Predicting fs-laser-induced NV centers with PCA and neural networks

Diamond hosts a variety of lattice defects, among which nitrogen-vacancy (NV) centers stand out due to their relevance in quantum photonics with optically addressable qubits. Yet, the complex laser–material interactions governing its formation are not fully understood, and the influence of laser parameters on NV generation still raises open questions. Here, we investigate the generation of NV centers using principal component analysis (PCA) and artificial neural networks (ANNs) as predictive tools based on femtosecond laser parameters. Experimental results from femtosecond laser micromachining of diamond provided the dataset for our analysis. We employed PCA to reduce data dimensionality and uncover dominant experimental trends, while a multilayer perceptron model was trained to predict NV center generation under simulated conditions. GridSearch optimization and Leave-One-Out cross-validation (LOOCV) ensured the best performance and robustness of the ANN. Our results reveal that NV center generation is directly proportional to laser peak fluence and inversely proportional to pulse duration and excitation wavelength. Notably, PCA and ANN modeling independently converged on consistent trends, reinforcing the reliability of the observed parameter–defect relationships. This convergence supports the development of predictive frameworks for controlled color center generation in diamond with greater precision. • This study applied PCA and artificial neural networks to predict NV center generation in diamond. • NV center formation increases with laser fluence and decreases with pulse duration and wavelength. • PCA and neural networks reached the same conclusions observed experimentally, confirming the reliability of the results.