One of the grand challenges in the physical sciences is to ‘design’ a material before it is ever synthesized. There has been fast progress in predicting new solid-state compounds with the help of quantum-mechanical computations and supervised machine learning, and yet such progress has largely been limited to materials with ordered crystal structures. In this Perspective, we argue that the computational design of entirely non-crystalline, amorphous solids is an emerging and rewarding frontier in materials research. We show how recent advances in computational modelling and artificial intelligence can provide the previously missing links among atomic-scale structure, microscopic properties and macroscopic functionality of amorphous solids. Accordingly, we argue that the combination of physics-based modelling and artificial intelligence is now bringing amorphous functional materials ‘by design’ within reach. We discuss new implications for laboratory synthesis, and we outline our vision for the development of the field in the years ahead.

物理科学的一大挑战是在材料合成之前“设计”材料。在量子力学计算和监督机器学习的帮助下，预测新的固态化合物取得了快速进展，但这种进展在很大程度上仅限于具有有序晶体结构的材料。在这个观点中，我们认为完全非结晶、无定形固体的计算设计是材料研究中一个新兴且有益的前沿领域。我们展示了计算建模和人工智能的最新进展如何提供非晶固体的原子尺度结构、微观特性和宏观功能之间以前缺失的环节。因此，我们认为，基于物理的建模和人工智能的结合现在使非晶功能材料“设计”触手可及。我们讨论了实验室合成的新影响，并概述了我们对未来几年该领域发展的愿景。