As quantum mechanics enters its second century, theoretical and computational chemistry stands at a pivotal transition. Traditional orbital-based approaches such as valence bond theory and molecular orbital theory and density-based frameworks like density functional theory have long provided the computational and conceptual foundations for the field. However, the advent of machine learning and quantum computers introduces completely new paradigms for representation, inference, and understanding. In this perspective, we examine how chemical understanding has been evolving in the past century through the lenses of ontology, epistemology, and emergence. We argue that chemical concepts, such as aromaticity, electronegativity, reactivity, and stereoselectivity, are not merely reducible to basic laws of physics but emerge as essential scaffolds linking chemical theories to chemical understanding. We propose a general scheme to obtain chemical understanding from the basic variables of chemical theories. Extending this scheme to deep learning and quantum computing, we suggest roadmaps to harvest chemical understanding from them and then advocate for hierarchical modeling as a new platform that moves beyond the constraints of multiscale modeling. Hierarchical modeling integrates abstraction across scales, captures emergent behaviors, and enables conceptual innovation for hierarchical systems. We conclude that the future of chemical understanding depends less on solving harder physical equations alone and more on epistemological shift characterized by conceptual pluralism, epistemic adaptability, and deeper appreciation of the multilayered ontological structure inherent to molecular systems.
Shubin Liu (Thu,) studied this question.