Infinite Dimensional Analysis, Quantum Probability and Related Topics Vol. 8, No. 4 (2005) 573–591 c World Scientific Publishing Company TWO-PHOTON ABSORPTION AND EMISSION PROCESS FRANCO FAGNOLA Dipartimento di Matematica “F. Brioschi”, Politecnico di Milano, Piazza Leonardo da Vinci 32, I-20133 Milano, Italy [email protected]ROBERTO QUEZADA Departamento de Matem´ aticas, UAM-Iztapalapa, Av. San Rafael Atlixco 186, Col Vicentina, 09340 M´ exico D.F., M´ exico [email protected]Received 25 March 2005 Communicated by L. Accardi We analyze the two-photon absorption and emission process and characterize the sta- tionary states at zero and positive temperature. We show that entangled stationary states exist only at zero temperature and, at positive temperature, there exists infinitely many commuting invariant states satisfying the detailed balance condition. Keywords : Quantum dynamical semigroup; detailed balance; KMS condition. AMS Subject Classification: 46L55, 82C10, 60J27 1. Introduction The two-photon absorption and emission process is one of the most basic radiation- matter interaction mechanisms. The first steps in the development of the physical theory go back to the work of M. G¨ oppert-Mayer 17 in 1931 but the phenomenon of two-photon absorption was not observed until 1961 (see Ref. 16), after the advent of the laser, in fact two-photon absorption is one of the first phenomena demon- strated with the aid of laser radiation. Since then, this phenomenon has been studied intensively. Gilles and Knight, 15 introduced a model based on a quantum Markov semigroup and found nonclassical stationary states. Here we consider a general- ization of this model. In Sec. 8 we outline the deduction of our model for the two-photon absorption or absorption and emission processes, from the stochastic limit of the evolution of the system (one-mode EM field) weakly coupled with a boson reservoir (see Ref. 3 for the general theory of the stochastic limit). 573
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
We analyze the two-photon absorption and emission process and characterize the sta-tionary states at zero and positive temperature. We show that entangled stationarystates exist only at zero temperature and, at positive temperature, there exists infinitelymany commuting invariant states satisfying the detailed balance condition.
When ν = λ/µ < 1 we can easily find two invariant states. Indeed, a straightforward
computation yields the states ρe, ρo defined by
ρe = (1 − ν2)∑
k≥0
ν2k|e2k〉〈e2k| , ρo = (1 − ν2)∑
k≥0
ν2k|e2k+1〉〈e2k+1| .
Proposition 3.1. The states ρe and ρo are invariant.
Proof. Let L∗ be the generator of the predual semigroup T∗ = (T∗t)t≥0, acting on
the Banach space L1(h) of trace class operators on h. Consider the approximations
ρe,n = (1 − ν2)∑n
k=0 ν2k |e2k〉〈e2k|, of ρe by finite range operators.
The operators ρe,n belong to the domain of L∗ and we have L∗(ρe,n) = λ2(2n+
1)(2n+2)ν2n(|e2n+1〉〈e2n+1|− |e2n〉〈e2n|). Thus, for all m < n, we have ‖L∗(ρe,n −ρe,m)‖1 = 2λ2((2n+1)(2n+2)ν2n +(2m+1)(2m+2)ν2m) which converges to zero
as n, m → ∞ since ν < 1. Similar computations and conclusion hold for ρo,n. This
proves that ρe, ρo ∈ Dom(L∗) and L∗(ρe) = 0 = L∗(ρo).
In Sec. 6 we shall prove that, when λ > 0, all the invariant states are convex
combination of ρo and ρe. When λ = 0, we can easily find all the invariant states.
November 9, 2005 11:49 WSPC/102-IDAQPRT 00211
576 F. Fagnola & R. Quezada
Proposition 3.2. If λ = 0 all invariant states have the form αρe + (1 − α)ρo +
z|e0〉〈e1| + z|e1〉〈e0|, with α ∈ [0, 1] and |z|2 ≤ α(1 − α).
Proof. Let σ be an invariant state. Then L∗(σ) = 0, i.e.
Now, since 2(3n+1) ≤ 8n for n ≥ 1, we can apply Theorem 3.1.34, p. 193 of Ref. 4
to conclude.
Starting from T , we can define in the same way T α and Lα by T αt (ρ
1−θ
2
α xρθ
2
α ) =
ρ1−θ
2
α Tt(x)ρθ
2
α . The linear manifold ι(M) is also a core for Lα (Lα and Lα differ only
by the sign of ω).
November 9, 2005 11:49 WSPC/102-IDAQPRT 00211
Two-Photon Absorption and Emission Process 581
The following result establishes the duality between T α and T α.
Theorem 5.1. The semigroups T α, T α are the dual on L2(h), i.e. for all t ≥ 0
and all x, y ∈ L2(h) we have
tr(T αt (y)x) = tr(yT α
t (x)) . (5.4)
In particular (5.3) holds for all x, y ∈ B(h).
Proof. Since ρα is a function of the number operator, it commutes with the oper-
ators G and G∗. Moreover, an algebraic computation using the canonical commu-
tation relations shows that
ρθ/2α L∗
1 = ν2θ−1L2ρθ/2α , L1ρ
(1−θ)/2α = ν1−2θρ
(1−θ)/2α L∗
2 ,
ρθ/2α L∗
2 = ν1−2θL1ρθ/2α , L1ρ
(1−θ)/2α = ν2θ−1ρ
(1−θ)/2α L∗
1 .(5.5)
Therefore, for all x, y ∈ M, we have the identities
tr(ρ(1−θ)/2α yρθ/2
α (G∗x + xG)) = tr(ρ(1−θ)/2α (Gy + yG∗)ρθ/2
α x) ,
tr(ρ(1−θ)/2α yρθ/2
α (L∗1xL1 + L∗
2xL2)) = tr(ρ(1−θ)/2α (L∗
1yL1 + L∗2yL2)ρ
θ/2α x)
showing that tr(ρ(1−θ)α yρθ
αL(x)) = tr(ρ(1−θ)α L(y)ρθ
αx) and, for all r > 0,
tr((r − Lα)(ι(y))ι(x)) = tr(ι(y)(r − Lα)(ι(x))) .
It follows then that, for all x ∈ Dom(Lα) and y ∈ Dom(Lα), we have
tr((r − Lα)(y)x) = tr(y(r − Lα)(x)) .
Taking the resolvents, we find the identity
tr(y(r − Lα)−1(x)) = tr((r − Lα)−1(y)x)
for all x, y ∈ L2(h). Therefore, for all t > 0 and n ≥ 1 we obtain
tr(y(nt−1 − Lα)−n(x)) = tr((nt−1 − Lα)−n(y)x) .
The duality formula (5.4) follows from the Trotter–Kato formula letting n tend to
infinity. Replacing the operators x, y by ρθ/2α xρ
(1−θ)/2α , ρ
(1−θ)/2α yρ
θ/2α with x, y ∈ M,
in this formula we find
tr(ρ(1−θ)α T α
t (y)ρθαx) = tr(ρ(1−θ)
α yρθαT α
t (x)) .
Now (5.3) follows from the weak* density of M in B(h).
November 9, 2005 11:49 WSPC/102-IDAQPRT 00211
582 F. Fagnola & R. Quezada
6. Characterization of the Invariant States for λ > 0
In this section we show that ρe and ρo determine all invariant states of T . Indeed,
we prove the following
Theorem 6.1. If λ > 0, then all invariant states of T are convex linear combina-
tions of ρe and ρo.
Let σ a fixed normal T -invariant state and let (σjk)j,k≥0 be its matrix elements
in the canonical orthonormal basis. As a first step we show that the diagonal part
of σ has the desired form.
Lemma 6.1. Let σ be a normal T -invariant state. Then we have
peσpe = tr(σpe)ρe , poσpo = tr(σpo)ρo . (6.1)
As a consequence, the diagonal part σd =∑
j σjj |ej〉〈ej | of σ is given by
σd = tr(σpe)ρe + tr(σpo)ρo .
Moreover, we have the inequalities |σjk | ≤ c(σ, ν)ν(j+k)/2 for all j, k, with c(σ, ν)
constant depending only on σ, ν.
Proof. We know from Theorem 4.5 that
w∗ − limt→∞
T et (pexpe) = tr(ρex)pe , w∗ − lim
t→∞T o
t (poxpo) = tr(ρox)po
for any x ∈ B(h) since Ae (resp. Ao) is the dual of states supported in pe (resp. po).
By the invariance of σ we have then
tr(σpexpe) = tr(σTt(pexpe)) = limt→∞
tr(σT et (pexpe)) = tr(ρex)tr(σpe) .
In the same way we find tr(σpoxpo) = tr(ρox)tr(σpo). Therefore, since x is arbitrary,
we obtain (6.1).
The positivity of σ implies |σjk |2 ≤ σjjσkk for all j, k. Therefore, since the
diagonal part of σ is a convex combination of ρe and ρo, the last claim follows.
We are now in a position to prove Theorem 6.1.
Proof of Theorem 6.1. Let σ be a T -invariant state. Then L∗(σ) = 0. Writing
σ =∑
j,k≥0 σjk |ej〉〈ek| we find
L∗(σ) =∑
j,k≥0
|ej〉〈ek|{
iω(k[k − 1] − j[j − 1])σjk − µ2
2(k[k − 1] + j[j − 1])σjk
+ µ2(k + 1)1
2 (k + 2)1
2 (j + 1)1
2 (j + 2)1
2 σj+2k+2
+ λ2k1
2 [k − 1]1
2 j1
2 [j − 1]1
2 σj−2k−2
− λ2
2((j + 1)(j + 2) + (k + 1)(k + 2))σjk)
}
November 9, 2005 11:49 WSPC/102-IDAQPRT 00211
Two-Photon Absorption and Emission Process 583
with the understanding that [k − 1] = max{(k − 1), 0}, [j − 1] = max{(j − 1), 0}.We shall prove that σjk = 0 for all j 6= k. Clearly, since σ is hermitian, it suffices
to check that σjk = 0 for all j > k. Letting j = k + n, with n ≥ 1, the identity
L∗(σ) = 0 yields
0 = −iωn(2k + n − 1)σk+n,k − µ2(k[k − 1] + (k + n)(k + n − 1))σk+n,k
−λ2((k + 1)(k + 2) + (k + n + 1)(k + n + 2))σk+nk
+ 2µ2(k + 1)1
2 (k + 2)1
2 (k + n + 1)1
2 (k + n + 2)1
2 σk+n+2k+2
+ 2λ2k1
2 [k − 1]1
2 (k + n)1
2 (k + n − 1)1
2 σk+n−2k−2
for all k, n ≥ 0. Fix n and put yk = ν−k/2σk+nk . Notice that the sequence (yk)k≥0
is square summable because, by Lemma 6.1, |yk| ≤ c(σ, ν)ν(k+n)/2. Multiplying by
yk, summing on k and taking the real part, the above equation reads as
0 = −µ2∑
k≥0
k[k − 1]|yk|2 − µ2∑
k≥0
(k + n)(k + n − 1)|yk|2
+ 2λµ Re∑
k≥0
(k + 1)1
2 (k + 2)1
2 (k + n + 1)1
2 (k + n + 2)1
2 ykyk+2
−λ2∑
k≥0
(k + 1)(k + 2)|yk|2 − λ2∑
k≥0
(k + n + 1)(k + n + 2)|yk|2
+ 2λµ Re∑
k≥0
k1
2 [k − 1]1
2 (k + n)1
2 (k + n − 1)1
2 ykyk−2
= −µ2n(n − 1)|y0|2 − µ2n(n + 1)|y1|2
−µ2∑
k≥2
k[k − 1]|yk|2 − µ2∑
k≥0
(k + n + 1)(k + n + 2)|yk+2|2
+ 2λµ Re∑
k≥0
(k + 1)1
2 (k + 2)1
2 (k + n + 1)1
2 (k + n + 2)1
2 ykyk+2
−λ2∑
k≥0
(k + 1)(k + 2)|yk|2 − λ2∑
k≥2
(k + n − 1)(k + n)|yk−2|2
+ 2λµ Re∑
k≥2
k1
2 [k − 1]1
2 (k + n)1
2 (k + n − 1)1
2 ykyk−2 .
Reconstructing a square from the three sums for k ≥ 0 and another from the three
sums on k ≥ 2 we find
0 = −µ2n(n − 1)|y0|2 − µ2n(n + 1)|y1|2
−∑
k≥0
∣
∣
∣λ(k + 1)
1
2 (k + 2)1
2 yk − µ(k + n + 1)1
2 (k + n + 2)1
2 yk+2
∣
∣
∣
2
November 9, 2005 11:49 WSPC/102-IDAQPRT 00211
584 F. Fagnola & R. Quezada
−∑
k≥2
∣
∣
∣λ(k + n − 1)1
2 (k + n)1
2 yk−2 − µk1
2 [k − 1]1
2 yk
∣
∣
∣
2
= −µ2n(n − 1)|y0|2 − µ2n(n + 1)|y1|2
−∑
k≥0
∣
∣
∣λ(k + 1)1
2 (k + 2)1
2 yk − µ(k + n + 1)1
2 (k + n + 2)1
2 yk+2
∣
∣
∣
2
−∑
k≥0
∣
∣
∣λ(k + n + 1)1
2 (k + n + 2)1
2 yk − µ(k + 2)1
2 (k + 1)1
2 yk+2
∣
∣
∣
2
.
Now, for n > 1, this implies y0 = y1 = 0 and then, by induction yk = 0 for all
k ≥ 0. If n = 1 then y1 = 0 and yk = 0 for all odd k’s by induction. Moreover, if
the two sums of squares are 0, then
yk+2 = ν(k + 1)1
2 (k + 3)−1
2 yk , yk+2 = ν(k + 1)−1
2 (k + 3)1
2 yk
for all k ≥ 0. This can happen only if yk = 0. Therefore σ is diagonal.
7. Approach to Equilibrium
In this section we study convergence of states T∗t(σ) towards an invariant state and
determine the domains of attraction of the invariant states. As a first step, with
the notation of Sec. 4, we prove the following.
Lemma 7.1. Suppose λ > 0. Then F(T ) = N (T ).
Proof. The QMS T has a faithful normal invariant state. Therefore the inclusion
F(T ) ⊆ N (T ) always holds. It suffices then to prove the opposite.
By Theorem 4.4 we have N (T ) = {a2, a+2}′ and F(T ) = {a2, a+2, a+2a2}′. A
x ∈ N (T ), satisfies then a2x ⊆ xa2 and a+2x ⊆ xa+2. It follows that
therefore f ′(σ, t) = 0 for every σ ∈ M if and only if c2m
c2m−2
= 2−1(2m − 1)(m −iωµ2 )−1, for m ≥ 0. Taking c0 = 1 we obtain after some computations that c2m =
2−2m (2m)!m! Πm
j=1(j − iωµ2 )−1. With this choice of the coefficients c2m, we have that
f(σ, t) is a constant function of t for every σ ∈ M.
Now using Wallis’s product formula one can show that the sequence√
2m + 1c2m
is bounded by a positive constant c. Then for a positive σ ∈ M, the inequality
|pjk(t)|2 ≤ pjj(t)pkk(t) yields
f(σ, t) ≤ c∑
m≥0
|p2m,2m+1(t)| ≤ c∑
m≥0
p1
2
2m(t)p1
2
2m+1(t)
≤ c
∑
m≥0
p2m(t)
1
2
∑
m≥0
p2m+1(t)
1
2
≤ c tr(σ) .
Since every positive σ ∈ L1(h) can be approximated by an increasing sequence of
positive elements (σn) ⊂ M, f(σ, t) can be extended continuously to the whole
L1(h). We obtain that f(σ, t) is also a constant function of t and consequently
z = limt→∞
θ0(t) = limt→∞
fσ(t) = f(σ, 0) =∑
m≥0
c2m
√2m + 1〈e2m, σe2m+1〉 .
8. Deduction from the Stochastic Limit
In this section we outline the deduction of the two-photon absorption quantum
Markov semigroup from the stochastic limit of the evolution of a system (one-mode
EM field) coupled with a boson, zero temperature reservoir.
The state space of the system, a one-mode electromagnetic field, is the complex
separable Hilbert space h = `2(N). The free evolution of the system is given by
a strongly continuous unitary group (e−itHS )t∈R, where HS = E(N) is a positive
function E (E : Z → [0, +∞[) of the number operator N = a+a. The state space
of the reservoir (zero temperature boson gas) is the boson Fock space F over a
complex separable Hilbert space k1 (the one-particle space of the reservoir). The
free evolution of one particle in the reservoir is given by a strongly continuous one
parameter group (S0t )t∈R of unitary operators on k1 enjoying the following property:
there exists a dense subspace k of k1 such that
∫
R
|〈g, S0t f〉|dt < ∞
for all f, g ∈ k. The free evolution of the reservoir is given by the unitary group
obtained by second quantization (Γ(S0t ))t∈R of the unitaries S0
t on k1. This is a
strongly continuous unitary group and its generator HR, the self-adjoint operator
on F such that e−itHR = Γ(S0t ) for all t ∈ R, is the Hamiltonian of the reservoir.
November 9, 2005 11:49 WSPC/102-IDAQPRT 00211
588 F. Fagnola & R. Quezada
The evolution of the whole system is given by the unitary group generated by
the total Hamiltonian
Hλ = HS ⊗ 1F + 1S ⊗ HR + λV ,
where λ is a real positive parameter and V is an interaction operator such that Hλ
is self-adjoint for all λ > 0.
Suppose that the interaction operator (of dipole type) has the form
Vg = i(ad ⊗ A∗(g) − a+d ⊗ A(g)) ,
where d ∈ N∗, A(g), A∗(g) are creation and annihilation operators on F with g ∈ k1.
A straightforward computation using the commutation relations
eitE(N)ad = adeitE(N−d)
shows that generalized rotating wave approximation
eitHSade−itHS = e−iω0tad ,
where ω0 > 0 (see Ref. 3, Definition 4.10.1 on p. 125) holds, with ω0 = E(n− d)−E(n) for all n ∈ N if and only if E is linear. This is the case, for example, when
HS = a+a.
Suppose that the generalized rotating wave approximation holds. The sesquilin-
ear form on k
(f |g) :=
∫
R
〈g, S0t f〉dt
is positive (see Ref. 3). Therefore it defines a pre-scalar product on k. We denote
by K the Hilbert space obtained by quotient and completion; the scalar product
will be denoted by (·|·). Defining
U(λ)t = eitH0e−itHλ
a straightforward computation shows that the family of unitaries (U(λ)t )t≥0 on h⊗F
satisfies the differential equation
d
dtU
(λ)t = −iλVg(t)U
(λ)t , U
(λ)0 = 1l ,
where Vg(t) = i(D ⊗ A∗(Stg) − D∗ ⊗ A(Stg)) and St = eitω0S(0)t .
Let W (f) (f ∈ k1) denote the unitary Weyl operators on F acting on exponen-
tial vectors as
W (f1)e(f2) = e−‖f1‖2/2−〈f1,f2〉e(f1 + f2) .
The basic idea of Accardi, Frigerio and Lu1 was to study the result of small
interactions (λ → 0) on a large time scale (time goes to infinity). This was realized
by scaling time by λ2 and space by λ and letting λ tend to 0. As a result of the
limiting procedure, the state space of the whole system also changes.
The following result allows us to find the structure of the space of the limit
evolution.
November 9, 2005 11:49 WSPC/102-IDAQPRT 00211
Two-Photon Absorption and Emission Process 589
Proposition 8.1. For all n, n′ ∈ N and all f1, . . . , fn, f ′1, . . . , f
′n′ ∈ K,
s1, t1, . . . , sn, tn, s′1, t′1, . . . , s
′n′ , t′n′ ∈ R with sk ≤ tk, s′k ≤ t′k for all k denote
W (f1, . . . , fn) = W
(
λ
∫ λ−2t1
λ−2s1
Sr1f1dr1
)
· · ·W(
λ
∫ λ−2tn
λ−2sn
Srnfndrn
)
.
We have then
limλ→0
〈W (f1, . . . , fn)0, W (f ′1, . . . , f
′n)0〉
= 〈W (f1 ⊗ 1[s1,t1]) · · ·W (fn ⊗ 1[sn,tn])0 ,
W (f ′1 ⊗ 1[s′
1,t′
1]) · · ·W (f ′
n′ ⊗ 1[s′
n′,t′
n′])e(0)〉 ,
where W (f1 ⊗ 1[s1,t1]), . . . , W (f ′n′ ⊗ 1[sn′ ,tn′ ]) are Weyl operators in the boson Fock
space over L2(R+;K).
It follows that the state space of the limit evolution is the tensor product of the
initial space h with the boson Fock space Γ(L2(R+;K)). The limit of unitaries is
given in the following theorem.
Theorem 8.1. For all v, u ∈ h, n, n′ ∈ N and all f1, . . . , fn, f ′1, . . . , f
′n′ ∈ K, s1,
t1, . . . , sn, tn, s′1, t′1, . . . , s′n′ , t′n′ ∈ R with sk ≤ tk, s′k ≤ t′k we have
limλ→0
〈vW (f1, . . . , fn)e(0), U(λ)
λ−2tuW (f ′1, . . . , f
′n)e(0)〉
= 〈W (f1 ⊗ 1[s1,t1]) · · ·W (fn ⊗ 1[sn,tn])e(0) ,
UtuW (f ′1 ⊗ 1[s′
1,t′
1]) · · ·W (f ′
n′ ⊗ 1[s′
n′,t′
n′])e(0)〉 ,
where U is the unique unitary process satisfying the quantum stochastic differential
9. F. Fagnola and R. Rebolledo, Lectures on the qualitative analysis of quantum Markovsemigroups, Quantum Interacting Particle Systems, QP–PQ: Quantum Probab. WhiteNoise Anal., Vol. 14 (World Scientific, 2002).
10. F. Fagnola and R. Rebolledo, Subharmonic projections for a quantum Markov semi-group, J. Math. Phys. 43 (2002) 1074–1082.
11. A. Frigerio, Quantum dynamical semigroups and approach to equilibrium, Lett. Math.
Phys. 2 (1977) 79–87.12. A. Frigerio, A. Kossakowski, V. Gorini and M. Verri, Quantum detailed balance and
KMS condition, Commun. Math. Phys. 57 (1977) 97–110.13. A. Frigerio and M. Verri, Long-time asymptotic properties of dynamical semigroups
on W∗-algebras, Math. Z. 180 (1982) 275–286.
14. J. C. Garcıa and R. Quezada, Hille–Yosida estimate and nonconservativity criteriafor quantum dynamical semigroups, Inf. Dim. Anal. Quantum Probab. Rel. Topics 7