With ground- and space-based telescopes, astronomers have now obtained high-resolution spectra for millions of stars and discovered thousands of new worlds — exoplanets — beyond our Solar System. Astrophysics is entering a new era in which elemental abundances can be measured with unprecedented precision, not only in stars but also in the atmospheres of their planets. These chemical fingerprints provide crucial clues to stellar evolution, Galactic chemical enrichment, and the formation and composition of planetary systems. However, interpreting stellar and planetary spectra remains challenging, as the simplifying assumptions traditionally adopted in spectral modeling often introduce systematic biases.

Neutral sodium (Na I) plays a particularly central role in this context, serving as both a tracer of Galactic chemical evolution in late-type stars and a diagnostic of atmospheric processes in exoplanets. It also acts as a key indicator of distinct stellar populations, such as those in globular clusters, where characteristic abundance variations reveal multiple stellar generations. Classical hydrostatic one-dimensional (1D) models assuming local thermodynamic equilibrium (LTE) can systematically overestimate Na I abundances — by up to 0.5 dex in giant stars — because the simplifying approximations break down. In high-resolution transmission spectroscopy of exoplanets, additional stellar phenomena such as center-to-limb variations (CLV) and the Rossiter–McLaughlin signal during transits must also be taken into account. These stellar effects can mimic or obscure planetary absorption features, leading to false detections if not modeled correctly. Accurately treating such processes requires realistic three-dimensional (3D) radiation-hydrodynamic (RHD) stellar atmospheres combined with non-local thermodynamic equilibrium (non-LTE) radiative transfer.

In this thesis, I develop a state-of-the-art grid of 3D non-LTE synthetic spectra for Na I lines in FGK-type stars, based on the extended and refined Stagger-grid of RHD models. This grid enables more accurate sodium abundance determinations in large spectroscopic surveys — such as GALAH DR4, which recently published parameters and abundances for nearly one million stars — and improves the interpretation of high-resolution exoplanet spectra from instruments such as ESPRESSO on the VLT and the forthcoming ANDES spectrograph on the ELT.

The thesis demonstrates several applications of these 3D non-LTE models: (i) an analysis of spatially resolved solar spectra from the Swedish 1-m Solar Telescope (Paper I); (ii) atmospheric characterization of four giant exoplanets observed with ESPRESSO (Paper II); and (iii) a detailed investigation of Na I abundances across Galactic stellar populations using GALAH DR4 data (Paper III). Together, these studies show that 3D non-LTE modeling provides a unified and more physically accurate framework for interpreting sodium lines in both stellar and planetary contexts.