Collective effects in the collapse-revival phenomenon and squeezing in the Dicke model

G. Ramon,∗ C. Brif,† and A. Mann‡

Department of Physics, Technion – Israel Institute of Technology, Haifa 32000, Israel

and revivals of the Rabi oscillations, which manifests it-

Resonant interaction of a collection of two-level atoms with self in the clearest way when the cavity field is prepared
a single-mode coherent cavity field is considered in the frame- initially in the coherent state [8]. The shape of collapses
work of the Dicke model. We focus on the role of collective
and revivals is determined by the initial photon-number
atomic effects in the phenomenon of collapses and revivals of
distribution. A similar behavior can be found also in the
the Rabi oscillations. It is shown that the behavior of the
many-atom case [9–11]. For a sufficiently strong coher-
system strongly depends on the initial atomic state. In the
case of the initial half-excited Dicke state we account for a ent field, the nonlinearity in the Rabi frequency is slight
arXiv:quant-ph/9809051v1 17 Sep 1998

number of interesting phenomena. The correlations between and the system exhibits regular dynamics in the form of
the atoms result in a suppression of the revival amplitude, and collapses and revivals of the oscillations. However, in the
the revival time is halved, compared to the uncorrelated fully- many-atom case there exist anharmonic collective cor-
excited and ground states. The phenomenon of squeezing of rections which modify the shape of the collapses and re-
the radiation field in the atom-field interaction is also dis- vivals related to the photon-distribution mechanism [12].
cussed. For the initial fully-excited and ground atomic states, In the present work we study in detail these collective
the field is squeezed on the short-time scale, and squeezing can effects for different initial atomic states. If atoms are
be enhanced by increasing the number of atoms. Some em- prepared initially in the ground state or in the fully-
pirical formulas are found which describe the behavior of the excited state, then the system behaves rather similarly
system in excellent agreement with numerical results. For the to the single-atom case, although collective effects man-
half-excited Dicke state, the field can be strongly squeezed on ifest themselves clearly in the length of the revival time.
the long-time scale in the case of two atoms. This kind of But, the results are different when the atoms are pre-
squeezing is enhanced by increasing the intensity of the ini- pared initially in the half-excited Dicke state. This state
tial coherent field and is of the same nature as revival-time is well known as the superradiant atomic state in the
squeezing in the Jaynes-Cummings model. The appearance context of collective spontaneous emission in open space
of this long-time squeezing can be explained using the factor- [1,2]. It was also found [13] that the half-excited Dicke
ization approximation for semiclassical atomic states. state strongly modifies the resonance fluorescence spec-
42.50.Ct, 42.50.Md, 42.50.Dv, 32.80.-t
trum of the atomic system. However, to the best of our
knowledge, this state was rarely considered in the con-
text of the Tavis-Cummings (cavity) version of the Dicke
model. When the atoms are prepared in the half-excited
I. INTRODUCTION Dicke state the system exhibits a number of interesting
phenomena. In particular, the amplitude of the Rabi os-
Since the pioneering work of Dicke [1] on cooperative cillations is strongly suppressed and relative intensities of
spontaneous emission, a great deal of attention has been revivals are essentially changed. This suppresion of the
devoted to the interaction of the radiation field with a revival amplitude is similar to the trapping phenomenon
collection of two-level atoms located within a distance occuring in the case of a single atom prepared in the
much smaller than the wavelength of the radiation. Such equally weighted superposition of the two levels [14].
a system is commonly referred to as the Dicke model It has been known for a long time [15,16] that the non-
(for a review see, e.g., Ref. [2]). A particular case of the linear character of the Jaynes-Cummings model leads to
Dicke model, when atoms interact with a single-mode squeezing in one of the quadratures of an initially coher-
radiation field inside a cavity, was considered by Tavis ent cavity field. It was also predicted [17] that strong
and Cummings [3]. The Tavis-Cummings Hamiltonian is squeezing can be obtained in the Jaynes-Cummings
mathematically equivalent to the trilinear boson Hamil- model near the revival times for large initial intensities
tonian describing various nonlinear optical processes [4]. of the field. This phenomenon of strong revival-time
The single-atom version, known as the Jaynes-Cummings squeezing was explained in Ref. [18] using the factoriza-
model [5], is the simplest and one of the most popular tion approximation for the so-called semiclassical atomic
models of quantum optics. In spite of its simplicity, the states [19]. Butler and Drummond [20] showed that
Jaynes-Cummings model shows a variety of interesting short-time squeezing can be enhanced in the Dicke model
nonclassical phenomena such as vacuum-field Rabi oscil- compared to the single-atom case. Seke [21] considered
lations, sub-Poissonian photon statistics, and squeezing field and atomic squeezing in the Dicke model without
of the radiation field (for reviews see, e.g., Refs. [6,7]). the rotating-wave approximation and in the presence of
One of the most interesting quantum features of the losses. Analytical approaches to squeezing in the strong-
Jaynes-Cummings model is the phenomenon of collapses field limit [22] and in the weak-field limit [23] were pre-

1
sented. Higher-order squeezing in the Dicke model was commutes with the Hamiltonian (2.1) and is an inte-
studied in Ref. [24]. For atoms prepared initially in the gral of motion. Another integral of motion is the SU(2)
ground state or in the fully-excited state, the uncertainty Casimir operator J2 = Jz2 + 21 (J+ J− + J− J+ ). We de-
of the field quadrature rapidly oscillates and squeezing scribe the state of the atomic system in terms of the
occurs during short periods of time at the very beginning SU(2) orthonormal basis |j, miat (m = j, j − 1, . . . , −j),
of the time evolution. On the long-time scale the tempo-
ral behavior of the quadrature uncertainty is correlated Jz |j, miat = m|j, miat , (2.5)
with times of collapses and revivals of the Rabi oscil- J2 |j, miat = j(j + 1)|j, miat . (2.6)
lations. On the other hand, for the initial half-excited
Dicke state, the behavior of squeezing is qualitatively In the context of the atomic realization (2.3), the states
different. The quadrature uncertainty oscillates on the |j, miat are the symmetric Dicke states:
long-time scale, with a period of the order of the revival
p
−1/2 X Y
half-time, and strong squeezing is obtained in the case of

N Y
two atoms. The value of squeezing is enhanced by the |j, miat = |+ilk |−il , (2.7)
p
intensity of the initial coherent field. This phenomenon k=1 l6=lk

is of the same nature as revival-time squeezing in the

where |+il and |−il are the upper and lower states, re-
Jaynes-Cummings model and can be explained using the
spectively, of the lth atom, and the summation is over
factorization approximation for the semiclassical atomic
all possible permutations of N atoms. If only symmetric
states. In general, squeezing is better and its duration is
atomic states are considered, then the ‘cooperative num-
much longer for the initial two-atom half-excited Dicke
ber’ j is equal to N/2 and p = m + j is just the number
state than for the ground and fully-excited states.
of excited atoms.
The Hilbert space H of the atom-field system can be
II. THE MODEL AND METHODS OF SOLUTION
decomposed into a direct sum of finite-dimensional in-
variant subspaces HL :
We consider the resonant interaction between N two- ∞
M
level atoms and the single-mode radiation field inside a H= HL . (2.8)
lossless cavity. In the rotating-wave approximation, the L=0
Tavis-Cummings interaction Hamiltonian reads (h̄ = 1):
Each invariant subspace HL is spanned by the orthonor-
†
H = g(a J− + aJ+ ). (2.1) mal basis |nif |j, L − j − niat , where |nif are the Fock
states of the radiation field, a† a|nif = n|nif . For L < N ,
Here and in the following we use the interaction picture. n = 0, 1, . . . , L and dim(HL ) = L + 1; for L ≥ N ,
In Eq. (2.1) g is the coupling constant, a and a† are the n = L − N, L − N + 1, . . . , L and dim(HL ) = N + 1. If the
annihilation and creation operators of the field mode, J+ field is initially in the Fock state |n0 if and the atoms are
and J− are the collective atomic raising and lowering in the Dicke state |j, m0 iat , the state of the system will
operators. They satisfy the su(2) Lie algebra, evolve in the invariant subspace HL with L = n0 +m0 +j.
For the field and/or atoms prepared initially in a super-
[J+ , J− ] = 2Jz , [Jz , J± ] = ±J± , (2.2) position state, one should take into account contributions
from different subspaces.
where Jz is the operator of atomic inversion. In terms The exact solution of the problem is obtained by the di-
of the standard Pauli matrices, describing each two-level agonalization of the interaction Hamiltonian (2.1) in each
atom, one obtains of the invariant subspaces HL involved [3]. It is known
[3,4] that in the basis |nif |j, L − j − niat the Hamiltonian
N N
1 X (i) 1 X (i) is given by a tridiagonal matrix with symmetric eigenval-
J± = σ , Jz = σ . (2.3)
2 i=1 ± 2 i=1 z ues and the corresponding characteristic equation can be
reduced to an algebraic equation of order [dim(HL )/2].
If the Schwinger realization of the SU(2) generators is Therefore, an analytical solution is possible when only in-
used, the Hamiltonian (2.1) becomes the trilinear bo- variant subspaces with dim(HL ) ≤ 9 are involved. How-
son Hamiltonian describing nonlinear optical processes ever, already for dim(HL ) > 3, analytical solutions are
such as parametric conversion and Raman and Brillouin rather complicated [26–28]. Note that for dim(HL ) ≥ 4
scattering [4]. An interesting physical realization of the the eigenvalues are not equidistant, so the time evolution
Hamiltonian (2.1) is given by the coupling of the inter- is not periodic even within a single subspace. Semiclas-
nal levels of atoms or ions to a mode of their quantized sical approximate solutions were proposed [29–31] which
oscillatory motion in a harmonic trap [25]. give the time evolution of systems governed by a trilinear
The total excitation operator Hamiltonian of type (2.1) in terms of elliptical functions.
A perturbative analytical approach to the problem with
L = a† a + Jz + N/2 (2.4) weak fields was developed by Kozierowski and co-workers

2
[27,28,32,33]. In the present work we use the exact solu- One should expect a similar behavior also for the Dicke
tion based on the numerical diagonalization of the inter- model in the strong-field domain n̄ > N [9–12]. Of
action Hamiltonian (2.1). course, the collapses and revivals related to the photon-
number distribution would be modified by the collec-
tive atomic effects due to the fact that the eigenvalues
III. COLLAPSES AND REVIVALS of the interaction Hamiltonian are not equidistant. In
the strong-field limit n̄ ≫ N the anharmonic correc-
We study the temporal behavior of the atom-field sys- tions to the eigenvalues become small, and one should
tem in the Dicke model for the cavity field prepared ini- expect a quite regular behavior similar to that of the
tially in the coherent state |αif : resonant Jaynes-Cummings model. However, we will see
that there exist reasons for an irregular behavior that
∞ are much more important than just the anharmonicity of
2 X αn
|αif = e−|α| /2
√ |nif . (3.1) the eigenvalues. In fact, the initial atomic state deter-
n=0 n!
mines how important will be various factors leading to
irregularities in the behavior of the system.
Without loss of generality we consider only real values of
α. The initial atomic state is supposed to be one of the
0.5
Dicke states |j, miat (recall that j = N/2). Two possi- a
bilities which are frequently considered in the literature 0
are the fully-excited state |j, jiat and the ground state
|j, −jiat . We are also interested in the half-excited Dicke −0.5
0 20 40 60 80 100 120 140 160 180 200
state |j, 0iat .
5
Since the coherent field state is a superposition of many b

< Jz >
Fock states |nif , the invariant subspaces HL with differ- 0
ent values of L = n + m + j contribute to the evolution.
The temporal behavior of the atomic inversion hJz i is −5
0 20 40 60 80 100 120 140 160 180 200
given by the sum of the appropriately weighted atomic
5
responses to each Fock state. (The mean photon num- c
ber ha† ai is connected with the atomic inversion by the 0
fact that hLi = ha† ai + hJz i + N/2 is a constant.) In
the resonant Jaynes-Cummings model (N = 1) with ini- −5
tially unexcited (m = − 21 ) or excited (m = 21 ) atom, one 0 20 40 60 80 100 120 140 160 180 200

obtains [8] τ
FIG. 1. The atomic inversion hJz i versus the scaled time
∞
X τ = gt for the initial coherent field state with n̄ = 36 in-
hJz i = m Pn cos Ωn,m τ, (3.2) teracting with (a) one excited atom (the Jaynes-Cummings
n=0 model) and with 10 atoms prepared in (b) the fully-excited
state |j, jiat and (c) the ground state |j, −jiat .
where τ = gt is the scaled time,
p We first consider the cases with the atoms prepared
Ωn,m = 2 n + m + 1/2 (3.3)
initially in the fully-excited and ground states. The tem-
is the scaled Rabi frequency corresponding to each sub- poral behavior of the atomic inversion hJz i is shown in
space, and Pn is the photon-number distribution. For the Fig. 1(b,c) for the case of N = 10 and n̄ = 36, with atoms
initial coherent state, Pn = exp(−n̄)n̄n /n! is the Poisso- prepared in (b) the fully-excited state |j, jiat and (c) the
nian distribution and n̄ = |α|2 is the initial mean pho- ground state |j, −jiat . The first feature to note is the de-
ton number. Then, due to the property of this distribu- pendence of the revival time on the initial atomic state.
tion, the most regular dynamics occurs for large values Similarly to the Jaynes-Cummings model, we can esti-
of the initial mean photon number n̄. Contributions cor- mate the revival time for the initial atomic state |j, miat
responding to different n’s interfere in such a manner as
that they initially go out of phase, after that acquire a p
τR = 2π n̄ + m + 1/2, (3.5)
common phase, and this process is repeated resulting in
a series of collapses and revivals, as shown in Fig. 1(a). where we use the strong-field limit expression for the Rabi
The revival time τR can be estimated using the condition frequency (i.e., neglect the anharmonic corrections to the
[8,34] eigenvalues). Formula (3.5) is in a good agreement with
our numerical results, as demonstrated in Fig. 2. The
τR (Ωn̄+1,m − Ωn̄,m ) = 2π, (3.4)
difference between the revival times for the fully-excited
p state and the ground state is particularly obvious when
which gives τR = 2π n̄ + m + 1/2.
N is not too small compared to n̄. The second feature is

3
that the spread of the revivals is determined only by the faster than for m = j. In particular, we see that the
photon statistics of the initial field state [34] and does not amplitude of the revivals for m = −j is smaller than for
depend on the value of m. These two observations may m = j, and this effect becomes more pronounced as N
explain why the temporal behavior loses its regularity increases. For n̄ = 36, as N increases from 2 to 16, the
for the initial ground state notably faster than for the relative amplitude A1 (the difference between the max-
initial fully-excited state. Indeed, we see in Fig. 1(b) imum and minimum values of hJz i in the first revival,
four regular revivals for m = j, while in Fig. 1(c) for divided by N ) decreases from 0.5459 to 0.4039 for m = j
m = −j the behavior starts to be irregular already at the and from 0.5362 to 0.2112 for m = −j. This effect occurs
end of the third revival. The reason is that the spread of because the anharmonic corrections partially destroy the
the revivals increases with time and neighboring revivals interference of the oscillating terms. Of course, as the
start to overlap, which leads to the loss of regularity. initial mean photon number n̄ increases, the behavior of
Since for m = −j the revival time is shorter than for m = the system becomes more regular.
j, the overlapping of the revivals and the corresponding Now we turn to the case when the atoms are prepared
irregularity occur earlier. in the half-excited Dicke state |j, 0iat (for even values of
N ). The temporal behavior of the atomic inversion hJz i
2 a is shown in Fig. 3 for n̄ = 36 and N = 2, 6, 10. Here
we see a number of interesting phenomena. First of all,
0
the amplitude of the Rabi oscillations is significantly sup-
−2 pressed. For n̄ = 36, as N increases from 2 to 16, the
25 30 35 40 45 50 relative amplitude A0 (the difference between the maxi-
b mum and minimum values of hJz i in the very beginning
< Jz >

0 of the evolution, divided by N ) increases linearly from

−0.2 0.006 to 0.027, according to the empirical formula
−0.4
25 30 35 40 45 50 A0 = 0.0015(N + 2). (3.6)
2 c For comparison, A0 is about 0.9 for m = j and about
0
0.95 for m = −j, being almost independent of N .
−2
0
25 30 35 40 45 50 a
τ −0.005

−0.01
FIG. 2. The atomic inversion hJz i versus the scaled time
τ = gt for the initial coherent field state with n̄ = 36 inter- −0.015
0 20 40 60 80 100 120 140 160 180 200
acting with with 10 atoms prepared in (a) the fully-excited
state |j, jiat , (b) the half-excited Dicke state |j, 0iat , and (c) b
< Jz >

0
the ground state |j, −jiat . The region of the first revival is −0.05
shown, demonstrating the dependence of the revival time on −0.1
the initial atomic state. −0.15
0 20 40 60 80 100 120 140 160 180 200

Another factor that leads to differences between the c

0
behavior of the system in the cases m = j and m = −j
is the dependence of the anharmonic corrections to the −0.2

eigenvalues on the value of m. In order to give a rep- −0.4

0 20 40 60 80 100 120 140 160 180 200
resentative example, we considered the eigenvalues hi τ
of the interaction Hamiltonian for the subspaces with
FIG. 3. The atomic inversion hJz i versus the scaled time
L = n + m + j for n = 36, N = 10, 20, 30 and m = ±j, 0.
τ = gt for the initial coherent field state with n̄ = 36 interact-
The ratio f51 = h5 /h1 of the 5th and 1st eigenvalues ing with N atoms prepared in the half-excited state |j, 0iat :
was chosen as a parameter representing the anharmonic- (a) N = 2, (b) N = 6, (c) N = 10.
ity (for equidistant eigenvalues f51 is exactly 5). The
numerical results are listed in Table I. We see that for Another important feature is the appearance of half-
given N the value of f51 increases with decrease of m, i.e., time revivals, i.e., for m = 0 the revival time is actu-
irregular effects related to the anharmonicity are most ally τR /2. In fact, tiny half-time revivals appear also for
important for m = −j. For m = j the value of f51 de- m = ±j, but their amplitudes are much smaller than the
creases slightly with increase of N , but the overall con- amplitudes of the first revivals at τR . As N increases
tribution of the anharmonic corrections increases with N from 2 to 16, the relative amplitude A1/2 (the difference
just because there are more eigenvalues. For m = −j between the maximum and minimum values of hJz i in
the value of f51 increases with N , so the importance of the half-time revival, divided by N ) increases monotoni-
the anharmonic corrections here increases with N much cally from 0.0018 to 0.0171 for m = j and from 0.0019 to

4
0.0354 for m = −j. The ratio A1/2 /A1 increases in the and for the half-excited Dicke state (m = 0):
same range of N from 0.0033 to 0.0424 for m = j and
∞
from 0.0035 to 0.1677 for m = −j. On the other hand, for X 1
m = 0 the half-time revival and the first revival are, for hJz i = − Pn [1 − cos(2Ωn,0 τ )] . (3.10)
n=0
Ω2n,0
small values of N , of the same order of magnitude. For
N = 2 we even find A1/2 /A1 > 1. In contradistinction to Now it is clear why the oscillations are strongly sup-
the cases m = ±j, for m = 0 the ratio A1/2 /A1 decreases pressed for m = 0. Here both the constant and oscillatory
from 1.3427 to 0.0473 as N increases from 2 to 16. In this terms have the prefactor Pn /(4n+2), while for m = ±j in
range of N the relative amplitude A1/2 of the half-time addition to the constant and oscillatory terms with pref-
revival increases slightly from 0.0037 to 0.0052, while the actors of the order Pn /n there exists an oscillatory term
relative amplitude A1 of the first revival increases much with a prefactor of the order Pn . For n̄ = 36 the suppres-
faster: from 0.0028 to 0.1093. Starting from N = 4 the sion is by two orders of magnitude. For m = 0 the oscil-
amplitude of the first revival exceeds that of the initial latory term has the frequency 2Ωn,0 = 2(n + 1/2)1/2, and
oscillations. The ratio A1 /A0 increases almost linearly the effective revival time is τR /2 = π(n̄+ 1/2)1/2 . On the
from 0.4555 to 4.0476 as N increases from 2 to 16. The other hand, for m = ±j the leading oscillatory term with
value hJz iC of the atomic inversion during the collapse the prefactor of the order Pn has the frequency Ωn,±1 ,
is always positive for m = ±j. As N increases from 2 to associated with the revival time τR , while the smaller os-
16, hJz iC increases from 0.0036 to 0.3569 for m = j and cillatory term with the prefactor of the order Pn /n has
from 0.0035 to 0.5195 for m = −j. On the other hand, the double frequency, leading to the half-time revivals.
for m = 0 the value of hJz iC is always negative and de- This explains why for m = ±j the half-time revival is
creases from −0.0070 to −0.2433 in the same range of N . much smaller than the first revival while for m = 0
We found that for m = 0 the value of hJz iC can be well both types of revivals are of the same order of magni-
approximated by the empirical formula tude. The more irregular dynamics in the case m = 0
 z(α) can be explained by two reasons.
√ First, the nonlinearity
N of the Rabi frequency (∼ n) is less important for larger
hJz iC = − ,
k1 α − k2 values of n, so the dynamics is more regular when the
z(α) = k3 − exp[−(k4 α + k5 )], (3.7) main contribution comes from larger n’s. For m = ±j
the prefactors of the leading terms are Poissonian, so for
where k1 = 7.45, k2 = 11.16, k3 = 1.773, k4 = 0.328, large enough n̄, the main contribution will come from
k5 = 1.681. We also see that in general the dynam- the high-frequency terms, resulting in a regular behav-
ics in the case m = 0 is much more irregular than for ior. However, for m = 0 the prefactors are Pn /(4n + 2),
m = ±j. All these observations show that the structure so lower frequencies also contribute, which results in a
of the phenomenon of collapses and revivals is essentially less regular behavior. The second reason is that the re-
different for the half-excited Dicke state relative to the vivals start to overlap much earlier if the revival time is
fully-excited state or the ground state. Inspecting the halved. The same reason also leads to an additional ir-
eigenvalues (see Table I), we see that the influence of regularity in the case m = −j when the number of atoms
the anharmonic corrections cannot explain this princi- is relatively large and the half-time revivals are not too
pal difference. As expected, the behavior becomes more small.
regular with increase of the mean photon number n̄. For The suppression of the revival amplitude for the
odd numbers of atoms the half-excited Dicke state |j, 0iat half-excited Dicke state is similar to the trapping phe-
does not exist. However, for sufficiently large odd values nomenon which occurs for a single atom prepared in the
of N the Dicke state |j, 21 iat exhibits properties very sim- equally weighted superposition state 2−1/2 (|+i±|−i) (see
ilar to those of the half-excited state. Ref. [14]). In the latter case the population inversion is
In order to explain the peculiar phenomena discussed given by
above, we consider the exactly solvable case N = 2. For
∞
the fully-excited state (m = 1) we find 1X √
hJz i = Pn [cos2 ( n + 1τ )
∞
2 n=0
X 2
hJz i = Pn [(n + 3) − (n + 1) cos(2Ωn,1 τ ) n̄ √ 1
Ω4n,1 + sin2 ( n + 1τ )] − . (3.11)
n=0 n+1 2
+8(n + 1)(n + 2) cos(Ωn,1 τ )] . (3.8)
For large n̄, the Poissonian distribution is sharply peaked
Analogously, we obtain for the ground state (m = −1): around n̄, and the two terms in the sum almost add up
to 1. The remaining oscillating term has the prefactor of
∞
X 2 the order Pn /n, so the amplitude of the Rabi oscillations
hJz i = Pn [(n − 2) − n cos(2Ωn,−1 τ ) is reduced by the factor of the order 1/n̄. As explained
n=0
Ω4n,−1
above, in such a situation the dynamics is less regular
−8n(n − 1) cos(Ωn,−1 τ )] , (3.9) than for the case of initially unexcited or fully excited

5
atom. However, the revival time is not halfed for the 12

single atom in the equally weighted superposition state. 10

10
In the single-atom case the population trapping occurs 8

due to the destructive interference between the contribu- 6

4
tions of the two levels, while for the two-atom half-excited
2 5
Dicke state this phenomenon can be explained by the de- 2 a
0
structive interference between the contributions of the 0 20 40 60 80 100 120 140 160 180 200

two correlated atoms. ξ

12
2
10
IV. SQUEEZING OF THE RADIATION FIELD 10
8
6

The coherent radiation field interacting with atoms 4

5
2
can acquire interesting nonclassical properties such as b
0
sub-Poissonian photon statistics and squeezing. In the 0 20 40 60 80 100 120 140 160 180 200

present paper we focus on the important quantum phe- τ

nomenon of squeezing. The initially coherent cavity field FIG. 4. The squeezing parameter ξ versus the scaled time
can be squeezed when it interacts with a single atom τ = gt for the initial coherent field state with n̄ = 36 in-
[15,16], and squeezing in the revival-time regime can be teracting with N atoms (N = 2, 5, 10) prepared in (a) the
very strong for large intensities of the field [17,18]. Butler fully-excited state |j, jiat and (b) the ground state |j, −jiat .
and Drummond discovered [20] that in the Dicke model
collective atomic effects can improve squeezing obtained As demonstrated in Fig. 4, this long-time behavior is
for short interaction times, compared to the single-atom fully correlated with the collapses and revivals of the Rabi
case. Squeezing in the Dicke model was also considered oscillations. For even values of N , the squeezing param-
recently in a number of works [21–23]. Here we present a eter ξ oscillates achieving its minima at integer multiples
detailed study of squeezing of the initially coherent cavity of τR and maxima at half-odd multiples of τR . On the
field, comparing between the ground, fully-excited and other hand, for odd values of N , we see that ξ has minima
half-excited initial atomic states. √ at both half-odd and integer multiples of τR . This differ-
The field√ quadratures, q = (a† + a)/ 2 and p = ence between even and odd values of N can be explained
i(a† − a)/ 2, satisfy the canonical commutation relation by the fact that squeezing depends on two-photon transi-
[q, p] = i. Then their uncertainties satisfy the Heisenberg tions, when a pair of photons is simultaneously absorbed
relation: ∆q∆p ≥ 1/2, where (∆q)2 = hq 2 i − hqi2 and or emitted by a pair of atoms. The oscillations of ξ decay
similarly for (∆p)2 . For the coherent field state (in par- with time and abandon their regular form. This decay
ticular, for the vacuum
√ state) the uncertainties are equal, is correlated with the loss of regularity in the behavior
∆q0 = ∆p0 = 1/ 2, and an equality is achieved in the of the atomic inversion, caused by the overlaps of neigh-
Heisenberg uncertainty relation. A field state is called boring revivals. The behavior of ξ is rather similar for
squeezed, if the uncertainty of one of √ the quadratures √ is m = j and m = −j, but in the latter case the decay of
below the vacuum level, i.e., ∆q < 1/ 2 or ∆p < 1/ 2. the oscillations and the irregularity occur earlier, as the
Here we consider the squeezing parameter revival time is shorter and the overlaps of revivals begin
earlier. Also, for m = −j the decay rate of the squeezing
∆q √ oscillations increases rapidly with N , while for m = j the
ξ= = 2∆q, (4.1)
∆q0 dependence of the decay rate on N is less pronounced.
Inspecting the minima of ξ in Fig. 4, we clearly see that
and search for ξ < 1 that manifests field squeezing. their period becomes longer for m = j and shorter for
m = −j as N increases, in full accordance with the for-
mula (3.5) for the revival time.
A. Squeezing for the fully-excited and ground states For m = ±j, the minima of ξ on the long-time scale are
above the vacuum level, i.e., ξ > 1. However, squeezing
1. The long-time behavior is achieved at short times, soon after the beginning of the
interaction. In this region the squeezing parameter ex-
We first consider the temporal behavior of the squeez- hibits fast oscillations, with ξ falling below 1. Therefore,
ing parameter ξ on the long-time scale, for the atoms for m = ±j we will focus on the behavior of squeezing
prepared initially in the fully-excited and ground states on the short-time scale.
(m = ±j).

6
2. Squeezing on the short-time scale: The fully-excited state ξm (i.e., the maximum of squeezing), achieved during the
time evolution, depends on N and n̄.
We first consider the case of the initial fully-excited
0.9
atomic state (m = j). Squeezing is achieved for τ < 2
and appears at shorter times for larger values of N . 0.85
2

1.4 0.8 3
a 10
5 4
1.2 0.75

1 2
ξm 0.7
6
0.8
0.65

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
9

ξ 0.6 14
20
30
0.55
3 b 40
2.5
60
0.5
2 2 2.5 3 3.5 4 4.5 5 5.5 6
100
1.5
10 α
1
0.5
FIG. 7. The minimum value ξm of the squeezing parameter
0
versus the coherent amplitude α for various values of N .
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
τ
FIG. 5. The squeezing parameter ξ versus the scaled time
τ = gt for the initial coherent field state with n̄ = 36 inter- 5.5
acting with N atoms prepared in the fully-excited state: (a) 5
a
N = 2, 5, 10, (b) N = 10, 60, 100. 4.5
αmN 4

In Fig. 5 we see the short-time behavior of the squeez- 3.5

ing parameter ξ for n̄ = 36 and various values of N . For 3

2.5
relatively small values of N (N ≤ 10), ξ exhibits quite 5 10 15 20 25 30 35 40

regular oscillations whose amplitude increases with N . N

However, for N ∼ n̄ and larger, the oscillations of ξ be-
come irregular. In Fig. 6 the short-time behavior of ξ is
0.8 b
0.75
shown for N = 14 and various values of n̄. The oscilla-
ξmN 0.65
0.7
tions of ξ become more regular as n̄ increases.
0.6
0.55
2 10 0.5
5 10 15 20 25 30 35 40
15
N
20 FIG. 8. (a) The value αmN of the coherent amplitude,
for which the squeezing parameter is minimized, versus N :
1.5 numerical results (circles) and empirical fitting of Eq. (4.2)
(line); (b) the corresponding value ξmN of the squeezing pa-
ξ rameter versus N : numerical results (circles) and empirical
fitting of Eq. (4.3) (line).

1
√
We first consider the dependence of ξm on α = n̄ for
given N . As shown in Fig. 7, ξm has a minimum as a
function of α, i.e., for given N there exists a value αmN
for which the minimum value ξmN of the squeezing pa-
0.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
rameter is achieved. Figure 8 shows that αmN increases
and ξmN decreases (squeezing improves) as N increases.
τ We found that the dependence of αmN and ξmN on N
FIG. 6. The squeezing parameter ξ versus the scaled time at the considered range can be well approximated by the
τ = gt for the initial coherent field state with n̄ = 10, 15, 20 following empirical formulas:
interacting with 14 atoms prepared in the fully-excited state.
αmN = a − be−cN , (4.2)
It is interesting to investigate how the minimum value ξmN = rN −s , (4.3)

7
where a = 6.21, b = 4.03, c = 0.0471, r = 0.909, s = 0.490. Note that the relations (4.2) and (4.4) are not the
0.156. inverse of each other. This fact can be easily understood,
if one imagine ξm as a two-dimensional function of N and
0.75
2.5 α. When taking a section of ξm along the α axis (i.e.,
for a given N ), one will find the minimum for a certain
0.7 3.0 value of α. However, when fixing this value of α and
going along the N axis, a minimum will be found, in
general, for a different N .
0.65
Finally, we would like to compare our numerical re-
3.5 sults with approximate analytical expressions derived by
Retamal et al. [22] in the strong-field limit n̄ ≫ N . They
ξm
0.6
4.0
found [22] the following expression for the minimum of
4.5 the squeezing parameter, achieved during the time evo-
0.55 5.0 lution,
5.5 1/2
N2

0.5 6.0 N
ξm = 1 − a + b 2 , (4.6)
α 8α
0 20 40 60 80 100 120

N
√
where a = 1/ e ≈ 0.606 and b = 1 + a − a2 − a4 ≈
FIG. 9. The minimum value ξm of the squeezing parameter 1.103. Then, for given α, the squeezing parameter will
versus N for various values of α. be minimized by

100
4a
Nmα = α ≈ 2.2α. (4.7)
80 b
60 0.95
Nmα
40
0.9
20 a 0.85
0 b
2 2.5 3 3.5 4 4.5 5 5.5 6
0.8
α
0.9 0.75

0.8 b ξm 0.7

ξmα 0.7 0.65

0.6 0.6

0.5 0.55
2 2.5 3 3.5 4 4.5 5 5.5 6 a
α
0.5

0.45
FIG. 10. (a) The value Nmα of the number of atoms, for 0 50 100 150
which the squeezing parameter is minimized, versus α: nu- N
merical results (circles) and empirical fitting of Eq. (4.4) FIG. 11. The minimum value ξm of the squeezing parame-
(line); (b) the corresponding value ξmα of the squeezing pa- ter versus N for α = 6: (a) numerical result, (b) approximate
rameter versus α: numerical results (circles) and empirical analytical formula (4.6).
fitting of Eq. (4.5) (line).
In Fig. 11 we compare our numerical results for ξm
We next consider the dependence of ξm on N for given with the approximate formula (4.6) for α = 6 (n̄ = 36).
α. As shown in Fig. 9, ξm has a minimum as a func- A good agreement is found only for very small values
tion of N , i.e., for given α there exists a value Nmα for of N , while the values of Nmα are absolutely different
which the minimum value ξmα of the squeezing parame- (Nmα ≈ 100 for our numerical calculations and Nmα ≈
ter is achieved. Figure 10 shows that Nmα increases and 13 for the analytical approximation). This discrepancy
ξmα decreases (squeezing improves) as α increases. We can be explained by the fact that the numerical values
found again that the dependence of Nmα and ξmα on α of Nmα shown in Fig. 10 do not satisfy the strong-field
at the considered range can be well approximated by the condition n̄ ≫ N . For the considered range of n̄ (≤
following empirical formulas: 36), the approximate solution (4.6) is not valid. From
Nmα = −a + becα , (4.4) Eq. (4.7) one can see that the strong-field condition will
be satisfied for the optimal value Nmα when Nmα /n̄ ≈
ξmα = rα−s , (4.5)
2.2/α ≪ 1. This means that Eq. (4.7) gives a true value
where a = 2.30, b = 1.216, c = 0.743, r = 1.159, s = of Nmα only for n̄ ∼ 104 or more. According to Eq. (4.6),

8
the absolute minimum of the squeezing parameter that
can be achieved in the strong-field limit is ξ ≈ 0.58. For 0.98

n̄ = 36 (the maximum n̄ we considered), the best value 0.96

of squeezing is ξ ≈ 0.49 and it is achieved for Nmα ≈ 100. 0.94

0.92

0.9 38
3. Squeezing on the short-time scale: The ground state
ξm 0.88 30
We next consider the case when the initial atomic state 0.86
24
is the ground state (m = −j). In this case we can dis- 0.84
tinguish three different regimes: the strong-field regime
0.82 20
(n̄ ≫ N ), the weak-field regime (n̄ ≪ N ), and the inter-
mediate regime (n̄ ∼ N ). 0.8

The short-time behavior of the squeezing parameter 0.78 16

ξ in the strong-field regime is shown in Fig. 12 (for 2 4 6 8 10 12 14 16

α = 30 and N = 2, 6, 16). We see that ξ exhibits fast N

regular oscillations, whose frequency is almost indepen- FIG. 13. The minimum value ξm of the squeezing parame-
dent of N (this is just the strong-field Rabi frequency ter versus N for various values of α: numerical results (circles)
(s) p
and linear fitting of Eq. (4.8) (line). The fitting is good in the
Ωn̄,−j = 2 n̄ − N/2 + 1/2, and the dependence on N is
very weak because n̄ is very large), but the amplitude in- strong-field regime n̄ ≫ N .
creases with N . Therefore, the minimum value ξm of the 1.3 1.3
squeezing parameter decreases as N increases. A similar a b
1.2 1.2
behavior is found also for the case m = j in the strong-
1.1 1.1
field regime. An interesting feature of this regime is that
the minimum of ξ is achieved after a relatively large num- 1 1

ber of oscillations (the same time for different values of 0.9 0.9

N ), while in the intermediate regime the minimum of ξ is 0.8 0.8

ξ
0 2 4 6 0 2 4 6
achieved, as a rule, in the first or second oscillation. As
can be seen from Fig. 13, in the strong-field regime the 1.3 1.3
decrease of ξm is linear and can be well approximated by c d
1.2 1.2
the empirical formula
1.1 1.1

ξm = −f (α)N + k, (4.8) 1 1

0.9 0.9
where the slope f (α) is a monotonically decreasing func-
0.8 0.8
tion of α and the free term k ≈ 1 is independent of α. 0 2 4 6 0 2 4 6

τ
1.3 FIG. 14. The squeezing parameter ξ versus the scaled time
1.25 16
τ = gt for the initial coherent field state with α = 1 interact-
ing with N atoms prepared in the ground state: (a) N = 10,
1.2 (b) N = 20, (c) N = 30, (d) N = 40.

1.15
It is well known [27,28,32,33] that in the weak-field
ξ 1.1 regime the behavior of the system is very regular (note
6 that for the case of the fully-excited initial state the weak-
1.05
field regime actually does not exist). In Fig. 14 we see the
1
evolution of the squeezing parameter ξ for relatively short
2 times (for α = 1 and N = 10, 20, 30, 40). The oscillations
0.95 of ξ are quite regular and their frequency increases with
(w)
0.9 N (this is just the weak-field Rabi frequency Ωn̄,−j =
p
2 N − n̄/2 + 1/2), while the amplitude does not change
0.85
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 significantly. The value of squeezing achieved in this limit
τ is rather modest (as a rule, ξ does not decrease below
FIG. 12. The squeezing parameter ξ versus the scaled time 0.85). A perturbative analytical approach to the Dicke
τ = gt for the initial coherent field state with α = 30 inter- model dynamics in the weak-field regime was developed
acting with N atoms (N = 2, 6, 16) prepared in the ground by Kozierowski and co-workers [27,28,32,33] and used for
state. the study of squeezing in Ref. [23].

9
 1
The value Nmα , which gives optimal squeezing for
given α, and the corresponding minimum value ξmα of
0.95
the squeezing parameter are shown in Fig. 16 as functions
0.9
of α. As α increases, squeezing improves (ξmα decreases)
and the minimum occurs at larger Nmα . The dependence
0.85 of Nmα and ξmα on α at the considered range can be well
ξm approximated by the following empirical formulas:
0.8 1
Nmα = kαγ , (4.9)
0.75
1.5 ξmα = x + ye−zα , (4.10)
2
0.7 where k = 1.382, γ = 1.639, x = 0.476, y = 0.420, z =
2.5
3.5 0.259. The largest value of α we considered is 6.5. If we
0.65 3
assume that Eq. (4.10) remains valid for arbitrarily large
5 10 15 20 25 30 35 40
α, then the absolute minimum of squeezing achievable
N with unexcited atoms is ξ ≈ 0.476.
FIG. 15. The minimum value ξm of the squeezing parame-
ter versus N for various values of α.
B. Squeezing for the half-excited state and the
We focus our attention on properties of squeezing in factorization approximation
the intermediate regime, where no analytical approxi-
mation can be used. We study the minimum value ξm , Finally, we discuss the case when the initial atomic
achieved by the squeezing parameter during the time evo- state is the half-excited Dicke state (m = 0). The tem-
lution, for N in the range between 2 and 40 and for α poral behavior of the squeezing parameter ξ on the long-
between 1.0 and 6.5. The results are presented in Fig. time scale is shown in Fig. 17. We see that there is a
15, where ξm is plotted versus N for various values of α. strong correlation between the collapses and revivals of
The typical behavior for given α is as follows. Initially, the atomic inversion and the behavior of squeezing.
ξm decreases slowly with N , but then steeply sinks down
and acquires a minimum at a certain value of N . After 10
the minimum, ξm slightly oscillates and then saturates 10
8
for large N ’s at an almost constant value. The region
6
of linear decrease, occurring in the strong-field regime,
4
appears only for α ≥ 4. This leads to an additional mini- 4
mum at small values of N , but it is less pronounced than 2 2 a
the main minimum at larger N ’s. 0
0 20 40 60 80 100 120 140 160 180 200

ξ
80 a 50
10 6 b
60 40
Nmα 40 30

20 4
20
10
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 2
α
0
0 20 40 60 80 100 120 140 160 180 200

0.8
τ
0.75 FIG. 17. The squeezing parameter ξ versus the scaled time
τ = gt for the initial coherent field state interacting with N
ξ mα 0.7
atoms prepared in the half-excited Dicke state: (a) n̄ = 36,
0.65
N = 2, 4, 10, (b) n̄ = 900, N = 2, 4, 6, 10.
0.6
b
0.55
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 For n̄ = 36 and N = 2, 4, we see in Fig. 17(a) that ξ os-
α cillates with a period of order τR /2 (recall that for m = 0
FIG. 16. (a) The value Nmα of the number of atoms, for we find relatively strong half-time revivals). However, af-
which the squeezing parameter is minimized, versus α: nu- ter a number of periods, the amplitude of the oscillations
merical results (circles) and empirical fitting of Eq. (4.9) decays and ξ becomes nearly constant. This decay is cor-
(line); (b) the corresponding value ξmα of the squeezing pa- related with the loss of regularity in the behavior of the
rameter versus α: numerical results (circles) and empirical atomic inversion that happens when neighboring revivals
fitting of Eq. (4.10) (line). overlap. For larger numbers of atoms, the behavior of the
atomic inversion is very irregular, and so is the behavior

10
of squeezing. In particular, for N = 10, the oscillations obtained for two atoms in the half-excited Dicke state
of ξ decay already after τR /2. The phenomenon of the has the same physical origin as revival-time squeezing
collapses and revivals is quite regular for very large values in the Jaynes-Cummings model [17,18]. This long-time
of n̄. We see in Fig. 17(b) that for n̄ = 900 the squeezing squeezing can be explained by means of the factorization
parameter also behaves very regularly. The structure of approximation for the semiclassical atomic states [19].
the oscillations of ξ becomes more complicated (but keeps This approximation is valid for a strong initial coherent
the regularity) as N increases. For all values of N , the field and for times short compared with τ0 ∼ n̄.
squeezing parameter reaches deep minima at times just In a semiclassical treatment, one replaces the boson op-
before integer multiples of τR /2. The larger the value of erators of the field in the interaction Hamiltonian with
N , the closer is the minimum of ξ to τR /2. (For N ≥ 6, c-numbers. The eigenstates |piat of this semiclassical
we also find additional minima of ξ, but they are not so Hamiltonian (called the semiclassical states) are just the
deep.) eigenstates of the operator J+ +J− = 2Jx with the eigen-
The phenomenon of field squeezing for the half-excited values λp = N − 2p (p = 0, 1, . . . , N ). An atomic state
Dicke state is drastically different from what we found for (and, in particular, a Dicke state) can be expanded in
the fully-excited and ground states. In the former case the basis of the semiclassical states. The factorization
(m = 0), for moderate values of n̄ (α < 30), the min- approximation means that if the atoms are prepared ini-
ima of ξ decrease below 1 (i.e., squeezing occurs) only tially in a semiclassical state, then the total wave function
for N = 2. This behavior is in contrast to the situa- of the system can be approximately written as a product
tion in the latter case (m = ±j), where squeezing can of its field and atomic parts [19,11]:
be achieved for any N (with a proper choice of α) and
is, moreover, enhanced by increasing N . However, just |Ψp (t)i ≈ |Ap (t)iat ⊗ |Φp (t)if . (4.11)
two atoms prepared in the half-excited Dicke state can
produce quite strong squeezing. Using the factorization approximation, one can ana-
In Fig. 18 we see the temporal behavior of ξ for N = 2 lytically estimate mean values of the field operators cor-
and α = 6, 10, 16, 30. The minima of ξ, which occur responding to a specific initial semiclassical atomic state
at times before integer multiples of τR /2, become deeper |piat . In √
particular, one obtains (as usual, we assume
(i.e., squeezing improves) as n̄ increases. It is very impor- that α = n̄ is real):
tant to note that here squeezing is achieved at minima of   
the oscillations on the long-time scale. (Fast oscillations iλp τ 1
haip = α exp − 1− 2
of ξ on the short-time scale, which produce squeezing 2α 4α
for m = ±j, are negligible for the case m = 0.) Conse-
  
2 iλp τ /4α3 iλp τ
quently, for m = 0 the duration of squeezing is essentially × exp α e −1− , (4.12)
4α3
longer than for m = ±j.
  
2 2 iλp τ 1
12 ha ip = α exp − 2− 2
10 2α α
  
2 iλp τ /2α3 iλp τ
10 × exp α e −1− . (4.13)
2α3
6
8
Also, ha† ip and ha†2 ip are given by complex conjugates
ξ of haip and ha2 ip , respectively, while ha† aip = n̄ can
6
be taken constant. Then an approximate expression for
the squeezing parameter can be easily obtained. For the
4 semiclassical state |piat , one finds [22]
h  2 2

2 ξp2 = 1 + 2n̄ e−Tp /8 − e−Tp /16
16 #
30 Tp2  −T 2 /16
 
2
−Tp /8
 λp τ
0
0 20 40 60 80 100 120 140 160 180 200 + e p − 4e cos √
16 n̄
τ √   
FIG. 18. The squeezing parameter ξ versus the scaled time Tp n̄ −Tp2 /8 −Tp2 /16
 λp τ
+ 2e −e sin √
τ = gt for the initial coherent field state with α = 6, 10, 16, 30 2 n̄
interacting with two atoms prepared in the half-excited Dicke 
−Tp2 /16

state: numerical results (solid line) and the factorization ap- +2n̄ 1 − e , (4.14)
proximation (dashed line).
where Tp = λp τ /n̄. It can be seen that squeezing is
Actually, this kind of squeezing on the long-time scale achieved for any semiclassical state, except for those with

11
λp = 0 (then ξp = 1 is constant as long as the factoriza- classical state with λp = 0. Consequently, the squeez-
tion approximation is valid). For λp 6= 0, the oscillations ing behavior of the half-excited Dicke state is very close
of the squeezing parameter achieve minima for times to the behavior of a semiclassical state with |λp | = 2,
while for the fully-excited and ground states squeezing
(p) (p) (p)
τsq ≈ 0.9τR , 1.95τR , . . . , (4.15) is spoiled by the influence of the semiclassical state with
λp = 0. These considerations explain why squeezing on
where the revival time for the state |piat is given by the long-time scale can be achieved only for the half-
√ excited Dicke state |1, 0i. The factorization approxima-
(p) 2π n̄ τR tion also explains the appearance of the minima of the
τR = ≈ . (4.16)
|λp | |λp | squeezing parameter at times close to integer multiples
of τR /2, whose depth increases with n̄. From Fig. 18 we
The minima of ξp become deeper (i.e., squeezing im- see that the factorization approximation describes very
proves) as n̄ increases. well the squeezing behavior of the two-atom half-excited
Using formulas (4.12) and (4.13), one can calculate Dicke state for large values of n̄. As shown in Fig. 19, the
an approximate expression for the squeezing parameter minimum value ξm of the squeezing parameter decreases
for an initial atomic state which is a superposition of monotonically with α. For n̄ > 10, the factorization ap-
the semiclassical states. In particular, we will be inter- proximation is in excellent agreement with the numerical
ested in approximate results for squeezing behavior of the results. Similar results are also obtained in the strong-
Dicke states. First, note that in the Jaynes-Cummings field regime of the Jaynes-Cummings model. For α = 38
model (N = 1) there are just two semiclassical states (the maximum value of α we considered), the best value
with λp = ±1, whose squeezing behavior is the same, of the squeezing parameter is 0.3285. This is much better
for a good degree of accuracy. Consequently, any ini- than the optimal values of squeezing which are obtained
tial atomic state in the Jaynes-Cummings model (and, in for the fully-excited or unexcited atoms when one takes
particular, the ground and excited states) will exhibit the N ∼ 102 .
same squeezing behavior (within the validity of the fac- For atom numbers N ≥ 3, any Dicke state will be
torization approximation). These considerations explain a superposition of the semiclassical states with different
the appearance of significant squeezing in the strong-field values of |λp |. Since the minima of the squeezing parame-
regime of the Jaynes-Cummings model for times near in- ter ξp occur at different times for different |λp |, squeezing
teger multiples of the revival time τR . on the long-time scale will be spoiled for the Dicke states
of three and more atoms. It can be shown [22], that for
1 a Dicke state of N -atom system the condition for the ex-
istence of squeezing in the revival regime is n̄ > (2N )4 .
0.9 Of course, the two-atom half-excited Dicke state is an ex-
ception to this rule, as it involves only two semiclassical
0.8 states with the same value of |λp |. However, the above
condition is in good agreement with our numerical results
0.7
for N ≥ 3.
ξm As we see, the factorization approximation can be very
0.6
useful for explaining many features of the field-atom in-
teraction in the Dicke model. However, this approxima-
0.5
tion fails to predict some interesting phenomena found
by using numerical calculations. In particular, squeez-
0.4
ing on the short-time scale, which clearly dominates for
0.3
the fully-excited and ground atomic states (especially, for
5 10 15 20 25 30 35 large values of N ), is not predicted by the factorization
α approximation. We also observe that the suppression of
FIG. 19. The minimum value ξm of the squeezing parame- the revival amplitude for the half-excited Dicke state can-
ter versus the coherent amplitude α for two atoms prepared not be described within this approximation. This can
in the half-excited Dicke state: numerical results (circles) and be readily understood by recalling the exact expression
the factorization approximation (line). (3.10) describing the evolution of the population inver-
sion for the half-excited state |1, 0i. The amplitude of
In the case of the Dicke model with two atoms, there the Rabi oscillations here is of the order of 1/n̄, which is
are three semiclassical states; two of these states (with neglected in the factorization approximation.
λp = ±2) lead to squeezing, while for the third state
(with λp = 0) no squeezing is found (ξp = 1 is constant).
The half-excited Dicke state |1, 0i is a superposition of
the two semiclassical states with λp = ±2, while the fully-
excited and ground states also include the third semi-

12
 V. CONCLUSIONS [10] G. Drobný and I. Jex, Opt. Commun. 102, 141 (1993).
[11] S. M. Chumakov, A. B. Klimov, and J. J. Sánchez-
In this paper we considered in detail properties of the Mondragón, Phys. Rev. A 49, 4972 (1994).
system of N two-level atoms interacting with a single- [12] S. M. Chumakov and M. Kozierowski, Quantum Semi-
class. Opt. 8, 775 (1996).
mode cavity field (the Dicke model). When the field is
[13] H. M. Castro-Beltran, S. M. Chumakov, and J. J.
initially in the coherent state with a sufficiently large
Sánchez-Mondragón, Opt. Commun. 129, 184 (1996).
mean photon number, the dynamics of the system is quite
[14] K. Zaheer and M. S. Zubairy, Phys. Rev. A 39, 2000
regular, and the Rabi oscillations of the atomic inversion (1989).
exhibit an interesting quantum phenomenon of collapses [15] P. Meystre and M. S. Zubairy, Phys. Lett. 89A, 390
and revivals. We studied how this phenomenon is influ- (1982).
enced by collective atomic effects. The main conclusion [16] A. S. Shumovsky, Fam Le Kien, and E. I. Aliskenderov,
is that the role of the collective effects is determined by Phys. Lett. A 124, 351 (1987).
the initial atomic state. We found that by preparing [17] J. R. Kukliński and J. L. Madajczyk, Phys. Rev. A 37,
just two atoms in the half-excited Dicke state one can 3175 (1988).
cause greater effect on the behavior of the system than [18] C. W. Woods and J. Gea-Banacloche, J. Mod. Opt. 40,
by collecting tens or even hundreds of excited or unex- 2361 (1993).
cited atoms. In the phenomenon of the collapses and [19] J. Gea-Banacloche, Phys. Rev. A 44, 5913 (1991).
revivals, the half-excited Dicke state causes two basic ef- [20] M. Butler and P. D. Drummond, Opt. Acta 33, 1 (1986).
fects: the revival amplitude is strongly suppressed (anal- [21] J. Seke, Quantum Semiclass. Opt. 7, 161 (1995); Physica
ogously to the trapping phenomenon for a single atom in A 213, 587 (1995); 240, 635 (1997).
the equally weighted superposition state) and the revival [22] J. C. Retamal, C. Saavedra, A. B. Klimov, and S. M.
time is halved. The two-atom half-excited Dicke state Chumakov, Phys. Rev. A 55, 2413 (1997).
also leads to very interesting squeezing behavior. It is the [23] M. Kozierowski and S. M. Chumakov, (unpublished).
only Dicke state for N ≥ 2 which exhibits strong squeez- [24] F.-L. Li, X.-S. Li, D. L. Lin, and T. F. George, Phys.
ing on the long-time scale, similarly to the behavior found Rev. A 41, 2712 (1990).
in the strong-field regime of the Jaynes-Cummings model. [25] D. J. Wineland, J. J. Bollinger, W. M. Itano, and D. J.
Heinzen, Phys. Rev. A 50, 67 (1994).
[26] I. R. Senitzky, Phys. Rev. A 3, 421 (1971).
[27] M. Kozierowski, S. M. Chumakov, J. Światlowski, and
ACKNOWLEDGMENTS
A. A. Mamedov, Phys. Rev. A 46, 7220 (1992).
[28] M. Kozierowski, S. M. Chumakov, and A. A. Mamedov,
G.R. and C.B. gratefully acknowledge the financial J. Mod. Opt. 40, 453 (1993).
help from the Technion. A.M. was supported by the [29] R. Bonifacio and G. Preparata, Phys. Rev. A 2, 336
Fund for Promotion of Research at the Technion and by (1970).
the Technion VPR Fund. [30] S. Kumar and C. L. Mehta, Phys. Rev. A 21, 1573 (1980);
24, 1460 (1981).
[31] J. Katriel and D. G. Hummer, J. Phys. A 14, 1211 (1981).
[32] M. Kozierowski, A. A. Mamedov, and S. M. Chumakov,
∗ Phys. Rev. A 42, 1762 (1990).
E-mail: ramon@physics.technion.ac.il [33] M. Kozierowski, S. M. Chumakov, and A. A. Mamedov,
†
E-mail: costya@physics.technion.ac.il Physica A 180, 435 (1992).
‡
E-mail: ady@physics.technion.ac.il
[34] G. J. Milburn, Opt. Acta 31, 671 (1984).
[1] R. Dicke, Phys. Rev. 93, 99 (1954).
[2] C. Leonardi, F. Persico, and G. Vetri, Riv. Nuovo Ci-
mento 9, 1 (1986). TABLE I. The ratio f51 = h5 /h1 , that represents the
[3] M. Tavis and F. W. Cummings, Phys. Rev. 170, 379 anharmonicity of the Hamiltonian eigenvalues, for n = 36,
(1968). N = 10, 20, 30 and m = ±j, 0.
[4] D. F. Walls and R. Barakat, Phys. Rev. A 1, 446 (1970).
[5] E. T. Jaynes and F. W. Cummings, Proc. IEEE 51, 89 N m f51
(1963). 10 5.0219
[6] L. Knoll and D. G. Welsch, Prog. Quantum Electron. 16, 20 j 5.0184
135 (1992). 30 5.0159
[7] B. W. Shore and P. L. Knight, J. Mod. Opt. 40, 1195 10 5.0382
(1993). 20 −j 5.0638
[8] J. H. Eberly, N. B. Narozhny, and J. J. Sánchez- 30 5.1852
Mondragón, Phys. Rev. Lett. 44, 1323 (1980); N. B. 10 5.0283
Narozhny, J. J. Sánchez-Mondragón, and J. H. Eberly, 20 0 5.0309
Phys. Rev. A 23, 236 (1981). 30 5.0360
[9] S. M. Barnett and P. L. Knight, Opt. Acta 31, 435
(1984); 31, 1203 (1984).

13