Você está na página 1de 9

PHYSICAL REVIEW A VOLUME 45, NUMBER 7

Dynamic Stark effect for the Jaynes-Cummings system


1 APRIL 1992
P. Alsing, D.-S. Guo,
*
and H. J. Carmichael
Department ofPhysics, Chemicai Physics Institute, University of Oregon, Eugene, Oregon 97403
and Institute of TheoreticaL Science, University of Oregon, Eugene, Oregon 97403
(Received 26 August 1991)
We calculate the quasienergies and steady states of a coupled two-level atom and quantized elec-
tromagnetic cavity mode with the cavity mode driven by a periodic classical field. The atom, the cavity
mode, and the classical field are all on resonance. The quasienergies give shifted Jaynes-Cummings level
splittings. These splittings are reduced by the interaction with the driving field and vanish at a threshold
value of the driving field strength. Above the threshold, discrete quasienergies and normalizable steady
states do not exist. Below the threshold, for weak driving fields, the steady states are squeezed and dis-
placed Jaynes-Cummings eigenstates. We discuss the relevance of these results to work in cavity quan-
tum electrodynamics.
PACS number(s): 42.50.Dv, 32.60.+i, 36.90.+f
I. INTRODUCTION
The Jaynes-Cummings Hamiltonian describes the in-
teraction of one mode of the electromagnetic field with a
two-state atom in the electric dipole and rotating-wave
approximations. This Hamiltonian is of fundamental im-
portance to the field of quantum optics; it is a central in-
gredient in the quantized description of any optical sys-
tem involving the interaction between light and atoms.
The problem of a two-state atom interacting with an
electric field is mathematically equivalent to the problem
of a spin- , particle in a magnetic field. The early history
of the Jaynes-Cummings model is therefore found in
work on magnetic resonance [1,2]. Jaynes and Cum-
mings presented their analysis in 1963 [3].
Since that
time, a large number of theoretical papers have appeared
dealing with various aspects of the Jaynes-Cummings
Hamiltonian and the evolution in time that it generates
[4,5]. For example, the Jaynes-Cummings Hamiltonian is
used widely in quantized theories of the laser [6,7]; this
alone accounts for a vast literature on the subject. How-
ever, in spite of the extensive theoretical attention it has
received, the full, quantized Jaynes-Cummings Hamil-
tonian has had little relevance to experiments in optics
until recently. This is because most interactions between
light and atoms involve highly populated modes of the
electromagnetic field where a semiclassical treatment is
all that is required. In the semiclassical limit the electric
field enters the Hamiltonian as a c number. The evolu-
tion of the atomic state is represented by a precession on
the Bloch sphere at a frequency determined
by the
strength of the electric field. The precession
frequency

the Rabi frequency

is produced by a split-
ting of the atom's
energy levels due to its interaction with
the time-periodic field

dynamic Stark splitting [8].


To obtain conditions where a semiclassical treatment is
inadequate, it is necessary either that the fundamental
coupling strength (the dipole coupling strength) be large
or that many weakly coupled modes contribute to the
physics in an important way. In both situations, single-
or few-photon excitations, requiring a quantized treat-
ment, can produce observable effects. Spontaneous emis-
sion is an example of a phenomenon involving many
weakly coupled modes where a quantized treatment is
called for [9]. Photon antibunching in resonance fluores-
cence is another [10,11]. For interactions with one mode
of the electromagnetic field, the size of the dipole cou-
pling strength is important; rates of coherent evolution
(Rabi frequencies) must be comparable to dissipation
rates for the single-mode interaction to be significant.
Traditionally, coherent effects are observed in the limit of
very large photon numbers and small dipole coupling
constant. In contrast, work in cavity quantum electro-
dynamics attempts to make the dipole coupling constant
itself large [12]. When this is achieved, dynamical effects
associated with the energy spectrum of the full quantized
Jaynes-Cummings Hamiltonian are observable [12

16].
Usually, the Jaynes-Cummings Hamiltonian will not,
on its own, provide a complete description of an experi-
ment, no matter how closely it approaches the two-state,
one-mode idealization of the Jaynes-Cummings model.
The Jaynes-Cummings Hamiltonian defines a
"mole-
cule,
"
a composite system formed from the coupling of a
two-state system and a quantized harmonic oscillator.
To interrogate the "molecule,
"
we must probe it in some
manner. The probe is a perturbation; we must analyze
the problem of "molecule"
plus probe to fully understand
the experiment. In this paper, we analyze the problem of
the Jaynes-Cummings "molecule"
probed by a classical
(external) coherent field. Recent experiments designed to
observe the so-called "vacuum"
Rabi splitting are of this
type [17,18]. Although the splitting for a single atom
coupled to an electromagnetic cavity mode has not yet
been observed, it appears that this observation is not very
far off. (Note that vacuum Rabi splitting refers to the
splitting of the first excited state of the Jaynes-Cummings
Hamiltonian. )
The complete Hamiltonian we consider consists of the
Jaynes-Cummings
Hamiltonian for a two-state atom in-
teracting on resonance with one mode of an electromag-
45 5135 1992 The American Physical Society
5136 P. ALSING, D.-S. GUO, AND H. J. CARMICHAEL 45
netic cavity

the free Hamiltonian


Ho

plus an interac-
tion Hamiltonian Ht(t) describing the coupling between
this system and the external field. The interaction Ham-
iltonian has an explicit periodic time dependence due to
the oscillation of the external field at the common reso-
nance frequency of the atom and cavity mode. We are
therefore interested in steady states (in the sense of
periodic states) and quasienergies rather than in energy
eigenstates and eigenvalues [19,20]. The external field
can either couple to the atom, by illumination through
the open sides of the cavity, or to the cavity mode by
il-
lumination through one of the mirrors. The most in-
teresting results are found in the latter case. For this case
we find that the resonance frequencies of the Jaynes-
Cummings
"molecule"
undergo a dynamic Stark shift; in
place of the Rabi splittings +&n
g,
we obtain quasiener-
gies
+v n
g[1

(2@/g)
],
n =0,
1, 2, . . .
where 8 is the amplitude of the external field and
g
is the
dipole coupling constant [21]. The associated steady
states are generally quite complicated. For weak excita-
tion they are squeezed and displaced Jaynes-Cummings
eigenstates. For an external field amplitude larger than
6 =g/2, no normalizable steady states exist. (Note that
8=g /2 is the threshold condition for spontaneous
dressed-state polarization [16]. )
We have used the word "molecule"
by analogy and it
becomes awkward with repetition. We therefore drop it
in the rest of the paper. We will refer to the Jaynes-
Cummings system, and the driven Jaynes-Cummings sys-
tem, when the external driving field is added.
To our knowledge, the method we use for finding the
quasienergies and steady states of the driven Jaynes-
Cummings system has never been applied before. We
therefore introduce the method in stages. In Sec. II we
derive the familiar eigenstates

the so-called dressed
states
[8]

and eigenenergies of the Jaynes-Cummings


Hamiltonian, in the absence of a driving field. In Sec. III
we treat the driven Jaynes-Cummings system with the
external field driving the atom. Here the quasienergies
are not shifted and the steady states are displaced dressed
states. We calculate the quasienergies and steady states
for the driven Jaynes-Cummings system with the external
field driving the cavity mode in Sec. IV. In Sec. V we
discuss the relevance of our results to work in cavity
quantum electrodynamics. A concluding summary is
provided in Sec. VI.
II. EIGENSTATES AND EIGENENERGIES
OF THE JAYNES-CUMMINGS SYSTEM
In this section we solve the eigenvalue problem for the
standard Jaynes-Cummings system. We formulate the
problem in the language used to treat the driven Jaynes-
Cummings system in Secs. III and IV. We therefore be-
gin with the definition of the driven Jaynes-Cummings
Hamiltonian:
dard Jaynes-Cummings system,
Ho= H

+Hg =%coo(a a+
,
'cr,
)+ifig(a cr

ao
~),
0
(2)
o
+
1Q)pt
a
e
0
ENpt
e
a
where we allow the driving field to couple either to the
atom (upper row inside the large parentheses) or to the
cavity mode (lower row inside the large parentheses). a
and a are creation and annihilation operators for the cav-
ity mode, satisfying the commutation relation
[a,
at]=1;
(4)
o.
+,
o.
,
and
o,
are atomic pseudospin operators, satisfy-
ing the commutation relations
[cr+, o
]
=2o
[oo+]
=+o+;
coo is the frequency of the driving field, in resonance with
the atom and the cavity mode; 8 is the amplitude of the
driving field; and
g
is the dipole coupling constant. Arbi-
trary phases for the driving field and the dipole coupling
constant may be absorbed into the definition of the
operators; thus, there is no loss of generality in using real
quantities 8 and g.
We will denote solutions to the Schrodinger equation
obtained from Hamiltonian (1) by ~
g'(t) ):
[H, +H, (t)]~
1(')
=
. [H. +H, +H, (t)]iq'),
where the superscript t on
g
refers to the Schrodinger
picture, the picture in which the Hamiltonian has an ex-
plicit time dependence. The Hamiltonian 0 generates
p
rotations that remove this time dependence. We there-
fore also define states
~g)
in the "interaction picture,
"
with
where the Schrodinger equation in the "interaction pic-
ture" is
=
.
(H,
+H, )~q),
dt i%
(8)
with
o+
0
H
+H@ =iong(a
cr ao+)+ifi6'
a

a
and Ht(t) is the Hamiltonian for the periodic interaction
with the driving field,
H,
(t):H~(t)
H =Ho+Ht(t),
where Ho is the (on-resonance) Hamiltonian for the stan-
We use quotes to remind us that the "interaction picture"
is defined here
by separating H and H +H@(t), rather
p
45 DYNAMIC STARK EFFECT FOR THE JAYNES-CUMMINGS. . . 5137
than
Ho
and Hl(t). Our objective is to solve the eigenval-
ue problem
EItP+)+ifigalg )
=0,
Elyz & ivigatlitz+ &
=o
.
(15a)
(15b)
lfz&=e
and solutions to the Schrodinger equation (1) in the form
lyt (r) ) e
i(zlzz
)tlat ) (12)
The states
Ilitz ) are periodic in time with period 2m/coo;
we follow Sambe
I
19] in referring to these states as steady
states. The quantities E are the quasienergies. Each
quasienergy defines a frequency shift E/A that is added
(in the Schrodinger picture) to every harmonic meio,
m =0,
1,2, ...
,
present in the periodic oscillation of the
steady state If'z).
From the definition of
Hg+H@
in Eq. (9), we write the
eigenvalue problem (10) in the form
0
i AC
a

a
+E
lyz
&
(Hg +Hz ) I yz
&
=E
lgz
&
(10)
The eigenstates
lgz
) in the "interaction picture" define
the Schrodinger picture states

i (H /A)t
(11}
atalqz
&=
E/fi
I~
)
This is the decoupled eigenvalue problem, and in this
case there is no need for further transformations to con-
vert it into an harmonic-oscillator problem. The quasien-
ergies are determined by the requirement
E/fi
=n, n =0,
1,2, . . .
which gives
ED
=0
and
(18a)
E=+&n
g
E
i
=
A&n

g
(18b)
(n =1,
2, . . .). The states
If@
) are proportional to the
Fock states
Using Eq. (15a) to substitute for
It/iz ) in Eq. (15b), we ob-
tain
'2
+
isa
+
0
+E
lli;)
I@o
&=c, lo&
~/i
)
=
I11(&=c(
n
&,
n =1,
2, . . .
(19)
iR
+

isa
t+
0
I1lz+ ) =(}, (13b)
where we have expanded the eigenstate as
I 0
&
=
I
0'
& I
+ &+
I g
& I

&; (14)
I
+ ) and
I

) are the upper and lower states of the two-


state atom, and
lgz
) and
lgz
) are field states normal-
ized so that
(/zlzz
)
=1. We will use the following gen-
eral approach to solve Eqs. (13). We first decouple the
equations to obtain an eigenvalue problem for
Igz
) or
Ifz
) alone in which an effective Hamiltonian appears
that is quadratic in the operators a and a. We then use
displacement and squeezing transformations (when neces-
sary) to convert the effective Hamiltonian into the Hamil-
tonian for an harmonic oscillator. From solutions to the
harmonic-oscillator problem, the solutions to Eqs. (13)
are---constructed
by inverting the transformations that
have been used.
Carrying out this program gets progressively more
diScult as we move from the standard Jaynes-Cummings
system to the driven Jaynes-Cummings system with the
external field driving the atom, and then to the driven
Jaynes-Cummings system with the external field driving
the cavity mode. We begin with the standard Jaynes-
Cuminings system where the conversion of Eqs. (13) into
an harmonic-oscillator problem is almost trivial.
For the standard Jaynes-Cummings system 8=0, and
Eqs. (13}reduce to
and corresponding to the quasienergies (18b},we obtain
Ig)
=(1W'2)(
n

1&I+ &+i In & I

&),
I q,
&
=(1W'2)(
In

1)I+ ) i ln ) I

)
)
(21b)
(n =1,
2, ...) where we have chosen the arbitrary phase for
these states by taking
c&
to be pure imaginary.
Using Eq. (11), the eigenstates (21) in the interaction
picture give steady states
(22a)
and

i [n

(1/2) ]capt
If'&)
=e
lg~),
n =1,
2, . . .
, g=u, l .
(22b)
Here the time dependence is contained in an overall
phase factor, and in the full solution to the Schrodinger
where g=u or l, and the constants
co
and
c&
will be
determined
by
the normalization. The states
lgz+)
are
obtained from Eqs. (15), (18), and (19) in the form
Iq+g&=
etc&In

1), n
=
1,2, . . .
where e&=+1,

1 for g=u, l. From Eqs. (14), (19), and


(20), we can now construct the normalized eigenstates
I fz
) . Corresponding to the quasienergy (18a), we obtain
(21a)
5138 P. ALSING, D.-S.
GUO, AND H. J. CARMICHAEL 45
equation [Eq. (12)] the phases (n

,
'
)coot and (E/A)t add
to give (E,
/A)t
=
,
'
coot, n =0, and
(E,/A)t
=
[(n
,
'
ko0+ep'n
g]t,
n =1,
2, ,
g=u, l (e~=+1) .
This rejects the fact that Hand H commute; there-
0
fore, the states (21) are simultaneous eigenstates of both
Hamiltonians. Of course, the "quasienergy" and
"steady-state'*
language is not really necessary for treat-
ing this case. But it is needed to handle the driven
Jaynes-Cummings system. When the driving field is not
zero, H does not commute with H +H@ and we do not
0
find simultaneous eigenstates; the time dependence of the
states
I 1ltz ) will then be nontrivial.
IP&=(1/&2)(I@/g;n 1)I+ &+il@/g;n &I

&),
(29b)
I y,
& =(1/&2)(
I @/g; n

1 & I+ &

il @/g; n & I

& ),
(n=1,2,...) where
Ib/g;n

1) and Ih/g;n ) are dis-
placed Fock states:
la;n ) =D(a)ln & . (30)
Ip')= 'I
'"'@/
o&l

& (31a)
and
From Eqs. (11) and (29), the steady states of the driven
Jaynes-Cummings system with the external field driving
the atom are
III. THE DRIVEN JAYNES-CUMMINGS
SYSTEM: COUPLING TO THE ATOM
We first consider the driven Jaynes-Cummings system
with the external field driving the atom. This problem is
solved
by
a minor extension of what we have just seen.
The eigenvalue problem in the "interaction picture"
takes the form [from the upper row in Eq. (13)]
Elq;)+tA(ga

@)lq, )
=O,
EI
q
& t A(ga

4')
I
y+
&

=0
.
(23a)
(23b)
(a

@/g)(a

@/g)
If'
)
=
E/A
I ) (24)
The only change, in comparison with Eq. (16), is that
here we have the eigenvalue problem for a displaced har-
monic oscillator. We can remove the displacement by
multiplying on the left
by
D ( 8/g), where
D (a )
=
exp(aa

a'a
), (25)
We solve Eq. (23a) for
IlltE ) and substitute the solution
into Eq. (23b) to obtain
2
X(le
'
'
@/g;n

1)I+ )
+ie(Ie
'6'/g;n
&I

&),
n =1,
2, . . .
, g=u, l (e&=+1) . (31b)
These states illustrate the features of the steady states for
a time-periodic Hamiltonian in a nontrivial way. They
carry the same time-dependent phases as the states (22);
but they also carry the periodic time dependence that
enters through the displaced Fock states. The full solu-
tions
I/0(t)) and IP'&(t)) to the Schrodinger equation
[Eq. (12)] do not involve single frequencies

, 'coo and
(n
,
')coo+e&&n
g;
they involve infinite sets of frequen-
cies: (m

, ')coo, m=0,
1,2, ...
,
for the state
Igo(t))
and
(m 2)t00+
e
&&ng,
m=0, 1,2, ...
,
for the states IP' t(t)),
n=l,
2,...
,
g=u, l (e&=+1). Thus, the quasienergies
ED=0 and
E&=e&A&ng characterize the steady states
by defining frequency shifts that are applied to the whole
series of harmonically spaced frequencies (m

,
'
)~o,
m =0, 1,2, . . . . Each steady state is also distributed in a
characteristic way across the Fock states. This distribu-
tion determines the relative strengths of the different fre-
quency components in the state.
and using
D (a)aD(a)=a+a .
Then,
a
alp~
&
E/A
l~
)
g
where
I g,
&
=D'(@/g)
I y;
& .
(26)
(27)
(28)
IV. THE DRIVEN JAYNES-CUMMINGS
SYSTEM: COUPLING TO THE CAVITY MODE
We now turn to our main interest

calculation of the
quasienergies and steady states of the driven Jaynes-
Cummings system with the external field driving the cavi-
ty mode. This calculation follows the same general steps
as the calculations in Secs. II and III. However, the de-
tails are considerably more complicated. In the "interac-
tion picture" we have the eigenvalue problem [from the
lower row in Eqs. (13)]
ly, &
=
I@/g;0& I

& (29a)
and
Equation (27) is solved by Fock states as before. The
quasienergies are given by Eqs. (18), and, after inverting
the displacement, the eigenstates in the "interaction pic-
ture" are
[
iA6(at

a)+E]I
QE
) +i
Agalg~ )
=0,
[

iAC(at

a
)+E)IPE ) i Aga I fE
)
=0

.
(32a)
Our first task is to decouple these equations and obtain an
equation for
IQE
) (or
IQE
) ) alone. To this end, we mul-
tiply Eqs. (32a) and (32b) on the left by
at
and a, respec-
tively, which gi~es
45 DYNAMIC STARK EFFECT FOR THE JAYNES-CUMMINGS. . . 5139
[
i
fi@(a
a
)+E]a
lPz )+isa alga )

iAlgz )
=0,
(33a)
[

iM(a

a)+E]alga )
i
Rgaa
lPz
&
i
kelgz &=0
.
0 (E)0 (E)lyz &
=0, (40)
where O~(E) and 0 (E) commute. The general solution
will take the form
(33b)
Ip;)=c, lp;
,
)+c.
lg; .
.
), (41)
Then, from Eqs. (32) we have
alpz )
=
(6/g)(a

a)+i
E/R
l&z
(34a)
(42a}
(42b)
where
lPz.
~
) and
l gz. ) are solutions to the equations
o, (E)ly;.
,
) =0,
o (E)lit;.
&=0.
a
lpz
)
=
(@/g)(a

a)+i
+
E/A
(34b)
(v jg)(at a)+i
Ejfi
g
+ata
Using Eq. (34b) to substitute for a tl
Pz
) in Eq. (33a), and
Eq. (34a) to substitute for al
fz
) in Eq. (33b), we obtain
It may happen that there are quasienergies E for which
Eq. (42a) has a solution and Eq. (42b} does not, or vice
versa. For these quasienergies one of the constants, c or
c, will vanish. But, in general, we allow for the possibil-
ity that both of Eqs. (42) have a solution for the same
value of E. When this is the case, the constants c and c
will be determined by the requirement that Eqs. (32) are
satisfied [this is not guaranteed for arbitrary solutions to
Eqs. (40)] and by
the normalization
( 8lg)(a

a)+i
E/R
g

(@jg) I
yz+ & =0, (35a)
+aat
lgz+ )
2
+(@
jg)lyz
&=0. (35b)
&
qz I gz
& &&z l&z

&+ &
qz I gz
&

1 .
Before we can determine the constants, we must find the
state
l gz
) that corresponds to the state
l fz
) given by
Eqs. (41) and (42). For this purpose we use Eq. (35a) and
Eqs. (39) to write
We can now use Eq. (35a) to eliminate
llitz ) from Eq.
(35b). This gives
lyz
&=(e
jg)
'[0

,
'[1++1 (2ejg)']]lyz &,
[[0(E)+, '][0(E)

,
']+(8
jg)
] lPz
)
=0, (36)
(43)
where
0(E)
=
(Bjg)(a

a )+i
Ejl
g
a a+aa
(37)
2
where we indicate two alternative forms for the operator
on the right-hand side; the + sign goes with the subscript
p,
and the

sign goes with the subscript m. From Eqs.
(42) and (43) we have
+ +,
'+1

(26/ ) (39a)
Equation (36) is the desired equation for
lpga
) alone.
But it is not related in an obvious way to the eigenvalue
problem for an harmonic oscillator. In particular, Eq.
(36) is quartic, rather than quadratic in the creation and
annihilation operators. We must therefore take an addi-
tional step before we can proceed as we did in the previ-
ous calculations. We observe that the operator on the
left-hand side of Eq. (36) factorizes in the form
j
[0(E)+, '][0(E)

, ']+(8
jg)
]
=0
(E)0 (E), (38)
where
2
0 (E)=
(8/g)(a
a)+i
Ejl
P
lgz )
=
(g/26)[c [1++1

(2C/g)2]lgz. )
+c.[1

&1

(2@jg)']]lq;.
.
& .
(44)
There are two steps left in our calculation: We must
solve Eqs. (42a) and (42b} to determine the allowed
quasienergies E and the states
l gz
) and
lPz ), and we
must substitute the solutions for the states into Eqs. (32)
and
apply
the normalization condition to determine the
constants c and c . We accomplish the first task by
us-
ing displacement and squeezing transformations to con-
vert Eqs. (42a) and (42b) into eigenvalue equations for an
harmonic oscillator. We multiply Eqs. (42a) and (42b) on
the left by S (ri)D (a), where D(a) is defined in Eq. (25)
and
0 (E)= (Cjg)(a
a)+i
E/fi
+

'+1

(2A'/g)
2
2
(39b}
Now we are looking for solutions to the factorized equa-
tion
S(ri) =exp[ ,
'(riat

ri'a
)];
we transform the operators 0 (E}and 0 (E) using
S (ri)D (a)aD(a}S(7})
=(a +a)coshri+(at+a*)sinhri .
Then, if we choose
(45)
(46)
5140 P. ALSING, D.-S.
GUO, AND H. J. CARMICHAEL
45
a=P(E):i
E/fi 2@/g
1

(28/g)
ata
Iy

) [I (2g/ )2)

3/2
ri=r, e
"=+I

(2@/g)
Eqs. (42) are replaced by the equations
2
(47a)
(47b)

1
ly,
,
),
[

itic'(a

a)+E]IPE )

ih'ga
lg~ )
=
Io) . (52b)
[Note that the vacuum state IO) appears on the right-
hand side of Eq. (52b), where 0 appears in Eq. (32b).
]
It
is for this reason that the arbitrary constants c and c
appear in Eqs. (41) and (44). The ratio of these constants
is determined
by requiring that Eq. (32b) is satisfied. The
separate constants are then fixed by the normalization
(48a)
ata
IPF.
)=.
[1

(2@/g)
]
' '
IQE ),
.
(48b)
where
(q, lq,
) =(q,+l1i,
+
&+(@;I@; &
=I
.
The calculation is straightforward, but tedious, and we
therefore just quote the results. Corresponding to the
quasienergy (soa), we obtain
ly,
.
,
&=S'(r)D
(P(E))
P
.
),
Iy,
.
.
&=S (r)D (P(E))lg . ) .
(49a)
Ix, &
=
Ir, o;o& IM &, (53a)
and corresponding to the quasienergies (sob), we obtain
Equations (48) are satisfied
by Fock states when the con-
stants on the right-hand sides are the non-negative in-
tergers. Thus, the quasienergies are
IX.
..
& =(I/&2)
[lr, p(E.
..
);n

I& IP &
+ilr, P(E);n& IM & ],
Eo
=0, (soa)
ly, (
& =(1/&2)[lr,
p(E,();n

1&IP &
(53b)
which is permitted by Eq. (48b) but not
by Eq. (48a), and
E=+R&ng[1

(2@/g)
]
~
EI= fi&ng[1

(2C/g)
]
~
(50b)
(n=1,2, ...) where
ilr, P(E&);n ) IM ) ],
(n=1,2, ...) which are permitted by both Eqs. (48a) and
(48b). The corresponding states are
I&F
,
& :
If&~&.
=ln

1&, n =1,
2, . . .
,
g=u, l
(Sla)
and
IP) =(I/&2)[[1++1

(2@/g)
]'~
I+ )

[1

&I

(26/g)']' 'I

)
j,
IM )
=( I/&2)
[
[1++I

(28/g)
]'
I

)
(S4a)
Ix.
, .
&=Io&
)=
'
0
IX.
. .
&
=
In &,
n =1,2, . . .
,
g=u, l .
(5 lb)

[1

+I

(2@/g)']' 'I+
)
j;
(54b)
the states
I r, p(E&);
n

1 ) and
I r, p(E&);
n ),
g
=
u, I,
are squeezed and displaced Fock states:
(52a)
Equations (41), (44), (49), and (51) define states
lyo
)
and
ly&),
n=1,
2, ...
, g=u, l, that satisfy Eqs. (33). We
seek solutions to Eqs. (32), and while these must satisfy
Eqs. (33), the converse is not true. This follows because
we multiplied Eq. (32b) by
a to obtain Eq. (33b); there-
fore, in addition to the states that satisfy Eqs. (32), the
solutions to Eqs. (33) also include states that satisfy
[

i'd@(a

a )+E]I fE
) +i figa I fE
)
=0,
Ig,
a;n ) =D(a)S(g)ln ) . (55)
From Eqs. (11), (53), and (54), the steady states of the
driven Jaynes-Cummings system with the external field
driving the cavity mode are
lyo)
=e
'
Ie
'r,
o;0) IM, )
and
where
+exile
r, e
P(E&);n ) IM, ) ],
n =1,
2, . . .
, (=u, l (e&= 1) (56b)
IP,
&=(IW'2)[[1++1

(26/g)
]'~
I+)

e
'[1

+I

(2e/g)
)'
I

&j,
(57a)
45 DYNAMIC STARK EFFECT FOR THE JAYNES-CUMMINGS. . . 5141
~M, ) =(1/&2)[[1++1

(2C/g)
]'
~

e [1

+1

(2A/g)
]'
~+ )],
(57b)
V. DISCUSSION
The quasienergies (50) and steady states (56) are the
central results of this paper. These results have various
applications to problems in cavity quantum electro-
dynamics which we will explore in future work. In this
section we point out some of the more obvious connec-
tions with cavity quantum electrodynamics.
The quasienergies (50}define shifted Jaynes-Cummings
level splittings. The shifts are relevant to proposed spec-
troscopic measurements on the Jaynes-Cummings sys-
tem. For example, there is interest in making direct
frequency-space measurements of the Jaynes-Cummings
spectrum, with particular emphasis on the splitting of the
first excited state

the so-called "vacuum"


Rabi split-
ting. Comparing Eqs. (18b) and (50b), we see that the size
of this splitting depends on the way it is observed; prob-
ing the atom or the cavity mode gives different results.
There are no observations yet of vacuum" Rabi splitting
for a single atom. But there are two observations in
many-atom systems [17,18]. Both of these experiments
coherently excite the cavity mode. In a single-atom ex-
periment the modulation technique used by Raizen et al.
[17] would observe the splitting 2g [1

(2g/g)2]
~,
where 8 is the amplitude of the carrier field. For a direct
transmission measurement like the one performed by Zhu
et al. [18] it is not possible to make a quantitative predic-
tion from the present results because this measurement
involves detunings that are not included in our calcula-
tion. We can say, however, that the quasienergies shift in
response to the driving field and that this shift will
change as the frequency of the driving field is swept. Of
course, in either of these methods the frequency shifts in-
duced by the driving Geld can be made small by reducing
the driving field amplitude; indeed, if 2C/g is not very
small ( &0.1), we observe numerically that spectra mea-
sured by coherent excitation include contributions from
multiphoton resonances involving states above the first
excited state.
It is interesting to compare our result for the shift of
the single-atom "vacuum"
Rabi peaks with what we
would expect for many-atom "vacuum"
Rabi splitting.
There has been some discussion of the fact that, for weak
excitation, the spectroscopic features for the single-atom
system and the many-atom system are the same and can
be understood in terms of a classical coupled harmonic-
oscillator model (linear dispersion theory) [17,18,22].
This equivalence does not extend to the frequency shifts
induced
by the driving field. For definiteness we compare
single-atom and many-atom systems with a resonant
coherent field driving the cavity mode; to probe the fre-
quency structure, this field will carry a small modulation
in the manner of the experiment of Raizen et al. [17].
We identify the quasienergies (single atom} and eigenval-
ues of linearized Bloch equations (many atoms) that
define the "vacuum"
Rabi peaks in the limit of weak
driving fields and ask how these quantities change as a
function of driving field strength. Of course, if the driv-
ing field is too strong, major differences between the spec-
tra will arise because of excitations beyond the first excit-
ed state. For a meaningful comparison, we therefore con-
sider only weak-field perturbations of the "vacuum"
Rabi
peaks (2C/g 1).
For the single-atom system, the vacuum" Rabi split-
ting changes according to the expression
(E) E,
(} lfi=2g
[1

(2@/g)
]
=2g[1

3(6'/g)
]
. (58)
For the many-atom system, we obtain the "vacuum"
Rabi peaks from eigenvalues of the linearized optical
Bloch equations. In the limit of weak driving fields, these
equations take the form of coupled oscillator equations:
a= ca+&
NgP,
P
=
(y /2)13

2&Ngm
ssa,
(59a)
(59b)
where a and
P
are the amplitudes of the field and polar-
ization oscillators, v and y/2 are the half-widths of the
cavity and atomic resonances, N is the number of atoms,
and
mss
is the steady-state inversion in the presence of
the coherent field driving the cavity. Normally, for weak
driving fields we set m
ss
=
,
',
'
then, when
&Ng a, y/2, the eigenvalues of the 2X2 matrix
defined
by the right-hand sides of Eqs. (59) give the
many-atom "vacuum"
Rabi splitting 2&Ng. The first-
order correction to this result is obtained using
mss=

,
'[1

2N '(@/v'Ng)
];
(60)
this is the saturated inversion calculated taking the intra-
cavity absorption into account (8 describes the external
field, not the field inside the cavity). The many-atom
"vacuum"
Rabi splitting is then
2&Ng+2~mss ~
=2v'Ng
[1
N'(v /v'Ng)~]

. (61)
Here the shift induced by the driving field is negligible in
the large-Nlimit. The N dependence is consistent with
the fact that the many-atom system behaves as a pair of
coupled harmonic oscillators
up to corrections of order
N
Another connection between our results and previous
work in cavity quantum electrodynamics comes from the
squeezing involved in constructing the steady states (56).
Carmichael showed that the field transmitted
by a
coherently driven atom containing a single atom is
squeezed [23). This squeezing is related to the squeezing
in absorptive optical bistability [24] which has been ob-
served in a many-atom system [25]. Rice and Carmichael
showed that the presence of squeezing induces a narrow-
ing of the "vacuum"
Rabi peaks in incoherent spectra,
replacing Lorentzians
by squared Lorentzians [26]. The
5142 P. ALSING, D.-S.
GUO, AND H. J. CARMICHAEL 45
steady states (56) provide a new view of squeezing-related
effects in the driven Jaynes-Cummings system. For ex-
ample, the "ground state" of the driven Jaynes-
Cummings system (the steady state corresponding to the
quasienergy Eo
=0) is the product of a squeezed state for
the field and the state
~M, ) for the atom. With spontane-
ous emission and cavity loss included, we can show that,
for weak driving fields, the driven Jaynes-Cummings sys-
tem settles (to lowest order) in this "ground state.
"
Thus,
our analysis of the Hamiltonian (1) identifies the basic
origin of the squeezing, and the steady states (56) provide
a natural basis for calculating effects such as squeezing-
induced linewidth narrowing [27], which is difficult to
calculate using bare-energy eigenstates (eigenstates of
H ) or standard dressed states (eigenstates of
0
Ho
=
H+Hs ).
Our results are also related to the phenomenon of
spontaneous dressed-state polarization [16]. In this
phenomenon the asymptotic quantum state of a driven
cavity mode interacting with a two-state atom undergoes
a novel symmetry-breaking transition as the strength of
the driving field is increased. Above the transition
threshold, the intracavity field shows a bimodality in
phase; the two-phase states result from a spontaneous po-
larization of the atom in one or the other of the two semi-
classical dressed states produced by the intracavity field.
To see the connection between this phenomenon and our
present work, we observe that the quasienergies (50) and
steady states (56) are only valid for 2C/g (1. For larger
values of 28/g, discrete quasienergies and normalizable
steady states do not exist, although a continuum of states
satisfying a 5-function normalization will exist [28]. The
boundary 28/g= 1 is the threshold found in our work on
spontaneous dressed-state polarization [16]. Moreover,
the atomic states
~M, ) and
~P, ) given by Eqs. (57) are
precisely the steady states found, below threshold, from
our semiclassical analysis of this phenomenon [Eqs. (24b)
and (24c) in Ref.
[16]]. In fact, when continued for
26
/g
& 1
~M, ) and
~ P, ) also reproduce the steady states
given by
our semiclassical analysis above threshold[Eqs.
(25b) and (25c) in Ref. [16]]. Thus, the basic physics
un-
derlying spontaneous dressed-state polarization is con-
tained in Hamiltonian (1) and can be understood without
considering the dissipative terms that are present in a full
treatment of this phenomenon.
Finally, we should say something about detunings.
There are two detunings that might be added to the
driven Jaynes-Cummings Hamiltonian: a detuning A~
between the external field and the atom, and a detuning
4C
between the external field and the cavity mode. It is
straightforward to generalize our results to include a de-
tuning b,
.
This changes E to E+A'b,

in Eq. (13a) and


to E fih
z
in Eq.

(13b). After these changes, the
methods used to solve Eqs. (13) carry through with minor
modifications. As an example of the results, the quasien-
ergies
E&=e&fi&ng[1

(26/g)
],
g'=u, l (p&=+I)
are replaced by
Eg=EgN 1

(2@/g)2
X [(b, /2)2+ng2V 1

(2g/g)~]'~2
(62)
A detuning between the driving field and the cavity mode
is more diScult to handle. A detuning A~ changes E to
E+A'b
caa in Eqs. (13). The term fihca a does not
commute with the other field operators in Eqs. (13) and,
as a result, the methods we used to solve Eqs. (13) for
he
=0
cannot be used when b, c%0. One property of the
solution in this case does seem fairly clear, however. We
expect that a detuning b, c%0 will remove the singular
behavior at 2C/g= 1 and allow discrete quasienergies for
all driving field strengths. This follows because the boson
term ifr8(a

a)+irihca a that dominates the Hamil-
tonian in the strong-driving-field limit and does not have
a discrete spectrum for b,
z
=0
does have a discrete spec-
trum for b,
CIAO.
VI. SUMMARY
We have calculated the quasienergies and steady states
of the driven Jaynes-Cummings system

a single atom
coupled to a single electromagnetic cavity mode with ei-
ther the atom or the cavity mode driven
by a coherent
external field. When the atom is driven by the external
field, we find that the quasienergies give the usual
dressed-state level splittings +fiV n
g,
n =0, 1,2, ,
where...
g
is the dipole coupling constant and the steady states are
displaced dressed states. When the cavity mode is driven
by the external field, we find that the quasienergies give
shifted level splittings
+&ng[1

(2A'/g)
),
n =0,
1, ,
. . .
,
where
6'
is the amplitude of the driving field. In this case
the steady states are formed from superpositions of atom-
ic states multiplied by squeezed and displaced Fock
states; for weak driving fields, they are just squeezed and
displaced dressed states. Above the threshold 28/g= 1,
discrete quasienergies and normalizable steady states do
not exist.
We have given explicit results with the atom, the cavi-
ty mode, and the driving field all on resonance. The alge-
braic method used to obtain these results is easily gen-
eralized to include a detuning between the driving field
and the atom. We have not found an analytically tract-
able solution when there is a detuning between the driv-
ing field and the cavity mode.
Our results are relevant to work in cavity quantum
electrodynamics. The quasienergies have obvious
relevance to spectroscopic measurements on the Jaynes-
Cummings system [17,18] and the steady states provide a
useful basis for analytical calculations of spectra. A par-
ticular attraction of the steady states as a basis is that
45 DYNAMIC STARK EwvaCT FOR THE JAYNES-CUMMINGS. . . 5143
they automatically incorporate the squeezing
e6'ects
found in earlier work on the driven Jaynes-Cummings
system [23,26]. Our results also provide insight into the
threshold behavior found in recent numerical calcula-
tions for the driven Jaynes-Cummings system with
dissipation

spontaneous dressed-state polarization [16].
There appears to be ample scope for the application of
our results.
ACKNOWLEDGMENTS
This work was supported by the National Science
Foundation under Grant No. PHY-9096137. We are
grateful to Dr. Marcus Lindberg for sending us his
derivation of the quasienergies (50b) and to Dr. Craig Sa-
vage for discussions on the subject of frequency shifts in
the driven Jaynes-Cummings system.
'Present address: Department of Physics, University of
Windsor, Windsor, Ontario, Canada N9B 3P4.
[1]
I. I. Rabi, Phys. Rev. 51, 652 (1937).
[2]
F. Bloch, Phys. Rev. 70, 460 (1946).
[3]
E. T. Jaynes and F. W. Cummings, Proc. IEEE 51, 89
(1963).
[4]
P. L. Knight and P. W. Milonni, Phys. Rep. 66, 21 (1980).
[5]
Y. I. Yoo and J. H. Eberly, Phys. Rep. 118, 239 (1985).
[6]
M. O. Scully and W. E. Lamb, Jr.
, Phys. Rev. 159, 208
(1967); 166, 246 (1968).
[7]
H. Haken, Licht und Materie Ic, Handbuch der Physik
Vol. XXV/2c, edited
by L. Genzel (Springer, Berlin,
1970).
[8]
M. Weissbluth, Photon Atom In-teractiosn (Academic, Bos-
ton, 1989), pp.
392-395.
[9]
V. G. Weisskopf and E. Wigner, Z. Phys. 63, 54 (1930).
[10] H. J. Carmichael and D. F. Walls, J. Phys. B 9, L43
(1976); 9, 1199(1976).
[11]H. J. Kimble and L. Mandel, Phys. Rev. A 13, 2123
(1976); H. J. Kimble, M. Dagenais, and L. Mandel, Phys.
Rev. Lett. 39, 691 (1977).
[12] S. Haroche and J. M. Raimond, in Advances in Atomic and
Molecular Physics, edited by D. Bates and B. Bederson
(Academic, New York, 1985), Vol. 20, pp.
347-411.
[13]J. H. Eberly, N. B. Narozhny, and J. J. Sanchez-
Mondragon, Phys. Rev. Lett. 44, 1323 (1980).
[14] J. J. Sanchez-Mondragon, N. B. Narozhny, and J. H. Eber-
ly, Phys. Rev. Lett. 51, 550 (1983).
[15] G. Rempe, H. Walther, and N. Klein, Phys. Rev. Lett. 58,
353 (1987).
[16]
P. Alsing and H. J. Carmichael, Quantum Opt. 3, 13
(1991).
[17] M. G. Raizen, R. J. Thompson, R. J. Brecha, H. J. Kim-
ble, and H. J. Carmichael, Phys. Rev. Lett. 63,240 (1989).
[18] Y. Zhu, D. J. Gauthier, S.E. Morin, Q. Wu, H. J. Carmi-
chael, and T. W. Mossberg, Phys. Rev. Lett. 64, 2499
(1990).
[19]H. Sambe, Phys. Rev. A 7, 2203 (1973).
[20] J. M. Okuniewicz, J. Math. Phys. 15, 1587 (1974).
[21]This result for the quasienergies was first derived
by M.
Lindberg using a generating function technique (private
communication).
[22] H. J. Carmichael, Phys. Rev. A 44, 4751 (1991).
[23] H. J. Carmichael, Phys. Rev. Lett. 55, 2790 (1985).
[24] L. A. Lugiato and G. Strini, Opt. Commun. 41, 67 (1982).
[25] M. G. Raizen, L. A. Orozco, M. Xiao, T. L. Boyd, and H.
J. Kimble, Phys. Rev. Lett. 59, 198 (1987); L. A. Orozco,
M. G. Raizen, M. Xiao, R. J. Brecha, and H. J. Kimble, J.
Opt. Soc. Am. B4, 1490 (1987).
[26] P. R. Rice and H. J. Carmichael, J. Opt. Soc. Am. B 5,
1661 (1988); H. J. Carmichael, R. J. Brecha, M. G. Raizen,
H.J. Kimble, and P. R. Rice, Phys. Rev. A 40, 5516 (1989).
[27] P. Alsing, Ph. D. thesis, University of Arkansas, 1991.
[28] C. F. Lo, Quantum Opt. 3, 333 (1991).

Você também pode gostar