⚛️ quantum physics

JCO: Optimization Framework for Nonlinear Superconducting Circuits Using a Lumped-Element Approach and Harmonic Balance

This paper introduces JosephsonCircuitsOptimizer.jl (JCO), a Julia-based framework that combines lumped-element modeling, harmonic balance, and Bayesian optimization to efficiently design and optimize nonlinear superconducting circuits, demonstrated through the systematic optimization of a SNAIL-based Josephson Traveling-Wave Parametric Amplifier.

Original authors: Emanuele Palumbo, Alessandro Alocco, Andrea Celotto, Luca Fasolo, Bernardo Galvano, Patrizia Livreri, Emanuele Enrico

Published 2026-03-31

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Emanuele Palumbo, Alessandro Alocco, Andrea Celotto, Luca Fasolo, Bernardo Galvano, Patrizia Livreri, Emanuele Enrico

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to build a super-precise radio amplifier out of tiny, frozen metal loops. These loops are made of superconducting materials, which means they conduct electricity with zero resistance, but they also have a quirky personality: they are nonlinear.

In the world of electronics, "nonlinear" is like a car that doesn't just speed up when you press the gas; it might suddenly jump to 100 mph, then slow down, then spin in circles, depending on exactly how hard you press. This makes them incredibly powerful for quantum computing, but also a nightmare to design. If you get the recipe wrong by a tiny fraction, the whole thing stops working.

This paper introduces a new tool called JCO (Josephson Circuits Optimizer) that acts like a super-smart, automated chef to help engineers cook up the perfect recipe for these circuits.

Here is how JCO works, broken down into simple steps:

1. The Problem: Too Many Ingredients, Too Many Choices

Imagine you are baking a cake, but you have 7 different ingredients (like flour, sugar, eggs, etc.), and for each one, you can choose from 10 or 20 different amounts. If you tried to bake every single combination just to see which one tastes best, you'd be baking cakes for the rest of your life.

In the world of superconducting circuits, the "ingredients" are things like the size of the metal loops, the thickness of the insulation, and the magnetic fields applied. The "cake" is a device called a JTWPA (Josephson Traveling-Wave Parametric Amplifier), which is used to amplify the faint whispers of quantum computers.

2. The Solution: A Three-Step Cooking Class

Instead of baking every cake, JCO uses a clever three-step strategy to find the perfect recipe quickly.

Step 1: The "Taste Test" (Linear Simulations)

First, the tool takes a quick, rough look at a grid of different recipes. It doesn't bake the full cake yet; it just checks the "flavor profile" (mathematically, this is called S-parameters).

The Analogy: Imagine a food critic tasting a tiny crumb of every possible cake combination. They aren't eating the whole thing; they just want to know: "Is this too salty? Is the texture right?"
The Goal: To filter out the terrible recipes and find the "good neighborhoods" where the cake might turn out great.

Step 2: The "Smart Detective" (Bayesian Optimization)

Once the critic finds the good neighborhoods, JCO switches to a Gaussian Process, which is like a super-smart detective.

The Analogy: Instead of tasting random crumbs, the detective looks at the crumbs they already tasted and says, "Okay, the ones with slightly more sugar and less flour were good. Let's try a combination right between those two."
The Magic: It learns as it goes. It stops wasting time on bad recipes and zooms in on the specific combination of ingredients that minimizes errors. It finds the optimal design ( $p^*$ ) much faster than a human could.

Step 3: The "Grand Tasting" (Nonlinear Simulations)

Now that they have the perfect recipe, they finally bake the full cake and serve it to the judges.

The Analogy: This is the full, real-world test. They turn on the "pump" (the energy source) and see how the amplifier actually performs. Does it amplify the signal? Is it quiet?
The Goal: They tweak the "serving temperature" (the operating point, $q^*$ ) to get the maximum gain (volume) without the cake collapsing.

3. The Result: A Perfect Amplifier

The team used JCO to design a specific type of amplifier called a SNAIL-based JTWPA.

The Challenge: This device needs to mix three waves of energy together perfectly (3-wave mixing) to amplify quantum signals. If the waves don't line up, the signal gets lost.
The Success: JCO found the perfect mix of loop sizes and magnetic fields. It tuned the device so that it amplifies weak quantum signals by 20 times (20 dB) with almost no extra noise.

Why This Matters

Before this tool, designing these circuits was like trying to find a needle in a haystack by looking at one straw at a time. It took forever and often failed.

JCO is like a metal detector that scans the whole haystack in seconds. It allows scientists to:

Save Time: What used to take weeks of trial and error now takes hours of computer time.
Handle Complexity: It can juggle dozens of variables at once without getting confused.
Future-Proof: Because it's built on a flexible system (Julia language), engineers can easily swap in new types of circuits or new optimization tricks later.

In short, this paper presents a GPS for quantum engineers. Instead of wandering aimlessly through a maze of complex physics, they can now drive straight to the exit, finding the perfect settings for their superconducting circuits every time.

1. Problem Statement

Superconducting quantum circuits, particularly those based on Josephson Junctions (JJs), are critical for quantum computing and signal processing due to their low dissipation and strong nonlinearity. However, designing these circuits presents significant computational challenges:

High Dimensionality: Circuits often involve large arrays of unit cells with many coupled design parameters.
Strong Nonlinearity: Traditional time-domain simulations (e.g., SPICE-like) are computationally expensive, requiring long integration times and fine temporal resolution to capture steady-state behavior.
Optimization Difficulty: Existing optimization methods often focus on parameter matching via electromagnetic simulations or Hamiltonian models but struggle to efficiently explore high-dimensional, strongly nonlinear design spaces (e.g., large-scale circuit arrays) to find optimal working points.

2. Methodology: The JCO Framework

The authors introduce JosephsonCircuitsOptimizer.jl (JCO), a simulation and optimization framework built on the Julia-based JosephsonCircuits.jl library. JCO utilizes a lumped-element approach combined with Harmonic Balance (HB), a frequency-domain technique that computes steady-state responses directly without simulating time evolution.

The framework operates via a three-stage modular pipeline:

Stage 1: Linear Simulations & Metric Definition

Process: Performs linear simulations on a uniformly sampled grid of the device parameter space ( $P$ ).
Output: Calculates S-parameters ( $S(p)$ ) over a broadband frequency range (near-DC to tens of GHz).
Metric ( $M$ ): A device-specific metric is defined to quantify circuit behavior (e.g., impedance matching, phase matching, suppression of unwanted harmonics). This metric guides the optimization.
Visualization: Results are visualized via correlation matrices and histograms to reveal parameter inter-dependencies.

Stage 2: Bayesian Optimization

Strategy: Uses Bayesian Optimization with a Gaussian Process (GP) surrogate model to minimize the metric $M(S(p))$ .
Advantage: Unlike uniform grid sampling, this method adaptively explores the continuous design space, focusing on regions surrounding local minima. This significantly reduces the number of required simulations to find the optimal parameter set $p^*$ .

Stage 3: Nonlinear Simulations & Working Point Selection

Process: With the optimal design parameters ( $p^*$ ) fixed, the framework sweeps physical drive quantities ( $Q$ ), such as input frequencies and tone amplitudes.
Output: Computes X-parameters (generalized S-parameters for strong tones) to evaluate the system's nonlinear response.
Performance Function ( $F$ ): A target function (e.g., gain, bandwidth, noise figure) is maximized to determine the optimal working point $q^*$ .
Iteration: The framework can apply nonlinear corrections (e.g., accounting for the Kerr effect) to the metric $M$ and repeat the cycle for higher accuracy.

3. Key Contributions

Unified Framework: Integration of parametric circuit generation, multi-regime (linear/nonlinear) simulations, and Bayesian optimization into a single, extensible Julia-based workflow.
Efficiency: By leveraging Harmonic Balance and Bayesian Optimization, JCO avoids the computational bottlenecks of time-domain simulations, making it feasible to optimize high-dimensional, nonlinear arrays.
Automation: Automates the search for optimal circuit parameters and operating points, replacing manual trial-and-error with a systematic, reproducible design process.
Scalability: Demonstrated capability to handle complex devices with hundreds of unit cells and multiple coupled parameters.

4. Results: Use Case (SNAIL-based JTWPA)

The framework was validated by optimizing a Josephson Traveling-Wave Parametric Amplifier (JTWPA) based on Superconducting Nonlinear Asymmetric Inductive eLements (SNAILs) operating in the three-wave mixing (3WM) regime.

Device Specs: A transmission line with $N=360$ SNAIL cells.
Parameter Space: 7 design parameters ( $A_J, \rho I_c, \alpha, t, L_\ell, C_\ell, P$ ) with discrete values ranging from 2 to 20 per parameter.
Optimization Goal: Minimize reflection ( $S_{11}$ ), ensure phase matching ( $k_p \approx 2k_{p/2}$ ), and suppress second-harmonic generation.
Performance:
- Linear Phase: The Bayesian optimization identified the optimal design parameters ( $p^*$ ) in approximately 2 hours (exploring a subset of the total $14 \times 10 \times \dots$ configurations), with each linear simulation taking ~2 seconds.
- Nonlinear Phase: Nonlinear simulations determined the optimal drive conditions ( $q^*$ ), achieving a gain of ~20 dB in the target bandwidth (4.75–6.75 GHz) centered at 5.75 GHz.
- Total Runtime: The linear phase took ~22 hours; the nonlinear phase took tens of minutes.
Outcome: The optimized device successfully achieved phase-matched amplification with near-quantum-limited noise performance.

5. Significance

Accelerated Design: JCO drastically reduces the time required to design complex superconducting circuits, enabling rapid iteration for quantum hardware development.
Handling Complexity: It effectively addresses the "curse of dimensionality" in nonlinear circuit design, which traditional methods often fail to solve efficiently.
Future-Proofing: The modular architecture allows for easy extension to other devices (e.g., qubits, mixers) and the integration of advanced corrections (e.g., Kerr nonlinearity) and dimensionality-reduction techniques (e.g., PCA).
Community Impact: As an open-source tool built on Julia, it provides a robust, high-performance foundation for the broader superconducting quantum community to systematically develop quantum-limited amplifiers and other critical components.