Rational design of photoactive molecules using "black box" quantum dynamics simulations

越来越多的工业、技术和医疗过程是围绕着使用吸收的光的能量来驱动化学反应或能量转移而建立的；例如，光催化剂使用吸收的光来执行原本可能不可能进行的化学反应，共轭聚合物正在被纳入新的轻型光伏设备中以将光转化为能量，以及几种光敏剂现在被批准用于光动力学疗法来治疗癌症。为了设计具有目标特性的下一代光活性分子，我们必须能够理解和合理化光诱导化学反应的机制；这就是计算机模拟可以产生变革的地方。不幸的是，光化学动力学模型是计算化学中最困难的前沿挑战之一；研究分子吸收光后所有电子和原子核的耦合运动需要高度专业化的计算机模拟方法(特别是多电子态上的量子波函数传播)，因此一直是高度专业化的专家的领域。我们最近的工作已经开始通过将机器学习策略与精确的波函数传播方法相结合来改变这一领域的格局；我们的新兴量子动力学策略现在使我们能够在几小时到几天内完成光化学动力学的模拟，而在过去二十年中流行的已建立的方法(例如基于网格的波函数在全球势能面上的传播)通常需要数月的模拟和用户时间。该提议的“大想法”是将这些新兴的量子模拟方法转化为一个真正的“黑匣子”工具，专家和非专家都可以使用它来执行精确的光化学动力学模拟。这将需要一个软件和方法开发的初始阶段，以提高我们现有方法的可用性和效率。为了进一步扩大我们的即时模拟方法的范围，我们将开发能够明确说明溶剂分子对光化学动力学的影响的策略；毕竟，最有趣的光驱动过程发生在溶液或固相中。这些方法的发展将为光化学动力学模拟的大量新应用打开大门。在这项提案中，我们确定了两个最先进的应用，一旦我们的“黑箱”量子动力学方法建立后，我们将能够解决这两个应用。首先，我们将设计新的通用荧光团标签，这些标签显示与环境相关的荧光光谱；这些标签可能会在生物成像应用中找到应用，为细胞环境提供新的见解，或者在检测水管中的污染物方面。其次，我们将设计新的光酸催化剂来活化烯烃，烯烃是原油中的一大部分，但通常商业价值较低；我们将研究一种新的光催化方法，可能能够将这些化学物质转化为更高价值的商品化学品。一个独特的转折是，量子模拟的这两个应用都将与实验合成和光谱表征同步运行。这将提供一条验证(和改进！)的途径。我们的模拟方法，也将作为一个反馈回路，以实现真正的计算机驱动的光活性分子的合理设计。总体而言，这项提议将无缝地整合计算方法开发和尖端光化学传感和催化方面的高影响应用。我们的研究团队汇聚了模拟、光谱学和合成领域的专家，在实现这些充满希望的新研究方向方面具有得天独厚的优势。

A rapidly-growing number of industrial, technological and healthcare processes are built around using the energy of absorbed light to drive chemical reactions or energy transfer; for example, photocatalysts use absorbed light to perform chemical reactions which might be otherwise impossible, conjugated polymers are being incorporated into new lightweight photovoltaic devices to convert light into energy, and several photosensitizing drugs are now approved for photodynamic therapies to treat cancers. To design the next-generation of photoactive molecules with targeted properties, it is essential that we are able to understand and rationalise the mechanism of light-induced chemical reactions; this is where computer simulations can be transformative. Unfortunately, modelling photochemical dynamics is one of the most difficult frontier challenges of computational chemistry; studying the coupled motions of all of the electrons and nuclei in a molecule after it absorbs light requires highly-specialized computer simulation methods (in particular, quantum wavefunction propagation on multiple electronic states), and so has remained the domain of highly-specialized experts.Our recent work has begun to transform the landscape of this field by combining machine-learning strategies with accurate wavefunction propagation methods; our emerging "on the fly" quantum dynamics strategy now enables us to perform simulations of photochemical dynamics in a matter of hours to days, whereas the established methodology which has prevailed during the last two decades (e.g. grid-based wavefunction propagation on global potential energy surfaces) typically requires months of simulation- and user-time.The "big idea" of this proposal is to take these new emerging quantum simulation methods and transform them into a true "black box" tool which can be used, by experts and non-experts alike, to perform accurate simulations of photochemical dynamics. This will require an initial period of software and methodology development to improve the usability and efficiency of our exisiting approach. To further increase the scope of our on-the-fly simulation methods, we will then develop strategies which can explicitly account for the influence of solvent molecules on photochemical dynamics; after all, most interesting photo-driven processes take place in solution or solid-phases.These method developments will then open the gateway to an enormous range of new applications of photochemical dynamics simulations. In this proposal, we identify two state-of-the-art applications which we will be able to address once our "black box" quantum dynamics methodology has been established. First, we will design new universal fluorophore tags which exhibit environment-dependent fluorescence spectra; these tags might find application in bioimaging applications, provide new insights into cellular environments, or in detecting contaminants in water pipes. Second, we will design new photo-acid catalysts for activating olefins, molecules which constitute a large fraction of crude oil but which generally have low commerical value; we will investigate a new photocatalysis method which might be able to transform these chemicals into much higher value commodity chemicals. In a unique twist, both of these applications of quantum simulations will run in tandem with experimental synthesis and spectroscopic characterisation. This will provide a route to validating (and improving!) our simulation methods, and will also function as a feedback loop to enable true computer-driven rational design of photoactive molecules. Overall, this proposal will seamlessly integrate computational method development and high-impact applications in cutting-edge photochemical sensing and catalysis. Bringing together experts in simulation, spectroscopy and synthesis, our research team is uniquely positioned to deliver on these promising new research directions.