Nature Materials访谈：伦敦 Thomas Young 中心材料理论模拟

材料牛注：材料建模对于材料科学的理论和实验发展非常重要，基于此有许多科学家和机构研究材料的计算与理论模拟，如：伦敦的Thomas Young 中心。他们试图通过计算和模拟材料中的电子运动来对材料的性能进行预测、调控，为先进新材料的制备以及其相关学科的发展提供理论支撑。

采访者（Q）：Nature Materials

访谈对象（A）：Angelos Michaelides ，伦敦大学学院理论化学教授，Thomas Young 中心联合主任

图为Angelos Michaelides

Q：你为何对材料建模感兴趣？

How did you become interested in modelling?

A：在 Queen’s University Belfast 读本科的时候就被建模所吸引，在原子尺度上得到清楚的信息，可以以一个非常明确的方式来测试事物，我喜欢这样的想法。理论和模拟吸引着我用最简单的方法来得到信息。

I was attracted by modelling during my undergraduate days at Queen’s University Belfast. I liked the idea of getting very clean information at the atomic scale and being able to test things in a very well-defined manner.Theory and simulation appeared to me to be the easiest way to access this information.

Q：你现在致力于研究什么？

What are you currently working on?

A：我和我的团队目前致力于研究材料建模方面的一系列项目，这包含很多问题。如对于DNA里面氢键的理解，表面分子参考资料数据的确定，高压材料。目前，我们有个欧洲研究委员会资助的项目，这个项目的主要目的是研究冰在分子尺度的形成及其杂质在这个过程中的作用。

超净水很难结冰，在0摄氏度以下很容易形成过冷状态。0摄氏度是热力学上相转变温度，通常情况下水在0摄氏度下会结冰，这是因为杂质在水的结冰过程中起到了催化剂的作用，降低了形核的能垒。目前，没有理论可以预测材料在成核过程中是好还是坏。我们利用模拟的办法来解决这一问题，筛选不同种类的材料，在基底上进行模型研究。发展这一项基础研究，在理论上可以使我们控制冰的形成，更好的理解云的形成。

My team and I are working on a range of projects in materials modelling, covering different questions, from the understanding of hydrogen bonding in DNA, to the determination of accurate reference data for molecules on surfaces, to high-pressure materials. At the moment, we also have a project funded by the European Research Council aimed at understanding ice formation at the molecular scale, and the role that impurities play in this process.

Ultraclean water is surprisingly difficult to freeze and can easily be supercooled to well below 0 °C. The reason that water typically freezes at 0 °C, the thermodynamic phasetransition temperature, is the presence of impurities acting as catalysts, which reduce the free-energy barrier of nucleation. At the moment, there is no theory able to predict which materials are good or bad at nucleating ice.

We are working on using simulation techniques to address this by,for example, screening different types of materials as well as doing model studieson generic substrates. Developing this fundamental insight could in principle also allow us to control ice formation, and better understand cloud formation.

Q：你和实验小组合作吗？

Do you collaborate with experimental groups?

A：当然，我们很大一部分的论文是和实验者合作完成的。这种合作在理解问题上互利互惠，例如，界面上的水的问题。水分子之间相互作用和环境之间的作业都很弱，通过调节氢键和范德华力之间的平衡，可以形成非常复杂的相图，相图上大约有17种相。这种丰富的行为意味着即使是单原子组成的一个平面我们很难提前知道水会形成什么样的结构。因此，实验者的对我们的指导非常有利，他们也经常揭示一些有趣或者令人惊讶的结构。我们的观点和理论对实验者来说也很有用，例如，通常最好的扫描探针测试手段都无法探测表面吸附层完全溶解的结构。

Yes, a large fraction of our papers are joint works with experimentalists. These collaborations are of great mutual benefit in understanding, for example, water at interfaces. Water molecules interact weakly with each other and their environment through a balance of hydrogen bonding andvan der Waals dispersion forces resulting in an extraordinary complex phase diagram, for example, with about 17 different ice phases. This rich behaviour means thateven on an atomically flat surface it is difficult to know in advance what kind of structures water will form.

Guidance from experimentalists is therefore of huge benefit to us, and it also often reveals interesting or surprising structures. We’re also of use to the experimentalists, for example, even the very best scanning probe measurements struggle in general to fully resolve structures of adsorbed layers on surfaces。

Q：什么是Thomas Young 中心，你在里面起着什么样的作用？

What is the TYC, and what is your role in it?

A：Thomas Young 中心是一个以伦敦为基础的研究中心，主要研究材料的理论和模拟。这个中心是将我工作的伦敦大学学院、伦敦大学玛丽皇后学院、伦敦国王学院、帝国理工学院联合起来的一个组织，实际上它是一个民间组织，今年是它组建十周年，它致力于让伦敦成为世界上材料理论计算和模拟最好的地方。利用在伦敦的一大批杰出的研究者，希望通过合作而不是竞争来达到这个目标。

The TYC is a London-based centre for the theory and simulation of materials. It brings together researchers from UCL, where I am based, King’s College London, Imperial College and Queen Mary University of London. It is essentially a grass-roots organization, now in its tenth year, that aims to make London one of the best places in the world to do theory and simulation of materials. We hope that by collaborating, rather than competing, we can achieve this, and in so doing, exploit the critical mass of outstanding researchers in London.

这个研究中心有着公平的管理团队，团队是由分别来自各个大学的联合主任组成的。Alessandro De Vita 是伦敦国王学院的联合主任，Arash Mostofi 是帝国理工学院方面的负责人，Martin Dove 是伦敦大学玛丽皇后学院方面的负责人，Alex Shluger和我是伦敦大学学院联合主任。Thomas Young 中心所有的管理结构有来自每一个成员大学的人，这点很赞！我们决定的策略和目标对于每一个成员大学都是公平的，杜绝不同大学之间不健康的竞争。很幸运的是，我们得到了很多极好的行政支援。

The TYC has a fairly flat management team formed of co-directors from the member universities. Alessandro De Vita is co-director at King’s College London, Arash Mostofi leads the work at Imperial College, and Martin Dove at Queen Mary. Alex Shluger and myself are co-directors at UCL. One of the nice things about the TYC is that all colleges are represented in this management structure. We all decide the strategies and goals for the centre on an equal basis and there is no (unhealthy) rivalry between the different colleges. We are also fortunate to have some excellent administrative support

Q：Thomas Young 中心实际上是怎么运行的？

And how does the TYC work in practice?

A：在日复一日的工作基础上，Thomas Young 中心致力于进行世界上一流的科研活动，以此来使我们的每一个成员大学，特别是我们的学生受益。我们注入了很多的精力与财力进去，给学生提供材料理论与模拟方面优质的教育和尽可能好的学术环境。他们在所有的大学里享受浓郁的学术气息，我们开设了Thomas Young 硕士课堂，在这个课堂里面讲课的老师都是世界上著名的研究者，这些老师会在这个课堂里面讲授他们所擅长的领域知识。在这片学术的沃土之上，很容易见到年轻的学者，这片沃土也发展了很多新的研究方法和研究计划。确实，在伦敦，Thomas Young 中心在培养不同领域的相互合作方面很有用处。

On a day-to-day basis, the TYC aims to run top-class events and activities for the benefit of all our members and in particular our students. We put a lot of value into giving our students an outstanding education in theory and simulation of materials and in providing the best possible intellectual environment for them. They get exposed to academics across all the colleges, and amongst other things we run the so-called TYC Master Classes, where world-leading academics fly in to London to teach specialist classes on their particular area of expertise. In this fertile environment young academics can meet people easily and develop new research ideas and projects. Indeed the TYC has been very useful in fostering collaborations between different groups across London.

Thomas Young 中心另外的一个核心作用就是提供了连贯性的方式，与工业生产紧密结合起来。许多有趣的问题需要从建模来入手解决，但是一些工业家不知道哪个学者能够解决这样的问题。Thomas Young 中心有大约100个研究团队组成，在材料的现在理论和模拟方面非常擅长，所以，在实质上我们可以实现一站式的服务，促进创造工业家与我们学者之间的合作伙伴关系。与此相类似的，我们可以同大的跨国公司和组织进行高度战略性的相互交流与合作，我们Thomas Young 中心在这方面做的很成功，例如同BP公司，Rolls-Royce公司和美国的很多国家实验室都有很好的合作伙伴关系。

Another core aspect of the TYC is that we provide a coherent way to engage with industries. There are many interesting problems that require input from modelling, but industrialists do not always know who the right academic to approach is. The TYC is formed of about 100 research groups with expertise in almost all areas of modern theory and simulation of materials. So, in essence we can act as a ‘one-stop shop’and facilitate the creation of partnerships between industry and our academics. Similarly, because we are a large entity, we can engage strategically at a high level with large international companies and organizations. The TYC has successfully done this, for example with BP, Rolls-Royce and several national labs in the US.

Q：你认为Thomas Young 中心会成功吗？

Do you consider the TYC to be successful?

A：当然会！我们已经做成了许多一开始没有预料到的事情，这十年的发展是表明我们做的事情是正确的，我们是有用的也是有需要的。但是，当然我们可以做得更好。特别地，我非常想看到Thomas Young 中心能够工业的项目中进行的更加深入，在英国能够作为一个杰出的区域中心被外界所认可。我们这个团体里面的研究者是非常有热情、有干劲的，为了中心的成功，我们希望大学能够提供更诱人的环境来吸引最杰出的青年研究学者来到伦敦。

中心的研究者经费很大程度还是来自国家研究委员会，欧洲计划和其他的基金组织。这些组织都认为材料建模是现代科学研究必不可少的一个部分，伦敦是世界上研究材料建模最好的地方之一。我们还很大程度上依赖高性能的计算，而大学和研究委员会需要在本地区的水平上继续投资。如果在英国，特别是伦敦，仍然想在理论和模型方面保持着竞争力，那么对于高性能计算的投资是很重要的。如果没有合适的投资的话，无疑国家流失最好的建模研究者，而这会使国家进行更大的投资。

Yes, and we have managed to do much more than anticipated at the outset. Simply having lasted for ten years is a sign that we are doingsomething right, that we are useful and that we are needed. But, of course, we can always do better. In particular I would like to see the TYC involved in more industrial projects and to get more recognition as a regional centre of excellence in the UK. Our community of researchers is enthusiastic, active and engaged, and for the success of the TYC we need to ensure that our universities provide an environment that attracts the best and the brightest young researchers to London.

Our members also rely heavily on support from the national research councils, European programmes and other funding agencies; such organizations are increasingly recognising that materials modelling is an indispensable part of modern scientific research and that London is one of the very best places internationally for this sort of research. We also rely very heavily on high-performance computing (HPC), and this requires continued investment from our universities at a local and regional level, and from the research councils. Investment in HPC is critical if the UK in general, and London in particular, are to remain competitive in theory and simulation; without appropriate investment there is no doubt that the country will lose more of our best modelling researchers to countries making greater investments.

Q：回到材料建模，它当前的局限性和挑战是什么？

Going back to materials modelling, what are its limitations and current challenges?

A：采用更加精确和更加计算有效的方法，我们通常的目的在于模拟更大更加真实的体系，探索长时间跨度的过程事件。20年前，你也许已经得到了一个类似的答案，但是由于理论和模型软件的发展，我们的能力得到显著的提高。显然，计算能力的提高对我们收益匪浅。试想一下我们运用密度泛函理论来进行电子结构的计算，我们总是在进步，现在我们使用所谓的线性缩放密度泛函理论使得计算成百上千甚至是上百万的原子的电子结构成为了可能。

通常情况下，线性缩放密度泛函理论的应用还是很少的，而密度泛函理论只能计算成百、成千的原子电子结构。例如，我们可以模拟，在一个平面很小的一部分上大约有100个水分子，或者在一个有着很明确的缺陷的表面。这是在真实物理系统里面很明显的一个简化，但是它抓住了足够的物理特征，给人们提供了很有用的信息。在这样的体系里面，可以进行一个短的从头算来进行分子动力学模拟，或许可以达到100 ps。偶然情况下，这足够用来获得关于这个体系特征的有用信息，但是，如果一个人对于模拟化学过程和晶体在水里的溶解很感兴趣的话，例如晶体在水里的溶解，则提高采样的技术是有必要的。当然，比系统尺寸大几个数量级的经典分子的动力学过程是可以看到且这个过程最长可达到毫秒级别。

将经典和量子方法结合起来的，就是所谓的QM/MM 技术。在未来，一个有趣的方面是它的应用和经典分子潜力的发展，这种潜能可以通过机器学习从头算数据来训练。在牺牲经典分子潜能的代价下，这些机器学习潜力提供量子准确性。

We generally aim to model larger and more realistic systems and to explore processes that happen on longer timescales, with more accurate and more computationally efficient approaches. Twenty years ago, you would have gotten a similar answer but our capability has increased dramatically, primarily due to the development of improved theories and simulation software. Obviously, improvements in computational capacity have also benefited us greatly. Advances are always being made, but let’s assume that we are talking about an electronic structure calculation with density functional theory (DFT). It is now possible to use so-called linear scaling DFT methods to do calculations of hundreds of thousands or even millions of atoms.

In general though, linear scaling DFT is still quite rare and DFT calculations on a routine basis typically involve a few hundreds or thousands of atoms. For instance, we can model a few hundred water molecules on a small segment of a flat surface or a surface with some very well-defined defects. This is obviously a major simplification of the real physical system, but the hope is that it captures enough of the physics to provide useful insight. With such a system one can run short ab initio moleculardynamics simulations for perhaps up to 100 ps. Occasionally, this can be sufficient to obtain useful information about the system of interest, but if one is interested in modelling chemical processes, for instance how a crystal might dissolve into water, thenenhanced sampling techniques are needed. With classical molecular dynamics it is, of course, possible to look at system sizes several orders of magnitude larger and to run dynamics up to milliseconds.

There are also hybrid techniques that combine classical and quantum approaches, socalled QM/MM techniques. An interesting area for the future is the application and development of classical potentials that have been trained through, for example, machine learning on ab initio data. These machine learning potentials offer the promise of quantum accuracy at the cost of a classical potential.

Q：这些局限性如何影响你们自己的研究？

How do these limitations affect your own research?

A：为了研究材料的电子性能，密度泛函理论是目前研究的重点。原则上它包含近似，这就意味着密度泛函理论无法解释范德华色散力。事实证明这些力对于水分子之间、水分子与界面的键合非常重要，许多研究小组对此做出了令人激动的发展，这些发展在某种程度上克服了范德华力无法应用密度泛函理论的问题。然而，在对待不同相的水使用密度泛函理论计算分析时，精确性依然不够好。

在经典分子模拟动力学基础上，我们处于仅仅能预测多相的冰形成速率的阶段。然而，在这个模拟过程中精细的电子结构和极化效应都没有考虑进来。我们在从头算的精度水平计算成核速率（利用精确的从头算理论），对待任何一个绝对成核速率的计算都要万分小心。目前，如果我们想解释一个特定的矿物在成核后性能好还是坏，就需要一个完整的博士学位。越准确的电子结构模拟，精细的机器学习潜力，再加上更加有效的采样技术，一个博士生也许在几年的时间内就能筛选出成百上千种材料，这将极大的增加我们探索有用而且有预测性的理论的机会。

The workhorse technique that we have in order to look at the electronic properties of materials is DFT. Although exact in principle, in practice it contains approximations which mean that traditionally DFT does not account for van der Waals dispersion forces. It turns out that these forces are particularly important to the binding between water molecules and between water molecules and surfaces. Various groups have made exciting developments that have allowed for the problem of van der Waals dispersion forces within DFT to be overcome to some extent. However, the accuracy of DFT for treating water in all its various phases and at surfaces is still not as good as we would like.

We are now just about at the stage where we can make predictions about heterogeneous ice formation rates on the basis of simulations with classical molecular dynamics. However, subtle electronic structure and polarization effects are not taken into account in such simulations. Untilwe can compute rates at an ab initio level (and with an accurate ab initio theory at that), we must treat any predictions of absolute nucleation rates with the utmost caution.

At the moment, if we want to explain why a specific mineral is good or bad at nucleating ice, it basically takes an entire PhD. With more accurate models of the electronic structure coupled with sophisticated machine learning potentials and more efficientenhanced sampling techniques, maybe the PhD student that comes along in a few years will be able to screen thousands of materials — this would certainly greatly increase our chances of developing a theory that is really predictive and useful.

Q：正如你前面所说的，Thomas Young 中心与工业有着紧密的联系，你感觉到材料建模对工业生产的影响了吗？

As you have explained before, the TYC actively engages with industry. Do you feel that materials modelling has had industrial impact?

A：新的证据表明，材料建模为工业的发展提供了助推剂。在许多领域都有这样的例子，多相催化就是其中一个例子。在这个领域里，材料建模（主要是密度泛函理论）被用于发展基础研究，这样的研究能够用来预测不同种类的反应催化剂、发开新合金来提高催化性能。还有很多工作与材料的发现相关，人们希望辨别和筛选出具有理想功能的材料。例如，这样的方法使得电池行业取得了许多进展，这些进展都已运用到了具体的生产实践当中了。Stefano Curtarolo 和他的同事发表了一篇很有趣的综述（Nature Mater.12, 191-201; 2013），这篇综述讲的是如何利用高通量的计算来影响不同领域的科学发展，还有人已经写了一个高通量的计算如何影响工业生产的文章。

There is now clear evidence that materials modelling is providing real insight leading to improvements in industry. There are examples in many fields, heterogeneous catalysis is one of them. In that field, materials modelling (mainly DFT) was used to develop the fundamental insight that allows quantitative predictions to be made about the reactivity of different catalysts, and this insight has since been used to develop new alloys that have improved catalytic performance. There is also a lot of work associated with materials discovery in which people hope to identify and screen materials with desired functionalities.

This approach has led, for example, to some improvements in battery technology that are being exploited in industry now. Stefano Curtarolo and colleagues, for example, wrote an interesting Review Article (Nature Mater. 12, 191–201; 2013) on how high-throughput computation has had an impact in different areas of science, and others have also written about this within an industrial context.

Q：在你的脑海里，有没有想过你研究的领域会影响未来的发展？

Do you have in mind any industry or field where you think your research may have an impact in the future?

A：BP公司已经资助我们在腐蚀和成核方面的研究工作，我认为他们已经深刻的意识到模拟的重要性，这将给他们带来巨大的价值。我目前的是更好的理解和控制冰的形成，这个将在许多领域有着潜在应用，如航空、食品加工、低温贮藏。我希望这项工作能够最终应用到上述的每一个领域里。在气候和大气科学领域，我希望我的建模可以有更大的影响，这就是气候的模拟。当前，大气科学家在宏观领域的研究收集的数据和表面科学家对于表面水原子结构的理解之间还存在着差距。我认为建模可以充当一个桥梁的作用，总体来说我认为这个领域的建模还有待开发。理解大气冰的成核是很重要的，因为这会影响全球云量和类型，本质上来说就是影响冰与水的相对含量，与这相关的是热量的反射与保存，这会影响地球的温度。

BP has funded some of our work on corrosion and nucleation, and I think already they see value in the fundamental understanding that our simulations have provided. My current research into a better understanding and subsequent control of ice formation could see application in several sectors, ranging from the airline to the food industries to cryopreservation. I hope that this work will eventually lead to useful insight that can be applied in each of these areas. Another very relevant area where I hope our modelling will have more impact is in climate and atmospheric sciences, and subsequently in climate modelling.

At the moment there is a serious gap between the extremely valuable but largely macroscopic data collected in field studies by atmospheric scientists and the atomistic insight obtained by surface scientists on the structure of water at interfaces. I think modelling can act as a bridge, and in general I think modelling is still underexploited in this area. Understanding atmospheric ice nucleation is important because it impacts on the amount of cloud cover across the globe and on the type of clouds; basically the fraction of water relative to ice. This is relevant to the amount of heat that is reflected out or kept in, and therefore to the temperature of the planet.

原文参考地址：Materials modelling in London

素材：朱德杰  编译：朱德杰

欢迎加入材料人科普团队，一起进行材料顶刊学术动态的跟踪和报道以及SCI相关知识科普等。加入方式：（1）加入材料人文献检索群（410109144），私信管理员“成都-小小（QQ: 763044722）”报名；（2）报名请戳我。

欢迎各大课题组到材料人宣传科技成果并对文献进行深入解读，投稿邮箱tougao@cailiaoren.com。