Quantum dynamics and quantum control of spins in diamond Viatcheslav Dobrovitski Ames Laboratory US DOE, Iowa State University Works done in collaboration with Z.H. Wang (Ames Lab – now USC), T. der Sar, G. de Lange, T. Taminiau, R. Hanson (TU Delft), G. D. Fuchs, D. Toyli, D. D. Awschalom (UCSB), D. Lidar (USC) Individual quantum spins in solid state Quantum dots NV center in diamond Donors in silicon Quantum spin coherence: valuable resource Quantum information processing Single-spin coherent spintronics and photonics High-precision metrology and magnetic sensing at nanoscale Grand challenge – controlling single quantum spins in solids Fundamental problems: 1. Understand dynamics of individual quantum spins 2. Control individual quantum spins 3. Preserve coherence of quantum spins 4. Generate and preserve entanglement between quantum spins Spins in diamond – excellent testbed for quantum studies • Long coherence time • Individually addressable • Controllable optically and magnetically Jelezko et al, PRL 2004; Gaebel et al, Nat.Phys. 2006; Childress et al, Science 2006 Dynamical decoupling protocols Traditional analysis and classification: Magnus expansion Uper exp[i T ( H (0) H (1) H ( 2) ...)] O (1) Simplest – Periodic DD : Symmetrized protocol: O (T ) O(T 2 ) Period τ -X- τ -X τ-X-τ-X-X- τ -X- τ = τ -X- τ - τ -X- τ CPMG sequence 2nd order protocol, error O(τ2) Concatenated protocols (CDD) level l=1 (CDD1 = PDD): τ -X- τ -X level l=2 (CDD2): etc. PDD-X-PDD-X Deficiencies of Magnus expansion: • Norm of H(0), H(1),… – grows with the size of the bath • Validity conditions are often not satisfied in reality (the UV cutoff is too large) but DD works • Behavior at long times – unclear • Accumulation of pulse errors and imperfections – unknown Assessing the quality of coherence protection 1. Exact numerical modeling H H S H B H SB (t ) exp(iHt)(0) Up to 32 spins (Hilbert space d = 4×109) on 128 processors Parallel code, 80 % efficiency 2. Approximate – but very accurate – numerics: coherent spin states 3. Analytical mean-field techniques Outline 1. Quantum control and dynamical decoupling of NV center: protecting coherence Spectacular recent progress: DD on a single NV spin de Lange, Wang, Riste, Dobrovitski, Hanson: Science 2010 Ryan, Hodges, Cory: PRL 2010 Naydenov, Dolde, Hall, Fedder, Hollenberg, Jelezko, Wrachtrup: PRB 2010 2. Decoherence-protected quantum gates 3. Decoherence-protected quantum algorithm: first 2-qubit computation with invidivual solid-state spins NV center in diamond Simplest impurity: substitutional N (P1 center) Environment (spin bath) S = 1/2 Long-range dipolar coupling Nitrogen meets vacancy: NV center Central spin S = 1, I = 1 HF coupling onsite Dipolar coupling to the bath Single NV spin can be initialized, manipulated and read out Single NV center – optical manipulation and readout Excited state: Spin 1 orbital doublet m = +1 m = –1 m=0 ISC (m = ±1 only) 1A 532 nm m = +1 m = –1 m = 0 – always emits light m = ±1 – not MW Ground state: Spin 1 Orbital singlet m=0 Initialization: m = 0 state Readout (PL): population of m = 0 Decoherence: NV center in a spin bath C N V C C Bath spin – N atom NV spin C ms = –1 C m = +1/2 C ms = +1 0 ms = 0 1 0 0 B ms = -1/2 B NV electron spin: pseudospin S = 1/2 (qubit) No flip-flops between NV and the bath: energy mismatch H 0 SZ SZ Ak S H B 0 SZ Bˆ (t )SZ Z k k ˆ (t ) – field created by the bath spins B Time dependence governed by HB Mean field picture: bath as a random field Gaussian, stationary, Markovian noise B(0) B(t ) b2 exp( t C ) b – noise magnitude (spin-bath coupling) τC – correlation time (intra-bath coupling) Direct many-spin modeling: confirms mean field simulation O-U fitting 1.0 0.8 B F2 B(0) B(t ) 0.6 (a) 0.4 0.2 0.0 0 10 20 30 40 Time Dobrovitski et al, PRL 2009 Hanson et al, Science 2008 0.5 *2 2 exp t T 2 Free decoherence Decay due to field inhomogeneity from run to run b 2 T2* T2* = 380 ns -0.5 0 0.2 0.4 0.6 t (µs) Modulation: HF coupling to 14N of NV 0.8 Spin echo: probing the bath dynamics exp(iHt) exp(iHt) 1 0.5 SZ SZ H H 0 τC = 25 μs T2 = 2.8 μs exp t 3 T23 1 10 free evolution time (ms) Quantum control and Dynamical decoupling: Extending coherence time of a single NV center Choice of the DD protocol: theory Signal(T ) exp[i] exp[W (T )] Long times (T >> τC): Short times (T << τC): PDD 4 2 WF (T ) (b C ) N 3 C τ -X- τ -X Fast decay CPMG τ-X- 2τ -X-τ 3 1 WF (T ) (b C ) 2 N 3 C 3 Slow decay Slow decay at all times, rate WS (T) optimal choice Concatenated PDD Fast decay at all times, makes things worse Concatenated CPMG Slow decay at all times, no improvement and many other protocols have been analyzed… Qualitative features • Coherence time can be extended well beyond τC as long as the inter-pulse interval is small enough: τ/τC << 1 • Magnus expansion (also similar cumulant expansions) predict: W(T) ~ O(N τ4) for PDD but we have W(T) ~ O(N τ3) Symmetrization or concatenation give no improvement Source of disagreement: Magnus expansion is inapplicable S ( ) Ornstein-Uhlenbeck noise: 1 2 C2 1 Second moment is (formally) infinite – corresponds to H B2 Cutoff of the Lorentzian: UV ~ m B2 a 3 ~ 2 5 GHz 1 C Pulses only along X: τ-X-2τ-X- τ X component – preserved well Y component – not so well State fidelity DD “as usual” 1.0 x y 0.6 simulation 0 What is wrong? Control pulses are not perfect 5 10 total time (ms) 15 Fast rotation of a single NV center Experiment Example pulse shape: Simulation 29 MHz 109 MHz 223 MHz Time (ns) Time (ns) • Rotating-frame approximation invalid: counter-rotating field • Pulse imperfections important Fuchs et al, Science 2009 1. Bootstrap protocol - characterize all pulse errors from scratch Dobrovitski et al, PRL 2010 2. Understand well the accumulation of the pulse errors Wang et al, arXiv:1011.6417; Khodjasteh et al PRA 2011 Pulses only along X: τ -X-2 τ -X- τ X component – preserved well Y component – not so well State fidelity Protecting all initial states 1.0 x y 0.6 simulation 0 5 10 total time (ms) 15 Pulses along X and Y: τ -X-2 τ -Y-2 τ -X-2 τ -Y- τ State fidelity Solution: two-axis control 1.0 x y simulation 0.6 Both components are preserved Coherence extended far beyond echo time 0 5 10 total time (ms) 15 Aperiodic sequences: UDD and QDD Are expected to be sub-optimal: no hard cut-off in the bath spectrum 20 State fidelity 1 0.5 UDD Np = 6 0 CPMG 1/e decay time (μs) CPMG UDD exp. sim. 5 5 10 15 Total time (ms) 5 10 Np 15 Robustness to errors: QDD6 vs XY4 1.0 B C QDD, SX QDD, SY 0.5 Np= 48 0.0 0 10 20 30 Total time (ms) 40 XY4, SX XY4, SY 1.0 UDD vs XY4 B C B CX 0.5 UDD, S UDD, SY 0.0 Np= 48 -0.5 0 10 20 30 Total time (ms) 40 XY4, SX XY4, SY Extending coherence time with DD Master curve: for any number of pulses 3 S (T ) exp t 3 / Tcoh Tcoh T2 N p2 / 3 100 1/e decay time (μs) State fidelity 1 SE N=4 N=8 N = 16 N = 36 0.5 N = 72 NV2 10 NV1 N = 136 0.1 1 Normalized time (t / T2 N 2/3) 10 1 10 100 number of pulses Np 136 pulses, coherence time increased by a factor 26 Tcoh = 90 μs at room temperature, and no limit in sight De Lange, Wang, Riste, et al, Science 2010 Using DD for other good deeds Single-spin magnetometry with DD de Lange, Riste, Dobrovitski et al, PRL 2011 Taylor, Cappellaro, Childress et al. Nat Phys 2008 Naydenov, Dolde, Hall et al. PRB 2011 Detailed probe of the mesoscopic spin bath SZ 0.50 0.25 0 0 1 2 3 4 time (ms) de Lange, van der Sar, Blok et al, arXiv 2011 Combining DD and quantum operation Gates with resonant decoupling Coupling NVs to each other – hybrid systems Hybrid systems: different types of qubits for different functions NV centers – qubits Nanomechanical oscillators – data bus Rabl et al, Nat Phys 2010 NV centers – qubits Spin chain (other spins) – data bus Cappellaro et al PRL 2010; Yao et al. PNAS 2011 Electron spins – processors Nuclear spins – memory Many works since Kane 1998, maybe before “Standard” quantum operation Unprotected quantum gate Bath Protected storage: decoupling Bath Contradiction: DD efficiently preserves the qubit state but quantum computation must change it Gates with integrated decouplind Unprotected quantum gate Protected storage: decoupling Bath Bath DD gate Tg DD Protected gate Bath Gate with resonant decoupling (GARD) for hybrid systems Different qubits have different coherence and control timescales One qubits decoheres before another starts to move Nuclear 14N spin: memory, Electronic NV spin: processing (quantum memory, quantum repeater, magnetic sensing, etc.) Childress, Taylor, Sorensen et al. PRL 2006 Taylor, Marcus, Lukin PRL 2003 Hint A SZ I Z C C N V C C Jiang, Hodges, Maze et al. Science 2009 Neumann, Beck, Steiner et al. Science 2011 C But control of nuclear spin takes much longer than T2* C Poor choice: either decouple the electron – no gates possible or gating without DD – no gates possible A way out: use internal resonance in the system How the GARD works - 1 Rotating frame (ωN << A ) Rotating frame: H A S Z I Z N I X 0, 1, A A = 2π ∙ 2.16 MHz ωN = 2π ∙ 18 kHz 1, 0, ωN 100 times smaller Electron: 1 Nuclear rotation around Z Electron: 0 Nuclear rotation around X How the GARD works - 2 Main problem: electron switches very frequently between 0 and 1 and slow nuclear spin should keep track of this Contradiction with the very idea of DD? Motion of the nuclear spin: conditional single-spin rotation 0-X-1-1-Y-0 H0 exp[i (n0 )] XY4 unit: τ -X- τ - τ -Y- τ H exp[ i ( n 1-X-0-0-Y-1 1 1 )] 2 smth 1 Axes n0 and n1 are both close to z (A >> ω1): n0 n1 1 ( N / A) smth 2 small Resonance: smth 2 also becomes small when (2n 1) A n0n1 1 How the GARD works - 3 IN: 1 1 0 OUT: 1, 1, X RZ(π) IN: 0 RX(2α) RZ(π) 0 1 OUT: 0, 0, X RX(α) XY-4 unit: RZ(2π) 2 2 A RX(α) Experimental implementation of GARD Resonances are very narrow, ~ (ωN /A)2 Timing jitter < 1 ns over 100 μs time span Error by 10 ns – fidelity drops by 10% Nuclear spin rotation conditioned on the electron: (2n 1) A Unconditional nuclear spin rotation: 2n A All nuclear gates are produced only by changing τ Experimental implementation: proof of concept CNOT gate Protected C-Rot gate electron Total gate time (us) 100μs200 TG0 = 60 >> 1.0 * T300 2 400 500 nucleus 0.5 P(Nucleus = Up)P(Nucleus = Up) Total gate time (us) IZ IZ 0.00 0.5 1.0 0.5 100 200 300 400 500 1 -0.5 0.0 1.0 1.0 0.5 0.5 0 mostly, T1 decay -0.5 0.0 1.00 0.5 4 8 12 16 20 24 Total # of XY2 Fidelity 97% How good is GARD: protected CNOT gate Controllable decoherence: inject a noise into the system Decoherence time: T2 = 50 μs; Gate time TG = 120 μs in 0 i 1 Overlap out Fidelity out 0 1 GARD implementation of Grover’s algorithm 2 qubits – Grover’s algorithm converges in one iteration Total time: 330 μs, T2 time only 250 μs First quantum computation on two individual solid-state spins GARD implementation of Grover’s algorithm 1 2 3 4 5 6 Fidelity: 95% for 1, For other states: 0.93, 0.92, 0.91 High fidelity beyond coherence time Conclusions 1. Diamond-based QIP becomes truly competitive 2. Coherence time can be extended, 25-fold demonstrated 3. DD can be efficiently combined with gates 4. GARD algorithms demonstrated, 50% longer than T2 Fidelity above 90% First 2-qubit computation on individual solid state spins