Quantum annealing with capacitive-shunted flux qubits

Yuichiro Matsuzaki1,∗ Hideaki Hakoshima1,† Yuya Seki1, and Shiro Kawabata1 1 Nanoelectronics Research Institute (NeRI), National Institute of Advanced Industrial Science and Technology (AIST) Umezono 1-1-1, Tsukuba, Ibaraki 305-8568 Japan

Quantum annealing (QA) provides us with a way to solve combinatorial optimization problems. In the previous demonstration of the QA, a superconducting flux qubit (FQ) was used. However, the flux qubits in these demonstrations have a short coherence time such as tens of nano seconds. For the purpose to utilize quantum properties, it is necessary to use another qubit with a better coherence time. Here, we propose to use a capacitive-shunted flux qubit (CSFQ) for the implementation of the QA. The CSFQ has a few order of magnitude better coherence time than the FQ used in the QA. We theoretically show that, although it is difficult to perform the conventional QA with the CSFQ due to the form and strength of the interaction between the CSFQs, a spin-lock based QA can be implemented with the CSFQ even with the current technology. Our results pave the way for the realization of the practical QA that exploits quantum advantage with long-lived qubits. arXiv:2001.09844v2 [quant-ph] 17 Feb 2020

∗ [email protected] † [email protected] 2

I. INTRODUCTION

Quantum annealing (QA) and adiabatic quantum computation (AQC) are attractive ways to solve combinatorial optimization problems where a cost function is minimized with a suitable parameter set [1–4]. It is known that we can map some optimization problems into a task of a ground-state search of the Ising Hamiltonian [5]. Quantum annealing and AQC can find the ground state of the Ising Hamiltonian by taking advantage of quantum fluctuations [6]. In particular, AQC is guaranteed to converge to the ground state if the quantum adiabatic condition is satisfied [4]. There are many experimental and theoretical studies along this direction [7–11].

A superconducting flux qubit (FQ) [12, 13] has been used to demonstrate the QA in the previous demonstration [14]. The superconducting qubit is an artificial atom, and so we have many degrees of freedom to control the parameters [15, 16]. The changes in the design of the superconducting circuit allow us to control the properties of the superconducting qubit in the fabrication. After the fabrication, control lines coupled with the superconducting qubits can tune the parameters such as the qubit resonant frequency [17, 18] and the coupling strength between qubits [19]. Such a controllability is essential for the implementation of the QA. A potential problem is that the FQs used in the QA have a short coherence time such as tens of nano seconds [20]. Such a short coherence time could make it difficult to exploit the quantum advantage for the QA. There are many ways to improve the coherence time of the FQ. Symmetric designs of the device is known to be one of the important ingredients for the long coherence time [21–24]. The use of the 3D cavity containing the FQ can improve the coherence time due to the change in the environment of the FQ [25]. However, from the view point of the scalability and reproducability, further improvement is required to use the FQ for the QA to solve the practical problems with high coherence.

Recently, a capacitive shunt flux qubit (CSFQ) was fabricated [26–29] where an additional capacitor shunted in parallel to the Josephson junction in the loop is introduced. This structure is known to suppress noise. At the optimal point, the CSFQ has a coherence time of tens of micro seconds [26, 27], which is a few order of magnitude longer than the FQs used in the QA. Due to this long coherence time, the CSFQ is considered as a promising device to realize quantum information processing.

However, there are two main difficulties to use the CSFQ for the QA. Firstly, the CSFQ has a much smaller coupling strength than the FQ used in the QA. The CSFQ has a persistent current of tens of nano ampere [27] while the FQs used in the QA has a few micro ampere of the persistent current [14]. This results in a few orders of magnitude smaller coupling strength of the CSFQs than that of the FQs. In the QA, it is necessary for the qubits to have a coupling strength that is comparable with the qubit resonant frequency, and the CSFQ may not satisfy such a condition due to the weak coupling strength. Secondly, near the optimal operating point where the CSFQ has a longer coherence time, there are not only Ising type interaction but also a flip-flop type interaction with the CSFQ [27]. In the QA, only Ising type interaction is required at the end of the calculation, and the residual flip-flop interaction could reduce the success probability to obtain an accurate solution.

Here, we propose to implement the QA with the CSFQs by using a spin-lock technique. The spin lock technique is a way to keep the quantum state in |+i (an eigenstate of σx) in a rotating frame. Firstly, we prepare a state of |0i (a ground state of σz). Secondly, we perform a π/2 pulse of the microwave along the y axis to prepare a state of |+i. Thirdly, we continuously perform the microwave driving along x axis. Since |+i is an eigenstate of the Hamiltonian in the rotating frame, we can keep the state in |+i due to the driving, which is called the spin lock. This is a common technique in the field of the magnetic resonance [30]. Importantly, during the spin lock, the effective frequency of the qubit is the detuning between the qubit bare frequency and microwave driving frequency. Also, if there is a large detuning between the qubit bare frequencies, the effect of the flip-flop type interaction between the qubits can be suppressed. Therefore, the spin lock technique can overcome the limitation of the CSFQ to be used for the QA. In the NMR, the spin-lock based QA has been discussed and demonstrated [31, 32]. While DC magnetic fields are applied along x axis in the conventional QA, the AC driving fields along x axis are applied in the spin lock-based QA. However, since it is difficult to scale up in the NMR, a practical advantage is unclear to use the NMR for the QA. On the other hand, we propose to use the CSFQs for the spin-lock based QA, which is considered as a scalable device in quantum information processing.

We theoretically study the performance of the spin-lock based QA with the CSFQs. The spin-lock based QA becomes equivalent to the conventional QA under the assumption that the rotating wave approximation (RWA) is valid. This means that, the performance of the spin-lock based QA can be degraded if the Rabi frequency of the driving is comparable with the qubit frequency. Since the qubit frequency (an order of GHz) is just a few orders of magnitude larger than the other typical frequencies (an order of MHz) [31], a careful assessment of the error accumulation due to the violation of the RWA is necessary to evaluate the practicality of the spin-lock based QA with the CSFQs. It is worth mentioning that, in our previous study at the SSDM [33], we implement a numerical simulation along this direction with a fixed number of the qubits. On the other hand, we investigate how the performance of the QA with spin lock changes when we increase the number of the qubits, and discuss how our scheme scales in this paper. Moreover, we analyze the expression of the violation of the RWA. 3

II. THE CONVENTIONAL QUANTUM ANNEALING WITH DC TRANSVERSE MAGNETIC FIELDS

Let us review the conventional quantum annealing where DC transverse magnetic fields are applied [1]. The Hamiltonian in the QA is described as follows.

−γ2t2 −γ2t2 HQA = e HTR + (1 − e )HIsing, (1) L L Õ h Õ J 0 0 H = − i σ(i) − i,i σ(i)σ(i ), (2) Ising 2 z 2 z z i=1 i,i0=1 L Õ Λ H = − σ(i), (3) TR 2 x i=1 where γ is the sweeping rate of the Hamiltonian, hi is the resonant frequency of the i-th qubits, Ji,i0 is the interaction strength between the i-th qubit and i0-th qubit, and Λ denotes the amplitude of the transverse magnetic fields. Here, we consider that the ground state of HIsing is the solution of optimization problems which we want to find. We first prepare a state of |+ + ··· +i, which is the ground state of HQA at t = 0, and we gradually turn off the transverse magnetic fields with a time scale determined by γ, while we adiabatically increase the term HIsing with the same time scale. If γ is much smaller than the energy gap between the ground state and first excited state of HQA for all t, the quantum state continues −1 the ground state of HQA for all t because of the quantum adiabatic theorem, and after the long time evolution (t  γ ), we can obtain a ground state of HIsing.

III. QA WITH SPIN LOCK TECHNIQUE

We explain the details of the QA with the spin-lock technique. Firstly, we prepare a state of |00 ··· 0i. Secondly, we apply a global π/2 pulse so as to prepare a state of |+ + ··· +i. Thirdly, we continue driving all these qubits with the AC magnetic fields along x direction, and gradually turn off the transverse driving fields while we gradually turn on HIsing. Finally, we measure the qubits. In the third step of this protocol, the state is governed by the unitary time evolution determined by the Hamiltonian H

−γ2t2 −γ2t2 H = H0 + e HD + (1 − e )(HIsing + Hxy), (4) L Õ ω + δω H = i σ(i), (5) 0 2 z i=1 L Õ (i) HD = − λ cos [(ω + δωi)t]σx , (6) i=1 L L Õ J 0 0 Õ J 0 0 H = − i,i σ(i)σ(i ) − i,i σ(i)σ(i ), (7) xy 2 x x 2 y y i,i0=1 i,i0=1 where (ω + δωi) denotes a bare frequency of the i-th qubit and λ denotes the Rabi frequency of the driving fields. In a rotating frame by the unitary transformation exp [iH0t], we neglect the fast oscillating terms with the frequencies (ω + δωi) in the Hamiltonian, which is called a rotating wave approximation (RWA). Under this approximation, the Hamiltonian H becomes equivalent to HQA in the Eq. (1) by setting λ = Λ This approximation becomes valid when (ω + δωi) is sufficiently large.

IV. EVALUATION OF THE PERFORMANCE OF THE QA WITH THE SPIN-LOCK TECHNIQUE

We investigate the potential errors during the spin-lock based QA derived from the violation of the RWA using numerical simulations. We solve a time-dependent Schrödinger equation with Hamiltonian (4) from t = 0 (the initial state is |+ + ··· +i) to t = Tend. We demonstrate a ferromagnetic one-dimensional Ising model with free boundary condition. In particular, we consider the homogeneous case hj = h > 0 and Ji,i+1 = J > 0 (i = 1, 2, ··· , L − 1) with the nearest neighbor interaction. In this case, the ground state of HIsing is an all up state of |11 ··· 1i. To quantify the accuracy to find the ground state of HIsing, we investigate a fidelity F(t) = | h11 ··· 1|Φ(t)i |2, where |Φ(t)i denotes a solution of the Schrödinger equation at a time t. F(t) means the overlap between the true ground state of HIsing and |Φ(t)i, which is the state created by the spin-lock based QA. If F(t) is close to 1, |Φ(t)i is close to the true ground state, and therefore our method works very well. 4

FIG. 1. The infidelity of the state plotted against time. Here, we set the parameters as L = 4, λ/2π = 1 GHz, h/2π = 0.03 GHz, γ = 0.01 GHz, J/2π = 0.05 GHz, Tend = 500 ns, ω/2π = 2.4, 3.6, 4.8 GHz, and δωj /2π = 1.9(j − 1) GHz. These parameters are typical in the CSFQ [34].

We plot an infidelity (1 − F(t)) against time with feasible parameters of the CSFQs in the Fig. 1. From this figure, we find that in all three cases, the infidelities are sufficiently close to 0 (less than 0.01), and we can achieve the high fidelity even by realistic parameters. Moreover, the infidelity decreases as the qubit frequency increases. This reveals that the larger frequency of the qubit becomes the high accuracy of the RWA, which reduces the error in the spin-lock based QA. Also, we plot the fidelity of the state at the end of the QA against the number of the qubits as shown in the Fig. 2. The fidelity decreases almost linearly with the number of the qubits. Such a linear scaling of the infidelity is typically observed when the independent noise acts on the qubit [35]. So our numerical results indicate that the noise due to the violation of the RWA does not show a spatial correlation but have independent noise properties.

V. ANALYTICAL EXPRESSION OF THE VIOLATION OF THE RWA

In this section, to understand the dynamics shown in Fig. 1, we derive an effective Hamiltonian and will show that the ground state of the effective Hamiltonian is a good approximation of the exact state numerically obtained by solving the time-dependent Schrödinger equation. We explain the effective dynamics under the periodic driving fields. In our analysis, we assume that γ−1 is much larger than any other time scale so that the adiabatic condition should be satisfied. By going to the rotating frame, the Hamiltonian in the Eq. (4) is written as follows.

−γ2t2 −γ2t2 H(t) = HQA + e HD0 (t) + (1 − e )Hxy0 (t), (8) L Õ Λ  (i) (i) H 0 = − cos [2(ω + δω )t]σ − sin [2(ω + δω )t]σ , (9) D 2 i x i y i=1 L Õ Ji,i0 h  (i) (i0) (i) (i0) H 0 = − cos [(δω − δω 0 )t] σ σ + σ σ xy 2 i i x x y y i,i0=1  (i) (i0) (i) (i0)i + sin [(δωi − δωi0 )t] σx σy − σy σx . (10) 5

FIG. 2. The fidelity of the state at the end of the QA plotted against the number of the qubits. Here, we set the parameters as λ/2π = 1 GHz, h/2π = 0.03 GHz, γ = 0.01 GHz, J/2π = 0.05 GHz, Tend = 500 ns, ω/2π = 4.8 GHz, and δωj /2π = 1.9(j − 1) GHz.

We can perform the Magnus expansion [36, 37] up to the second order, and obtain the following

∫ t −i H(t0)dt0 −i(Ω (t)+Ω (t)) U(t) = T{e 0 }' e 1 2 , (11) ¹ t 0 0 Ω1(t) = H(t )dt , (12) 0 i ¹ t ¹ t2 Ω2(t) = − dt2 dt1[H(t2), H(t1)]. (13) 2 0 0 where U(t) is the time evolution operator and T denotes the time ordering product.

In order to obtain a simple analytical form, we especially consider the case of L = 2. Let us define a period of the oscillation −1 of our system as T = 2πn1/(ω + δω1) = 2πn2/(ω + δω2) where n1 and n2 are positive small integers with T  γ . We can define the average Hamiltonian as follows

1 H = Ω (T), (14) 1 T 1 1 H = Ω (T). (15) 2 T 2

Starting from a ground state of the Hamiltonian at t = 0, the system remains in a ground state of the Hamiltonian of (H1 + H2) for a small γ. Note that H1 corresponds to HQA (RWA) while H2 corresponds to the violation of the RWA. We can calculate H2 as follows. 6 1 0.500

0.100 0.050

0.010 0.005

0.001 0 100 200 300 400 500

FIG. 3. The infidelity between |11i and the ground state of the Hamiltonian H1 + H2 plotted against time. Here, we set the same parameters as Fig. 1 ( ω/2π = 2.4 GHz).

2 (1) (2) ! Λ −2γ2t2 σz σz H2 = − e + 8 (ω + δω1) (ω + δω2) (1) (2) (1) (2) (1) (2) !! Λ 2 2 2 2 h σ h σ σ σ σ σ + e−γ t (1 − e−γ t ) 1 x + 2 x + J x z + z x 4 (ω + δω1) (ω + δω2) (ω + δω1) (ω + δω2)

JΛ −γ2t2 −γ2t2  (1) (2) (1) (2) + e (1 − e ) σx σz − σz σx 2(δω1 − δω2) 2 −γ2t2 2 J (2) (1) + (1 − e ) (σz − σz ) (16) 2(δω1 − δω2)

Now, we consider a ground state of the Hamiltonian H1 + H2 at each t. In Fig. 3, we plot the infidelity between the state and |11i against time, and compare this result with the infidelity obtained by numerically solving the time-dependent Schrödinger equation. There is a good agreement between the results from the effective Hamiltonian and that from numerical solutions. This shows that our effective Hamiltonian approximately explains the dynamics shown in Fig. 1, which provides us with the intuition to understand our scheme.

VI. CONCLUSION

In conclusion, we propose to implement the QA with CSFQs. Although it is difficult to perform the conventional QA by using the CSFQs because of the weak coupling strength and residual flip-flop interactions, we show that a use of the spin-lock based QA can overcome these problems. Our numerical simulations show that the spin-lock based QA can be implemented even with the current technology.

ACKNOWLEDGMENTS

This work was supported by Leading Initiative for Excellent Young Researchers MEXT Japan, and is partially supported by MEXT KAKENHI (Grant No. 15H05870) and the New Energy and Industrial Technology Development Organization (NEDO), 7

Japan.

Y.M and H.H equally contributed to this paper.

[1] Tadashi Kadowaki and Hidetoshi Nishimori. Quantum annealing in the transverse ising model. Physical Review E, 58(5):5355, 1998. [2] Edward Farhi, Jeffrey Goldstone, Sam Gutmann, and Michael Sipser. Quantum computation by adiabatic evolution. arXiv preprint quant-ph/0001106, 2000. [3] Edward Farhi, Jeffrey Goldstone, Sam Gutmann, Joshua Lapan, Andrew Lundgren, and Daniel Preda. A quantum adiabatic evolution algorithm applied to random instances of an np-complete problem. Science, 292(5516):472–475, 2001. [4] Tameem Albash and Daniel A Lidar. Adiabatic quantum computation. Reviews of Modern Physics, 90(1):015002, 2018. [5] Andrew Lucas. Ising formulations of many np problems. Frontiers in Physics, 2:5, 2014. [6] Satoshi Morita and Hidetoshi Nishimori. Mathematical foundation of quantum annealing. Journal of Mathematical Physics, 49(12):125210, 2008. [7] Giuseppe E Santoro, Roman Martoňák, Erio Tosatti, and Roberto Car. Theory of quantum annealing of an ising spin glass. Science, 295(5564):2427–2430, 2002. [8] Giuseppe E Santoro and Erio Tosatti. Optimization using quantum mechanics: quantum annealing through adiabatic evolution. Journal of Physics A: Mathematical and General, 39(36):R393, 2006. [9] Trevor Lanting, Anthony J Przybysz, A Yu Smirnov, Federico M Spedalieri, Mohammad H Amin, Andrew J Berkley, Richard Harris, Fabio Altomare, Sergio Boixo, Paul Bunyk, et al. Entanglement in a quantum annealing processor. Physical Review X, 4(2):021041, 2014. [10] Sergio Boixo, Troels F Rønnow, Sergei V Isakov, Zhihui Wang, David Wecker, Daniel A Lidar, John M Martinis, and Matthias Troyer. Evidence for quantum annealing with more than one hundred qubits. Nature Physics, 10(3):218, 2014. [11] Hayato Goto. Bifurcation-based adiabatic quantum computation with a nonlinear oscillator network. Scientific reports, 6:21686, 2016. [12] JE Mooij, TP Orlando, L Levitov, Lin Tian, Caspar H Van der Wal, and Seth Lloyd. Josephson persistent-current qubit. Science, 285(5430):1036–1039, 1999. [13] T. P. Orlando, J. E. Mooij, L. Tian, C. H. van der Wal, L. S. Levitov, S. Lloyd, and J. J. Mazo. Superconducting persistent-current qubit. Phys. Rev. B, 60(22):15398, 1999. [14] Mark W Johnson, Mohammad HS Amin, Suzanne Gildert, Trevor Lanting, Firas Hamze, Neil Dickson, R Harris, Andrew J Berkley, Jan Johansson, Paul Bunyk, et al. Quantum annealing with manufactured spins. Nature, 473(7346):194, 2011. [15] Yuriy Makhlin, Gerd Schön, and Alexander Shnirman. Quantum-state engineering with josephson-junction devices. Reviews of modern physics, 73(2):357, 2001. [16] John Clarke and Frank K Wilhelm. Superconducting quantum bits. Nature, 453(7198):1031, 2008. [17] FG Paauw, A Fedorov, CJP M Harmans, and JE Mooij. Tuning the gap of a superconducting flux qubit. Physical review letters, 102(9):090501, 2009. [18] Xiaobo Zhu, Alexander Kemp, Shiro Saito, and Kouichi Semba. Coherent operation of a gap-tunable flux qubit. Applied Physics Letters, 97(10):102503, 2010. [19] AO Niskanen, K Harrabi, F Yoshihara, Y Nakamura, S Lloyd, and JS Tsai. Quantum coherent tunable coupling of superconducting qubits. Science, 316(5825):723–726, 2007. [20] Isil Ozfidan, Chunqing Deng, AY Smirnov, T Lanting, R Harris, L Swenson, J Whittaker, F Altomare, M Babcock, C Baron, et al. Demonstration of nonstoquastic hamiltonian in coupled superconducting flux qubits. arXiv preprint arXiv:1903.06139, 2019. [21] Guido Burkard, David P DiVincenzo, P Bertet, I Chiorescu, and JE Mooij. Asymmetry and decoherence in a double-layer persistent-current qubit. Physical Review B, 71(13):134504, 2005. [22] Patrice Bertet, Irinel Chiorescu, Guido Burkard, Kouichi Semba, CJPM Harmans, David P DiVincenzo, and JE Mooij. Dephasing of a superconducting qubit induced by photon noise. Physical review letters, 95(25):257002, 2005. [23] F Yoshihara, K Harrabi, AO Niskanen, Y Nakamura, and JS Tsai. Decoherence of flux qubits due to 1/f flux noise. Physical review letters, 97(16):167001, 2006. [24] Jonas Bylander, Simon Gustavsson, Fei Yan, Fumiki Yoshihara, Khalil Harrabi, George Fitch, David G Cory, Yasunobu Nakamura, Jaw-Shen Tsai, and William D Oliver. Noise spectroscopy through dynamical decoupling with a superconducting flux qubit. Nature Physics, 7(7):565, 2011. [25] Michael Stern, Gianluigi Catelani, Yuimaru Kubo, Cecile Grezes, Audrey Bienfait, Denis Vion, Daniel Esteve, and Patrice Bertet. Flux qubits with long coherence times for hybrid quantum circuits. Physical review letters, 113(12):123601, 2014. [26] JQ You, Xuedong Hu, S Ashhab, and Franco Nori. Low-decoherence flux qubit. Physical Review B, 75(14):140515, 2007. [27] Fei Yan, Simon Gustavsson, Archana Kamal, Jeffrey Birenbaum, Adam P Sears, David Hover, Ted J Gudmundsen, Danna Rosenberg, Gabriel Samach, Steven Weber, et al. The flux qubit revisited to enhance coherence and reproducibility. Nature communications, 7:12964, 2016. [28] Antonio D Córcoles, Jerry M Chow, Jay M Gambetta, Chad Rigetti, James R Rozen, George A Keefe, Mary Beth Rothwell, Mark B Ketchen, and Matthias Steffen. Protecting superconducting qubits from radiation. Applied Physics Letters, 99(18):181906, 2011. [29] Matthias Steffen, Shwetank Kumar, David P DiVincenzo, JR Rozen, George A Keefe, Mary Beth Rothwell, and Mark B Ketchen. High-coherence hybrid superconducting qubit. Physical review letters, 105(10):100502, 2010. [30] M Loretz, T Rosskopf, and CL Degen. Radio-frequency magnetometry using a single electron spin. Physical review letters, 110(1):017602, 2013. 8

[31] Hongwei Chen, Xi Kong, Bo Chong, Gan Qin, Xianyi Zhou, Xinhua Peng, and Jiangfeng Du. Experimental demonstration of a quantum annealing algorithm for the traveling salesman problem in a nuclear-magnetic-resonance quantum simulator. Physical Review A, 83(3):032314, 2011. [32] Mikio Nakahara. Lectures on Quantum Computing, Thermodynamics and Statistical Physics, volume 8. World Scientific, 2013. [33] Yuichiro Matsuzaki, Yuya Seki, and Shiro Kawabata. Quantum annealing with capacitive-shunted flux qubits. SSDM2019 : International Conference on Solid State Devices and Materials, 5 E-1-05, 2019. [34] Steven J Weber, Gabriel O Samach, David Hover, Simon Gustavsson, David K Kim, Alexander Melville, Danna Rosenberg, Adam P Sears, Fei Yan, Jonilyn L Yoder, et al. Coherent coupled qubits for quantum annealing. Physical Review Applied, 8(1):014004, 2017. [35] M Hein, W Dür, and H-J Briegel. Entanglement properties of multipartite entangled states under the influence of decoherence. Physical Review A, 71(3):032350, 2005. [36] M Matti Maricq. Application of average hamiltonian theory to the nmr of solids. Physical Review B, 25(11):6622, 1982. [37] Eugène S Mananga and Thibault Charpentier. Introduction of the floquet-magnus expansion in solid-state nuclear magnetic resonance spectroscopy. The Journal of chemical physics, 135(4):044109, 2011.