• Aucun résultat trouvé

Lyapunov control of Schrödinger equations: beyond the dipole approximation

N/A
N/A
Protected

Academic year: 2021

Partager "Lyapunov control of Schrödinger equations: beyond the dipole approximation"

Copied!
1
0
0

Texte intégral

(1)

Bibliography [1] M. MIRRAHIMI, P. ROUCHON, AND G.TURINICI, Lyapunov control of bilinear Schr¨odinger

equations AUTOMATICA, 41 :1987 1994, 2005.

[2] H.K. KHALIL. Nonlinear Systems. MACMILLAN, 1992.

[3] H. RABITZ. AND W. ZHU. Quantum control design via adaptive tracking. J. CHEM. PHYS., 119(7), 2003.

Convergence analysis

Theorem 2. Consider (2) with Ψ ∈ S2n−1C and an eigen-state φ ∈ S2n−1C of H0 associated to the eigenvalue λ. Take the feedback (9)

with c, k > 0. If H0 is not degenerate and either hΦα|H1φi 6= 0 or hΦα|H2φi 6= 0 ∀α, Φα 6= φ, , then the Ω-limit set of S1Ψ0 reduces to a solution of the uncontrolled system, with |I1|, |I2| ≤ Cδ with a constant C depending only on H0.

Proof of Theorem 2

• Observe that neither S1 or S2 define a classical dynamical system, that is the semigroup property is lost. One has instead S1(t + s)Ψ0 = S1(t)S1(s)Ψ0 or S1(t+s)Ψ0 = S2(t)S1(s)Ψ0 according to the control we have when touching S1(s)Ψ0.

• On the limit set Ω(Ψ0) corresponding to S1, V is constant and Ω(Ψ0) is invariant either to S1 or to S2, that is if Ψ1 ∈ Ω(Ψ0) then S1(t)Ψ1 ∈ Ω(Ψ0), t > 0 or S2(t)Ψ1 ∈ Ω(Ψ0), t > 0. But then, since V is constant along the trajectory, u = 0 and we have that Ω(Ψ0) is in-variant under the Schr¨odinger flow for H0.

• On Ω(Ψ0) we have that |I1| ≤ δ and I2 ≥ −δ. Since Ω(Ψ0) is a trajectory to the uncontrolled equation, we have that there exists a constant C > 1 depending only on H0 such that I2 ≤ Cδ.

Remark 5. Now, the stabilizing strategy is obvious since, when δ → 0, the trajectories in Ω(Ψ0) = Ωδ0) converge to a trajectory of the uncontrolled system with I1 = I2 = 0.

FIG. 4 – Population |Ψ1|2(solid line) and control u (dashed line) ;

ini-tial condition : Ψ(t = 0) = (0, 1/√2, 1/√2) ; system defined by (8) with feedback (9). We take k = c = 0.8 and δ = 10−3.

FIG. 5 – Time evolution of I1,I2 ; initial condition : Ψ(t = 0) =

(0, 1/√2, 1/√2) ; system defined by (8) with feedback (9). We take k = c = 0.8 and δ = 10−3 ; I1, I2 oscilates and its absolute value converges to zero.

Examples and simulations

We use the previous Lyapunov control in order to reach the first eigen-state

φ = Φ1 = (1, 0, 0) of energy λ = 0, at the final time T .

Remark 2. For the system defined by (8) ,we note that hΦ3|H1φi = 0.

FIG. 2 – Population |Ψ1|2 (blue line) and control u (green line) ; initial

condi-tion Ψ(t = 0) = (0, 1/√2, 1/√2) ; system defined by (8) with feedback (6). We take k = c = 0.8 The feedback (6) fails to reach the target (quality is only 30%).

In order to overcome the lack of convergence for cases similar to that pre-sented above, we consider the feedback :

u(I1, I2) =    kI2, in A = {|I1| < δ and I2 < −δ} 0 , in B = { |I1| < δ and I2 > δ} −kI1/(1 + kI2), in C = {|I1| > δ/2 or |I2| < 2δ}. ω = −λ − cIm(hΨ(t) | φi). (9)

Remark 3. With c > 0, k > 0 and δ > 0 we ensure : dV /dt ≤ 0, i.e V is decreasing.

Remark 4. The regions A, C, respectively B, C are overlapping such that ∂A ⊂ intB, ∂B ⊂ intA and, respectively, ∂B ⊂ intC, ∂C ⊂ intB.

• We define the evolutor S1Ψ0 by solving the feedback equation such that :

• if the initial state Ψ0 ∈ A∩C, we initiate with the feedback corresponding to C

• and if the initial state Ψ0 ∈ B ∩ C, we initiate also with the feedback corresponding to C.

• We define also the evolutor S2Ψ0 by solving the feedback equation such that :

• if the initial state Ψ0 ∈ A∩C, we initiate with the feedback corresponding to A

• and if the initial state Ψ0 ∈ B ∩ C, we initiate also with the feedback corresponding to B.

FIG. 3 – u(I1, I2), u defined by (9), for δ = 0.5.

Examples and simulations

Numerical simulations have been performed for a three-dimensional test sys-tem with H0, H1 and H2 given by :

H0 =   0 0 0 0 1 0 0 0 32   , H1 =   0 1 1 1 0 0 1 0 0  , H2 =   0 0 1 0 0 0 1 0 0   . (7)

In this case the wave function is Ψ = (Ψ1, Ψ2, Ψ3)T . We use the previous Lyapunov control in order to reach the first eigen-state φ = Φ1 = (1, 0, 0) of

energy λ = 0, at the final time T .

Remark 1. We note that the conditions of T heorem 1 are fulfilled since : hΦ2|H1φi 6= 0 and hΦ3|H1φi 6= 0.

FIG. 1 – Population |Ψ1|2 (blue line) and control u (green line) ; initial

condi-tion Ψ(t = 0) = (0, 1/√2, 1/√2) ; system defined by (8) with feedback (6). We take k = c = 0.8 .

We consider another three-dimensional test system with H0, H1 and H2 gi-ven by : H0 =   0 0 0 0 1 0 0 0 32   , H1 =   0 1 0 1 0 0 0 0 0  , H2 =   0 0 1 0 0 0 1 0 0   . (8)

Convergence analysis

We denote by λi, with i = 1, ..., N the eigenvalues of the matrix H0. We say that H0 has non degenerate spectrum if for all (i, j), with i 6= j, i, j = 1, ..., N , λi 6= λj.

Theorem 1. Consider (2) with Ψ ∈ S2n−1C and an eigen-state φ ∈ S2n−1

C of

H0 associated to the eigenvalue λ. Take the feedback (6) with c, k > 0. Then the two following propositions are true :

1. If the spectrum of H0 is not degenerate, the Ω-limit set of the closed loop system is the intersection of S2n−1 with the vector space E = Rφ Sα α where Φα is any eigenvector of H0 not co-linear to φ such that hΦα|H1φi = 0.

2. If H0 is not degenerate and hΦα|H1φi 6= 0, ∀α, Φα 6= φ, , the Ω-limit set reduces to {φ, −φ}.

Lyapunov control design

For convenience we denote : I1 = Im(hH1Ψ(t)|φi) and I2 = Im(hH2Ψ(t)|φi. Then (5) becomes the feedback :

u = −kI1/(1 + kI2)

ω = −λ − cIm(hΨ(t) | φi). (6)

Lyapunov control design

Take the following time varying function V (Ψ, t) :

V (Ψ, t) = hΨ − Ψr|Ψ − Ψri (3)

where

• h.|.i denotes the Hermitian product,

• t 7→ (Ψr(t), ur(t), ωr(t)) a reference trajectory i.e., a smooth solution of (2). The fonction V

is positive for all t > 0 and all Ψ ∈ S2n−1

C ,

• vanishes when Ψ = Ψr.

The derivative of V is given by : dV

dt = 2(u − ur) 

Im(hH1Ψ(t)|Ψri) + (u + ur)Im(hH2Ψ(t)|Ψri)+

2(ω − ωr)Im(hΨ(t) | Ψri), (4)

where Im denotes the imaginary part. When for instance we take :

u = ur(t) − kIm(hH1Ψ(t)|Ψri)(1 + kIm(hH2Ψ(t)|Ψri))

ω = ωr(t) − cIm(hΨ(t) | Ψri), (5)

with k and c strictly positive parameters, we ensure dV /dt ≤ 0, i.e. V is de-creasing.

In the following we will focus on the important case when the reference trajectory corresponds to an equilibrium :

ur = 0, ωr = −λ and Ψr = φ

where φ is an eigen-vector of H0 associated to the eigenvalue λ ∈ R.

General settings

We consider a n-level quantum system (~ = 1) evolving under the equation : i d

dtΨ(t) = (H0 + u(t)H1 + u

2(t)H

2)Ψ(t), (1)

where

Ψ is the wave function, Ψ ∈ S2n−1

C = {Ψ ∈ C n

kΨk

Cn = 1};

• H0, H1 and H2 are n × n Hermitian matrices with complex coefficients ;

u(t) ∈ R is the control (for instant the intensity of the laser field) ; • H0 is the internal Hamiltonian ;

• H1, H2 are operators that couple the system with the control u.

Since Ψ and eıθ(t)Ψ describe the same physical state for any global phase t 7→ θ(t) ∈ R, i.e Ψ1 and Ψ2 are identified when exists θ ∈ R such that

Ψ1 = eıθΨ2.

• we add a second control ω corresponding to ˙θ (see also [1]) ;

• we consider the following control system i d

dtΨ(t) = (H0 + u(t)H1 + u

2(t)H

2 + ω(t))Ψ(t), (2)

where ω ∈ R is a new virtual control .

Introduction

• We analyse the Lyapunov trajectory tracking of the Schr¨odinger equation for a second order coupling operator ;

• We present a theoretical convergence result ;

• For situations not covered by the first theorem we propose a numerical approach and complement it with a second theoretical result.

Lyapunov control of Schr¨odinger equations : beyond the dipole approximation

Andreea Grigoriu, CEREMADE - Universit´e Paris Dauphine, France ; Catalin Lefter, Faculty of Mathematics, University ”Al.I.Cuza”, Iasi, Romania ;

Gabriel Turinici, CEREMADE - Universit´e Paris Dauphine, France.

Références

Documents relatifs

In contrast, we here associate the state observer to an adaptive version of so-called Markovian feedback (essentially, proportional control) and we show that this leads to a

As a result, Theorem 1 gives no information about the control time if the controls are allowed to be arbitrarily large; in particular, it does not preclude the possibility

Thus, approximate controllability in arbitrarily small time (allowing potentially large controls) is an open problem.. The goal of this article is to prove that, for poten- tials

We consider discrete-time quantum systems subject to Quantum Non-Demolition (QND) measurements and controlled by an adjustable unitary evolution between two successive QND measures..

Connection between the controllability of the linear tangent approximation around the reference trajectory and asymptotic tracking is studied.. When the linear tangent approximation

Since the design of (memoryless) homogeneous continu- ous or discontinuous feedback controllers using the Implicit Lyapunov Function (ILF) method [6] greatly simplifies the gain

In Section III we recall the time adiabatic theorem and some results on the regularity of eigenvalues and eigenstates of parameter-dependent Hamiltonians.. In Section IV we deepen

Décret présidentiel n° 04-141 du 28 avril 2004, portant ratification des amendements à la convention pour la protection de la mer Méditerranée contre la pollution, adoptée