4

Emerging Areas in the Chemical Sciences

Success in the chemical sciences relies on our ability to observe, measure, and understand the fundamental chemical interactions that happen when we create or alter molecules and materials. Tools and techniques such as computation, automation, and advanced analytics are fueling advances in our fundamental understanding of chemistry. The fact that the discovery of such tools is also reliant on fundamental chemical principles emphasizes the continually evolving and interdependent nature of fundamental research. Although it is difficult to predict which experiment or project will lead to the next major breakthrough in chemistry, it is clear that the evolution of new tools and technologies is essential to advance our capabilities and further our understanding of basic chemistry.

Among the various tools and technologies that are available to scientists at present, a few are emerging that are particularly impactful to understanding the molecular world and promoting real-world discovery. These are measurement, automation, computation, and catalysis. Investments in these four areas will enable chemistry and related fields to advance more rapidly and facilitate research discoveries likely to help us solve global challenges in the fields of energy, human health, national security, and environmental stewardship. As the chemical sciences have matured, advances in these tools and technology are promoting exciting new opportunities for convergent research. The synergies among them are being harnessed to develop resource-efficient chemistry and processes that conserve feedstock and energy while minimizing adverse effects on human and planetary health. These advances in the chemical sciences are spawning interdisciplinary work both among various subdisciplines of chemistry, including resurgent areas such as photochemistry and electrochemistry, and with researchers in other disciplines in science and engineering such as biology, chemical engineering, and data sciences.

The remainder of this chapter looks more deeply at measurement, automation, computation, and catalysis.

4.1 MEASUREMENT

By enabling us to accurately and precisely ascertain the composition, structure, properties, and quantities of a variety of materials, measurement science fosters innovation across the chemical enterprise. State-of-the-art measurement technologies allow us to probe everything from subatomic particles to human health to ocean sustainability to our solar system, providing real-time or near-real-time data even when looking back to the beginnings of the universe. The development of all measurement technologies is an interdisciplinary feat that requires expertise in physics, engineering, chemistry, biology, data science, computation, and many other fields to create accurate and precise tools. We focus on the topic of measurement because it is critically important to every aspect of the chemical sciences, and advances in measurement are both driven by and help to drive chemical discovery.

There have been many reports and review articles that cover different aspects of measurement in the chemical sciences, and every area of chemical research benefits from frequent new advances in measurement. The committee chose to focus on advances in measurement and analytical chemistry that benefit all disciplines of the chemical sciences. These areas include improvements in the visualization of matter, the enhanced speed of taking and analyzing measurements, and the increased accessibility of measurement technologies. All of these improvements also play a role in making the practice of chemistry more sustainable and contribute to advancements that will help address grand challenges such as climate change.

4.1.1 Improvements in Visualization and Imaging

The ability to sensitively and accurately image chemical matter has increased dramatically over the past several years. To understand imaging and visualization, a Natioal Academies’ report from 2006 on Visualizing Chemistry: The Progress and Promise of Advanced Chemical Imaging defined chemical imaging as “the spatial (and temporal) identification and characterization of the molecular chemical composition, structure, and dynamics of any given sample” (NRC, 2006b). This definition outlines the fact that chemical imaging has contributed to visualizing the very small (single atoms or molecules) to the very large (entire organs and ecosystem dynamics) by decoding the chemical and atomic composition of a sample.

Some of the most publicized improvements to measurement have been in the area of microscopy, for which Nobel Prizes were awarded in 2014 and 2017 for, respectively, the “development of super-resolved fluorescence microscopy” (Nobel Prize Outreach, 2022b) and “developing cryo-electron microscopy for the high-resolution structure determination of biomolecules in solution” (Nobel Prize Outreach, 2022c). Both of these techniques enabled measurement of different types of materials, reactions, chemical compounds, and biological molecules in unprecedented ways. With cryo-electron microscopy (cryo-EM), for example, computation-enabled improvements in resolution of structural images of biomolecules improved from around 15 angstroms (Å) before 2014 to as low as 1.2 Å in 2021 (Figure 4-1) (Peplow, 2020). Single-molecule fluorescence microscopy has not only allowed for a clearer spatial and temporal understanding of molecules but also provided an important technology for tracking reaction rates, including those driven by electrocatalysis (Li et al., 2018). This technique, along with other measurements of electrocatalysis such as surface plasmon resonance, will help drive the basic chemical research needed for the optimization of fuel cells.

Increased imaging sensitivity has allowed researchers to deconvolute the chemical composition of complex samples, including in environmental research. One example is identifying and quantifying different plastic polymers from water samples (Ivleva, 2021). It can be quite difficult to determine which plastic is being measured, due to structural similarities in their polymers. The most common method for interpreting the composition of these samples is by pyrolysis-gas

chromatography/mass spectrometry (Py-GC/MS), in which the sample is decomposed under heat and separated using gas-phase chromatography before the mass is analyzed. The full image developed from a Py-GC/MS data set helps researchers to understand the amount of different types of plastic in the water and can even lead to source identification. Similar types of MS methods have been applied to measuring the chemical interactions between different species within microbial communities in order to get a snapshot of community interactions (Dunham et al., 2017).

In addition to optical imaging, EM, and MS, there are a number of other imaging techniques that are rapidly improving and contributing to chemistry. These include atomic and molecular spectroscopies, other types of optical imaging, nuclear magnetic resonance (NMR), noninvasive imaging, and quantum imaging. Continued improvement in our understanding of basic chemical principles will enable analytical chemistry to better utilize all of these different techniques, and those insights will continue to improve imaging technologies for chemical research, as well as research in a variety of other fields.

One area on this list that has generated a lot of interest is the use of quantum principles to enhance sensing and imaging technologies. In a recent review, Yu and colleagues (2021) note that “a molecular approach to quantum sensing offers the unmatched combination of atomic structural control and tuanbility, enabling transformative discovery in fields spanning biology to astrophysics.” The impacts that quantum physics can have on chemical discovery as well as the ways that chemistry can contribute to our understanding of quantum mechanics include a plethora of ideas and avenues, especially in a well-explored area such as quantum sensing. A National Academies’ study is currently exploring this topic, specifically how quantum information science and measurement and modeling in chemistry can inform and advance one another.¹

4.1.2 Real-Time Chemical Measurements

Real-time chemical measurements and analytics have become critical for everything from monitoring the time courses of reactions to ensuring acceptable air, food, and water quality (more detail on the latter is in Section 3.4.4). In this report, the concept of real-time chemical measurements includes near-real-time chemical measurements where a sampling phase is required in order to gather an appropriate amount of analyte to produce an accurate measurement. The process of near-real-time chemical measurement still happens quickly and provides time-dependent and actionable information. The speed and accuracy of real-time measurements regularly improve, and those improvements are reliant on advances in computation, automation, data analysis methods, engineering, and separation science (see Box 4-3 for more information on separation science). The discoveries in analytical chemistry that drive these processes create new tools and technologies that are used to identify differences in an established baseline, make experimental comparisons, and identify and intervene in dangerous or life-threatening situations.

Real-time chemical imaging and measurement can be applied to monitor changes in the environment such as monitoring volcanic activity (see Box 4-1) (Bi and Han, 2019) or surveying water sources (Yaroshenko et al., 2020) for pollutants and other unusual activity. One notable example is the Ocean Observatories Initiative, funded by the National Science Foundation (NSF), that takes data measurements on “physical, chemical, geological and biological properties and processes from the seafloor to the air-sea interface” and makes the data available to any researcher who wants to use them.² Another example is the mobile chemical analysis stations that have been deployed to identify pollution in rivers and are, in one particular case, able to take measurements of “temperature,

___________________

¹ See https://www.nationalacademies.org/our-work/identifying-opportunities-at-the-interface-of-chemistry-and-quantuminformation-science.

² See https://oceanobservatories.org/.

total phosphorous, pH and ammonium ions, dissolved oxygen, conductivity, nitrate ions, and total organic carbon” (Yaroshenko et al., 2020). These measurements rely on different types of electrochemical and optical sensors, and due to the extended duration of measurement (sometimes decades), there is a strong established baseline, meaning that abnormal levels of various pollutants are discovered instantly. Real-time measurements of the environment can also utilize other types of instrumentation such as MS (Zuth et al., 2018). Much research is still needed to improve sensitivity and deconvolute complex samples using rapid separation science (NASEM, 2019a). These types of systems are also critical for chemical manufacturing, enabling workers to monitor product quality and ensure that pollutants are not spilling into the surrounding environment (Schmitz, 2015).

Measurement speed is related to the instrumentation and how well analytical technologies can accurately quantify or image a subject. But it is also directly tied to the availability of rapid data analysis. High-speed data analysis requires both computing power and efficient algorithms that can quickly process a data input (see Section 4.3 for more detail on computation). Analysis of chemical measurements has benefited greatly from the use of data interpretation algorithms that use ML and AI principles. For example, a variety of ML algorithms can be paired with electrochemical sensors to “extract complex relationships between chemical structures and their electrochemical properties and to analyze complicated electrochemical data to improve calibration and analyte classification” (Puthongkham et al., 2021). In these cases, ML methods help to calibrate the sensors and enhance the sensitivity of measurement, creating faster measurement capacity without the need to adjust the setup by hand (Figure 4-2). The incorporation of advanced computational techniques, such as AI and ML, into real-time measurement tools will increase the speed with which researchers and other users can make decisions based on new measurement data.

4.1.3 Increased Accessibility of Chemical Measurement

Many chemical measurement tools are getting smaller, more portable, and cheaper. These developments contribute to the increased accessibility of chemistry. Although not every analytical technology has benefited from increased accessibility, this is a definitive trend in a number of different areas. In the 2017 Annual Review Issue of Analytical Chemistry, it was noted that, “[c] learly, analytical tools are getting cheaper and more available as seen in reviews on paper microfluidics, 3-D printing, digital assays, and point of care diagnostics” (see Box 4-2) (Kennedy, 2017). Miniaturization of chemical instrumentation is partly enabled by the silicon revolution, and in part because chemical knowledge of sample preparation, separation, and sensor technology has facilitated the engineering of smaller devices. Frequently, decreasing the size of an instrument comes at a cost of reduced sensitivity and accuracy, but some recent advances have shown that this is not always the case.

Some instrumentation such as MS has recently undergone a transformative miniaturization process. Mass spectrometers that used to require entire rooms and specialized setups now can fit in the palm of your hand. Of course, full-room MS setups still exist and provide state-of-the art analysis of chemical samples, but there is a critical need for MS capabilities in places where entire rooms cannot be set aside for instrumentation. For example, companies such as 1st Detect Corporation and 908 Devices are building MS instruments that weigh less than 4 lbs for use in quality control of food and water as well as many security and forensics applications (Perkel, 2014). The scientific community also needs analytical capabilities at places that are challenging to get to, such as the bottom of the ocean or outer space. MS technologies have already been deployed for use in space research, but their capabilities are limited and there is a need for continually miniaturizing MS systems, so that single instruments can be used for purposes such as geological, chemical, and biological testing in space (Wainerdi, 1970).

4.1.4 Future of Measurement

Analytical chemistry and associated measurement technologies have made significant advances in the past couple of years, accompanied by advances in data acquisition, computation, sample preparation, microfluidics, and other enabling technologies. These tools can measure compounds in complex samples, create 3-D images of material by multiplexing different analytical methods, and trace material back to its original source with surprising accuracy. Two of the most important things that have facilitated these rapid advancements in analysis and measurement are the availability of data and the computational speed with which we can analyze data (Adams and Adriaens, 2020). As measurement continues to advance, so will the complexity of the data, which will require both better ways to integrate measurement and analysis and demand for new ways to store, retrieve, and use data. One increasingly popular concept in this regard is “democratic analytical chemistry” (Figure 4-3), which relates closely to the notion of accessible chemistry. The idea of democratic analytical chemistry is that measurements and instrumentation are available to any user directly (taking the measurements yourself) or indirectly (analyzing the measurement data from others) (de la Guardia and Garrigues, 2020). This concept is empowered by open-access software and readily available data repositories. As more systems like this develop, the network of chemists who can use and analyze data from measurements will expand significantly.

The concept of democratic analytical chemistry is also inherently a green chemistry concept. This is because instrumentation that already exists will be used to its full potential, mitigating the

need to use materials and energy to make and operate new instrumentation. Additionally, in some cases, no use of instruments is required. Rather, existing data, analyzed using the latest analysis techniques, will provide new insights.

Another future area for chemical measurement is the ability to perform real-time remote analytics. As discussed above, most real-time analytics rely on sample collection at a particular site and the movement of that sample to an on-site or nearby measurement device. There are, however, some unique capabilities that would be afforded by completely remote sensing. Some remote sensing techniques are already a reality (Bogue, 2018), but there is much development still to pursue in this area. Real-time remote chemical sensing would afford extensive opportunities to environmental and forensic science, national security, manufacturing, and many other areas.

Elsewhere in this report, there are other examples of chemical measurement’s role in advancing environmental stewardship and sustainability (see Sections 1.3.3 and 3.4.4). As noted earlier, many new tools and technologies are improving our ability to analyze air and water quality so as to be good stewards of the environment. Beyond the more direct impacts, the idea of “green analytical chemistry” is not a new one; it evolved simultaneously with the concept of green chemistry. These ideas are mostly related to reducing waste and preventing destructive interactions with the environment. As analytical chemistry increases in complexity, there is also an opportunity to be more innovative in the implementation of green and sustainable practices in chemical measurement.

4.2 AUTOMATION

The earliest mention of laboratory automation is from 1875 when chemists developed a device to “wash filtrates unattended” (Olsen, 2012). Over the next century, a number of unattended processes were developed for the chemical sciences. Laboratory automation in its current form, which utilizes robotics and computation, stems from the “demands in the life sciences in the 1980s for more productive means to aliquot and dilute biological samples for testing and analysis” (Selekman et al., 2017). This led to the development of automated liquid handlers capable of generating arrays of samples with minimal manual intervention. Further advances in equipment to automatically

dose both liquids and solids led to its use in high-throughput screening assays for drug discovery and combinatorial chemistry for material discovery. The unifying element behind applications of automation and high-throughput experimentation (HTE) is the ability to rapidly generate large diverse arrays of experimental data where first principles and rational design are difficult, owing to the innate complexity of the systems and responses under study.

High-throughput techniques built upon laboratory automation technology, coupled to in-line analytical monitoring and measuring technologies, statistic experimental design, and parallel experimentation, are accelerating the discovery of new molecules (e.g., catalysts) and new reactions, and optimizing processes that generate an array of chemical products such as pharmaceuticals, agrichemicals, and materials (Selekman et al., 2017). The capabilities of automation increased significantly in the past decade with advances in software, computational power, robotics, analytical measurements, data science, and ML now incorporated into automated experimentation (Coley et al., 2019).

Advances in science are made by iterative cycles of experimentation, data analysis, and hypothesis formulation. These advances progress at a rate that is commensurate with the rates at which experiments can be performed and data can be analyzed. Thus, scientists have been motivated to automate the execution and analyses of experiments to accelerate discovery. The power of automation in research is readily apparent from recent breakthroughs at the interface of biology, chemistry, physics, and engineering. Some examples are the automated synthesis and analysis of DNA, RNA, and polypeptides. Automation technologies for these biopolymers have played a critical role in most of the transformative discoveries and advances in biology and medicine. Preparation of the mRNA vaccines for SARS-CoV-2 and sequencing of the human genome are just two recent landmark achievements in synthesis and analysis, respectively, that were enabled by automation.

While chemists can certainly claim partial ownership of the invention of automated methods for the synthesis and analysis of biopolymers in biological research, scientists are now using automation in new ways to advance research in the chemical sciences. This section highlights how chemistry is benefiting from automation technologies.

4.2.1 Automation for High-Throughput Experimentation

With automation, it is possible to execute millions of experiments on small molecules in parallel and in a short period of time. For example, pharmaceutical companies and academic and national research centers have facilities for HTE on libraries containing millions of small molecules. Robots are used to rapidly dispense molecules into 96-, 384-, and 1536-well plates in which assays of their binding to biomolecules of interest or their capacity to perturb enzymatic or cellular activity are measured (Mennen et al., 2019). Because of these robust automated systems, the synthesis of small molecules, rather than their evaluation in biological assays, has become the rate-limiting component in drug discovery. In addition to its use in drug discovery, chemical companies have recognized the value of automation, and understand its benefits in boosting the productivity of research and increasing worker safety (Bardin, 2021).

The technology for automating experimentation has been repurposed for many research areas in the chemical sciences. Robots enable parallel experiments in screens for catalysts for chemical transformations and for optimizing conditions for chemical reactions of interest. This is exemplified by the Merck Center for Catalysis at Princeton University, where a robotic system facilitates the setup, monitoring, and characterization of thousands of reactions in parallel.³ It is equipped to dispense desired quantities of solid reagents and solutions in ambient atmospheres or in the inert atmosphere of a glovebox. Reaction analyses are enabled by automated, in-line GC/MS analysis.

___________________

³ See https://chemistry.princeton.edu/research-facilities/merck-catalysis-center.

With this infrastructure, it is possible for skilled analysts to quickly evaluate a large number of reactants and catalysts such that data sufficient for model building can be generated. An early success of this technology was the discovery of a novel amino acid C—H arylation reaction (McNally et al., 2011). The research team noted that automation and rapid measurement and monitoring of successful reactions helps to “exploit [the] serendipity” present in scientific experimentation.

4.2.2 Automation for High-Throughput Synthesis

Historically, chemical synthesis was performed on an ad hoc basis focused on the optimal synthesis of a particular small molecule or its closely related analogs. Early deviations from this molecule-centric approach for synthesis was the preparation of polymers. In those cases, chemists recognized that the synthesis of polymers of defined structures from discrete building blocks was amenable to automation. Indeed, visionary chemists, such as Bruce Merrifield and Marvin Caruthers, developed methodologies for the automated synthesis of peptides (Kent, 2006) and nucleic acids (Caruthers, 2013), respectively. More recently, Peter Seeberger and others have implemented methods for the automated synthesis of carbohydrates, which are among the most complex of the polymers, due to the diversity of the monomeric building blocks, the complex branching of many carbohydrates, and the need to control stereospecificity (Plante et al., 2001).

Whereas polymers are particularly well suited for automated synthesis, it is a very different matter for molecules with molecular weights under 800 Daltons (Trobe and Burke, 2018). The structural diversity and complexity of small molecules is immense and there are no general solutions for their preparation. Conditions for chemical transformations in the synthesis of a small molecule are selected or screened for based on the peculiarities of the reactants, and approaches for the isolation of reaction products are either chosen by analogy or determined empirically. These facts have underscored arguments that chemical synthesis is tedious, challenging, and equally art and science. When there is an interest in the syntheses of closely related molecules, as in a medicinal chemistry program or ligand screening in a catalyst study, the ad hoc approach to small-molecule synthesis can slow progress. Accordingly, there has been growing interest in automating the syntheses of small molecules. Some of the earliest cases of small-molecule synthesis via automation date back to the 1970s and 1980s wherein volumetric transfer of solvents and solutions containing reagents to resin-bound reactants enabled parallel and combinatorial syntheses. The solid-phase methodology for chemical synthesis is highly amenable to automation but is limited to certain types of reactants and transformations (Leznoff, 1978). Accordingly, there has been much interest in and effort spent developing automated platforms for executing solution-phase reactions.

4.2.2.1 Automated Chemical Synthesis Using Flow Chemistry

Some of the most promising platforms for automating solution-phase chemistry are based on a subdiscipline of chemistry called “flow chemistry,” in which chemical reactions are effected by the controlled pumping of the input stream, and the mixture of solutions and reagents is developed and studied (Figure 4-4) (Hartman, 2020). Developing industrial-scale flow chemistry (i.e., continuous manufacturing) is under active investigation throughout the United States and the world, prominently at the Novartis-MIT Center for Continuous Manufacturing.⁴ In flow chemistry, conventional bench operations of batch approaches can be automated (reagent dispensing, reaction mixing and work up, product purification and analysis, etc.). In this regime, reaction rates and productivity are enhanced by thorough mixing of reagents with reactants and by efficient heat and/or mass transfer. Moreover, reactions can be pressurized at superheated conditions wherein reactivity is greater

___________________

⁴ See https://novartis-mit.mit.edu/.

(Hartman, 2020). Importantly, reactions performed in flow are often safer than those in round-bottom flasks or other conventional reaction vessels because hazardous substances and combustible reagents are contained within stable devices.

Reflecting the promise of the technology, there are now commercially available flow synthesizers in use by both academic and industrial groups. These devices have peristaltic pumps, mixing elements, residence-time loops, and separation units that are controlled by a computer (See Box 4-3 for more information on separation science). The hardware can be coupled to process analytical technology and can assess “in-line” multiple physicochemical properties of the reactants and products via thermocouple, spectroscopy (ultraviolet, infrared, and Raman), MS, NMR, crystallization monitoring, and particle size analysis (Browne et al., 2017). There are autonomous flow devices that utilize computational tools and software such as “design of experiments” and algorithms (e.g., evolutionary, self-optimizing, and ML) to monitor, manage, and fine-tune the operation of flow systems (Coley et al., 2019). Continuous-flow systems have proven to be an enabling technology for the efficient synthesis of small molecules of all kinds, including the manufacturing of active pharmaceutical ingredients. Indeed, the U.S. Food and Drug Administration has declared continuous manufacturing as one of the most important tools in the modernization of pharmaceutical chemistry (FDA, 2019). The combination of automated flow chemistry with comparably automated product and data analysis promises to change drug discovery and other areas within the chemical industry by giving us the ability to synthesize a wide range of organic matter (Li et al., 2015).

4.2.2.2 Automation and the Accessibility of Chemical Synthesis

Automation, and particularly flow chemistry, are changing the way in which academic chemistry functions. Martin D. Burke, a leader in the field, contends that automated flow chemistry will increase the accessibility of synthetic organic chemistry and make small molecules more readily available to all scientists (Burke, 2021). Dr. Burke cofounded the Molecule Maker Lab Institute, which is “an interdisciplinary initiative with leaders in AI and organic synthesis intensively collaborating to create frontier AI tools, dynamic open access databases, and fast and broadly accessible

small molecule manufacturing and discovery platforms.”⁵ To make organic synthesis more accessible, they are making advances in modular synthesis by treating small molecules like Legos or building blocks (Figure 4-5). To reach a broader audience, the institute is also running “make-athons,” where students develop ideas to solve health-related problems and design molecules that can be synthesized using the institute’s infrastructure. This provides an in-depth opportunity to excite students about chemistry and its practical applications (Burke, 2021).

4.2.3 Future of Automation in Chemistry

Automation is transforming the chemical sciences and manufacturing, and there are undoubtedly more advances ahead that will push against conventional boundaries. Importantly, automation, HTE, and data capture can play an important role in sustainability, especially in processes where automation can help optimize green and sustainable reaction conditions. These technologies can also help with sustainability in areas such as early-stage drug discovery where microscale or nanoscale HTE uses minute amounts of chemicals to find the desired reaction condition, and data science will enable researchers to get to the end point with far fewer experiments. Additionally, automation is enabling small-molecule syntheses on scales where computational data can be stored in the form of small molecules (Cafferty et al., 2019). Because automation is a central part of a rapidly changing landscape in the chemical sciences, academic institutions may want to reconsider both their curricula and research avenues to minimize the gaps among basic science, translational research, and manufacturing. In this regard, Carnegie Mellon University is at the vanguard. The

___________________

⁵ See https://moleculemaker.org/.

university has partnered with Emerald Cloud Lab to establish the Carnegie Mellon University Cloud Lab (CMU, 2021). Similar to the Molecule Maker Lab Institute, this remote-controlled lab is providing a universal platform for AI-driven experimentation in the biological and chemical sciences. It contains state-of-the-art instrumentation that can execute all aspects of daily work, from experiment design to data acquisition and analysis. The university envisions it as a resource not only for its faculty, students, and staff but for scientists around the world. It is easy to imagine that the successes of these laboratory platforms will inspire other institutions to break out of conventional ways of doing science and to embrace automation.

4.3 COMPUTATION

Building on concepts and theories from chemistry and physics developed in the late 18th, 19th, and early 20th centuries, scientists began turning theory into computable algorithms around the early 20th century, inaugurating the field that became computational chemistry. Early discoveries in computational chemistry include development of quantum chemistry and theories of chemical bonding of atoms in molecules (Esposito and Naddeo, 2014), the chemical ensembles of statistical thermodynamics that describe interactions between molecules (Gibbs, 2010), and elucidation and calculations of rates of transformation in chemical reactions (Eigen, 1961; Klippenstein et al., 2014). Even ML has an early precursor from chemistry in the work of Louis Hammett, who elucidated a mathematical relationship between equilibrium constants and descriptors of varying substitutions on a benzene core (Hammett, 1937).

Initially, solutions to these algorithms were calculated manually, using numerical methods for integration, differentiation, enumeration of ensembles of states, and curve fitting, with the result that algorithms were applied only to small, more tractable systems such as the hydrogen atom. Coming up with these solutions could take thousands of person-hours to calculate. The advent of mechanical, electromechanical, and vacuum-based computers vastly increased the speed of calculations and concomitantly expanded the size and complexity of the systems that could be studied. ENIAC, or Electronic Numerical Integrator and Computer, the first programmable, general-purpose computer, could perform thousands of computations per second (Kopplin, 2002; Levy, 2013; Mangalindan, 2021). Hand-in-hand with this computational speed, ENIAC’s architecture of branching and stored memory enabled programmers to move beyond a linear sequence of operations and incorporate conditional execution (if-then-else for loops) in their codes. This innovation in computer architecture led directly to computational innovation: the first-ever computational Monte Carlo simulation, a calculation run on ENIAC on behalf of Los Alamos National Laboratory that simulated the movement of neutrons through the fissile material of a nuclear bomb (Haigh et al., 2014). Since this first calculation, Monte Carlo simulations have become a widespread and valuable tool across a wide range of computational chemistry applications (Amar, 2006; Doll and Freeman, 1994; Earl and Deem, 2008; Saito, 1997). This synergy among computational speed, computer architecture, and algorithmic innovation is a recurring theme throughout the development of computational chemistry as a discipline.

There have been many reports and review articles that cover the current state of the art for computational chemistry, and every area of chemical research benefits from frequent new advances in computational power and algorithm development. While this chapter and many other sections in the report point out some specific and important examples, the committee chose to focus on chemistry-enabled advances in computational hardware that benefit all disciplines of the chemical sciences and that have the potential for transformative impact on the types of chemistry challenges that can be addressed by computational chemistry algorithms.

4.3.1 Present-Day Computational Chemistry

As noted in Section 2.3.3.3 of this report, fundamental chemistry, along with advances from many other fields, have led to a steady increase in computing power and provided an improved architecture (Clancy, 2012; NRC, 2003). Research on photolithography, in particular, led to monumental advances in the miniaturization of microchips to allow more transistors per microprocessor (see Section 2.3.3.3).

Because of these steady increases in computational power and speed, and the decreases in environmental impact, computation can be applied to larger and more complex chemical systems. Computational chemistry has become a mature discipline that is used across all chemical subdisciplines to derive understanding, enable prediction, and drive experimentation. Quantum mechanics and statistical thermodynamics theories are regularly applied to biological macromolecules such as proteins, DNA, and RNA and to complex materials such as catalysts and batteries. ML and AI methods have gone far beyond Hammett’s examinations of a few substituted benzoic acids to model building for data sets containing millions of molecules. These advances have become a driving reason for why the collection, storage, and standardization of chemical data are so critical. These more advanced computational processes only work with reliable data and training sets, which come from a wide breadth of information collected through experimentation and computational modeling. These new computational tools are also being used with supercomputers to analyze complex chemical models. In more complex simulations with many increasing atoms and variable space, supercomputers are the best option for computation, especially because of the modern-day increase in computational power associated with these machines (Figure 4-6).

Evidence of this breadth of application can be seen in usage statistics for supercomputing facilities available through the NSF’s Extreme Science and Engineering Discovery Environment (XSEDE) computing environment and the Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program for access to the Department of Energy’s (DOE’s) Leadership Computing Facilities. Chemists and materials scientists regularly account for approximately 20–35% of the principal investigators (PIs) who received INCITE allocations between 2017 and 2021, and chemists also made up approximately 22–24% of the PIs that used the XSEDE environment over the same time period (Figure 4-7). Within the field of chemistry, all subdisciplines are represented among XSEDE users, with materials scientists accounting for the biggest portion of users of this resource (Figure 4-7).

While computing is enormously helpful to solving complex chemistry questions, it comes at a price: Computing exacts a huge energy burden. While some programs, including one in DOE,⁶ are working to solve this issue, the energy usage from computers and supercomputers is still an issue. One article noted that “electricity demand by data centers in 2018 was an estimated 198 terawatt hours, or almost 1% of demand for electricity in the world” (Ayanoglu, 2019), and the numbers will certainly keep increasing. Groups have started to measure the energy usage of supercomputers, and there is a push toward vastly increased energy efficiency for high-performance computing. The most efficient supercomputers are tracked using the Green500 List.⁷ However, the energy efficiency seen in supercomputing has not been incorporated into all aspects of computing. This is especially true for data analytics of increasingly large data sets, which is a critical aspect of chemical research and will only grow in importance. As with advances in computing power and miniaturization, chemistry will be a vital contributor to improving energy efficiency for supercomputing, data centers, and other high-performance computing needs, both through the creation of new, more sustainable architectures such as neuromorphic computing (see Section 4.3.2.1) and in materials chemistry innovations that lead to greater sustainability for all computational systems.

___________________

⁶ See https://www.energy.gov/eere/femp/energy-efficiency-data-centers.

⁷ See https://www.top500.org/lists/green500/2021/11/.

4.3.2 Looking Toward the Future of Computational Chemistry

Miniaturization, increased computational power, and faster computational speeds have vastly increased the size and complexity of chemistry problems accessible to computation, but we are reaching the limits of miniaturization. As transistors are reaching scales at the size of atoms, we can go no smaller. In addition, as computation has become a regular and important component of chemical research, and data science becomes embedded across a wide range of disciplines, energy consumption becomes a limiting factor. Developments for the future will need to focus on improving sustainability and energy efficiency across all scales of computation—supercomputers, data centers, and personal computing devices—and on new computing paradigms that allow us to move beyond Moore’s Law. As we explore new materials and computer architectures, chemistry and chemical engineering will continue to play a role in making fundamental discoveries to drive sustainability in computation, just as they drove miniaturization in the past. As was the case with development of our current semiconductor architectures, new materials and computer architectures will drive algorithmic innovation that enables us to ask new chemistry questions in entirely new ways.

4.3.2.1 Quantum Simulation with Quantum Computers for Chemistry

Nonclassical computing architectures such as quantum computers represent a new class of computational tools for chemistry simulations. These computer architectures will complement classical tools since performing certain calculations with quantum computers provides a computational resource advantage for specific problems in quantum chemistry. Quantum computers behave fundamentally differently from classical computers because they are made of different types of materials than semiconductors and store and encode information based on the principles of quantum mechanics (Figure 4-8). Though the development of quantum computers is an active area of research, several classes of them have begun to emerge. At a high level, these include ion trapping, transmon-based technologies involving superconductors, and theoretically topological quantum computers. Quantum bits, “qubits,” are being produced today that leverage ion-trapping materials and superconductors. (At the time of this report, topological quantum computers had not been actualized.)

Since quantum computers operate under fundamentally different principles than do classical computers, they are best suited to tackle certain types of problems. As a concrete example of the differences in problem solving, there is no advantage (speed up) to using a quantum computer for addition or multiplication operations, but there is a theoretical advantage for number factoring (e.g., Shor’s algorithm). Through the construction of quantum circuits, programmers leverage quantum mechanical properties of the qubits—such as superposition, entanglement, and interference—for computation. Thus, quantum computers are best suited to handle complex, interconnected information.

Chemists may intuitively associate complex, interconnected system problems with electron correlations, and indeed, the behavior of electrons in a molecular system is well described by quantum algorithms. Quantum chemistry calculations using Schrödinger’s equation are typically exponential in resource requirements. As the problem size gets larger, the calculation quickly becomes intractable on classical devices. So, “exact” calculations on molecules, where the entire set of particle interactions in a system is simulated, become impossible. Furthermore, the Pauli exclusion principle adds to the cost because two electrons cannot sit at the same energy with the same spin, which is referred to as the “sign problem.” State-of-the-art full-configuration interaction calculations are intractable past approximately 20 orbitals and approximately 20 electrons, even with today’s most powerful supercomputers. The calculations can take more than a week to run, and calculations for reactions with popular approximate computational chemistry methods such as density functional theory (DFT) can be inaccurate. To mitigate these problems, simulations

involving DFT invoke approximations to avoid high computational cost––for example, modeling the distinct electrons in an “electron field” compared to calculating all of the distinct electron–electron interactions. Determining the energy landscape for a reaction mechanism can take months or years of trial and error.

Although all of these issues with classical computers might be solved using quantum computers, there is still a large overhead associated with quantum chemistry calculations on quantum computers. A quantum chemistry circuit can require millions of gates to do a full configuration interaction calculation of a molecule. For quantum algorithms, such as quantum phase estimation for chemistry simulations, error-corrected devices (e.g., universal fault-tolerant quantum computers) will be necessary in order to carry out the operations (Gambetta, 2020). Quantum computing in a fault-tolerant regime promises more accurate chemistry calculations (possibly leading to predictive chemistry) (Motta and Rice, 2022). However, while research continues in quantum hardware, software, and theory, researchers can test small examples on devices available today. Researchers will often use a simulator that runs on a classical computer to test resources for calculations before trying the calculation on a noisy real device. The quantum computers all have differences in connectivity, number of qubits, and noise profiles, which enables customization for a calculation of interest. How to best leverage the hardware for chemistry calculations remains an open research question.

Beyond quantum chemistry, quantum computing might have advantages for chemistry broadly, through quantum ML, optimization, and Monte Carlo calculations. Applied to chemistry examples, it is possible to imagine exploring quantum computers for classification, anomaly detection, conformation optimization, solvation, and kinetics, to name a few. Finally, it is feasible that due to hardware performance differences, different types of quantum computers may be better suited for different classes of problem solving. Thus, incorporating quantum computers into computational chemistry workflows is a promising use of this emerging technology.

4.3.2.2 Biology-Inspired Computer Architecture

Neuromorphic computers, also referred to as “brain-inspired computing,” aim to maximize the energy efficiency of computational processing and communication, thereby offering a low-energy computing platform, especially when combined with accelerators designed with the same principles (Schuman et al., 2017). New and emerging hardware for neuromorphic computing can operate with energy input on the order of milliWcm^-2, whereas complementary metal oxide semiconductor (CMOS) technologies operate in the 50–100 Wcm^-2 range (D. Liu et al., 2021). Given the significant energy input required to run the world’s silicon-based supercomputers, offloading key computations to low-energy requirement architectures would improve the sustainability of computation. Additionally, since the neuromorphic architecture handles data input differently than conventional computers, neuromorphic computers may excel at certain tasks for chemistry applications. A key research question will be to determine which tasks should be managed by neuromorphic, quantum, and conventional architectures.

Neuromorphic computing is a new architecture made up of artificial neurons and synapses and inspired by the current understanding of how brains work (Schuman et al., 2017). These architectures aim to capture properties of how brains function as learning machines that operate efficiently. Current semiconductor architectures are made up of large numbers of separate computational (CPU) and memory components, and current computational algorithms aim to maximize the simultaneous usage of all CPU and memory components to attain computational complexity and speed. In contrast, neuromorphic architectures make use of a vast number of relatively simple processing components that contain both memory and computational properties in one device. Algorithms for these architectures sparsely activate only a small number of the processing components at any time.

Neuromorphic computers are incarnated as neurons and synapses on traditional CMOS chips, in mixed analog and digital systems known as MEMristors,⁸ and, increasingly, in future devices that are modeled after actual chemistry-driven synapses. Examples of the latter include water–lipid mixtures (DOE Office of Science, 2020b) and electron transport through ion channels (Lee et al., 2021; Tang et al., 2019). These shifts in architecture hold promise for the creation of computers that can use energy more efficiently by orders of magnitude.

A key challenge for making neuromorphic computing a reality is the feedback between architecture and algorithm. All aspects of computing are being rebuilt at the same time, and neuromorphic hardware needs to exist to motivate creation of new algorithms, while an understanding of neuromorphic-appropriate algorithms is necessary to drive the design of appropriately matched algorithms. Solving this challenge will require true multidisciplinary research at the intersection of chemistry, biology, neuroscience, and computer science.

4.4 CATALYSIS

Catalysts are used in most of the chemical reactions that make commonplace products, such as fuels, food, pharmaceuticals, synthetic fibers, and plastics. They drive chemical processes by facilitating the rearrangement of atoms and molecules. It is estimated that more than 95% (by volume) of everyday products contain ingredients that are made using catalysts, accounting for >80% of added value in the chemical industry (Hagen, 2015). Despite the well-established importance of catalysis, feedstock changes, energy transitions, and other environmental, health, and safety concerns will bring about fundamental changes in the chemical economy requiring new approaches for catalytic systems as well as a reexamination of previously explored technologies. In particular, the fields of electrocatalysis, photocatalysis, and biocatalysis are expected to play significant roles in the resource-efficient conversion of emerging feedstocks.

___________________

⁸ More information on MEMreistors available online at https://www.nanowerk.com/memristor.php.

4.4.1 Homogeneous Versus Heterogeneous Catalysts

Two main types of catalysts drive reactions: homogeneous and heterogeneous. Homogeneous catalysts are in the same phase as the reaction they are catalyzing, usually meaning they are in the liquid phase, while heterogeneous catalysts and reactants are in different phases. Both types of catalysts are employed in the chemical economy; however, because homogeneous catalysts are less thermally stable at the elevated temperatures used in many high-volume petrochemical applications, heterogeneous catalysts are more commonly used. More than 80% of industrial catalytic processes are based on heterogeneous catalysis where the catalyst exists as a solid and the reactants are present in a gas and/or a liquid phase surrounding the catalyst (Wacławek et al., 2018).

Even in systems where heterogeneous catalysts dominate commercially, there are often homogeneous catalyst systems as well. For example, while heterogeneous Ziegler–Natta catalysts produce the majority of polyolefins, the commercialization of single-site catalysts in the 1980s has allowed exceptional control of polymer molecular weight and weight distributions, branching, stereochemistry, and a wide range of other properties not possible using heterogeneous catalytic systems (Stürzel et al., 2016). Continued progress in ligand design has expanded the thermal stability of some homogeneous polyethylene catalysts to 160°C and greatly extended the potential application space (Klosin et al., 2015). Design of new homogeneous polymerization catalyst systems remains a robust area of research (Chen, 2018).

Homogeneous catalysts typically have superior selectivity over heterogeneous catalysts because ligand selection can be used to tune the electronic and steric properties at the metal site. As such, homogeneous catalysts are used in a number of fine chemical and pharmaceutical processes (Howard et al., 2006; Zecchina and Califano, 2017). The desire to improve the sustainability of pharmaceutical preparations is leading to the design of novel metal catalytic systems (Hayler et al., 2019). Application of techniques such as NMR and infrared spectroscopy enables direct investigation of the catalytic reaction (Howard et al., 2006) which when combined with computational methodologies promises to deliver even more detailed mechanistic understanding with the potential for increased selectivity (Durand and Fey, 2021).

One downside of homogeneous catalysts is that the recovery of the catalyst often requires additional materials and energy that could have adverse environmental impacts, while the separation of heterogeneous catalysts from the reaction mixture is usually quite simple. Because homogeneous catalysts are present at low levels, many of the standard separations techniques are not well suited for this application (Schnoor et al., 2019). Furthermore, near-quantitative catalyst recovery is often demanded by economic and/or product purity considerations, especially when precious metals are used. As a result, developing new approaches for catalyst separation is an active area of research (Cole-Hamilton, 2003; Vural Gürsel et al., 2015). Use of membrane separation (Janssen et al., 2011; Schnoor et al., 2019; Xie et al., 2020), soluble polymer supports (Bergbreiter et al., 2009; Xie et al., 2015), magnetically separable catalysts (Kazemi and Mohammadi, 2020), and mimicking homogeneous metal sites through metal organic frameworks (Drake et al., 2018) are a few of the many approaches under exploration to address this challenge. These advances in homogeneous catalysis may have applications in processing biorenewable sources of carbon, such as conversion of lignocellulosics to chemicals, where low-volume high-value products are required for economic viability of biorefineries (Bender et al., 2018).

The importance of catalysis has long been recognized. Beginning with F. Wilhem Ostwald’s 1909 Nobel Prize for the discovery of fundamental principles of equilibria and reaction rates, the scientific contributions of both homogeneous (including enzymatic catalysis) and heterogeneous catalysis have been recognized with Nobel Prizes more than 15 times (Thayer, 2013). Probably one of the more cited examples is the 1918 Nobel Prize awarded to Fritz Haber for “the synthesis of ammonia from its elements” (Nobel Prize Outreach, 2022a).

The conversion of atmospheric nitrogen into ammonia via the Haber-Bosch process, used today primarily to produce synthetic fertilizers (see Section 1.3.1), is a well-known example of a heterogeneous catalytic reaction. While the natural process of nitrogen fixation uses soluble molecules and enzymes as homogeneous catalysts, the Haber-Bosch process uses powdered iron particles as a heterogeneous catalyst. The key reaction steps between nitrogen and hydrogen occur on the surface of the iron particles.

The evolution of the field of heterogeneous catalysis is shown in Figure 4-9. Most heterogeneous catalysts comprise metal particles anchored to high-surface-area oxide or carbon supports. To maximize the number of active sites per mass of metal, the metal particle diameter is ideally on the order of a few nanometers. However, industrial catalyst synthesis and development are still largely based on trial-and-error approaches; accumulated practical knowledge; and the basic principles of physical chemistry, solid-state chemistry, and advanced experimental tools for characterizing surface reactions. New methods and tools in surface science are helping to address some of these gaps (see Box 4-4), but if catalysts can be improved with precision design to maximize activity, selectivity, and durability, the projected cost savings could be between $3 billion and $6 billion/year, with corresponding energy savings of 300–600 trillion BTU/year (Thayer et al., 2006).

Bridging homogeneous and heterogeneous catalysis continues to be a grand challenge (Cui et al., 2018). In heterogeneous single-metal-site catalysts (also referred as single-atom catalysts), the electronic properties of the atomically dispersed metal centers and their catalytic activity are tuned by the interaction between the metal and the neighboring surface atoms such as nitrogen, oxygen, or sulfur. This is similar to how the activity of metal centers in homogeneous catalysis is tuned by the choice of ligands. In this way, heterogeneous single-metal-site catalysis introduces new opportunities for bridging homogeneous and heterogeneous catalysis. Supported ionic liquid phases (SILPs) represent another advance in attempts to “heterogenize” homogeneous catalysts (Selvam et al., 2012). In SILP catalyst materials, a thin film of ionic liquid (IL), containing the dissolved (homogeneous) transition metal catalyst complex, is deposited within a porous solid, such as silica. Given that ILs have vanishingly low vapor pressure at typical reaction conditions, the IL film

remains relatively stable during continuous fixed-bed reactor operation (Riisager et al., 2005). SILP catalysis has been demonstrated for industrially important reactions, such as hydroformylation, hydrogenation, carbonylation, hydroaminomethylation, metathesis, water-gas shift reaction, and oxidations (Fehrmann et al., 2014). The SILP catalytic materials have been demonstrated to surpass heterogeneous counterparts in terms of activity and selectivity. A better fundamental understanding of the physicochemical processes underlying SILP catalysis as well as its stability in multiphase environments is key to broader practical application of this uniquely tunable system.

4.4.2 Reemergence of Photocatalysis, Electrocatalysis, and Biocatalysis

While alternative methods to classical metal-based catalysis, such as photocatalysis, electrocatalysis, and biocatalysis, have been under consideration for decades as useful methods to drive chemical reactions, they are recently experiencing a resurgence in popularity as alternative methods to classical metal-based catalysis (Figure 4-10). The literature describing new transformations via electrochemistry, photochemistry, and biocatalysis portends their potential for vast applications in materials synthesis, chemical production, and new pharmaceutical and agrochemical molecules. Though there are many challenges to be overcome, these methods provide an alternative to

energy-intensive synthesis methods that use high heat, high pressure, or both to perform chemical transformations in the production of some industrial chemicals.

4.4.2.1 Photocatalysis

Photochemistry is the use of light (i.e., radiant energy) to excite the electrons of a molecule to drive chemical change. The most ubiquitous photochemical reaction is photosynthesis, performed by phototrophic microorganisms and plants that use solar energy to convert carbon dioxide (CO₂) and water into glucose and oxygen.

The emergence of photochemistry as a sustainable technology for the synthesis of complex organic molecules can be attributed to developments over the past decade in light sources, photocatalysts, and flow reactors. Advancements in light-source technology, specifically light-emitting diodes (LEDs), fabricated from layered crystalline semiconductor material, offer advantages over compact fluorescent light sources. LEDs are smaller, more energy efficient, and able to generate light with very narrow emission spectra. Though the energy content of visible light is often insufficient to directly cleave bonds and induce chemical reactions, this can be addressed by adding a catalyst molecule that can more effectively adsorb the radiation, and then exchange the energy, electrons, or hydrogen atoms with other organic molecules. In the past decade, a large library of photocatalysts has been developed that can catalyze a wide array of reactions (Djurišić et al., 2020). These include photochemical transformations like C–H activation, C–H arylation, C–H alkylation, C–O bond formation, C–N bond formation, fluorination, dehalogenation, isomerization, rearrangements, cycloadditions, and many more (Djurišić et al., 2020). Some other specific advantages of photocatalysis include the ability to access highly saturated molecules or to modify chemically obstructed areas of molecules, which are common needs in many pharmaceutical syntheses.

Photochemical reaction rates are dependent on the local light intensity; thus, the reactor configuration and light source placement are critical to reaction performance. Reactors with shallow dimensions provide efficient irradiation of the entire reaction medium and would be favored for photochemical reactions. These requirements can be met by using continuous flow reactors as

discussed in Section 4.2.2.1. Custom flow microreactors for performing photocatalytic organic chemistry have proliferated thanks to 3-D printing and inexpensive LEDs and materials. Commercial flow reactors for reaction screening are now available as well. With these advances in light sources, catalysts, and flow technology, photocatalytic reactions can be performed with precise control over the reaction progress (Buglioni et al., 2022). Although there are a few industrial processes using photochemistry (e.g., see Box 4-5), large-scale photocatalysis of complex organic molecules is still an emerging area.

4.4.2.2 Electrocatalysis

Electrochemistry uses electrical potential to drive the arrangement of electrons to create a chemical change. Within an electrolytic cell, electricity is used to drive an oxidation-reduction (redox) reaction to generate a flow of electrons from the anode to the cathode. One of the first electrochemical reactions was the decomposition of water into hydrogen and oxygen by electrolysis, discovered in 1800 (Smolinka, 2009). This soon led to the discovery of electroplating (addition of metal to surfaces) and electropolishing (removal of metals from surfaces). These are examples of direct electrolysis, in which molecules undergo electron transfer directly at the electrode surface. Electrochemistry also includes indirect electrolysis, in which a catalyst serves as a mediator of the electron transfer. Indirect electrolysis as well as advances in laboratory equipment for these reactions are now enabling complex molecular transformations. These transformations include carboxylation, C–N formation, functionalization of C–O, methoxylation, oxidative intermolecular coupling, and oxidative intramolecular cyclization. Electrochemical reactions using common feedstocks such as CO₂, ammonia, and water to supply carbon, nitrogen, and oxygen, respectively, can form the basis to build other chemical feedstock molecules. These are essential transformations as we move toward electrification and decarbonization of the chemical industry (Schiffer and Manthiram, 2017).

Until this century, electrochemistry had not been broadly investigated by organic chemists as a reaction for high chemoselectivity to create new molecules. The setup for these reactions was believed to be too complex, with too many reaction variables to consider, such as electrolytes, electrode composition, cell type, and potentiostat. In addition, there was a lack of standard instrumentation and protocols to run these reactions. The fundamental research work of several academic organic chemistry groups, starting around 2000, showed that these types of electrochemical transformations were feasible (Moeller, 2000). The discovery of new electrocatalysts that can lower the overpotential of electron transfer and impart chemo-, regio- and stereoselectivity during the transformation of reactive intermediates is another significant advancement in the reemergence of electrochemistry. Strategies to provide precise control over the reaction progress now include externally controlling the electrical input, establishing new reactivity by coupling multiple redox events, and using new electroanalytical tools to understand the mechanisms of redox transformation. The growing interest in electrochemistry by organic chemists is driving the integration of electrochemical reactions into flow systems and the development of commercially available reactor systems to screen electrochemical reactions, such as the IKA ElectraSyn 2.0.⁹ Electrochemistry is poised to expand from the academic laboratory to industrial R&D, as industrial organic chemists adopt electrocatalysis for the development of new molecules.

4.4.2.3 Biocatalysis

As nature’s catalysts, driving biochemical reactions for billions of years, it may seem peculiar to highlight enzymes as an emerging area of chemical research leading to new and impactful applications. Furthermore, biologically driven catalytic processes used for food and beverages date back to the earliest known civilizations. However, it is only quite recently that biocatalysis has proven to be useful for large-scale, targeted synthesis of a wide array of chemical compounds.

At first glance, the scope of reactions catalyzed by enzymes may seem too narrow to be suitable for broad use in the chemical industry. These specialized proteins are classified by the Enzyme Commission of the International Union of Biochemistry and Molecular Biology into six classes: oxidoreductases, transferases, hydrolases, lyases, isomerases, and ligases. Each class is further delineated based on the functional groups on which the enzyme acts, the reaction being performed,

___________________

⁹ See https://www.ika.com/en/Products-Lab-Eq/Electrochemistry-Kit-csp-516/ElectraSyn-20-Package-cpdt-20008980/.

and the nature of the substrate. It is true that unmodified enzymes have been used for many decades in industries that range from food ingredients (e.g., glucose isomerase to produce high-fructose corn syrup) to pharmaceuticals (e.g., lipase-catalyzed hydrolysis to resolve enantiomers), proving that biocatalytic production systems are scalable.

But a significant breakthrough, based on fundamental chemical and biological research, has been the development of methods for precisely modifying the structure, and hence the function, of enzymes. These tools allow researchers to engineer new enzymes. In one example, researchers at Merck and Codexis engineered a transaminase to replace a rhodium-catalyzed reaction step in the production of sitagliptin (see Section 2.3.3.2 for more details). The resulting process increased productivity and yield and decreased waste. Additional innovations in enzyme design, including using computational approaches, have produced biocatalysts that perform new reactions (Siegel et al., 2010), create new molecules (Coelho et al., 2013), and incorporate new atoms (Kan et al., 2016).

The advantages of biocatalysis are well known, including moderate reaction temperatures and pressures, minimal organic solvent waste due to the use of largely aqueous solutions, and exquisite regio-, stereo-, and substrate specificity. Naturally, there are disadvantages, too, though these vary widely and are enzyme specific. They include low catalytic rate constants, poor enzyme stability, low total turnover numbers, and a limited substrate range, with respect to both the molecular structure surrounding the target functional group and the types of atomic bonds that can be formed. Recent advances in addressing these limitations, however, suggest that biocatalysis is primed to play a much larger role in chemical transformations, especially as work moves toward more sustainable chemical production.

4.4.2.4 Scaling Up Photocatalysis, Electrocatalysis, and Biocatalysis

Advances in electrocatalysis, photocatalysis, and biocatalysis have created new reactions with promising applications in many different industries. However, these chemistries differ from traditional organic chemistry in that scale-up equipment and strategies are more complex. For traditional organic reactions, the general reaction parameters (e.g., solvent, concentration, additives, catalysts, and temperature) and reactor-related parameters (mixing conditions, heat transfer, chemical compatibility, and residence time) are well understood and quantifiable. However, for electrocatalysis, photocatalysis, and biocatalysis, industrial-scale conditions and processes have to be developed, tested, and optimized.

Scaling up electrocatalytic reactions requires consideration of the standard potential, over-potentials, faradaic efficiency, electrolyte choice and concentration, resistance of the electrolyte, and variations in voltage and current. Existing large-scale electrochemical reactor units are either optimized for a specific reaction type or built for aqueous systems and not compatible with organic solvents.

In the case of photocatalysis, there are radiation-related reaction parameters to optimize, such as wavelength-dependent quantum yield, molar extinction coefficient of the reaction species, and wavelength selectivity of the desired reaction. Additionally, there are reactor-related parameters to work out, such as the wavelength and intensity of the light source, the reactor geometry (which impacts light distribution), the photon flux on the reactor walls, and the reactor material absorbance. A range of scale-up units for photochemistry have been custom developed using capillary and tubular plug flow reactors, immobilized photocatalysts, thin-film reactors, rotor-stator reactors, and continuous stirred tank reactors (Cambié et al., 2016). Customization has been aided by using 3-D printing to rapidly prototype designs.

Many of the reaction, energy source, and reactor parameters are coupled in electrochemistry and photochemistry. The challenge will be in decoupling these parameters to ensure that electrochemical and photochemical processes can be transferred across different platforms and scales.

To facilitate that, fundamental research in electrochemistry and photochemistry would be tied to research and development on the platform and reactor equipment that will permit scaling up these chemistries. Standardization of equipment for photochemistry and electrochemistry will enable the manufacturing supply chain to produce the new products and chemicals enabled by the chemistry.

Although there have been many successful translations of biological enzyme-mediated reactions into commercial processes, significant gaps remain that currently limit the ability of biocatalysis to play a larger role in sustainable chemical synthesis. The scope of known enzymatic reactions is far less than the full slate of reactions accessible through nonbiological catalysis. While multiple reaction schemes are possible for any specified target, it stands to reason that broader replacement of thermochemical catalysis necessarily requires an expansion of the chemistry that can be performed by enzymes. However, it is also the case that these new biological catalysts must perform well outside of academic laboratories. New insights are needed to enable not just proof-of-concept synthesis, but large-scale production. Directed evolution, for example, has proven to be a powerful tool to improve the function of both naturally occurring and de novo designed enzymes. A significant limitation to its use, though, is the need to generate extremely large libraries (typically on the order of 10⁹ variants) to search for the best-performing enzyme. Advances in measurement could launch new screening methods, but an alternative approach is reliable computational design. New knowledge leading to greater understanding of sequence–structure–function relations could lead to predictable enzyme engineering, requiring tractable library sizes on the order of 10² rather than 10⁹.

4.4.3 Future Challenges in Catalysis

Fundamental challenges in catalysis and surface science are articulated in two DOE reports (De Yoreo et al., 2016; DOE Office of Science, 2017). To address these challenges, DOE established five research priorities: considering both the binding site and allosteric effects when designing catalysts, understanding the dynamic evolution of catalysts, manipulating reaction networks in complex environments to selectively steer catalytic transformations, designing electrocatalysts that are highly selective and energy efficient, and driving new catalyst discoveries by coupling data science, theory, and experiment. Research programs and centers funded by the DOE Office of Basic Energy Sciences are addressing several of these challenges. The goal is that fundamental science advances and tools made available by this research will lead to precision design of catalysts that maximize yields of desired products under mild conditions, thus conserving both feedstocks and energy. While catalysis has been instrumental in major industrial success stories such as ammonia synthesis (Section 1.3.1) and the catalytic converter (Section 2.3.3.5), there are processes that have eluded a catalytic solution for decades. An example is the steam cracking of naphtha to produce ethylene, a major platform chemical with an annual capacity of 160 million tons. Among petrochemical processes, this one consumes the largest amount of energy with the endothermic cracking reaction accounting for roughly 13–22% of overall energy and the downstream separation of ethylene from ethane accounting for the rest (Wong and van Drill, 2020). Ongoing efforts to power the steam cracking furnaces with renewable electricity (the so-called electrification of steam crackers) are expected to significantly reduce the carbon footprint of this energy-intensive process (Hydrocarbon Processing, 2021). Efforts to develop a catalyst that further reduces energy requirements for this endothermic process have been challenged by rapid catalyst deactivation due to coking and catalyst stability issues. Advances in heterogeneous catalysis, materials science, and process engineering are needed to effectively decarbonize current ethylene manufacturing (Gao et al., 2019).

Many industrial heterogeneous catalytic reactions, including the most relevant reactions related to energy and decarbonization, are limited by the Sabatier principle (Elnabawy et al., 2020; Montoya et al., 2015). The activity of a catalytic site is dictated by one or more rate-determining steps in the catalytic sequence of adsorption, surface reaction, and desorption. Overcoming this limitation

is a fundamental research challenge in heterogeneous catalysis. Recently, Ardagh and colleagues (2020) put forward the catalytic resonance theory to understand and potentially enhance rate and selectivity in heterogeneous catalysis beyond the Sabatier limit. The theory recognizes that surface binding energy and transition state energies oscillate over time such that at a certain point, it may be easier for a reactant to adsorb on, or a product to desorb from, the catalyst surface. By superimposing an external wave at the catalyst surface that resonates with these oscillations, the simulation work shows that it may be possible to accelerate a surface-catalyzed reaction by several orders of magnitude compared to static conditions. The resonance theory was demonstrated experimentally for intensified formic acid formation via dynamic electrocatalysis (Gopeesingh et al., 2020). However, creating the required frequency of temperature or concentration oscillations in thermally driven conversions may be more challenging. This is an example of fundamental chemical research where experimentation and theory intersect to produce some groundbreaking possibilities for catalytic processes.

Plasma-driven heterogeneous catalysis is another area that shows promise for the chemical transformation of hard-to-activate molecules such as N₂, CO₂, and CH₄, all relevant in ammonia synthesis and in mitigating greenhouse gas emissions (Mehta et al., 2019). Mehta et al. (2019) note that there is a challenge in discovering catalytic materials

For example, using ozone, a powerful oxidant, at mild conditions is being tried in organic syntheses in flow reactors (Polterauer et al., 2021) and light alkane activation (Zhu et al., 2021). Electrification of the chemical industry using affordable renewable energy could spur increased use of plasmas and ozone in chemical transformation schemes.

Multiphase catalysis, involving gas, liquid, and solid phases, may be key to profitably processing emerging feedstocks such as biomass, CO₂, recycled plastics, and natural gas liquids. For example, catalytic hydroprocessing of biomass to produce fuel and chemical precursors is often performed in a liquid phase (Bagnato et al., 2021). There is growing interest in performing catalytic CO₂ hydrogenation in the liquid phase to produce fuels and chemicals (Mitchell et al., 2019). Similarly, chemical upcycling of plastic waste involves catalytic hydroprocessing in a melt phase (Celik et al., 2019). To understand how the multiphase environment around a catalytic site influences its elementary reaction steps, advanced experimental operando/in situ tools are needed to probe catalytic surfaces in condensed media under pressurized environments. Recent examples include the development of high-pressure operando x-ray absorption spectroscopy and NMR techniques for probing heterogeneous catalytic reaction mechanisms involving conversion of biomass model compounds (Walter et al., 2018).

The broad field of catalysis, encompassing heterogeneous, homogeneous, bio-, electro-, and photocatalysis, will be key to solving many of the challenges related to climate change and ensuring a sustainable supply of energy and materials. Success will require building a bridge between theory and experiments to understand how intrinsic material properties determine catalyst performance: what some people call the quest for the “catalyst genome” (Nørskov and Bligaard, 2013). This will require fundamental chemical research that integrates expertise in materials, measurement, and computation.

Research on catalysis covers broad length and time scales (Figure 4-11). Ideal catalytic conditions on a small scale may not produce the desired product at larger scale. To bridge this “scale gap,” it is vital to consider reaction and process engineering aspects along with life-cycle analysis

during early stages of catalyst and process development (Figure 4-11). This will ensure that energy efficiency is considered across all manufacturing steps, including separation processes.

The urgent need to tackle sustainability-related challenges is driving the need to improve the pace and efficiency of discovering new catalysts and developing new catalytic processes. Toward this end, available materials, adsorption, and reaction data are being harnessed using high-throughput computation and ML to gather further insights and predict new materials (Bo et al., 2018). However, this knowledge extraction approach is challenged by a data deficit, or non-uniform reporting of materials-related data that are not computer readable (Himanen et al., 2019). Wulf et al. (2021) note that “in order to make data widely useful, rather advanced and well-coordinated approaches are needed that are beyond what a single group or institution can develop and sustain.” Digitalization of the catalysis field is essential to “enable efficient data-driven interdisciplinary development of catalysts and catalytic processes” (Wulf et al., 2021). The NFDI4Cat (National Research Data Infrastructure for Catalysis-Related Sciences), supported by the German government, is an example of such a large-scale effort.¹⁰ Establishment of an “internet of catalysis” will guide research along the development chain from molecules to chemical processes. The creation of digital workflows that bridge theory and experimental studies in catalyst design, characterization, kinetics, and related engineering aspects will accelerate discovery and innovation in the catalysis sciences.

4.5 CONCLUSIONS

This chapter explored state-of-the-art tools and technologies in the chemical sciences that are enabling, and enabled by, fundamental research. Advances in measurement, automation, computation, and catalysis have already driven chemical research forward, and there are many examples where advances in these tools have created subsequent advances in chemistry. Though there are a large number of enabling technologies that move chemistry forward, the committee identified these four because they are considered to have the biggest immediate impact and offer enormous potential for the future. The committee also noted some similar needs in these four emerging areas, especially around the use of chemical data. The committee reached a number of conclusions about

___________________

¹⁰ See http://gecats.org/NFDI4Cat.html.

the importance of measurement, automation, computation, and catalysis for enhancing the chemical economy.

Conclusion 4-1: Chemistry is an enabling scientific discipline that will continue to have the largest impact on society when chemists collaborate with experts from other areas such as engineering, biology, physics, computation, and data science to generate new fundamental knowledge and create translational impact at larger scales.

Conclusion 4-2: Measurement, automation, computation, and catalysis are the enabling tools and technologies of fundamental chemical research that will have a substantial impact on both the adoption of novel methodologies and future discoveries in the chemical economy.

Conclusion 4-3: The ability to collect, document, store, share, and use chemistry-related data is needed to advance the use of new tools, such as computation and automation in fundamental chemical research, and increase the accessibility of chemical research to a larger community of practitioners. This information architecture will produce an indispensable tool for the chemical sciences research community to increase the pace and efficiency of innovation by fully harnessing advances made with previous research investments.

In considering each of the four areas outlined in this chapter, important conclusions emerged related to measurement, automation, computation, and catalysis.

Conclusion 4-4: Analytical chemistry will continue to play a substantial role in driving fundamental chemistry and having an important impact on numerous sectors including medicine, environmental and forensic monitoring, and national security, especially as analytical tools increase in speed and accuracy and instrumentation miniaturizes.

Conclusion 4-5: Automation has the potential to change the way chemistry research is designed by increasing the number of syntheses, measurements, or process steps that can be done in rapid succession and by producing data that can be analyzed to help move research forward.

Conclusion 4-6: Fundamental research in computational chemistry is fundamental research in chemistry. A synergistic relationship among advances in computer architectures, computational chemistry algorithms, and the application of computational chemistry enables innovation in all chemistry disciplines.

Conclusion 4-7: Fundamental multidisciplinary research in chemistry, physics, and engineering has played a critical role in the ongoing development of modern computer architecture, and will continue to do so with the continued miniaturization of computers and the emergence of new architectures such as quantum and neuromorphic computing.

Conclusion 4-8: Advances in catalysis remain critically important for driving new products and processes, and the subfields of catalysis covering heterogeneous, homogeneous, biocatalysis, electrocatalysis, and photocatalysis will all play key roles in advancing the fundamental science and technologies needed for making renewable energy, decarbonizing the chemical industry, and promoting a circular economy.