An autonomous laboratory for the accelerated synthesis of novel materials

Abstract

To close the gap between the rates of computational screening and experimental realization of novel materials1,2, we introduce the A-Lab, an autonomous laboratory for the solid-state synthesis of inorganic powders. This platform uses computations, historical data from the literature, machine learning (ML) and active learning to plan and interpret the outcomes of experiments performed using robotics. Over 17 days of continuous operation, the A-Lab realized 41 novel compounds from a set of 58 targets including a variety of oxides and phosphates that were identified using large-scale ab initio phase-stability data from the Materials Project and Google DeepMind. Synthesis recipes were proposed by natural-language models trained on the literature and optimized using an active-learning approach grounded in thermodynamics. Analysis of the failed syntheses provides direct and actionable suggestions to improve current techniques for materials screening and synthesis design. The high success rate demonstrates the effectiveness of artificial-intelligence-driven platforms for autonomous materials discovery and motivates further integration of computations, historical knowledge and robotics.

Synthesis and Structure of Vacancy-Ordered Perovskite Ba6Ta2Na2X2O17 (X = P, V): Significance of Structural Model Selection on Discovered Compounds

“…The mixtures were calcined in air at 600 °C for 12 h and heated at 800 °C for 20 h with intermediate grinding. Figure 3 shows XRD patterns of Ba 6 Ta 2 Na 2 X 2 O 17 (X = P, V) together with a simulation pattern of Ba 6 Ta 2 Na 2 V 2 O 17 reported by Szymanski et al 6 The simulated pattern captures the features of the experimental patterns. However, some peaks are absent in the experimental pattern (e.g., a peak at 17.4°), and the relative intensities of the peaks are different, indicating the structure reported is slightly different from the real.…”

“…22 We synthesized the Sb-analogue, Ba 6 Sb 2 Na 2 V 2 O 17 , since Szymanski et al also reported 12H-type Ba 6 Sb 2 Na 2 V 2 O 17 (they reported it as Ba 6 Na 2 V 2 Sb 2 O 17 ). 6 Note that Quarez et al reported 12H-type Ba 6 Sb 2 Na 2 V 2 O 17 without structural refinement. 4 Polycrystalline samples of Ba 6 Sb 2 Na 2 V 2 O 17 were prepared by a standard high-temperature solid-state reaction from stoichiometric mixtures of BaCO 3 (99.95% Kojundo Chemical Lab.…”

“…Figure 3. Laboratory XRD patterns of Ba 6 Ta 2 Na 2 X 2 O 17 (X = P, V)with a simulation pattern of reported CIF 6. …”

A simple solution to the Rietveld refinement recipe problem

“…Likewise, the AutoFP expert system allows for automation of FullProf (Cui et al, 2015;Rodrı ´guez-Carvajal, 1993). More recently, Szymanski et al (2023) described a robotically enabled self-driving inorganic synthesis laboratory that includes an 'automated approach to multiphase Rietveld refinement' based on GSAS-II. Many of the refinement plots provided in that work appear as if they would benefit from further refinement progress, indicating that further work on automating refinements is still needed (Peplow, 2023).…”

FAIR Data and Software: Improving Efficiency and Quality of Biocatalytic Science

“…Biocatalysis is considered as a cornerstone of a sustainable bioeconomy. , However, it is not widely applied in the fine chemical industry, yet, because the success of biocatalysis is limited by the speed of development and implementation of bioprocesses . In order to accelerate development, novel concepts and integrated approaches to process design have been successfully explored, such as synthetic biology, retrobiosynthesis, , and biofoundries. , They all take advantage of decreasing sequencing costs, which reduced by 6 orders of magnitude in the last 20 years (), high throughput experimental techniques, such as liquid handling and microfluidics techniques for screening of enzyme libraries, − substrate libraries, or reaction conditions, − and automated robotic platforms for screening and design. − As a consequence, a rapidly increasing stream of data is generated, which feeds data-driven modeling methods that depend on large volumes of data. Machine learning is applied to guide directed evolution of enzymes and to efficiently explore reaction space .…”