Introduction Chemical sensor arrays have attracted significant attention as a powerful tool for detecting, discriminating, and identifying target analytes, in particularly, various smells consisting of a complex mixture of gaseous molecules. In a chemical sensor array, sensing signals are obtained by measuring interactions induced by the sorption of target analytes in sensing materials designed to respond to a wide range of chemical classes. Although a wide range of applications has been demonstrated in a variety of fields, these pattern recognition-based analyses are basically limited to gaseous analytes.In this study, we propose the reverse approach, that is, pattern recognition of solid materials. As the sensing signals of chemical sensors are based on the interaction between gases and solids, a sensing element and a target analyte can be exchanged, i.e. solid materials as target analytes and gaseous molecules as sensing probes, leading to the pattern recognition of solid materials (Figure 1a). To demonstrate this new approach, we focus on a nanomechanical Membrane-type Surface stress Sensors (MSS) [1] as an example of sensing platform. A nanomechanical sensor detects mechanical information derived from the interactions between gaseous molecules and solid materials with high sensitivity [2]. Since it has been confirmed that almost all kinds of solid materials including organic small molecules, polymers and inorganic nanoparticles provide some signals as a result of the gas-solid interaction [1,3], a nanomechanical sensor is an ideal platform to examine various kinds of solid materials. We have demonstrated successful discrimination of polymers having different molecular weights as well as those composed of different monomers by means of pattern recognition [4]. Results and Conclusions We performed identification of 4 different polymers (polystyrene PS; poly(4-methylstyrene) P4MS; polycaprolactone PCL; poly(vinylidene fluoride) PVF) through the pattern recognition. Twelve different vapors are used as probes to acquire the signals for each gas-solid interaction. Upon exposure of each vapor, the polymers exhibited unique responses, reflecting the differences in chemical and physical affinity between each polymer and vapor. For the obtained dataset, we conducted principal component analysis (PCA) and linear discriminant analysis (LDA), respectively. With the all features from the 12 vapors, the 4 different polymers can be clearly distinguished by forming well-separated clusters by PCA as well as LDA (Figure 1b and c). On the LDA plane, PS and P4MS form clusters close to each other, reflecting their similarity in the chemical and physical affinity to each probe gas.We also developed machine learning models based on support vector machine (SVM) classifier. To optimize and evaluate the models, we employed 5 × 2 cross validation. All combinations of each probe gas were calculated to create SVM model. Identification accuracy depending on the combination of probe gases is shown as a dot plot in Figure 2c. Remarkably, almost a quarter of all combinations of probe gases resulted in 100% identification accuracy with 2 to 12 kinds of probe gases by the SVM analysis (Figure 1d and e). These results clearly indicate that the appropriate selection of probe gases depending on the target solid samples leads to the highly accurate and efficient identification.To evaluate further applicability of the pattern recognition of solid materials, we demonstrated the identification of molecular weight of polymers. The LDA provided clear discrimination of polystyrenes including PS and P4MS in terms of molecular weight. These results indicate that the pattern recognition can be also effectively applied to solid materials to identify each analyte even with similar chemical and physical properties. Furthermore, we also demonstrated the practical solids, in this case, differences of rice. Rice was grinded, and the resulting powder was coated on the MSS. By using water and ethanol as a set of probe gases, rice originated from different places can be clearly discriminated by PCA. Since the major components of rice are amyloid and amylopectin, this approach can be discriminate very minute differences. Conclusion In this study, we have demonstrated that solid specimens can be identified by pattern recognition, achieving 100% accuracy in the case of some specific combinations of probe gases. As proved by means of SVM as well as PCA and LDA, even slight differences in material properties such as rice can be also discriminated through this approach. Since any kind of gaseous or volatile molecules can be potentially utilized as a probe for this pattern recognition-based solid materials identification, this approach possesses unlimited possibilities to differentiate solid materials. The potential target solid materials of this concept include inorganic nanoparticles, functional organic materials and biomolecules. While nanomechanical sensors provide a versatile sensing platform, this concept is not limited to nanomechanical sensors but can be expanded to a variety of chemical sensors. Therefore, the presented concept of the pattern recognition-based analysis of solid materials will open a new world of chemical sensors and materials science.