Quantum circuit simulation with high-order optical tensor processing

Abstract

Classical simulation is essential for developing and testing quantum circuits, but the exponential growth of Hilbert space challenges conventional digital processors. Here, we introduce a classical optical quantum simulation architecture that treats gates and batches of input states as independent tensor dimensions rather than separate simulation instances, while mapping eligible tensor network contractions onto the same processor. Using co-designed multidimensional encoding and readout across spatial and wavelength degrees of freedom, we experimentally realize high-order tensor-matrix multiplication in a single propagation and process all computational-basis inputs for each 4-qubit gate. Using a physics-based optical propagation model, we further numerically evaluate the architecture for the Deutsch algorithm and for the tensor network of all 128 target instances of a 7-qubit Grover algorithm. These results establish a unified framework for gate-state batch and tensor network simulation, providing a route to exploit optical concurrency across quantum simulation tasks.

The paper addresses how classical optical hardware can support quantum-circuit simulation by processing batches of gates and input states and by executing tensor-network contractions. Its architecture represents gates and batches of input states as independent tensor dimensions, rather than treating each gate–input combination as a separate simulation instance. Spatial and wavelength degrees of freedom provide multidimensional encoding and readout, with eligible tensor-network contractions mapped onto the same optical processor. The experimental result is high-order tensor–matrix multiplication in a single optical propagation, including processing all $16$ computational-basis inputs for each demonstrated $4$-qubit gate. The architecture is also evaluated numerically using a physics-based optical propagation model for the Deutsch algorithm and for tensor networks representing all $128$ target instances of a $7$-qubit Grover algorithm. These algorithm-level results are simulations of the optical architecture, rather than experimental circuit demonstrations. The stated scope is a common architecture for gate–state batch processing and eligible tensor-network contractions. The abstract does not specify which contractions are eligible, nor does it report numerical accuracy, runtime, energy consumption, or comparisons with digital simulation methods. It also does not establish scaling beyond the reported small circuits or a reduction in the exponential resources generally required for classical quantum simulation. With only the abstract available, further implementation constraints and limitations cannot be assessed.