Three strongly coupled Kerr parametric oscillators forming a Boltzmann machine
Gabriel Margiani, Orjan Ameye, Oded Zilberberg, Alexander Eichler · 2025
Read Three strongly coupled Kerr parametric oscillators forming a Boltzmann machine by Gabriel Margiani, Orjan Ameye, Oded Zilberberg, Alexander Eichler (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-04254.txt.
Introduction
Gabriel Margiani, Orjan Ameye, Oded Zilberberg, and Alexander Eichler show in 2025 how three coupled oscillators can act as a Boltzmann machine. Coupled Kerr parametric oscillators are a promising tool for analog computation. Their state space, though, is complex and can hold states that do not map to spin problems.
The team uses three strongly coupled oscillators to simulate an Ising Hamiltonian, a model of interacting spins. Boltzmann sampling measurements estimate its ground state. The system is fully classical, but the work speaks directly to quantum versions that run on coherent states. Conditions for successful analog optimization are clarified.
Catalog record
| CREATOR | Gabriel Margiani, Orjan Ameye, Oded Zilberberg, Alexander Eichler |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 22 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-04254
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Three strongly coupled Kerr parametric oscillators forming a Boltzmann machine Preprint: APS/123-QED Gabriel Margiani, Orjan Ameye, Oded Zilberberg, Alexander Eichler Abstract Coupled Kerr parametric oscillators (KPOs) are a promising resource for classical and quantum analog computation, for example to find the ground state of Ising Hamiltonians. Yet, the state space of strongly coupled KPO networks is very involved. As such, their phase diagram sometimes features either too few or too many states, including some that cannot be mapped to Ising spin configurations. This complexity makes it challenging to find and meet the conditions under which an analog optimization algorithm can be successful. Here, we demonstrate how to use three strongly coupled KPOs as a simulator for an Ising Hamiltonian, and estimate its ground state using a Boltzmann sampling measurement. While fully classical, our work is directly relevant for quantum systems operating on coherent states. The use of nonlinear oscillators for computation is a long-standing concept that has recently gained renewed interest, driven by the growing variety of available resonator platforms Csaba et al. 2016 . Of special interest is the Kerr parametric oscillators (KPO), a bistable driven system that can be implemented in optical, electrical, and mechanical platforms Giordmaine and Miller 1965 ; Ryvkine and Dykman 2006 ; Mahboob and Yamaguchi 2008 ; Wilson et al. 2010 ; Eichler et al. 2011 ; Gieseler et al. 2012 ; Lin et al. 2014 ; Marandi et al. 2012 ; Puri et al. 2017 ; Nosan et al. 2019 ; Frimmer et al. 2019 ; Grimm et al. 2019 ; Wang et al. 2019 ; Puri et al. 2019 ; Miller et al. 2019 ; Yamaji et al. 2022 ; Yamaji et al. 2023 ; Frattini et al. 2024 . In a KPO, a modulation of the resonator potential energy, which we refer to as parametric pumping, gives rise to two possible ‘phase states’ that have identical amplitudes and opposite phases Dykman et al. 1993 ; Lifshitz and Cross 2008 ; Eichler and Zilberberg 2023 . Early on, these states were proposed and used as a physical basis for digital computing Goto 1959 ; Neumann 1959 ; Sterzer 1959 . The KPO is currently a major research focus due to the analogy between its two phase states and the ‘up’ and ‘down’ polarization states of an Ising spin. In particular, it was proposed that networks of KPOs [see Fig. 1 (a)] can be used to find the ground state of Ising Hamiltonians, that is, the energetically preferred configuration of a spin network Ising 1925 . Such resonator-based Ising solvers Mahboob et al. 2016 ; Goto 2016 ; Dykman et al. 2018 ; Puri et al. 2019 ; Bello et al. 2019 ; Okawachi et al. 2020 ; Yamamoto et al. 2020 ; Han et al. 2024 are of high interest because the corresponding calculations are hard to tackle with conventional computers Mohseni et al. 2022 . At the same time, they map to many key optimization problems, such as the traveling salesman problem Lucas 2014 , the MAX-CUT problem Inagaki et al. 2016 ; Goto et al. 2019 , and the number partitioning problem Nigg et al. 2017 . Interestingly, in the quantum regime, a KPO network can function as a quantum annealer, potentially outperforming its classical counterpart in finding optimal solutions Kadowaki and Nishimori 1998 ; Goto 2016 . To maintain quantum coherence during the adiabatic annealing procedure, the coupling between the N nodes must be stronger than the decay rate. Strong coupling, in turn, leads to a complicated phase diagram that does not necessarily map to an Ising Hamiltonian Heugel et al. 2019 ; Heugel et al. 2022 ; Álvarez et al. 2024 . So far, there exists no experimental confirmation that strongly coupled KPO resonators can function as Ising simulators. In this work, we demonstrate Ising simulations on a network of three strongly coupled KPOs. To do so, we explore the phase diagram of our network, and identify where in parameter space it possesses the correct number of solutions to manifest an Ising Hamiltonian. In that region, we perform Boltzmann sampling Goto et al. 2018 ; Margiani et al. 2023 and extract the total dwell times the system spends in each state. This procedure allows us to find the correct ground state of the corresponding Ising Hamiltonian, both without and with an applied external bias field. Finally, we highlight the crucial role of the nonlinearity and its relation to the rotating-frame Hamiltonian. Our work thus establishes that Ising simulation is viable in the strong coupling limit and provides guidelines for optimized experimental setups. Figure 1: Ising network built from three coupled KPOs. (a) The parametric phase states are represented as double-well quasi-potentials whose minima are our artificial Ising spins. (b) Single KPO realized as an RLC circuit. We use a varactor diode as nonlinear capacitance C and a coil as inductance L . The capacitor C_{\mathrm{b}} and resistor R_{\mathrm{b}} decouple coil and bias source. We use the voltage U=U_{\mathrm{F}}\cos(\omega t)+U_{\mathrm{p}}\cos(2\omega t) to drive the resonator, while we measure the voltage signal x . Our experimental setup consists of three inductively coupled nonlinear RLC-resonators, cf. Fig. 1 (b). Each resonator uses a varactor diode as a nonlinear element, which allows tuning the (angular) resonance frequency \omega_{0} via a reverse DC voltage U_{\mathrm{b}} . External forcing and parametric pumping is implemented via a wire loop close to the main inductor of the circuit. Applying an AC voltage tone U(\omega) to the drive loop induces a current in the resonator, which acts as a near-resonant forcing term with amplitude U_{\mathrm{F}} when \omega\approx\omega_{0} . At the same time, the second-order nonlinearity \beta_{2} introduced by the varactor diode (analogous to an optical \chi^{(2)} nonlinearity) enables three-wave mixing and parametric pumping with a modulation depth \lambda Nosan et al. 2019 ; Eichler and Zilberberg 2023 . When a voltage with amplitude U_{\mathrm{p}} is applied close to 2\omega\approx 2\omega_{0} , parametric oscillation can arise if \lambda exceeds a threshold \lambda_{\mathrm{th}} . These oscillations induce a voltage x in a second readout loop. The voltage is measured using an Intermodulation Products Multifrequency lock-in amplifier. We model our system by the following set of coupled differential equations, each describing a nonlinear parametric oscillator: \ddot{x}_{i}+\omega_{0}^{2}\left[1-\lambda\cos\left(2\omega t\right)\right]x_{i}+\beta x_{i}^{3}+\Gamma\dot{x}_{i}-\sum_{j\neq i}J_{ij}x_{j}=F_{i}\, (1) where x_{i} is the voltage measured from each of the resonators i\in\{1,2,3\} with an effective Duffing nonlinearity constant \beta=-10\beta_{2}^{2}/(9\omega_{0}^{2}) Lifshitz and Cross 2008 ; Eichler and Zilberberg 2023 , a damping rate \Gamma , and an external forcing term F_{i}=F\cos(\omega t) with F\propto U_{\mathrm{F}} Nosan et al. 2019 ; Heugel et al. 2022 . The parametric pump modulates the resonator’s potential with a modulation depth \lambda=\frac{2U_{\mathrm{p}}}{QU_{\mathrm{th}}} , where U_{\mathrm{p}} is the amplitude of the parametric drive signal, U_{\mathrm{th}} the parametric threshold voltage, and Q the quality factor of the resonator. Finally, each resonator couples to the two others via symmetric coupling constants J_{ij}=J_{ji} . We individually characterize each of the resonators while keeping the others far detuned (by setting their bias to U_{\mathrm{b}i}=0 ). This allows us to extract their bare parameters independently of coupling effects, and to calibrate the values of U_{\mathrm{b}i} required to tune all resonators to the same frequency \omega_{0} (and roughly the same \Gamma_{i}=\Gamma ). Using