Motivation & Objective

To measure lithium abundances in extremely metal-poor stars with [Fe/H] < -3 to test the universality of the Spite plateau.
To assess whether lithium abundance exhibits a metallicity-dependent trend below [Fe/H] = -3, contrary to the classical plateau model.
To determine the impact of stellar parameters and non-LTE 3D effects on lithium abundance determinations in metal-poor dwarfs.
To evaluate whether the observed lithium distribution can be explained by a single depletion mechanism or requires multiple processes.
To provide a publicly available analytical fit of A(Li) as a function of equivalent width, T_eff, log g, and [Fe/H] for future use.

Proposed method

Stellar parameters (T_eff, log g, [Fe/H]) were derived using four independent T_eff scales, including the infrared flux method and Hα wing fitting with 1D LTE models.
Hα profiles were modeled using two synthetic grids based on different self-broadening theories in 1D LTE, and a final 3D hydrodynamical model was used to compute Hα line profiles.
Lithium abundance A(Li) was measured via 3D NLTE spectral synthesis of the Li i 670.8 nm doublet, accounting for non-LTE and convective motions.
A 3D,NLTE analytical fit of A(Li) was derived as a function of equivalent width, T_eff, log g, and [Fe/H], and made publicly available.
The analysis was repeated across four T_eff estimators to test the robustness of the results to stellar parameter uncertainties.
Stars with T_eff < 6250 K were examined separately to assess their role in the observed trends.

RQ1Does the lithium abundance in metal-poor stars remain constant below [Fe/H] = -3, as predicted by the classical Spite plateau?
RQ2Is there a significant correlation between lithium abundance and metallicity in the most metal-poor stars?
RQ3How do 3D hydrodynamical and NLTE effects influence lithium abundance determinations in metal-poor dwarfs?
RQ4Are the observed lithium abundances consistent with a single depletion mechanism across all metallicities?
RQ5Can the observed scatter and trend in lithium abundance be explained by stellar evolution or observational biases?

The Spite plateau does not extend below [Fe/H] ≈ -3; instead, a significant positive correlation between A(Li) and [Fe/H] is observed, with a slope of 0.30 dex per dex in [Fe/H] at 2-3σ significance.
The observed slope is robust across four different T_eff estimators, with all slopes consistent within 1σ, indicating the result is not an artifact of parameter uncertainty.
The scatter in A(Li) increases by a factor of two toward lower metallicities, and the plateau at [Fe/H] > -2.8 is very thin, with a mean A(Li)_3D,NLTE = 2.199 ± 0.086.
Stars with T_eff < 6250 K are systematically Li-poor and drive the A(Li)–T_eff trend, but removing them does not significantly alter the A(Li)–[Fe/H] slope.
No star below [Fe/H] = -3 has A(Li) above the plateau level, suggesting they formed at plateau abundance and subsequently experienced depletion.
The analytical fit of A(Li) as a function of equivalent width, T_eff, log g, and [Fe/H] is provided for public use in future studies.