arXiv:2610.29901
Superconducting parametric amplifiers based on arrays of unequal Josephson junction elements
Boris Chesca, Tom Dixon, Narendra Acharya, Renuka D. Pothuraju, Jacob W. Dunstan, Connor D. Shelly, Oscar W. Kennedy
Abstract
Impedance matching is a crucial feature of an amplifier used to increase its instantaneous bandwidth and is typically achieved using static LC resonators. Here we show an approach to impedance matching in parametric amplifiers using arrays of unequal Josephson-junction-based elements (rf/dc SQUIDs, SNAILs, etc). The elements act either as small-footprint flux-tunable inductors forming an impedance-matching circuit, or 'active' amplifier elements which result in amplifier gain. We develop an analytical theory describing these systems, predicting a quasi-non-periodic amplifier frequency and gain modulation with applied flux and the broadening of gain bandwidths. We demonstrate this in devices using arrays of flux-focused SNAILs where flux biasing varies across the array. We find instantaneous bandwidths of 200 MHz with $>$17 dB of gain centred at $\sim$5.4 GHz in a device without a static impedance matcher. This broadening of amplifier bandwidth is stable with $>$100 MHz bandwidth at 17 dB gain across $>$1.2 GHz tunable bandwidth. Our analytical model captures flux dependence and qualitatively reproduces gain-bandwidths giving confidence in our understanding of these systems and allowing operational figures of merit to be optimised in future devices.
AI Summary
The paper asks whether deliberately unequal Josephson-junction elements can provide impedance matching within a parametric amplifier, without a separate static matching circuit. The proposed mechanism assigns different roles to elements in a series array: some provide three-wave-mixing gain, while others contribute flux-tunable inductance that, together with circuit capacitance, modifies the impedance environment. An analytical model treats unequal elements through a multidimensional Josephson potential, eliminates internal phase coordinates, and calculates resonance frequencies and nonlinear coefficients. Gain is modeled using input–output theory with an effective pump amplitude accounting for unequal phase responses and frequency-dependent self-energies representing impedance matching. Experiments use arrays of three and five SNAILs embedded in transmission-line resonators, with integrated flux focusers producing different effective loop areas under a common magnetic bias. Resonance spectroscopy establishes the non-$\Phi_0$-periodic flux response; fitted circuit parameters also qualitatively reproduce the flux dependence of gain. The three-SNAIL device reaches approximately $200\,\mathrm{MHz}$ instantaneous bandwidth above $17\,\mathrm{dB}$ gain near $5.4\,\mathrm{GHz}$, compared with below $33\,\mathrm{MHz}$ for measured single-SNAIL devices. Bandwidth above $100\,\mathrm{MHz}$ at $17\,\mathrm{dB}$ is reported across $4.9$–$6.4\,\mathrm{GHz}$; a five-SNAIL device reaches approximately $100\,\mathrm{MHz}$. Noise-visibility measurements provide a consistency check rather than an absolute added-noise calibration. Agreement with calculated gain profiles is qualitative and does not reproduce their detailed structure. The maximum fractional bandwidth is $4\%$, below the approximately $10\%$ cited for purpose-built impedance matchers; further optimization and individually biased elements are proposed.
Central claim may not hold: The bandwidth theory assumes an LC matching circuit without deriving its parameters or topology from the unequal-array device model, leaving the proposed self-impedance-matching mechanism only conditionally supported. The reported experimental bandwidth comparison uses a single-SNAIL baseline rather than a comparable identical-element array, so it does not isolate element inequality as the cause of broadening. The effective-pump formula adds a phase-participation factor to a Hamiltonian coupling that already incorporates unequal phase responses, without deriving the amplitude-normalization change needed to justify that factor.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 15 pages, 8 figures, main text and appendices