Techno-Economic Analysis of Lithium Extraction from Geothermal Brines

The United States has a large, domestic source of lithium in geothermal fluids, especially at the Salton Sea region of southern California, where estimates of lithium pass-through at geothermal plants exceed 24,000 metric tons per year, based on 2019 geothermal plant operations. Lithium extraction from geothermal brines offers the potential to provide the United States with a secure, domestic supply of lithium to meet the increasing demands of electric vehicles, grid energy storage, portable electronics, and other end-use applications. Additionally, the use of direct extraction technologies allows for a more sustainable lithium supply relative to current evaporative brine and hardrock mining operations in terms of land use, water use, time to market with lithium products, and carbon intensity of operations. Review of projects focused on lithium extraction from geothermal and other brine indicates that expected costs to produce end products lithium carbonate (Li2CO3) and lithium hydroxide monohydrate (LiOH·H2O) are near $4,000/metric ton of lithium carbonate equivalent (LCE). Reported internal rates of return suggest this production cost target is economically feasible with estimated prices of =$11,000/mt LCE. For comparison, market prices since mid-2018 have ranged from approximately $20,000/mt to $7,500/mt LCE. Many techniques and process strategies have been proposed to extract lithium directly from geothermal and other brines, and these can be generally categorized into adsorption, ion exchange, and solvent extraction techniques. Of these technologies, the ones currently advancing to pilot- and near-commercial-scale demonstrations involve adsorption and ion exchange techniques. Future advancements are likely to come from improved sorbent and solvent performance, with goals to increase lithium selectivity relative to competing ions, increase operating cycles between regeneration and replacement, and lower costs of sorbent and solvent manufacturing.