Empirical Optimization of Two-Qubit Boundary Disentanglers for Entanglement Entropy Reduction in Quantum Tensor Networks on 156-Qubit Hardware

Simulating strongly correlated many-body quantum systems using Quantum Tensor Networks (QTNs) is fundamentally constrained by entanglement growth and the resulting computational requirements. This research presents an empirical optimization framework for two-qubit boundary disentanglers parameterized in SU(4), using an unconstrained coordinate-descent search over local unitary rotations. The proposed approach achieves an 87.75% reduction in bipartite von Neumann entanglement entropy, from 2.0000 to 0.2450 bits. The resulting Schmidt spectrum enables a 50% bond-dimension truncation to χ = 2, with a reported state fidelity of 99.85%. The optimized four-qubit circuit was compiled using Qiskit 2.5.2 for IBM Quantum's 156-qubit Heron r2 processor (ibm_fez) and executed on physical quantum hardware using 1,024 measurement shots. The compiled circuit has a depth of 36, with 9 native CZ gates and 46 single-qubit SX/RZ gates. The reported measurement results show over 88.6% probability mass concentration in the theoretical Schmidt eigenspaces. This work investigates the practical potential of boundary disentanglement for reducing tensor-network representation costs and supporting the simulation of strongly entangled quantum systems on near-term superconducting quantum hardware.