You are on page 1of 10

15 August 1999

Optics Communications 167 1999. 115124 www.elsevier.comrlocateroptcom

Quantum-phase properties of the Kerr couplers

J. Fiurasek a , J. Krepelka b, J. Perina


a Department of Optics, Palacky Uniersity, 17 Listopadu 50, 772 07 Olomouc, Czech Republic Joint Laboratory of Optics, Palacky Uniersity and Institute of Physics of Academy of Sciences of the Czech Republic, 17 Listopadu 50, 772 07 Olomouc, Czech Republic

Received 2 March 1999; accepted 3 June 1999

Abstract We use the concept of phase space and Husimi quasidistribution to derive joint-phase probability distribution and quantum-phase properties for the Kerr couplers. The exact numerical as well as approximate analytical solutions of the Schrodinger equation are found. The spatial development of the single-mode phase distributions and phase-difference distribution is demonstrated. The Fourier coefficients of the phase distributions are introduced and employed to describe quantum-phase behaviour. It is shown that the phase-difference evolution is closely connected to an energy exchange between two waveguides, which form the coupler. The collapses and revivals of the mean photon number oscillations are due to the bifurcation of the phase-difference probability distribution, which has a two-fold symmetry in the interval of collapse. q 1999 Elsevier Science B.V. All rights reserved.
PACS: 42.50.Dv Keywords: Quantum phase; Non-linear coupler; Kerr effect

1. Introduction The aim of this paper is to study phase properties of the optical modes propagating in the Kerr non-linear coupler. This non-linear optical device consists of two parallel waveguides whose modes are mutually linearly coupled and each mode is non-linearly coupled to itself by means of the Kerr effect. We also include the non-linear cross-interaction. The

Kerr coupler can be described with the help of the interaction momentum operator w1,2x

G int s " k ab q a b . q "ga 2 a 2 q "gb 2 b 2

q "gaab b .


Corresponding author. Fax: 420-68-5225246; e-mail:

Here a a . and b b . are annihilation creation. operators of the modes in the first and second waveguides in the interaction picture, k is the linear coupling constant between the two modes, g and g are non-linear coupling constants proportional to the third-order susceptibility characterizing the strength of the self-action and cross-action processes, respec-

0030-4018r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved. PII: S 0 0 3 0 - 4 0 1 8 9 9 . 0 0 2 8 6 - 2


J. Fiurasek et al.r Optics Communications 167 (1999) 115124

tively. It is assumed that the propagation constants of both modes are the same. Analysis of such a device, based on a classical wave optical description, reveals that the coupler exhibits several different regimes of operation in dependence on the values of linear and non-linear coupling constants and input intensities w35x. Also the soliton propagation in the Kerr couplers has been investigated and regimes of bifurcations were found and discussed w6,7x. In standard considerations the energy is injected into the first waveguide and its transmission to the second waveguide is analyzed. A certain threshold value of the input intensity exists; the energy is periodically fully exchanged between the waveguides below the threshold; the intensities are asymptotically uniformly distributed at the threshold; and, finally, when the input intensity exceeds the threshold value, the non-linear effect locks the energy in the waveguide, where it was initially injected, and the exchange of the energy decreases. The equivalent quantum model has been discussed in Ref. w1x. The Schrodinger equation in the basis of Fock states was solved numerically and the evolution of mean numbers of photons for initial Fock and coherent states was studied. The quantum Kerr coupler behaves quite differently from its classical counterpart for small input intensities. Particularly, collapses and revivals of the photon number oscillations take place if the coupler operates below the threshold. To explain such a behaviour, an approximate analytical solution was found for the initial Fock state. This solution is based on a transformation of the basis new annihilation operators defined as linear combination of the old ones are introduced. and neglection of the part of the momentum operator leading to its diagonalization in the new basis. This idea was fully unfolded in Ref. w2x, where the evolution of the annihilation operators in the Heisenberg picture was obtained in analytical approximation. Assuming initial coherent states, the analytical expressions for mean photon numbers, principal squeeze variance and integrated intensity variance were derived and a possibility of generation of non-classical states of light was discussed. The influence of the z-dependence of the linear coupling constant on the switching properties of the coupler was also investigated within the framework of these

approximate analytical expressions w8x. The analytical results give us important general information about the behaviour of the coupler. However, these predictions fail for long distance z or high values of non-linear coupling constants. In those cases, the numerical calculation has to be employed. In contrast to the classical description, the transition from operation below the threshold to the above-threshold regime is smoothed in the quantum model when only small numbers of input photons are assumed. The quantum dynamics of the coupler is determined by a countable set of the eigenvalues of the momentum operator and effectively only a finite number of them contributes substantially to the evolution of the state vector <C :. Nevertheless, the notion of the threshold is still useful because, with increasing g, the behaviour of the coupler substantially changes w1x. One can use classical equations to determine the threshold w3x. When all energy is initially injected into the first waveguide b s 0., then the operation of the coupler depends on an effective non-linearity parameter h s <2 g y g < < a < 2rk . The classical threshold is reached at h s 4. We are below the threshold for h - 4 and above the threshold for h ) 4. Notice that for g s 2 g the parameter h s 0 regardless of the value of a and k and the coupler behaves like a linear one. We have fixed a s 2, b s 0 and k s 1 in our calculations and changed g and g to obtain various regimes of the coupler operation. In this paper we investigate the phase properties of the modes propagating in the coupler. The theory of quantum phase has attracted much attention during past years when several approaches to the quantum phase were proposed w912x for a review, see Ref. w13x.. Quantum-phase properties of elliptically polarized light propagating in a Kerr medium have been investigated within the framework of the PeggBarnett Hermitian phase formalism w1416x. In the present work we use the concept of a phasespace approach w17x, which has been recently applied to the investigation of phase properties of contradirectional non-linear coupler operating by means of degenerate parametric down-conversion w18x. We calculate the phase distribution P w . from the Husimi quasidistribution FA a , a ) . related to the antinormal ordering of field operators. We show that the function P w . can give us an important informa-

J. Fiurasek et al.r Optics Communications 167 (1999) 115124


tion about the quantum state of the mode and about the behaviour of the quasidistribution. Particularly, several peaks of this function signalize quite complex form of the quasidistributions having also several peaks, as will be shown below.

help of two-mode density matrix elements rm 1 m 2 , n1 n 2 s m1 ,m 2 < r < n1 ,n 2 : similarly as the single-mode quasidistribution 2.. The direct integration leads to P w1 , w2 . s 1

2 p .

q1 G

rm 1 m 2 , n1 n 2
m2 q n2 2

m1 , n 1s0 m 2 , n 2s0

2. Phase-space approach to the quantum phase = 2.1. Phase distributions The single-mode Husimi quasidistribution may be expressed in terms of single-mode density matrix r ,

m1 q n1


m1 ! n1 ! m 2 ! n 2 ! .
=e i n1ym 1 . w 1 e i n 2ym 2 . w 2 .


FA a , a . s

1 p

m, ns0

rm n a ) ma n

'm! n!

ey< a < .

Further we define phase difference D w s w 1 y w 2 whose probability distribution can be easily determined, P Dw . s


Here rm n s m < r < n: are density matrix elements in Fock basis. The FA function is a real non-negative function normalized to the unity. These properties are transferred to the phase distribution. The definition of P w . is w17x

H0 H0



P w1 , w2 .

=d D w y w 1 q w 2 . d w 1 d w 2 ,


where d is the Dirac delta function. Hence we have P Dw . 1 s 2p

` `

Pw. s

H0 F

< a <e i w , < a <eyi w . < a < d < a < ,

H02 p P w .


q1 G

rm 1 m 2 , n1 m 2qm 1yn1 .
2 m 2 q m1 y n1 2

and it holds that P w . 0 0 and d w s 1. The substitution of 2. into 3. and the integration over < a < yield 1 2p

m1 , n 1s0 m 2smax 0, n 1ym 1 .


m1 q n1 2



G rm n

Pw. s

m, ns0


q1 2 e i nym. w . 'm! n!

m1 ! n1 ! m 2 ! m 2 q m1 y n1 . ! =e i n1ym 1 .D w .


In order to calculate P w . we have to determine the z-dependent density matrix r z . of the single modes. We can also introduce the two-mode joint-phase probability distribution, defined in terms of two-mode FA -function
` `

2.2. Phase dispersions and Fourier coefficients We conclude this section with some remarks about the phase-distribution variance. The usual definition s 2 s w y w :. 2 : is not suitable for our purposes, because P w . is a periodic function and the variance depends on the choice of the integration interval. Thus the so-called phase dispersion has been introduced in Ref. w19x, using the mean values of periodic functions sin w and cos w ,

P w1 , w2 . s

H0 H0 F

a 1 , a 1) ; a 2 , a 2) .

=< a 1 < < a 2 < d < a 1 < d < a 2 < ,

where a j s < a j <expi w j ., j s 1,2. The two-mode Husimi quasidistribution can be expressed with the

s 2 s 1 y sin w :2 y cos w :2 s 1 y <e i w :< 2 .



J. Fiurasek et al.r Optics Communications 167 (1999) 115124

This expression is suitable for calculations. The mean value expi w .: can be calculated with the help of the general formula 4. for the phase distribution P w ., yielding

formally equivalent, if we define an effective density matrix for the latter case,

rm n s
e i w : s

rm s, n sqmyn.

2 sqmyn 2



rnq 1, n G n q
n!'n q 1

3 2

s! s q m y n . !


10 .

m0n ;

rm n s rn m , )
Using 9. we can easily interpret the meaning of the phase dispersion. The quantity <expi w .:< represents an amplitude of one Fourier coefficient in the spectral Fourier decomposition of the phase-probability distribution. Thus the quantity s provides information about the appearance of the fundamental frequency in the spectral decomposition of the phaseprobability distribution P w .. The fundamental component induces one peak maximum. of the function P w .. Similarly the higher harmonics support n peaks of the probability distribution. We can generalize the definition of s and introduce higher-order dispersions s k .,

m-n .

13 .

This effective density matrix may be inserted in 12., which yields the required Fourier coefficients of the phase-difference distribution.

3. Numerical and analytical solutions of the Schrodinger equation We assume only pure states. The state vector <C z .: is a solution of the Schrodinger equation, yi " d dz

<C z . : s G <C z . :

14 .

s 2 k . s 1 y <e i k w :< 2 ,

kgN .

11 .

and may be expressed in the basis of the Fock states < n1 ,n 2 : ' < n1 :< n 2 :,

They allow us to study the appearance or disappearance of n-fold symmetry in P w . and FA a , a ) . functions. When s n. decreases the n-fold modulation of the distributions is more pronounced and vice versa. Therefore it is useful to study directly the spatial or temporal development of the Fourier coefficients F k . s <expi k w .:<. A simple generalization of the expression 10. yields 2nqk 2

<C z . : s

n1 , n 2s0

c n1 n 2 z . < n1 ,n 2 : ,

c n1 n 2 z . s n1 ,n 2 <C z . : .

15 .

rnq k , n G

F k. s



n! n q k . !

12 .

The density operator can be written as r s <C : C < and its matrix elements are r m 1 m 2 , n 1 n 2 z . s ) c m 1 m 2 z . c n1 n 2 z .. The solution 15. of Eq. 14. can be easily found numerically. This was described in detail in Ref. w1x and we point out only the most important facts. The coefficients c n1 n 2 z . have to be calculated. The Hilbert space of all states splits into a direct sum of invariant subspaces due to the conservation of the total number of photons operator n s aa q b b commutes with the momentum operator.,
` H s [ns 0 Hn ,

The Fourier coefficients of the phase-difference distribution can be introduced in a similar way. A closer inspection of the formulas 4. and 8. reveals that the expressions for the single-mode phase distribution P w . and phase-difference distribution P D w . are

16 .

the dimension of each subspace Hn being n q 1 and the basis can be chosen in the form < j,n y j :, j s 0, . . . ,n. The Schrodinger equation is rewritten into

J. Fiurasek et al.r Optics Communications 167 (1999) 115124


matrix form for each subspace, the eigenvalues l n j and eigenvectors <Cn j : of the corresponding matrices Gn ,

G n . j jX s j,n y j < G < jX ,n y jX : ,


j, jX s 0, . . . ,n ,

These states differ from the Fock states < n1 ,n 2 :. To clarify the notation, we use N for the states defined in 23.. An inversion of the definition 22. and substitution into 1. give the momentum operator expressed in terms of the operators A, B,

17 .
are found and the initial conditions are applied to obtain the coefficients c n1 n 2 z .. Defining vectors < Cn z . : s c j, nyj z .

G s " k A A y B B .
" q 4

2 g q g . A 2A 2 q B 2 B 2 .

' c 0, n z . , . . . ,c n ,0 z .

18 .
we can write

q 2 "gA AB B
" q 4

2 g y g . A 2 B 2 q B 2A 2 . .

24 .

< Cn z . : s

ei l


Cn j < Cn 0 . :<Cn j : .

19 .

If initial coherent states are assumed with complex amplitudes a , b , then

a n 1b n 2
c n 1 n 2 0 . s

(n ! n !
1 2

exp y

< a <2q< b <2 2

20 .

The density matrices of single modes can be expressed as traces over the other mode,

To obtain the analytical results we omit the rotating term proportional to A 2 B 2 q B 2A 2 in the momentum operator 24.. This approximation is equivalent to the neglection of corresponding rotating terms in the Heisenberg equations for the operators A, B, which was done in Ref. w2x and the two approaches give identical results. The neglection of the above two terms leads to the diagonalization of the momentum operator G in the basis < j, N y j :, corresponding eigenvalues being

r 1, m n z . s r 2, m n z . s

) cm r z . cn r z . ,

LN j s k 2 j y N . q 1 2 g q g . N N y 1. 4
1 q 2 2 g y g . j N y j . .

rs0 `

25 .

c r m z . c r)n z . .

21 .

We can also employ the approximate analytical solutions to obtain the analytical expressions for the coefficients c n1 n 2 z .. Following Ref. w1x we introduce new operators

Now we can write the solution in this new basis and transform it back to the former basis to obtain the coefficients c n1 n 2 z .. We have
` N
Nj z

<C z . : s

ei L
Ns0 js0

d N j < j, N y j : ,



a qb. , B s

'2 a y b . .

22 .

d N j s j, N y j <C 0 . : .

26 .

These new operators obey the same commutation relations as a and b. We define Fock states in terms of them < j, N y j : s j N j A j B Nyj <0,0: ,

We assume initial coherent states in the following considerations, i.e. <C 0.: s < a , b :. With the help of the definition 23. we arrive at d N j s j N j 2yN r2 a q b . a y b .
j Nyj

jNj s

(j! N y j . !

23 .

=exp y

< a <2q< b <2 2

27 .


J. Fiurasek et al.r Optics Communications 167 (1999) 115124

A straightforward calculation based on the definition of c n1 n 2 z . in Eq. 15. yields c n 1 n 2 z . s

n1 ! n 2 ! 2 n1qn 2

based on numerical results. The approximate formulas are generally useful only for small non-linearities and short distances z. However, they become exact when g s 2 g and this special case will be discussed in detail. 4.1. Single-mode phase distributions We will confine ourselves to the first mode because the behaviour of the second mode is similar. To investigate spatial evolution of the distribution P1 w ., we can employ Fourier coefficients F1 k, z . as a powerful tool subscript refers to the first mode.. They are in close relation with the phase dispersions s 2 k . according to 11.; if a function F k, z . decreases, then the corresponding dispersion s 2 k . increases and vice versa. It is sufficient to deal with the first four or five coefficients which comprise almost all information about the phase distribution. For the coherent state < j :, P w . has one peak at w s arg j . and the coefficient F11, z . prevails. The spatial development of F1 k, z . for two different values of non-linear parameters is shown in Fig. 1. It is useful to remember the results obtained for the light propagation in Kerr media w1416x. If two polarization modes are taken into account, the description is equivalent to our model without linear coupling, i.e. with k s 0. The analytical results, obtained in this case, reveal that the spatial evolution can be periodic this takes place if grg can be expressed as a fraction of integers. and that the state of the field becomes a discrete superposition of finite number of coherent states for t s MTrN, T is a period and M, N are primes. The appearance of N-fold structure can be observed in Fig. 1 and it corresponds to peaks of F1 N, z .. Simultaneously N peaks also appear in the phase distribution P1 w . and quasidistribution FA ,1 a , a ) . see Figs. 2 and 3.. It is obvious from Fig. 1a that the linear coupling constant k breaks the periodicity down. The peak of F11, z . in Fig. 1a at z f 20 p is connected with the revival of the energy exchange between the waveguides which will be discussed in greater detail in the next subsection. The structure of joint-phase distribution P w 1 , w 2 . can be quite complex. Particularly the N-fold symmetry is exhibited there clearly. The total number of peaks may differ from N when

n1qn 2


exp i L n1qn 2 , j z . y1 .

n 2 yj

=j n1qn 2 , j d n1qn 2 , j R n1 n 2 j ,
min j, n 1 .

R n1 n 2 j s

ksmax 0, jyn 2 .


j k

n1 q n 2 y j . n1 y k

28 .
Of course, we cannot sum over the whole Hilbert space, but we must choose the upper limit of the sum, n1 q n 2 s Nmax . The same holds for numerical solutions discussed above. We must choose Nmax so that the Hilbert subspace is large enough to contain all important information about the process under discussion. The upper limit depends significantly on the total mean number of photons n s n:, for our calculations n s 4 and Nmax s 20 has proved to be sufficient. Both numerical and analytical calculations permit the independent tests of their validity because a mean photon number in each Hilbert subspace Hn should be conserved during the evolution.

4. Discussion The same initial conditions were chosen in Refs. w1,2x to investigate a spatial evolution of the mean photon numbers and this allows us to use the results obtained there. In the first part of the discussion we analyze the spatial evolution of the single-mode phase distributions and in the second part we focus ourselves on the spatial evolution of the phase difference D w . We show that the latter is closely associated with the spatial development of the mean photon numbers and that the phase distributions can shed some light on the processes arising from simultaneous influence of linear coupling and non-linear self- and cross-phase modulations. The discussion is

J. Fiurasek et al.r Optics Communications 167 (1999) 115124


Fig. 2. Spatial evolution of the phase-probability distribution P1 w , z . of the first mode; the parameters are the same as in Fig. 1b.

Fig. 1. Spatial evolution of the Fourier coefficients F1 k, z . of the first-mode phase distribution. The linear coupling constant k s1; a. g s 0.1, g s 0; b. g s 0.1, g s 0.2. The first mode is initially in the coherent state with complex amplitude a s 2 and the second mode is initially in vacuum state, b s 0.

energy is entirely periodically transferred between the waveguides with the period Z2 w2x. This follows from the analytical results presented in the previous section, which are exactly valid for g s 2 g. Notice that F1 k, z . has k peaks in the interval 0,Z1 .. However, some peaks are split into two shifted peaks. This is a consequence of the energy exchange

g s 2 g to N 2 when g s 0 w15x. An example is given in Fig. 4. The distance z s 20.7 was chosen to coincide with the peak of F12, z . in Fig. 1a. Indeed, the two-fold structure is present. Note that due to the influence of linear coupling the function P w 1 , w 2 . is not completely symmetric. In a special case when g s 2 g the periodic be haviour of the spatial evolution can be restored, if a relative value grk can be expressed as a quotient of two integers. This is shown in Figs. 1 and 2. We can see the evolution arising from the non-linear phase modulation with the period Z s prg, which is superimposed with oscillations induced by k . This holds even if grk is not a ratio of two integers. Then beats between the two oscillations having periods Z1 s prg and Z2 s prk appear and the initial state is never fully reconstructed. It should be pointed out that in this case the energy exchange between the waveguides is equivalent to the linear waveguide, the

Fig. 3. Quasidistribution FA ,1 a , a ) . for the first mode; the parameters are the same as in Fig. 2 and z s10. This choice of z corresponds to the peak of F13, z . in Fig. 1b and it slightly differs from the value p r3 g . s10 p r3. This shift is induced by linear coupling. It holds that n110.: ) n110 p r3.: and thus the 3-fold modulation is more pronounced for z s10.


J. Fiurasek et al.r Optics Communications 167 (1999) 115124

first and second modes is "pr2. The function P D w . reflects it as a peak at pr2 or 3pr2. The spatial development of P D w . can again be extracted using appropriate Fourier coefficients. Their spatial evolution for two different cases is shown in Fig. 5. In Fig. 5a we can see the spatial evolution of the Fourier coefficients F k, z . when only self-phase modulation is present and g s 0. It is obvious that with increasing z the coefficient F 2, z . becomes dominant and the two-fold symmetry of the phasedifference distribution is established. This bifurcation implies the appearance of collapse of the energy exchange between the waveguides when n1 : f n 2 : f n:r2 for a quite long interval of z w1,2x. Notice that a similar bifurcation of single-mode phase distribution has been obtained for highly squeezed states w20,18x. Finally, the revival takes place and the energy exchange is re-established. However, it is weaker and only a part of the energy is periodically exchanged.

Fig. 4. Phase distribution P w 1 , w 2 . for z s 20.7. a. 3-D view. b. Topographic plot. The parameters are the same as in Fig. 1a.

between the waveguides. The peaks are most pronounced if the amplitude is maximal, the missing peaks correspond to the zero amplitude vacuum state. of the first mode. If g ) 2 g, then the periodic ity of the spatial evolution disappears again and the behaviour is similar to g - 2 g. 4.2. Phase-difference distribution The relative phase D w whose distribution is defined by 8., is closely associated with a spatial behaviour of the mean photon numbers in two modes of the coupler. The condition for an optimal energy exchange between the waveguides is that the phase difference between the complex amplitudes of the

Fig. 5. Spatial evolution of the Fourier coefficients F k, z . of the phase-difference distribution. All parameters are the same as in Fig. 1a,b.

J. Fiurasek et al.r Optics Communications 167 (1999) 115124


The phase-difference distribution P D w . can be seen in Fig. 6. For z s 0 it is just a straight line, P D w . s 1r2 p, because the second mode is initially in a vacuum state. For z s pr4 the distribution has one peak at D w s 3pr2 and for z s 3pr4 it has one peak at D w s pr2, which corresponds to the almost complete periodic energy exchange for small z; z s 10 p holds in the region of the collapse of the oscillations. The distribution has two-fold symmetry, the position of the peaks being pr2 and 3pr2. We move to the revival area for z s 20 p. We

can see that the distribution has now only one peak but it is flatter, because the revival is only partial. The increase of g towards the threshold value leads to a qualitative change of the spatial development of all the assumed quantities. The coefficient F 1, z . becomes dominant for all z. Thus no collapses and revivals appear but oscillations take place. With a further increase of g above the threshold the energy becomes trapped in the waveguide where it was initially injected and thus the amplitude of the energy exchange decreases. This is reflected in the phase-difference distribution, with increasing g the average values of F k, z . decrease and tend to zero, which signalizes the flattering of phase-difference distribution. This behaviour takes place not only when g - 2 g but also for g ) 2 g. The case g s 2 g regardless of the absolute value of these constants. forms an exception as it can be seen in Fig. 5b. The evolution of P D w . is periodic, the period being Z2 s prk . As a matter of fact the phase difference behaves in the same way as if the coupler was linear and the two modes were in the coherent states. This result directly follows from the analytical solution. If g s 2 g, then the momentum operator eigenvalues L N j defined in 25. are simplified and we have

LN j s k 2 j y N . q g N 2 y N . .

29 .

This expression is precise. The first term in the sum corresponds to the linear coupler. The second term, which is a function of N only, results in a non-linear phase modulation of the single modes but is cancelled in the formula 8. for P D w ..

5. Conclusions We have investigated quantum-phase properties of optical fields propagating in the Kerr couplers. We have employed the phase-space approach working with the Husimi quasidistribution related to antinormal ordering of field operators. The numerical and approximate analytical solutions of the corresponding Schrodinger equation, describing quantum dynamics of the coupler, were found. Fourier coefficients of the phase-probability distribution were in-

Fig. 6. 3-D plot of a spatial evolution of the phase-difference distribution P D w , z . a. and 2-D plot of P D w . for z s 0 ., z s p r4 `., z s 3p r4 I., z s10 p ^. and z s 20 p e. b.. The parameters are the same as in Fig. 1a.


J. Fiurasek et al.r Optics Communications 167 (1999) 115124

troduced and utilized to examine the spatial development of the phase distributions. It was shown that the mutual interaction of linear coupling and non-linear phase modulation destroy the periodicity of the spatial development, which can take place for k s 0. However, such a periodicity is re-established if nonlinear self- and cross-interactions compensate each other, i.e. g s 2 g. The peaks of the Fourier coefficients F k, z . indicate a k-fold symmetry of the phase distribution and the corresponding single-mode anti-normal quasidistribution. The two-fold symmetry of the phase-difference distribution i.e. a phase-difference bifurcation. is responsible for the collapse of photon-number oscillations. If the above mentioned compensation of non-linear interactions takes place, the phase difference behaves like that of a linear coupler with the same coupling constant k and input coherent states. Thus the energy is periodically exchanged between the two modes of the coupler in this case. Acknowledgements This work was partly supported by Grant No. VS96028 of the Czech Ministry of Education.

w1x w2x w3x w4x w5x w6x w7x w8x w9x w10x w11x w12x w13x w14x w15x w16x w17x A. Chefles, S.M. Barnett, J. Mod. Opt. 43 1996. 709. N. Korolkova, J. Perina, Opt. Commun. 136 1996. 135. S.M. Jensen, IEEE J. Quantum Electron. QE 18 1982. 1580. A. Ankiewicz, J. Opt. Quantum Electron. 20 1988. 329. A.W. Snyder, D.J. Mitchell, L. Poladian, D.R. Rowland, Y. Chen, J. Opt. Soc. Am. B 8 1991. 2102. B.A. Malomed, I.M. Skinner, P.L. Chu, G.D. Peng, Phys. Rev. E 53 1996. 4084. B.A. Malomed, I.M. Skinner, R.S. Tasgal, Opt. Commun. 139 1997. 247. N. Korolkova, J. Perina, J. Mod. Opt. 44 1997. 1525. D.T. Pegg, S.M. Barnett, Europhys. Lett. 6 1988. 483. S.M. Barnett, B.J. Dalton, Phys. Scr. T 48 1993. 13. J. Noh, J.W. Fougeres, L. Mandel, Phys. Scr. T 48 1993. 29. A. Luks, V. Perinova, Phys. Scr. T 48 1993. 94. V. Perinova, A. Luks, J. Perina, Phase in Optics, World Scientific, Singapore, 1998. Ts. Gantsog, R. Tanas, J. Mod. Opt. 38 1991. 1537. Ts. Gantsog, R. Tanas, Quantum Opt. 3 1991. 33. R. Tanas, Ts. Gantsog, J. Opt. Soc. Am. B 8 1991. 2505. R. Tanas, A. Miranowicz, Ts. Gantsog, in: E. Wolf Ed.., Progress in Optics, vol. 35, Elsevier, Amsterdam, 1996, p. 355. L. Mista Jr., J. Rehacek, J. Perina, J. Mod. Opt. 45 1998. 2269. A. Brandilla, H. Paul, Ann. Phys. Leipzig 23 1969. 323. W. Schleich, R.J. Horowicz, S. Varro, Phys. Rev. A 40 1989. 7405.

w18x w19x w20x