UC Berkeley

Semiconductor nanocrystals exhibit size-, shape-, and composition-dependent electronic and optical properties governed by quantum confinement. Predictive computational modeling of these effects remains challenging because accurate first-principles excited-state methods such as GW/BSE scale too steeply for nanocrystals containing hundreds to thousands of atoms, while traditional semi-empirical approaches sometimes struggle to achieve the flexibility and transferability needed across diverse semiconductor materials and alloy compositions. At the same time, despite major advances in machine learning for ground-state materials modeling, its application to excited-state properties, electron-phonon interactions, optical response, and charge carrier dynamics of large nanomaterial systems remains limited.This dissertation addresses this challenge through the development of DeepPseudopot, a machine-learned atomistic semi-empirical pseudopotential framework for nanomaterials. By combining a neural-network description of local screened pseudopotentials with physically motivated nonlocal and spin-orbit terms, DeepPseudopot retains the physical grounding of pseudopotential theory while extending its flexibility and predictive capability. Trained on high-level theory reference data from a compact set of bulk system calculations, the model enables accurate atomistic predictions of electronic structure across a broad range of semiconductors. Importantly, it generalizes beyond the materials and environments explicitly included in training, allowing predictive treatment of unseen crystal phases, continuous alloy compositions, chemically distinct systems, and nanostructures of experimentally relevant sizes without retraining.Using this framework, this dissertation investigates the excited-state properties and photophysics of colloidal III-V nanocrystals newly accessible through recent synthetic advances. In close collaboration with experiment, DeepPseudopot is applied to GaAs quantum dots to predict excitonic fine structure, radiative lifetimes, Stokes shifts, and ensemble optical properties in quantitative agreement with measurements. The framework is further extended to describe single-nanocrystal photoluminescence spectral lineshape, providing a microscopic interpretation of fluorescence line narrowing and photoluminescence excitation spectra. It is then used to reveal how alloying in ternary III-V quantum dots, including In1−xGaxP and GaP1−xAsx, continuously reshapes band-edge wavefunction character and symmetry, thereby tuning oscillator strengths and photoluminescence quantum yield. Combined with open quantum system dynamics simulation and two-dimensional electronic spectroscopy, these atomistic calculations further clarify how alloying controls ultrafast hot-exciton relaxation pathways and electronic coherences. Together, these results highlight the need for a unified atomistic description of nanocrystal photophysics and establish a foundation for data-driven design of semiconductor nanomaterials.