STANDARD TEST DATA FOR XPS - 12 August 1996 J. M. Conny, C. J. Powell, and L. A. Currie Surface and Microanalysis Science Division, National Institute of Standards and Technology, Gaithersburg, MD 20899 Standard test data (STD) are simulations of analytical instrument responses that can be used to assess data analysis algorithms for extracting relevant information from analytical data. The STD on these disks are well- defined data sets for testing procedures that operators would choose for detecting, locating, and measuring the intensities of overlapping peaks in a doublet as measured by x-ray photoelectron spectroscopy (XPS). The STD are constructed to simulate a range of analytical situations that are encountered in practice, such as varying degrees of peak overlap, varying relative peak intensities, and varying levels of random noise. We aim first to determine the variability in peak parameters as determined by a single operator using his or her software and a particular set of data- analysis procedures. Second, we seek to determine the variability in peak parameters that might be obtained from different operators with different software systems and different data-analysis procedures for the same set of analytical situations. This initial set of STD for XPS is being sent to a small number of volunteers as a pilot study to assess the usefulness of the STD design described below. The STD were constructed from selected carbon 1s spectra of polymers that were measured on a Scienta XPS instrument and supplied by courtesy of Dr. D. Briggs. Each of the selected spectra had a single "main" carbon 1s peak. These spectra were smoothed and then fitted with fifth-degree spline functions. In most cases, the STD were constructed by adding the spectra for two polymers according to the design summarized below to simulate a spectrum that might be measured from a specimen consisting of two separate phases; that is, the carbon 1s peaks for the two component materials overlap. Different pairs of polymer spectra were used to represent different types of binary specimens, and adjustments were made in the energy scales so that the peak positions would not necessarily be the same from one binary specimen to another. For a spectrum with a given set of peak parameters, random (Poisson) noise was then added in replicate to simulate multiple experimental measurements of the same specimen. Although the STD have been derived from experimental data, the peak locations and intensities in each data set do not now correspond to the parent materials. This set of STD for XPS is based on the following factorial design which was developed following a workshop held last year and later suggestions from a number of users: Factor 1: Separation of component peaks in the doublet (a) No shoulder present in spectrum (based on the numerical second derivative). (b) Shoulder present in the spectrum (based on the numerical second derivative). (c) Valley present in the spectrum (based on the numerical first derivative). Factor 2: Relative intensities of peaks in the doublet (a) Weaker peak on the high binding-energy side of the doublet (b) Approximately equal peak intensities (c) Stronger peak on the high binding-energy side of the doublet Factor 3: Absolute intensity (level of Poisson noise) (a) High intensity (low fractional noise) (b) Low intensity (high fractional noise) This factorial design gives 18 combinations. In addition, two spectra with a single carbon 1s peak serve as null cases for factors 1 and 2. For these null cases, spectra were generated at the same two intensity levels. There is thus a total of 22 different combinations. Ten replicates were made for each combination (that is, different distributions of random noise to simulate repeated measurements for a given set of peak parameters) to give a total of 220 spectra in the STD set. We also provide two reference spectra for a single polymer so that participants can, if they wish, investigate a typical carbon 1s lineshape. One reference spectrum contains the measured data, and the other is the spline-modeled spectrum without noise for the same polymer. DIRECTORY STRUCTURE AND FILE FORMATS The 220 STD spectra and the two reference spectra each contain 261 intensity values that span the same binding-energy range, 294.45 eV to 281.45 eV in steps of 0.05 eV. The enclosed high-density disk (PC formatted) contain the following files with the indicated formats: \REF\ RAW.REF Reference spectrum of measurements. MODEL.REF Reference spectrum of spline-modeled data. For each reference file, format is ASCII text with three header lines followed by spectral data as energies and intensities (counts), tab delimited, in pairs (columns) on separate rows. There are 264 rows in each file. \SEPARATE.STD\ FILE001.DAT ... FILE220.DAT ASCII text with three header lines followed by the spectral data as energies and intensities (counts), tab delimited, in pairs (columns) on separate lines. Each of the 220 spectra is in a separate file. There are 264 rows in each file. INSTRUCTIONS FOR ANALYSIS OF STD We ask that you analyze as many of the 220 spectra as possible using procedures that you believe are suitable for the data. The first 22 spectra contain an example of each of the 22 combinations adopted in the design, the second 22 spectra contain other examples (with different noise distributions) for each of the 22 combinations (in a different random order), and so on. Repeated analyses of spectra with the same peak parameters but with different distributions of random noise will enable us to determine the uncertainties in the peak parameters obtained by each participant for each of the 22 combinations. Comparisons of the mean peak parameters determined by each participant for a particular combination with those determined by other participants will enable us to determine the variability in measurements based on different data-analysis procedures for each combination. For each spectrum, we wish to know: 1. The number of component peaks in each spectrum, 2. The binding energies for the identified peaks, and 3. The intensities (areas) of the identified peaks. This information can be tabulated in an ASCII file or otherwise using the Excel 4.0 worksheet (ANALYSIS.XLS) on Disk 1. Hardcopy of the worksheet is also provided. In addition, as indicated on the Data Return Form enclosed in this file and in hardcopy, we ask for information on the software you used, all steps of your data analysis procedure, the reasons for your precedure choices and uncertainty estimates. We wish particularly to know the fitting function you used and the type of background and end points if a background was subtracted. It would be useful if you could send several figures illustrating typical fits to representative spectra; the figures should show the background function and the end points. We would also be interested in any comments you may have on the STD design and on your experiences. The completed forms (or the completed files on the disk) should be returned to us by mail. Please report your results to either Joe Conny or Cedric Powell (to whom questions can also be addressed). We will then report back to you on the variability in your peak parameters for each combination in the design. After a sufficient number of returns have been received, we will report back to you on the variability in the peak parameters determined by different participants for each combination. We expect to prepare a report for publication on our findings, but the individual results will not be identified with the corresponding participants. We thank you for your cooperation in advance and look forward to hearing from you. Joe Conny Chemistry B-364 National Institute of Standards and Technology Gaithersburg, MD 20899 Phone: 301-975-3932 Fax: 301-216-1134 Email: jconny@enh.nist.gov Cedric Powell Chemistry B-248 National Institute of Standards and Technology Gaithersburg, MD 20899 Phone: 301-975-2534 Fax: 301-926-6689 Email: cpowell@enh.nist.gov ____________________________________________________________________________ STANDARD TEST DATA FOR XPS -- Data Return Form Name: __________________________________________________________________ Address: __________________________________________________________________ __________________________________________________________________ __________________________________________________________________ Phone: _______________ Fax: _______________ E-Mail: __________________ Type of software used for data analysis. Give vendor and version number. ____________________________________________________________________________ By what means did you decide whether there were one or two peaks in a spectrum (e.g., visual inspection, statistical test - please specify, etc.)? ____________________________________________________________________________ ____________________________________________________________________________ Please give details of all steps of the data analysis procedure, e.g., type of background and end points if a background was subtracted, type of fitting function, etc. Please attach figures with backgrounds showing typical fits for several representative spectra. ____________________________________________________________________________ ____________________________________________________________________________ ____________________________________________________________________________ ____________________________________________________________________________ ____________________________________________________________________________ ____________________________________________________________________________ ____________________________________________________________________________ What means did you use to determine the quality of the fit? ____________________________________________________________________________ ____________________________________________________________________________ What is a typical statistical uncertainty in your determination of: Peak position (eV) _________________________________________ Intensity (high intensity data) ____________________________ (low intensity data) ____________________________ How were the uncertainty estimates obtained (e.g., visual inspection, statistical tests - please specify, etc.)? ____________________________________________________________________________ ____________________________________________________________________________ Did you find the analyses easy, straightforward or hard? ___________________ Other comments on your ___________________________________________________ experiences in the analysis of the STD ___________________________________________________ ____________________________________________________________________________ ____________________________________________________________________________ Other comments on the ___________________________________________________ design of the STD set ____________________________________________________________________________ ____________________________________________________________________________ ____________________________________________________________________________ Do you think it would be useful to have additional STD for other XPS conditions (e.g., different analyzer pass energies or no x-ray monochromator) to simulate measurements with different energy resolutions or different types of instruments? Would it be useful to have STD for other purposes (e.g., assessment of detection limits) or for other techniques of surface analysis? ____________________________________________________________________________ ____________________________________________________________________________ ____________________________________________________________________________ ____________________________________________________________________________