c + r t + r/c) therefore must have a positive initial slope,
G(2) ( ) > G(2) (0) = 0 .

(19.12)

To understand the origin of photon anti-bunching in resonance uorescence, we can invoke

the following intuitive picture. A driven two-level atom can only emit a photon if it had
been previously excited by the driving eld. Once it emits a photon, the atom jumps back
to its ground state. Hence, in other to emit a second photon, it must be re-excited by the
driving eld which takes a certain amount of time. Therefore, the atom cannot emit two
photons at the same time, and the scattered light shows perfect photon anti-bunching.
Squeezing in resonance uorescence: We have already seen that the scattered light
from a resonantly driven two-level atom shows nonclassical behaviour in terms of photon
75

anti-bunching. But there is even more to it. The resonance uorescence light can even
be squeezed. To see that, recall that a measure for the occurence of squeezing is that the
variance of the electric-eld strength drops below the vacuum level for a particular choice of
phase or, equivalently, if the normally-ordered variance of the electric-eld strength becomes
t)]2 : < 0.
negative, : [E(r
Using the source-quantity representation (19.5), we can write this variance as
t + r/c)]2 : = 2|g(r rA)|2 ee (t)
: [E(r

2|g(r rA)|2 |(t)|2 + g2 (r rA ) 2 (t) + g2 (r rA ) 2 (t) .

(19.13)

Now we introduce again the slowly-varying amplitudes (t) = eiA t

(t) and write the
(t)| eiA t+i(t) where (t) denotes the sum of the phase
product g(r rA )(t) as |g(r rA)
of
(t) and the propagator function. Note that for vanishing detuning we have A = .
Hence it follows that the variance is

t + r/c)]2 : = 2|g(r rA)|2 ee (t) |

: [E(r
(t)|2 [1 + cos 2( t (t))] .

(19.14)

Recall from Eqs. (17.10) that the stationary values for large times t are simply
ee (t) =

2
1 + w(t) t

2
22 + 2

|
(t)|2 =

u2 (t) + v 2 (t) t
2 2

4
(22 + 2 )2

(19.15)

so that nally
t + r/c)]2 : t

: [E(r

22
2
2
2
2

cos
2(
t

())
.
|g(r
r
)|

A
(22 + 2 )2

(19.16)

Since the trigonometric function in Eq. (19.16) is bounded by one, we can infer that the
uorescence light will be squeezed (for particular values of the phase) if
t + r/c)]2 : < 0 22 < 2 .
: [E(r

(19.17)

This means that the driving eld should not be too strong if one is to observe squeezing.
The reason is that squeezing is a coherent eect which can only be built up if the atomic
coherences (t) play a signicant role. However, in the strong-eld limit the steady-state
value of () vanishes and the resonance uorescence radiation is purely incoherent and

squeezing is not observed. The maximally obtainable squeezing value is, for = / 6,
t + r/c)]2 : = |g(r rA )|2 /8 (see right panel in Fig. 24).
: [E(r
76