1	Expressing Measurements and Chemical Systems for Physical Property Data Peter Linstrom National Institute of Standards and Technology Gaithersburg, MD, USA
2	Outline Nature of physical property data Historical record or interpretation Limitations of automated systems Problem areas Summary
3	Physical Property Data A numeric tuple which applies to a physical system Describing how the numeric value was obtained from the system is difficult Identification of techniques, equipment, ancillary data used in calculations and calibrations.
4	Physical Property Data Describing the system is difficult Identification of sample: chemical species and concentrations Recording the numeric tuple is easy
5	Historical Record or Interpretation Two goals (non-exclusive) goals: 1.) Historical record What was measured, computed, or estimated? How was this done? 2.) Basic knowledge What do we know about this property? What is the probable range of the numeric value?
6	Historical Record or Interpretation Historical record Does not change Applies to specific physical systems and measurements Basic knowledge Built on analysis of the historical record Applies to an “idealized” physical system Improves through scientific processes
7	Example 1998 Roux et al, Fraday Trans. Stability of dimethyl benzenedicarboxolates
8	Example 2 C₆H₅CO₂Me ® C₆H₄(CO₂Me) + C₆H₆ Endothermicity of gas phase reaction: ortho 52.3 kJ/mol meta 29.2 kJ/mol para 30.4 kJ/mol Quite different from dinitro and dicyano benzenes
9	Example 2002 Roux et al, Phys. Chem. Chem. Phys. D_fHº methyl benzoate gas (kJ/mol): 1971 Hall et al -299.8 1980 Guthrie et al -269.3 ± 5.1 1994 Pedley -287.9 ± 2.4 1998 Maksimuk et al -277.74 ± 1.2 2002 Roux et al -276.1 ± 4.0
10	Example 2 C₆H₅CO₂Me ® C₆H₄(CO₂Me) + C₆H₆ Revised endothermicity of gas phase reaction: ortho 28.7 kJ/mol meta 5.6 kJ/mol para 6.8 kJ/mol Similar to dinitro and dicyano benzenes
11	Automated Systems Automated systems require data with well defined semantics Portions of physical property data are recorded in natural language (literature) Need procedures to map information to a form appropriate for automated systems
12	Automated Systems Mapping of information to computer friendly semantics may involve Loss of information A judgement on the part of the archivist (introduction of information not explicitly contained in the original source) Blurring of the line between historical record and interpretation
13	Automated Systems Some options for expressing information Develop taxonomy of codes Token value pairs Incorporate into database design Text comments (loss of data processing capability) Ignore the information
14	Automated Systems
15	Automated Systems Proper design of systems for expressing data requires significant domain knowledge Definition of appropriate taxonomies, codes, etc. Knowledge of what will be important to future investigators Knowledge of what can be safely ignored
16	Some Problem Areas Chemical identification Taxonomies for methods Describing domain-specific meta-data
17	Chemical Identification Identification of pure species can be difficult Identification of mixtures is a superset of the problem for single species Chemical nomenclature is too complex for most data systems to handle
18	Chemical Identification Registry of species Simplifies identification to an integer number Maintained by third parties Species may not be in registry Identification may not be precise (isomers) Deprecated entries Users consult secondary sources – errors propagate
19	Chemical Identification Chemical structure No third party Less ambiguity, but more complex semantics Expensive to draw or look up Costs decreasing with modern technology
20	Chemical Identification Purity / uncertainty of composition May not be known Purification / synthesis technique may be provided Often omitted from database
21	Taxonomies for Methods Classification of the manner in which a value was obtained Instrument type, model form natural divisions Appropriate resolution determined by archivist How does one handle unique methods? Science is not static – taxonomies will grow
22	Taxonomies for Methods Lias, et al, Ionization Potential Database Compiled over many years Taxonomy for basic measurement types Additional codes added to supplement supplement taxonomy for new methods which cross existing hierarchical boundaries (e.g. electron impact and laser spectroscopy)
23	Domain Specific Meta-Data Meta-data recognized by archivist (domain specialist) as significant Need method to encode in computer friendly format Taxonomies Token value pairs
24	Domain Specific Meta-Data Affefy, Liebman, and Stein – Neutral Thermochemistry Archive Meta-data options expanded as archive grew Correction to current CODATA heats of formation: done, not-done, or not-possible Data disagrees with previously published data: acknowledged by author(s), or not acknowledged
25	Summary Two pairs of trade-offs Historical record vs. interpretation Semantic complexity vs. loss of information Important for archivists and researchers to be aware of the compromises that are made