Traversing Gaussian and non-Gaussian regimes in photonic quantum simulators: a review

Quantum photonic simulators are pivotal for designing and analyzing quantum photonic circuits that utilize both continuous-variable (CV) and discrete-variable (DV) encodings. This review critically examines the transition from Gaussian to non-Gaussian photonic models and the computational overhead that emerges from simulating systems with more than 20 modes or photon numbers exceeding 10 per mode. Gaussian states; such as coherent and squeezed states, are modeled efficiently using first-moment vectors and covariance matrices of dimension 2n × 2n 2 n × 2 n , where n is the number of modes. Non-Gaussian elements like the cubic phase gate U = exp (i γ x̂³) U = exp ( i γ x ˆ 3 ) and photon-subtracted states demand full Fock space representations and scale exponentially with photon number d² d 2 , where d is the Hilbert space dimension. We analyze simulators including , , and , benchmarking their support for tensor networks (bond dimension D >100), GPU acceleration, thus NVIDIA A100 with 1555 GB/s bandwidth, and SoC implementations such as Edge TPU with 8 TFLOPS. Special focus is given to noise models such as photon loss with transmissivity η , thermal noise modeled with n̄ = 1.5 n ¯ = 1.5 , and dark count Poisson distributions ( λ = 0.1 λ = 0.1 ). The review underscores scalable techniques and hybrid quantum-classical workflows vital for near-term photonic quantum computing.