Alán Aspuru-Guzik's lab was a pioneer in early research in quantum computing. Today, his work reaches further to include both chemistry and machine learning. Sitting down with MRS TV, Dr. Aspuru-Guzik shares highlights from his 2023 Plenary Session, along with his hopes and expectations for the future in science and tech.
Alán Aspuru-Guzik, University of Toronto
The Future of Materials Is Self-Driving
Written by Alison Hatt
In Monday’s plenary session, Alán Aspuru-Guzik gave an overview of his work on self-driving laboratories, in which he uses machine learning and robotics to accelerate materials discovery. He noted that 10 years ago at the MRS meeting, he and other researchers were focused on high-throughput virtual screening of materials. In a study from that time, Aspuru-Guzik used high-throughput screening to identify new materials for organic light-emitting diodes. His research team started with millions of candidates and gradually narrowed the list down to 1000 or so using machine learning and high-throughput calculations, but ultimately found that only 30 of them could be synthesized.
“It didn’t really move the needle,” said Aspuru-Guzik of the study. “We needed to actually accelerate discovery by not only calculating the materials but testing them.”
These efforts led to his vision for a self-driving lab. Instead of going serially through materials discovery process, he imagined a cycle of identifying candidate materials, synthesizing batches of them, and feeding the results back into the dataset continuously, so the machine learning tool can focus on synthesizable areas of parameter space.
But how to identify the candidate compounds? Using modern computational chemistry, we can determine the function of a material based on its structure, but a challenge for this century, according to Aspuru-Guzik, is to go the other way and determine the structure of a material to achieve a particular function.
In work published in 2018, Aspuru-Guzik and collaborators developed a tool for solving inverse design problems, using neural networks to encode a chemical structure, compress it into a bi-directional latent space, and then decompress it. In the latent space, the composition can be optimized by following gradients or doing sampling methods and, once a maximum is found, the material can be decoded to reveal the optimized structure. Many researchers are now using this generative model approach with various machine learning models, like genetic algorithms, generative adversarial networks, and, of course, large language models.
Demonstrating the power of the generative model approach, Aspuru-Guzik discussed a study in which he and collaborators developed a promising new drug candidate using artificial intelligence and generative models, going from ideation to testing in animals in just 45 days, usually a one-year process. He also shared examples from his work using a similar approach to design injectable materials for drug delivery and materials for carbon capture.
However, generative models can also dream up materials that are not synthesizable. With self-driving labs, Aspuru-Guzik hopes to better focus explorations on synthesizable materials space. As an example, he discussed efforts to discover new molecules for organic lasers, for which the screening criteria are a list of desired optical properties. Working with collaborators in British Columbia, Urbana Champaign, and Glasgow, his Toronto-based team took a modular approach, developing a library of molecular blocks and iteratively coupling them in varying configurations. The four collaborating teams all have self-driving labs, all of them slightly different, that can do synthesis, identification, and optical characterization of organic compounds. As a “hello world” experiment, the team produced 40 new organic laser compounds in just one weekend, several of them better than the molecule they started from.
More recently, the team demonstrated running the four geographically dispersed labs in parallel on a single discovery effort, sharing data and shipping materials between them. To achieve this, the researchers developed an asynchronous Bayesian optimization tool. They calculated 190,000 molecules with DFT, creating a model which they then calibrated experimentally by synthesizing and characterizing a hypercube sample of 500 compounds. Then, using their Bayesian optimization tool, the labs closed in on the most promising areas of materials space and ultimately identified 21 new compounds, now the brightest organic molecules in the world.
Looking forward, Aspuru-Guzik asked, what will the laboratory of the future look like? Anticipating a laboratory operated by independent robots, his group is training robotic arms to recognize transparent objects, which are particularly hard for robots to do but is an essential skill for working in a messy chemistry lab. They are also developing a computer program named Organa that integrates ChatGPT-4 technology and can plan and execute chemistry experiments given minimal input in natural language.
Aspuru-Guzik closed with a vision that computers can be agents of understanding, not just computational workhorses, a once-fantastical future that now seems almost within reach.
The Kavli Foundation is dedicated to advancing science for the benefit of humanity, promoting public understanding of scientific research and supporting scientists and their work.
We start day 2 with the plenary session titled “The Future of Materials is Self-Driving” by Alan Aspuru-Guzik.
Upon entering the ballroom on the 2nd floor of the Sheraton, we were greeted with a beautiful blue backdrop accompanied by the sparkling crystal illuminated by a gentle lighting on the ceiling. The attendance amazing with full attendance.
Having an interview Saabir Petker, a 1st Year PhD student working on using computational methods to simulate and predict barocaloric properties in polymers, I gained some insight into how people from different backgrounds benefitted from the talk. “It was very interesting and inspired me to work hard in my research, especially thinking about how I can adapt the sampling methods for relevant materials to my work. It gave me a lot to think about in terms of human input incorporated with AI”, he said.
It was great listening to Prof. Alán Aspuru-Guzik at the first plenary talk today, where he talked about how the field of chemistry could be automated using robots. One of the things Prof. Aspuru-Guzik talked about towards the end is about how one can "talk" an experiment to the AI robot in his lab, where the current capabilities expand to making margaritas!!! Since making Cocktails is quite similar for a robotic arm to identify, pick and mix solutions from different beakers, it is quite possible for the automated robotics to also prepare margaritas on command. Imagine a group social where your friendly neighborhood robot chemist is making margaritas for everyone!
Connect with Professor Alán Aspuru-Guzik from University of Toronto at MRS Fall 2023 to see if the self-driving chemistry robots could make your favorite cocktails!
Sir Konstantin ‘Kostya’ Novoselov, National University of Singapore & The University of Manchester
Materials for the Future
Written by Cullen Walsh
In his keynote talk, “Materials for the Future,” Konstantin Novoselov from the National University of Singapore and the University of Manchester explored what materials of the future will look like and how we will go about producing them. He started his talk by discussing graphene, which he successfully isolated in 2004, resulting in the 2010 Nobel Prize in physics. Since then, graphene has become commonplace in commercial applications, being used in Ford cars for noise cancelation and in Huawei phones for device cooling. To facilitate these new technologies, numerous advances have occurred in the production of graphene. For instance, we can now produce high-quality graphene at production scale using techniques like chemical vapor deposition, in which gas particles are condensed into a solid, and liquid phase exfoliation.
After reviewing these advances in graphene production, the remainder of Novoselov’s talk focused on the potential of other two-dimensional (2D) materials beyond graphene. There are now dozens of these 2D crystals that range from semiconductors to insulators to superconductors. By stacking these 2D materials, we can create customizable structures with unique properties which we call van der Waals heterostructures (due to the van der Waals forces holding the layers together). We can then further customize these heterostructures by changing the relative orientation of the stacked layers, resulting in dramatic changes in the physics and chemistry of the structure. For instance, if we rotate the topmost layer of bilayer hexagonal boron nitride (hBN) by 180 degrees, we get a ferroelectric material due to a change in the dipole moment between the stacked boron and nitrogen atoms.
Now that we can create these layered materials systems on demand and control their properties, what’s next? To answer this question, Novoselov proposed a new paradigm for creating these materials systems that involves “bottom-up functionality at the material level.” Current technologies are typically top-down, meaning the components are not functional until they are assembled into a system. As a result, complexity typically comes from the top-level. This is in stark contrast to biological systems, where functionality is spread across all scales, from proteins to cells to organs. What Novoselov and his group are now researching is whether we can produce materials similarly so that we can create technologies in which the composite materials provide the functionality.
Architecting this complexity is difficult and “I don’t have the solution,” said Novoselov. However, he outlined some principles of materials design to help achieve this goal. These included creating systems with a degenerate energy landscape, similar to that of proteins, that can allow for easy conformational or property changes. To explore these complex energy landscapes, Novoselov and his collaborators have begun using custom robotic systems and artificial intelligence to predict the non-equilibrium dynamics of novel materials systems. For instance, he highlighted predictive studies being performed on smart nanocontainers that open and close based on the pH of the environment.
Overall, Novoselov emphasized the need to think about how to design functional and intelligent materials out of equilibrium for new applications. This will allow us to take materials of different dimensionalities (from 0D to 3D) and assemble them into systems that can demonstrate new properties. Using this bottom-up design approach, we could one day revolutionize two-dimensional materials technologies.
The Kavli Foundation is dedicated to advancing science for the benefit of humanity, promoting public understanding of scientific research and supporting scientists and their work.
Jenny Nelson, Imperial College London
Optimizing Solar Energy Conversion in Molecular Electronic Materials
Written by Rahul Rao
Today, the solar energy world is ruled by crystalline silicon. But recent advances in materials have spawned a host of alternative materials seeking to challenge silicon’s dominance. One class of challengers are molecular electronic materials: carbon-based organic semiconductors that can be excited by visible light.
At Monday’s Plenary Session, Jenny Nelson spoke about the titular materials, how they’ve rapidly developed, and how her group’s work has contributed to bring these organic materials closer to viable solar cells.
Just several decades ago, these materials barely registered on solar researchers’ radars. Most of them had been fashioned from fullerenes, limiting their variety. Additionally, such materials tend to be disordered and suffer from a poor dielectric constant, hurting their prowess as semiconductors.
But by the 1990s, researchers had learned how to circumvent those restrictions by interfacing two different disordered materials—one to donate electrons, another to accept them. Since then, the materials themselves have diversified from fullerenes into a wide range of other materials that researchers could easily process and readily customize. As a bonus, researchers have found that their coupled molecular electronic materials’ properties are easy to computationally predict.
The result: solar cells made from molecular electronic materials have blossomed in efficiency, from less than 2.5% in 2002 to 11% in 2016 to over 19% in 2022.
Still, any organic solar cells have a long way to go before they can match their crystalline silicon counterparts, many of which now clear the 30% mark. Nelson said that researchers need to quantify and understand the loss in materials.
One way of doing that is to take a solar cell and run it in reverse, like an LED—pushing electrons in and watching the light that comes out. Because a physical relationship exists between that emitted light and the light absorbed when the device is operating as designed, researchers can study how close their device’s efficiency is to its ideal limit.
Through this method, by 2015, Nelson’s group had helped demonstrate that organic solar cells shed much of their energy through non-radiative loss—something that researchers could address by improving their materials. Recent efforts to do just that have helped drive much of the rapid efficiency jumps these solar cells experienced from the late 2010s.
Their efficiency could rise even farther if researchers can understand which molecular properties lead to which changes in a solar cell’s performance. Nelson’s group probed this computationally, outfitting a model of a solar cell device with a molecular model. They found that varying just four molecular parameters controlled a device’s voltage, current density, and fill factor.
And molecular electronic materials may not need separate donor and acceptor components forever. A smattering of recent research has suggested that single-molecule materials—for instance, polymers built with donor and acceptor subunits—have already reached at least 11% efficiency. Rearranging and reshaping polymers might boost that even further.
Finally, Nelson mentioned that optimal solar cell materials already exist in nature: in the photosystems of photosynthesizing plants. If materials scientists studied them, she said, their findings could light possible paths into a bright future of bountiful organic solar cells.
The Kavli Foundation is dedicated to advancing science for the benefit of humanity, promoting public understanding of scientific research and supporting scientists and their work.
Today is the last day of the 2022 MRS Spring Meeting & Exhibit. A great benefit of the hybrid design is that some of the talks are available on the virtual platform through June 30, 2022. This includes the Plenary and Featured Talks - which means all of the presentations given in Symposium X: Frontiers of Materials Research!
As a preview to the Symposium X speakers, here are two of the interviews done with MRS-TV.
"From Atom to System—Tera-Scale Energy Transition with Better Batteries” – Symposium X speaker Y. Shirley Meng
The Light Stuff: Enabling Sustainable, Product-Selective Photocatalysts with Plasmonics – Symposium X speaker Jennifer Dionne of Stanford University
Blogger: Judy Meiksin
Sossina Haile of Northwestern University discusses her Plenary address to the 2022 MRS Spring Meeting, titled “Vignettes in Solid State Electrochemistry for Sustainable Energy Technologies.” Haile's presentation can be viewed online through June 30, 2022.
Sossina Haile, Northwestern University
Vignettes in Solid State Electrochemistry for Sustainable Energy Technologies
Written by Don Monroe
Warming and precipitation changes will hit equatorial regions the hardest, said Sossina Haile of Northwestern University, and developed countries have the responsibility and resources to explore technologies to deal with the changes they are largely responsible for. Her group looks for overlaps between energy technologies, fundamental materials properties, and fabrication of precise structures to get meaningful measurements. Haile presented four examples of solid-state phenomena at successively higher temperatures, which draw on understanding of materials chemistry, thermochemistry, and electrochemistry.
Her first vignette concerned the use of superprotonic conductors at “warm” temperatures. Specifically, the electrical conductivity of cesium dihydrogen phosphate jumps up by three or four orders of magnitude when heated above about 250°C, reflecting rapid reorientation of proton-bearing phosphate groups and the transfer of the proton between neighboring phosphates. One challenge with these materials is that they are water soluble, so they need to be stabilized using a high-humidity environment.
Haile’s research team exploited the proton conduction of this solid acid as the central layer of fuel cells that generate electricity directly from the reaction of hydrogen with oxygen. In this device every electron brings a hydrogen atom, she said, meaning a 100% faradaic efficiency. Adding a reforming catalyst at the anode side creates a cell that uses alcohols as the starting source for hydrogen. There may even be ways to use liquid ammonia as a hydrogen source. “Superprotonic conductors are pure proton conductors,” she said, and “have lots of intriguing technological possibilities.”
Haile’s second, “hot” vignette involved protonic ceramic electrolytes, such as doped barium zirconate, a perovskite. She described a reversible cell, which can act either as a fuel cell or for electrolysis, similar to a storage battery. The conductivity of this material is good at temperatures around 500°C.
The proton conducting material avoids many of the downsides seen in oxygen-conducting systems, she said, including degradation of materials and dilution of the fuel source. However, current materials are hampered by some electronic conduction.
In her third example, Haile described studies of ceria-zirconia materials, which are commonly used in catalysis at “hotter” temperatures. Although the oxidation state of zirconium is largely fixed, that of cerium can be 3+ or 4+. The researchers used glancing-angle x-ray absorption near-edge structure (XANES) to find that the 3+ state of cerium is much more prominent at the surface. Somewhat surprisingly, this result was largely independent of the zirconium concentration and of the surface orientation. “The relationship between the Ce3+ on the surface and catalytic activity remains a mystery,” she said.
The final example involves not electrochemistry but thermochemistry, thermally cycling non-stoichiometric perovskites to produce hydrogen. At “fiery” temperatures as high as 1500°C, entropy drives thermal reduction and oxygen release from the materials. Upon quenching to a lower temperature, perhaps 800°C, the material reacts with steam and CO2 to form H2 and CO.
The second process is driven primarily by enthalpy. This driver led Haile and her team to explore materials such as “STM55,” SrTi0.5Mn0.5O3. To achieving the right balance between reduction and oxidation to improve the production rate per cycle, she said, “It’s really important to recognize that the thermodynamics govern the fuel production rate. The whole thing is about thermodynamics.”
Still, “[we’re] waiting for the mechanical engineers to design the reactors so we can design the materials, and the mechanical engineers are saying ‘What can your material do for us, so we can design the reactor,’” she said. “It’s all a bit of a Catch-22.”
The Kavli Foundation is dedicated to advancing science for the benefit of humanity, promoting public understanding of scientific research and supporting scientists and their work.
Sir Kostya Novoselov, The University of Manchester
Materials for the Future
Written by Sophia Chen
Two-dimensional structures like graphene and hexagonal boron nitride make a versatile palette for designing next-generation materials, physicist Kostya Novoselov said in Wednesday’s keynote, titled, “Materials for the Future.”
Novoselov, now at the National University of Singapore, won the 2010 Nobel Prize in Physics for his seminal experiments characterizing the two-dimensional (2D) material graphene. Since then, graphene’s profile has risen among commercial products, both because of its desirable properties and because of its ease of production. Graphene can be found under the hood of all Ford cars, said Novoselov, where they serve noise absorption purposes. Montreal-based company ORA has made headphones with graphene for its stiffness and lightness. Huawei uses graphene films to cool their phones. Graphene is also used as an electrical resistance standard, he said.
Researchers are now pursuing new materials from stacking layers of 2D structures—not just graphene, but also boron nitride, molybdenum diselenide, and tungsten disulfide. This class of materials are broadly known as van der Waals heterostructures, as each atomically thin layer is bonded to the next via van der Waals forces. Researchers have already created LEDs with this type of structure.
A heterostructure’s properties derive from the strong interaction of many electrons between different layers. Thus, researchers can tune a heterostructure’s electronic properties by changing the order and number of layers. Researchers have found surprising materials properties from rotating one layer of atoms with respect to the other slightly. In one 2018 study, researchers created a superconductor by stacking two layers of graphene and rotating one layer by about 1.1 degrees.
Novoselov also discussed several promising future applications for two-dimensional materials. They could be used to design adaptive intelligent materials that respond to external triggers. One example might be a capsule that can open and close to deliver medicine inside the body depending on environmental conditions. Two-dimensional materials could also be used as components in a new type of computer known as neuromorphic computing, he said.
The use of two-dimensional structures could simplify the materials for our electronics. Novoselov pointed out that modern silicon-based devices use a dizzying number of elements ranging from semiconductors to metals to rare earth elements—more than those found in the human body. The increase in material complexity began around 1990, when manufacturers strived to cram more transistors on chips. Two-dimensional materials could potentially reduce the need for rare elements, as they can be oriented and stacked in many configurations to achieve a range of properties.