amplitude
c 2
assoiated with thecos 2kx
exitation mode of the BEC. By introduingthe quadratures (real and imaginary parts) of
c 2
,X = 1
√ 2 (c † 2 + c 2 ) ; Y = i
√ 2 (c † 2 − c 2 ),
(5.6)the Hamiltoniantakes the form
H = − δ C a † a + iη (a † − a) + 2ω R (X 2 + Y 2 ) + u a † a X .
(5.7)Here
δ C = ∆ C − NU 0 /2
is the shifted avity detuning, andu = √
NU 0 /2
, i.e., thesingle-atom oupling strength
U 0
is magnied by the square root of the atom number,whih an mean several orders of magnitude. Even if the dispersive oupling is very
weak due to large detuning,the bosoni enhanement an lead to asigniant oupling
betweenthetwomodes
a
andc 2
.TheHamiltonian(5.7)orrespondstothewidelystudied radiationpressureouplinginopto-mehanialsystems[128℄.There,intheordinaryopto-mehanial sheme, the avity photon number is oupled to the position operator of a
mirror. Here, in the BEC-avity system, the
X
quadrature operator of the atom eldexitation mode an be interpreted only as a titious `position'. Although the formal
analogy is omplete between the two systems, the typial parameters lead to dierent
regimes.Inpartiular,hereintheaseoftheBECthe`mehanial'vibrationalfrequeny
ω R
is typiallymuh less than other frequenies of the system, whih is not the ase inusualopto-mehanialsetups.Therefore,itislegitimatetomakeafurtherapproximation
for the BEC-avity system, whih onsists in using an expansion in terms of the small
parameter
ω R /κ
.The avity noise is taken into aount by the noise operator
ξ
. It has a vanishingexpetation value:
h ξ i = 0
, meanwhile its seond order orrelation is assumed to be aDira-
δ
funtion,deribing white noise:h ξ(t) ξ † (t ′ ) i = 2κ δ(t − t ′ ) ,
(5.9)allthe other orrelations vanish,
h ξ † ξ † i = h ξ † ξ i = h ξ ξ i = 0 .
5.2.1 Adiabati elimination of the eld
On integrating Eq.(5.8a), weget the timedependent solutionof the eld operator
a(t)
.The time sale of the atomimotion whih weare interested in is muh longerthan
1/κ
,thus the homogeneoussolution with the initialvaluea(0)
vanishes. The partiularsolutionis
a(t) = η κ − i∆ +
Z t 0
e (i∆ − κ)(t − t ′ ) ξ(t ′ )dt ′ ,
(5.10)where
∆ ≡ ∆(X) = δ C − uX
. The rst term is the steady-state oherent amplitude of the eld,α(t) = η/(κ − i∆(X(t))) ,
(5.11)whihisnowanoperator-valuedfuntionoftheposition
X
.TheseondterminEq.(5.10)representsan`aggregate' noise,denoted by
Σ(t)
.Its expetation value isobviouslyzero:h Σ(t) i = 0
.The seond orderorrelation funtionan bereadily derived from Eq.(5.9),h Σ(t 1 )Σ † (t 2 ) i = e iδ(t 1 − t 2 ) e − κ | t 1 − t 2 | − e − κ(t 1 +t 2 )
.
(5.12)The`aggregate'noise,havinganexponentiallydeayingorrelationfuntion,isaoloured
noisewithaspetralwidth
κ
.However,onatimesalemuhlongerthan1/κ
,theseondterm inthe braket vanishes,and the remainingpart an be approximated by aDira-
δ
funtion. Fulllingproper normalization,the approximate orrelationfuntion beomes
h Σ(t 1 )Σ † (t 2 ) i ∼ = 2 κ
∆ 2 + κ 2 δ(t 1 − t 2 ) .
(5.13)The dipole fore to be substituted into equation (5.8) is the sum of an adiabati
potentialand a noise term:
ua † a = u | α(t) | 2 + Ξ(t) ,
(5.14)with the intensity noise dened as
Ξ(t) = u α(t)Σ † (t) + α † (t)Σ(t) + Σ † (t)Σ(t)
.
(5.15)The noise orrelation funtionis
h Ξ(t 1 )Ξ(t 2 ) i = u 2 h α † (t 1 )Σ(t 1 )α(t 2 )Σ † (t 2 ) i = D(X) δ κ (t 1 − t 2 ) ,
(5.16)wherethe diusion operatoris
D(X) = u 2 | α | 2 2 κ
∆ 2 + κ 2 = u 2 2κη 2
(∆ 2 (X) + κ 2 ) 2 .
(5.17)Thisdiusionproessrepresentsthequantumnoisetransmittedfromthephotonnumber
utuationsintothe momentumviathe optomehanialoupling. Wenotethat herethe
diusion isan operator-valuedfuntion of the positionoperator
X
.5.2.2 First-order orretion to the adiabati elimination
WhengettingtheadiabatieldamplitudeinEq.(5.10),theondensatedynamisis
as-sumedtobefrozen.Thisassumptionanberelaxedand theatomimotionanbetaken
intoaountbymeansofasystematiexpansionintermsofthe titiousmomentum
Y
.Thesmallparameterinourmodelis
ω R /κ
,beausethereovery timeofthephotoneld1/κ
is smallompared to the harateristi time sale of the atomi motion determined by the reoil frequenyω R
. At the same time wewillnegletthe eet ofmotion ontheutuating fore
Σ
sine it would ontribute to the variation of momentum orrelationfuntions inhigher orders.
Inpriniple,theeldderivativeinEq.(5.8a)anbeintegratedandtheinstantaneous
eld an be obtained from the knowledge of the full trajetory
X(t)
. The adiabatiapproximation means that the eld depends only on the atual position. Close to the
adiabatiregime,thehistorywellinthe pastisirrelevant,andjusttheloalbehaviouris
important.Thetrajetoryloallyanbedesribedbythepositionanditsrstderivative,
i.e., the momentum.Thus welook forthe eld amplitude inthe form
α(X, t) ∼ = α 0 (X) + 1
2 { Y, α 1 (X) } .
(5.18)where
{ , }
denotes antiommutation.Wemust takeare ofoperatorordering, beauseY
does not ommute with
α 1 (X)
.Throughout the alulation we use symmetri ordering.The time-derivativeof a funtion of anoperator, suh as
α(X, t)
, isd
dt α(X, t) = ∂
∂t α + i [H, α(X, t)] .
(5.19)In the Hamilton operator of Eq. (5.7), it is only the term
2ω R Y 2
that gives nonzeroommutator with
X
. On using that[Y, α(X)] = − i ∂α(X,t) ∂X
, one getsd
dt α(X, t) = ∂
∂t α + 4ω R 1 2
Y, ∂α(X, t)
∂X
,
(5.20)whihissymmetriallyordered.Usingthisderivativeontheleft-hand-side,andinserting
theansatzEq.(5.18)totheright-hand-sideofEq.(5.8a),wegetahierarhy ofequations
indierentpowersof theoperator
Y
.Inzerothorder,wegetthe adiabatioherenteldamplitude
α 0 (X) = η
− i∆(X) + κ .
(5.21)In rst order one gets the orretion:
α 1 (X) = 4ω R i∆ − κ
∂α 0 (X)
∂X = i 4ω R uη
(κ − i∆(X)) 3 .
(5.22)We express the photon number up to rst order in
Y
with the ansatz Eq. (5.18) asa † a = 1
2 (αα ∗ + α ∗ α) = | α 0 | 2 + 1
2 { Y, α ∗ 0 α 1 + α 0 α ∗ 1 } .
(5.23)This expansion gives rise to a frition fore in the equation of motion of
Y
Eq. (5.8),that is inthe form
F f = − 1
2 { Y, Γ(X) } ,
(5.24)wherethe frition oeient isan operator-valuedfuntion of the positionoperator
X
:Γ(X) = − 16ω R u 2 ∆(X)κη 2
(∆ 2 (X) + κ 2 ) 3 .
(5.25)Note that damping of the atomi motion is possible via the photon loss hannel, if the
detuning is anegative-valued funtion,i.e.,
h ∆(X) i < 0
, forall states.5.2.3 Eetive master equation
Puttingtheabovealulateddiusion(5.14)andfrition(5.24)intoaneetive
Heisenberg-Langevin equations,we get the replaementof the Eqs. (5.8b,), namely,
X ˙ = 4ω R Y ,
(5.26a)Y ˙ = − 4ω R X − u | α 0 (X) | 2 − 1
2 { Y, Γ(X) } + Ξ ,
(5.26b)Given the eetiveHeisenberg-Langevin equations,inthe nextstep wean onstrut an
equivalentquantum masterequationforthe densitymatrix oftheBECexitationmode,
˙
ρ = − i [H eff , ρ] + L
diρ + L
friρ .
(5.27)IntegratingthedipoleforeinEq.(5.26b)wegettheadiabatidipolepotentialwhih,
added tothe harmoni osillator energy, formsthe eetiveHamilton operator
H eff = 4ω R 1
2 (X 2 + Y 2 ) + η 2
κ arctan
uX − δ C κ
.
(5.28)It has to be emphasized, that the adiabati dipole potential is not proportional to the
intensity as inthe ase of anexternal laser potential[102℄. The reasonlies in the
bak-ationof the atoms onthe avity photoneld.
The sum of the harmonipotentialand the arus tangentterm an lead totwo loal
energy minimafor ertain settings of the parameters. This is the bistabilityregime and
the eet is losely related to the dispersive optial bistability [83, 132℄. Here, the two
minima(
X c
andX h
)areatoppositesidesoftheavityresonane.Theoneontheoolingside,
∆(X c ) < 0
, is generally situated farther from the resonane than the one on theheating side,
∆(X h ) > 0
, i. e.,| ∆(X c ) | > | ∆(X h ) |
. Therefore low photon number andshallowoptialpotentialorrespondtotheminimumontheoolingside,aordingly,the
BECis hardlymodulated. Whereasat theheatingside minimum,the photonnumberis
larger,hene the deeperoptial potentialleads toa moremodulatedondensate ground
state.
We generalize the standard diusion and frition terms that appear in the quantum
Brownian motion [133℄ for position dependent diusion and frition oeients. The
momentumdiusionduetothenoisesoure
Ξ
an bedesribedbyadoubleommutatorwith
d(X)
,L
diρ = − [d(X), [d(X), ρ]] ,
(5.29)provided the funtion
d(X)
obeysD(X) = 2
∂d(X)
∂X 2
.
(5.30)Thisdierentialequationisobtained bymathingthevariationofthemomentumsquare
mean,
h Y 2 i
in the Heisenberg-Langevin and in the master equation approahes. After integration,d(X) = η
√ κ arctan
∆(X) κ
.
(5.31)Thediusion anbeinterpretedas ameasurement-induedbak-ation onthe quantum
state of a BEC. The diusion eet of the eld on the BEC has been inluded in a
masterequation for adierentsheme where the optialeld freely propagates through
the dispersive mediumof BEC [117℄. In our sheme the measurementan be assoiated
with the irreversible detetion of photons leaking out fromthe avity.
Finally,the fritionterm in Eq. (5.26b)an bereprodued by a termlike
L
friρ = − i
2 [g(X), { Y, ρ } ]
(5.32)inthe master equation, provided
Γ(X) = ∂
∂X g(X) ,
(5.33)whih follows frommathing the variationof
h Y i
in the twopitures. The result isg(X) = − 4ω R u κη 2
(∆ 2 (X) + κ 2 ) 2 .
(5.34)Withthis,wehaveentirelydenedtheeetivemasterequation(5.27)forthebosoni
exitationmode of the BEC. The equationrelies on that the ratio
ω R /κ
is small.5.2.4 Relation with linearized models
Quantum utuationsinopto-mehanialsystemsare usuallydesribed inmodelsbased
on linearization: the state of the system is assumed to remain in a small viinity of a
stable stationary solution [134, 135, 136℄. Although the master equation (5.27) derived
in the previous subsetion does not rely on any restrition onerning the range of the
variable
X
, itis instrutive to redue our more generalmodeltothe speial ase wherethestateofthe
c 2
modeiswellloalizedaroundastationarypositionX 0
andlinearization anbeinvoked. On theother hand,our modelisrestrited tosmallvalues ofω R /κ
,thusweanmakeadiretonnetiontothe resultsof,e.g.,the referene[134℄,providedthese
latter are expanded up toleading orderin
ω R /κ
.Linearization of the master equation an beperformedaround the energy minimum
oftheeetiveHamiltonianEq.(5.28).Weassumethatweareoutofthebistableregime,
thus, in position representation, the eetive potential has a well-dened minimum at
X 0
.Considering utuations aroundthis stationary position,X = X 0 + δX ; Y = δY ,
(5.35)the linearized Hamiltoniantakes the form:
H eff ′ = 2ω R (2X 0 δX + δX 2 + δY 2 ) + uη 2 δX
(uX 0 − δ C ) 2 + κ 2 .
(5.36)In the linearized regime,the optialpotentialterm must have the only eet of shifting
theenter oftheharmoni potentialinto
X 0
.As thepotentialhas aminimumatX 0
,theterms linear in
δX
should anel. Aordingly, we get a self-onsistent equation for the equilibriumposition:X 0 = − u 4ω R
η 2
(uX 0 − δ C ) 2 + κ 2 .
(5.37)Outside the bistable regime, this third order equation has only one real solution. If
| uX 0 | ≪ κ, δ C
, then anapproximate value ofX 0
an be obtained by negleting theX 0
-dependeneofthe denominatorontherighthandside.Linearizationofthediusionand
frition terms isreadily availablefrom Eq. (5.30)and Eq. (5.33),thus
d(X) = d(X 0 ) +
r D(X 0 )
2 δX,
(5.38)g(X) = g (X 0 ) + Γ(X 0 ) δX.
(5.39)The onstant terms drop from the ommutators, and we reover the famous
Caldeira-Legett master equation [137℄ with diusion and frition oeients depending on the
stationaryposition
X 0
:˙
ρ = − i [H eff , ρ] − D(X 0 )
2 [δX, [δX, ρ]] − i
2 Γ(X 0 ) [δX, { δY, ρ } ] .
(5.40)Notethatafterthelinearization,the diusionand thefritionarenolongerdesribed by
operators,but by realnumbers. Hene,we aneasilyexpress the steady-state exitation
number of the osillator:
¯
n(X 0 ) = D(X 0 )
Γ(X 0 ) = − ∆ 2 (X 0 ) + κ 2
8ω R ∆(X 0 ) .
(5.41)Let us analyze the weak oupling limit, where
| uX 0 | ≪ κ, δ C
, and one an neglet theX 0
-dependene of the avity detuning, thus∆(X 0 ) ≈ δ C = ∆ C − NU 0 /2
. In this ase,the expression Eq. (5.41)simplies to
¯
n = (∆ C − NU 0 /2) 2 + κ 2
8ω R ( − ∆ C + NU 0 /2) .
(5.42)ThisagreeswithEq.(6)ofRef.[134℄,byusingtheorrespondenebetweenthenotations,
ω m → 4ω R
,4τ 2 → 1/κ 2
,∆ L → δ C
. Let us also note, that for a BEC, the exitationnumberof the osillator orresponds tothe populationin the
c 2
mode, thatis nothingelse but the ondensate depletion already alulated in Chapter 4. The steady-state
exitation number of the osillator Eq. (5.42), whih we obtained from the linearized
master equation, is equal to the ondensate depletion Eq. (4.12) derived in the weak
We an dedue the riterion of linearization, sine the mean exitation number
de-terminesthe magnitudeof the utuations
δX
,asn ¯ = h δX 2 i /2
.The smallparameter inthe Taylor-seriesexpansion of the arus tangent funtion was
δX u κ
"
1 +
∆(X 0 ) κ
2 # − 1
≪ 1 .
(5.43)Raising the expression to seond power, taking the average and replaing
h δX 2 i
byEq. (5.41),one gets arestrition to the strength of the atom-eldinteration:
u 2 ≪ 8ω R | ∆(X 0 ) |
"
1 +
∆(X 0 ) κ
2 #
.
(5.44)The eetive avity detuning
∆(X 0 )
needs tobein the order of the avity deay rateκ
for the avity to have onsiderableeets, henethe ondition of linearizationbeomes,
roughly,