Digital Quantum Simulation of Spin Systems in Superconducting Circuits
U. Las Heras,1 A. Mezzacapo,1 L. Lamata,1 S. Filipp,2 A. Wallraff,2 and E. Solano1, 3 1Department of Physical Chemistry, University of the Basque Country UPV/EHU, Apartado 644, 48080 Bilbao, Spain 2Department of Physics, ETH Z¨urich,CH-8093 Z¨urich,Switzerland 3IKERBASQUE, Basque Foundation for Science, Alameda Urquijo 36, 48011 Bilbao, Spain (Dated: May 13, 2014) We propose the implementation of a digital quantum simulator for prototypical spin models in a circuit quantum electrodynamics architecture. We consider the feasibility of the quantum simula- tion of Heisenberg and frustrated Ising models in transmon qubits coupled to coplanar waveguide microwave resonators. Furthermore, we analyze the time evolution of these models and compare the ideal spin dynamics with a realistic version of the proposed quantum simulator. Finally, we discuss the key steps for developing a toolbox of digital quantum simulators in superconducting circuits.
PACS numbers: 03.67.Ac, 03.67.Lx, 42.50.Pq, 85.25.Cp
The quantum coherent control of superconducting superconductor-based qubit with long enough coherence qubits has improved dramatically in the last years [1]. In time, we focus on a transmon qubit setup. Superconduct- this sense, circuit quantum electrodynamics (cQED) [2] ing transmon qubits are commonly used because of their is considered as a potential scalable platform for quan- low sensitivity to offset charge fluctuations [27]. However, tum computing. Basic quantum algorithms [3] and tests depending on the targeted physics, other superconduct- of fundamentals in quantum mechanics [4] have been al- ing qubits may be considered for quantum simulations ready realized. Single and two qubit gates [5], prepa- or quantum information processing. First, we show that ration of complex entangled states [6], and basic pro- a variety of spin dynamics can be retrieved by a digital tocols for quantum error correction [7], are among the decomposition in a generic quantum simulator. Then, quantum information tasks that can be performed with we consider prototypical spin models, simulation times, good fidelities. Moreover, superconducting circuits have and fidelities with current circuit QED technology, show- reached sufficient complexity and potential scalability to ing the computational power of superconducting qubits be considered as quantum simulators. in terms of digital quantum simulations. In this way, we A quantum simulator is a platform that allows us to analyze the required resources in a realistic setup for a reproduce the behavior of another quantum system. The multipurpose quantum simulator of spin dynamics capa- original idea of quantum simulation can be traced back ble of emulating a general many-qubit spin Hamiltonian. to Feynman [8], while the first mathematical formula- Most physical Hamiltonians can be written as a sum of PN tion using local interactions was proposed some years local terms, H = k=1 Hk, where each Hk acts on a local later [9]. So far, initial steps for quantum simulations in Hilbert space. The dynamics of a generic Hamiltonian H circuit QED have been taken, where a few analog quan- can be approximated by discrete stepwise unitaries, up tum simulators have been proposed in superconducting to arbitrary small errors, according to the formula (~ = 1 qubits [10–17]. On the other hand, an experiment of here and in the following) [9], discrete-time gate sequences to reproduce the dynam- l iHt iH1t/l iHN t/l e− = e− e− ics of a given spin Hamiltonian has been recently real- ··· ized in ion-trap [18] and photonic [19] systems, together 2 X [Hi,Hj]t X∞ with proposals for the emulation of interacting fermionic- + + E(k), (1) 2l bosonic models [20, 21]. The digital decomposition of i
Heisenberg interaction.— Digital methods can be used a) Rx† Rx Ry† Ry to simulate the Heisenberg spin model with available re- Hxy Hxy Hxy sources in superconducting circuits. We consider a setup Rx† Rx Ry† Ry made of several transmon qubits coupled to a single coplanar microwave resonator [27], b) Rx† Rx Ry† Ry N Hxy Hxy Hxy T X h 2 Rx† Rx Ry† Ry H =ω a†a + 4E (n n ) E cos φ r C,i i − g,i − J,i i Hxy Hxy Hxy i=1 Rx† Rx Ry† Ry i +2βieVrmsni(a + a†) . (2)
c) Rx Rx Here, ni, ng,i and φi stand, respectively, for the quantized Hxy Hxy Hxy Hxy charge on the superconducting island, the offset charge Rx Hxy Hxy and the quantized flux of the ith transmon qubit. The op- erators a(a†) act on the resonator field, whose first mode has frequency ωr. EC,i is the charging energy of the su- FIG. 1. Protocols for digital quantum simulations with trans- perconducting island, while E = Emax cos(πΦ /Φ ) is J,i J,i | i 0 | mon qubits. (a) Heisenberg model of two qubits. (b) Heisen- the Josephson energy of the dc-SQUID loop embedded in berg model of three qubits. (c) Frustrated Ising model of x(y) x the ith qubit. The latter can be tuned from small values three qubits. Here, Rx(y) ≡ R (π/4) and Rx ≡ R (π/2). max Note that exchanging each R matrix with its adjoint does not up to EJ,i by changing the ratio between the external affect the protocols. magnetic flux Φi, that threads the loop, and the elemen- tary flux quantum Φ0. Here, βi are renormalization coef- ficients of the couplings due to circuit capacitances, Vrms xy is the root mean square voltage of the resonator, and e is qubits interact for a time t with the H12 Hamiltonian. x the electron charge. Typical transmon regimes consider Step 4.— Application of single qubit rotation R12†(π/4) ratios of Josephson to charging energy E /E 20. to both qubits. Step 5.— Application of single qubit ro- J C & y Notice that cavity and circuit QED platforms do not tation R12(π/4) to both qubits. Step 6.— The qubits xy feature the Heisenberg interaction from first principles. interact for a time t according to the H12 Hamiltonian. y Nevertheless, one can consider a digital simulation of the Step 7.— Application of single qubit rotation R12†(π/4) model. We show that the coupled transmon-resonator to both qubits. Consequently, the total unitary evolution system, governed by the Hamiltonian in Eq. (2), can sim- reads ulate Heisenberg interactions of N qubits, which in the xy xz yz H H iH12 t iH12 t iH12 t iH12t case of homogeneous couplings reads U12(t) = e− e− e− = e− . (4)
N 1 This evolution operator simulates the dynamics of Eq. (3) H X− x x y y z z H = J σi σi+1 + σi σi+1 + σi σi+1 . (3) for two qubits. Arbitrary inhomogeneities of the cou- i=1 plings can be achieved by implementing different sim- ulated phases for different digital steps. Notice that, in Here the Pauli matrices σj, j x, y, z refer to the i ∈ { } this case, just one Trotter step is needed to achieve a sim- subspace spanned by the first two levels of the ith trans- ulation without digital errors, due to the commutativity mon qubit. We begin by considering the simplest case, xy xz yz of H12 , H12 , and H12 . Thus, from a practical point of in which two qubits are involved. The XY exchange in- view, the only source of errors will come from accumu- teraction can be directly reproduced by dispersively cou- lated gate errors. One can assume two-qubit gates with pling two transmon qubits to the same resonator [28–30], an error of about 5% and eight π/4 single qubit rotations xy + + x x y y H12 = J σ1 σ2− + σ1−σ2 = J/2 (σ1 σ2 + σ1 σ2 ). The with errors of 1%. This will give a total fidelity of the XY exchange interaction can be transformed via local ro- protocol around 77%. Moreover, the total execution time tations of the single qubits to get the effective Hamiltoni- for a π/4 simulated XYZ phase will be of about 0.10 µs. xz x xy x x x z z ans H12 = R12(π/4)H12 R12†(π/4) = J/2 (σ1 σ2 + σ1 σ2 ) Throughout the Letter, we compute the execution times yz y xy y y y z z and H12 = R12(π/4)H12 R12†(π/4) = J/2 (σ1 σ2 + σ1 σ2 ). by summing the corresponding times of all the employed x(y) x(y) x(y) Here, R12 (π/4) = exp[ iπ/4(σ1 + σ2 )] represents gates, where we consider typical circuit QED values. a local rotation of the first− and second transmon qubits Now, we consider a digital protocol for the simulation along the x(y) axis. The XYZ exchange Hamiltonian of the Heisenberg interaction for a chain of three spins. xyz H12 can therefore be implemented according to the pro- When considering more than two spins, one has to take tocol shown in Fig. 1(a) with the following steps. Step into account noncommuting Hamiltonian steps, involv- 1.— The qubits interact for a time t according to the ing digital errors. This three-spin case is directly extend- xy XY Hamiltonian H12 . Step 2.— Application of single able to arbitrary numbers of spins. We follow a digital x qubit rotations R12(π/4) to both qubits. Step 3.— The approach for its implementation, as shown in Fig. 1(b). 3
0.0.0606 a) b) TABLE I. Execution times and error bounds for the 00.2.2 Heisenberg(H) and Ising(I) models with open(o) and 0.0.0404 periodic(p) boundary conditions for N qubits. Here θ ≡ Jt, F J/2 and gφ are, respectively, the coupling strength of the XY 00.1.1 and single-qubit gates, and τs is the pulse time required for a 0.02 1 0.02 single qubit rotation.
00 00 0 ⇡Pi/8/8 ⇡/4 0 ⇡Pi/8/8 ⇡/4 Execution time Error bound 2 Ho 4lτs + 6(N − 1)θ/J 24(N − 2)(Jt) /l c) d) 2 0.0.0707 Hp 4lτs + 6Nθ/J 24N(Jt) /l 0.2 0.2 2 Io 2(N − 1)lτs + θ/gφ + 4(N − 1)θ/J 2(N − 1)(Jt) /l 0.0.0505
F 2 Ip 2Nlτs + θ/gφ + 4Nθ/J 2N(Jt) /l 0.0.0303 00.1.1 1
0.0.0101 error and time domains in which the largest part of the 00 00 0 ⇡Pi/8/8 ⇡/4 0 ⇡Pi/8/8 ⇡/4 error in the quantum simulation is due to experimental ✓ ✓ gate errors. One can consider interactions with open and closed boundary conditions, adding an extra term cou- FIG. 2. Fidelity loss for simulated Hamiltonians for three pling the first and last spin. Extending this protocol to qubits, in the interval θ = [0, π/4], θ ≡ Jt. Curved lines show N qubits with open or periodic boundary conditions, we digital errors, while horizontal lines show the accumulated compute an upper bound on the second order Trotter er- error due to a single step error of . Red solid (black dotted) 2 2 ror Eopen = 24(N 2)(Jt) /l and Eperiodic = 24N(Jt) /l. lines stand for lower (higher) digital approximations l. (a) − −2 Ising interaction.— Here, we consider a generic Heisenberg model, with = 10 , l = 3, 5, and (b) = 5 × P x x 10−2, l = 2, 3. (c) Transverse field Ising model, with = 10−2, N qubit Ising interaction J i σi σi+1, with periodic l = 3, 5 and (d) = 5 × 10−2, l = 2, 3. boundary conditions. Considering a three site model is sufficient to show the effect of frustration in the system. The antiferromagnetic interaction is inefficiently solvable Step 1.— Qubits 1 and 2 interact for a time t/l with the in a classical computer, while it is efficient for a quantum XY Hamiltonian. Step 2.— Qubits 2 and 3 interact for a simulator [32]. We consider the isotropic antiferromag- I P x x time t/l with the XY Hamiltonian. Step 3.– Application netic case between three sites , H123 = J i Ising model with transverse magnetic field, considering 11 an error for each step due to the imperfect gates. The time for simulating the transverse field Ising model for 00.75.75 the considered dynamics is about 190 ns per Trotter step. 00.5.5 The protocol can also be extended to N qubits with open and periodic boundary conditions, where we compute an 00.25.25 FidelityF x upper bound to the second order error in Jt/l of Eopen = Spin1 1 h xi 2 2 0 Spin1Time 1 ⇢ 2(N 1)(Jt) /l and E = 2N(Jt) /l. We report in 0 h xi periodic Spin2 2 − h xi Table I execution times and error bounds for the models Spin2Time 2 ⇢ −00.25.25 h i proposed, for N qubits. In general, given the nonlocal 0 \Pi/8⇡/8 \Pi/4⇡/4 character of the microwave resonator acting as a quantum ✓ bus, one can emulate 2D and 3D interaction topologies. In order to estimate the feasibility of the protocols in FIG. 3. Dynamics for the simulated Heisenberg model a superconducting circuit setup, we perform a numer- for√ two transmon qubits, which are initialized in the state 1/ 5(|↑i + 2 |↓i) ⊗ |↓i. Fidelity F = Tr(ρ|ΨI ihΨI |) shows ical simulation for the Heisenberg interaction between the behavior of the protocol for a given simulated phase θ. two transmon qubits coupled to a coplanar waveguide x The ideal spin dynamics hσi i for both qubits is plotted ver- x resonator. We compute the effect on the protocol of a re- sus mean values hσi iρ obtained with the qubit Hamiltonian alistic XY interaction, given as an effective second order Ht. Hamiltonian, obtained from the first order Hamiltonian, 2 2 X X j where we have defined the Lindblad superoperators Ht = ω i, j i, j + ωra†a ˆ ˆ ˆ ˆ ˆ ˆ ˆ i | i h | L(A)ρ = (2AρA† A†Aρ ρA†A)/2. We have set a i=0 j=1 decay rate for the− resonator− of κ = 2π 10 kHz, and a 2 2 × X X dephasing and decay rate for the single transmon qubit + gi,i+1( i, j i + 1, j + H.c.)(a + a†). (8) | i h | of Γφ = Γ = 2π 20 kHz. We perform a numerical i=0 j=1 simulation− for the Heisenberg× protocol for two transmon j qubits, following the steps as in Fig. 1(a), using for the Here, ωi is the transition energy of the ith level, with XY interaction steps the result of the dynamics obtained respect to the ground state, of the jth qubit, and ωr by solving Eq. (9), and ideal single-qubit rotations. The is the transition frequency of the resonator. We con- result is plotted in Fig. 3. The evolution for the density sider the first three levels for each transmon qubit, and j j j matrix ρ, that encodes the dynamics of the two transmon a relative anharmonicity factor of αr = (ω 2ω )/ω = 2 1 1 qubits, is compared to the exact quantum evolution Ψ I , 0.1, typical for the transmon regime [27].− We assume | i − that evolves according to the Hamiltonian in Eq. (3), identical transmon devices, with transition frequencies 2 2 2 1,2 with J = g01ω1/(ω1 ωr ) 2π 6 MHz. One can ob- ω ω1 = 2π 5 GHz. The resonator frequency is − ≈ × 1 serve that good simulation fidelities F = Tr(ρ ΨI ΨI ) set to≡ω = 2π ×7.5 GHz. We consider the coupling | ih | r × are achieved for nontrivial dynamics. Note that the ac- between different levels of a single transmon qubit [27] tion of the Heisenberg Hamiltonian on an initial state, gi,i+1 = √i + 1g0, where g0 = 2βeVrms = 2π 200 MHz. z z × which is also an eigenstate of the σ1 σ2 operator, would The chosen experimental parameters are typical for su- be equivalent to the one of the XY exchange interac- perconducting circuit setups and they can be optimized tion. To show signatures of the Heisenberg interaction, for each platform. The resonator-transmon coupling we choose in our simulation an initial state which does Hamiltonian, in the interaction picture with the free en- P j not have this property. One can also notice the typical ergy ω i, j i, j +ω a†a, results in an effective cou- i,j i | i h | r small time-scale fidelity oscillations due to the first order pling between the first two levels of the two transmon part of the dispersive exchange interaction. By further qubits H = [g2 ω /(ω2 ω2)](σxσx + σyσy), where we eff 01 1 1 − r 1 2 1 2 detuning the qubits from the resonator, one can reduce have considered negligible cavity population a†a 0 h i ≈ the contribution of the nondispersive part of the inter- and renormalization of the qubit frequencies to cancel action, and increase the global fidelity of the protocol. Lamb shifts. Here we have defined a set of Pauli matri- ces for the subspace spanned by the first two levels of In conclusion, we have proposed a digital quantum each transmon, e.g. σx 0, 1(2) 1, 1(2) + H.c. In 1(2) simulation of spin chain models in superconducting cir- order to estimate the effect≡ of | decoherencei h in| a realistic cuits. We have considered prototypical models such as setup, we consider the master equation dynamics, the Heisenberg and frustrated Ising interactions. Fur- 2 thermore, we have shown the feasibility of the simulation X z with state-of-the-art technology of transmon qubits cou- ρ˙ = i[Ht, ρ] + κL(a)ρ + ΓφL(σi )ρ + Γ L(σi−)ρ , − − i=1 pled to microwave resonators. In the near future, these (9) protocols may be extended to many-qubit spin models, 5 paving the way towards universal quantum simulation of and E. Solano, Phys. Rev. X 2, 021007 (2012). spin dynamics in circuit QED setups. [16] J. S. Pedernales, R. Di Candia, D. Ballester, and E. We acknowledge funding from the Basque Gov- Solano, New. J. Phys. 15, 055008 (2013). ernment IT472-10, Spanish MINECO FIS2012-36673- [17] O. Viehmann, J. von Delft, and F. Marquardt, Phys. Rev. Lett. 110, 030601 (2013). C03-02, Ram´ony Cajal RYC-2012-11391, UPV/EHU [18] B. P. Lanyon, C. Hempel, D. Nigg, M. M¨uller,R. Ger- UFI 11/55, CCQED, PROMISCE and SCALEQIT EU ritsma, F. Z¨ahringer, P. Schindler, J. T. Barreiro, projects. M. Rambach, G. Kirchmair, M. Hennrich, P. Zoller, R. Blatt, and C. F. Roos, Science 334, 57 (2011). [19] B. P. Lanyon, J. D. Whitfield, G. G. Gillet, M. E. Goggin, M. P. Almeida, I. Kassal, J. D. Biamonte, M. Mohseni, B. J. Powell, M. Barbieri, A. Aspuru-Guzik, and A. G. [1] M. H. Devoret and R. J. Schoelkopf, Science 339, 1169 White, Nat. Chem. 2, 106 (2009). (2013). [20] J. Casanova, A. Mezzacapo, L. Lamata, and E. Solano, [2] A. Wallraff, D. I. Schuster, A. Blais, L. Frunzio, R.-S. Phys. Rev. Lett. 108, 190502 (2012). Huang, J. Majer, S. Kumar, S. M. Girvin, and R. J. [21] A. Mezzacapo, J. Casanova, L. Lamata, and E. Solano, Schoelkopf, Nature (London) 431, 162 (2004). Phys. Rev. Lett. 109, 200501 (2012). [3] A. Fedorov, L. Steffen, M. Baur, M. P. da Silva, and A. [22] M. Suzuki, Phys. Lett. A 146, 319 (1990). Wallraff, Nature (London) 481, 170 (2012). [23] D. W. Berry, G. Ahokas, R. Cleve, and B. C. Sanders, [4] A. A. Abdumalikov, J. M. Fink, K. Juliusson, M. Perchal, Commun. Math. Phys. 270, 359 (2007). S. Berger, A. Wallraff, and S. Filipp, Nature (London) [24] S. Chen, L. Wang, S.-J. Gu, and Y. Wang, Phys. Rev. E 496, 482 (2013). 76, 061108 (2007). [5] J. M. Chow, J. M. Gambetta, A. D. C´orcoles, S. T. [25] L. F. Santos, F. Borgonovi, and F. M. Izrailev, Phys. Merkel, J. A. Smolin, C. Rigetti, S. Poletto, G. A. Keefe, Rev. Lett. 108, 094102 (2012). M. B. Rothwell, J. R. Rozen, M. B. Ketchen, and M. [26] P. W. Anderson, Science 235, 1196 (1987). Steffen, Phys. Rev. Lett. 109, 060501 (2012). [27] J. Koch, T. M. Yu, J. Gambetta, A. A. Houck, D. I. [6] M. Neeley, R. C. Bialczak, M. Lenander, E. Lucero, M. Schuster, J. Majer, A Blais, M. H. Devoret, S. M. Girvin Mariantoni, A. D. O’Connell, D. Sank, H. Wang, M. Wei- and R. J. Schoelkopf, Phys. Rev. A 76, 042319 (2007). des, J. Wenner, Y. Yin, T. Yamamoto, A. N. Cleland, and [28] A. Blais, R.-S. Huang, A. Wallraff, S. M. Girvin, and R. J. M. Martinis, Nature (London) 467, 570 (2010). J. Schoelkopf, Phys. Rev. A 69, 062320 (2004). [7] M. D. Reed, L. DiCarlo, S. E. Nigg, L. Sun, L. Frunzio, S. [29] J. Majer, J. M. Chow, J. M. Gambetta, Jens Koch, B. M. Girvin, and R. J. Schoelkopf, Nature (London) 482, R. Johnson, J. A. Schreier, L. Frunzio, D. I. Schuster, A. 382 (2012). A. Houck, A. Wallraff, A. Blais, M. H. Devoret, S. M. [8] R. P. Feynman, Int. J. Theor. Phys. 21, 467 (1982). Girvin, and R. J. Schoelkopf, Nature (London) 449, 443 [9] S. Lloyd, Science 273, 1073 (1996). (2007). [10] J. J. Garc´ıa-Ripoll, E. Solano, and M. A. Martin- [30] S. Filipp, M. G¨oppl,J. M. Fink, M. Baur, R. Bianchetti, Delgado, Phys. Rev. B 77, 024522 (2008). L. Steffen, and A. Wallraff, Phys. Rev. A 83, 063827 [11] L. Tian, Phys. Rev. Lett. 105, 167001 (2010). (2011). [12] E. J. Pritchett, C. Benjamin, A. Galiautdinov, M. R. [31] C. Rigetti, J. M. Gambetta, S. Poletto, B. L. T. Plourde, Geller, A. T. Sornborger, P. C. Stancil, and J. M. Mar- J. M. Chow, A. D. C´orcoles,J. A. Smolin, S. T. Merkel, tinis, arXiv:1008.0701. J. R. Rozen, G. A. Keefe, M. B. Rothwell, M. B. Ketchen, [13] Y. Zhang, L. Yu, J.-Q. Liang, G. Chen, S. Jia, and F. and M. Steffen, Phys. Rev. B 86, 100506(R) (2012). Nori, Sci. Rep. 4, 4083 (2014). [32] K. Kim, M.-S. Chang, S. Korenblit, R. Islam, E. E. Ed- [14] F. Mei, V. M. Stojanovi´c,I. Siddiqi, and L. Tian, Phys. wards, J. K. Freericks, G.-D. Lin, L.-M. Duan, and C. Rev. B 88, 224502 (2013). Monroe, Nature (London) 465, 590 (2010). [15] D. Ballester, G. Romero, J. J. Garc´ıa-Ripoll, F. Deppe,