Examples of time-dependent control targets Objective: Determine a laser pulse which achieves as prescribed goal that a) the wave function follows a given path in Hilbert space (i.e. a given TD wave function) b) the density should follow a given classical trajectory r(t) c) a given peak in the HHG spectrum is enhanced Control the path of the current with laser left lead right lead Control the path of the current with laser left lead right lead Optimal control of time-dependentTHANKS targets OUTLINE OUTLINE • Optimal Control Theory (OCT) of static targets -- OCT of current in quantum rings -- OCT of ionization -- OCT of particle location in double well with frequency constraints • Optimal Control of time-dependent targets -- OCT of path in Hilbert space -- OCT of path in real space -- OCT of harmonic generation THANKS Alberto Castro Esa Räsänen Angel Rubio (San Seb) Kevin Krieger Jan Werschnik Ioana Serban Optimal control of static targets (standard formulation) For given target state Φf , maximize the functional: J1 T f 2 ˆ T T f f T T O Optimal control of static targets (standard formulation) For given target state Φf , maximize the functional: J1 T f 2 ˆ T T f f T T O Ô Optimal control of static targets (standard formulation) For given target state Φf , maximize the functional: J1 T f 2 ˆ T T f f T T O Ô with the constraints: T 2 J 2 dt t E 0 0 E0 = given fluence Optimal control of static targets (standard formulation) For given target state Φf , maximize the functional: J1 T f 2 ˆ T T f f T T O Ô with the constraints: T 2 J 2 dt t E 0 0 T E0 = given fluence ˆ t t J 3 , , 2 Im dt t t Tˆ V 0 Optimal control of static targets (standard formulation) For given target state Φf , maximize the functional: J1 T f 2 ˆ T T f f T T O Ô with the constraints: T 2 J 2 dt t E 0 0 T E0 = given fluence ˆ t t J 3 , , 2 Im dt t t Tˆ V 0 TDSE Optimal control of static targets (standard formulation) For given target state Φf , maximize the functional: J1 T f 2 ˆ T T f f T T O Ô with the constraints: T 2 J 2 dt t E 0 0 T E0 = given fluence ˆ t t J 3 , , 2 Im dt t t Tˆ V 0 GOAL: Maximize J = J1 + J2 + J3 TDSE Set the total variation of J = J1 + J2 + J3 equal to zero: Control equations Algorithm 1. Schrödinger equation with initial condition: J 0 i (t ) Hˆ (t ) (t ), (0) Forward propagation t 2. Schrödinger equation with final condition: J 0 it (t ) Hˆ (t ) (t ), (T ) Oˆ (T ) Backward propagation 3. Field equation: J 0 (t ) 1 Im (t ) ˆ (t ) New laser field Algorithm monotonically convergent: W. Zhu, J. Botina, H. Rabitz, J. Chem. Phys. 108, 1953 (1998)) Control of currents 2 |t| and j (t) l=1 l = -1 l=0 I ~ A E. Räsänen, A. Castro, J. Werschnik, A. Rubio, E.K.U.G., PRL 98, 157404 (2007) OCT of ionization • Calculations for 1-electron system H2+ in 3D • Restriction to ultrashort pulses (T<5fs) nuclear motion can be neglected • Only linear polarization of laser (parallel or perpendicular to molecular axis) • Look for enhancement of ionization by pulse-shaping only, keeping the time-integrated intensity (fluence) fixed Control target to be maximized: ˆ T J1 T O with ˆ 1ˆ O bound i i i Standard OCT algorithm (forward-backward propagation) does not converge: ˆ before the backward-propagation eliminates the Acting with O smooth (numerically friendly) part of the wave function. Instead of forward-backward propagation, parameterize the laser pulse to be optimized in the form t f t cos 0 t , with ω0 = 0.114 a.u. (λ = 400 nm) 2 2 f t fn c os n t gn sin n t , T T n 1 N with ωn = 2πn/T Choose N such that maximum frequency is 2ω0 or 4ω0 . T is fixed to 5 fs. Maximize J1 (f1…fN, g1…gN) directly with constraints: i f 0 f T 0 N f n 1 n 0 ii 0 dt 2 (t) E0 . T using algorithm NEWUOA (M.J.D. Powell, IMA J. Numer. Analysis 28, 649 (2008)) of initial pulse of initial pulse Ionization probability for the initial (circles) and the optimized (squares) pulse as function of the peak intensity of the initial pulse. Pulse length and fluence is kept fixed during the optimization. Control of electron localization in double quantum dots: t = 0 ps t = 1.16 ps t = 2.33 ps t = 3.49 ps t = 4.66 ps t = 5.82 ps E. Räsänen, A. Castro, J. Werschnik, A. Rubio, E.K.U.G., Phys. Rev. B 77, 085324 (2008). target state: f = first excited state (lives in the well on the right-hand side) Optimization results Optimized pulse 1 (T ) 2 99.91% Occupation numbers Spectrum 3 0 1 2 1 0 2 1 0 3 OCT finds a combination of several transition processes E algorithm Forward propagation of TDSE (k) Backward propagation of TDSE (k) new field: ~ k 1 t 1 Im k t ˆ k t (W. Zhu, J. Botina, H. Rabitz, J. Chem. Phys. 108, 1953 (1998)) algorithm Forward propagation of TDSE (k) Backward propagation of TDSE (k) ~ k 1 t 1 Im k t ˆ k t new field: (W. Zhu, J. Botina, H. Rabitz, J. Chem. Phys. 108, 1953 (1998)) With spectral constraint: k 1 filter function: or t : F f F ~ k 1 t 2 2 f ω exp γω ω0 exp γω ω0 2 2 f ω 1 exp γω ω0 exp γω ω0 J. Werschnik, E.K.U.G., J. Opt. B 7, S300 (2005) E Frequency constraint: Only direct transition frequency 0 allowed Spectrum of optimized pulse 1 (T ) 2 0.9997 occupation numbers Time-Dependent Density Frequency constraint: Selective transfer via intermediate state 2 E ω0 2 ω2 1 0 2 1 Spectrum of optimized pulse occupation numbers Time-Dependent Density Frequency constraint: Selective transfer via intermediate state 3 ω0 3 ω3 1 0 3 1 E Frequency constraint: All resonances excluded Spectrum of optimized pulse occupation numbers All pulses shown give close to 100% occupation at the end of the pulse OPTIMAL CONTROL OF TIME-DEPENDENT TARGETS Maximize J J1 J 2 J 3 T 1 ˆ t t J1 dt t O T0 T 2 J 2 dt t E 0 0 T ˆ t t J 3 , , 2 Im dt t t Tˆ V 0 Set the total variation of J = J1 + J2 + J3 equal to zero: Control equations Algorithm 1. Schrödinger equation with initial condition: J 0 i (t ) Hˆ (t ) (t ), (0) Forward propagation t 2. Schrödinger equation with final condition: Inhomogenous TDSE : i ˆ J 0 ˆ i t H (t ) (t ) T O(t ) (t ), (T ) 0 Backward propagation 3. Field equation: J 0 (t ) 1 Im (t ) ˆ (t ) New laser field Y. Ohtsuki, G. Turinici, H. Rabitz, JCP 120, 5509 (2004) I. Serban, J. Werschnik, E.K.U.G. Phys. Rev. A 71, 053810 (2005) Control of path in Hilbert space ˆ t t t O t 0 t e with 0 t 2 o t 0 1 t e given target occupation, and 1t t 1 2 1 1 Goal: Find laser pulse that reproduces |αo(t)|2 I. Serban, J. Werschnik, E.K.U.G. Phys. Rev. A 71, 053810 (2005) t 0 2 Control path in real space ˆ t r r0 t O 1 r r t e 2 2 2 0 2 2 with given trajectory r0(t) . Algorithm maximizes the density along the path r0(t): I. Serban, J. Werschnik, E.K.U.G. Phys. Rev. A 71, 053810 (2005) J. Werschnik and E.K.U.G., in: Physical Chemistry of Interfaces and Nanomaterials V, M. Spitler and F. Willig, eds, Proc. SPIE 6325, 63250Q(113) (ISBN: 9780819464040, doi: 10.1117/12.680065); also on arXiv:0707.1874 Control of time-dependent density of hydrogen atom in laser pulse Control of charge transfer along selected pathways Trajectory 1 Trajectory 2 Time-evolution of wavepacket with the optimal laser pulse for trajectory 1 Trajectory 1: Results Start Lowest six eigenstates Populations of eigenstates ground state first excited state second excited state fifth excited state Trajectory 2 Optimization of Harmonic Generation Harmonic Spectrum: H dte it 2 d dt t 3 d r z r, t 2 Lmax Maximize: J1 L max H L L 1 0 To optimize the 7th harmonic of ω0 , choose coefficients as, e.g., α7= 4, α3 = α5 = α9 = α11 = -1 Harmonic generation of helium atom (TDDFT calculation in 3D) 3 5 7 9 11 13 15 17 19 21 Enhancement of 7th harmonic xc functional used: EXX Research&Training Network SFB 450 SFB 658 SPP 1145