PRL 113, 043001 (2014) week ending 25 JULY 2014 PHYSICAL REVIEW LETTERS Correction of Arbitrary Field Errors in Population Inversion of Quantum Systems by Universal Composite Pulses 1 Genko T. Genov,1,2 Daniel Schraft,2 Thomas Halfmann,2 and Nikolay V. Vitanov1 Department of Physics, St. Kliment Ohridski University of Sofia, 5 James Bourchier boulevard, 1164 Sofia, Bulgaria 2 Institut für Angewandte Physik, Technische Universität Darmstadt, Hochschulstraße 6, 64289 Darmstadt, Germany (Received 19 March 2014; revised manuscript received 22 May 2014; published 22 July 2014) We introduce universal broadband composite pulse sequences for robust high-fidelity population inversion in two-state quantum systems, which compensate deviations in any parameter of the driving field (e.g., pulse amplitude, pulse duration, detuning from resonance, Stark shifts, unwanted frequency chirp, etc.) and are applicable with any pulse shape. We demonstrate the efficiency and universality of these composite pulses by experimental data on rephasing of atomic coherences in a Pr3þ∶Y2 SiO5 crystal. DOI: 10.1103/PhysRevLett.113.043001 PACS numbers: 32.80.Qk, 32.80.Xx, 42.50.Md, 82.56.Jn Performing high-fidelity and robust population transfer by external fields remains an important challenge in various areas of physics, e.g., atomic and molecular physics [1], nuclear magnetic resonance [2], and quantum information processing [3–5]. Most of the available techniques provide either high population transfer efficiency or robustness against variations in the field parameters, but not both. For example, resonant techniques deliver high efficiency but are not robust. Adiabatic passage techniques [6–8] are robust, but in nearly all of them population transfer is incomplete and very long interaction times and/or high intensities are required to satisfy the adiabaticity criteria. Optimal control techniques [1,9], including the emerging field of “shortcuts to adiabaticity” [10], achieve high fidelity by tailoring the time dependences of the amplitude and the phase of the driving field. However, their implementation is often challenging because they usually prescribe rather demanding time dependences, while various experimental limitations (field inhomogeneities, spatial pulse shape distortion, etc.) can limit their efficiency. The technique of composite pulses (CPs) is far easier to implement and has been used for decades in nuclear magnetic resonance [2] and, since recently, in quantum information processing [3–5,11] and quantum optics [12–16] for highly accurate and robust qubit rotation. CPs are unique in combining the advantages of resonant techniques (very high fidelity) and adiabatic techniques (robustness to errors). The basic idea of CPs is to correct the imperfect interaction of a quantum system with a single pulse by using a sequence of pulses with suitably chosen relative phases, which are usually easy to control experimentally. The errors introduced by the constituent pulses are canceled by destructive interference up to a certain order. In this Letter, we describe a general theoretical procedure to derive universal CPs for complete population inversion, which compensate deviations in any experimental parameter and work with any pulse shape. Essentially, all of the existing CPs compensate errors in a single interaction parameter, or at most 0031-9007=14=113(4)=043001(5) two interaction parameters, and usually assume a rectangular pulse shape. The only assumptions made here are those of a two-state system, coherent evolution, and identical pulses in the CP sequence [17]. As a basic example of relevance to many applications in quantum physics, we experimentally demonstrate the concept by rephasing of atomic coherences for optical data storage in a Pr3þ∶Y2 SiO5 crystal. We consider a coherently driven two-state quantum system. Its dynamics obeys the Schrödinger equation, iℏ∂ t cðtÞ¼HðtÞcðtÞ, where the vector cðtÞ¼½c1 ðtÞ;c2 ðtÞT contains the probability amplitudes of the two states. The Hamiltonian in the rotating-wave approximation reads HðtÞ ¼ ðℏ=2ÞΩðtÞe−iδðtÞ j1ih2j þ H.c., with δðtÞ ¼ Rt 0 0 0 Δðt Þdt , where Δ ¼ ω0 − ω is the detuning between the field frequency ω and the Bohr transition frequency ω0 . The Rabi frequency ΩðtÞ ¼ −d · EðtÞ=ℏ defines the coupling of the two states, induced by the electric field EðtÞ and the transition dipole moment d. In general, both ΩðtÞ and ΔðtÞ are time dependent. We assume that the CP duration is shorter than the decoherence times in the system. Then, the evolution is described by the propagator U, which connects the amplitudes at the initial and final times ti and tf : cðtf Þ ¼ Ucðti Þ. It is conveniently parametrized with the three real Stückelberg variables q (0≦q≦1), α, and β as qeiα peiβ U¼ ; ð1Þ −pe−iβ qe−iα pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ where p ¼ 1 − q2 ; p2 is the transition probability from state j1i to state j2i, and q2 is the probability for no transition. A constant phase shift ϕ in the Rabi frequency, ΩðtÞ → ΩðtÞeiϕ , is imprinted in the propagator UðϕÞ by taking β → β þ ϕ [13]. The propagator for a composite sequence of n pulses, each with a phase ϕk , reads UðnÞ ¼ Uðϕn Þ Uðϕ2 ÞUðϕ1 Þ; ð2Þ where the phases ϕ1 ; …; ϕn are free control parameters. 043001-1 © 2014 American Physical Society PRL 113, 043001 (2014) PHYSICAL REVIEW LETTERS Our objective is to transfer all population from state j1i to state j2i, even when the properties of the driving pulse are unknown. This requires robustness to deviations in all pulse parameters, i.e., pulse shape, Rabi frequency, duration, detuning from resonance, dynamic Stark shifts, residual frequency chirp, etc. Mathematically, our goal is ðnÞ to maximize the transition probability PðnÞ ¼ jU21 j2, i.e., to minimize the CP infidelity (the probability for no ðnÞ transition) QðnÞ ¼ jU11 j2 for any values of q, α, and β. We make no assumptions about the constituent pulses, i.e., how q, α, and β depend on the interaction parameters. This justifies the term “universal” for these CPs because they will compensate imperfections in any interaction parameter. We assume only that the constituent pulses are identical and that we have control over their phases ϕk . In order to determine the phases ϕk of a universal CP sequence of n pulses we calculate the P propagator element ðnÞ ðnÞ U11 from Eq. (2). We find U11 ¼ nj¼1 cnj qj , where the coefficients cnj depend on α and ϕk only. Following arguments in Refs. [13,14] we use symmetric CPs, the phases of which obey the anagram relation ϕnþ1−k ¼ ϕk , with k ¼ 1; 2; …; ðn − 1Þ=2. Such symmetric sequences automatically nullify all cnj for even j. We choose phases such that the coefficients cnj are nullified for any α up to the highest possible order, denoted j0 . This generates a system of trigonometric equations for the phases ϕk . In general, there are multiple solutions. We choose solutions, which minimize the sum of the absolute values of the coefficients of the first nonzero order j0 þ 1. The transition probability for ðnÞ such CP reads PðnÞ ¼ 1 − jU11 j2 ¼ 1 − Oðq2j0 þ2 Þ. Since 0≦q≦1, we can make the error term arbitrary small (unless q ¼ 1, which means zero transition probability for a single pulse) by nullifying cnj to an arbitrary high order j0 by using longer composite sequences. For example, for a five-pulse ð5Þ sequence, we have U 11 ¼ f½1 þ 2 cosð2ϕ2 − ϕ3 Þeiα þ −iα 2 cosðϕ2 − ϕ3 Þe gq þ Oðq3 Þ. The q term vanishes for two distinct sets: fϕ2 ¼ 5π=6; ϕ3 ¼ π=3g and fϕ2 ¼ 11π=6; ϕ3 ¼ π=3g. Then, we have j0 ¼ 2 ð5Þ and PðnÞ ¼ 1 − jU11 j2 ¼ 1 − Oðq6 Þ. Several universal CPs are listed in Table I. The CPs labeled “a” perform better against variations in the pulse area (i.e., pulse duration and amplitude), whereas those labeled “b” perform better against the static detuning. Either of these, however, permit compensation against both parameters, as well as against any other parameters. The CPs U5a and U5b in Table I were derived earlier, aiming at compensation of pulse area [18] and detuning errors [13] for specific pulse shapes. To the best of our knowledge, all other universal CPs in Table I were not known before. Figure 1 depicts the theoretical performance of a single pulse and universal CPs up to order 25 of types “a” and “b” against the static detuning and the duration T of each constituent pulse. As it is well known, the transition probability for a single pulse quickly drops when T does not match to a perfect π pulse, i.e., T ¼ τ, or when the pulse carrier week ending 25 JULY 2014 TABLE I. Phases of universal CPs with n pulses (indicated by the number in the label of the CP). We have j0 ¼ 0 for n ¼ 3, j0 ¼ 2 for n ¼ 5 to 9, j0 ¼ 4 for n ¼ 13, and j0 ¼ 8 for n ¼ 25. Each phase is defined modulo 2π. The excitation dynamics remain the same when we simultaneously change the sign of all phases or add a constant shift to all phases. Pulse Phases U3 U5a U5b U7a U7b U9a U9b U13a U13b U25a ð0; 1; 0Þπ=2 ð0; 5; 2; 5; 0Þπ=6 ð0; 11; 2; 11; 0Þπ=6 ð0; 11; 10; 17; 10; 11; 0Þπ=12 ð0; 23; 10; 5; 10; 23; 0Þπ=12 ð0; 0.635; 1.35; 0.553; 0.297; 0.553; 1.35; 0.635; 0Þπ ð0; 1.635; 1.35; 1.553; 0.297; 1.553; 1.35; 1.635; 0Þπ ð0; 9; 42; 11; 8; 37; 2; 37; 8; 11; 42; 9; 0Þπ=24 ð0; 33; 42; 35; 8; 13; 2; 13; 8; 35; 42; 33; 0Þπ=24 ð0; 5; 2; 5; 0; 11; 4; 1; 4; 11; 2; 7; 4; 7; 2; 11; 4; 1; 4; 11; 0; 5; 2; 5; 0Þπ=6 ð0; 11; 2; 11; 0; 5; 4; 7; 4; 5; 2; 1; 4; 1; 2; 5; 4; 7; 4; 5; 0; 11; 2; 11; 0Þπ=6 U25b ðnÞ frequency is off resonance. We plotted the error jU 11 j2 and we used a logarithmic scale to verify the suitability of these CPs for applications requiring ultrahigh fidelity, e.g., as in quantum information processing. The universal CP sequences are robust with respect to variations along both axes. Furthermore, the 10−4 error regions are far broader than for a single pulse, and they expand steadily with the CP order. We experimentally verified the performance of the universal CP sequences by rephasing atomic coherences for optical data storage. In the experiment, we generate the atomic coherence on a magnetic, radio-frequency (rf) transition between two inhomogeneously broadened hyperfine levels of Pr3þ∶Y2 SiO5 crystal. The atomic coherence between the two quantum states is prepared and read out by electromagnetically induced transparency (EIT) [19]. This enables straightforward, on-demand optical readout. The EIT scheme couples states j1i and j2i by a strong control field and a weak probe field via an excited state j3i. By simultaneously and adiabatically turning off the control and probe fields, we convert the probe field into an atomic coherence, i.e., a coherent superposition of states j1i and j2i. This is the “write” process of optical information encoded in the probe field, often termed “stopped light” or “stored light” [19]. To “read” the optical memory after an arbitrary storage time, we apply the strong control field again to beat with the atomic coherence and thereby generate a signal field with the same properties as the stored field. The concept and the experimental setup for EIT-based light storage in Pr3þ∶Y2 SiO5 are described in detail elsewhere [8,15,20]. In such a coherent optical memory it is crucial to reverse the effect of dephasing of the atomic coherences during the storage time. The dephasing is due to inhomogeneous broadening of the hyperfine levels, while rephasing is implemented usually by resonant rf π pulses. However, 043001-2 Single Pulse Duration (in units of τ) 1.5 single U5a 1 U7a 1 1 1 week ending 25 JULY 2014 PHYSICAL REVIEW LETTERS PRL 113, 043001 (2014) 2 1 2 U9a 4 4 2 4 U13a 1 2 4 2 4 2 0.5 1.5 U3 1 U5b 1 1 2 2 4 U7b 1 U9b 2 4 U13b 1 1 4 2 4 -1 0 1 -1 0 1 U25b 1 4 2 0.5 U25a 1 2 -1 0 1 -1 0 1 Static Detuning (in units of Ω) -1 0 1 -1 0 1 ðnÞ FIG. 1 (color online). Numerically simulated infidelity jU 11 j2 vs static detuning and duration T of each constituent pulse (referred to as “single pulse duration”) for a single pulse and several universal CPs (with rectangular shapes) from Table I. The labels m ¼ 1; 2; 4 indicate the error level 10−m . such pulses do not work efficiently in systems with large inhomogeneous broadening, as the transition frequency varies for different ions. The efficiency is further reduced by the spatial inhomogeneity of the rf field over the crystals dimensions. In order to permit a much broader operation bandwidth, we replace now the single π pulses by our universal CPs, similarly to Ref. [15]. In the experiment we set the storage time to 600 μs, which is much larger than the dephasing time of about 20 μs. During this storage time we rephase the atomic coherence by applying a pair of (composite) pulses, where the first (composite) pulse is centered at 150 μs and the second at 450 μs. The optical “write” and “read” sequences were kept fixed, while the rf rephasing sequences were varied; therefore, the energy of the retrieved signal measures the rephasing efficiency, and hence, the efficiency of the driving single pulses or universal CPs. Figure 2 compares the theoretical performance and experimental results for a single pulse and the universal CP U3 versus variations in the single pulse duration and static detuning. We note that in Figs. 2(a) and 2(b), in contrast ðnÞ to Fig. 1, we have simulated jU21 j4 , i.e., the rephasing efficiency, averaged over all possible stored coherences. The experimental data [Figs. 2(c) and 2(d)] agree remarkably well with the numerical simulations, and fully confirm the pronounced robustness of the universal CP sequences. Next, we experimentally compare the behavior of universal CP sequences of orders n ¼ 3, 5, 7, 9 in Fig. 3. We apply CPs with rectangular (R) and Gaussian (G) temporal shape (left and right columns in Fig. 3). The data for the universal CP sequence U3 R, already shown in Fig. 2(d), serve as a benchmark. As the data clearly show, the robustness of all universal CP sequences to variations in the experimental parameters is much larger compared to a single pulse [Fig. 2(c)]. As expected, the region of large rephasing efficiency increases with the order of the universal CP sequences. The behavior is the same for rectangular and Gaussian pulses; i.e., the optimal phases in the CP sequence do not depend on the pulse shape. The experimental data clearly confirm the theoretical predictions: indeed, the overall shapes of the high-efficiency regions in Fig. 3 look remarkably similar to the theoretical plots in Fig. 1. The “duck-foot” pattern for U3, the “boomerang” pattern for U5b, and the “pine-tree” pattern for U7b are very nicely reproduced. FIG. 2 (color online). Numerically simulated (a),(b) and experimentally measured rephasing efficiency of stored light (c),(d) vs static detuning and single pulse duration T for a single rectangular pulse (a),(c) and rectangular CP U3 (b),(d) with phases (0, π=2, 0) from Table I. The Rabi frequency is kept fixed at Ωrf ≈ 2π × 155 kHz; i.e., a single π pulse has a duration of τ ¼ 3.2 μs. 043001-3 PRL 113, 043001 (2014) PHYSICAL REVIEW LETTERS week ending 25 JULY 2014 FIG. 4 (color online). Numerically simulated (a),(b) and experimentally measured rephasing efficiency of stored light (c),(d) vs static detuning and chirp range, symmetric around resonance, of each constituent pulse for a single rectangular pulse (a),(c) and universal CP U9b with phases from Table I (b),(d). The Rabi frequency is kept fixed at Ωrf ≈ 2π × 155 kHz and the duration of each (constituent) pulse is T ¼ 3.2 μs. FIG. 3 (color online). Experimentally measured rephasing efficiency of stored light vs static detuning and single pulse duration for different universal CP sequences from Table I (labeled in each frame). The left column is for sequences of rectangular pulses and the right column is for sequences of Gaussian pulses. In either case the peak Rabi frequency is Ωrf ≈ 2π × 155 kHz. Hence, for rectangular pulses, the pulse duration of a perfect π pulse is τ ¼ 3.2 μs and for Gaussian pulses, it is 2.8 μs (FWHM). We also verified by extensive simulations the robustness of the universal CPs against variations in other interaction parameters, e.g., Stark shifts, unwanted frequency chirp, pulse shape, etc. All simulations confirm that our universal CPs are amazingly robust to any such variation. For example, Fig. 4 shows a comparison of the rephasing efficiency of a pair of single pulses and a pair of U9b CPs, given in Table I, versus the frequency chirp range and the static detuning. In this example, we applied a linear chirp, symmetric around resonance, on each constituent pulse. Figures 4(a) and 4(c) show that the efficiency of a single pulse quickly drops with static detunings larger than 50 kHz and chirp ranges larger than 700 kHz. In comparison, the CP U9b sequence in Figs. 4(b) and 4(d) features a much better stability against variations in both axes. The experimental data agree remarkably well with the numerical simulations and fully confirm the pronounced robustness of the universal CP sequences. In conclusion, we theoretically developed and experimentally demonstrated universal broadband CP sequences for robust high-fidelity population inversion. These CPs compensate simultaneous variations in all experimental parameters (e.g., intensity or Rabi frequency, pulse duration, static detuning, Stark shifts, frequency chirp, etc.) and are applicable with any pulse shape. We experimentally confirmed the theoretical predictions by rephasing atomic coherences in a Pr3þ∶Y2 SiO5 crystal. In particular, our data demonstrate robust rephasing in a broad range of experimental parameters and for different pulse shapes. The efficiency and operation bandwidth of universal CPs are significantly larger compared to conventional π pulses or nonuniversal CPs. The universal CPs will be a highly accurate and robust tool for quantum control, particularly valuable in the presence of significant experimental uncertainties. This work is supported by the Deutsche Forschungsgemeinschaft, the Alexander von Humboldt Foundation, and the European Union’s Seventh Framework Programme under REA Grant No. 287252. G. T. G. and D. S. contributed equally to the results presented in this Letter. 043001-4 PRL 113, 043001 (2014) PHYSICAL REVIEW LETTERS [1] C. Brif, R. Chakrabarti, and H. Rabitz, New J. Phys. 12, 075008 (2010); S. Rice and M. Zhao, Optimal Control of Quantum Dynamics (Wiley, New York, 2000); M. Shapiro and P. Brumer, Principles of Quantum Control of Molecular Processes (Wiley, New York, 2003); D. J. Tannor, Introduction to Quantum Mechanics: A Time-Dependent Perspective (University Science Books, Sausalito, CA, 2007). [2] M. H. Levitt, Prog. Nucl. Magn. Reson. Spectrosc. 18, 61 (1986); R. Freeman, Spin Choreography (Spektrum, Oxford, 1997). [3] F. Schmidt-Kaler, H. Häffner, M. Riebe, S. Gulde, G. P. T. Lancaster, T. Deuschle, C. Becher, C. F. Roos, J. Eschner and R. Blatt, Nature (London) 422, 408 (2003); M. Riebe, T. Monz, A. S. Villar, P. Schindler, M. Chwalla, M. Hennrich, and R. Blatt, Nat. Phys. 4, 839 (2008); T. Monz, K. Kim, W. Hänsel, M. Riebe, A. S. Villar, P. Schindler, M. Chwalla, M. Hennrich, and R. Blatt, Phys. Rev. Lett. 102, 040501 (2009). [4] M. Pons, V. Ahufinger, C. Wunderlich, A. Sanpera, S. Braungardt, A. Sen(De), U. Sen, and M. Lewenstein, Phys. Rev. Lett. 98, 023003 (2007); N. Timoney, V. Elman, S. Glaser, C. Weiss, M. Johanning, W. Neuhauser, and C. Wunderlich, Phys. Rev. A 77, 052334 (2008); C. Piltz, B. Scharfenberger, A. Khromova, A. F. Varon, and C. Wunderlich, Phys. Rev. Lett. 110, 200501 (2013). [5] C. D. Hill, Phys. Rev. Lett. 98, 180501 (2007). [6] N. V. Vitanov, T. Halfmann, B. W. Shore and K. Bergmann, Annu. Rev. Phys. Chem. 52, 763 (2001). [7] M. F. Pascual-Winter, R. C. Tongning, R. Lauro, A. Louchet-Chauvet, T. Chanelière, and J.-L. Le Gouët, Phys. Rev. B 86, 064301 (2012). week ending 25 JULY 2014 [8] S. Mieth, D. Schraft, T. Halfmann, and L. P. Yatsenko, Phys. Rev. A 86, 063404 (2012). [9] L. S. Pontryagin, V. G. Boltyanskii, R. V. Gamkrelidze, and E. F. Mishchenko, The Mathematical Theory of Optimal Processes (Mir, Moscow, 1974); B. Bonnard and M. Chyba, Singular Trajectories and Their Role in Control Theory (Springer, New York, 2003), Vol. 40. [10] M. Demirplak and S. A. Rice, J. Phys. Chem. A 107, 9937 (2003); M. V. Berry, J. Phys. A 42, 365303 (2009). [11] S. S. Ivanov and N. V. Vitanov, Phys. Rev. A 84, 022319 (2011). [12] S. S. Ivanov and N. V. Vitanov, Opt. Lett. 36, 1275 (2011); G. T. Genov, B. T. Torosov, and N. V. Vitanov, Phys. Rev. A 84, 063413 (2011); G. T. Genov and N. V. Vitanov, Phys. Rev. Lett. 110, 133002 (2013). [13] B. T. Torosov and N. V. Vitanov, Phys. Rev. A 83, 053420 (2011). [14] B. T. Torosov, S. Guérin, and N. V. Vitanov, Phys. Rev. Lett. 106, 233001 (2011). [15] D. Schraft, T. Halfmann, G. T. Genov, and N. V. Vitanov, Phys. Rev. A 88, 063406 (2013). [16] B. T. Torosov and N. V. Vitanov, Phys. Rev. A 87, 043418 (2013). [17] We point out that this method compensates only errors in the driving field and should not be confused with quantum error correction, which protects from errors due to decoherence and other quantum noise stemming from the environment. [18] R. Tycko and A. Pines, Chem. Phys. Lett. 111, 462 (1984). [19] M. Fleischhauer, A. Imamoglu, and J. P. Marangos, Rev. Mod. Phys. 77, 633 (2005). [20] G. Heinze, C. Hubrich, and T. Halfmann, Phys. Rev. Lett. 111, 033601 (2013). 043001-5

© Copyright 2021 Paperzz