In this research, an efficient method of lithium extraction from α-spodumene by thermal treatment with NH₄HF₂ was optimized. Temperature (T), α-spodumene:NH₄HF₂ molar ratio (m), and reaction time (t) were studied using a two-level univariate strategy. The results were modeled using least squares (LS) and artificial neural networks (ANN) and then compared to obtain a predictive model of the system. Both models showed good concordance with the experimental data (R² of 0.9881 and 0.9957, respectively) and with each other. The ANOVA of the cubic model indicated that T, m, t, and the interactions Tt, T², and T³ were significant. Finally, the system was optimized using response surface methodology to maximize Li extraction and minimize operational parameters. The desirability function predicted an extraction value of 95.48 ± 2.50% for T = 156.7 °C, m = 1:17.5, and t = 100.6 min. Experimental lithium extractions of 96.45 ± 3.68% were obtained at 157 °C using a molar ratio of 1:17.5 for 100 min. The products of the thermal treatment were LiF, (NH₄)₃SiF₆·F, (NH₄)₃AlF₆, NH₃, and H₂O. After a water leaching step, the silicon in the sample was separated, obtaining (NH₄)₃SiF₆·F as a by-product. Finally, the solid products were leached with H₂SO₄ 10% (v/v) to solubilize all lithium.

在这项研究中，优化了一种通过NH ₄ HF ₂热处理从α-锂辉石中提取锂的有效方法。使用两级单变量策略研究了温度（T），α-锂辉石：NH ₄ HF ₂摩尔比（m）和反应时间（t）。使用最小二乘法（LS）和人工神经网络（ANN）对结果进行建模，然后进行比较以获得系统的预测模型。两种模型均与实验数据（R²分别为0.9881和0.9957）以及彼此之间显示出良好的一致性。三次模型的方差分析表明T，m，t，并且相互作用Tt，T²和T³很重要。最后，使用响应面方法对系统进行了优化，以最大程度地提取锂并最小化操作参数。期望函数预测，对于T  = 156.7°C，m  = 1：17.5和t  = 100.6 min ，提取值为95.48±2.50％。在157°C下使用1：17.5的摩尔比进行100分钟的实验锂提取率为96.45±3.68％。热处理的产物为LiF，（NH ₄）₃ SiF ₆ ·F，（NH ₄）₃ AlF ₆，NH ₃和H ₂ O的水沥滤步骤之后，将样品中的硅分离，得到（NH ₄）₃的SiF ₆ ·女作为副产物。最后，将固体产物用10％（v / v）的H ₂ SO ₄浸提以溶解所有锂。