As the global economy evolves, modernization and urbanization accelerate and various global problems followed. Research in many fields, such as environment, climate and healthcare, is facing increasing challenges. These problems are too complex to be systematically solved by a single discipline. In recent years, artificial intelligence (AI) technology has developed rapidly, as the core discipline of scientific research, when AI technology is applied into chemistry research, great potential will be shown in breaking disciplinary boundaries, promoting multidisciplinary integration and solving various global problems. This paper systematically reviews a number of literatures in the field of AI in the past decade, introduces some AI systems or models that have been applied to various application scenarios in the field of chemistry, such as AlphaFold, SynthReader, dZiner, and AlphaFlow, from reading, deeply learning and analyzing literature, designing synthesis routes, to conducting experiments independently. It predicts and discusses the mechanism of breaking down disciplinary barriers by building an interdisciplinary real-time fusion platform and dynamic knowledge graph, as well as the future path of AI assisted chemical researchu2014"AI+Chem" and other disciplines. It also analyzes and looks forward to the possible development direction of AI and its challenges. At present, whether it is AI+Chem or AI+Chem+multidisciplinary research is still in its infancy, preliminary research has shown that AI technology is very likely to be a powerful tool to break down disciplinary barriers and connect knowledge islands, and also provides great possibilities for solving complex global problems, and the play of tools and problem solving still need to be explored by all scientific researchers.
Controlled/living polymerization,owing to its excellent capability for regulating relative molecular mass and topological architecture,has become an important strategy for the synthesis of high-performance polymeric materials.However,such polymerization processes are typically characterized by multi-factor coupling,multiscale reaction dynamics,and highly nonlinear mappings between reaction conditions and polymer properties,which render conventional polymer design approaches reliant on empirical screening or mechanistic modeling limited by high computational cost,poor cross-system transferability,and restricted global optimization capability.In recent years,machine learning(ML)methods,leveraging multi-source data from experiments,literature reports,and computational simulations,have demonstrated strong capability in efficiently uncovering nonlinear relationships between reaction conditions,monomer/catalyst structures,and polymer properties.Even under limited data availability,ML enables rapid property prediction and inverse reaction design,providing a new paradigm for the rational design of controlled/living polymerization processes.This review summarizes recent advances in the application of ML to atom transfer radical polymerization,reversible addition-fragmentation chain transfer polymerization,ring-opening polymerization,and controlled/living polymerization,covering key aspects such as dataset construction,feature engineering,model training and optimization,as well as physical constraints and model interpretability.Furthermore,the integration of ML with high-throughput synthesis,online/in situ characterization techniques,and self-driving experimental platforms is discussed.Finally,in view of challenges including data scarcity,limited model interpretability,and insufficient automation,future perspectives are proposed,emphasizing the development of standardized polymerization databases,physically informed ML models,and self-driving experimental systems.
This paper proposes the Quaternion Hilbert Transform(QHT)and applies it to the encryption of color red,green and blue(RGB)images.Firstly,by decomposing an arbitrary quaternion signal into the linear sum of two complex signals and based on the Hilbert transform of complex signals,a Hilbert transform suitable for quaternion signals is constructed.Secondly,various commonly used properties satisfied by the Quaternion Hilbert Transform are derived.Finally,color RGB images are represented as quaternion signals,and a color RGB image encryption method based on the Quaternion Hilbert Transform is proposed.Experiments show that the proposed method has excellent visual effects and index performance,exhibits robustness against attacks such as noise and cropping,and has high sensitivity to keys,which verifies its effectiveness.
Electrocardiogram(ECG)anomaly detection plays a critical role in the early diagnosis and timely treatment of patients with heart disease.This is particularly important for medical devices such as implantable cardioverter defibrillators(ICD),where it is essential to have detection algorithms that are not only highly accurate but also consume minimal power.Traditional detection methods often face challenges in achieving optimal performance on embedded devices with limited resources,which necessitates the exploration of new and innovative solutions.In this paper,we present the design of an ECG anomaly detection device that leverages the capabilities of artificial intelligence algorithms.For this purpose,the STM32F303K8T6 development board has been selected as the platform for model design and deployment.Additionally,we have implemented an evaluation system that thoroughly tests and assesses the performance of the deployed algorithms.The study undertakes a comprehensive comparison between deep learning algorithms and traditional machine learning algorithms,culminating in the proposal of a novel classification algorithm that combines hybrid features.Experimental results highlight that this algorithm not only demonstrates superior memory efficiency and reduced latency on resource-constrained ICD devices but also exhibits higher precision and robust generalization capabilities.This research provides a novel and effective solution for ECG anomaly detection,offering significant implications for the design of medical devices and the optimization of detection algorithms.
Solution adsorption theory, as the core teaching content of surface chemistry, serves as a crucial link between fundamental chemistry and application fields such as industrial separation, environmental governance, and material preparation. The quality of its teaching is of vital importance to the cultivation of talents in chemistry, chemical engineering, environmental science, and other related fields. However, there are significant contradictions in current teaching and research. For instance, when traditional gas-solid adsorption theories (such as Langmuir's molecular layer adsorption theory and Brunauer-Emmett-Teller (BET) multi-molecule adsorption theory, etc.) are directly applied to solution systems, logical contradictions arise due to the neglect of the solvent effect, substitution effect and other special properties of solutions. Moreover, the theoretical derivation is complex and lacks coverage of practical systems such as non-uniform adsorbents and multi-component solutions, resulting in a disconnection between theory and practice. To fill the current gap in teaching resources, this paper takes statistical thermodynamics as the core theoretical basis and constructs a systematic and complete teaching system for solution adsorption theory. In teaching, it is essential to first clarify the three core assumptions. (1) The essence of solution adsorption is the monolayer adsorption of solutes at the interface, and the chemical potential of solutes is equal at equilibrium. (2) The intermolecular interaction energy is composed of van der Waals forces and hydrogen bonds, and the coordination number is directly proportional to the volume fraction of the components. (3) The surface of the non-uniform adsorbent is a collection of multiple types of active centers, and the total adsorption capacity is the linear superposition of the adsorption capacities of each center. Based on the uniform potential model (liquid phase) and the ideal potential well model (adsorption state), the derivation process of the monolayer adsorption equation for binary solutions is simplified, making the physical meaning of the equation clearer. At low concentrations, it conforms to Henry's law, and at high concentrations, it approaches the saturated adsorption capacity, which is in line with the essential characteristics of monolayer adsorption. On this basis, the teaching content of the theoretical boundary was further expanded. The adsorption equation of non-uniform adsorbents and the adsorption equation of multi-component solutions were established. To verify the validity of the theory, multi-dimensional experimental verification was carried out simultaneously. The adsorption experiments covered 7 types of adsorbates, including aniline, phenol and cyclohexanol, and 6 types of carbon material adsorbents, and combined a large number of chromatographic experiments, temperature effect studies and system verifications of adsorption of organic acids, alcohols and alkali metal ions. The results show that the theoretical calculated values are highly consistent with the experimental values, confirming that the solution adsorption is monolayer adsorption rather than the multi-molecular adsorption as traditionally understood. In addition, the universality of the theory in different adsorption systems was further verified through the adsorption of acetic acid by bone charcoal, the adsorption of alkali metal ions by silica gel, and the adsorption of n-butanol by blood charcoal, and the influence mechanisms of factors such as temperature, molecular structure, and the active center of the adsorbent on the adsorption behavior were clarified. The theoretical system of solution adsorption constructed in this paper not only provides a concise and systematic teaching tool for related majors in colleges and universities, but also helps students quickly master the core principles. At the same time, it provides scientific guidance for practical applications and demonstrates significant application value in fields such as liquid chromatography retention value prediction and adsorption separation process optimization.
