Quandela publishes a Systematic review of Perceval, a Software Platform for Discrete Variable Photonic Quantum Computing intended for Photonic Experimentations

Quandela Publishes A Systematic Review Of Perceval, A Software Platform For Discrete Variable Photonic Quantum Computing Intended For Photonic Experimentations

Quandela, a French company founded in 2017, published an in-depth systematic overview of Perceval, an open-source software platform for simulating and interfacing with discrete-variable photonic quantum computers, in Quantum Journal on the 20th of February, 2023, through several Perceval experiments, such as photonic experiments and photonic simulations of various quantum algorithms, ranging from Grover’s and Shor’s to examples of quantum machine learning.

Perceval’s Python front-end enables the creation of photonic circuits from basic photonic building elements such as photon sources, beam splitters, phase shifters, and detectors. Several computational back-ends are available, each optimized for a certain use case. They employ cutting-edge simulation approaches that cover both weak simulation, sampling, and strong simulation.

In this study, the researchers offered an idealized discrete variable (DV) model, which is still emphasized to have numerous noise, flaws, and errors that will occur in real implementations. Perceval is designed to be a tool for developing, modeling, simulating, and optimizing realistic DV linear optical circuits and experiments and for including realistic noise models. The study aims to provide an in-depth overview of how Perceval handles defective photon purity at the source and distinguishability due to imperfect synchronization.

Moreover, researchers have also identified that Perceval is intended to be a useful toolkit for experimentalists who want to easily model, design, simulate, or optimize a discrete-variable photonic experiment, theorists who want to design algorithms and applications for discrete-variable photonic quantum computing platforms and application designers who want to evaluate algorithms on state-of-the-art photonic quantum computers.

The Fock state representations of photons by Perceval

Perceval is a fully efficient software framework for photonic quantum computing’s discrete variable (DV) model. It extensively uses Fock state representations of photons emitted by sources, developing through linear optical networks made up of beam splitters, phase shifters, waveplates, and other linear optical components, and then be detected.

Perceval’s capabilities can build, optimize, simulate, and eventually transpile DV linear optical circuits for execution on cloud-based physical processors. Although Perceval connects to existing open-source quantum computing toolkits like Qiskit, it also allows users to operate at a level closer to photonic hardware than these toolkits’ normal gate-based qubit quantum circuit model already the focus of a number of other software platforms. It is also more adaptable and has many more features than previous linear optical quantum systems packages.

Perceval is designed to be useful for experimentalists designing photonic experiments, including realistic noise and imperfection modeling, and computer scientists and theorists developing algorithms and applications for photonic quantum computers. It is an open-source platform for community development that includes GitHub, the project forum website, and updated documentation.

In this study, the researchers primarily used Perceval to simulate linear optical circuits with universal function approximators. They demonstrated how these might be used with Quantum Machine Learning approaches to compute differential equation solutions accurately. The precision of these function approximators is determined, among other things, by the number of photons injected into the linear optical circuits.

Perceval’s Structure

Perceval is a modular object-oriented Python code with optimized C routines that use SIMD vectorization. The following section provides a summary of the main classes that are available to the user. Perceval incorporates various cutting-edge methods for performing simulations optimized with a low-level single instruction, multiple data (SIMD) implementations, allowing users to push the boundaries of traditional simulability with desktop computers.

Furthermore, few framework additions are planned for high-performance computing (HPC) cluster deployment, allowing simulation to scale even further.

Perceval’s Modular Design

In a linear optical circuit, information is encoded in the state of photons in specific “modes” established by the circuit designer. The following two classes implement states in Perceval. BasicState is used to express Fock states of n photons over m modes. Photons are indistinguishable by default. However, each photon can be labeled to govern its distinguishability. Then, StateVector extends BasicState to represent state superposition.

Perceval’s Backends

The simulation backends of Perceval are designed to run on local desktop machines, with modifications for HPC clusters. They can be used to execute computational experiments to fine-tune algorithms, compare experimental data from real-world experiments and photonic quantum computing platforms, and duplicate published studies in code.

Perceval to continue the development of realistic noise models.

Perceval is meant to be accessible to both experimental and theoretical physicists, as well as computer scientists, to build a bridge between these fields. To maintain a strong connection between software and hardware for photonic quantum computing, a significant focus of future development will be on the continuing development of realistic noise models that can describe the operation of individual hardware components with increasing accuracy.

The study recommends that future versions of Perceval will include more optimized simulators, noisy simulators, features for working with density matrices and cluster states, more advanced detector features and options to cover both threshold and photon-number resolving detectors, and features for treating feedforward circuits.

Using Perceval and their developed methodologies to solve other differential equations of major practical significance is an intriguing future direction that the researchers hope to pursue.

Read the full white paper here.