Using machine learning (ML) approach, we unearthed a new III–V semiconducting material having an optimal bandgap for high-efficient photovoltaics with the chemical composition of Gallium–Boron–Phosphide ($$\hbox {GaBP}_{2}$$, space group: $$\hbox {Pna2}_{1}$$). ML predictions are further validated by state-of-the-art ab initio density functional theory simulations. The stoichiometric Heyd–Scuseria–Ernzerhof bandgap of $$\hbox {GaBP}_{2}$$ is noted to be 1.65 eV, a close ideal value (1.4–1.5 eV) to reach the theoretical Queisser–Shockley limit. The calculated electron mobility is similar to that of silicon. Unlike perovskites, the newly discovered material is thermally, dynamically and mechanically stable. Above all the chemical composition of $$\hbox {GaBP}_{2}$$ is non-toxic and relatively earth abundant, making it a new generation of PV material. Using ML, we showed that with a minimal set of features, the bandgap of III–III–V and II–IV–V semiconductor can be predicted up to an RMSE of less than 0.4 eV. We have presented a set of scaling laws, which can be used to estimate the bandgap of new III–III–V and II–IV–V semiconductor, with three different crystal phases, within an RMSE of $$\approx $$ 0.4 eV.

中文翻译：

---

镓-硼-磷化物 ($$\hbox {GaBP}_{2}$$)：一种用于光伏的新型 III-V 半导体

使用机器学习 (ML) 方法，我们发现了一种新的 III-V 半导体材料，该材料具有用于高效光伏的最佳带隙，其化学成分为镓-硼-磷 ($$\hbox {GaBP}_{2}$$ , 空间群：$$\hbox {Pna2}_{1}$$)。ML 预测通过最先进的 ab initio 密度泛函理论模拟得到进一步验证。注意到 $$\hbox {GaBP}_{2}$$ 的化学计量 Heyd–Scuseria–Ernzerhof 带隙为 1.65 eV，这是达到理论 Queisser–Shockley 极限的接近理想值 (1.4–1.5 eV)。计算出的电子迁移率与硅相似。与钙钛矿不同，新发现的材料在热、动态和机械方面都是稳定的。最重要的是，$$\hbox {GaBP}_{2}$$ 的化学成分是无毒且相对地球丰富的，使其成为新一代光伏材料。使用 ML，我们表明，使用最少的特征集，可以预测 III-III-V 和 II-IV-V 半导体的带隙高达小于 0.4 eV 的 RMSE。我们提出了一组标度定律，可用于估计具有三种不同晶相的新型 III-III-V 和 II-IV-V 半导体的带隙，RMSE 为 $$\approx $$0.4 eV .