Analysis of Interlaboratory Measurements on the Vapor Pressure of Gold (Certification of Standard Reference Material 745) Robert C. Paule and John Mandel A detailed statistical analysis has been made of results obtained from a series of interlaboratory measurements on the vapor pressure of gold. The gold Standard Reference Material 745 which was used for the measurements has been certified over the pressure range 10-8 to 10-3 atm. The temperature range corresponding to these pressures is 1300-2100 K. The gold heat of sublimation at 298 K and the associated standard error were found to be 87,720±210 cal/mol (367,040±900 J/mol). Estimates of uncertainty have been calculated for the certified temperature-pressure values as well as for the uncertainties expected from a typical single laboratory's measurements. The statistical analysis has also been made for both the second and third law methods, and for the within- and betweenlaboratory components of error. Several notable differences in second and third law errors are observed. Key words: Components of error (within and between-laboratories); gold; heats of sublimation (second and third law); interlaboratory measurements; standard errors; standard reference materials; vapor pressure. 1. Introduction This report is part of a program to establish five standard reference materials. The materials, Cd, Ag. Au, Pt, and W, are being certified by the National Bureau of Standards for their vapor pressures as a function of temperature. Certification covers the 10- to 10-3 atm range. For the complete series of materials, the temperatures corresponding to the above pressures will range from 600 to 3000 K. Gold, the first material to be certified, covers a temperature range from 1300 to 2100 K. The gold standard reference material is now available for sale to the public [1].1 Experience in high-temperature vapor-pressure measurements has shown that large systematic errors in pressure of 30, 50, or even 100 percent are not uncommon, even among experienced investigators. The vapor pressure standard reference materials will allow workers in the field to detect such systematic errors and to evaluate the precision and accuracy of their measurements. The materials should be most useful for checking low vapor pressure measurement methods such as the Knudsen, torque Knudsen, Langmuir, and mass spectrometric methods. This report will give estimates of the uncertainty of the certified temperature-pressure values as well as estimates of the uncertainties of a "typical" Single laboratory's measurements. These uncertainties summarize results obtained from interdboratory tests made in 1968 (see list of cooperative laboratories). The uncertainties represent current practice and should not be considered fixed with respect to time and progress. We believe the uncertainties will be reduced in the future through the use of the vapor pressure standard reference materials. Figures in brackets indicate footnotes and references beginning on p. 7. The detailed temperature-pressure data from the 11 laboratories which measured the vapor pressure of gold are given in table 2 (see sec. 6.3). Plots of the data are given in figures 1 through 5 of section 6.3. The solid line in these figures represents the pooled curve for all accepted data from all laboratories. A total of 41 sets of data (runs) with over 350 temperature-pressure points were available for consideration. Each temperature-pressure run has been used to obtain both the second and third law heats of sublimation at 298 K. Equation (1) was used to calculate the individual third law AH sub298 values corresponding to each temperature-pressure point and the average AH sub298 value was calculated for each run. The evaporation coefficient for gold ha been assumed to be unity. In agreement with thi assumption, we observed no evidence of trend i third law heats with changing orifice area. The second law heat for each vapor pressure temperature run was obtained by least-squares fitting the A and B constants in the equation: 3. Statistical Analyses Two OMNITAB programs were written to perform the statistical analyses, the ultimate purpose of which was to obtain overall weighted average values of the second and third law heats of sublimation and estimates of the uncertainties. The first OMNITAB program performed least-squares fits for each run to obtain the second law heats and the average third law heats. The program also made a preliminary test to detect laboratories that exhibited excessive scatter of points about the fitted curves (see sec. 6.1). The authors then examined the results and made tentative decisions regarding the data to be excluded from the weighted averages and the estimated uncertainties [7]. The secon OMNITAB program was then run to determine (1 the weighted average values of the second an third law heats of sublimation, (2) the uncertaintie associated with the heats, and (3) the uncertaintie expected for a typical in-control laboratory's meas urements (see sec. 6.2). In the second OMNITA program the rejected data were not used for the calculation of the weighted averages and standar deviations, but were used in all other statistica tests. This procedure avoids distorting the overal results, but still allows for further evaluation of al of the data. The statistical analyses indicate the weighte |