Technologies such as batteries, biomaterials, and heterogeneous catalysts have functions that are defined by mixtures of molecular and mesoscale components. As yet, this multi-length scale complexity cannot be fully captured by atomistic simulations, and the design of such materials from first principles is still rare 1-5. Likewise, experimental complexity scales exponentially with the number of variables, restricting most searches to narrow areas of materials space. Robots can assist in experimental searches 6-14 but their widespread adoption in materials research is challenging because of the diversity of sample types, operations, instruments and measurements that is required. Here we use a mobile robot to search for improved photocatalysts for hydrogen production from water 15. The robot operated autonomously over 8 days, performing 688 experiments within a 10-variable experimental space, driven by a batched Bayesian search algorithm 16-18. This autonomous search identified photocatalyst mixtures that were six times more active than the initial formulations, selecting beneficial components and deselecting negative ones. Our strategy uses a dexterous 19,20 free-roaming robot 21-24 , automating the researcher rather than the instruments. This modular approach could be deployed in conventional laboratories for a range of research problems beyond photocatalysis. Leverhulme Research Centre for Functional Materials Design, the Engineering and Physical Sciences Research Council (EPSRC) (EP/N004884/1), the Newton Fund (EP/R003580/1), and CSols Ltd. X.W. and Y.B. thank the China Scholarship Council for a Ph.D. studentship. We thank KUKA Robotics for help with gripper design and initial implementation of the robot. Author contributions. B.B. developed the workflow, developed and implemented the robot positioning approach, wrote the control software, designed the bespoke photocatalysis station, and carried out experiments. P.M.M. and V.V.G. developed the optimiser and its interface to the control software. X.L. advised on the photocatalysis workflow. C.M.A., Y.B. and X.L. synthesized materials. Y.B. performed kinetic photocatalysis experiments. X.W. performed NMR analysis and synthesized materials. B.L. carried out initial scavenger screening. R.C. and N.R. helped to build the bespoke stations in the workflow. B.H. analysed the robustness of the system, assisted with the development of control software, and operated the workflow during some experiments. B.A. helped to supervise the automation work. R.S.S. helped to supervise the photocatalysis work. A.I.C. conceived the idea, set up the five hypotheses with BB, and coordinated the research team. Data was interpreted by all authors and the manuscript was prepared by A.I.

Conjugated polymers are an emerging class of photocatalysts for hydrogen production where the large breadth of potential synthetic diversity presents both an opportunity and a challenge. Here, we integrate robotic experimentation with high-throughput computation to navigate the available structure–property space. A total of 6354 co-polymers was considered computationally, followed by the synthesis and photocatalytic characterization of a sub-library of more than 170 co-polymers. This led to the discovery of new polymers with sacrificial hydrogen evolution rates (HERs) of more than 6 mmol g–1 h–1. The variation in HER across the library does not correlate strongly with any single physical property, but a machine-learning model involving four separate properties can successfully describe up to 68% of the variation in the HER data between the different polymers. The four variables used in the model were the predicted electron affinity, the predicted ionization potential, the optical gap, and the dispersibility of the polymer particles in solution, as measured by optical transmittance.

Three series of conjugated microporous polymers (CMPs) were studied as photocatalysts for hydrogen production from water using a sacrificial hole scavenger. In all cases, dibenzo[b,d]thiophene sulfone polymers outperformed their fluorene analogues. A porous network, S-CMP3, showed the highest hydrogen evolution rates of 6076 μmol h −1 g −1 (λ > 295 nm) and 3106 μmol h −1 g −1 (λ > 420 nm), with an external quantum efficiency of 13.2% at 420 nm. S-CMP3 outperforms its linear structural analogue, P35, whereas in other cases, nonporous linear polymers are superior to equivalent porous networks. This suggests that microporosity might be beneficial for sacrificial photocatalytic hydrogen evolution, if suitable linkers are used that do not limit charge transport and the material can be wetted by water as studied here by water sorption and quasi-elastic neutron scattering.

Covalent organic frameworks (COFs) are distinguished from other organic polymers by their crystallinity1–3, but it remains challenging to obtain robust, highly crystalline COFs because the framework-forming reactions are poorly reversible4,5. More reversible chemistry can improve crystallinity6–9, but this typically yields COFs with poor physicochemical stability and limited application scope5. Here we report a general and scalable protocol to prepare robust, highly crystalline imine COFs, based on an unexpected framework reconstruction. In contrast to standard approaches in which monomers are initially randomly aligned, our method involves the pre-organization of monomers using a reversible and removable covalent tether, followed by confined polymerization. This reconstruction route produces reconstructed COFs with greatly enhanced crystallinity and much higher porosity by means of a simple vacuum-free synthetic procedure. The increased crystallinity in the reconstructed COFs improves charge carrier transport, leading to sacrificial photocatalytic hydrogen evolution rates of up to 27.98 mmol h−1 g−1. This nanoconfinement-assisted reconstruction strategy is a step towards programming function in organic materials through atomistic structural control.

Here we study how the introduction of nitrogen into poly(pphenylene) type materials affects their ability to act as hydrogen evolution photocatalysts. Direct photocatalytic water splitting is an attractive strategy for clean energy production, but understanding which material properties are important, how they interplay, and how they can be influenced through doping remains a significant challenge, especially for polymers. Using a combined experimental and computational approach, we demonstrate that introducing nitrogen in conjugated polymers results in either materials that absorb significantly more visible light but worse predicted driving force for water/sacrificial electron donor oxidation, or materials with an improved driving force that absorb relatively less visible light. The latter materials are found to be much more active and the former much less. The trade-off between properties highlights that the optimization of a single property in isolation is a poor strategy for improving the overall activity of materials.