SchroSIM: A Scalable Photonic Quantum Circuit Simulator

SchroSIM is a scalable quantum photonic circuit simulator designed to unify device-level physics, architectural modeling, and differentiable quantum programming within a single computational framework. Photonic quantum computing has emerged as one of the most promising approaches to scalable, fault-tolerant quantum information processing, largely due to its room-temperature operation, compatibility with integrated chip platforms, and the natural ability of photons to propagate long distances without decoherence. Yet, despite advances in photonic hardware, the ecosystem still lacks a mature simulation tool that simultaneously captures nonlinear wave dynamics, Gaussian and non-Gaussian transformations, loss channels, detection physics, and hybrid variational workflows. SchroSIM was created to fill this gap, offering a multi-resolution framework that connects the underlying physics of photonic elements to high-level quantum software abstractions.

The simulator begins at the physical layer using the Nonlinear Schrödinger Equation (NLSE) as its core dynamical model. Traditional quantum photonic simulators often treat waveguides and propagation as idealized linear channels, ignoring phenomena such as self-phase modulation, cross-phase modulation, dispersion, nonlinear absorption, and spatiotemporal reshaping of pulses. SchroSIM instead implements a carefully optimized Split-Step Fourier Method (SSFM), enabling it to model pulse propagation over centimetre- or metre-scale waveguides with realistic dispersion engineering. This makes it possible to explore how photonic pulses deform as they traverse integrated circuits, how soliton-like behaviour emerges under specific parameter regimes, and how nonlinearity introduces transformations that deviate significantly from purely Gaussian evolutions.

Beyond the physical propagation, SchroSIM supports a wide range of Gaussian operations—beam splitters, phase shifters, squeezers, homodyne detection, and displacement operators—mirroring the standard formalism used in continuous-variable (CV) quantum computing. These operations are implemented using both symplectic transformation libraries and differentiable backends compatible with modern machine learning tools. The differentiability is a deliberate design choice, motivated by the growing adoption of quantum-classical hybrid optimization algorithms such as VQE, VQS, gradient-based photonic neural networks, and differentiable quantum architecture search. With a differentiable engine, users can compute gradients of circuit outputs with respect to hardware parameters, enabling automated tuning, optimal control, and gradient-based learning of photonic circuit configurations.

A major innovation in SchroSIM is its explicit handling of non-Gaussian transformations, which are essential for universal photonic quantum computation. Non-Gaussianity typically arises from conditional measurements such as photon subtraction, photon addition, nonlinear detectors, or Kerr-type interactions. Most simulators struggle to bridge Gaussian and non-Gaussian regimes efficiently, yet SchroSIM provides a modular representation that allows researchers to plug in custom non-Gaussian operators, approximate them using series expansions, or simulate their effects using stochastic sampling. This flexibility is crucial for exploring architectures that incorporate cat states, cubic phase states, bosonic codes, or device-specific nonlinearities—topics highly relevant for quantum error correction and fault-tolerant resource state generation.

SchroSIM also integrates a differentiable quantum-classical workflow inspired by frameworks like PennyLane and JAX, but designed specifically for photonic systems. Users can run variational circuits, compute gradients through NLSE propagation, and embed the simulator into machine learning pipelines. This unified architecture turns SchroSIM into more than a physics engine; it becomes a platform for algorithm-hardware co-design. Researchers can optimize circuit parameters, train photonic neural networks, or tune physical device geometries directly through simulation. This functionality is forward-looking, anticipating the growing need for automated design of photonic circuits as systems scale from a few modes to hundreds or thousands of spatial-temporal channels.

On the architectural side, SchroSIM simulates multi-photon interference, bosonic sampling configurations, and Gaussian boson sampling (GBS) circuits with tunable loss, mode mismatch, and detector noise. This makes it useful for both near-term photonic processors and future large-scale architectures. By supporting both time-bin and spatial encoding, SchroSIM reflects real design strategies used by companies such as Xanadu, PsiQuantum, and Quandela. Additionally, the simulator incorporates simplified models of integrated beam splitters, directional couplers, Mach–Zehnder interferometers, phase modulators, and squeezed-light sources, allowing users to prototype architectural layouts before fabrication or experimental implementation.

The anticipated results from SchroSIM are multifaceted. At the physical layer, we expect to generate detailed maps of pulse deformation, nonlinear spectral broadening, soliton dynamics, and energy dissipation across photonic waveguides. These simulations will allow us to benchmark SSFM results against FDTD-based methods such as MEEP, enabling hybrid validation across frequency and time domains. At the circuit layer, we anticipate accurate predictions of mode interference patterns, photon-number distributions, and the effect of loss or noise on quantum advantage demonstrations such as GBS sampling tasks. At the variational layer, we anticipate discovering new photonic ansätze, efficient embeddings for continuous-variable machine learning, and data-driven optimization of circuit topologies.

In its final implementation, SchroSIM aims to serve as a bridge between device physics, quantum architectures, and quantum compiler stacks. As photonic processors grow in scale, the ability to co-simulate nonlinear propagation, Gaussian circuits, measurement-induced non-Gaussian transformations, and differentiable optimization will become increasingly important. SchroSIM is positioned to become a foundational tool for researchers developing photonic algorithms, exploring fault-tolerant continuous-variable encodings, or designing next-generation photonic quantum hardware.

The long-term vision is to integrate SchroSIM into a fully autonomous photonic design pipeline, where physical parameters, circuit architectures, and error-mitigated variational strategies are jointly optimized by a combination of physics-based simulation and machine learning. This will support not only research groups in academia but also industry teams working on fabrication-aware quantum photonic systems. Ultimately, SchroSIM aspires to accelerate the transition from theoretical continuous-variable models to real, scalable quantum photonic processors.