Combinatorial Materials Science for Energy Applications

R. Bruce van Dover
Department of Materials Science and Engineering
Cornell Universitys
Ithaca, NY 14853-1501
Email: rbv2@cornell.edu

Materials are the fundamental basis for solutions to the most pressing issues in energy generation, transport, and utilization, as well as more general issues in sustainability. In many cases, long-term solutions to these problems will depend on breakthrough innovations in materials. As emphasized by the U.S. Department of Energy panel on “New Science for a Secure and Sustainable Energy Future”:

“…existing energy approaches-even with improvements from advanced engineering and improved technology based on known concepts will not be enough to secure our energy future. Instead, meeting the challenge will require new technologies for producing, storing and using energy with performance levels far beyond what is now possible. Such technologies spring from scientific breakthroughs in new materials and chemical processes that govern the transfer of energy between light, electricity and chemical fuels.”1

Can there be an effective strategy for finding breakthrough materials, since they are, by definition, unpredictable?

One answer is found in Combinatorial Materials Science techniques, which represent a powerful approach to identifying new and unexpected materials. The high cost of single-sample synthesis/ characterization, along with the need for reduced research and development time are driving the materials community to explore high-throughput methodologies. Increasing the number of materials that are studied can increase the understanding of composition/ property relationships and the probability of a breakthrough materials discovery. To be effective, this method must balance the need for robust and authoritative measures of properties with the need for rapid characterization.

Figure 1 delineates some of the considerations that must be met for the high-throughput synthesis/characterization approach to be viable. It is not necessary that materials made by the chosen high-throughput synthesis technique be identical to those made in one-off experiments, but it is necessary that the resulting properties are similar. Materials properties depend on morphology, microstructure and other processdependent variables. Thin films are often particularly convenient to work with and sometimes have properties that differ significantly from their bulk counterparts. Fortunately, it is found, in many cases, that compositional trends observed in thin films closely approximate those of the bulk.

figure1

Figure 1. Is the combinatorial/high-throughput approach the right choice to find a new
material that can solve an outstanding problem? This decision tree illustrates some of
the key questions that should be asked. When the right conditions are met, highthroughput experiments can be impressively effective.

Many methods for high-throughput synthesis have been developed,2,3 but the Codeposited Composition Spread (CCS) technique has proven to be an especially versatile method for forming a wide range of compositions in a single experiment. In this method, thin films are deposited by physical vapor deposition on a substrate simultaneously, from two or more sources that are spatially separated and chemically distinct, producing a film with an inherent composition gradient and intimate mixing of the constituents. With three sources, an entire ternary phase diagram may be produced in a single experiment.


Composition spreads may also be synthesized using a traveling shutter4 or shaped mask5 to create a film with a thickness gradient (wedge). A composition gradient can then be obtained by rotating the sample with respect to the shutter and depositing a new overlapping wedge of a second or third film. Atomic mixing is achieved by depositing many wedges each of submonolayer thickness. This approach has the advantage of the composition/position dependence being well-defined (ignoring resputtering effects), although surface reorganization during the short time between wedge depositions can lead to adventitious artifacts.

 

CCS synthesis is distinct from conventional “combinatorial chemistry” as employed, for example, in drug discovery6 as well as the discrete combinatorial synthesis (DCS) approach pioneered by Xiang and Schultz.7 In the latter, mixtures of inorganic components are created by sequential deposition of discrete layers of precursors followed by moderate- or high-temperature diffusion and reaction steps. An important advantage of the DCS technique is that arbitrary compositions with a large number of constituents can be prepared as desired. A key advantage of the composition-spread approach is the opportunity to prepare materials with no subsequent processing, which means that low-temperature or metastable phases may be prepared. The CCS approach also allows properties to be determined with very fine composition resolution, often limited by the resolution of the property measurement itself. For example, it is generally straightforward to sample composition space at 1 mol% intervals, equivalent to investigating thousands of materials in a single experiment. The CCS technique has been used to create both alloys and compounds in chemical systems, including intermetallics, nitrides, oxides, and carbides. A large number of thin film deposition techniques are available for synthesis of composition spread thin films, including evaporation,8,9 sputtering,10-12 pulsed-laser deposition (PLD),13 chemical vapor deposition,14 and cold-plasma processing,15 among others. Of these, sputtering offers a unique combination of advantages:

  • In most cases sputtering rates are constant and reproducible16 when powered by modern dc, rf, or pulsed-dc regulated power supplies, allowing composition spreads with a targeted range of compositions to be synthesized.

  • When magnetron sputter guns are used for sputtering there is little interaction between the sources. That is, the deposition rate from one gun is independent of the operation of the other guns. As a result the composition profile (composition of the codeposited film as a function of position on the substrate) can be predicted quantitatively once the deposition rate for each individual source is determined. However, in some circumstances the magnetron gun may interact significantly with the growing films, leading to resputtering and deviation of the composition from that predicted by simple summation of independent deposition rates. This effect can be predicted quantitatively using known sputter yields and the conditions of the sputter system, leading to accurate adjustments and an excellent quantitative prediction of the composition.

  • The composition gradients achieved by codeposition (using two-inchdiameter sources) are typically about 1 atomic percent per mm, a convenient scale that allows many measurement techniques to resolve trends with about 1 atomic percent resolution.

  • Sputtering is convenient for deposition of metals, oxides (by reactive sputtering in an oxygen-containing ambient), nitrides (sputtering in Ar-N2), carbides (sputtering Ar-CH4), and mixed anion (e.g., oxynitride) systems. Targets of most metals and many unary and binary oxides are readily available, along with many other common materials.

Sputter codeposition also has inherent disadvantages, including:

  • It is not simple to change the composition gradient. The gradient is affected by changing the total inert gas pressure in the system, but many other aspects of sputtering change as well, confounding systematic studies.

  • Targets of the alkali metals, alkaline earth metals, and the lighter lanthanide metals are problematic, due to rapid oxidation/hydrolysis/ hydration of the metal target in air.

Determining the structure of the materials formed in a composition spread is important for understanding composition/property relations. Thin films are well suited for X-ray diffraction studies for phase identification. Using automated data acquisition, in either a conventional stationary-anode diffractometer17 or a synchrotron-based system,18 hundreds of diffraction patterns can be acquired on a single composition-spread substrate. An unsolved challenge is to develop automated techniques for the identification of unique diffraction patterns and clustering these into contiguous phase fields corresponding to regions (compositions) of the CCS film. Some progress has been made in this respect,17,19 but a fully robust algorithm has yet to be developed.

Fuel Cell Studies

Combinatorial techniques have been used in a wide range of energyrelated studies, both to identify new higher-performance materials and to elucidate the composition-dependent properties of known systems. One challenge for which the CCS technique has been found to be particularly well-suited is that of identifying superior electrocatalysts for Polymer Electrolyte Membrane (PEM) fuel cells.20 Catalysis is a complex phenomenon that is essentially impossible to predict with confidence; incremental improvements may be obtained using rational design, but the only way to identify new active compositions is by screening a large number of compositions. Mallouck and coworkers20 developed a viable technique for qualitative optical screening of catalytic activity, which has been further developed into a semiquantitative method.21 Alternative methods for identifying active materials include multiple independent electrode arrangements22 and scanning electrochemical microscopy.23

As an illustrative example, the CCS technique was recently used to gain insight into the most catalytically active material in the Pt-Ta and related systems. After exploring hundreds of chemical systems, the Pt-Ta system was identified as showing interesting activity, and it was selected for close study of the onset potential for methanol oxidation determined by optical fluorescence as a function of position on the binary composition spread. X-ray diffraction data was collected in a high-throughput automated experiment using a synchrotron beamline and used to identify phase fields, as illustrated in Figure 2. It was immediately clear that the best catalyst behavior (lowest value for the half-wave potential E1/2) is strongly correlated with the presence of the orthorhombic Pt2Ta structure, and that the activity is optimized at the approximately stoichiometric composition Pt0.71Ta0.29.24

Figure2

Figure 2. The catalytic activity of Pt-Ta compositions is measured using a fluorescence test: a lower half-wave potential, E1/2, implies greater activity for methanol oxidation. This plot shows that strong activity is associated with the presence of the orthorhombic Pt2Ta phase, is best in the single phase region (18 -35 at% Ta), and is optimized at the
composition Pt0.71Ta.29, close to the stoichiometric value.24

Figure 2 shows that the fine compositional resolution offered by the CCS technique permits two important conclusions about the internal consistency of the data. First, the close agreement between values at adjacent compositions indicates that random variations in the measurement are small compared to the overall trend. Second, the smooth trend with composition in the Pt2Ta phase field allows the optimum composition to be identified with confidence. Internal consistency does not prove that the data is accurate in composition or in E1/2 value; the absolute accuracy must be validated by comparison with detailed one-off studies. The key advantage of the combinatorial method is that it allows the researcher to efficiently identify particular compositions for further study based on data rather than on conservative or radical speculation.

Predicting catalytic properties is not reliable—neither from first principles nor from accumulated experience—so catalyst development has always relied on an empirical approach. High-throughput studies have proven especially useful in facilitating rapid optimization of catalyst composition using factorial designs as well as more exotic approaches, such as genetic algorithms.25 Solution-based synthesis is often employed for optimization because it most closely approximates the processing of realistic formulations. The high-throughput approach has been applied successfully to primary screening and optimization for a wide range of catalyst functions, including polymerization catalysts, enantioselective catalysts, oxidation catalysts, reduction catalysts, dehydrogenation catalysts, and many others.

Transparent Conductivity Oxides

Another example of the usefulness of the combinatorial materials science approach is provided in the study of transparent conducting oxides-materials that serve a wide range of energy-related optoelectronic functions, from low-emissivity window coatings to frontside current collectors in photovoltaics.26,27 Established n-type transparent oxides, such as In1-xSnxOy, SnO2:F, and ZnO:Al, offer adequate performance, but with concomitant drawbacks, such as low thermal stability or high cost. More complex materials based on Ga2O3 (Aldrich Prod. No. 215066), SnO2 (Aldrich Prod. No. 518174), ZnO (Aldrich Prod. No. 204951), CdO (Aldrich Prod. No. 202894), and In2O3 (Aldrich Prod. No. 203424) and their multinary mixtures offer the prospect of improved performance and have been the subject of high-throughput as well as conventional studies.

ZnO is an inexpensive wide-bandgap semiconductor with excellent optical transmission. Highly conductive n-type ZnO has been achieved through doping with Al, In, or Ga. The electrical properties of ZnO are highly dependent on native point defects such as oxygen vacancies, zinc interstitials, and hydrogen, all of which act as electron donors. In order to obtain high conductivity ZnO, these point defects must be deliberately induced by doping. The highest conductivity is achieved after annealing in reducing conditions, while exposure to oxidizing conditions (typically air at moderate temperatures) degrades the conductivity significantly. Extensive studies on the substitution of Al3+ on the Zn2+ site have shown that doping at the level of a few atomic percent leads to a moderately high conductivity that is fairly stable. Doping with In3+ has a similar effect. Since Al3+ has an ionic radius 32% smaller than that of Zn2+, and In3+ has an ionic radius 9% larger, it is not surprising to find that the introduction of either impurity degrades the electron mobility of ZnO. A composition spread study of the conductivity, mobility, and crystal structure of the (Zn, Al, In, O) system allowed the effect of these impurities to be clarified. Figure 3 shows that codoping with both Al and In degraded mobility less than doping with either element alone.28 The lattice strain inferred from x-ray diffraction shows a similar trend; the average lattice constant matches that of undoped ZnO for materials with the highest mobility. The conductivity is also maximum for this condition. Use of a composition spread sample allowed the experiment to be executed without the confounding associated with typical run-to-run variations that accompany one-off studies, thereby allowing robust conclusions to be drawn.

Figure3

Figure 3. ZnO with Al or In forms a transparent conductor. Compared to Zn2+, the Al3+ ion is much smaller while In3+ is much larger. Co-doping yields an “average” dopant size that is a better fit in the ZnO crystal, reducing scattering. A Zn-Al-In-oxide composition spread allows the effect of varying levels of both dopants to be measured. The binary spreads Zn-Al-oxide and Zn-In-oxide are prepared in separate experiments under nominally identical conditions. For a given overall doping level, the highest mobility is observed in co-doped compositions.28

Other Energy-related Studies

Combinatorial materials science techniques have great potential for many other studies of energy-related materials. Thermoelectrics offer the potential to revolutionize technologies such as high energy refrigeration or energy scavenging from low-grade heat sources, a prospect that has driven the search for a breakthrough in thermoelectric materials.29 Many of the ingredients needed for a successful combinatorial search are in place. There is a well-accepted Figure of Merit (FOM) for thermoelectric materials that can, in principle, be measured in a thin film geometry. Recent concepts regarding the factors that might lead to high FOM materials could provide guidance regarding materials systems worthy of investigation.30 Perhaps the most challenging aspect of a search for new thermoelectrics is the sensitivity of the thermoelectric FOM to doping, which implies that a suitable dopant introduced at the optimum concentration must be identified for any prospective candidate. This both increases the number of materials combinations dramatically and introduces the possibility of strong sensitivity to process conditions.

Another example is that of piezoelectric materials, which have a variety of commercial applications and are proving useful in harvesting lowgrade energy. Only a small number of piezoelectric materials are commonly used: others are known but barely characterized. Undoubtedly, many more have yet to be synthesized or characterized. Highlytextured crystalline thin films are needed for reliable characterization of piezoelectric properties, so microstructure control may be a key synthesis issue. Screening approaches should be straightforward, using optical techniques or atomic force microscope (AFM) cantilever probing. New phases, perhaps with a wider processing window or larger piezoelectric Figure of Merit, could be quite exciting, both from a fundamental point of view as well as from the perspective of commercial importance.

Outstanding Issues

While development of synthesis and characterization techniques is basic to the concept of high-throughput experimentation, these steps lead to data and information, not knowledge or insight. In fact, the massive quantity of data, coupled with the modest level of accuracy or precision associated with the speed/quality tradeoff, quickly leads to a new challenge-how to make use of the data. Tracking only the latest topperforming material in a high-throughput investigation discounts the insights and new directions that could be gleaned from lesserperforming materials.

Efficient methods are urgently needed for transforming an overwhelmingly data-rich environment into actionable insights regarding structure-processing-property associations and relationships in materials. Machine learning and statistical approaches, such as cluster analysis and multivariate regression, are important components for this task, but they do not incorporate the relationships that are inherent in the physics and chemistry of materials. Thus, perhaps the most essential difference between conventional research and the combinatorial materials science approach is in the role of informatics—the processing, management and retrieval of information.

Conclusions

The high-throughput approach to inorganic materials discovery improves the likelihood of discovering new materials with useful properties because it dramatically lowers the cost—in terms of financial resources, human effort, and time—of examining unexplored regions of composition space, including regions that might be avoided as unlikely candidates for an expensinve one-off study. Experience has shown that high-throughput screening can be a useful tool for solving real-world problems in materials discovery if three broad criteria are met: In general, the approach is suitable for a well-defined problem for which samples can be prepared using parallelized synthesis and evaluated using a suitable high-throughput screen. High-throughput techniques are also valuable for investigations of known materials systems, where the goal is to elucidate the composition dependence of materials properties. The codeposited composition spread technique has proven particularly effective for exploring energy-related materials in a wide range of investigations.

Materials

     

References

  1. Hemminger, J.; Crabtree, G. W.; Kastner, M. "New science for a secure and sustainable energy future," Office of Basic Energy Sciences, Department of Energy, 2008.
  2. Xiang, X.-D.; Takeuchi, I., Eds. Combinatorial Materials Synthesis; Marcel Dekker: New York, 2003.
  3. High-throughput Synthesis: Principles and Practices; Marcel Dekker: New York, 2001.
  4. Xiang, X.-D. Bull. Am. Phys. Soc. 1999, 44, 103.
  5. Dahn, J. R.; Trussler, S.; Hatchard, T. D.; Bonakdarpour, A.; Mueller-Neuhaus, J. R.;Hewitt, K. C.; Fleischauer, M. Chem Mater 2002, 14, 3519.
  6. Doyle, P. M. J Chem. Tech. 1995, 64, 317.
  7. Xiang, X.-D.; Sun, X.; Briceno, G.; Lou, Y.; Wang, K.-A.; Chang, H.; Wallace-Freedman, W. G.; Chen, S.-W.; Schultz, P. G. Science 1995, 268, 1738.
  8. Kennedy, K.; Stefansky, T.; Davy, G.; Zackay, V.; Parker, E. R. J Appl Phys 1965, 36, 3808.
  9. Hammond, R. H.; Ralls, K. M.; Meyer, C. H.; Snowden, D. P.; Kelly, G. M.; Pereue, J. H., Jr. J. Appl. Phys. 1971, 43, 2407.
  10. Hanak, J. J. J. Mat. Science 1970, 5, 964.
  11. van Dover, R. B.; Hong, M.; Gyorgy, E. M.; Dillon, J. F., Jr.; Albiston, S. D. J Appl Phys 1985, 57, 3897.
  12. van Dover, R. B.; Schneemeyer, L. F.; Fleming, R. M. Nature (UK) 1998, 162.
  13. Lippmaa, M.; Koida, T.; Minami, H.; Jin, Z. W.; Kawasaki, M.; Koinuma, H. Applied
    Surface Science
    2002, 189, 205.
  14. Smith, R. C.; Hoilien, N.; Roberts, J.; Campbell, S. A.; Gladfelter, W. L. Chem Mater 2002, 14, 474.
  15. Terajima, T.; Koinuma, H. Applied Surface Science Proceedings of the Second Japan-US Workshop on Combinatorial Materials Science and Technology 2004, 223, 259.
  16. Gregoire, J. M.; Dale, D.; Kazimirov, A.; DiSalvo, F. J.; van Dover, R. B. Journal of Vacuum Science & Technology A 2010, 28, 1279.
  17. Long, C. J.; Hattrick-Simpers, J.; Murakami, M.; Srivastava, R. C.; Takeuchi, I.; Karen, V. L.; Li, X. Review of Scientific Instruments 2007, 78.
  18. Gregoire, J. M.; Dale, D.; Kazimirov, A.; DiSalvo, F. J.; van Dover, R. B. Review of Scientific Instruments 2010, 80, 123905.
  19. Long, C. J.; Bunker, D.; Li, X.; Karen, V. L.; Takeuchi, I. Review of Scientific Instruments 2009, 80.
  20. Reddington, E.; Sapienza, A.; Gurau, B.; Viswanathan, R.; Sarangapani, S.; Smotkin, E. S.; Mallouk, T. E. Science 1998, 280, 1735.
  21. Gregoire, J. M.; Kostylev, M.; Tague, M. E.; Mutolo, P. F.; van Dover, R. B.; DiSalvo, F. J.; Abruna, H. D. J Electrochem Soc 2009, 156, 160.
  22. Strasser, P.; Fan, Q.; Devenney, M.; Weinberg, W. H.; Liu, P.; Norskov, J. K. Journal of Physical Chemistry B 2003, 107, 11013.
  23. Jayaraman, S.; Hillier, A. C. Journal of Physical Chemistry B 2003, 107, 5221.
  24. Gregoire, J. M.; Tague, M. E.; Cahen, S.; Khan, S.; Abruna, H. D.; DiSalvo, F. J.; van Dover, R. B. Chem Mater 2009, 22, 1080.
  25. Potyrailo, R. A.; Maier, W. F., Eds. Combinatorial and High-Throughput discovery and Optimization of Catalysts and Materials; Taylor & Francis: Boca Raton, 2007.
  26. Ginley, D. S.; Bright, C. MRS Bulletin 2000, 25, 15.
  27. Perkins, J. D.; del Cueto, J. A.; Alleman, J. L.; Warmsingh, C.; Keyes, B. M.; Gedvilas, L.M.; Parilla, P. A.; To, B.; Readey, D. W.; Ginley, D. S. In 2nd International Symposium on Transparent Oxide Thin Films for Electronics and Optics (TOEO-2), 8-9 Nov. 2001; Elsevier: Switzerland, 2002; Vol. 411, pp 152.
  28. Kirby, S. D.; van Dover, R. B. Thin Solid Films 2009, 517, 1958.
  29. DiSalvo, F. J. Science 1999, 285, 703.
  30. Mahan, G. D. J Appl Phys 1989, 65, 1578.

 

Acknowledgments

This work was supported by the U.S. Department of Energy (DOE), Office of Science, Office of Basic Energy Sciences (under Award Number DE-FG02-07ER46440). The author would like to gratefully acknowledge formative discussions with Lynn Schneemeyer, Héctor Abruña, Francis DiSalvo and John Gregoire.

Related Links