PHYSICAL REVIEW A VOLUME 43, NUMBER 1 1 JANUARY 1991

Resonators in quantum optics: A first-principles approach

L. Knoll
Sektion Physik, Friedrich-Schiller- Universitat Jena, Max-8'ien-Platz 1, Jena, 0-6900, Germany

W. Vogel
Sektion Mathematik/Physi k, Padagogische Hochschule "Liselotte Herrmann, "
Gustrom, 0-2600, Germany

D.-G. Welsch
Sektion Physik, Friedrich-Schiller-UniUersitat Jena, Max-8'ien-Platz 1, Jena, 0-6900, Germany
(Received 9 March 1990)
Input-output relations in resonators are studied on the basis of a recently developed field-
theoretical approach to the action of passive systems in quantum optics [Knoll, Vogel, and Welsch,
Phys. Rev. A 36, 3803 (1987)]. This method allows the rigorous proof of recently adopted quantum
stochastic approaches, including the presence of sources that form the active medium in the cavity.
Quantum Langevin equations are obtained in a coarse-graining approximation. Moreover, we
derive all required commutation relations that have been postulated so far. Correlation functions of
the field quantities measured outside the cavity are related to correlation functions of field operators
of the intracavity field and the incoming field.

I. INTRODUCTION mirrors of a resonator cavity. This multimode field usu-

ally interacts with atomic sources that form the active
Passive optical systems such as beam splitters, systems medium within the resonator. There are several articles
of lenses, spectral filters, etc. play an important role both based on such a rigorous quantum field approach to laser
in classical and quantum optics. Resonators are fre- theory dealing with the problem of the action of resona-
quently used devices in particular, e.g. , in lasers, optical tors with output coupling. '
parametric oscillators (OPO) and four-wave mixers Solving the equations of motion for the correct mul-
(FWM). Although the radiation properties of lasers can timode radiation field automatically yields the field both
mostly be understood using classical theory as a basis, it inside and outside the resonator, without any additional
has been demonstrated that lasers also may generate non-
classical light.' Moreover, FWM and OPO have been
hypothetical premises. Because of the multimode charac-
ter of the radiation field, the equations of motion to be
used for generating squeezed light fields. In such situ- solved are in general much more complicated than those
ations a consistent quantum-physical description of the in a quantum stochastic approach, in which one can often
action of the resonators is required. For idealized, loss- confine oneself to equations of motion for a very few
less resonators the situation is rather simple. In reality, (damped) intracavity modes only. From their solution
however, the losses of the resonator cannot be disregard- the properties of the output field can then be calculated
ed. The field of interest is usually the output field of the in terms of input and intracavity field quantities in a
resonator that represents nothing but the losses of the straightforward way. Recently, some aspects of the con-
internal field. nection between a multimode quantum field theory and
Some approaches to deal with such situations have
been developed. " They are based on generalized quan-
the quantum stochastic input-output theory have been
studied' with special emphasis on squeezing inside and
tum noise theories (quantum Langevin equations) in or- outside a cavity.
der to derive relations between the output field and the The aim of the present paper is to derive the quantum
internal and/or the input field. The starting point is a stochastic concepts for resonatorlike systems from a
given radiation mode of the idealized (lossless) resonator rigorous quantum-field-theoretical point of view. Our
that is coupled to a reservoir. Physically, this reservoir is starting point is the general theory of the action of pas-
assumed to represent the field modes outside the resona- sive, lossless optical systems. ' In the following we refer
tor, and the coupling of the reservoir with the internal to this work as I. In I we have performed the field quant-
mode should represent the partial transparency of one or ization in the presence of both lossless, passive optical
both mirrors. In reality, however, there exists neither a systems and sources, the passive optical systems being de-
cavity mode nor a reservoir. The actual situation is the scribed by a space-dependent refractive index. This ap-
existence of a (continuous) multimode radiation field, the proach has successfully been applied to the problem of
mode structure of which is more or less modified (in com- spectral filtering of light. ' Such a general field-
parison to the free-space case) due to the presence of (pas- theoretical concept allows to derive all the required com-
sive) optical devices, such as the fractionally transparent mutation relations in a rigorous manner, cf., e.g. , I and

43 543 1991 The American Physical Society

 L. KNOLL, W. VOGEL, AND D.-G. WELSCH 43

Ref. 19. One of the obstacles of the quantum noise

theories "
is that fundamental commutation relations /I (k, x)=
4mI' coco
1/2

are postulated rather than derived. This fact is most ob-

vious in Ref. 11, where Carmichae1, through great efFort,
demonstrates that the commutation relations are not T(~)(e'kx —e ' x) ()&x &L
X (2.3)
violated by certain expectation values. ikx —ikx
Using the same concepts used in I we will derive quan- T*(co )
tum stochastic equations, including all fundamental com-
mutation rules (for a preliminary report see Ref. 21). The F being the mirror area. In Eq. (2.3), T(co) is the spectral
limits of applicability of such quantum Langevin equa- response function of the cavity field, which reads as
tions approach become evident, as they are connected
with a coarse-graining assumption. Clearly, the general t(~o)
T(co)= ', l =L+d/2 . (2.4)
1+—(
.
concept also works when the quantum stochastic 2ilcu/c
)
methods fail (cf. Refs. 12 —14.) In Sec. II the field modes It is expressed in terms of the spectral transmission and
are derived for a simple, realistic resonator mode1. For reflection response functions t(io) and r(co), respectively,
demonstrating the essential features we deal with a one- of the (partially transparent) right-hand mirror. In our
dimensional, single-ended cavity. Such resonators are dielectric layer model we derive
used, e.g. , in lasers and OPO. Section III is devoted to
the representations of the field operators and the corre- 2) i in —1)den/c
sponding propagation functions. The approximations t(co) = 2 2indcu/c (2. 5)
leading to the concept of damped cavity modes are
developed in Sec. IV. Fundamental commutation rela- 2i nd co/c
e i cod /c— (2.6)
tions required for a quantum stochastic theory are de- 1 y 2g
2i @dc'/c

rived from first principles. In Sec. V correlation func-

tions of the external field are studied and their relations where r =(n —1) /'(n +1) . It is worth noting that the
to the correlation functions of the internal field are de- structure of Eq. (2.4) is more general than might be ex-
rived. A summary and some conclusions are given in pected from our simple model. The only essential point is
Sec. VI. that t(co) and r(io) must satisfy the conditions
/t(io)i + ir(to)i =1, (2.7a)
II. MODES OF THE RADIATION FIELD t (co)r(co)+t(co)r*(co) =0 . (2.7b)
For demonstrating the essential features we confine In the following we therefore need not return to Eqs. (2.5)
ourselves to a one-dimensional cavity of length L, bound- and (2. 6). Clearly, both the dependence of t(to), r(co) on
ed by a perfectly rejecting plane mirror at x =0 and a frequency and the mode structure of the radiation field
partly transparent plane mirror at x =I.
. In order to inside the right-hand mirror (L &x &L+d) depend on
simplify the calculations we describe the latter by a the actual mirror used, which is, in general, a highly
dielectric layer of refractive index n and thickness d complicated multislab configuration. To circumvent this
(d «L). Moreover, we consider a given polarization problem we assume the mirror to be suKciently thin, so
direction of the radiation field, so that the vector charac- that in Eq. (2.3) we may write L =L+d =l. In this
ter of the operator of the electric field strength E(x, t) sense, we may renounce the mode structure inside the
may be omitted. The field quantization is performed with mirror and Eq. (2. 3) is complete.
continuous mode functions, since the resonator is embed- In many cases of practical interest we may assume that
ded in the infinite space. In this case the mode functions within the bandwidth of the light under study the spec-
(of the vector potential) 3 (k, x) obey the differential tral transmission and reflection functions t(co) and r(ro),
equation respectively, are slowly varying in frequency, so that their
frequency dependence may be disregarded: t(co)=t,
Gr2 r(io) =r In this c. ase the (complex) poles of the spectral
/I (k, x)+n (x)k /I (k, x)=0, k =co /c (2. 1) response function of the cavity field T(co), defined by the
Jx equation
where 2l I Oc
0 (2. 8)
1, O~x ~I., I +d ~x
L «L+d, (2 2) can easily be evaluated.
r =fr fe'@,
Making the ansatz
(2.9)
together with the corresponding boundary conditions at we derive
the surfaces of discontinuity (x =0, x =L, x =L+d) —iI
and the normalization condition [cf. Eq. (2.53) of I]. The
/2, , (2. 10)
evaluation of the mode functions is a straightforward where
procedure. We may restrict the consideration to non-
77C
negative values of the wave number (k 0). We derive =m +5co, m integer (2. 11)
43 RESONATORS IN QUANTUM OPTICS: A FIRST-. . . 545

5a) = —
21
(~ — ), tt (2. 12) Making use of the orthogonality
we easily derive
of the mode functions

(2.13) Et,+„)(x,t)= 2e—

oF f0
dx'K'+'(x, t;x', t')E'+'( ', t') .

In the particular case when the transmission is (3.5)

su%ciently weak (~t~ ((I)
we may simplify Eq. (2. 13) as Note that in the equations given above t' is an arbitrarily
follows: chosen initial time, which may be smaller or larger than
t. The propagation function is defined by the relation
(2. 14)
2l K'+'(x„t„'x, t )
Equation (2. 14) allows a simple, intuitive (quantum-
mechanical) interpretation. We imagine that a photon = —— dk f co A (k, x, ) A *(k,x2)e (3.6)
0
which enters the cavity leaves it after the time of Aight
To study the field outside the cavity it is reasonable to
2I
(2. 15) decompose it into two parts, viz. ,

with probability ~t . Hence, provided that times shorter

E'+'(x, t)=E;'„+'(x, t)+E', „+,'(x, t), x ) l, (3.7)
than the time of flight are not resolved, I = t lr„ is just
~

~ ~
where
the transition probability per unit time for a photon to
pass out of the cavity. As we will see later this interpre-
tation is closely related to the damped-mode concept to
E;„(,„,)(x, t)=i f 0
dk coA;„(,„,)(k, x)a(k, t), (3.8)
' 1/2
be developed. Clearly, in the case when Eq. (2. 14) is valid
A;„(k,x) =— 4m.
—ikx
the "damping rate" becomes small compared with the e (3.9)
Fso~
frequency spacing hco=co +1 — co =~c/l, where, apart
from the frequency shift 5co, the co coincide with the 1/2
T(co) e'
well-known frequencies of the radiation mode inside the A, „,(k, x) = (3.10)
T*(co)
ideal-resonator cavity ( t =0). From Eqs. (2.8) and ~ ~

(2. 10) together with Eq. (2. 14) the behavior of T(co) for Note that A;„(k, x)+ A „,(k, x) = A (k, x), A (k, x) being
frequencies in the vicinity of a given resonance frequency the complete mode functions as given in Eq. (2.3) (for
can readily be derived. The result is x & l). The information about the field incident on the
1/2 cavity is contained in E;'„+'(x, t), whereas E',+„,'(x, t) de-
T(a))= —C

2l
I
I /2
ip
—i(a~ —a) )
(2. 16) scribes the outcoming field available for practical applica-
tions. Clearly, the field operators E „+ '(x, t) and
where P is defined by the relation E',+„,'(x, t) may also be expressed in terms of source- and
free-field parts. Analogous to Eqs. (3.1) —(3.4) we may
(2.17) write
We see that for suSciently small values of I (that is,
((I)
the spectral response function of the cavity be-
(+)(out)(x~t)
in
(+)
in (out), s(x& )+Ein(+)(out), free(x& (3.11)
~t~
comes efFective in discriminating against values of co not where
equal to co
E,'„+(,'„,), (x, t)=F f d~f 0
dx'K, '„+(',„,)(x, t;x', r)
III. REPRESENTATIONS OF FIELD OPERATORS X J(x', r), (3. 12)
AND PROPAGATION FUNCTIONS

the results of I we may represent the opera-

Applying
E,'„+(',„, t„„(x,t) =i
) f 0
dk a) A;„(,„,)(k, x)

tor of the electric field strength Xe '"" ' 'Q(k t') (3.13)
E'+'(x, t) =i f 0
dk ai A (k, x)a (k, t) (3. 1) Analogous to Eq. (3.6), the propagation
defined by the relation
function is now

in terms of source- and free-field parts, viz. ,

„t );x~, = —— A;„(,„, (k, x, )
t) —E( (', , ) (x t2 ) dk to
E( )( )( t)+E( )( t) (3.2) 0
)

where X A*(k, x~)e

E, ' '(x, t) =F f dr f dx'K(+'(x, t;x', r)J(x', r)
r' 0
(3.14)
[the sources described by the operator J(x, t) are assumed Since the sources are assumed to be located inside the
to be inside the cavity] and cavity, the field outside the cavity fulfills the homogene-
" ous Maxwell equations, the solution of which can be writ-
Et,+„'(x, t)=i f0
dk A (k, x)e ' ' 'a(k, t') . (3.4) ten in the form
546 L. KNOLL, W. VOGEL, AND D.-G. WELSCH 43

E'+'(x, t) =E,'„+'(t +x lc)+E.'+„,'(t —x lc), x & I . is trivial for the free-field parts [see Eq. (3. 13) together
with Eqs. (3.9) and (3. 10)]. For the source-field parts we
(3. 15) will see this somewhat later after having calculated the
We note that the fields E,'„+'(t +x /c) and E,'+„,'(t —x/c), propagation functions. Substituting Eq. (2.3) (remember
respectively, are just the fields E „+'(x,i) and E,'+„,'(x, t) L =L+d =l) for the mode functions in Eq. (3.6) yields
defined in Eq. (3.7) together with Eqs. (3.8) —(3. 10). This the following result:

(i) 0&x, &l, 0&x, &l:

1
G, (t, —tz —(x, —xz)/c)+G, *(tz —t, —(x, —xz)/c)
&. Epc [
—G, (t, —tz —(x, +xz)/c ) —G,*(tz —t, —(x, +xz)lc )], (3.16)
Gi(t)= f deal T(co)l e (3.17)
(ii) 0&xz &l, l &x, :
1
[ Gz(t, tz —(x, —x—
z)/c)+Gz (tz t, —(x——
, xz)lc)
Epc
—Gz(t, tz —(x,—
+xz)/c) —Gz (tz —t, —(x, +xz)/c)], (3. 18a)
1
[Gz (tz t, —(x, —x—
z)/c) —Gz (tz —t) —(xi+xz)/c)], (3. 18b)
7T CpC

: — —(x, —xz—
)/c) —Gz(t, —tz —(x, +xz)/c)],
1
K,'+„,'( x„t ); xzt z)=
7T Epc
[Gz(t, tz (3.18c)

Gz(t)—T(t) = f dc@ T(co)e (3. 19)

0
(iii) I & xz, I & x, :

X' '(x„t, ;xz, tz ) =— 1

[ Go(t, tz
—(x, —xz)/c)+G() (tz —t, —(x, —xz)/c)
EpC

—G3(t, —tz —(x, +xz)/c) —G3 (tz —t, —(x, +xz)/c)], (3.20)
Go(t)= f den e (3.21)

G (t) f d
T(~) i~i- (3.22)
T*(~)

Note that in the case where 0&x, &l, l &x, is valid, or disregarding the frequency dependence of r(co) and
K'+'(x), t), xz, tz) can be derived from Eq. (3. 18a) by t ( co ) yields
means of the relation
G3(t) = rGO(t +2l/c)+—tGz(t) . (3.25)
(x)&tli zt)z[+
x& (xz&tz&x)&t) )] (3.23)
Here, the first term obviously describes the reAection at
The functions 6,
with difFerent arguments reAect the the right-hand mirror. The second term is related to the
various kinds of allowed light propagation (between the reAection at the left-hand mirror; after having passed
two space points x, and xz). Since in ease (ii) the light through the right-hand mirror into the cavity [transmis-
must necessarily pass through the fractionally transpar- sion coefficient t in the second term in Eq. (3.25)] a light
ent right-hand mirror we may expect that 62 ~ t as t goes signal can return to outside the cavity [propagation func-
to zero. In case (iii) the light propagation via refiection at tion Gz in the second term in Eq. (3.25)]. We therefore
the cavity as a whole is described by the function 63 in may expect tGz ~ t for t tending to zero. In case (i) fur-
Eq. (3.20). Substituting in Eq. (3.22) for T(co) [and for ther decomposition of G, analogously to Eq. (3.24) for G3
T*(co)] Eq. (2.4) and making use of Eqs. (2.7a) and (2.7b) is impossible, because light having left the cavity cannot
we may rewrite Eqs. (3.22) as follows: return.
— —ice(t+zllc) Equations (3. 16) —(3.24) are exact. For their further use
G 3 (t) d ( ) with the aim to derive eA'ective equations of motion for
0
the intracavity field and input-output relations some as-
+ f 0
dm t(co)T(co)e (3.24) sumptions and approximations are required.
43 RESONATORS IN QUANTUM OPTICS: A FIRST-. . . 547

(i) From the arguments given in Sec. II we disregard At this point we note that from substituting in Eq.
the frequency dependence of t(co) and r(co) and assume (3. 12) for KI„+(),„,)(x, t;x', r) [Eqs. (3.18b) and (3.18c)] to-
the inequality t ~ ~
((1
is valid. In this case we know that gether with Eq. (3.28), in the same manner as the free-
the spectral response function of the cavity field T(co) has field parts, the source-field parts E,'„+,'(x, t) and E,'+„,', (x, t)
poles in the complex co plane [see Eq. (2. 10) together with depend on x and t in the combination t + x /c and
Eqs. (2. 11) and (2. 14)] and moreover, we can use G3 in t— x/c, respectively, so that Eqs. (3.7) and (3.15) are
Eq. (3.25) instead of that in Eq. (3.24). The only func- identical:
tions G; to be evaluated are therefore Go, G, and G2.
(ii) As usual, we extend the (t) integrations to minus E(,„+)(x,t) =E,'„+'(t+x/c), (3.29a)
infinity. Although this approximation might seem to be
somewhat artificial it can be justified from the argument
E,'+„„'(x,t):E,'„+— —
, '(t x/c) . (3.29b)
that in the case of optical radiation fields we are interest-
ed in, slowly varying (in comparison with the mean opti- IV. THE CONCEPT OF DAMPED CAVITY MODES
cal frequency) amplitudes may be separated from the field
variables, so that all calculations to be carried out may be A. I.angevin equations
thought to refer to slowly varying quantities. The fre-
quency extension then gives rise to negligibly small off- Now we suppose t — t' & 0 and study the field inside the
resonant contributions. cavity, that is, 0(x (l, in Eqs. (3. 1) —(3.5). For the fol-
lowing it is convenient to decompose Et„,'(x, t) given in
Now making use of assumptions (i) and (ii) and applying Eq. (3.5) into two parts, viz. ,
standard methods of complex plane integration we easily
Ettee'(x, t) =E I+tt'ee(x, t)+E2+tt'„(x, t) t (4. 1)
arrive at the following result:
G, (t) = 2r).5( t), (3.26)
EI+I,'„(x, t)= —2eoF I dx'K'+'(x, t;x', t')E'+'(x', t'),
0
—ice t —(I (4.2a)
G, t= e
/2)~t~
(3.27)
—dx'K'+'(x, t;x', t')E
l
E2+t„„(x,t) = 2eoF '(x', t') .
'
1/2 I

+ 2~ —
c —i [co i ( I /2—
G2(t) =
)}t (4.2b)
I (3.28)
2l
From a careful inspection of Eq. (4.2b) together with Eqs.
where e(t) is the unit step function. (3.18a), (3.23), and (3.28) we see that

2eocF J — dr K'+'(x, t;r)E'+'(c (r —t'), t'), t t'& (l —x)/c, — (4.3a)

E(+) (x t)=' 0, t t &(l x)/c (4.3b)

where mately) shows the following spatial structure:

K(+'(x, t;r) =- 4~F roc [G2(t +x /c —r)

j, K' '(x, t;x', t')=/K'+'(x, t;x', t'), (4.5a)

—G2(t —x/c —r)] . (4.4) K' '(x, t;x', t') ~sin(co x/c)sin(co x'lc), (4.5b)

Note that in the x' integral we have let x'=(r —t')c and (0&x & l, 0 (x' & l), which in general is not possible, cf.
dx — c dv. Eqs. (3. 16) and (3.27). Now imagine we are only interest-
We saw that in the case where the transparency of the ed in resolving time elements A~ large compared with the
right-hand mirror is small ( t && 1) the spectral ~ ~
time of fiight rs [cf. Eq. (2. 15)] but, of course, small com-
transmission response function of the cavity field is pared with I
sharply peaked (with linewidth I o- t ) near the frequen- )
~

~„«h«& I (4.6)
cies of the normal modes of the resonator bounded by
two perfectly reAecting walls. The question arises of so that in slowly varying functions of time
whether the radiation field inside the cavity might (ap- t=t+xlc —
=t x/c (0&x &l), which means disregard-
proximately) be expanded in normal modes of this ideal ing time-of-Aight effects due to the finite cavity length or,
resonator, the small but finite value of )t giving rise to a ~
in other words, neglecting terms of the order of magni-
damping effect in the usual sense. Looking, for example, tude I r„. (Note that «rI ' is equivalent to )t && 1.) ~

at Eq. (4.2a) we see that attempting to express the cavity Only in this case, from Eq. (3. 16) together with Eq. (3.27)
field in terms of standing waves might only be successful we see that K' '(x, t;x', t') has indeed the form suggest-
if the propagation function K'+'(x, t;x', t') (approxi- ed in Eq. (4.5), viz. ,
54S L. KNOLL, W. VOGEL, AND D.-G. WELSCH 43

IC'+'(x, t;x', t')= — +exp i—to i —

— (t t—
') sin(to x/c)sin(co x'/c), 0&x &1,0&x'&1 . (4.7)

Clearly, the coarse-graining condition (4.6) implies that performing time integrations with higher precision than Ar
makes no sense. Hence, by making use of Eqs. (3. 18a), (3.23), (3.28), we can treat Eqs. (4.3) as
E2+t„'„(x,t) = —2eocF d~ K'+'(x, t;r)E,'„+'(r), (4.8)

where
1/2
~'+'(x t r)= l
y 2c r l
e'&exp i —
co i ——
.I
(t —r) sin I co x /c I . (4.9)
eoF1

Note that in Eq. (4. 8) the contribution of E,'+„,' to Ez+t, '„ is [which are orthonormalized in the sense of Eq. (2. 54) of I]
omitted because of its off-resonant character. Further, and combining Eqs. (3.2), (3.3), (4. 1), (4.2a), and
we also can substitute Eq. (4.7) for K( xt; x', v) in Eq. (4.7) —(4.9) eventually yields
(3.3) (0&x &1).
Introducing the standing-wave mode functions E'+'(x, t)=i g co A (x) a (t), (4. 11)
'
1/2
sin(co x/c I, 0&x &1, (4. 10)
IFEoCOm where

a (t) = a (t')exp i to —i (t t') —

—— + Fdr —
dx'2 (x')exp i to i ——(t—
—w) J(x', r)

—i —r—
+(I )' e'~
f dr exp i co (t ) b (r), (4. 12)

2 EDFc
b (t)=— 2eoFC E,'„+'(t) . (4. 13) c (t)= — E,'+'(t) . (4. 16)
&~m &~m

From Eq. (4. 12) the operators a are seen to obey the
differential equations B. Commutation relations

a (t)= i co
.
i — it
r + F
(t)— —f dx'3—(x')J(x', t) In order to treat practically the Langevin equations as
given in Eq. (4. 14) the commutation relations must be de-
rived. In particular, it must be proved that the mul-
+(r)'"e'~b. (t) . (4. 14) timode operators a and a satisfy, in coarse-graining
approximation, the boson commutation relations for
These equations may be regarded as quantum-stochastic equal times:
Langevin equations for the multimode photon destruc-
tion and creation operators & and a in the presence of [&,a ]=5 (4. 17a)
light-matter interaction, where the (I )'~ e'~b represent
the operator Langevin noise sources, and the — (I /2)& [8,a .]=0=[a,a .] . (4. 17b)
give the damping effects.
%'e would finally like to mention that in the case when The simplest way is to start from the reversal of Eq.
t & t' is valid in close analogy to the way outlined above (4. 11), viz. ,
the time-reversal Langevin equations can be derived to be
2eDF
a (t)= — co +i it —F f dx'2
(t)+ ')J(x', t)
(x—
& (t)= .
sA f dxo
A (x)E'+'(x, t) . (4. 18)
2 A 0

+(1 )' e '~c (t), (4. 15) Since for equal times [E' ', E'+']=0=[E' ', E' '], we
readily see that the commutation relations in Eq. (4. 17b)
where are valid. To calculate the commutator
43 RESONATORS IN QUANTUM OPTICS: A FIRST-. . .

4~OF [a (t), a, (t)]=5 +O(l rs) . (4.22)

[a (t), 8 (t)]=
fg
f dx f dx'2
I

0
I

0
(x)A, (x')
Apart from corrections of the order of magnitude of I wz
X[E'+'(x, t), E' (x', t)], (that is, in coarse-graining approximation), the operators
a and a indeed satisfy Bose commutation relations
(4. 19) and can therefore be regarded as photon destruction and
we expand the field with respect to the normal modes creation operators, so that the Langevin equations now
A (k, x) [cf. Eq. (3. 1)] and use the Bose commutation rela- may be written as
tions for the photon operators a(k, t) and a (k, t), which
yields a = —i co i—a ———[a,Hz]+(I )' e'~b (t),
[E'+'(x, t), E' '(x', t)]=iiric K(x, x'), (4.20) (4.23)
HI being the Hamiltonian of the intracavity light-matter
where the quantity K(x, x') is given by
interaction.
Now, let study the (time-dependent) commutation rela-
K (x, x') = — —(x —x')/c
1
4m. Froc [ 6, ( ) tions for the operators b (t) and b (t). Applying Eq.
(4.22) of I to the incoming field we may write
—G*, ( —(x —x')/c } r
(j&) (j&)
[E;„' (x, t, ), E;„' (x2, t2)]
—6, ( —(x +x')/c) ~(jl
= [E;„f ) (j2)
(xl tl) E' f (x2 t2)]
+G*, ( —(x +x')/c )] . (4.21) (j[,j&) (j2, j& )
+Dl~ l~ (xl, tl, x2, t2) D;„;„(x2,t2, x ltl)
Substituting here for 6, Eq. (3.27) and combining Eqs. (4.24)
(4. 19)—(4.21) we arrive at where, in analogy to Eq. (4. 16) of I,

D;„'„' (x„t,;x, t )= F f d—x', f dx' f

0 0 oo
dt', f oo
dt'e(t
(j&),
t' )e(t'

X K;„(xl, t l, x l, t ,l
—
t', )e(—

(j2)
t', t, )—
)K;„(x2,t2, x 2, t 2 )

X[J ' (x', , t', ), J ' (x2, t2}], (4.25)

l &x„ l &x, . Remember the symbols stand for + or j [E,'+ '(t, ), E,'„'(t, )] = 2eoFc |i(t, t, ), —28a)
—.Suppose )
t, . From Eq. (4.25) we easily see that in
t2 f)t,
(4.
this case D;„",„' (x2, t2,'x, t, )=0. On the other hand,
from Eqs. (3. 18b) and (3.28) together with Eq. (4.25) we [E,'„+'(t, ), E,'„+'(t, )] =0= [E,'„'(t, ), E', „'(t, )], (4.28b)
~(j[,j2)
see that nonvanishing D;„;„(x„t,
;x2, t2) requires the so that according to Eq. (4. 13)
following conditions to be fulfilled:

t2 &t2) ti ) ti (4.26)
[b (t, ), b (t, )]= „, b(tl —t2), (4.29a)
x2+x 2 xi+x i
t2 t2& )
i' —t
1 1
[b (t, ), b, (t, )]=0=[b (t, ), b (t, )] . (4.29b)
C

)
Since (x2+x 2 )/c 0, (x l+x l )/c 0, these conditions ) For the sake of transparency let us introduce slowly vary-
cannot be fulfilled simultaneously, so that ing photon operators,
D;„;„(x„t, ;x2, t2)=0. From analogous arguments we
conclude that the D terms also vanish in the case t2 & t, .
Hence we find that
a (t)=a (t)e™, (4.30)
~(jl ) ~( j))
[E;„' (x, , t, ), E;„' (x2, t2)] b (t)=b (t)e (4. 31)
(j, )(j, )
(x„t, ), E;„
=[E;„f„, f (x2 t2)] . (4.27)
so that the Langevin equations read
In other words, the incoming field can be treated as free
field. The free-field commutator in Eq. (4.27) is calculat- a = ——a ——[a,Hz]+(I )'~ e'~b (t) (4.32)
ed in a straightforward way. Combining Eqs. (4.27),
(3.11), (3.13), and (3.15) yields and Eq. (4.29a) becomes
550 L. KNOLL, W. VQGEL, AND D.-G. WELSCH

scale. (Remember that r„~ Ace ', Aa] being the frequency
[b (t, ), b' (t, )] spacing of the cavity modes. ) In particular, using the
l m'] m'z commutator in Eq. (4.33) (together with slowly varying
] z (4.33) quantities) in time integrals, we must integrate over a
)]gz gt
time which is long compared with Ace ' and we see that
(

the integrand goes through many cycles and averages to

As already pointed out, the Langevin equation approach
zero unless m = m'. In this sense, we have
to the problem of determining the intracavity field is
based on a coarse-graining approximation. The results [b (t]), b'+](tz)]=5 6(t] t, ) —
. (4. 34)
derived in this manner can only be regarded as being
correct on a time scale which is long compared with ~&, Next we consider commutation relations for cavity pho-
because in the derivation of Eq. (4.32) we have smoothed ton operators and Auctuation operators. Applying Eq.
out the fluctuations of the intracavity field on this time (4.24) [cf. also Eq. (4.22) of I] in a modified form yields

~(j ) ~(j, ~(j ~(j),j2) —D;„(xz,

~(j2,J)
~(j,' tz;x„t, },
) ) ) )
[E (x»t, ), E;„' (xz, tz)]=[E««(x»t]), E]„ xz) tz)] +D „]( x„ t]; xz)tz) (4.35)

),
fus ee(

where
~(j &j2)
D ';„' (x„t,;xz, tz)= Fdx',
2
0
1 t —Idx'
0
t +OO

OO
dt',
+OO

QO
dt' e(t —t' )e(t' —t', )e(t', —t])
XI (jl
„(
(x], t„x', , t',, )K,(j2
)
zx, tz; zx, tz )

X[J ' (xi, ti), J ' (xz, tz)] . (4.36)

In contrast to Eq. (4.25) in Eq. (4.36) only one IC;„, name- mode expansion [cf. Eqs. (3.4) and (3. 11)]. We derive
ly, K;„' (xz, tz;x z, tz ) appears. Clearly, if tz t, the two ) ~(j& ) ~(j&)
[Ef„,(x, , t, ), E;„f„,(xz, tz)]
D terms again vanish, in analogy to the arguments fol-
lowing the inequalities (4.26). [Note that the second con- — (j2)
dition in the second row of the inequalities (4.26) is now
i% E]„' (xz, tz, x], t]), j]Wjz (4.41)
Bt,
)
missin~. ] Now suppose t, tz. In this case, the relevant
(0 &x] (
1, 1 (xz ). From Eqs. (3.18b) and (3.28) we
D;„' ' (xz, tz, x„t,
) may only be expected to contribute
readily see that in coarse-graining approximation this
to the commutator in Eq. (4.35) if the conditions
commutator vanishes if t, & t2+x2/c. In this manner we
t] )t] )t2) t2 (4.37)
eventually arrive at the result
t, ) t, +(x, +x,' )IC
' ~(j]
[E
)m(j2)
(x], t]), E,„(xz,tz)]=0, t] &tz+xz/c (4.42)
are fulfilled. In coarse-graining approximation, these
conditions cannot be fulfilled as long as t, tz+xzlc, so ( (0&x, (1, 1 &xz). Remembering
(4.18) we therefore derive
Eqs. (3. 15), (4. 13), and
that
[a (t, },b .(t, )]=0=[a (t]), bt (t, )],
[E ' (x, , t, ), E;„' (xz, tz)]
(j, ) (j ) (4.43)
=[Efp«(x]&t])&E]g fusee(xz)tz)], t, &tz+xzlc In relaxation theory, these commutation relations are
(4.38) usually postulated from the argument that the system
(0 & x, ( I, 1 & xz ), and hence
variables a and & do not depend on the input variables
b and b in the future, cf., e.g. , Refs. 6 —11.
~ ~

[E]+](x,, t, ), E]„+](x,, t, )]=0 Summarizing the results of this section we see that the
problem of solving Maxwell equations for determining
= [E' '(x „t, ), E', „'(xz, tz )], the radiation field inside a high-Q cavity naturally leads,
in coarse-graining approximation, to quantum-stochastic
t, &t, +x, lc . (4.39)
Langevin equations. This renders it possible to apply the
Moreover, applying Eqs. (3. 15) and (4. 18) yields concepts of quantum stochastics, in which the method of
Langevin equations represents one in a series of
[a (t] ), E]+](tz )] =0= [& (t, ), E]„'(t,)], t, & t, . equivalent descriptions. Finally, it should be pointed out
(4.40) that in the system considered the losses of the internal ra-
diation field only arise from its contact to the external
In the cases of [E'+', E, '„'] and '„']
[E' ', E, we calculate field via the fractionally transparent mirror. However,
the free-field commutators by the familiar procedure of there will be additional losses arising, for example, from
43 RESONATORS IN QUANTUM OPTICS: A FIRST-. . . 551

some wall heating. Describing these e6'ects in relaxation [Ee„', (xi, ti), E()„', (x2, t2)]
approximation leads to additional damping terms and ~(
~(j) )
operator Langevin noise sources in the Langevin equa- =[E, ] f (x] t]) E j2)] f (x2 t2)] (5.5)
tions. Clearly, in this case only some part of the overall
noise operator can be identified with the incoming radia- (l&x„ l &x2) and
tion field.
[P ' (x„t, ), E,„', (x2, t2)]
V. CORRELATION FUNCTIONS (j& ) ( j2)
=[Efeee(x]&t])&E«] f ee(x2tt2)]t ti & t2 x2/c
OF THE EXTERNAL FIELD
(5.6)
A. Commutation relations (0 (x, & l, l &x2). The free-field commutators in Eqs.
(5.5) and (5.6) may now be evaluated in the standard way,
We showed that, under certain circumstances, the by making use of Eqs. (3.4) and (3.11). In close analogy
problem of determining the radiation field inside the cavi- to Eqs. (4.28a) and (4.28b), we derive
ty may be treated within the framework of quantum sto-
chastics. On the other hand, experimental studies are (5.7a)
usually performed on the external field, the properties of 2Fe c Bt,
which are therefore the desired information. To relate [E'.„+,'(t, ), E'.+„,](t2) ] =0= [E.
'„,'(t, ), E'.„,'(t, )] . (5.7b)
the radiation field outside the cavity to the intracavity
field we must study Eqs. (3.2) —(3.4) together with Eqs. The evaluation of the free-field commutator on the right-
(3.18a), (3.20), (3.25), (3.26), and (3.28). The calculations hand side of Eq. (5.6) is analogeous to Eq. (4.41), with
performed in coarse-graining (and resonance) approxima- K,"„', instead of K J'. We therefore derive [cf. Eq. (4.42)]
tion are very similar to those in Sec. IV, so that we may
forego describing them in detail. Assuming t — ]«(x2, t2)] —
~(j&') ~(j&)
x/c & t [E (x»t]), E 0, ti & t2 x2/c
we arrive at the following result:
(0 (xi & l, l & x2), or equivalently
E'+ '(x, t ) = E,'„+ '( t + x /c ) + E',+„,'( t —x /c ), (5. 1)
'
1/2 [a (t, ), E,' +, '(t, )] =0= [a (t, ), E,'„,'(t, )], t, & t, (5.9)
E'+'(t
Out
—x/c) = g (I")'~ e'~ ACOm
a (t —x/c) [cf. Eq. (4. 18) and the second relation in Eq. (3. 15)].
2eoFc
m At this point we note that the commutation relations
+ rE .„+ '( t —x /c ) (5.2) given in Eqs. (5.7) and (5.9) are just the commutation re-
lations needed for complementing the time-reversal
For practical applications of Eqs. (5. 1) and (5.2) commu- Langevin equations given in Eq. (4. 15). Making use of
tation relations involving operators of the outgoing field
are needed. Applying Eq. (4.24) (in a modified form)
yields
[E, „', (x], t]), P, „', (x2, t2)]
[c (t, ), c .(t, )]=,
Eq. (4. 16) we may represent Eqs. (5.7) in the form
l
,
C)
5(t] —t, ), (5. 10a)

[c (t, ), c (t2)]=0=[c (t, ), c .(t2)] . (5. 10b)

[Eo t, f (x] tl ) Eo t f ( 2' 2)]
In particular, introducing slowly varying quantities [cf.
~(jl,'.j2)
+D. „, „,(x ], t]., x „t, —D,„,,„,(x„t, ;x],t,
) ) Eqs. (4.30) and (4.31)] we may [analogous to Eq. (4. 34)]
rewrite Eq. (5. 10a) as
(5.3)
(l &x„ l &x2), [c (t, ), c (t2)]=5,5(t] —t, ) . (5. 11)
Further, Eq. (5.9) may be represented in the form
[E ' (x»t, ), E«', (x2, t2)]= [Epee(x]yt]))E ex'
«f„]e( t2)]
[a (t, ), c (t, )] =0= [a (t, ), c .(t, )], t, & t, . (5. 12)
D,
+ ~(jl, j&)„,(x 1, t, ;x2, t2 )
Equations (5. 10), (5. 11), and (5. 12) which complement the
D,(j2, j& )
„, (x2, t2;x—
i, t, ) time-reversal Langevin equations given in Eqs. (4. 15), ob-
(5.4) viously correspond to Eqs. (4.29), (4.34), and (4.43), which
complement the (ordinary) Langevin equations given in
(0(x, &l, 1&x2). In Eqs. (5.3) and (5.4), D,„", ,'„, and Eq. (4. 14).
D ', '„, are defined according to Eqs. (4.25) and (4.36), Finally, let us consider the commutator:
with K,'j„',instead of K;'„". Comparing Eq. (3. 18b) with
[E;„' (x, , t, ), E,„', (x2, t2)]
Eq. (3.18c) and taking into account Eq. (3.28) we see that
in the unit step functions of K;'j'(x„t, ;x2, t2) and (J, ) -(J, )
,
(f 1 1) Eo t, free(x2 t2)l
K'~„', (x 1, t, ;x2, t2 ) the time arguments are exchanged
from t2 —t, into t, — (x„ti,x2, t2) —D,
~(J),'„,
Jp) w(J) J) )
t2. +D;„, ~
„, ;„(x2,t2;x„t, )
Remembering the derivations of Eq. (4.27) and Eq. (5. 13)
(4.38) for the corresponding in-field commutation rela-
tions we readily find that [cf. Eq. (4.24)], where
552 L. KNOLL, W. VOGEL, AND D.-G. WELSCH 43

D;„,'„, (x&, t&, xz, tz)= F—0 dxI 0

dxz —co dtI —oo dt'zB(tz
—tz)B(tz —t&)B(t'& —t&)
XI(C;„(x),t), xI, tI )E,„', (xz, tz,'xz, tz
X[Jn(jl ),
(x', , t',, ),
~(j2)
J '(xz, tz)]
)

(5. 14)

)
(l &x&, l &x~). Suppose t& tz. In this case we easily (m, n)
(Xi)tl) ' ' )Xm+n) m+n )
find that D;„",„', (x&, ti, xz, tz)=0. From Eq. (5. 14) to- m+n
gether with Eqs. (3. 18b), (3. 18c), and (3.28) we see that
(j2, )
E' '
x, , t, T
j=m+1
Q E' '(x, , t, ) ),
nonvanishing D, „,j& ;„(xz,tz, x„t,
) requires the following i =1
conditions to be fulfilled: (5.21)
t1) t1 ) t2) t2, )
(x, , x l) in terms of correlation functions of the intra-
(5. 15)
x 2+x x 1+ cavity field and the incoming field. To illustrate the pro-
cedure let consider the simplest case, when the incoming
field is in the vacuum state, viz.
Clearly, these conditions cannot be fulfilled simultaneous-
( E,'„+') =o=(E,'„-' ) . (5.22)
ly, so that D,J2'~l
„, ; (xz, tz;x, , t, ) =0. Now suppose t, & tz,
so that D, „, ;„(xz,tz , x &, t, ) ='0. Nonvanishing Substituting Eq. (5. 1) and its Hermitian conjugate, re-
~(j&,j2)
D;„,„, (x, , t, ;xz, tz) requires the conditions spectively, for E'+'(x, t ) and E' '(x;, t; ) in Eq. (5.21)
and remembering the commutation relations given in Eq.
(5. 18), in the T+ (T ) products the operators E „+' and
(5. 16) „' ',
E,'„+, ' (E', and E, ') may be rearranged in such a way
f1 —t1 )
X1+X 1
t2 —f2 )
x2+x 2 that the operators E,'„+'(E, '„')
are on the right (left) of the
operators EI+„,'(E, „,'). Taking into account Eq. (5.22) we
to be fulfilled. In coarse-graining approximation, these therefore derive
conditions cannot be fulfilled simultaneously as long as
t2 —x2/c (t,+x1/c. we therefore derive m m+n
~(j& )
[E;„' (x, , t, ), E,(j2)
„', (xz, tz)]
(5.23)
i =1 j=m+1
= [E;„'t„,
~(j& )
(x „t, ),
~(j2)
E, „', t„,(x z, tz ) ], where r= t —x Ic. Note that in Eq. (5.23) time ordering
t, +x, Ic ) tz —xz/c (5. 17)
prescriptions are needless because of the commutation re-
lations given in Eq. (5.7b). In particular, we therefore
(1&x„l &xz), so that may rearrange the operators E,' t'(r ) [E,'
t'(w; )] in such a
way that
[E', +'(t, ), E', +, '(t, )]=0=[E',„'(t, ), E', „,'(t, )], t, )t, .
(5. 18) G (m, n) (x„t, ;. . . ;x „,t „)
m m+n
The remaining free-field commutators in Eq. (5. 17) may
again be evaluated by the familiar procedure of mode ex- i
g=1 E,'„,'(;) j =m+1
E', „,'(r ) (5.24)
pansion [cf. Eq. (3. 11)]. We obtain
where the time-ordering symbols T and T+ now refer
to the time arguments r,. and r, . Substituting Eq. (5.2)
and its Hermitian conjugate, respectively, for E,'+„, ' and
(tz —ti), A &iz
lA 8 (j2)
G3' (5. 19)
4~Fe0c Bt, E', „,' in Eq. (5.24), and taking into account the commuta-
tion relations given in Eq. (4.40), in the T+(T ) prod-
where G3+' =G3, G3 ' =G3. From an inspection of ucts the operators EI„' and a (E, and a ) may be '„'
Eqs. (3.25), (3.26), and (3.28) we see that G3(tz —t& )=0, rearranged in such a way that the operators E,'„+'(E,'„')
)
t, tz (in coarse-graining approximation). Hence, are on the right (left) of the operators a (a ). Together
[E,„' (t&), E,„', (tz)]=0, t& ) t,- . (5.20) with Eq. (5.22) we therefore obtain
G (m, n) (x„t, ;. . . ;x +„,t +„)
m+n
B. Field correlation functions a(., ) ), (5.25)
j=m+1
The commutation relations derived above enable us to
express observable correlation functions of the type where the abbreviation
43 RESONATORS IN QUANTUM OPTICS: A FIRST-. . . 553

1/2 ing field is not in the vacuum state, so that Eq. (5.22) is
&~m
&(t) =g (I )' e'~
~

2eoF
& (t) (5.26) not valid. From the considerations outlined above it is
evident that after substituting Eq. (5. 1) together with Eq.
is used. Thus we have expressed the measured field (5.2) and its Hermitian conjugate, respectively, for
correlation functions outside the cavity in terms of nor- E'+~(x, t ) and E' '(x;, t; ) in Eq. (5.21) all the operator
mally and time-ordered correlation functions of intracav- rearrangements described and the operator ordering
ity (damped-mode) photon operators. The latter ones can prescriptions related to them remain valid. Clearly, the
be calculated from the quantum Langevin equations. result found differs from Eq. (5.25) in nonvanishing con-
Finally, let us brieAy study the case, when the incom- tributions of mixed correlation functions of the type

We may represent the correlation function given in Eq. is a quantum-field-theoretical approach, in which the los-
(5.21) in the following form: sy resonator is embedded in the infinite free space.
(m, n) (x Sources are assumed to be present inside the resonator,
1r li ' ~ m+n& rn+n )
throughout they are included in our considerations.
m+n
m
Quantum Langevin equations are derived, based on the
E' '
x, , t, 8 rt F. ' '(„t;) l, neglection of time-of-Aight effects of a photon through
i =1 j=m+1
the resonator (coarse-graining approximation). All the
(5.27) commutation relations required in practical applications
where [after substituting in Eq. (5.27) for E'+' and E' of the quantum-stoachastic equations are rigorously de-
respectively, Eq. (5. 1) together with Eq. (5.2) and its Her- rived. This is one of the strong points of our theory; so
mitian conjugate] the ordering symbol 8+ (6 ) intro- far the corresponding commutators have been postulated
duces the following operator ordering in products of more or less. Moreover, the inclusion of sources implicit-
operators F. „+ ' and a(E~„~ and a ): ly gives the proof of the applicability of quantum
Langevin equations for nonlinear optical processes inside
of the operators E', „+' and a (E';„' and a )
(i) Ordering the resonator, which also may be approximately de-
with the operators EI„+'(EI„') to the right (left) of the scribed by effective Hamiltonian concepts.
operators a(a ). Correlation functions of the field outside the resonator,
(ii) T+[T ] time ordering of the operators a(r. ) which are usually measured, can be expressed in terms of
[d'(r;)]. operators of the internal field and the input field. Espe-
In the particular case when Eq. (5.22) is valid, Eq. (5.27) cially when the input field is in the vacuum state, normal-
of course reduces to Eq. (5.25). ly and time-ordered correlation functions of intracavity
field operators are the relevant physical quantities.
VI. SUMMARY AND CONCLUSIONS Our first-principles approach for dealing with resona-
tors in quantum optics is not restricted to the resonator
Input-output relations have been studied for a one- model under consideration. It is straightforward to gen-
dimensional, single-ended resonator. The starting point eralize this method to other resonator configurations.

S. Machida and Y. Yamamoto, Opt. Commun. 57, 290 (1986). Am. B 4, 1683 (1987).
S. Machida, Y. Yamamoto, and Y. Itaya, Phys. Rev. Lett. 58, ' M. J. Carmichael, J. Opt. Soc. Am. B 4, 1588 (1987).
1000 (1987). K. Ujihara, Phys. Rev. A 12, 148 (1975); 16, 652 (1977).
3R. E. Slusher, L. W. Hollberg, B. Yurke, J. C. Mertz, and J. F. J. C. Penaforte and B. Baseia, Phys. Rev. A 30, 1401 {1984).
Valley, Phys. Rev. Lett. 55, 2409 (1985). J. Guedes, J. C. Penaforte, and B. Baseia, Phys. Rev. A 40,
4Ling-An Wu, H. J. Kimble, J. L. Hall, and H. Wu, Phys. Rev. 2463 (1989).
Lett. 57, 2520 (1986). J. Gea-Banacloche, Ning Lu, L. M. Pedrotti, S. Prasad, M. O.
~See also the papers in J. Opt. Soc. Am. B 4, Special Issue No. Scully, and K. Wodkiewicz, Phys. Rev. A 41, 369 (1990).
10 (1987). L. Knoll, W. Vogel, and D.-G. Welsch, Phys. Rev. A 36, 3803
M. J. Collett and C. W. Gardiner, Phys. Rev. A 30, 1386 {1987).
(1984) ~ L. Knoll, W. Vogel, and D.-G. Welsch (unpublished).
7C. W. Gardiner and M. J. Collett, Phys. Rev. A 31, 3761 SL. Knoll, W. Vogel, and D.-G. Welsch, J. Opt. Soc. Am. B 3,
(1985). 1315 (1986).
sY. Yamamoto and N. Imoto, IEEE J. Quantum Electron. QE- L. Knoll, W. Vogel, and D.-G. Welsch, Phys. Rev. A 42, 503
22, 2032 (1986). (1990).
90. Nilsson, Y. Yamamoto, and S. Machida, IEEE J. Quantum J. D. Cresser, J. Phys. B 20, 4915 (1987).
Electron. QE-22, 2043 (1986). 2 L. Knoll, W. Vogel, and D.-G. Welsch, J. Mod. Opt. (to be

C. W. Gardiner, A. S. Parkins, and M. J. Collett, J. Opt. Soc. published).