Geochemical Assessment and Resource Potential of the Critical Minerals in Salt Range Formation, Punjab, Pakistan: An Integrated Approach Using AI and Ml for Renewable Energy Resources

Abstract

The increasing demand for critical minerals required in renewable-energy technologies has encouraged the reassessment of underexplored sedimentary and evaporitic successions. This study evaluates the reconnaissance-level critical-element potential of the Salt Range Formation, Punjab, Pakistan, with emphasis on lithium and associated strategic elements in the Billianwala Salt, Sahwal Marl, and Bandarkas Gypsum members. A total of 23 representative samples were collected from salt, marl, and gypsum lithologies and analyzed through Inductively Coupled Plasma Mass Spectrometry (ICPMS). The results are interpreted as aqua-regia-extractable geochemical concentrations where aqua regia digestion was used, rather than total-rock abundances, because this method mainly dissolves soluble salts, carbonates, sulfides, oxides, organic matter, claybound fractions, and weakly adsorbed phases, but may not completely dissolve resistant silicate minerals. The geochemical dataset was assessed through lithology-wise comparison, correlation analysis, principal component analysis, K-Means clustering, Isolation Forest anomaly detection, and integrated exploration-priority ranking. The results show that critical-element enrichment is not uniformly distributed across the evaporitic succession. Instead, enrichment is mainly controlled by lithology. Marl samples show the strongest relative enrichment in lithium, boron, rare earth elements, uranium, vanadium, and zinc, whereas rock salt and gypsum display more selective evaporite-related signatures, particularly for strontium, barium, and potassium. The highest aqua-regia-extractable lithium value is recorded in Marl-7, followed by other marl-dominated samples, indicating localized enrichment within fine-grained reactive facies rather than broad enrichment throughout the formation. The AI/ML component was used as an exploratory decision-support tool, not as definitive proof of mineralization. Because the dataset is limited and lacks independent mineralization labels, supervised model accuracy is not emphasized. Instead, multivariate and anomalydetection methods are used to recognize geochemical patterns and identify priority samples for follow-up work. The results do not establish an economically viable criticalmineral deposit. Future work should include denser sampling, total digestion, XRD, SEM-EDS, petrography, and mineral-host validation to assess the true resource significance of the identified anomalies