Search for: All records

Abstract Machine learning models are susceptible to being misled by biases in training data that emphasize incidental correlations over the intended learning task. In this study, we demonstrate the impact of data bias on the performance of a machine learning model designed to predict the likelihood of synthesizability of crystal compounds. The model performs a binary classification on labeled crystal samples. Despite using the same architecture for the machine learning model, we showcase how the model’s learning and prediction behavior differs once trained on distinct data. We use two data sets for illustration: a mixed-source data set that integrates experimental and computational crystal samples and a single-source data set consisting of data exclusively from one computational database. We present simple procedures to detect data bias and to evaluate its effect on the model’s performance and generalization. This study reveals how inconsistent, unbalanced data can propagate bias, undermining real-world applicability even for advanced machine learning techniques. 

Abstract Large-density functional theory (DFT) databases are a treasure trove of energies, forces, and stresses that can be used to train machine-learned interatomic potentials for atomistic modeling. Herein, we employ structural relaxations from the AFLOW database to train moment tensor potentials (MTPs) for four carbide systems: CHfTa, CHfZr, CMoW, and CTaTi. The resulting MTPs are used to relax ~6300 random symmetric structures, and are subsequently improved via active learning to generate robust potentials (RP) that can relax a wide variety of structures, and accurate potentials (AP) designed for the relaxation of low-energy systems. This protocol is shown to yield convex hulls that are indistinguishable from those predicted by AFLOW for the CHfTa, CHfZr, and CTaTi systems, and in the case of the CMoW system to predict thermodynamically stable structures that are not found within AFLOW, highlighting the potential of the employed protocol within crystal structure prediction. Relaxation of over three hundred (Mo_1−xW_x)C stoichiometry crystals first with the RP then with the AP yields formation enthalpies that are in excellent agreement with those obtained via DFT. 

In light of breakthroughs in superconductivity under high pressure, and considering that record critical temperatures (T_cs) across various systems have been achieved under high pressure, the primary challenge for higher T_cshould no longer solely be to increase T_cunder extreme conditions but also to reduce, or ideally eliminate, the need for applied pressure in retaining pressure-induced or -enhanced superconductivity. The topological semiconductor Bi_0.5Sb_1.5Te₃(BST) was chosen to demonstrate our approach to addressing this challenge and exploring its intriguing physics. Under pressures up to ~50 GPa, three superconducting phases (BST-I, -II, and -III) were observed. A superconducting phase in BST-I appears at ~4 GPa, without a structural transition, suggesting the possible topological nature of this phase. Using the pressure-quench protocol (PQP) recently developed by us, we successfully retained this pressure-induced phase at ambient pressure and revealed the bulk nature of the state. Significantly, this demonstrates recovery of a pressure-quenched sample from a diamond anvil cell at room temperature with the pressure-induced phase retained at ambient pressure. Other superconducting phases were retained in BST-II and -III at ambient pressure and subjected to thermal and temporal stability testing. Superconductivity was also found in BST with T_cup to 10.2 K, the record for this compound series. While PQP maintains superconducting phases in BST at ambient pressure, both depressurization and PQP enhance its T_c, possibly due to microstructures formed during these processes, offering an added avenue to raise T_c. These findings are supported by our density-functional theory calculations.