- Email: [email protected]
The Thermodynamic Database Database
Calphad 61 (2018) 173–178
Contents lists available at ScienceDirect
Calphad journal homepage: www.elsevier.com/locate/calphad
The Thermodynamic Database Database a,⁎
a
Axel van de Walle , Chiraag Nataraj , Zi-Kui Liu a b
T
b
School of Engineering, Brown University, Providence, RI 02912, USA Department of Materials Science and Engineering, Pennsylvania State University, University Park, PA 16802, USA
A R T I C LE I N FO
A B S T R A C T
Keywords: Database Materials informatics Search engine Thermodynamic data
This paper describes an effort to organize the condensed phases thermodynamic data freely available in electronic form within the scientific literature through the design of a special-purpose search engine indexing data electronically available in the “thermodynamic database” (TDB) format. This focus is motivated by the fact that it is widely used and can be readily read or imported into most thermodynamic modeling software. This form of data also provides a rather complete thermodynamic description of a given system and thus enables researchers to generate any phase diagram cross-section of interest, a capability typically not available in traditional phase diagram handbooks. For convenience, users can quickly preview selected cross sections directly online. In designing this system, special emphasis was devoted to ensuring that the bibliographic references of the original data sources are transparently reported and to providing links to the original data sources, rather than the data itself, in order to enforce access rights. This effort was made possible by combining and building upon a number of key components, such as the CALPHAD journal's supplementary information section, the NIMS database, the NIST materials data repository, the Crossref bibliographic service, and various thermodynamic (OpenCalphad, ATAT) and graphical (gnuplot, XTK) software.
1. Introduction
database [12], the NIST materials data repository [13,14], the Crossref bibliographic service [15], and various thermodynamic (OpenCalphad [8,16], ATAT [17,18]) and graphical (gnuplot [19], XTK [20,21]) software.
While there exist various comprehensive commercial sources of thermodynamic data (e.g. [1–3]), this paper describes an effort to organize the condensed phases thermodynamic data freely available in electronic form within the scientific literature. This is accomplished via a special-purpose search engine, the Thermodynamic DataBase DataBase (TDBDB) [4], that indexes data that is electronically available in the “thermodynamic database” (TDB) format. This capability aims (i) to facilitate the development of future CALPHAD models, (ii) to enable CALPHAD-based high-throughput computational materials discovery and optimization and (iii) to provide information that is complementary to widely used phase diagram handbooks [3]. Indeed, merely reporting phase diagrams becomes less effective for higherorder systems — researchers rather need the ability to generate specific sections on demand, which demands a comprehensive thermodynamic model. Such models can be readily read or imported into standard thermodynamic software (e.g., [2,5,6,1,7–11]). In designing this system, emphasis was put on simplicity, ensuring that access rights are followed and transparent attribution of proper credit to the authors of the original data. This effort was made possible by combining and building upon a number of key components, such as the CALPHAD journal's supplementary information section, the NIMS
⁎
2. Methodology The Thermodynamic DataBase DataBase (TDBDB) effort combines a number of existing resources in the form of data and software libraries. This section describes each aspect, summarized in Fig. 1, in more detail. The raw data was obtained by first scanning the entire publication history of the CALPHAD journal in search of electronic supplementary information in the form of TDB files. This process also incidentally gathered databases containing some kinetic information. This gave a database of thermodynamic data somewhat biased towards the most recently assessed systems (because the systematic inclusion of electronic supplementary information is a relatively new phenomenon). To complete this database with systems that have been well-assessed for a long time, we queried the NIMS database [12], which covers a large number of binary systems as well as a few higher-order multicomponent systems. We also searched the NIST materials data repository [13,14] for CALPHAD assessments. Finally, we have manually added a few well-known openly available databases, such as the
Corresponding author. E-mail address: [email protected] (A. van de Walle).
https://doi.org/10.1016/j.calphad.2018.04.003 Received 17 January 2018; Received in revised form 30 March 2018; Accepted 3 April 2018 0364-5916/ © 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).
Calphad 61 (2018) 173–178
A. van de Walle et al.
Data sources
Software
Output
Off-line uses
Calphad J.
TDBDB
Biblio. data
Latex, Word, ...
NIMS
ATAT
TDB files
Open Calphad ThermoCalc FactSage Pandat ...
NIST
... Crossref
XTK gnuplot
Phase diagrams
Open Calphad
Graphic files (VTK, gnuplot)
preference for those near axes or close to a previously known equilibrium). Third, the probability of properly “capturing” a phase is proportional to the size of the set of points in temperature-composition space that involve an equilibrium with that phase, i.e. if the algorithm “misses” a phase, it is likely to have been small anyway. Finally, this approach can be easily parallelized and the current implementation exploits that fact to deliver a faster feedback. This method does have some drawbacks. It does not currently apply to the calculation of sections that involve composition constraints (but such capabilities will be included in the future). This method is also less efficient than boundaryfollowing techniques because each equilibrium is re-computed “from scratch” without using information from nearby equilibria, although this too could be improved in the future, since OpenCalphad has the ability to re-use previously calculated equilibria to speed up the calculations. Nevertheless, the robustness and scalability of the current method seem ideal to deliver phase diagram previews with minimal user input. This data is then plotted and posted on the website in one of a number of ways. For binary systems, the phase diagram data is plotted using gnuplot [19]. It is rendered online using gnuplot's “canvas” terminal (see Fig. 2). One can also download the gnuplot script to generate a plot in any graphic format off-line. Ternary isothermal sections can similarly be plotted within a standard Gibbs triangle by specifying identical minimum and maximum temperatures (see Fig. 3). For ternary systems, a 3D plot (including temperature as the third axis) can be generated. The website displays a static 3D view but, by downloading the gnuplot script and running it on a local copy of gnuplot, one can interactively rotate the phase diagram off-line or generate a plot in any graphic format. For quaternary systems, only isothermal sections are implemented and are shown within a 3D Gibbs tetrahedron, again using gnuplot. While the gnuplot-generated 3D plots show the phase boundaries as discrete points (with tie lines connecting them), another option is available that provides higher-quality and interactive 3D graphics. This feature makes use of the XTK package [20,21], which enables 3D graphics via WebGL that can be interactively rotated. Phase boundaries are shown as partially transparent triangulated surfaces with tie-lines connecting them (see Fig. 4). Both Gibbs triangle and Gibbs tetrahedron (see Fig. 5) representations are available. The underlying graphic files can also be download in standard VTK
Paraview gnuplot
Fig. 1. Overall organization of the Thermodynamic DataBase DataBase effort.
NIST solder database [22], some open databases supplied with MatCalc [11] and some contributed TDB files (e.g. [23]). The end result is (currently) a total of 766 data entries arising from 528 publications. This database will be updated periodically by scanning new issues of the CALPHAD journal and re-scanning the NIMS and NIST databases, among other web crawling efforts. We then matched each paper to its Digital Object Identifier (DOI), which provides a permanent pointer to the data source. This also enables a simple mechanism to generate appropriate bibliographic citations in any format via the Crossref service [15]. While the DOI assignment for CALPHAD Journal articles and NIST database entries was straightforward, for articles cited in the NIMS database, DOI matching was achieved through queries to the Crossref service. If no correct match is found (which may happen, for instance, for older papers), a standard URL (Uniform Resource Locator) is used instead. In designing this system, we wanted to ensure that copyrights or the terms of service of the data sources were respected. For these reasons, the TDBDB website does not actually provide the content of the TDB files or the papers. It acts as a search engine and yield links to the requested resource. It is up to the users to ensure they have the credential to access the data. For data hosted in the NIMS database this simply involves creating an account on the NIMS site (which is free). For data or articles hosted in scientific journals, one needs an electronic subscription to the journal (which most researchers have via their respective institutions). As a result, the TDBDB service will most likely increase the visibility and access statistics of the data providers rather than reduce them. The user interface is deliberately minimalist, as this provides crisper user feedback, eases code development and maintenance and makes it easier to access the website on devices with smaller screens (e.g. smart phones). Users simply enter the elements of interest and the system displays a list of databases containing data on all of the requested elements. Along with each database entry are reported links to the paper, its bibliographic reference and to the TDB file. The TDBDB web interface also provides a quick preview of the system's phase diagram (based on the database selected). This feature is made possible by the OpenCalphad [8,16] software. As the plotting capabilities of OpenCalphad currently require some user guidance, we opted for a simple robust brute-force method to generate the phase diagram preview. A large number of randomly chosen temperaturecomposition coordinates are generated and the corresponding phase equilibria are calculated with OpenCalphad. The points yielding singlephase equilibria are discarded and the remaining multiphase equilibria are sorted by phases and the temperature composition data points belonging to the same multiphase equilibrium are re-connected to form the phase boundaries. For 2D sections, this re-connection algorithm reduces to a simple nearest neighbor search. In higher dimension, this amounts to a type of Delaunay triangulation [26]. The advantages of this scheme are fourfold. First, it is easily scalable from a fast lowprecision representation to a slower high-precision one by merely adjusting the number of randomly chosen points. Second, it can easily capture phase fields located anywhere in the chosen section (without
Fig. 2. Example of binary phase diagram using the data from Reference [24]. 174
Calphad 61 (2018) 173–178
A. van de Walle et al.
Fig. 3. Example of isothermal section of a ternary phase diagram using the data from Reference [25]. Fig. 5. Example of an isothermal section of quaternary phase diagram using the data from Reference [28].
suitable for gnuplot or 3D viewers supporting the widely used VTK format. The ocutil command fixes common issues that prevent some TDB files from being readable by OpenCalphad. It cleanly truncates text lines to the right length, eliminates nonprintable stray characters arising from file format conversions, re-orders some keywords in the order OpenCaphad expects them, etc. The web front end of TDBDB interacts with the back end server via a standard RESTful (Representational state transfer) API (Application Programing Interface) and via data structures in the standard JSON (JavaScript Object Notation) format. This feature is of independent interest and the protocol and formats currently used are described in the Appendix. 3. Discussion The TDBDB service complements existing materials thermodynamic resources. While the aim of the TDBDB effort has some similarity with NIST's materials data repository [13,14] or the NIMS database [12], it exhibits some key differences. Most importantly, the system actively fetches the data rather than relying on researchers to upload or type in data. Of course, both the NIST and the NIMS databases contribute significantly to TDBDB's usefulness, as they provide extensive sources of data that would otherwise be unavailable. In particular, the NIMS database represents a tremendous resource by providing data in electronic format based on the thermodynamic information that would otherwise only be available in tabular form within the original papers. TDBDB does offer some additional features such as allowing users to interactively view isothermal sections of interest in multicomponent systems and providing easy-to-use links to papers and citation information. It should be noted that, compared to the TDBDB, NIST's materials data repository [13] is far broader in scope, contains richer metadata and allows users to upload their data. Also, at this point, the phase diagram plots available in the NIMS database [12] are generally of higher quality than the previews generated within the TDBDB. One of TDBDB's
Fig. 4. Example of 3D representation of a ternary phase diagram using the data from Reference [27].
(Visualization ToolKit) format [29] for off-line viewing with more powerful software, such as Paraview [30,31], which enables the calculation of general cross-sections interactively (see Fig. 6). A convenient by-product of these efforts is a set of tools, available within the ATAT package [17,18], that facilitates handling and plotting of phase diagrams with OpenCalphad. The ocplotpd command implements all of the website's phase diagram plotting capabilities. It calls OpenCalphad to calculate equilibria and generates graphical output 175
Calphad 61 (2018) 173–178
A. van de Walle et al.
Fig. 6. Example of interactive viewing of general sections of a ternary phase diagram obtained with Paraview based on a 3D phase diagrams exported from TDBDB using the data from Reference [32].
triangulation of the phase boundaries in general n-dimensional space has already been implemented but cross-sectioning operations enabling plotting as 2D or 3D graphs has not.) The TDBDB could also be expanded to include links to commercial databases, so that researchers can at least be made aware of the existence of ready-made databases for their systems of interest (even if the data itself is not openly available). Another way to extend TDBDB's coverage would be to mine older papers to extract thermodynamic solution parameters provided in tabular form to build ready-to-use TDB files (although this has already been accomplished in large part thanks to the NIMS database [12]). With the large amount of thermodynamic data now readily available through a single interface, one could imagine allowing users to more easily perform CALPHAD “extrapolation” by combining known binaries or ternaries to generate reasonable predictions for higher order systems. This task would entail a standardization of phase names as well as other aspects of the TDB files (e.g. mutual consistency of the free energy models), which would require a significant effort. It would also be possible to fill gaps in the current set of available assessments with automated high-throughput CALPHAD methods [43,44], made possible in part by the recent development of a formal approach to ensuring that ab initio and experimental energies are mutually consistent [45]. Finally, we hope that this effort will provide further incentives to include thermodynamic information in electronic format as supplementary
novel contributions is to make the thermodynamic data available as supplementary information to CALPHAD Journal articles much more readily accessible. There also exists a number of sources of thermodynamic information primarily in the form of ab initio-calculated absolute-zero formation energies [33–42]. In contrast, TDBDB's focus is on finite-temperature thermodynamics properties arising primarily from experimental data, with increasingly more data arising from electronic structure calculations in recent years. 4. Conclusion The current extent of the TDBDB project represents the beginning of a larger scale effort that could include some of the following aspects. The system currently only indexes files in the “TDB” format, but one could also include other formats as well [1,10]. The phase diagram preview capabilities are currently limited and could benefit from a better algorithm to connect the calculated data points to form smooth phase boundaries. The website's preview capabilities could also offer the option to invoke the pycalphad toolkit [9] for plotting purposes. As an increasing number of thermodynamic models include 4 or more components, the ability to represent higher-dimensional phase diagrams via interactive cross-sections would be a plus. (Currently, 176
Calphad 61 (2018) 173–178
A. van de Walle et al.
information in CALPHAD journal publications and encourage other journals to follow suit in requesting the submission of supplementary thermodynamic information in electronic form.
Acknowledgments This research was supported by ONR under grant N00014-17-12202, and by Brown University through the use of the facilities at its Center for Computation and Visualization. This work uses the Extreme Science and Engineering Discovery Environment (XSEDE) resource Stampede 2 at the Texas Advanced Computing Center through allocation TG-DMR050013N, which is supported by National Science Foundation grant number ACI-1548562.
TDBDB then returns a JSON-formatted object with the following keys:
Appendix A. Application Programing Interface (API) In describing the API, the following conventions are used. The string http://tdbdb/ stands for the website's URL. Sections marked as [some expression] need to be replaced by the actual parameters needed. Any text to be included verbatim is indicated by a courier font. Interaction with the server typically starts by a GET request to list the available database containing data regarding the desired elements of the form: http://tdbdb/getdbid.php?element=[comma-separated list of element symbols] The returned information is a JSON-formatted object with the following keys:
References [1] C. Bale, E. Blisle, P. Chartrand, S. Decterov, G. Eriksson, A. Gheribi, K. Hack, I.H. Jung, Y.-B. Kang, J. Melanon, A. Pelton, S. Petersen, C. Robelin, J. Sangster, P. Spencer, M.-A.V. Ende, FactSage thermochemical software and databases, 2010–2016, Calphad 54 (2016) 35. [2] J.O. Andersson, T. Helander, L. Höglund, P.F. Shi, B. Sundman, Thermo-Calc & DICTRA, computational tools for materials science, Calphad 26 (2002) 273. [3] P. Villars, H. Okamoto, K. Cenzual (Eds.), Alloy Phase Diagrams Center, ASM International, 2007. URL 〈http://www.asminternational.org〉. [4] A. van de Walle, The thermodynamic database database (TDBDB) (2017). URL 〈https://alum.mit.edu/www/avdw/tdbdb.html〉. [5] G. Eriksson, K. Hack, ChemSage– A computer program for the calculation of complex chemical equilibria, Metall. Trans. B 21B (1990) 1013. [6] C. Bale, P. Chartrand, S. Degterov, G. Eriksson, K. Hack, R.B. Mahfoud, J. Melançon, A. Pelton, S. Petersen, FactSage thermochemical software and databases, Calphad 26 (2002) 189. [7] W. Cao, S.-L. Chen, F. Zhang, K. Wu, Y. Yang, Y. Chang, R. Schmid-Fetzer, W. Oates, PANDAT software with PanEngine, PanOptimizer and PanPrecipitation for multi-
The user would then typically initiate a download of the desired TDB file whose URL is given in the tdburl key. This is done without relying on the TDBDB service, typically with a simple wget [URL] command. The user can also request a preview of the phase diagram with the following GET request (line breaks inserted for clarity): 177
Calphad 61 (2018) 173–178
A. van de Walle et al.
[8] [9] [10]
[11]
[12] [13]
[14] [15] [16]
[17]
[18]
[19] [20]
[21] [22] [23]
[24] [25] [26]
[27] [28]
sb melts, Calphad 52 (2016) 159. [29] L.S. Avila, U. Ayachit, S. Barré, et al., The VTK Users Guide, Kitware, USA, 2010. URL 〈https://www.vtk.org/vtk-users-guide/〉. [30] J. Ahrens, B. Geveci, C. Law, Paraview: An end-user tool for large data visualization, in: C.D. Hansen, C.R. Johnson (Eds.), Visualization Handbook, Elsevier, New York, 2005. [31] U. Ayachit, The ParaView Guide: A Parallel Visualization Application, Kitware, USA, 2015. [32] S. Sridar, R. Kumar, K.C.H. Kumar, Thermodynamic modelling of ti-zr-n system, Calphad 56 (2017) 102. [33] Y. Wang, S. Curtarolo, C. Jiang, R. Arroyave, T. Wang, G. Ceder, L.-Q. Chen, Z.K. Liu, Ab initio lattice stability in comparison with CALPHAD lattice stability, Calphad 28 (2004) 79. [34] A. Jain, S.P. Ong, G. Hautier, W. Chen, W.D. Richards, S. Dacek, S. Cholia, D. Gunter, D. Skinner, G. Ceder, K.A. Persson, The materials project: a materials genome approach to accelerating materials innovation, APL Mater. 1 (2013) 011002. [35] K. Mathew, J.H. Montoya, A. Faghaninia, S. Dwarakanath, M. Aykol, H. Tang, I.H. Chu, T. Smidt, B. Bocklund, M. Horton, J. Dagdelen, B. Wood, Z.-K. Liu, J. Neaton, S.P. Ong, K. Persson, A. Jain, Atomate: a high-level interface to generate, execute, and analyze computational materials science workflows, Comput. Mater. Sci. 139 (2017) 140. [36] A. Jain, G. Hautier, C.J. Moore, S.P. Ong, C.C. Fischer, T. Mueller, K.A. Persson, G. Ceder, A high-throughput infrastructure for density functional theory calculations, Comp. Mater. Sci. 50 (2011) 2295. [37] J.E. Saal, S. Kirklin, M. Aykol, B. Meredig, C. Wolverton, Materials design and discovery with high-throughput density functional theory: the open quantum materials database (oqmd), JOM 65 (2013) 1501. [38] S. Kirklin, J. Saal, B. Meredig, A. Thompson, J. Doak, M. Aykol, S. Rühl, C. Wolverton, The open quantum materials database (oqmd): assessing the accuracy of dft formation energies, NPJ Comput. Mater. 1 (2015) 15010. [39] M. Mihalkovic, M. Widom, First-principles prediction of a decagonal quasicrystal containing boron, Phys. Rev. Lett. 93 (2004) 095507. [40] M. Mihalkovic, M. Widom, Widom Alloy Database, 2004. URL 〈http://alloy.phys. cmu.edu/〉. [41] S. Curtarolo, W. Setyawan, G.L. Hart, M. Jahnatek, R.V. Chepulskii, R.H. Taylor, S. Wang, J. Xue, K. Yang, O. Levy, M.J. Mehl, H.T. Stokes, D.O. Demchenko, D. Morgan, Aflow: an automatic framework for high-throughput materials discovery, Comput. Mater. Sci. 58 (2012) 218. [42] C.E. Calderon, J.J. Plata, C. Toher, C. Oses, O. Levy, M. Fornari, A. Natan, M.J. Mehl, G. Hart, M.B. Nardelli, S. Curtarolo, The AFLOW standard for highthroughput materials science calculations, Comput. Mater. Sci. 108 (2015) 233. [43] A. van de Walle, R. Sun, Q.-J. Hong, S. Kadkhodaei, Software tools for highthroughput calphad from first-principles data, Calphad 58 (2017) 70, http://dx.doi. org/10.1016/j.calphad.2017.05.005. [44] R. Otis, Z.-K. Liu, High-throughput thermodynamic modeling and uncertainty quantification for ICME, JOM 69 (2017) 886, http://dx.doi.org/10.1007/s11837017-2318-6. [45] A. van de Walle, reconciling SGTE and ab initio enthalpies of the elements, Calphad 60 (2018) 1, http://dx.doi.org/10.1016/j.calphad.2017.10.008.
component phase diagram calculation and materials property simulation, Calphad 33 (2009) 328. B. Sundman, U.R. Kattner, M. Palumbo, S.G. Fries, OpenCalphad – a free thermodynamic software, Integr. Mater. Manuf. Innov. 4 (2015) 1. R.A. Otis, Z.-K. Liu, pycalphad: calphad-based computational thermodynamics in python, J. Open Res. Softw. 5 (2017) 1, http://dx.doi.org/10.5334/jors.140. R. Davies, A. Dinsdale, J. Gisby, S. Hodson, R. Ball, Thermodynamic modelling using MTDATA: A description showing applications involving oxides, alloys and aqueous solutions, in: P. Nash, B. Sundman (Eds.), Applications of Thermodynamics in the Synthesis and Processing of Materials, TMS, 1995, p. 371. E. Kozeschnik, B. Buchmayr, MatCalc - A simulation tool for multicomponent thermodynamics, diffusion and phase transformation kinetics, Math. Model. Weld. Phenom. 5 (2001) 349. T. Abe, K. Hashimoto, Y. Sawada, M. Palumbo, Y. Ogura, Nims Database, 2007. URL 〈http://cpddb.nims.go.jp/cpddb/periodic.htm〉. C. Campbell, U. Kattner, Z.-K. Liu, File and data repositories for next generation CALPHAD, Scr. Mater. 70 (2014) 7, http://dx.doi.org/10.1016/j.scriptamat.2013. 06.013. NIST Materials Data Repository. URL 〈https://materialsdata.nist.gov/〉. Crossref. URL 〈https://www.crossref.org/〉. B. Sundman, U.R. Kattner, C. Sigli, M. Stratmann, R.L. Tellier, M. Palumbo, S.G. Fries, The Open Calphad thermodynamic software interface, Comput. Mater. Sci. 125 (2016) 188. A. van de Walle, M.D. Asta, G. Ceder, The alloy theoretic automated toolkit: a user guide, Calphad 26 (2002) 539–553, http://dx.doi.org/10.1016/S0364-5916(02) 80006-2. A. van de Walle, Multicomponent multisublattice alloys, nonconfigurational entropy and other additions to the Alloy Theoretic Automated Toolkit, Calphad 33 (2009) 266–278, http://dx.doi.org/10.1016/j.calphad.2008.12.005. T. Williams, C. Kelley, et al., Gnuplot. URL 〈http://www.gnuplot.info/〉. D. Hähn, N. Rannou, B. Ahtam, P.E. Grant, R. Pienaar, Neuroimaging in the browser using the X toolkit, in: Front. Neuroinform.: 5th INCF Congress of Neuroinformatics., 2014. URL 〈http://f1000.com/posters/browse/summary/ 1092491〉. D. Hähn, N. Rannou, B. Ahtam, P.E. Grant, R. Pienaar, The X Toolkit, 2014. URL 〈https://goXTK.com〉. U.R. Kattner, The NIST Solder Database, 2017. URL 〈http://www.metallurgy.nist. gov/phase/solder/solder.html〉. G. Ghosh, A. van de Walle, M.D. Asta, G. Olson, Phase stability of the Hf-Nb system: from first-principles to CALPHAD, Calphad 26 (2002) 491, http://dx.doi.org/10. 1016/S0364-5916(02)80003-7. F. Zhang, S. Liu, J. Wang, Y. Du, L. Sun, Experimental investigation and thermodynamic assessment of the li-sb system, Calphad 57 (2017) 28. C. Guo, T. Wu, C. Li, Z. Du, Thermodynamic description of the al-fe-nb system, Calphad 57 (2017) 78. H. Chen, J. Bishop, Delaunay triangulation for curved surfaces, in: Proceedings of the 6th International Meshing Roundtable, Sandia National Laboratories, 1997, p. 115. P. Lafaye, C. Toffolon-Masclet, J.-C. Crivello, J.-M. Joubert, Thermodynamic modelling of the cr-nb-sn system, Calphad 57 (2017) 37. W. Chen, L. Zhang, Y. Du, Diffusivities and atomic mobilities in sn-ag-in and sn-ag-
178
Contents lists available at ScienceDirect
Calphad journal homepage: www.elsevier.com/locate/calphad
The Thermodynamic Database Database a,⁎
a
Axel van de Walle , Chiraag Nataraj , Zi-Kui Liu a b
T
b
School of Engineering, Brown University, Providence, RI 02912, USA Department of Materials Science and Engineering, Pennsylvania State University, University Park, PA 16802, USA
A R T I C LE I N FO
A B S T R A C T
Keywords: Database Materials informatics Search engine Thermodynamic data
This paper describes an effort to organize the condensed phases thermodynamic data freely available in electronic form within the scientific literature through the design of a special-purpose search engine indexing data electronically available in the “thermodynamic database” (TDB) format. This focus is motivated by the fact that it is widely used and can be readily read or imported into most thermodynamic modeling software. This form of data also provides a rather complete thermodynamic description of a given system and thus enables researchers to generate any phase diagram cross-section of interest, a capability typically not available in traditional phase diagram handbooks. For convenience, users can quickly preview selected cross sections directly online. In designing this system, special emphasis was devoted to ensuring that the bibliographic references of the original data sources are transparently reported and to providing links to the original data sources, rather than the data itself, in order to enforce access rights. This effort was made possible by combining and building upon a number of key components, such as the CALPHAD journal's supplementary information section, the NIMS database, the NIST materials data repository, the Crossref bibliographic service, and various thermodynamic (OpenCalphad, ATAT) and graphical (gnuplot, XTK) software.
1. Introduction
database [12], the NIST materials data repository [13,14], the Crossref bibliographic service [15], and various thermodynamic (OpenCalphad [8,16], ATAT [17,18]) and graphical (gnuplot [19], XTK [20,21]) software.
While there exist various comprehensive commercial sources of thermodynamic data (e.g. [1–3]), this paper describes an effort to organize the condensed phases thermodynamic data freely available in electronic form within the scientific literature. This is accomplished via a special-purpose search engine, the Thermodynamic DataBase DataBase (TDBDB) [4], that indexes data that is electronically available in the “thermodynamic database” (TDB) format. This capability aims (i) to facilitate the development of future CALPHAD models, (ii) to enable CALPHAD-based high-throughput computational materials discovery and optimization and (iii) to provide information that is complementary to widely used phase diagram handbooks [3]. Indeed, merely reporting phase diagrams becomes less effective for higherorder systems — researchers rather need the ability to generate specific sections on demand, which demands a comprehensive thermodynamic model. Such models can be readily read or imported into standard thermodynamic software (e.g., [2,5,6,1,7–11]). In designing this system, emphasis was put on simplicity, ensuring that access rights are followed and transparent attribution of proper credit to the authors of the original data. This effort was made possible by combining and building upon a number of key components, such as the CALPHAD journal's supplementary information section, the NIMS
⁎
2. Methodology The Thermodynamic DataBase DataBase (TDBDB) effort combines a number of existing resources in the form of data and software libraries. This section describes each aspect, summarized in Fig. 1, in more detail. The raw data was obtained by first scanning the entire publication history of the CALPHAD journal in search of electronic supplementary information in the form of TDB files. This process also incidentally gathered databases containing some kinetic information. This gave a database of thermodynamic data somewhat biased towards the most recently assessed systems (because the systematic inclusion of electronic supplementary information is a relatively new phenomenon). To complete this database with systems that have been well-assessed for a long time, we queried the NIMS database [12], which covers a large number of binary systems as well as a few higher-order multicomponent systems. We also searched the NIST materials data repository [13,14] for CALPHAD assessments. Finally, we have manually added a few well-known openly available databases, such as the
Corresponding author. E-mail address: [email protected] (A. van de Walle).
https://doi.org/10.1016/j.calphad.2018.04.003 Received 17 January 2018; Received in revised form 30 March 2018; Accepted 3 April 2018 0364-5916/ © 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).
Calphad 61 (2018) 173–178
A. van de Walle et al.
Data sources
Software
Output
Off-line uses
Calphad J.
TDBDB
Biblio. data
Latex, Word, ...
NIMS
ATAT
TDB files
Open Calphad ThermoCalc FactSage Pandat ...
NIST
... Crossref
XTK gnuplot
Phase diagrams
Open Calphad
Graphic files (VTK, gnuplot)
preference for those near axes or close to a previously known equilibrium). Third, the probability of properly “capturing” a phase is proportional to the size of the set of points in temperature-composition space that involve an equilibrium with that phase, i.e. if the algorithm “misses” a phase, it is likely to have been small anyway. Finally, this approach can be easily parallelized and the current implementation exploits that fact to deliver a faster feedback. This method does have some drawbacks. It does not currently apply to the calculation of sections that involve composition constraints (but such capabilities will be included in the future). This method is also less efficient than boundaryfollowing techniques because each equilibrium is re-computed “from scratch” without using information from nearby equilibria, although this too could be improved in the future, since OpenCalphad has the ability to re-use previously calculated equilibria to speed up the calculations. Nevertheless, the robustness and scalability of the current method seem ideal to deliver phase diagram previews with minimal user input. This data is then plotted and posted on the website in one of a number of ways. For binary systems, the phase diagram data is plotted using gnuplot [19]. It is rendered online using gnuplot's “canvas” terminal (see Fig. 2). One can also download the gnuplot script to generate a plot in any graphic format off-line. Ternary isothermal sections can similarly be plotted within a standard Gibbs triangle by specifying identical minimum and maximum temperatures (see Fig. 3). For ternary systems, a 3D plot (including temperature as the third axis) can be generated. The website displays a static 3D view but, by downloading the gnuplot script and running it on a local copy of gnuplot, one can interactively rotate the phase diagram off-line or generate a plot in any graphic format. For quaternary systems, only isothermal sections are implemented and are shown within a 3D Gibbs tetrahedron, again using gnuplot. While the gnuplot-generated 3D plots show the phase boundaries as discrete points (with tie lines connecting them), another option is available that provides higher-quality and interactive 3D graphics. This feature makes use of the XTK package [20,21], which enables 3D graphics via WebGL that can be interactively rotated. Phase boundaries are shown as partially transparent triangulated surfaces with tie-lines connecting them (see Fig. 4). Both Gibbs triangle and Gibbs tetrahedron (see Fig. 5) representations are available. The underlying graphic files can also be download in standard VTK
Paraview gnuplot
Fig. 1. Overall organization of the Thermodynamic DataBase DataBase effort.
NIST solder database [22], some open databases supplied with MatCalc [11] and some contributed TDB files (e.g. [23]). The end result is (currently) a total of 766 data entries arising from 528 publications. This database will be updated periodically by scanning new issues of the CALPHAD journal and re-scanning the NIMS and NIST databases, among other web crawling efforts. We then matched each paper to its Digital Object Identifier (DOI), which provides a permanent pointer to the data source. This also enables a simple mechanism to generate appropriate bibliographic citations in any format via the Crossref service [15]. While the DOI assignment for CALPHAD Journal articles and NIST database entries was straightforward, for articles cited in the NIMS database, DOI matching was achieved through queries to the Crossref service. If no correct match is found (which may happen, for instance, for older papers), a standard URL (Uniform Resource Locator) is used instead. In designing this system, we wanted to ensure that copyrights or the terms of service of the data sources were respected. For these reasons, the TDBDB website does not actually provide the content of the TDB files or the papers. It acts as a search engine and yield links to the requested resource. It is up to the users to ensure they have the credential to access the data. For data hosted in the NIMS database this simply involves creating an account on the NIMS site (which is free). For data or articles hosted in scientific journals, one needs an electronic subscription to the journal (which most researchers have via their respective institutions). As a result, the TDBDB service will most likely increase the visibility and access statistics of the data providers rather than reduce them. The user interface is deliberately minimalist, as this provides crisper user feedback, eases code development and maintenance and makes it easier to access the website on devices with smaller screens (e.g. smart phones). Users simply enter the elements of interest and the system displays a list of databases containing data on all of the requested elements. Along with each database entry are reported links to the paper, its bibliographic reference and to the TDB file. The TDBDB web interface also provides a quick preview of the system's phase diagram (based on the database selected). This feature is made possible by the OpenCalphad [8,16] software. As the plotting capabilities of OpenCalphad currently require some user guidance, we opted for a simple robust brute-force method to generate the phase diagram preview. A large number of randomly chosen temperaturecomposition coordinates are generated and the corresponding phase equilibria are calculated with OpenCalphad. The points yielding singlephase equilibria are discarded and the remaining multiphase equilibria are sorted by phases and the temperature composition data points belonging to the same multiphase equilibrium are re-connected to form the phase boundaries. For 2D sections, this re-connection algorithm reduces to a simple nearest neighbor search. In higher dimension, this amounts to a type of Delaunay triangulation [26]. The advantages of this scheme are fourfold. First, it is easily scalable from a fast lowprecision representation to a slower high-precision one by merely adjusting the number of randomly chosen points. Second, it can easily capture phase fields located anywhere in the chosen section (without
Fig. 2. Example of binary phase diagram using the data from Reference [24]. 174
Calphad 61 (2018) 173–178
A. van de Walle et al.
Fig. 3. Example of isothermal section of a ternary phase diagram using the data from Reference [25]. Fig. 5. Example of an isothermal section of quaternary phase diagram using the data from Reference [28].
suitable for gnuplot or 3D viewers supporting the widely used VTK format. The ocutil command fixes common issues that prevent some TDB files from being readable by OpenCalphad. It cleanly truncates text lines to the right length, eliminates nonprintable stray characters arising from file format conversions, re-orders some keywords in the order OpenCaphad expects them, etc. The web front end of TDBDB interacts with the back end server via a standard RESTful (Representational state transfer) API (Application Programing Interface) and via data structures in the standard JSON (JavaScript Object Notation) format. This feature is of independent interest and the protocol and formats currently used are described in the Appendix. 3. Discussion The TDBDB service complements existing materials thermodynamic resources. While the aim of the TDBDB effort has some similarity with NIST's materials data repository [13,14] or the NIMS database [12], it exhibits some key differences. Most importantly, the system actively fetches the data rather than relying on researchers to upload or type in data. Of course, both the NIST and the NIMS databases contribute significantly to TDBDB's usefulness, as they provide extensive sources of data that would otherwise be unavailable. In particular, the NIMS database represents a tremendous resource by providing data in electronic format based on the thermodynamic information that would otherwise only be available in tabular form within the original papers. TDBDB does offer some additional features such as allowing users to interactively view isothermal sections of interest in multicomponent systems and providing easy-to-use links to papers and citation information. It should be noted that, compared to the TDBDB, NIST's materials data repository [13] is far broader in scope, contains richer metadata and allows users to upload their data. Also, at this point, the phase diagram plots available in the NIMS database [12] are generally of higher quality than the previews generated within the TDBDB. One of TDBDB's
Fig. 4. Example of 3D representation of a ternary phase diagram using the data from Reference [27].
(Visualization ToolKit) format [29] for off-line viewing with more powerful software, such as Paraview [30,31], which enables the calculation of general cross-sections interactively (see Fig. 6). A convenient by-product of these efforts is a set of tools, available within the ATAT package [17,18], that facilitates handling and plotting of phase diagrams with OpenCalphad. The ocplotpd command implements all of the website's phase diagram plotting capabilities. It calls OpenCalphad to calculate equilibria and generates graphical output 175
Calphad 61 (2018) 173–178
A. van de Walle et al.
Fig. 6. Example of interactive viewing of general sections of a ternary phase diagram obtained with Paraview based on a 3D phase diagrams exported from TDBDB using the data from Reference [32].
triangulation of the phase boundaries in general n-dimensional space has already been implemented but cross-sectioning operations enabling plotting as 2D or 3D graphs has not.) The TDBDB could also be expanded to include links to commercial databases, so that researchers can at least be made aware of the existence of ready-made databases for their systems of interest (even if the data itself is not openly available). Another way to extend TDBDB's coverage would be to mine older papers to extract thermodynamic solution parameters provided in tabular form to build ready-to-use TDB files (although this has already been accomplished in large part thanks to the NIMS database [12]). With the large amount of thermodynamic data now readily available through a single interface, one could imagine allowing users to more easily perform CALPHAD “extrapolation” by combining known binaries or ternaries to generate reasonable predictions for higher order systems. This task would entail a standardization of phase names as well as other aspects of the TDB files (e.g. mutual consistency of the free energy models), which would require a significant effort. It would also be possible to fill gaps in the current set of available assessments with automated high-throughput CALPHAD methods [43,44], made possible in part by the recent development of a formal approach to ensuring that ab initio and experimental energies are mutually consistent [45]. Finally, we hope that this effort will provide further incentives to include thermodynamic information in electronic format as supplementary
novel contributions is to make the thermodynamic data available as supplementary information to CALPHAD Journal articles much more readily accessible. There also exists a number of sources of thermodynamic information primarily in the form of ab initio-calculated absolute-zero formation energies [33–42]. In contrast, TDBDB's focus is on finite-temperature thermodynamics properties arising primarily from experimental data, with increasingly more data arising from electronic structure calculations in recent years. 4. Conclusion The current extent of the TDBDB project represents the beginning of a larger scale effort that could include some of the following aspects. The system currently only indexes files in the “TDB” format, but one could also include other formats as well [1,10]. The phase diagram preview capabilities are currently limited and could benefit from a better algorithm to connect the calculated data points to form smooth phase boundaries. The website's preview capabilities could also offer the option to invoke the pycalphad toolkit [9] for plotting purposes. As an increasing number of thermodynamic models include 4 or more components, the ability to represent higher-dimensional phase diagrams via interactive cross-sections would be a plus. (Currently, 176
Calphad 61 (2018) 173–178
A. van de Walle et al.
information in CALPHAD journal publications and encourage other journals to follow suit in requesting the submission of supplementary thermodynamic information in electronic form.
Acknowledgments This research was supported by ONR under grant N00014-17-12202, and by Brown University through the use of the facilities at its Center for Computation and Visualization. This work uses the Extreme Science and Engineering Discovery Environment (XSEDE) resource Stampede 2 at the Texas Advanced Computing Center through allocation TG-DMR050013N, which is supported by National Science Foundation grant number ACI-1548562.
TDBDB then returns a JSON-formatted object with the following keys:
Appendix A. Application Programing Interface (API) In describing the API, the following conventions are used. The string http://tdbdb/ stands for the website's URL. Sections marked as [some expression] need to be replaced by the actual parameters needed. Any text to be included verbatim is indicated by a courier font. Interaction with the server typically starts by a GET request to list the available database containing data regarding the desired elements of the form: http://tdbdb/getdbid.php?element=[comma-separated list of element symbols] The returned information is a JSON-formatted object with the following keys:
References [1] C. Bale, E. Blisle, P. Chartrand, S. Decterov, G. Eriksson, A. Gheribi, K. Hack, I.H. Jung, Y.-B. Kang, J. Melanon, A. Pelton, S. Petersen, C. Robelin, J. Sangster, P. Spencer, M.-A.V. Ende, FactSage thermochemical software and databases, 2010–2016, Calphad 54 (2016) 35. [2] J.O. Andersson, T. Helander, L. Höglund, P.F. Shi, B. Sundman, Thermo-Calc & DICTRA, computational tools for materials science, Calphad 26 (2002) 273. [3] P. Villars, H. Okamoto, K. Cenzual (Eds.), Alloy Phase Diagrams Center, ASM International, 2007. URL 〈http://www.asminternational.org〉. [4] A. van de Walle, The thermodynamic database database (TDBDB) (2017). URL 〈https://alum.mit.edu/www/avdw/tdbdb.html〉. [5] G. Eriksson, K. Hack, ChemSage– A computer program for the calculation of complex chemical equilibria, Metall. Trans. B 21B (1990) 1013. [6] C. Bale, P. Chartrand, S. Degterov, G. Eriksson, K. Hack, R.B. Mahfoud, J. Melançon, A. Pelton, S. Petersen, FactSage thermochemical software and databases, Calphad 26 (2002) 189. [7] W. Cao, S.-L. Chen, F. Zhang, K. Wu, Y. Yang, Y. Chang, R. Schmid-Fetzer, W. Oates, PANDAT software with PanEngine, PanOptimizer and PanPrecipitation for multi-
The user would then typically initiate a download of the desired TDB file whose URL is given in the tdburl key. This is done without relying on the TDBDB service, typically with a simple wget [URL] command. The user can also request a preview of the phase diagram with the following GET request (line breaks inserted for clarity): 177
Calphad 61 (2018) 173–178
A. van de Walle et al.
[8] [9] [10]
[11]
[12] [13]
[14] [15] [16]
[17]
[18]
[19] [20]
[21] [22] [23]
[24] [25] [26]
[27] [28]
sb melts, Calphad 52 (2016) 159. [29] L.S. Avila, U. Ayachit, S. Barré, et al., The VTK Users Guide, Kitware, USA, 2010. URL 〈https://www.vtk.org/vtk-users-guide/〉. [30] J. Ahrens, B. Geveci, C. Law, Paraview: An end-user tool for large data visualization, in: C.D. Hansen, C.R. Johnson (Eds.), Visualization Handbook, Elsevier, New York, 2005. [31] U. Ayachit, The ParaView Guide: A Parallel Visualization Application, Kitware, USA, 2015. [32] S. Sridar, R. Kumar, K.C.H. Kumar, Thermodynamic modelling of ti-zr-n system, Calphad 56 (2017) 102. [33] Y. Wang, S. Curtarolo, C. Jiang, R. Arroyave, T. Wang, G. Ceder, L.-Q. Chen, Z.K. Liu, Ab initio lattice stability in comparison with CALPHAD lattice stability, Calphad 28 (2004) 79. [34] A. Jain, S.P. Ong, G. Hautier, W. Chen, W.D. Richards, S. Dacek, S. Cholia, D. Gunter, D. Skinner, G. Ceder, K.A. Persson, The materials project: a materials genome approach to accelerating materials innovation, APL Mater. 1 (2013) 011002. [35] K. Mathew, J.H. Montoya, A. Faghaninia, S. Dwarakanath, M. Aykol, H. Tang, I.H. Chu, T. Smidt, B. Bocklund, M. Horton, J. Dagdelen, B. Wood, Z.-K. Liu, J. Neaton, S.P. Ong, K. Persson, A. Jain, Atomate: a high-level interface to generate, execute, and analyze computational materials science workflows, Comput. Mater. Sci. 139 (2017) 140. [36] A. Jain, G. Hautier, C.J. Moore, S.P. Ong, C.C. Fischer, T. Mueller, K.A. Persson, G. Ceder, A high-throughput infrastructure for density functional theory calculations, Comp. Mater. Sci. 50 (2011) 2295. [37] J.E. Saal, S. Kirklin, M. Aykol, B. Meredig, C. Wolverton, Materials design and discovery with high-throughput density functional theory: the open quantum materials database (oqmd), JOM 65 (2013) 1501. [38] S. Kirklin, J. Saal, B. Meredig, A. Thompson, J. Doak, M. Aykol, S. Rühl, C. Wolverton, The open quantum materials database (oqmd): assessing the accuracy of dft formation energies, NPJ Comput. Mater. 1 (2015) 15010. [39] M. Mihalkovic, M. Widom, First-principles prediction of a decagonal quasicrystal containing boron, Phys. Rev. Lett. 93 (2004) 095507. [40] M. Mihalkovic, M. Widom, Widom Alloy Database, 2004. URL 〈http://alloy.phys. cmu.edu/〉. [41] S. Curtarolo, W. Setyawan, G.L. Hart, M. Jahnatek, R.V. Chepulskii, R.H. Taylor, S. Wang, J. Xue, K. Yang, O. Levy, M.J. Mehl, H.T. Stokes, D.O. Demchenko, D. Morgan, Aflow: an automatic framework for high-throughput materials discovery, Comput. Mater. Sci. 58 (2012) 218. [42] C.E. Calderon, J.J. Plata, C. Toher, C. Oses, O. Levy, M. Fornari, A. Natan, M.J. Mehl, G. Hart, M.B. Nardelli, S. Curtarolo, The AFLOW standard for highthroughput materials science calculations, Comput. Mater. Sci. 108 (2015) 233. [43] A. van de Walle, R. Sun, Q.-J. Hong, S. Kadkhodaei, Software tools for highthroughput calphad from first-principles data, Calphad 58 (2017) 70, http://dx.doi. org/10.1016/j.calphad.2017.05.005. [44] R. Otis, Z.-K. Liu, High-throughput thermodynamic modeling and uncertainty quantification for ICME, JOM 69 (2017) 886, http://dx.doi.org/10.1007/s11837017-2318-6. [45] A. van de Walle, reconciling SGTE and ab initio enthalpies of the elements, Calphad 60 (2018) 1, http://dx.doi.org/10.1016/j.calphad.2017.10.008.
component phase diagram calculation and materials property simulation, Calphad 33 (2009) 328. B. Sundman, U.R. Kattner, M. Palumbo, S.G. Fries, OpenCalphad – a free thermodynamic software, Integr. Mater. Manuf. Innov. 4 (2015) 1. R.A. Otis, Z.-K. Liu, pycalphad: calphad-based computational thermodynamics in python, J. Open Res. Softw. 5 (2017) 1, http://dx.doi.org/10.5334/jors.140. R. Davies, A. Dinsdale, J. Gisby, S. Hodson, R. Ball, Thermodynamic modelling using MTDATA: A description showing applications involving oxides, alloys and aqueous solutions, in: P. Nash, B. Sundman (Eds.), Applications of Thermodynamics in the Synthesis and Processing of Materials, TMS, 1995, p. 371. E. Kozeschnik, B. Buchmayr, MatCalc - A simulation tool for multicomponent thermodynamics, diffusion and phase transformation kinetics, Math. Model. Weld. Phenom. 5 (2001) 349. T. Abe, K. Hashimoto, Y. Sawada, M. Palumbo, Y. Ogura, Nims Database, 2007. URL 〈http://cpddb.nims.go.jp/cpddb/periodic.htm〉. C. Campbell, U. Kattner, Z.-K. Liu, File and data repositories for next generation CALPHAD, Scr. Mater. 70 (2014) 7, http://dx.doi.org/10.1016/j.scriptamat.2013. 06.013. NIST Materials Data Repository. URL 〈https://materialsdata.nist.gov/〉. Crossref. URL 〈https://www.crossref.org/〉. B. Sundman, U.R. Kattner, C. Sigli, M. Stratmann, R.L. Tellier, M. Palumbo, S.G. Fries, The Open Calphad thermodynamic software interface, Comput. Mater. Sci. 125 (2016) 188. A. van de Walle, M.D. Asta, G. Ceder, The alloy theoretic automated toolkit: a user guide, Calphad 26 (2002) 539–553, http://dx.doi.org/10.1016/S0364-5916(02) 80006-2. A. van de Walle, Multicomponent multisublattice alloys, nonconfigurational entropy and other additions to the Alloy Theoretic Automated Toolkit, Calphad 33 (2009) 266–278, http://dx.doi.org/10.1016/j.calphad.2008.12.005. T. Williams, C. Kelley, et al., Gnuplot. URL 〈http://www.gnuplot.info/〉. D. Hähn, N. Rannou, B. Ahtam, P.E. Grant, R. Pienaar, Neuroimaging in the browser using the X toolkit, in: Front. Neuroinform.: 5th INCF Congress of Neuroinformatics., 2014. URL 〈http://f1000.com/posters/browse/summary/ 1092491〉. D. Hähn, N. Rannou, B. Ahtam, P.E. Grant, R. Pienaar, The X Toolkit, 2014. URL 〈https://goXTK.com〉. U.R. Kattner, The NIST Solder Database, 2017. URL 〈http://www.metallurgy.nist. gov/phase/solder/solder.html〉. G. Ghosh, A. van de Walle, M.D. Asta, G. Olson, Phase stability of the Hf-Nb system: from first-principles to CALPHAD, Calphad 26 (2002) 491, http://dx.doi.org/10. 1016/S0364-5916(02)80003-7. F. Zhang, S. Liu, J. Wang, Y. Du, L. Sun, Experimental investigation and thermodynamic assessment of the li-sb system, Calphad 57 (2017) 28. C. Guo, T. Wu, C. Li, Z. Du, Thermodynamic description of the al-fe-nb system, Calphad 57 (2017) 78. H. Chen, J. Bishop, Delaunay triangulation for curved surfaces, in: Proceedings of the 6th International Meshing Roundtable, Sandia National Laboratories, 1997, p. 115. P. Lafaye, C. Toffolon-Masclet, J.-C. Crivello, J.-M. Joubert, Thermodynamic modelling of the cr-nb-sn system, Calphad 57 (2017) 37. W. Chen, L. Zhang, Y. Du, Diffusivities and atomic mobilities in sn-ag-in and sn-ag-
178