Study of Threshold Behavior of Rotated and Unrotated Surface Codes under Anisotropic Noise

Surface codes are the most promising quantum error correction codes, and their performance can be improved by slight modifications based on noise models. We study this by comparing a vast set of parameters on rotated and unrotated surface codes under realistic anisotropic noise. Our goal is to identify modest tailoring such as basis rotation, crosstalk-aware CNOT order that improves code threshold for hardware with strong X-biased and correlated noise. We compared the standard ZX surface code with a ZY variant with code distances ranging from 5 to 17, physical error rate ranging from 10^-4 to 10^-1, and the Pauli X-biased noise ranging from 0.5 to infinity. We used the stabilizer simulator Stim with Monte Carlo sampling via Sinter and PyMatching 2.3 decoder. At low bias, both codes exhibit conventional crossing with similar thresholds. As the bias parameter increases, ZX becomes asymmetric, indicating directional bias sensitivity. Preliminary results indicate that, in contrast to this, ZY maintains balanced suppression. These trends persist across tested CNOT schedules.