Chapter IV: Atoms coupled to a quasi-1D nanostructure
4.1 The low-saturation atomic system
In Sec. 3.4, we derived an expression for the electric field operator and a low- saturation system of equations for the atomic dipole moments. These expressions were in terms of the free-field (the field not coming from the atoms) and the full EM Green’s function. We use these two expressions, but make one change of convention. As discussed in Chapter 2, the Green’s function can be decomposed into a part coming from the guided modes and a part coming from the non-guided (i.e. free-space) modes:
G(ri,rj, ω)= Gwg(ri,rj, ω)+G0(ri,rj, ω). (4.1) Here the first term corresponds to the guided modes that propagate along the struc- ture, and the second term accounts for all other modes (e.g. emission into free- space). For many systems,G0(ri,rj, ω)is similar to the free-space Green’s function from Sec. 2.2, and therefore has a fast spatial decay. This is because the quasi-1D dielectric structure does not strongly alter the free-space modes. When the atoms are spaced by more than a wavelength, we can approximate
G0(ri,rj, ω) ≈ G0(ri,ri, ω)δi j, (4.2)
or equivalentlyΓi j = Γi jwg +Γii0δi j and Ji j = Ji jwg +Jii0δi j, whereδi j is the Kronecker delta. We assume that J0 is identical for every atom and incorporate it into the definition of ωA. That is the case for example if all the atoms are at the same position relative to the dielectric structure (see Ref. [105] for more detail).
In the Fourier domain, the electric field operator is still (see Eq. (3.51)) Eˆ+(r, ω)=Eˆ+free(r, ω) −
N
Õ
j=1
µ0ω2G(r,rj, ω)djσˆj(ω). (4.3)
The first term is the free-field from some input source, and the second term is the emitted field from each of the N atoms. It is important to remember that this expression was derived without using the Markov approximation, and the results that follow in this chapter are equally valid in the non-Markovian limit, e.g. in strong-coupling limit of CQED. As long asr,rj, we can replace the total Green’s function with the waveguide Green’s functionGwg.
In the low saturation regime the atoms are represented by a linear system of equations (see Eq. 3.52),
∆A+ i 2Γ0
σˆi(ω)+
N
Õ
j=1
gi j σˆj(ω)= 1
~di·Eˆ+free(ri, ω). (4.4) Here, the detuning from the atomic resonance frequency is∆A =(ω−ωA), and the complex waveguide coupling rates are
gi j = Ji j + i
2Γi j = µ0ω2
~ d∗i Gwg(ri,rj, ω)dj, (4.5) where the real and imaginary parts are the decay ratesΓi j and frequency shifts Ji j
given by
Ji j = µ0ω2
~ Re[di∗Gwg(ri,rj, ω)dj] (4.6) Γi j = 2µ0ω2
~ Im[d∗i Gwg(ri,rj, ω)dj]. (4.7) Note that we have omitted the waveguide superscripts onJandΓ, which will be the convention from now on. The decay rateΓ0in Eq. (4.4) represents the non-collective decay of the atoms due to non-guided modes.
The linear systems of equations in Eq. (4.4) can be solved by inverting the left side.
The solution for the atomic operators ˆσj can then be substituted into the electric
field operator in Eq. (4.3) to obtain the solution for the field at any position. But before proceeding, we simplify the notation of the system of equations and electric field operator. First, we express the free fieldE+free(rk, ω)as if it is being emitted by some classical dipoledinat a positionrinσinwhich is far outside the system,
Eˆ+free(r, ω)= µ0ω2G(r,rin, ω)dinσin. (4.8) For example, plane-wave illumination can be achieved by having the dipole very far away in free-space. Or an incoming guided mode field can be achieved by having the dipole far away and coupled to the guided mode (since the free-space component is shorter range and all we are left with is the guided component). The atomic system of equations in Eq. (4.4) is then (usinggk,in = µ0~ω2dkG(rk,rin)din)
(∆A+iΓ0/2)σk+
N
Õ
j=1
gk jσj = −gk,inσin, (4.9) and the electric field operator in Eq. (4.3) is
Eˆ+(r, ω)= µ0ω2
G(r,rin, ω)dinσin−
N
Õ
j=1
G(r,rj, ω)djσˆj(ω)
. (4.10)
The second change of notation is to express the electric field operator in a scalar form. We can accomplish this by multiplying the both sides of Eq. (4.10) by an arbitrary dipole vector d∗out. We can interpret this as if we are using a negligible dipoledoutat positionroutin order to to measure the electric field. The vector can be removed at the end of a calculation, but including it lets the output electric field be conveniently expressed in terms of the complex coupling ratesgi j as
gout,inσin+
N
Õ
j=1
gout,jσj = 1
~d∗out·E+(rout, ω). (4.11) Or written in matrix form and defining the vector of atomic operatorsσ=(σˆ1, · · ·, σˆN), Eq. (4.9) and (4.10) can be expressed as
h
(∆A+iΓ0/2)1+g i
·σ= −ginσin (4.12)
gout,inσin+gout·σ= 1
~d∗out·E+(rout, ω), (4.13) where the matrix(g)j k = gj k contains the propagators between the atoms, the vector gin = (g1,in, · · ·, gN,in)contains the propagators between the source and the atoms,
and the vectorgout=(gout,1, · · ·, gout,N)contains the propagators from the atoms to the output.
The solution for the atomic operator for a given input field is found by inverting Eq. (4.12) ,
σ= −
1
(∆A+iΓ0/2)1+g
·ginσin. (4.14) For a system with a single atom and a drive with amplitude proportional tog1,in, the atom’s dipole operator is
σ1= − g1,in
∆A+J11+ 2i(Γ0+Γ11). (4.15) The resonance frequency is shifted by J11and the total decay rate isΓtot = Γ0+Γ11. When multiple atoms interact with each other, g is not diagonal, and the atoms instead behave collectively. We can no longer talk about the decay rate or frequency shift of a single atom. However, we will show that we can instead use the frequency shifts and decay rates for the collective atomic modes which are uncoupled from each other. Due to the reciprocity identity from Eq. (2.15) [GT(r,r0, ω) = G(r0,r, ω)], the the matrixg is symmetric (gT =g) when the dipole matrix elements are real.1It is interesting to note that because of the dissipation, the matrixg is not Hermitian and the eigenvectors are not orthogonal in the more traditional sense which uses the conjugate transpose. Instead, as shown in Appendix C, complex symmetric matrices have eigenvectors defined by the eigenvalue equation
gvξ = λξvξ (4.17)
that satisfy the orthogonality and completeness relation vTξ ·vξ0 =δξξ0 and
N
Õ
ξ=1
vξ⊗vTξ = 1. (4.18) Here T is the transpose, rather than the more customary conjugate transpose for Hermitian or normal matrices. The matrix in Eq. (4.14) shares the same eigenvectors
1 The orthogonality and completeness relation of the eigenmodes rely on (g)i j =
µ0ω2
~ d∗iG(ri,rj, ω)dj being a complex symmetric matrix, gT =g. Taking the transpose ofgand using the reciprocity identity for the Green’s function,GT(ri,rj, ω)=G(rj,ri, ω), gives
(gT)i j = µ0ω2
~
diG(ri,rj, ω)d∗j. (4.16) Comparing this to(g)i j, we see that the only difference is that the complex conjugate is on the second dipole moment. The matrix is symmetric (g=gT)when 1) all the dipole moments are real 2) the Green’s function is linearly polarized.
asg, since the other term in the denominator is diagonal. Inserting the completeness relation into the right side of Eq. (4.14), using the eigenvalue equation gives an expansion of the dipole operator in terms of the Ncollective modes,
σ= −
1
∆A+iΓ0/2+g
·©
«
N
Õ
ξ=1
vξ⊗vTξª
®
¬
ginσin (4.19)
= −
N
Õ
ξ=1
vξ ·ginσin
∆A+iΓ0/2+λξvξ. (4.20) Defining the real and imaginary parts of the eigenvalue as λξ = Jξ + iΓξ/2 as the frequency shift and decay rate for the collective mode ξ, where the Greek letter signifies a collective mode rather than a single atom, the final expression for expansion is
σ=−
N
Õ
ξ=1
vξ·ginσin
∆A+ Jξ +i(Γ0+Γξ)/2vξ. (4.21) The elements of eigenvectorvξgive the relative amplitude and phase for the atoms in the collective mode. As a result of the interactions, the collective mode is shifted by Jξ, and the total decay rate is either inhibited or enhanced toΓtot= Γ0+Γξ. Finally, the coupling between the input field and a particular collective mode is given by the scalar productvξ·ginσin.