Available online at www.sciencedirect.com

ScienceDirect

Journal of Materiomics 1 (2015) 85e91www.ceramsoc.com/en/ www.journals.elsevier.com/journal-of-materiomics/

Combinatorial screening of thin film materials: An overview

Samuel S. Mao a,*, Paul E. Burrows b

a Department of Mechanical Engineering, University of California at Berkeley, Berkeley, CA 94720, USAb Samuel Mao Institute of New Energy, Science Hall, Shenzhen 518031, China

Received 20 March 2015; revised 13 April 2015; accepted 13 April 2015

Available online 24 April 2015

Abstract

Over the past several decades, technological advancement has grown increasingly dependent on new and advanced materials. Acceleratingthe pace of new material discovery is thus critical to tackling global challenges in areas of energy, health, and security, for example. There is apressing need to develop and utilize high throughput screening technologies for the development of new materials, as material discovery hasfallen behind the product design cycles in many sectors of industry. This article describes techniques of high throughput combinatorial thin filmmaterial growth and characterization developed over the past several years. Although being adopted in selected industries, combinatorialscreening technologies for thin film materials are still in their infancy. Caution must be exercised in selecting relevant combinatorial libraries andextrapolating from small-scale deposition techniques to industrially relevant processes. There are tremendous opportunities in the field ofcombinatorial discovery of thin film materials, as it enters its golden age along with the Materials Genome Initiative, which aims to change thepathway of materials discovery.© 2015 The Chinese Ceramic Society. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: High throughput screening technologies; Thin film materials; Materials discovery

1. Introduction

A great number of disruptive advances in technology overthe past century resulted from the discovery of new materials.Furthermore, many of today's global challenges in energy,health, and security will also require the discovery of newclasses of materials, including multi-component alloys andartificially structured materials or composites that havebecome known as advanced materials. It is becomingincreasingly clear that new materials will be a key drivingforce behind a more competitive manufacturing sector andeconomic growth. In response to this challenge, the MaterialsGenome Initiative is a multi-stakeholder effort involvingseveral funding agencies in the USA [1]. It aims to accelerate

* Corresponding author.

E-mail address: ssmao@berkeley.edu (S.S. Mao).

Peer review under responsibility of The Chinese Ceramic Society.

http://dx.doi.org/10.1016/j.jmat.2015.04.002

2352-8478/© 2015 The Chinese Ceramic Society. Production and hosting by Elsevi

creativecommons.org/licenses/by-nc-nd/4.0/).

the pace of discovery of advanced materials, reducing the timerequired to bring new materials to market by at least 50% fromthe current 10e20 years.

While the Initiative is based on a strong computational andmodeling approach, it is recognized that effective models ofmaterials behavior can only be developed from accurate andextensive sets of data on materials properties. Indeed, theInitiative declares that “In the discovery stage it is crucial thatresearchers have access to the largest possible data set uponwhich to base their models, in order to provide a more com-plete picture of a material's characteristics [1].” It is importantto understand how to best obtain such detailed data sets in thecontext of advanced materials that can be highly complex bothin composition and structure.

In reality, the pace of new materials development remainsslow in virtually all industry sectors. The time frame forbringing new materials to market is typically 10e20 yearsfrom laboratory discovery to first practical use. Fig. 1 showsexamples of the time frame for a few widely used materials

er B.V. This is an open access article under the CC BY-NC-ND license (http://

Fig. 1. Time frame of selected materials from initial research to market [2].

Fig. 2. Total number of experiments required to screen doped TiO2 as in-

dustrial photocatalysts.

Fig. 3. General procedures of high throughput material synthesis and char-

acterization, integrated with theoretical material theory and modeling, and

material database development [8,9].

86 S.S. Mao, P.E. Burrows / Journal of Materiomics 1 (2015) 85e91

[2]. Such a lengthy time frame for materials to move fromdiscovery to market is partly due to the time-consuming, trial-and-error style repetitive synthesis and characterization ex-periments that guide development. A frequently quotedexample is that of Thomas Alva Edison who, it is said, workedthrough thousands of materials in searching for a light bulbfilament more than a century ago, giving rise to the term “anEdisonian approach” to describe a linear process of materialsdevelopment.

In his search for a filament, however, Edison was restrictedto a relatively small number of materials, particularlycarbonized natural fibers [3]. In contrast, modern advancedmaterials are often ternary, quaternary or yet higher order al-loys. Furthermore, the properties of these materials arefrequently modified by the adjustment of, for example, crys-talinity, mesostructure and layering schemes. The total numberof experiments needed to screen materials for modern dayapplications, therefore, can be orders of magnitude greaterthan a century ago. As an example, consider the discovery ofthe photocatalytic splitting of water on a TiO2 electrode [4].This ushered in a new era for heterogeneous photocatalysisbased on semiconductor materials, which has been a subject ofvigorous scientific research over the past 40 years [5,6]. Inaddition to material composition changes through elementaldoping, however, variations in material synthesis also includedifferent synthesis processes such as precursor chemical se-lection, reaction temperature and time. To illustrate the po-tential scale of the problem, consider a thought experiment tooptimize a practical TiO2-based industrial photocatalyst for aparticular application. It is not unreasonable to expect tosearch 50 codopants (to optimize absorption) and 30 co-catalysts (to replace expensive platinum) at, for example, 5and 3 different concentrations, respectively. In addition, pre-cursor chemical selection can influence the end result, as canreaction temperature and reaction time. Adding 2 precursors, 4reaction temperatures and 2 reaction times to the experimentalmatrix leads to 360,000 individual synthesis experiments asshown in Fig. 2 even for this simple optimization matrix. Thisis equivalent to about a thousand man-years of work if onedifferent material is synthesized and characterized every day.

To address this limitation the Initiative includes highthroughput combinatorial material synthesis technology thathas already found applications in industry. In particular, itrecognizes the need for further development of new high

throughput technologies to characterize relevant materialproperties efficiently and quantitatively over a range of oper-ating conditions and environments. Fig. 3 schematically il-lustrates the general procedures of high throughput materialsynthesis and characterization, integrated with theoreticalmaterial theory and modeling and material database develop-ment. Accelerating the fabrication and analysis loop is criticalfor reducing the time to market for new materials. The focus ofthis article is combinatorial thin film fabrication, firstdemonstrated [7] in 1995, which is among the most maturehigh throughput material synthesis technologies that haveyielded new functional materials for a number of applications[8,9].

2. High throughput combinatorial thin film synthesis

Combinatorial synthesis of chemicals was developed in themid-20th Century by Merrifield for the synthesis of peptidechains; work which formed the basis for his Nobel Prize inchemistry [10]. In the 1990s the synthesis technique wasextended to commercial applications for drug discovery. Incombinatorial chemistry, molecules are attached to a solid

87S.S. Mao, P.E. Burrows / Journal of Materiomics 1 (2015) 85e91

support such as a small bead and synthesized step-by-step in aseries of reactant solutions. The molecular building blocks areinitially protected by blocking moieties at all reactive sites. Adesired reaction between reactants in the solution and thesubstrate on the bead can then be controlled by deprotectingthe relevant sites in the correct order. By sequentially sepa-rating and dividing a distribution of beads into a matrix ofreactant solutions, it is possible to synthesize a large library ofstructurally-related molecules in a relatively short time. Inconcept, this is a relatively simple process since the aim isonly to synthesize the correct chemical structure. Some,although not all, combinatorial methods have shown value forthe discovery of new pharmaceutical compounds [11].

With advanced materials, however, combinatorial librariescan potentially be vastly more complex. The correct chemicalstructure may be a starting point, but variations in the crystalstructure, surface functionalization, and (for porous or nano-structured materials) the mesostructure can critically impactthe performance of an advanced material. In this paper, wefocus on the combinatorial synthesis of thin film optoelec-tronic materials, illustrating both the potential and some of thepitfalls of high throughput techniques for advanced materials.

Combinatorial thin film synthesis was developed almosttwenty years ago [7]; it has since been utilized for screeningmaterials for applications such as high critical temperaturesuperconductors and phosphors [12]. Combinatorial thin filmmaterial synthesis is best exemplified by a parallel integrationof thin film fabrication process in the form of spatiallyaddressable arrays of samples, as schematically shown inFig. 4(a). The core of the method is spatially selective depo-sition to create individual combinatorial material librariesusing a combination of designed masks to delineate growthregions with different compositions. It thus allows hundreds ofindividually separated thin film samples to be synthesized ineach fabrication cycle. The first-generation combinatorial thinfilm fabrication systems, however, were not capable ofgrowing thin films with the high crystalline quality requiredfor narrow or wide band-gap semiconductors in a highthroughput fashion. This is largely due to the more criticalrequirement of in situ and post-growth crystalline quality

Fig. 4. (a) Combinatorial thin film material synthesis exemplified by a parallel in

arrays of samples. (b) High throughput semiconductor thin film material screening

control for thin film semiconductors in order to achieve thedesired properties. Equally challenging is the lack of highthroughput tools to measure compositional, structural, optical,and electrical properties of semiconductor thin films grown ina combinatorial library.

We developed a second-generation high throughputcombinatorial semiconductor thin film growth and character-ization platform [8,9], with improved control over crystalinityby controlling the substrate temperature both during and afterdeposition. We also developed a number of high throughputsemiconductor thin film property characterization tools,dedicated to making and screening semiconductor thin filmmaterials for targeted applications. As illustrated in Fig. 4(b),the high throughput material screening process started with thefabrication of a semiconductor material library of an array ofindividually separated thin films. The composition of eachindividual thin film material in the library was measured,followed by measurement of the band-gap. Some of the ma-terials in the library were found to have the desired band-gapvalue for the targeted application, thus down selection wasmade for the next step, which was to characterize the transportproperties of the material. After carrier mobility and lifetimemeasurements, a set of candidate semiconductor materials wasselected for further examination.

For our combinatorial thin film fabrication system, a seriesof four masks with self-similar patterns (Fig. 5(a)) were usedfor making 256 individual materials, each a few hundrednanometers thick, into a 16 � 16 matrix in a single thin filmfabrication experiment. Applying masks sequentially, thin filmmaterials can be selectively deposited in self-similar patternsof quadrants on the substrate, and 44 different samples can beobtained in a library if four different precursor target materialsare used for each mask run, thus there are 4 � 4 growth stepsfor each growth experiment. In situ heating was applied toensure the quality of the thin films, while post-growthannealing under vacuum or in a selected gas backgroundwas used when necessary.

The semiconductor material libraries are fabricated bypulsed laser-assisted epitaxy method in a vacuum of ~1.33 �10�6 Pa. As illustrated in Fig. 5(b), multiple precursor target

tegration of thin film fabrication process, in the form of spatially addressable

process (down selection) [8,9].

Fig. 5. (a) A set of four quaternary self-similar masks used for growing thin film material libraries each consisting of 256 different samples on a single substrate. (b)

Schematic design of the combinatorial semiconductor thin film growth system consisting of three chambers. (c) Photo of a combinatorial semiconductor thin film

library grown on a quartz substrate.

88 S.S. Mao, P.E. Burrows / Journal of Materiomics 1 (2015) 85e91

materials are mounted on a rotating target holder during thedeposition process to provide composition variations in thematerial libraries. An excimer laser operating at a wavelengthof 248 nm, pulse energy on the order of 1 J, and a repetitionrate of 10 Hz is used for vaporizing the prefabricated targetmaterials. The resulting library consists of 256 thin film ma-terials each having an area of 800 mm � 800 mm, deposited ona quartz or single crystal silicon substrate. Laser epitaxy has anumber of advantages over other thin film fabrication tech-niques, such as vaporization at relatively low temperature,maintenance of stoichiometry of the target material indeposited films, and flexibility toward alloying using multipletarget precursors. To achieve homogeneous thin films in asequential multi-target deposition process, the substrate has tobe maintained at an elevated temperature to ensure optimizeddiffusion, whereas each deposition is limited to yield athickness of approximately a monolayer. Fig. 5(c) is photo of acombinatorial semiconductor thin film library grown on aquartz substrate.

3. High throughput combinatorial thin filmcharacterization

Characterization of semiconductor thin film libraries in ahigh throughput fashion is the key to realizing the full po-tential of the combinatorial thin film material discoverytechnology. We developed a number of high throughput thinfilm characterization techniques for measuring composition,microstructure, electronic band-gap, and carrier mobility andlifetime of thin film material libraries.

Fig. 6(a) shows schematically the principles of highthroughput thin film composition and microstructure charac-terization. The composition and microstructure of the materialon each semiconductor thin film library are determined by anintegrated micro-beam X-ray fluorescence (XRF) and

diffraction (XRD) system operated in a two-dimensionalautomatic scanning mode. Fig. 6(b) shows an example ofcolor-coded material compositions (relative concentration ofCe) from a thin film metal oxide library made from indiumoxide, tin oxide, zinc oxide, zirconium oxide, and ceriumoxide targets, which may find applications visible-lighttransmitting solar-heat reflective films.

With the second-generation combinatorial material dis-covery technology, we have also explored a number of thinfilm semiconductor materials for applications ranging fromtransparent conducting oxides to room temperature radiationdetection. For example, we fabricated and characterized anumber of material libraries based on oxide, telluride, andselenide semiconductors for gamma- and x-radiation detec-tion. For these detectors, which are important to medical, in-dustrial, security and laboratory applications, currenttechnology uses cadmium zinc telluride (CZT), which directlyconverts X-ray or gamma-ray photons into electrons. Unlikesilicon and germanium detectors, CZT can operate at roomtemperature and has a higher resolution than commerciallyavailable scintallators. In order to achieve this, a semi-conductor material must have an appropriate band gap toensure low radiation detector leakage current. Furthermore,the product of the carrier mobility and charge carrier lifetimemust be high to allow efficient charge collection. Finally, el-ements with high atomic number (high Z) or large neutronabsorption cross-section are desired. It is, however, difficult toproduce high quality crystalline CZT that has the desiredcarrier transport properties due to, for example, large sizedifferences of atoms. To search for a possible replacement, wecreated a ternary library of Ga, Ag and Te. We used GaTe andelemental Ag2Te targets in the system described above togenerate a library of GaxAgyTe1-x-y materials for rapid char-acterization, where the typical growth temperature was 650 �Cwith overnight annealing at 500 �C.

Fig. 6. (a) Schematic illustration of high throughput composition and microstructure characterization based on micro-beam X-ray fluorescence and diffraction. (b)

Example of color-coded thin film material compositions (relative concentration of Ce) from a thin film oxide library.

89S.S. Mao, P.E. Burrows / Journal of Materiomics 1 (2015) 85e91

To characterize the compositional library at a pacecommensurate with the deposition speed, we also designedtwo high-throughput tools to measure the bandgap, carriermobility and charge carrier lifetime. To measure the opticalband-gap, the transmittance (T ) and reflectance (R) of indi-vidual materials on the thin film library deposited on a quartzsubstrate were recorded simultaneously using a scanningmicro-beam optical spectrometer shown schematically inFig. 7. The absorption coefficient, a, was calculated ac-cording to, a ¼ (1/d )ln[T/(1�R)2], where d is the thicknessof the thin film material. The optical band gap was thencalculated from the dependence of a on photon energy.Fig. 7(a) shows schematically the technique of highthroughput optical characterization (optical transmission),

Fig. 7. (a) Schematic illustration of high throughput band-gap characterization ba

oxide thin film library.

Fig. 8. (a) Schematic illustration of thin film material transport property characte

mobility data plot of a sulfide thin film material library.

and Fig. 7(b) is an example of band-gap measurement for anoxide thin film library.

For charge carrier (electrons and holes) transport proper-ties, particularly the carrier mobility (m) and lifetime (t), wedevised an ultrafast photo-induced charge probing method tomeasure m and t of the library. Two co-planar electrodes weredeposited on the sides of each sample in the thin film materiallibrary via evaporation. When two computer-controlledconductive probes are in contact with the deposited elec-trodes, a circuit loop is formed. As the thin film is irradiated byan ultraviolet femtosecond laser pulse (100 fs, 266 nm), atransient photocurrent is produced and recorded by the digitaloscilloscope. The charge collection transient as a function ofbias voltage can be fit to a model with the charge carrier

sed on optical transmission. (b) Example of color-coded band-gap map of an

rization based on ultrafast pulsed laser excitation. (b) An example of carrier

Fig. 9. Band gap measurement of GaAgTe with varying concentration of Ga

and Ag, measured in a single experiment from a combinatorial library.

90 S.S. Mao, P.E. Burrows / Journal of Materiomics 1 (2015) 85e91

mobility and lifetime as parameters as described previously[9]. Fig. 8(a) illustrates schematically the semiconductortransport property characterization approach based on ultrafastpulsed UV laser excitation. Fig. 8(b) shows carrier mobilitydata plot of a sulfide thin film material library.

In the x-radiation detection library, we found that certaincompositions of GaAgTe yielded suitable electronic band gaps(between 1.9 and 2.4 eV) as shown in Fig. 9. Downselectingthe appropriate subset of materials with the correct bandgap,we found that some compositions also have good electronmobility-lifetime product values (on the order 10�4 cm2/V),comparable or better than that of CdZnTe crystals currentlyused in radiation detection industry [9]. Further optimizationof these compositions by exploring the dimensions of substratetemperature and growth rate is the subject of ongoinginvestigation.

We caution, however, that the properties of thin films grownusing techniques such as laser-assisted epitaxy or pulsed laserdeposition, while ideal for building combinatorial libraries, arenot always replicated by thin films of identical compositiongrown by more scalable techniques such as chemical vapordeposition or magnetron sputtering. This is of particularconcern in the field of thin film electronic devices, where smalldifference in thin film morphology can cause significant de-viations in device performance. For example, a range ofcompositions of ZnO doped with Ga was previously investi-gated [13] as a potential transparent conducting anode fororganic light emitting devices (OLEDs). The aim here was tofind replacements for indium tin oxide (ITO); replacing theexpensive and limited abundance In with lower cost, moreabundant materials for high volume applications such as solid

Fig. 10. Summary of electrical properties for TCO anodes and

state lighting. Indeed, compositions of Ga:ZnO were foundfrom a pulsed laser deposition library which formed anodecontacts to OLEDs with an external quantum efficiency andoperating voltage as good as or better than commerciallyavailable ITO. The composition with 2% doping of Ga gaveparticularly good results (Fig. 10). However, deposition offilms of identical composition over large areas by rf-magnetron sputtering did not produce working OLEDs. Thereasons for the differences, likely due to film microstructure orinterfacial properties, will be the subject of futureinvestigations.

4. Conclusions

By developing combinatorial thin film material fabricationand characterization methodology, we have been utilizingcombinatorial material screening approach to new materialsdiscovery featuring a deliberate and systematic highthroughput exploration of the composition-structure-propertyrelationships of new thin film oxide, telluride and selenidematerials. Recent development of a second-generationcombinatorial semiconductor discovery technology hasenabled the creation of arrays of individually addressable thinfilm materials with the possibility of controlling film growthconductions such as temperature and background pressure,which would benefit to a greater range of applications.

The high throughput thin film growth and characterizationapproaches discussed in this overview article are importantnew tools. Firstly, they enable rapid screening of new prop-erties in a parallel “hyper-Edisonian” approach, as illustratedby the selection of a GaAgTe composition as a candidate forradiation detection. Here, high-throughput characterizationtools, capable of quickly scanning a library for one or moredesired physical properties, are just as important as high-throughput synthesis. Secondly, they can generate extremelylarge data sets from multiple, systematically varying parame-ters. Such data is critically important for validating computa-tional results from multi-scale models. Effective models ofmaterial behavior can only be built on accurate and extensivesets of experimental data of materials properties, rapidlypopulating material property databases, validating key pre-dictions, and extending the range of validity and reliability ofthe models. This is a key objective of the Materials GenomeInitiative.

For applications requiring particular bulk optical properties,combinatorial synthesis is indeed an ideal discovery tool. Foradvanced optoelectronic or catalytic applications, whereinterfacial chemistry and mesostructure can dominate func-tionality, current generations of combinatorial synthetic tools

resultant OLED transport properties. Data from Ref. [13].

91S.S. Mao, P.E. Burrows / Journal of Materiomics 1 (2015) 85e91

are still inadequate to fully explore materials space and somedegree of linear investigation remains necessary. In particular,thin films deposited using one technique may not mirror theproperties of films with the identical chemical compositiondeposited using another, more scalable technique. This reflectsthe extreme complexity of advanced materials, and givesmotivation to explore further developments in the field.

Acknowledgments

This work has been supported by the U.S. Department ofEnergy, and the Shenzhen Council of Science, Technology,and Innovation. The author acknowledges Z. Ma, P. Xiao, H.Hao, D. Liu, X. Zhang, L. Oehlerking, D. Speaks, K.M. Yu, W.Walukiewicz, and P.Y. Yu, for their contribution in variousstages of the research.

References

[1] Materials Genome Initiative for Global Competitiveness, National Sci-

ence and Technology Council, 2011.

[2] Eagar TW. Bringing new materials to market. MIT Technol Rev

1995;98:42.

[3] Thomas Alva Edison, U.S. Patent 223898 (1889).

[4] Fujishima A, Honda K. Electrochemical photolysis of water at a semi-

conductor electrode. Nature 1972;238:37.

[5] Chen X, Mao SS. Titanium dioxide nanomaterials: synthesis, properties,

modifications, and applications. Chem Rev 2007;107:2891.

[6] Chen X, Shen S, Guo L, Mao SS. Semiconductor-based photocatalytic

hydrogen generation. Chem Rev 2010;110:6503.

[7] Xiang X-D, Sun X, Briceno G, Lou Y, Wang K-A, Chang H, et al. A

combinatorial approach to materials discovery. Science 1995;268:1738.

[8] Mao SS. High throughput combinatorial screening of semiconductor

materials. Appl Phys A 2011;105:283.

[9] Mao SS. High throughput growth and characterization of thin film ma-

terials. J Cryst Growth 2013;379:123.

[10] Merrifield RB. Solid phase peptide synthesis. I. The synthesis of a tet-

rapeptide. J Am Chem Soc 1963;85(14):2149.

[11] Ecker DJ, Crooke ST. Combinatorial drug discovery: which methods will

produce the greatest value? Nat Biotechnol 1995;13:351e60.

[12] Koinuma H, Takeuchi I. Combinatorial solid state chemistry of inorganic

materials. Nat Mater 2004;3:429.

[13] Berry JJ, Ginley DS, Burrows PE. Organic light emitting diodes using a

Ga:ZnO anode. Appl Phys Lett 2008;92:193304.

Samuel S. Mao, After receiving his Ph.D. degree

from the University of California at Berkeley in

2000, Samuel S. Mao started his career at Lawrence

Berkeley National Laboratory, where he remained as

a career staff scientist until 2013, when he launched a

private institution to promote international collabo-

ration and global commercialization of energy and

environmental technologies. He has been an adjunct

professor at the University of California at Berkeley

since 2004. Having published 130 research articles

that have received more than 22,000 citations, he is

also an inventor of 30 patents in the U.S. and abroad. He delivered nearly 100

invited, keynote or plenary speeches at international conferences, co-chaired

Materials Research Society (MRS) annual meeting in the spring of 2011, and

the International Conference on Clean Energy in 2012. He received a “R&D

100” Technology Award (2011) for his technological innovation and a Ber-

keley MEGSCO Faculty Teaching Award (2008) for his dedication to higher

education.

Paul E. Burrows obtained the Ph.D. degree in

Physics from Queen Mary College, University of

London, in 1989. From 1990 to 1991 he was a

Research Scientist in the Frontier Research Program

at the Riken Institute in Japan. He also held research

appointments at the University of Southern Califor-

nia and Princeton University, where he was a

Research Scholar from 1997 to 2000. He subse-

quently joined Pacific Northwest National Labora-

tory as a Laboratory Fellow in the Energy Science

and Technology Directorate where he won a Federal

Laboratory Consortium for Technology Transfer award for excellence in

technology transfer in 2002. He founded Reata Research In 2008, and in 2013,

he was appointed Senior Vice President for Research at the Samuel Mao

Institute for New Energy, Shenzhen. He has published over 110 publications in

refereed journals and is a co-inventor of 104 issued U.S. Patents in the field of

thin film technology.