Organic solar cells (OSCs) are a third type of solar cell technology [[1], [2], [3]]. OSCs are an emerging solar power advancement that uses organic polymeric materials as light-absorbing layers. [4]. Organic photovoltaic cells (OPVs) are thin-film oriented cells that can store large amounts of solar energy compared to other photovoltaic technologies.Inherited properties of organic materials such as charge transport and light absorption/retention abilities are key advantages in OSC development [[5], [6], [7]]. Organic semiconductors are π-conjugated organic molecules composed of carbon and hydrogen atoms and heteroatoms such as sulfur, oxygen, and nitrogen. [[8], [9], [10]]. They have efficient conductive and electroluminescent properties. [11,12]. Organic semiconductor materials include metal complexes, polycyclic aromatic hydrocarbons, polythiophenes, polyphenylamines, and other materials produced from conductive polymers such as polyacenes. [[13], [14], [15]].
In particular, the bandgap is an important parameter for controlling the photovoltaic properties of semiconductor materials. [[16], [17], [18]].Low bandgap organic semiconductors require low energy to move electrons from the valence band to the conduction band [19,20]. A lower bandgap increases the electron mobility from the valence band to the conduction band of a semiconductor, improving its conductivity. [21,22]. Furthermore, low bandgap semiconductors can collect more photons compared to wide bandgap semiconductors, thus enabling higher solar energy conversion efficiency. High bandgap semiconductors require higher energy to move electrons, which can lead to lower conductivity. Therefore, low bandgap organic semiconductors show immense potential in a wide range of applications. [23,24].
To further expand the range of applications of organic semiconductors, materials with improved properties are needed. [25,26]. Applications of these low bandgap organic semiconductors include organic field-effect transistors (OFETs), organic light-emitting diodes (OLEDs), OSCs, supercapacitors, and thin-film batteries. [27,28]. Moreover, the inability to obtain high charge carrier mobilities in a reproducible manner limits the industrial use of semiconductors. Intensive research efforts are developing a variety of materials with higher mobilities that may exceed amorphous silicon. [29]. However, due to the presence of a wide variety of organic constituents, these few developed materials are thought to be of no use. Several new organic molecules can be designed by assembling potential organic building blocks.
To date, experimental approaches require a complete cycle for the discovery of new materials, including synthetic procedures, device fabrication, performance evaluation, molecular tuning, and trial-and-error experience. [30,31]. Computational/theoretical studies can play an important role in the advancement of OPV by providing further insight into the electronic properties (charge transport mechanisms). [[32], [33], [34]]. These studies are even more useful when combined with experiments. [[35], [36], [37]]. In recent years, various groups have been working on the discovery of high-performance materials that can guide experimenters to synthesize potential materials/targets. [38,39].
Machine learning, a sub-field of artificial intelligence, uses large-scale data mining approaches to extract common designs, potentially for various applications such as drug discovery, superhard materials, and heterogeneous catalysis. has emerged as a powerful approach for screening and designing targeted targets.standard [40,41]. In these screening/design approaches, easily calculable or easily listed properties of the molecule (system) known as molecular descriptors are used to identify important properties of the molecule/material/system (e.g. carrier mobility, bandgap, etc.). ) is determined to be correlated with These strategies are useful for screening potential candidates for targeted applications. [42,43]. Although various efforts have been made to discover potential materials for diverse applications, there are limited approaches focused on organic semiconductors for photovoltaic applications. [44,45]. By leveraging the potential benefits of machine learning approaches, screening of multiple databases can be rapidly achieved with highest accuracy and low computational cost. [46].
In this study, we use machine learning to design new low bandgap semiconductors and predict the properties of OSCs. For this purpose, data is collected. Different regression models are tested and the best one is selected, i.e. the additional tree regression model. To support the cock flight distance of the final model, t-SNE, learning curve and validation curve are drawn. Structure-based similarity analysis and library enumeration are performed to provide useful information about semiconductor molecules. This study will help researchers design potential candidates for improved solar power performance.