High-throughput virtual screening for organic electronics: a comparative study of alternative strategies

Nonadiabatic Simulation of Exciton Dynamics in Organic Semiconductors Using Neural Network-Based Frenkel Hamiltonian and Gradients

In this study, we present a multiscale method to simulate the propagation of Frenkel singlet excitons in organic semiconductors (OSCs). The approach uses neural network models to train a Frenkel-type Hamiltonian and its gradient, obtained by the long-range correction version of density functional tight-binding with self-consistent charges. Our models accurately predict site energies, excitonic couplings, and corresponding gradients, essential for the nonadiabatic molecular dynamics simulations. Combined with the fewest switches surface hopping algorithm, the method was applied to four representative OSCs: anthracene, pentacene, perylenediimide, and diindenoperylene. The simulated exciton diffusion constants align well with experimental and reported theoretical values and offer valuable insights into exciton dynamics in OSCs.

From Chemical Drawing to Electronic Properties of Semiconducting Polymers in Bulk: A Tool for Chemical Discovery

A quantum chemistry (QC)/molecular dynamics (MD) scheme is developed to calculate electronic properties of semiconducting polymers in three steps: (i) constructing the polymer force field through a unified workflow, (ii) equilibrating polymer models, and (iii) calculating electronic structure properties (e.g., density of states and localization length) from the equilibrated models by QC approaches. Notably, as the second step of this scheme is generally the most time-consuming one, we introduce an alternative method to compute thermally averaged electronic properties in bulk, based on the simulation of a polymer chain in the solution of its repeat units, which is shown to reproduce the microstructure of polymer chains and their electrostatic effect (successfully tested for five benchmark polymers) 10 times faster than state-of-the-art methods. In fact, this scheme offers a consistent and speedy way of estimating electronic properties of polymers from their chemical drawings, thus ensuring the availability of a homogeneous set of simulations to derive structure-property relationships and material design principles. As an example, we show how the electrostatic effect of the polymer chain environment can disturb the localized electronic states at the band tails and how this effect is more significant in the case of diketopyrrolopyrrole polymers as compared to indacenodithiophene and dithiopheneindenofluorene ones.

Precise Regulation on the Bond Dissociation Energy of Exocyclic C-N Bonds in Various N-Heterocycle Electron Donors via Machine Learning

Heterocycles with saturated N atoms (HetSNs) are widely used electron donors in organic light-emitting diode (OLED) materials. Their relatively low bond dissociation energy (BDE) of exocyclic C-N bonds has been closely related to material intrinsic stability and even device lifetime. Thus, it is imperative to realize fast prediction and precise regulation of those C-N BDEs, which demands a deep understanding of the relationship between the molecular structure and BDE. Herein, via machine learning (ML), we rapidly and accurately predicted C-N BDEs in various HetSNs and found that five-membered HetSNs (5-HetSNs) have much higher BDEs than almost all 6-HetSNs, except emerging boron-N blocks. Thorough analysis disclosed that high aromaticity is the foremost factor accounting for the high BDE of 5-HetSNs, and introducing intramolecular hydrogen-bond or electron-withdrawing moieties could also increase BDE. Importantly, the ML models performed well in various realistic OLED materials, showing great potential in characterizing material intrinsic stability for high-throughput virtual-screening and material design efforts.

Ground-State Orbital Descriptors for Accelerated Development of Organic Room-Temperature Phosphorescent Materials

Organic materials with room-temperature phosphorescence (RTP) are in high demand for optoelectronics and bioelectronics. Developing RTP materials highly relies on expert experience and costly excited-state calculations. It is a challenge to find a tool for effectively screening RTP materials. Herein we first establish ground-state orbital descriptors (πFMOs ) derived from the π-electron component of the frontier molecular orbitals to characterize the RTP lifetime (τp ), achieving a balance in screening efficiency and accuracy. Using the πFMOs , a data-driven machine learning model gains a high accuracy in classifying long τp , filtering out 836 candidates with long-lived RTP from a virtual library of 19,295 molecules. With the aid of the excited-state calculations, 287 compounds are predicted with high RTP efficiency. Impressively, experiments further confirm the reliability of this workflow, opening a novel avenue for designing high-performance RTP materials for potential applications.

Perspectives on development of optoelectronic materials in artificial intelligence age

Optoelectronic devices, such as light-emitting diodes, have been demonstrated as one of the most demanded forthcoming display and lighting technologies because of their low cost, low power consumption, high brightness, and high contrast. The improvement of device performance relies on advances in precisely designing novelty functional materials, including light-emitting materials, hosts, hole/electron transport materials, and yet which is a time-consuming, laborious and resource-intensive task. Recently, machine learning (ML) has shown great prospects to accelerate material discovery and property enhancement. This review will summarize the workflow of ML in optoelectronic materials discovery, including data collection, feature engineering, model selection, model evaluation and model application. We highlight multiple recent applications of machine-learned potentials in various optoelectronic functional materials, ranging from semiconductor quantum dots (QDs) or perovskite QDs, organic molecules to carbon-based nanomaterials. We furthermore discuss the current challenges to fully realize the potential of ML-assisted materials design for optoelectronics applications. It is anticipated that this review will provide critical insights to inspire new exciting discoveries on ML-guided of high-performance optoelectronic devices with a combined effort from different disciplines.

Accelerated Organic Crystal Structure Prediction with Genetic Algorithms and Machine Learning

We present a high-throughput, end-to-end pipeline for organic crystal structure prediction (CSP)─the problem of identifying the stable crystal structures that will form from a given molecule based only on its molecular composition. Our tool uses neural network potentials to allow for efficient screening and structural relaxation of generated crystal candidates. Our pipeline consists of two distinct stages: random search, whereby crystal candidates are randomly generated and screened, and optimization, where a genetic algorithm (GA) optimizes this screened population. We assess the performance of each stage of our pipeline on 21 molecules taken from the Cambridge Crystallographic Data Centre's CSP blind tests. We show that random search alone yields matches for ≈50% of targets. We then validate the potential of our full pipeline, making use of the GA to optimize the root-mean-square deviation between crystal candidates and the experimentally derived structure. With this approach, we are able to find matches for ≈80% of candidates with 10-100 times smaller initial population sizes than when using random search. Lastly, we run our full pipeline with an ANI model that is trained on a small data set of molecules extracted from crystal structures in the Cambridge Structural Database, generating ≈60% of targets. By leveraging machine learning models trained to predict energies at the density functional theory level, our pipeline has the potential to approach the accuracy of ab initio methods and the efficiency of empirical force fields.

Generative organic electronic molecular design informed by quantum chemistry

Generative molecular design strategies have emerged as promising alternatives to trial-and-error approaches for exploring and optimizing within large chemical spaces. To date, generative models with reinforcement learning approaches have frequently used low-cost methods to evaluate the quality of the generated molecules, enabling many loops through the generative model. However, for functional molecular materials tasks, such low-cost methods are either not available or would require the generation of large amounts of training data to train surrogate machine learning models. In this work, we develop a framework that connects the REINVENT reinforcement learning framework with excited state quantum chemistry calculations to discover molecules with specified molecular excited state energy levels, specifically molecules with excited state landscapes that would serve as promising singlet fission or triplet-triplet annihilation materials. We employ a two-step curriculum strategy to first find a set of diverse promising molecules, then demonstrate the framework's ability to exploit a more focused chemical space with anthracene derivatives. Under this protocol, we show that the framework can find desired molecules and improve Pareto fronts for targeted properties versus synthesizability. Moreover, we are able to find several different design principles used by chemists for the design of singlet fission and triplet-triplet annihilation molecules.

Guided diffusion for inverse molecular design

The holy grail of materials science is de novo molecular design, meaning engineering molecules with desired characteristics. The introduction of generative deep learning has greatly advanced efforts in this direction, yet molecular discovery remains challenging and often inefficient. Herein we introduce GaUDI, a guided diffusion model for inverse molecular design that combines an equivariant graph neural net for property prediction and a generative diffusion model. We demonstrate GaUDI's effectiveness in designing molecules for organic electronic applications by using single- and multiple-objective tasks applied to a generated dataset of 475,000 polycyclic aromatic systems. GaUDI shows improved conditional design, generating molecules with optimal properties and even going beyond the original distribution to suggest better molecules than those in the dataset. In addition to point-wise targets, GaUDI can also be guided toward open-ended targets (for example, a minimum or maximum) and in all cases achieves close to 100% validity of generated molecules.

Virtual Screening and Binding Analysis of Potential CD58 Inhibitors in Colorectal Cancer (CRC)

Human cell surface receptor CD58, also known as lymphocyte function-associated antigen 3 (LFA-3), plays a critical role in the early stages of immune response through interacting with CD2. Recent research identified CD58 as a surface marker of colorectal cancer (CRC), which can upregulate the Wnt pathway and promote self-renewal of colorectal tumor-initiating cells (CT-ICs) by degradation of Dickkopf 3. In addition, it was also shown that knockdown of CD58 significantly impaired tumor growth. In this study, we developed a structure-based virtual screening pipeline using Autodock Vina and binding analysis and identified a group of small molecular compounds having the potential to bind with CD58. Five of them significantly inhibited the growth of the SW620 cell line in the following in vitro studies. Their proposed binding models were further verified by molecular dynamics (MD) simulations, and some pharmaceutically relevant chemical and physical properties were predicted. The hits described in this work may be considered interesting leads or structures for the development of new and more efficient CD58 inhibitors.

Identification of Unknown Inverted Singlet-Triplet Cores by High-Throughput Virtual Screening

Molecules where the energy of the lowest excited singlet state is found below the energy of the lowest triplet state (inverted singlet-triplet molecules) are extremely rare. It is particularly challenging to discover new ones through virtual screening because the required wavefunction-based methods are expensive and unsuitable for high-throughput calculations. Here, we devised a virtual screening approach where the molecules to be considered with advanced methods are pre-selected with increasingly more sophisticated filters that include the evaluation of the HOMO-LUMO exchange integral and approximate CASSCF calculations. A final set of 7 candidates (0.05% of the initial 15 000) were verified to possess inversion between singlet and triplet states with state-of-the-art multireference methods (MS-CASPT2). One of them is deemed of particular interest because it is unrelated to other proposals made in the literature.

Reverse Designing the Wavelength-Specific Thermally Activation Delayed Fluorescent Molecules Using a Genetic Algorithm Coupled with Cheap QM Methods

Genetic algorithm (GA) optimization coupled with the semiempirical intermediate neglect of differential overlap (INDO)/CIS method is presented to inversely design the red thermally activation delayed fluorescent (TADF) molecules. According to the predefined donor-acceptor (DA) library to build an AD_n-type TADF candidate, we utilized the chemical notation language SMILES code to generate a TADF molecule and apply the RDKit program to produce the initial 3D molecular structure. A combined fitness function is proposed to evaluate the performance of the functional-lead TADF molecule. The fitness function includes three key parameters, i.e., the emission wavelength, the energy gap (ΔE_ST) between the lowest singlet (S₁)- and triplet (T₁)-excited states, and the oscillator strengths for electron transition from S₀ and S₁. A cheap QM method, i.e., INDO/CIS, on the basis of an xTB-optimized molecular geometry is applied to quickly calculate the fitness function. Finally, the GA approach is utilized to globally search for the wavelength-specific TADF molecules under our predefined DA library, and the optimum 630 nm red and 660 nm deep red TADF molecules are inversely designed according to the evolution of molecular fitness functions.

Deep transfer learning for predicting frontier orbital energies of organic materials using small data and its application to porphyrin photocatalysts

A deep transfer learning approach is used to predict HOMO/LUMO energies of organic materials with a small amount of training data.

Wijaya JB, Grünewald F, de Vries AH, Marrink SJ, Broer R, Havenith RWA. Strategies for Enhancing the Dielectric Constant of Organic Materials

High dielectric constant organic semiconductors, often obtained by the use of ethylene glycol (EG) side chains, have gained attention in recent years in the efforts of improving the device performance for various applications. Dielectric constant enhancements due to EGs have been demonstrated extensively, but various effects, such as the choice of the particular molecule and the frequency and temperature regime, that determine the extent of this enhancement require further understanding. In this work, we study these effects by means of polarizable molecular dynamics simulations on a carefully selected set of fullerene derivatives with EG side chains. The selection allows studying the dielectric response in terms of both the number and length of EG chains and also the choice of the group connecting the fullerene to the EG chain. The computed time- and frequency-dependent dielectric responses reveal that the experimentally observed rise of the dielectric constant within the kilo/megahertz regime for some molecules is likely due to the highly stretched dielectric response of the EGs: the initial sharp increase over the first few nanoseconds is followed by a smaller but persistent increase in the range of microseconds. Additionally, our computational protocol allows the separation of different factors that contribute to the overall dielectric constant, providing insights to make several molecular design guides for future organic materials in order to enhance their dielectric constant further.

Feasibility of p-Doped Molecular Crystals as Transparent Conductive Electrodes via Virtual Screening

Transparent conducting materials are an essential component of optoelectronic devices. It is proven difficult, however, to develop high-performance materials that combine the often-incompatible properties of transparency and conductivity, especially for p-type-doped materials. In this work, we have employed a large set of molecular semiconductors extracted from the Cambridge Structural Database to evaluate the likelihood of transparent conducting material technology based on p-type-doped molecular crystals. Candidates are identified imposing the condition of high highest occupied molecular orbital (HOMO) energy level (for the material to be easily dopable), high charge carrier mobility (for the material to display large conductivity when doped), and a high threshold for energy absorption (for the material to absorb radiation only in the ultraviolet). The latest condition is found to be the most stringent criterion in a virtual screening protocol on a database composed of structures with sufficiently wide two-dimensional (2D) electronic bands. Calculation of excited-state energy is shown to be essential as the HOMO-lowest unoccupied molecular orbital (LUMO) gap cannot be reliably used to predict the transparency of this material class. Molecular semiconductors with desirable mobility are transparent because they display either forbidden electronic transition(s) to the lower excited states or small exchange energy between the frontier orbitals. Both features are difficult to design but can be found in a good number of compounds through virtual screening.

Organic materials repurposing, a data set for theoretical predictions of new applications for existing compounds

We present a data set of 48182 organic semiconductors, constituted of molecules that were prepared with a documented synthetic pathway and are stable in solid state. We based our search on the Cambridge Structural Database, from which we selected semiconductors with a computational funnel procedure. For each entry we provide a set of electronic properties relevant for organic materials research, and the electronic wavefunction for further calculations and/or analyses. This data set has low bias because it was not built from a set of materials designed for organic electronics, and thus it provides an excellent starting point in the search of new applications for known materials, with a great potential for novel physical insight. The data set contains molecules used as benchmarks in many fields of organic materials research, allowing to test the reliability of computational screenings for the desired application, "rediscovering" well-known molecules. This is demonstrated by a series of different applications in the field of organic materials, confirming the potential for the repurposing of known organic molecules.