The path that chemists typically follow to discover new materials is paved with decades of trial-and-error experiments. Guided by chemical intuition, researchers synthesize substances and then tweak experimental conditions, generating new versions until a material arises with the properties they’re looking for. Nowadays, however, some scientists are trying to cut down on the time, effort, and molecular building blocks consumed during materials discovery by taking an alternative route—one that’s guided by computational chemistry.
To help find better solar-cell components, water-splitting catalysts, and other materials with properties that underpin critical applications, scientists are developing high-level theoretical search methods, commonly based on quantum mechanics calculations. In some cases, the materials pinpointed by these techniques don’t yet exist. In other cases, chemists are using these methods to screen enormous databases of known materials in search of candidates endowed with previously unrecognized combinations of critical properties.
Modern technologies for harvesting, transforming, and storing energy increasingly rely on specific materials with specialized functionality, says Alex Zunger of the University of Colorado, Boulder. “We understand the functionality needed for many technologies, but often we do not have the materials that provide those functionalities,” he adds.
One goal that Zunger, a specialist in computational materials science, and other theoreticians have is to find materials with specified sets of properties to enable new technologies. Another one is to improve existing technologies by finding new materials that can boost device performance, extend lifetime, lower cost, or enable rare elements to be replaced with Earth-abundant alternatives.
“Digging” for new materials with a computer and then attempting to synthesize the winners in the lab offers several advantages over traditional lab-only searches. Computation-first searches can be less expensive, safer, and far broader than standard lab searches, even ones based on combinatorial methods.
And as Stanford University’s Jens K. Nørskov explains, searching with a computer enables scientists to explore the way a material’s properties change as a function of parameters that cannot be controlled experimentally. Nørskov, a leading figure in computational surface catalysis, adds that by using a computer to run these “virtual” experiments and by combing through enormous data sets, researchers can uncover predictive and sometimes hidden trends among classes of materials.
But that’s not all computations bring to the table. Northwestern University’s Kenneth R. Poeppelmeier, a veteran of inorganic synthesis, points out that “closely coupling synthesis with theoretical prediction accelerates each step of the experimental discovery process.” Not only does theory guide experimentalists’ choices of chemical or material targets and synthesis conditions, it speeds analysis and confirmation that the target chemical or material was actually made.
In the past, chemists relied on empirical rules and lab experience to identify chemically plausible targets and rule out a large number of seemingly implausible ones. “But chemical intuition may not always hit the mark,” Poeppelmeier says. And when intuition-driven experiments fail, the experiments may not provide indications as to the source of failure.
That’s why when it comes to searching for new materials, Poeppelmeier stresses the need to tie prediction and validation together. Predictions indicate where to focus lab effort, and experiment results, successful or not, provide feedback for theoreticians—a two-part process he refers to as “closing the loop.”
Given the number of elements in the periodic table, the staggering number of proportions in which they can be combined to form molecules and compounds, and the number of structural variations such as those that define isomers and polymorphs, the range of possible materials is astronomical.
“To search effectively for new materials, we have to navigate a multidimensional landscape of bewildering complexity,” says Aron Walsh, a materials theoretician at the University of Bath. But the motivation for doing so is strong, he adds. And as the stories that follow indicate, researchers are making fast progress in applying these search methods to finding a variety of new materials including solar-cell components, solid catalysts, novel inorganic compounds, and polymeric dielectric materials.
“This is an exciting time for materials chemistry,” Walsh says. The challenge now is not simply to make new compounds, he adds, but to enable new functionality.
As Nørskov puts it, “Computations open totally new possibilities for people who are imaginative and let them test ideas on a timescale that, until recently, you couldn’t even dream about.”
Solar cells are “hot” right now—pun intended—because they can transform some of the nearly limitless power of the sun to electricity and potentially help meet rapidly growing global energy needs. Dye-sensitized solar cells (DSSCs) attract significant R&D interest because they can be made inexpensively and used in devices such as “smart” windows, which admit light to buildings while simultaneously generating power.
The energy-harvesting step in DSSCs begins when sunlight excites electrons in a layer of light-absorbing dye molecules known as sensitizers. The molecules inject the excited electrons into semiconductor particles such as TiO2, to which the molecules are anchored. From there, the electrons migrate to an electrode to produce electric current.
Although DSSCs have been around for some 20 years, today’s versions are not substantially better energy converters than the original ones, which used a ruthenium-based dye. According to Jacqueline M. Cole, a research group leader at the University of Cambridge, “A shortage of suitably efficient dyes is stifling progress in this area.”
Seeking to boost DSSC performance and avoid the expense and toxicity of metal-based dyes, Cole and coworkers decided not to follow the standard approach—repeatedly synthesizing and testing molecular variations of some of the best dyes. That approach has no capacity to reveal entirely new classes of suitable dyes, Cole explains.
Instead, the team devised a set of molecular dye design rules based on structure-property relationships that indicate how the arrangement of molecular groups on the dyes affects DSSC performance. One such rule, for example, calls for separating electron donor and acceptor groups with a π-conjugated unit to enhance the electron injection process.
The team incorporated these rules into a computer algorithm and used it to search more than 118,000 organic molecules in the Cambridge Structural Database, a repository for small-molecule crystal structures. Ultimately, the team identified the best dye candidate—a metal-free, N- and O-containing triaryl member of a class of molecules not previously known to be DSSC dyes.
The group synthesized the compound and a closely related one, tested them in DSSCs, and found that the solar conversion efficiencies are roughly one-third those of the best-performing metal-based dyes (Phys. Chem. Chem. Phys. 2014, DOI: 10.1039/c4cp02645d).
Although these compounds aren’t going to dramatically improve progress in DSSC research in their current form, the team’s results validate the novel application-specific data-mining method applied. They’ve also uncovered a new class of metal-free dyes that Cole says researchers can now optimize by applying chemically intuitive substitution strategies.
According to industry estimates, more than 80% of today’s large-scale chemical processes depend on solid catalysts, which are often based on transition metals. Historically discovered through trial and error, these critically important solids facilitate transformations in petroleum refining, pollution abatement, and production of fuels and chemicals.
The traditional approach to finding new catalysts is starting to yield to computational methods for designing them more rationally. At Stanford University, for example, a research group led by Jens K. Nørskov is developing theoretical techniques for identifying essential catalyst “descriptors.” These are fundamental properties, such as the strength with which molecules bind to a catalyst’s surface.
These binding energies and other parameters strongly, and perhaps unexpectedly, affect how efficiently a catalyst transforms reactants to products and how selectively it forms a target product relative to reaction by-products.
By using this strategy, the Stanford team can uncover predictive trends among large classes of materials and use that information to design new high-performance catalysts.
In one study along those lines, Nørskov, Stanford’s Frank Abild-Pedersen, and coworkers found that a group of nickel-gallium alloys shows promise as catalysts for reacting CO2 and hydrogen under mild conditions to form methanol.
Methanol could serve as a low-cost, sustainable fuel and chemical feedstock, especially if the hydrogen used to make it is supplied via sun-driven water splitting. Currently, methanol is made by using a Cu/ZnO catalyst and a high-pressure mixture of CO, CO2, and petroleum-derived hydrogen. Altering the pressure and other reaction conditions reduces the methanol yield and can lead to a high concentration of unwanted CO.
The Stanford team synthesized and tested several Ni-Ga alloys and found that Ni5Ga3 converts CO2 and hydrogen to methanol at ambient pressure as well as or better than Cu/ZnO and produces lower levels of the by-product CO (Nat. Chem. 2014, DOI: 10.1038/nchem.1873).
In another recent study, the Stanford team, which includes staff scientist Jakob Kibsgaard and chemical engineering professor Thomas F. Jaramillo, searched for new catalysts to facilitate the hydrogen evolution reaction (HER). That reaction electrochemically combines hydrogen ions to produce H2, a valuable fuel. Platinum-group metals are highly active HER catalysts, but they are costly.
So the team used its computational techniques to find substitutes. The researchers predicted that among a series of promising-looking low-cost transition-metal phosphides, the hydrogen adsorption strength (ΔGH) of one of them, Fe0.5Co0.5P, would be optimum for driving HER (Energy Environ. Sci. 2015, DOI: 10.1039/c5ee02179k). Then they confirmed their prediction by synthesizing a series of metal phosphides and conducting catalysis tests.
With the entire periodic table of elements at their disposal, inorganic chemists by now ought to have cobbled together gargantuan libraries of compounds. To be sure, they have synthesized quite a large number of inorganic compounds—but it could be even bigger.
For example, in the family of three-element compounds with 1:1:1 stoichiometry and 18 valence electrons, 483 members are possible, but only 83 of them have been created in a lab.
“That’s astonishing,” says Alex Zunger of the University of Colorado, Boulder. “Why are there 400 missing compounds?” Zunger proposes that the compounds may be missing for a good reason—perhaps they are unstable and decompose. “But maybe there’s no good reason. Maybe we simply have not gotten around to making them yet.”
Considering that some of the known members of this three-element family are thermoelectrics, piezoelectrics, and materials endowed with other technologically valuable properties, it behooves materials chemists to search for the missing ones.
For that reason, Zunger, a materials theoretician, embarked on a chemistry scavenger hunt with Northwestern University inorganic chemist Kenneth R. Poeppelmeier. The researchers used first-principles thermodynamics and other computational methods to evaluate an enormous range of chemical compositions and crystal structures. They concluded that 54 of the 400 missing compounds should be stable. Then they synthesized 15 of those compounds and found that for all of them, the experimentally verified structures matched the predicted ones.
In addition to stability, these compounds had other desirable properties. For example, ZrNiPb is a small-band-gap semiconductor suitable for use in thermoelectrics. And ZrIrSb is a rare example of a transparent positive-charge-carrying (p-type) conductor, a material that could be valuable for display technologies (Nat. Chem. 2015, DOI: 10.1038/nchem.2207). A related study led to the prediction and lab verification of another stable transparent p-type conductor—TaIrGe (Nat. Commun. 2015, DOI: 10.1038/ncomms8308).
Another approach to finding novel inorganic compounds calls for subjecting familiar molecules to extreme conditions. High pressures, as encountered in planetary chemistry, for instance, can be used to access unusual stoichiometries and crystal structures, as well as novel electronic and magnetic properties. A case in point is table salt. An international team reported that high pressures in a diamond anvil cell convert NaCl to oddball compounds such as Na3Cl and NaCl3 (Science 2013, DOI: 10.1126/science.1244989).
“For extreme conditions, we need computational prediction because chemical intuition goes out the window,” says Eva Zurek, a theoretician at the University at Buffalo, SUNY.
Zurek’s group has developed such methods and used them to evaluate the structure and high-pressure properties of several compounds. Among other findings, the studies predict new superconductors.
For example, the group finds that depending on pressure, BaGe3 can adopt three phases, one of which has been synthesized. The other two are predicted to be stable at atmospheric pressure and superconducting, but only at very low temperatures (Inorg. Chem. 2015, DOI: 10.1021/ic5030235). On the basis of a related study, the team predicts that under pressure, H2I and H4I will also be stable and superconducting at low temperature (J. Phys. Chem. Lett. 2015, DOI: 10.1021/acs.jpclett.5b01839).
Superconductors are a small group of materials that conduct electricity without losing energy in the form of heat. Compared with ordinary equipment, superconductor-based gear would, in principle, operate at extreme savings—energetically and financially. Like a large fraction of superconductors, however, the ones uncovered in these studies remain superconductors only at impractically low temperatures. Yet they give researchers clues about where to look for other new superconducting materials.
Electrically insulating organic polymers, also known as polymer dielectrics, play a major role in modern electronics, especially in capacitors. These charge storage devices are found in consumer electronics, electric-grid utility equipment, and hybrid electric vehicles.
The dielectric film at the core of a capacitor serves to insulate and maintain charge separation between electrically conducting plates. How well a capacitor stores energy depends on how well the dielectric can stabilize the charge separation, which in turn depends on its intrinsic properties.
To boost the performance of high-energy-density capacitors, researchers generally use intuition, plus a trial-and-error approach, to look for polymers that may outperform the standard polypropylene dielectric. But that kind of search is slow and limited in scope.
So a team led by Ramamurthy Ramprasad, a materials scientist at the University of Connecticut, is developing a multistep computational search technique to identify promising polymer leads. The method begins with a quantum-mechanics-based combinatorial search to identify molecular repeat units that could be inserted into polymers to give them desirable dielectric properties. The repeat units consist of four blocks selected from moieties such as CH2, C6H4, C4H2S, CO, NH, O, and CS.
Then the method determines stable three-dimensional arrangements of polymers built from the most promising repeat units and evaluates a variety of properties critical to capacitor use, including dielectric constant and band gap.
In a study based on the new computational method, the technique identified a few hundred promising leads, including polyureas, polyimides, polythioureas, and polyamines. Ramprasad notes that one feature common to the most promising candidates is the presence of at least one polar block and one aromatic block.
To validate the method, the team synthesized and characterized three of the top candidates. It found close agreement between the predicted and measured values for structural, electronic, and dielectric properties (Nat. Commun. 2014, DOI: 10.1038/ncomms5845).
Ramprasad points out that the study represents only an initial attempt at using a computer-based method to accelerate identification of promising polymer dielectrics. The group has since extended the technique to include polymers containing metals and other noncarbon species. And now they are working to improve the search method by broadening it to include additional electronic, mechanical, and other properties to more selectively identify promising candidates.