INSTITUTE OF PHYSICS PUBLISHING
MEASUREMENT SCIENCE AND TECHNOLOGY
Meas. Sci. Technol. 16 (2005) 309–316
doi:10.1088/0957-0233/16/1/041
Elements of informatics for combinatorial solid-state materials science S Meguro1,3 , T Ohnishi1, M Lippmaa1 and H Koinuma2,4 1 Institute for Solid State Physics, University of Tokyo, 5-1-5 Kashiwanoha, Kashiwa 277-8581, Japan 2 Materials and Structures Laboratory, Tokyo Institute of Technology, 4259 Nagatsuta, Midori-ku, Yokohama 226-8503, Japan
E-mail:
[email protected]
Received 21 April 2004, in final form 22 October 2004 Published 16 December 2004 Online at stacks.iop.org/MST/16/309 Abstract The main purpose of using combinatorial techniques for materials science studies is to achieve higher experimental throughput than what is possible when samples are synthesized and characterized one at a time. The instrumentation needed for performing high-throughput synthesis and characterization has seen rapid development in recent years. The software tools needed to connect all parts of the materials development process are still largely lacking. In this paper we discuss the requirements of a combinatorial informatics system for materials science experiments. Specifically, we focus on solid-state thin film synthesis. We also describe an implementation of such a system that is based on widely-available open-source software. The system offers features such as remote access via a Web browser, an electronic notebook-style Web interface, automatic upload of new measurement or processing results and rapid preview of experimental data. Keywords: combinatorial synthesis, thin films, informatics, database, web
interface, high throughput
1. Introduction Informatics is an integral part of the combinatorial or high throughput materials development process. The general purpose of using combinatorial techniques for solidstate synthesis is to expedite the process of creating and characterizing new material compositions or structures [1]. In most cases, obtaining high throughput in experiments is the most important goal, because it improves the odds of making serendipitous discoveries [2] in exploratory research and brings greater efficiency to optimizing or mapping the properties of known material families [3]. These benefits, either for pure scientific research or for commercial research and development, can only be realized if the information processing is as efficient as the production and characterization of samples. The informatics aspects of high-throughput 3
Permanent address: Nanomaterials Laboratory, National Institute for Materials Science, 1-1 Namiki, Tsukuba 305-0044, Japan. Also at: CREST, Japan Science and Technology Agency.
4
0957-0233/05/010309+08$30.00
experiments are often overlooked, and so far the available tools are far less developed than the hardware that is available for materials synthesis and characterization. A diagram that illustrates the basic steps of the combinatorial materials development cycle is shown in figure 1. Although the diagram refers specifically to combinatorial libraries, the same basic steps are followed when working with single samples. The biggest differences appear in the library design and visualization or analysis stages. Much larger volumes of data are produced rapidly in combinatorial experiments, complicating data visualization and analysis. A data management system is therefore needed to link together the various stages of the materials development process. An informatics system for combinatorial materials science can address a number of different problems. At the most basic stage, a system is needed for collecting, storing, distributing and presenting raw or processed data. Higher-level functions can also be added, such as integration of various data sources, e.g. existing materials databases, complex data processing and visualization, project management, etc. In this context,
© 2005 IOP Publishing Ltd Printed in the UK
309
S Meguro et al
Library design
Selection of processing conditions
Fabrication
Web application HTTP
Database Storing results
Web server
Characterization
HTTPS Firewall
User access Analysis
Visualization
Data collection
Instrument access Analysis tools
Results
Figure 1. The basic steps of a combinatorial materials development cycle. The informatics system provides data for or obtains data from each step in the process.
informatics is understood to mean a collection of software tools for handling diverse forms of information. The informatics tools described here are being designed to handle the experimental data obtained from thin-film experiments within the COMET collaboration [4]. It covers most of the basic functions that are necessary for running high-throughput experiments, namely data collection, storage, searching and distributed access in the form of an electronic notebook. Similar projects are also underway at other research laboratories, specifically aimed at polymer synthesis [5] and glass synthesis [6]. The software tools described in this paper handle the most basic subset of informatics functions, namely the database, a Web front end and interface functions for linking the system to automated data collection devices. Other informatics tools that perform more advanced functions all rely on this basic functionality, while the specific software, e.g. library design and data analysis, is being developed by other research groups. The main development goal is to minimize user intervention and ensure that as much of the experimental data as feasible is moved directly from instruments to the database server. The role of the database server is to store the data and to update in real time an electronic notebook, detailing the progress of each experiment and providing a quick graphical preview of all acquired data. Further analysis, processing and visualization tasks can then be built on top of this data handling system.
2. Main functions of the informatics system 2.1. Data management The basic structure of the informatics system is shown in figure 2. All data are accumulated in a central relational database. The database is not directly accessible by any of the users. Instead, the users and other applications communicate with a web-based interface of a web application. The web application communicates with the database, storing or retrieving the necessary information. The advantage of a centralized database lies in the ease of maintenance, searching and backup of data. Users who wish to view or add data to the system can do so using a Web-based electronic notebook interface, or 310
Internet
Figure 2. The structure of the informatics system. A relational database stores all the information, but is not directly accessible to users. All data entry and retrieval from the database is mediated by a Web application layer, which talks to client programs through public networks over secure sockets layer connections. This allows the system to operate in an otherwise firewalled environment.
by using dedicated data access tools. Such tools would usually be installed at each instrument that needs to perform automatic data transfers without direct user intervention. Data visualization and analysis tools also access the data using the Web interface. Most of the logic of the informatics system is contained in the Web application layer and in the database itself. 2.2. Data access Many combinatorial experiments rely on a collaboration of different research groups, some specializing in synthesis, others developing novel analysis instruments. The informatics system therefore needs to function also as a communication tool between collaborating groups. Data that have been stored in the database must be accessible over public networks. This presents a challenge, because data integrity and safety must also be guaranteed while opening the informatics system to the Internet. These goals can be achieved by using a Web-based front end, as shown in figure 2, where the standard Web-based tools are used for user authentication and the HTTPS protocol handles data encryption. Information stored in the database can be accessed in several ways, depending on the type of use. A human user would usually wish to search or browse the data stored in electronic notebooks using a simple web browser. Automated tools also need access to the database, either to retrieve information about a particular library, or to store measurement or processing results into the database. This functionality is also handled by using the same Web interface as a form of a web service. Regardless of the method or purpose of data access, in each case the network session between the data client and the database system occurs over an encrypted link using secure sockets layer (SSL) connections. A user or program would first log into the system with a user name and password or a digital certificate, obtaining access to data that belong to a particular user. The necessary capabilities are built into most existing web browsers and can be added to other software tools using existing communication libraries. Conventional
Elements of informatics for combinatorial solid-state materials science
(a)
(a)
C (d )
(b)
(b)
(e)
A0 (c)
Figure 3. Types of combinatorial solid-state libraries: (a) individual cells in a two-dimensional array, (b) composition gradient library, (c) electronic device library, (d ) one-dimensional library, (e) individually addressable bulk synthesis library.
access control lists have also been implemented to allow only selected users to modify or view data belonging to specific experiments. Several groups can thus use a single server for data storage and access.
3. Functionality of the informatics system 3.1. Library design Combinatorial library design involves two major steps: selection of the physical library layout and the composition or structure of the materials in the library. The physical layout of a library depends on the capabilities of the synthesis and characterization equipment. Some examples of solid-state libraries are shown in figure 3. Thin film libraries offer a high level of freedom in selecting the library type, i.e. individual cells, composition spreads or device libraries can all be fabricated in the same deposition system by simply selecting a suitable set of masks. Linear capillary-based libraries [6] or crucible sets for bulk synthesis [7] do not have this flexibility. The shape, size and layout of a library would usually be mapped onto a certain phase diagram that is being studied. An example of how parts of a ternary phase diagram might be translated into a composition spread or individual cell library is shown in figure 4. While binary and ternary phase diagrams are relatively easy to map to a two-dimensional library, this task becomes gradually more difficult as the number of dimensions in the phase diagram increases. Since the purpose of combinatorial synthesis is to explore complex materials with many elements, this is a problem that needs to be solved by the informatics system when translating the phase diagram points into detailed mask or target movement commands for the thin film deposition systems or into robot movement commands for bulk synthesis [8]. In order to do this, the informatics system has to present the user with information on the capabilities of a particular deposition system, deposition rate data and available mask patterns. The task of the informatics system is to simplify the library design and also to store the design parameters in a database
10
20
30
40
50
60
70
80
90
100
B
Figure 4. Mapping a ternary phase diagram onto a practical library. Composition gradient libraries (a) can map out sections of the phase diagram [9]. Two-dimensional arrays (b) can map out arbitrary sets of points selected from the phase diagram.
for later reference. The library structure is later used during characterization and data analysis steps for mapping the library characterization results onto the original phase diagram. Another function of the informatics system is to provide access to reference data for the constituent materials in a library. The precise needs vary, depending on the synthesis technique. For thin film pulsed laser deposition experiments, for example, it is necessary to know the deposition rates and deposition profiles for all target materials in order to select the appropriate pulse laser energy, repetition rate and pulse count for each mask or target position. 3.2. Processing parameters Each synthesis technique has a set of process parameters that are common to all parts of a combinatorial library. In many cases, for example, the processing temperature is equal in all parts of the library, unless deliberate temperature gradient experiments are used [10]. Another parameter that usually does not vary across a library is the ambient gas pressure and composition. The role of the informatics system is to provide reference data for deciding what the appropriate values might be in each case. Since these values are specific to a certain synthesis machine, the necessary data must be acquired in separate experiments before library synthesis can begin. 3.3. Library fabrication A common feature of all combinatorial synthesis systems is a high level of automated control for manipulating samples, robotic manipulators, ablation targets, masks or other components. The detailed sequence of movements or other control commands are generated in the library design step. The procedure can involve automated tools, but for moderately small libraries the synthesis sequences are often generated by hand. It is desirable for the information management system, however, to store the actual sequence that is used during the synthesis. Any monitoring data gathered during the synthesis procedure, e.g. true sample temperatures, actual chamber pressures, etc, can also be stored in a central database for later access. A particularly important part of the synthesis step is to establish reference points for library position calibration [11, 12]. In the case of individual samples, as shown in figure 3(e), the only issue is to identify a particular set of 311
S Meguro et al (a)
(b)
markers
Figure 5. Positioning of marker points at the edge of a library. The markers are deposited using the same masks that are used to define the actual cells in the library. By determining the positions of the markers during sample characterization, it is possible to map the characterization results onto the actual library cells and from there onto the original phase diagram.
samples. The measurement system only needs to know how large the samples are. Thin film libraries present a number of additional difficulties. Films are deposited in a vacuum chamber using contact or non-contact masks. It is difficult and in practice usually impossible to position the sample accurately in the deposition chamber. This means that the actual position of the library varies from sample to sample. The solution to this problem is to deposit a set of markers on the sample using the same masks that are used for defining the geometry of the library (figure 5(a)). The material, shape and position of the marker need to be chosen carefully, depending on which types of characterization techniques will be used later. For example, thin films are often transparent, the markers should be clearly visible in an optical microscope but also during x-ray diffraction mapping [13], in a microwave or SQUID (superconducting quantum interference device) microscope, etc. A metal layer, such as gold, is often a good candidate material. The marker shape should be such that it defines clearly the position of the library origin and the direction of x and y-axes. 3.4. Characterization and data collection Characterizing the physical properties of combinatorial libraries presents some of the greatest instrumentation challenges of the whole materials development process. Many of the instruments designed for high-throughput characterization of solid-state samples perform measurements in parallel, e.g. using image sensors, or by rapidly scanning a probe over the library surface. There are three different measurement geometries that are commonly met in combinatorial experiments, shown in figure 6. Imaging or scanning type instruments produce a two-dimensional array of data points (figure 6(a)). Each measurement point can produce more than one value. A microwave microscope, for example, gives a Q-value and a frequency shift for each point. Other instruments, such as profilometers or concurrent x-ray diffractometers, produce line sections through a library, as shown in figure 6(b). Most microsampling techniques, such as atomic force microscopy (AFM), or cross-sectional transmission electron microscopy (TEM) provide essentially single-point measurements at predetermined positions in the library (figure 6(c)). The selection of point positions can be automatic if an x–y stage is included with the instrument. However, many instruments, e.g. AFM, require manual sample 312
(a)
(b)
(c)
Figure 6. Different characterization instruments use different scanning techniques: (a) raster scans, i.e. optical camera-based measurements and library-scale scanning probe measurements, (b) linear section scans obtained from x-ray mapping or other slow scanning instruments, and (c) individual point measurements, such as narrow-scan scanning probe measurements or rapid-sampling transmission electron microscope analysis.
positioning before a scan of a small area can be made. The same limitations apply for the extraction of TEM samples by focused ion beam cutting [14]. The informatics system needs to perform two basic functions during sample characterization. The user performing the measurements needs to know the sample geometry and processing conditions. Those data can be obtained from the database that stores the library design and processing data from previous steps. A particularly important point is to note the location and type of library position calibration markers. During measurement it is critical to ensure that the markers are visible in the measured data set in the case of twodimensional scans. The marker positions can then be used to locate the correct points in the library even when there are small positioning errors, as illustrated in figure 5(b). Other positioning techniques have to be used for one-dimensional sections through the library or single-point or local-probe measurements. In the worst case, when the markers cannot be included in the measured data automatically, the operator at least needs to mark and store information regarding the location of each scan in the database. The second function of the informatics system is to transfer the actual measurement data into a centralized database, processing the raw data so that they are suitable for simple preview and on-line distribution is possible. Many of the instruments used for characterizing thin film libraries in particular are unique or tied to a specific location. Examples are TEM and synchrotron radiation sources. If such methods need to be used to characterize the material properties in a library, the samples have to be transferred from one location to another, i.e. from a synthesis lab to the characterization lab. This presents a problem for actually achieving high throughput in the materials development cycle. Rapid transfer of results between different laboratories via a centralized database is therefore essential. The greatest challenge is to link these measurement instruments into an integrated informatics system [5]. This task is relatively straightforward in cases where the software that controls the instrument has been developed in house. Commercial measurement equipment that is used or adapted for combinatorial library characterization is often driven by proprietary software, which cannot be modified to work in an automated data collection system. Most software packages only allow manual saving of data to disk. In many cases the details of the format used in raw data files are not readily available either. These issues can be solved by collaboration
Elements of informatics for combinatorial solid-state materials science
with equipment vendors or by using automated operating system level tools that work together with the prepackaged commercial software. Regardless of which particular solution is used, the purpose is to move measurement data immediately into a database with minimal user intervention. The purpose of the database system is to improve efficiency of the measurement process. Automatic data transfer from the measurement systems to the database is a key to achieving this efficiency by allowing the operator to concentrate on the experiment, rather than the logistics of converting data formats or manually transferring data between computers. 3.5. Visualization and analysis A variety of general-purpose visualization packages and specific purpose-built tools exist for visualizing multidimensional data. A general Web interface is not particularly suitable for this purpose because analysis is an inherently interactive process and would require client-side browser plug-ins or local software. A purely server-side solution would not be acceptable due to the large round-trip time between a user event and display updates. In general, the visualization, data processing and data mining tasks are the most difficult from the point of view of software development. At this stage we are not dealing with these issues directly. Instead, we allow easy and automated access to the stored information so that external applications can be integrated into a common workflow. For example, the only visualization task that the informatics system itself needs to handle internally is to generate graphic previews of uploaded raw or processed measurement data. In the present design we accommodate characterization equipment that produces data in various formats by writing small programs that upload these raw files automatically into the informatics system. It is critically important for the users to access and browse experimental results as quickly as possible and therefore desirable that the web application would automatically create graphic views of the uploaded data. These functions have a certain amount of overlap with the tasks performed by the analysis and visualization software, although the preview functionality is more restricted in terms of the features that it offers. The automatic preview generation should be able to produce a reasonable representation of a wide variety of different data sets without having detailed knowledge of what the data mean. The preview images are automatically added to an electronic notebook that holds all the data for a particular sample. 3.6. Database design The database has to store many different forms of experimental data. In addition to being a simple repository for raw measurement data, the database also holds information on the identities and access rights of users, the structure and content of electronic notebooks and the results of processing the analysis results. The types of files that need to be stored and processed by the data management system also change constantly, as new instruments are added. The database structure thus needs to be very flexible. Conventional relational databases store numeric or text data in table form. The databases can be queried using the
structured query language (SQL), which gives a standard way to access the data. The database itself manages queries, indexing, concurrent accesses, transactions, etc. The table-oriented structure of the database, however, is not particularly well suited for storing variable data, such as library descriptions or measurement results, particularly because it is impossible to define a rigid data structure during system design. As new types of libraries are designed and new measurement instruments become available, new data formats have to be handled by the same basic management and database system. A solution to this problem is to store the variable part of information in a general-purpose format, such as the extensible markup language (XML) [15]. A database entry for an AFM image, for example, would be divided into two parts, one containing the actual measured values and the other holding metadata, i.e. measurement parameters. An AFM entry could be structured as shown below. TY277 37763 friction 2E−6 256 256 60 21-10-2003 12777 12792 12922 13144 . . . The metadata part can be searched and retrieved by a variety of XML processing tools that work together with a traditional relational database. Various forms of indexing can be used to improve the efficiency of search functions even when large numbers of such XML entries need to be processed. The XML entries can easily be converted into a suitable form for display in the Web-based electronic notebook environment.
4. An implementation example A test bed Web-based data management system has been built, based on the characteristics outlined above. The system offers the following features: secure data transfer over SSL connections, user authentication by conventional user name and password or a client certificate, user grouping and access control lists for all data items. The web interface is also responsible for generating data preview images and an electronics notebook representation of the stored data. Our implementation is based mostly on open-source software, e.g. the Apache Web server [16] and the PostgreSQL relational database [17], running on a Linux server. The electronics notebook style web pages are dynamically generated using the PHP scripting language [18], which offers a high-level programing environment. The PHP scripts also handle the gateway functions to the database. All data are stored in the 313
S Meguro et al
AFM
PC transfer
Probe
WEB browser LabVIEW + libcurl library
Server Upload
Upload
Receive Analyze Generate View
Figure 8. The structure of a typical data upload tool. Data from the measurement instrument, e.g. AFM, are usually saved locally. They can then be transmitted to the database system either via a Web browser or by using an automated upload tool. The server stores the data and generates suitable preview images for the electronic notebook.
Figure 7. Screen image of the electronic notebook interface, showing a list of available tools, a project header and sample AFM thumbnail images that have been uploaded into the database.
file system or in PostgreSQL, either as numeric tables or as XML fragments. The choice of open-source and freely available software packages for this project is based on the understanding that it is impossible to design and build an integrated application to handle all materials science data processing needs. The system needs to operate in a variety of different hardware and operating system environments and, most importantly, has to be easily extendable by users who wish to integrate new functionalities specific for their experiments. Open-source software offers this flexibility at a cost that is affordable even for small research groups. For the same reasons data are always stored in standardized formats, such as XML, even when the data processing is done with commercial tools that use proprietary internal data formats. The specific choice of software packages is not critical for functionality. Most packages, such as the web server, the database, or the image processing libraries, can be replaced with relatively little effort. Only PHP, the scripting language used for building the Web interface, cannot be replaced without a major software rewrite effort. If necessary, however, the existing functionality can be extended using any other language that can be used to build either a shared library or a standalone application. Figure 7 shows an example image of an electronic notebook-style Web interface. New projects are created in a top level hierarchy for a particular user. Each project can have a tree of sub-projects. The leaf nodes of the tree structure correspond to actual combinatorial libraries. Each node in the project tree can accept experimental data files. A preview image is created automatically when a data file is 314
uploaded. Text, image files (such as BMP, JPEG, TIFF, etc), PDF document files and some vendor-specific formats, such as the Seiko Instruments AFM data files, are recognized by the preview generating functions. The graphic formats are handled by the ImageMagick library [19], preview thumbnails for PDF files are generated with Ghostscript [20] and vendorspecific formats are handled by custom PHP functions. Some AFM images are shown in the notebook example in figure 7. The AFM preview images are generated using the GD library. The scalable vector graphics [21] format is used for presenting plot data in the web interface. As a test case, we have integrated the Seiko AFM machines into the database system, including a file upload function and image preview generation on the server and an instrument-specific data upload tool that runs concurrently with the commercial AFM software on a client computer. Figure 8 shows a diagram of the data transfer software. The AFM machine (SII NanoTechnology Inc., Nanopics 1000) is connected to a personal computer. Measurements are done on the probing station and data are transferred to the PC using a vendor-supplied software package. Image processing, such as surface tilting or colourmap selection, is also handled by the program. The AFM data can be uploaded to the database system either by manually logging into the server and performing the upload or by using an automated tool. The manual method is more time consuming, but works even when it is not possible to install new software on the client computer and only a web browser is available. The automated tool removes the need for manually uploading individual files. The upload program has been implemented in LabVIEW from National Instruments, which provides a high-level graphical programing system that integrates well with a variety of measurement and control interfaces. Although LabVIEW has built-in tools for network communications, the software necessary for forming SSL connections is not a part of the base package. We therefore use an external libcurl communication library [22] for opening HTTPS connections to the database server. Libcurl supports many network protocols, such as HTTP, HTTPS and FTP. Compared to manual login procedures, a higher level of convenience can be achieved by using client certificates for establishing a connection to the Web interface. All lowlevel network-related functions have been placed in an external
Elements of informatics for combinatorial solid-state materials science
dynamic link library (DLL), supporting HTTP GET and POST methods over an SSL link. This DLL file is called from the LabVIEW program. User account and sample identification information are passed to the DLL file as arguments. The DLL file is not specific to the AFM upload function and can be used by other instrument control programs as well. Preparing and selecting the AFM data for upload is performed by the AFM-specific LabVIEW program, which monitors file creation in the PC file system and either notifies the user when new files are available, or transfers them directly to the database without further user intervention. On the server side, a static web page is used as a user interface for manual uploads. The user can specify which sample is being measured and enter the AFM file name in an entry box on the screen. A PHP script on the server receives the library information and the actual data file. This functionality is common to all manual upload functions, regardless of data type. The uploaded file is analysed and image metadata are extracted for storing in the database. The actual image data are converted into graphic format and stored in the file system. Two images are stored, one is full-sized, typically a 512 pixel by 512 pixel image, the other is a reduced size thumbnail for use in the electronic notebook interface. The newlyuploaded data become immediately available on the notebook interface. The data analysis and preview generation script is common for both manual and automated uploads. Each new measurement function that is added to the informatics system requires a specific script for analysing the new data, storing the metadata, generating preview images and building an electronic notebook fragment for each uploaded file.
5. Conclusions We have described the design considerations, requirements and an example implementation of a simple informatics system for managing combinatorial solid-state experiments. The system described here has been specifically designed to handle thin film deposition and characterization experiments. Most of the data handling functions, however, are quite universal and can be used for a much wider variety of materials science experiments. When compared to related fields, such as bioinformatics or medical informatics, the system described here is expected to handle only modest volumes of data. Instead of having to process huge volumes of structurally simple data, a materials science informatics system needs to accept and integrate into a single process a very wide range of diverse types of information. The most important features are the open nature of the software, as all the server side components and communication library are freely available in source form. The LabVIEW application can be distributed and executed with the LabVIEW runtime engine, which is freely available from National Instruments, although the LabVIEW development environment is commercial. All information is stored in a centralized relational database and can be accessed over the Web, both for data entry and retrieval. Data access is based on the concept of an electronic notebook and the details of communications, authentication, data protection and database access or structure are conveniently hidden from the user.
Information is presented to the user in the form of a hierarchical tree of experiments and reference data. The Web interface also provides a variety of management functions for limiting or granting access to specific data items to different users or user groups. The programing is based on widelyused software tools and scripting languages, enabling a diverse group of users to add new functionality to the system in the form of modular plug-ins.
Acknowledgments This work was carried out under the Visiting Researcher’s Program of the Institute for Solid State Physics, the University of Tokyo. Financial support from the Asahi Glass foundation is acknowledged.
References [1] Lippmaa M, Kawasaki M and Koinuma H 2003 Combinatorial Materials Synthesis ed X-D Xiang and I Takeuchi (New York: Dekker) chapter 5 [2] Matsumoto Y, Murakami M, Shono T, Hasegawa T, Fukumura T, Kawasaki M, Ahmet P, Chikyow T, Koshihara S and Koinuma H 2001 Room-temperature ferromagnetism in transparent transition metal-doped titanium dioxide Science 291 854 [3] Fukumura T, Ohtani M, Kawasaki M, Okimoto Y, Kageyama T, Koida T, Hasegawa T, Tokura Y and Koinuma H 2000 Rapid construction of a phase diagram of doped Mott insulators with a composition-spread approach Appl. Phys. Lett. 77 3426 [4] Koinuma H 2002 Combinatorial materials research projects in Japan Appl. Surf. Sci. 189 179 [5] Zhang W, Fasolka M J, Karim A and Amis E J 2004 An open source informatics system for combinatorial materials research ACS Pol. Mat.: Sci. Eng. 90 341 [6] Todoroki S and Inoue S 2004 Multi-dimensional data management by virtual sample library written in object-oriented script language Ruby Trans. Mat. Res. Soc. Japan 29 293 [7] Yanase I, Ohtaki T and Watanabe M 2002 Combinatorial study on nano-particle mixture prepared by robot system Appl. Surf. Sci. 189 292 [8] Yamamoto Y, Takahashi R, Matsumoto Y, Chikyow T and Koinuma H 2004 Mathematical design of linear action masks for binary and ternary composition spread film library Appl. Surf. Sci. 223 9 [9] Yoo Y-K, Duewer F, Yang H, Yi D, Li J-W and Xiang X-D 2000 Room-temperature electronic phase transitions in the continuous phase diagrams of perovskite manganites Nature 406 704 [10] Koida T, Komiyama D, Koinuma H, Ohtani M, Lippmaa M and Kawasaki M 2002 Temperature-gradient epitaxy under in situ growth mode diagnostics by scanning reflection high-energy electron diffraction Appl. Phys. Lett. 80 565 [11] Kim S-W, Sehgal A, Karim A and Fasolka M J 2003 Techniques for combinatorial and high-throughput microscopy, part 2: automated optical microscopy platform for thin film research Microsc. Today November/ December 30 [12] Hasegawa K, Ahmet P, Okazaki N, Hasegawa T, Fujimoto K, Watanabe M, Chikyow T and Koinuma H 2004 Amorphous stability of HfO2 based ternary and binary composition spread oxide films as alternative gate dielectrics Appl. Surf. Sci. 223 229 [13] Ohtani M et al 2001 Concurrent x-ray diffractometer for high throughput structural diagnosis of epitaxial thin films Appl. Phys. Lett. 79 3594
315
S Meguro et al
[14] Chikyow T, Ahmet P, Nakajima K, Koida T, Takakura M, Yoshimoto M and Koinuma H 2002 A combinatorial approach in oxide/semiconductor interface research for future electronic devices Appl. Surf. Sci. 189 284 [15] Extensible Markup Language (http://www.w3.org/XML/) [16] Apache HTTPD project (http://httpd.apache.org/)
316
[17] [18] [19] [20] [21] [22]
PostgreSQL (http://www.postgresql.org/) PHP scripting language (http://www.php.net/) Imagemagick (http://www.imagemagick.org/) GNU ghostscript (http://www.ghostscript.com/doc/gnu/) Scalable Vector Graphics (http://www.w3.org/Graphics/SVG/) cURL communication library (http://curl.haxx.se/)
