Access the full text.
Sign up today, get DeepDyve free for 14 days.
Loading next page...
/lp/springer-journals/recent-progress-on-coherent-computation-based-on-quantum-squeezing-qkdwD13s59
- Publisher
- Springer Journals
- Copyright
- Copyright © The Author(s) 2023
- ISSN
- 2309-4710
- eISSN
- 2309-4710
- DOI
- 10.1007/s43673-023-00077-4
- Publisher site
- See Article on Publisher Site
Abstract
Squeezed vacuum state of light is an important concept of quantum optics which has an uncertainty reduction in a specific quadrature compared to the coherent vacuum state. The coherent Ising machines (CIMs) based on the squeezed state are capable of searching the ground state of the Ising model, which can be used to solve combina‑ torial optimization problems and have been experimentally demonstrated to have excellent computational perfor‑ mance. This review introduces the recent progress of a CIM hardware solver based on optical parametric oscillators, including the delayed optical path scheme and a measurement feedback scheme. Also, the basic principles, unique advantages, and potential challenges are described. We expect that the applications of large‑ scale CIM hardware solv‑ ers will have a huge impact on the acceleration of the computation power. The combinatorial optimization problem requires find - 1 Introduction ing an optimal solution that minimizes the loss function In the past decade, the development of integrated elec- among the combinations of all options, which belongs to tronic circuits characterized by Moore’s law has slowed the non-deterministic polynomial time (NP)-hard prob- down, and various forms of non-silicon physical com- lem. There are many combinatorial optimization prob - puting have been highly anticipated [1–4]. Unlike digi- lems in various important domains, such as financial tal computers of von Neumann architecture, computers portfolios [18], drug discovery [19], integrated circuit in non-silicon forms take the task of simulating physical design [20], and Max-Cut problems [21]. Ising models are systems., such as superconducting quantum computers mathematical abstractions of spin glasses that describe [5–8], quantum annealers [9, 10], physical neural net- disordered magnetic systems. The Hamiltonian of Ising works [3], and coherent Ising machines (CIMs) [11–13]. machine could be expressed as H =− J σ σ , Among them, CIMs have been extensively studied due Ising ij i j i,j to their excellent performance on combinatorial optimi- where J denote the coupling coefficients between the i ij zation problems and have been implemented by various th and j th spins and σ ∈ {−1, 1} . The Max-Cut problem physical systems, such as optics [11–13], nanomagnetic- is equivalent to searching for the ground state energy of net arrays [14, 15], single-atom [16], and complementary an Ising model [21] oxide semiconductor devices [17]. Edges = J − J σ σ . ij ij i j (1) i,j i,j *Correspondence: Chuan Wang Therefore, the solution of the ground states of the Ising wangchuan@bnu.edu.cn School of Artificial Intelligence, Beijing Normal University, model is also an NP-hard problem. Beijing 100875, China Quantum computers should be able to simulate quan- Beijing QBoson Quantum Technology Co., Ltd., Beijing 100015, China tum systems more efficiently than classical computers Applied Optics Beijing Area Major Laboratory, Beijing Normal University, Beijing 100875, China [22, 23], but achieving large-scale quantum computers is still challenging at this stage. Many approximation © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. Lu et al. AAPPS Bulletin (2023) 33:7 Page 2 of 8 algorithms, such as Monte Carlo-based simulated solved. We believe this review can provide a reference annealing, can be applied to solve these combinato- for CIM research in the future. rial optimization problems. However, the solution time increases exponentially on conventional computers 2 Coherent Ising machines based on squeezed depending on the scale of the problem [24]. Therefore, states it is necessary to consider whether physical simulators Photons have been seen as a good medium for comput- can be used to replace traditional computers. There - ing and are often used in the field of quantum computing fore, the CIM hardware solver is the best choice for [27–29] due to their inherent advantages in information implementing a physical simulator to solve combinato- processing, such as highly parallel computing due to no rial optimization problems [25]. interference between different bands [2, 30–34], wide The noise of squeezed states exhibits the squeezing bandwidth from additional orbital angular momentum and anti-squeezing of the quadratures in the orthogo- and polarization [35, 36, 36, 37], and low loss over long nal components through nonlinear effects, respec - distances [38, 39]. tively. The squeezed light that anti-squeezed in-phase Compared with coherent light, the squeezed vacuum noise can help us to realize CIM with a quantum par- state of light has reduced uncertainty in a specific quad - allel search process. As a hardware solver, the experi- rature and can be widely used in quantum information mentally demonstrated excellent performance of CIM processing [40–42], gravitational wave detection 43, in combinatorial optimization has aroused extensive 44], quantum sensing [45–47], and optical parametric research interests [11, 12, 26]. In this brief review, we amplification [48]. Figure 1a [48] shows an experimental focus on the recent progress of squeezed-light-based scheme for generating squeezed light. The experiment CIMs, comparing their differences and the unique produces squeezed noise below the shot noise limit and advantages of their special structures. The main struc - anti-squeezed noise above the shot noise limit, as shown ture of the review is as follows: Section 2 mainly intro- in Fig. 1b [49]. Noise above the shot noise limit helps duces the principle and optical implementation of CIM to achieve the quantum parallel search. degenerate optical parametric oscillator (DOPO)-based In addition, the mutual injection between the optical CIM. Section 3 summarizes some challenges to be pulse networks can simulate the interaction relationship Fig. 1 Experimental scheme for generating squeezed light. a Experimental scheme for generating bright squeezed light and squeezed vacuum with an optical parametric oscillator (OPA). EOM, electro‑ optic modulator; DM, dichroic mirror; SHG, second harmonic generator; HR, high reflector. Figure 1a reproduced from Ref. [48] with permission of the Springer Nature. b Typical vacuum squeezing noise measurement (red) compared to the shot‑noise limit (black). Figure 1b reproduced from Ref. [49] with permission of the American Physical Society. c The anti‑ferromagnetic computing process of CIM (N = 2) and the success rate of finding two degenerate ground states varies with time Lu et al. AAPPS Bulletin (2023) 33:7 Page 3 of 8 in the Ising model. Therefore, squeezed light implemen - nonlinear crystals [60–66], as shown in Fig. 2a. OPO is an tation of CIM based on degenerate optical parametric open dissipative system with a second-order phase tran- oscillator (DOPO) is an effective way. We present a cal - sition at the oscillation threshold, with gain competition culation process of CIM based on a degenerate paramet- among the multiple modes in the laser [67]. Moreover, ric oscillator in anti-ferromagnetic 2-spin by numerical it has been experimentally proved that an Ising Ham- simulation in Fig.1c. The probabilities of both degener - iltonian can be simulated in a mutual injection-locked ate ground states(|↑↓� and|↓↑�) at the beginning of the laser network [68]. A DOPO, under coherent excitation † −iω t iω t d d calculation are both rising under the influence of noise ˆ ˆ , the interacting Hamilto- H = i ǫ(a e − a e ) d p †2 2 † until the collective symmetry-breaking stage. Then, the nian is H = i (aˆ aˆ − aˆ aˆ ) , and a ˆ (a ˆ ) is the pump int p p s s s p system randomly selects a ground state, and a quantum- (signal) mode. Considering the linear dissipation of the to-classical crossover occurs. Recently, the experimental pump (signal) mode as γ (γ ) , the Heisenberg-Langevin p s progress of the optical CIM has been listed as follows. equations of the system are [69] daˆ κ p √ 2.1 CIM based on DOPO =ǫ − γ aˆ − aˆ + γ F , p p p 1 dt 2 The Ising model is used to simulate the random process (2) daˆ κ of the phase transition of matter. It is a binary spin sys- † =− γ aˆ − aˆ aˆ + γ F ,. s s p s 2 dt 2 tem, and each bit is randomly flipped under the influence of the potential energy field [50]. Therefore, the artifi - cially analog spin in the Ising machine that simulates the Here, F is the time-dependent noise operators for 1(2) evolution of the Ising model should also have this prop- † ′ ′ ˆ ˆ the pump (signal) field, and F (t)F (t ) = δ δ(t − t ) . i ij erty. Coincidentally, such a binary value occurs when the 3 When γ is large, we can adiabatically eliminate the pump evolution of nonlinear dynamics satisfies x˙ = αx − x . operator a ˆ by daˆ /dt = 0 . Therefore, the Heisenberg- p p Nonlinear optics, which has been widely studied in opti- Langevin equation for the signal mode becomes cal frequency combs [51, 52], optomechanics [53, 54], squeezed light [48, 55, 56], and optical frequency conver- daˆ √ † † 2 † ˆ ˆ sion [57–59]. Wang Zhe et al. proposed that CIM can be = −γ aˆ + Raˆ − K aˆ aˆ + 2K aˆ F + γ F . s s 1 s 2 s s s s dt implemented using optical parametric oscillators (OPO) (3) [21]. Here, R = κǫ/γ is gain coefficient and K = κ /(2γ ) p p Optical parametric oscillation (OPO) describes the is saturation coefficient. Its equivalent master equation is nonlinear optical process that converts a high-energy [69] (2) photon into two low-energy photons through χ Fig. 2 Schematic diagram of the analog spin system based on DOPO. a DOPO process. A high‑ energy photon (indicated in blue) incident on the PPLN crystal is converted into two identical low‑ energy photons (indicated in red). PPLN: Periodically polarized lithium niobate crystal. b Quantum noise distribution of DOPO, b is below the threshold, and c is above the threshold. d The trend of the potential function does not oscillate below the threshold (red) and a bistable potential above the threshold (gray line). The pump energy rises, and the analog spin selects a state with lower potential energy Lu et al. AAPPS Bulletin (2023) 33:7 Page 4 of 8 principle of CIM [73]. The CIM based on DOPO has dρˆ γ † †2 2 =γ aˆ , ρaˆ + H .c. + p aˆ − aˆ , ρˆ s s s s s been proven to have excellent performance in dealing dt 2 (4) with combinatorial optimization problems, such as Max- 2 2 †2 + g aˆ , ρˆaˆ + H .c. . s s Cut, in numerical simulation [21]. Based on DOPO, these schemes successfully imple- where parametric gain p = K /γ , and two-photon ment a CIM, which implements the coupling between absorption coefficient is g = R/γ . Then, we extend the different pulses through a delayed optical path [73–75]. total field density operator ρ using the positive-P repre- The earliest scheme shown in Fig. 3a to implement opti- sentation P(α, β) and get the Fock-Planck equation [70]. cal CIM is based on a femtosecond laser-driven DOPO We ignore the third- and higher-order terms of the Fock- [73]. A fixed-length delay optical path can implement Planck equation and obtain the c-number SDEs of i − th different pulse couplings, and different coupling struc - DOPO by the Ito rules tures can be implemented by different length delay opti - cal paths, as shown in Fig. 3c. Later, the scheme was dα 2 2 2 2 ˆ =− α + pβ − g β α + p − g α f , i i i i,α i i extended to the 16-spin system to solve the more com- dt (5) plex combinatorial optimization problem, as shown in dβ 2 2 2 2 ˆ =− β + pα − g β α + p − g β f . i i i i,β Fig. 3d. T. Inagaki et al. used a 10,000-OPO CIM to simu- i i dt late a 1D anti-ferromagnetic Ising chain. They observed Here, f is real random noise number and the formation of spin domains in Fig. 3e, indicating that i,α(β ) ′ ′ ˆ ˆ f (t)f (t ) = δ δ δ(t − t ) . Under the positive-P the DOPO network can simulate the behavior of low- i,a ij ab j,b temperature Ising spins [75]. representation, α, β is generally a complex number, but In addition, the optical CIM provides a new solution to the parametric oscillations ensure that α, β can be the combinatorial optimization problem and a reference restricted to real numbers [21]. By ignoring the quantum for other ways to implement CIM. noise term, the amplitude evolution of the ith DOPO can be expressed as 2.2 CIM based on the measurement feed‑back dµ 3 Although the optical CIM based on the optical delay path = (p − 1)µ − µ + εJ µ . i ij j (6) dt performs well in some of these optimization problems, j=1,j�=i the fixed-length delay optical path can only map the com - Here,µ is the amplitude of i th DOPO pulse, andε binatorial optimization problem with a single structure is the coupling strength between pulses. The evolu - [74]. However, the combinatorial optimization problem dV (µ) tion of the amplitude of DOPO satisfies µ ˙ = − , is complex, and the CIM that can be connected arbitrar- dµ 2 4 andV (µ) = (1 − p)µ /2 + µ /4 is the potential func- ily becomes a new challenge. tion [71]. The evolution of DOPO is shown in Fig.2b and Therefore, a measurement feedback system (MFB) is c. Below the pump threshold, the pulses are amplified proposed by first measuring the pulse amplitude and then and unamplified in the x−axis andp−axis in a squeezed modulating the pulse through the optical modulator and vacuum state. When the parametric oscillation gain is injecting it into the target pulse at the appropriate time greater than the dissipation of the cavity, the pulse can [11, 12]. CIM with MFB adds a field programmable gate have a stable non-zero amplitude on the x-axis. How- array (FPGA) based on the DOPO fiber-ring cavity and ever, the probability of selecting phase 0 or phaseπ is implements the measurement-calculation-feedback pro- 50% each. This phenomenon is the collective symmetry cess through FPGA [26], as shown in Fig. 4a. McMahon breaking [72]. P. L. et al. first implemented a fully programmable opti - In addition, the relationship between V(x) and the cal CIM that can easily map arbitrarily connected Ising pump threshold is given in Fig.2d. The system is only models, even including the case of asymmetric coupling stable near the amplitude zero point when the pump J = J [11], as shown in Fig. 4b. CIM with MFB is imple- ij ji intensity is below the oscillation threshold. Moreover, mented by taking parts (10% ) of each pulse for measure- by increasing pump intensity to the oscillation thresh- ment and then feedback injection. The machine can find old, the potential energy of the system becomes a para- exact solutions, or good approximate solutions, to vari- digmatic bistable potential. The system chooses positive ous difficult combinatorial optimization problems under amplitude or negative amplitude with equal probability. 100-spins. In Fig. 4c, the scheme has been extended to By adjusting the pump gain, the existence of the coupling 2000-spins to solve more complex combinatorial optimi- term makes the depth of the potential difference. The zation problems by increasing the fiber loop length and system will select the ground state with lower energy, the firing rate of the pulses [12]. The experimental results the optimal spin configuration, which is the operating show that the machine can achieve the same calculation Lu et al. AAPPS Bulletin (2023) 33:7 Page 5 of 8 Fig. 3 Example of CIM based on DOPO. a Schematic diagram of 4‑ OPO composition coherent Ising machine from Ref. [73]. b The interval of the output OPO pulse and the coupling of the injection optical path is realized at c, from Ref. [73]. Figure 3a, b and c reproduced from Ref. [73] with permission of the Springer Nature. d Schematic diagram of 16‑ OPO composition coherent Ising machine reproduced from Ref. [74] with permission of the Springer Nature. e 1D Ising model simulation observed defect densityn and correlation lengthx as a function of normalized 1551 nm pump d 0 amplitude reproduced from Ref. [75] with permission of the Springer Nature Fig. 4 Example of CIM with MFB and applications. a The schematic diagram of CIM with MFB reproduced from Ref. [26] with permission of the American Association for the Advancement of Science. b 100‑spin CIM realizes full connection reproduced from Ref. [11] with permission of the American Association for the Advancement of Science. c 2000‑spin CIM machines for solving combinatorial optimization problems reproduced from Ref. [12] with permission of the American Association for the Advancement of Science. d CIM simulation of 2D Ising problem evolution versus low‑temperature Monte Carlo simulation reproduced from Ref. [76] with permission of the Springer Nature. e CIM to simulate spiking neuron dynamics reproduced from Ref. [77] with permission of the Springer Nature. f CIM simulates the Potts model schematic and solves the clustering and coloring problems reproduced from Ref. [78] with permission of the Springer Nature. Lu et al. AAPPS Bulletin (2023) 33:7 Page 6 of 8 result in 1% of the time of the simulated annealing (SA) Moreover, the CIM-inspired algorithm combined with implemented in the state-of-the-art central processing the FPGA-implemented machine also showed excellent unit (CPU), showing the accelerated calculation effect of computational performance in solving combinatorial optical CIM [12]. The latest research about 100,000-spins optimization problems, even better than optical CIM [87, CIM shows that by optimizing the parallel FPGA, the 88]. Implementing CIM with better computing perfor- computation time of CIM with MFB only depends on the mance is also a challenge for simulating quantum systems time which the light pulse runs in the fiber-ring [79]. using a physical system. CIM with MFB shows computational performance The introduction of nonlinear filtering to CIM reduces advantages over current digital computer [80–82]. R. amplitude heterogeneity [89–91]. Also, it connects CIM Hamerly et al. also compared the performance of CIM with quantum neural networks and quantum perceptrons with MFB and quantum annealer. They found that CIM due to the wide application of nonlinear functions in arti- with MFB is weaker than quantum annealer in the cal- ficial intelligence. The neural network and quantum per - culation of simple graphs (sparse) while having obvious ceptron based on the time multiplexing optical fiber-ring advantages over quantum annealer in more complex of CIM is worth looking forward to [92, 93]. For example, graphs (dense) [26]. Furthermore, to understand the evo- the spiking neural network has been simulated in CIM lution process of CIM searching for the ground state, F. [77]. Meanwhile, CIMs based on nonlinear optics are also Böhm et al. used CIM with MFB to simulate the evolu- expected to promote the research of nonlinear optics due tion process of 2D anti-ferromagnetic Ising networks. to their fully connected properties and system scalability. They found that CIMs behave like low-temperature spin In summary, we review the CIM based on squeezed systems, making them suitable for solving combinato- light in this brief review, summarizing the advantages, rial optimization tasks, as shown in Fig. 4d [76]. Spik- challenges, and potential applications of the existing ing neurons, specialized neurons that process signals in schemes. Combined with the recent rapid development human brain, have been shown to have better comput- of electro-optic modulators and photonic chips [94–96], ing performance than traditional artificial neurons and it is reasonable to expect CIMs with better computing are regarded as the next computational neural network performance. [83, 84]. T. Inagaki et al. studied the dynamic behavior of spiking neurons in neural clusters simulated by CIM. Authors’ contributions They found that it helps to improve the computational C.W. is the lead author in organizing this paper. B.L. and L.L. wrote the draft of performance of CIM, as shown in Fig. 4e [77]. CIM is this paper. All authors contributed to writing and polishing the manuscript. All the authors read and approved the final manuscript. also considered to be able to simulate the Potts model of multi-valued spins, as shown in Fig. 4f [78], and it is Funding shown, through experiments, that clustering and coloring This research was funded by the National Natural Science Foundation of China (Grants Nos. 62131002, 62071448) and the Fundamental Research Funds for problems can be achieved. the Central Universities (BNU). Availability of data and materials 3 Discussion and outlook Not applicable. The optically implemented CIM shows the advantages of physical simulation machines in simulating physical Declarations systems. This novel form of optical computing provides Competing interests a way to develop new computer structures, but there are The authors declare that they have no competing interests. also potential challenges that we need to address. Large-scale CIM based on DOPO networks needs Received: 3 October 2022 Accepted: 21 October 2022 to increase the spin scale by increasing the length of the fiber-ring cavity or increasing the pump repeti - tion rate. However, the nonlinear and thermal effects of long-distance optical fibers will bring instability to the References system [79]. How to achieve high stability under large- 1. A. Adamatzky, V. Erokhin, M. Grube, T. Schubert, A. Schumann et al., Phys‑ scale requirements is a challenge. In addition, the study arum chip project: growing computers from slime mould. Int. J. Unconv. on improving the performance of CIM hardware solver Comput. 8(4), 319–323 (2012) 2. X. Xu, M. Tan, B. Corcoran, J. Wu, A. Boes, T.G. Nguyen, S.T. Chu, B.E. Little, improvement has been widely existed [85, 86]. Moreo- D.G. Hicks, R. Morandotti et al., 11 tops photonic convolutional accelera‑ ver, what role quantum noise and thermal noise play in tor for optical neural networks. Nature 589(7840), 44–51 (2021) the calculation process of CIM and how to improve the 3. L.G. Wright, T. Onodera, M.M. Stein, T. Wang, D.T. Schachter, Z. Hu, P.L. McMahon, Deep physical neural networks trained with backpropagation. CIM hardware solver to improve performance will also Nature 601(7894), 549–555 (2022) be interesting questions. Lu et al. AAPPS Bulletin (2023) 33:7 Page 7 of 8 4. R. Barends, A. Shabani, L. Lamata, J. Kelly, A. Mezzacapo, U.L. Heras, 28. P. Kok, W.J. Munro, K. Nemoto, T.C. Ralph, J.P. Dowling, G.J. Milburn, Linear R. Babbush, A.G. Fowler, B. Campbell, Y. Chen et al., Digitized adi‑ optical quantum computing with photonic qubits. Rev. Mod. Phys. 79(1), abatic quantum computing with a superconducting circuit. Nature 135 (2007) 534(7606), 222–226 (2016) 29. L.S. Madsen, F. Laudenbach, M.F. Askarani, F. Rortais, T. Vincent, J.F. Bulmer, 5. L. DiCarlo, J.M. Chow, J.M. Gambetta, L.S. Bishop, B.R. Johnson, D. F.M. Miatto, L. Neuhaus, L.G. Helt, M.J. Collins et al., Quantum compu‑ Schuster, J. Majer, A. Blais, L. Frunzio, S. Girvin et al., Demonstration of tational advantage with a programmable photonic processor. Nature two‑ qubit algorithms with a superconducting quantum processor. 606(7912), 75–81 (2022) Nature 460(7252), 240–244 (2009) 30. A. Biberman, K. Bergman, Optical interconnection networks for high‑ 6. Y. Wu, W.S. Bao, S. Cao, F. Chen, M.C. Chen, X. Chen, T.H. Chung, H. Deng, performance computing systems. Rep. Prog. Phys. 75(4), 046402 (2012) Y. Du, D. Fan et al., Strong quantum computational advantage using a 31. M.A. Taubenblatt, Optical interconnects for high‑performance comput ‑ superconducting quantum processor. Phys. Rev. Lett. 127(18), 180501 ing. J. Light. Technol. 30(4), 448–457 (2011) (2021) 32. T. Szymanski, “hypermeshes’’: optical interconnection networks for paral‑ 7. J. Clarke, F.K. Wilhelm, Superconducting quantum bits. Nature lel computing. J Parallel Distrib. Comput. 26(1), 1–23 (1995) 453(7198), 1031–1042 (2008) 33. X. Xu, M. Tan, B. Corcoran, J. Wu, T.G. Nguyen, A. Boes, S.T. Chu, B.E. Little, R. 8. R. Barends, J. Kelly, A. Megrant, A. Veitia, D. Sank, E. Jeffrey, T.C. White, Morandotti, A. Mitchell et al., Photonic perceptron based on a kerr micro‑ J. Mutus, A.G. Fowler, B. Campbell et al., Superconducting quantum comb for high‑speed, scalable, optical neural networks. Laser Photonics circuits at the surface code threshold for fault tolerance. Nature Rev. 14(10), 2000070 (2020) 508(7497), 500–503 (2014) 34. S.T. Le, V. Aref, H. Buelow, Nonlinear signal multiplexing for communica‑ 9. A. Das, B.K. Chakrabarti, Colloquium: Quantum annealing and analog tion beyond the kerr nonlinearity limit. Nat. Photonics 11(9), 570–576 quantum computation. Rev. Mod. Phys. 80(3), 1061 (2008) (2017) 10. T. Kadowaki, H. Nishimori, Quantum annealing in the transverse ising 35. A.N.Z. Rashed, M.S.F. Tabbour, Best candidate integrated technology for model. Phys. Rev. E 58(5), 5355 (1998) low‑noise, high‑speed, and wide bandwidth‑based transimpedance 11. P.L. McMahon, A. Marandi, Y. Haribara, R. Hamerly, C. Langrock, S. amplifiers in optical computing systems and optical fiber applications. Tamate, T. Inagaki, H. Takesue, S. Utsunomiya, K. Aihara et al., A fully Int. J. Commun. Syst. 31(17), e3801 (2018) programmable 100‑spin coherent ising machine with all‑to ‑all connec‑ 36. R.C. Devlin, A. Ambrosio, N.A. Rubin, J.B. Mueller, F. Capasso, Arbitrary spin‑ tions. Science 354(6312), 614–617 (2016) to‑ orbital angular momentum conversion of light. Science 358(6365), 12. T. Inagaki, Y. Haribara, K. Igarashi, T. Sonobe, S. Tamate, T. Honjo, A. 896–901 (2017) Marandi, P.L. McMahon, T. Umeki, K. Enbutsu et al., A coherent ising 37. Z. Ji, W. Liu, S. Krylyuk, X. Fan, Z. Zhang, A. Pan, L. Feng, A. Davydov, R. machine for 2000‑node optimization problems. Science 354(6312), Agarwal, Photocurrent detection of the orbital angular momentum of 603–606 (2016) light. Science 368(6492), 763–767 (2020) 13. D. Pierangeli, G. Marcucci, C. Conti, Large‑scale photonic ising machine 38. I.A. Young, E. Mohammed, J.T. Liao, A.M. Kern, S. Palermo, B.A. Block, M.R. by spatial light modulation. Phys. Rev. Lett. 122(21), 213902 (2019) Reshotko, P.L. Chang, Optical i/o technology for tera‑scale computing. 14. B. Sutton, K.Y. Camsari, B. Behin‑Aein, S. Datta, Intrinsic optimization IEEE J. Solid State Circ. 45(1), 235–248 (2009) using stochastic nanomagnets. Sci. Rep. 7(1), 1–9 (2017) 39. J.H. Han, F. Boeuf, J. Fujikata, S. Takahashi, S. Takagi, M. Takenaka, Efficient 15. M. Saccone, F. Caravelli, K. Hofhuis, S. Parchenko, Y.A. Birkhölzer, S. low‑loss ingaasp/si hybrid mos optical modulator. Nat. Photonics 11(8), Dhuey, A. Kleibert, S. Van Dijken, C. Nisoli, A. Farhan, Direct observation 486–490 (2017) of a dynamical glass transition in a nanomagnetic artificial Hopfield 40. H. Yuen, J. Shapiro, Optical communication with two‑photon coherent network. Nat. Phys. 18(5), 517–521 (2022) states‑part I: quantum‑state propagation and quantum‑noise. IEEE Trans. 16. B. Kiraly, E.J. Knol, W.M. van Weerdenburg, H.J. Kappen, A.A. Khajetoo‑ Inf. Theory 24(6), 657–668 (1978). https:// doi. org/ 10. 1109/ TIT. 1978. 10559 rians, An atomic boltzmann machine capable of self‑adaption. Nat. 58 Nanotechnol. 16(4), 414–420 (2021) 41. Y.H. Chen, W. Qin, X. Wang, A. Miranowicz, F. Nori, Shortcuts to adiabatic‑ 17. C. Yoshimura, M. Yamaoka, M. Hayashi, T. Okuyama, H. Aoki, K.I. ity for the quantum rabi model: efficient generation of giant entangled Kawarabayashi, H. Mizuno, Uncertain behaviours of integrated circuits cat states via parametric amplification. Phys. Rev. Lett. 126, 023602 improve computational performance. Sci. Rep. 5(1), 1–12 (2015) (2021).https:// doi. org/ 10. 1103/ PhysR evLett. 126. 023602 18. A. Soler‑Dominguez, A.A. Juan, R. Kizys, A survey on financial applica‑ 42. W. Qin, A. Miranowicz, P.B. Li, X.Y. Lü, J.Q. You, F. Nori, Exponentially tions of metaheuristics. ACM Comput. Surv. (CSUR) 50(1), 1–23 (2017) enhanced light‑matter interaction, cooperativities, and steady‑state 19. D.B. Kitchen, H. Decornez, J.R. Furr, J. Bajorath, Docking and scoring in entanglement using parametric amplification. Phys. Rev. Lett. 120, virtual screening for drug discovery: methods and applications. Nat. 093601 (2018).https:// doi. org/ 10. 1103/ PhysR evLett. 120. 093601 Rev. Drug Discov. 3(11), 935–949 (2004) 43. W. Qin, A. Miranowicz, F. Nori, Beating the 3 db limit for intracavity 20. F. Barahona, M. Grötschel, M. Jünger, G. Reinelt, An application of com‑ squeezing and its application to nondemolition qubit readout. Phys. Rev. binatorial optimization to statistical physics and circuit layout design. Lett. 129, 123602 (2022).https:// doi. org/ 10. 1103/ PhysR evLett. 129. 123602 Oper. Res. 36(3), 493–513 (1988) 44. J. Abadie, B.P. Abbott, T. Abbott, M. Abernathy, M. Benacquista, T. 21. Z. Wang, A. Marandi, K. Wen, R.L. Byer, Y. Yamamoto, Coherent ising Creighton, H. Daveloza, M.E. Diaz, R. Grosso, S. Mohanty et al., A gravita‑ machine based on degenerate optical parametric oscillators. Phys. Rev. tional wave observatory operating beyond the quantum shot‑noise limit: A 88(6), 063853 (2013) squeezed light in application. Nat. Phys. 7, 962–965 (2011) 22. R.P. Feynman, in Feynman and computation. Simulating physics with 45. B.J. Lawrie, P.D. Lett, A.M. Marino, R.C. Pooser, Quantum sensing with computers (CRC Press, Florida, USA, 2018), pp. 133–153 squeezed light. Acs Photonics 6(6), 1307–1318 (2019) 23. S. Lloyd, Universal quantum simulators. Science 273(5278), 1073–1078 46. W. Qin, A. Miranowicz, F. Nori, Beating the 3 db limit for intracavity (1996) squeezing and its application to nondemolition qubit readout. Phys. Rev. 24. S. Arora, B. Barak, Computational complexity: a modern approach (Cam‑ Lett. 129, 123602 (2022). https:// doi. org/ 10. 1103/ PhysR evLett. 129. 123602 bridge University Press, Cambridge, U.K., 2009) 47. W. Qin, Y.H. Chen, X. Wang, A. Miranowicz, F. Nori, Strong spin squeezing 25. T. Byrnes, K. Yan, Y. Yamamoto, Accelerated optimization problem induced by weak squeezing of light inside a cavity. Nanophotonics 9(16), search using bose‑ einstein condensation. New J Phys 13(11), 113025 4853–4868 (2020) (2011) 48. G. Breitenbach, S. Schiller, J. Mlynek, Measurement of the quantum states 26. R. Hamerly, T. Inagaki, P.L. McMahon, D. Venturelli, A. Marandi, T. of squeezed light. Nature 387(6632), 471–475 (1997) Onodera, E. Ng, C. Langrock, K. Inaba, T. Honjo et al., Experimental 49. N.V. Corzo, Q. Glorieux, A.M. Marino, J.B. Clark, R.T. Glasser, P.D. Lett, investigation of performance differences between coherent ising Rotation of the noise ellipse for squeezed vacuum light generated via machines and a quantum annealer. Sci. Adv. 5(5), eaau0823 (2019) four‑ wave mixing. Phys. Rev. A 88(4), 043836 (2013) 27. J.L. O’brien, Optical quantum computing. Science 318(5856), 50. S.G. Brush, History of the lenz‑ising model. Rev. Mod. Phys. 39, 883–893 1567–1570 (2007) (1967). https:// doi. org/ 10. 1103/ RevMo dPhys. 39. 883 Lu et al. AAPPS Bulletin (2023) 33:7 Page 8 of 8 51. T.J. Kippenberg, R. Holzwarth, S.A. Diddams, Microresonator‑based optical 78. K. Inaba, T. Inagaki, K. Igarashi, S. Utsunomiya, T. Honjo, T. Ikuta, K. Enbutsu, frequency combs. Science 332(6029), 555–559 (2011) T. Umeki, R. Kasahara, K. Inoue et al., Potts model solver based on hybrid 52. S.T. Cundiff, J. Ye, Colloquium: Femtosecond optical frequency combs. physical and digital architecture. Commun. Phys. 5(1), 1–8 (2022) Rev. Mod. Phys. 75(1), 325 (2003) 79. T. Honjo, T. Sonobe, K. Inaba, T. Inagaki, T. Ikuta, Y. Yamada, T. Kazama, K. 53. M. Aspelmeyer, T.J. Kippenberg, F. Marquardt, Cavity optomechanics. Rev. Enbutsu, T. Umeki, R. Kasahara et al., 100,000‑spin coherent ising machine. Mod. Phys. 86(4), 1391 (2014) Sci. Adv. 7(40), eabh0952 (2021) 54. M. Metcalfe, Applications of cavity optomechanics. Appl. Phys. Rev. 1(3), 80. Y. Haribara, S. Utsunomiya, Y. Yamamoto, in Principles and Methods of 031105 (2014) Quantum Information Technologies. A coherent ising machine for max‑ cut 55. H. Vahlbruch, M. Mehmet, S. Chelkowski, B. Hage, A. Franzen, N. Lastzka, problems: performance evaluation against semidefinite programming S. Gossler, K. Danzmann, R. Schnabel, Observation of squeezed light with and simulated annealing (Springer, Berlin/Heidelberg, Germany, 2016), 10‑ db quantum‑noise reduction. Phys. Rev. Lett. 100(3), 033602 (2008) pp. 251–262 56. R. Loudon, P.L. Knight, Squeezed light. J. Mod. Opt. 34(6–7), 709–759 81. Y. Haribara, S. Utsunomiya, Y. Yamamoto, Computational principle and (1987) performance evaluation of coherent ising machine based on degenerate 57. M.M. Fejer, Nonlinear optical frequency conversion. Phys. Today 47(5), optical parametric oscillator network. Entropy 18(4), 151 (2016) 25–33 (1994) 82. Y. Haribara, H. Ishikawa, S. Utsunomiya, K. Aihara, Y. Yamamoto, Perfor‑ 58. R. Lifshitz, A. Arie, A. Bahabad, Photonic quasicrystals for nonlinear optical mance evaluation of coherent ising machines against classical neural frequency conversion. Phys. Rev. Lett. 95(13), 133901 (2005) networks. Quantum Sci. Technol. 2(4), 044002 (2017) 59. A.W. Bruch, X. Liu, J.B. Surya, C.L. Zou, H.X. Tang, On‑ chipχ (2) microring 83. E.M. Izhikevich, Simple model of spiking neurons. IEEE Trans. Neural Netw. optical parametric oscillator. Optica 6(10), 1361–1366 (2019) 14(6), 1569–1572 (2003) 60. J.A. Giordmaine, R.C. Miller, Tunable coherent parametric oscillation in 84. R. Brette, M. Rudolph, T. Carnevale, M. Hines, D. Beeman, J.M. Bower, M. linb o 3 at optical frequencies. Phys. Rev. Lett. 14(24), 973 (1965) Diesmann, A. Morrison, P.H. Goodman, F.C. Harris et al., Simulation of 61. L.A. Wu, M. Xiao, H. Kimble, Squeezed states of light from an optical networks of spiking neurons: a review of tools and strategies. J Comput. parametric oscillator. JOSA B 4(10), 1465–1475 (1987) Neurosci. 23(3), 349–398 (2007) 62. P. Powers, T.J. Kulp, S. Bisson, Continuous tuning of a continuous‑ wave 85. L. Luo, H. Liu, N. Huang, Z. Wang, Speed‑up coherent ising machine with periodically poled lithium niobate optical parametric oscillator by use of a squeezed feedback system. Opt. Express 28(2), 1914–1926 (2020) a fan‑ out grating design. Opt. Lett. 23(3), 159–161 (1998) 86. F. Böhm, G. Verschaffelt, G. Van der Sande, A poor man’s coherent ising 63. B. Lu, C.R. Fan, J.Y. Song, C. Wang, Optical parametric oscillation with ultra‑ machine based on opto‑ electronic feedback systems for solving optimi‑ low power threshold in a dimer of active‑passive cavities. Crystals 11(5), zation problems. Nat. Commun. 10(1), 1–9 (2019) 566 (2021) 87. H. Goto, K. Endo, M. Suzuki, Y. Sakai, T. Kanao, Y. Hamakawa, R. Hidaka, M. 64. W. Qin, A. Miranowicz, G. Long, J. You, F. Nori, Proposal to test quantum Yamasaki, K. Tatsumura, High‑performance combinatorial optimization wave‑particle superposition on massive mechanical resonators. npj based on classical mechanics. Sci. Adv. 7(6), eabe7953 (2021) Quantum Inf. 5(1), 1–8 (2019) 88. T. Leleu, F. Khoyratee, T. Levi, R. Hamerly, T. Kohno, K. Aihara, Scaling 65. W. Qin, V. Macrì, A. Miranowicz, S. Savasta, F. Nori, Emission of photon advantage of chaotic amplitude control for high‑performance combina‑ pairs by mechanical stimulation of the squeezed vacuum. Phys. Rev. A torial optimization. Commun. Phys. 4(1), 1–10 (2021) 100, 062501 (2019).https:// doi. org/ 10. 1103/ PhysR evA. 100. 062501 89. F. Böhm, T.V. Vaerenbergh, G. Verschaffelt, G. Van der Sande, Order ‑ of‑ 66. W. Qin, A. Miranowicz, H. Jing, F. Nori, Generating long‑lived macroscopi‑ magnitude differences in computational performance of analog ising cally distinct superposition states in atomic ensembles. Phys. Rev. Lett. machines induced by the choice of nonlinearity. Commun. Phys. 4(1), 127, 093602 (2021).https:// doi. org/ 10. 1103/ PhysR evLett. 127. 093602 1–11 (2021) 67. A. Roy, S. Jahani, C. Langrock, M. Fejer, A. Marandi, Spectral phase transi‑ 90. S. Reifenstein, S. Kako, F. Khoyratee, T. Leleu, Y. Yamamoto, Coherent ising tions in optical parametric oscillators. Nat. Commun. 12(1), 1–9 (2021) machines with optical error correction circuits. Adv. Quantum Technol. 68. S. Utsunomiya, K. Takata, Y. Yamamoto, Mapping of ising models onto 4(11), 2100077 (2021) injection‑locked laser systems. Opt. Express 19(19), 18091–18108 91. B. Lu, Y.P. Gao, K. Wen, C. Wang, Combinatorial optimization solving by (2011).https:// doi. org/ 10. 1364/ OE. 19. 018091 coherent ising machines based on spiking neural networks. 2022. arXiv 69. Y. Inui, M.D.S.H. Gunathilaka, S. Kako, T. Aonishi, Y. Yamamoto, Control of preprintarXiv: 2208. 07502 amplitude homogeneity in coherent ising machines with artificial zee ‑ 92. C. Leefmans, A. Dutt, J. Williams, L. Yuan, M. Parto, F. Nori, S. Fan, A. man terms. Commun. Phys. 5(1), 1–7 (2022) Marandi, Topological dissipation in a time‑multiplexed photonic resona‑ 70. P. Drummond, C. Gardiner, Generalised p‑representations in quantum tor network. Nat. Phys. 18(4), 442–449 (2022) optics. J. Phys. A Math. Gen. 13(7), 2353 (1980) 93. B. Bartlett, A. Dutt, S. Fan, Deterministic photonic quantum computation 71. T. Leleu, Y. Yamamoto, S. Utsunomiya, K. Aihara, Combinatorial optimiza‑ in a synthetic time dimension. Optica 8(12), 1515–1523 (2021) tion using dynamical phase transitions in driven‑ dissipative systems. 94. Q. Xu, B. Schmidt, S. Pradhan, M. Lipson, Micrometre‑scale silicon electro ‑ Phys. Rev. E 95(2), 022118 (2017) optic modulator. Nature 435(7040), 325–327 (2005) 72. Y. Yamamoto, K. Aihara, T. Leleu, K.I. Kawarabayashi, S. Kako, M. Fejer, K. 95. C.T. Phare, Y.H. Daniel Lee, J. Cardenas, M. Lipson, Graphene electro‑ optic Inoue, H. Takesue, Coherent ising machines—optical neural networks modulator with 30 ghz bandwidth. Nat. Photonics 9(8), 511–514 (2015) operating at the quantum limit. npj Quantum Inf. 3(1), 1–15 (2017) 96. P. Dong, S. Liao, D. Feng, H. Liang, D. Zheng, R. Shafiiha, C.C. Kung, W. Qian, 73. A. Marandi, Z. Wang, K. Takata, R.L. Byer, Y. Yamamoto, Network of time‑ G. Li, X. Zheng et al., Low v pp, ultralow‑ energy, compact, high‑speed multiplexed optical parametric oscillators as a coherent ising machine. silicon electro‑ optic modulator. Opt. Express 17(25), 22484–22490 (2009) Nat. Photonics 8(12), 937–942 (2014) 74. K. Takata, A. Marandi, R. Hamerly, Y. Haribara, D. Maruo, S. Tamate, H. Publisher’s Note Sakaguchi, S. Utsunomiya, Y. Yamamoto, A 16‑bit coherent ising machine Springer Nature remains neutral with regard to jurisdictional claims in pub‑ for one‑ dimensional ring and cubic graph problems. Sci. Rep. 6(1), 1–7 lished maps and institutional affiliations. (2016) 75. T. Inagaki, K. Inaba, R. Hamerly, K. Inoue, Y. Yamamoto, H. Takesue, Large‑ scale ising spin network based on degenerate optical parametric oscilla‑ tors. Nat. Photonics 10(6), 415–419 (2016) 76. F. Böhm, T. Inagaki, K. Inaba, T. Honjo, K. Enbutsu, T. Umeki, R. Kasahara, H. Takesue, Understanding dynamics of coherent ising machines through simulation of large‑scale 2d ising models. Nat. Commun. 9(1), 1–9 (2018) 77. T. Inagaki, K. Inaba, T. Leleu, T. Honjo, T. Ikuta, K. Enbutsu, T. Umeki, R. Kasahara, K. Aihara, H. Takesue, Collective and synchronous dynamics of photonic spiking neurons. Nat. Commun. 12(1), 1–8 (2021)
Journal
AAPPS Bulletin – Springer Journals
Published: Mar 1, 2023