Currently, the energies of the highest occupied molecular orbitals (HOMO) and lowest unoccupied molecular orbitals (LUMO) calculated by quantum chemical methods are important for the discovery of new materials, i.e., estimating the optoelectronic properties of candidate organic molecules and filtering databases. For this purpose Alfa Chemistry offers HOMO and LUMO forecasting services. To swiftly estimate HOMO and LUMO orbital energies, we have tested a number of machine learning techniques. These algorithms can take the place of computationally taxing DFT computations to provide frontier orbital energies and other useful features of whole molecules, bonds, or atoms.
Ultra-fast estimates can be produced by machine learning using data that has already been calculated using DFT or ab initio techniques. For this, reasonable machine learning methods, suitable machine learning descriptors, and well-designed data sets are needed. Machine learning models are expected to provide early filters that can select a set of promising molecules for further screening by other computationally more demanding methods.
Machine Learning Methods
- Random Forest (RF): An RF is a collection of unpruned regression or classification trees created using bootstrap samples from the training set. At Alfa Chemistry, RF is used to develop regression models to estimate HOMO, LUMO, and GAP energies. We generated RFs using the R program of the RandomForest library.
- Support Vector Machines (SVMs): SVMs map multidimensional data to hyperspace via a nonlinear transformation and then apply linear regression in that space. Boundaries are located using examples from the training set, which are called support vectors. We use the Weka implementation of the LIBSVM software to explore SVM models.
- Multilayer Perceptron (MLP): MLP is a feedforward neural network (NN) that we implement specifically in Weka. The perceptron computes a single output from multiple real-valued inputs - it forms a linear combination with the input values - and then predicts the output through a nonlinear decision surface. The MLP can be optimized using a backpropagation algorithm.
Fig 1. Prediction of the HOMO (a), LUMO (b), and HOMO-LUMO gap (c) energies in the test set by Random Forests trained on the basis of different molecular descriptors. (Pereira F, et al. 2017)
|Project Name||HOMO and LUMO Prediction Service|
|Deliverables||We provide all raw data and analysis services to our customers.|
|Samples Requirement||Our services require specific requirements from you.|
|Timeline Decide||According to customers' needs|
|Price||Please contact us for an inquiry|
The demand for ultra-thin, lightweight and flexible electronic devices has led to the exploration of organic materials that may have a unique combination of electronic, chemical, and mechanical properties. For example, organic materials have important applications in organic light-emitting diodes (OLEDs), organic photovoltaic devices (OPVs), and organic thin-film transistors (OTFTs). In addition to materials, HOMO and LUMO energy calculations can be used to evaluate chemical reactivity and to derive molecular descriptors for QSAR and QSPR models.
Alfa Chemistry provides fast, professional and high-quality HOMO and LUMO forecasting services to global customers at competitive prices. Our customers can directly contact our staff and provide timely feedback on their inquiries. If you are interested in our services, please contact us for more details.
- Pereira F, et al. (2017). "Machine Learning Methods to Predict Density Functional Theory B3LYP Energies of HOMO and LUMO Orbitals." J. Chem. Inf. Model. 57(1): 11-21.