DB2dI: Extracting Peak Height d-1 Data for the Powder Diffraction File from Rietveld Analyses

D. K. Smith, Department of Geosciences and Materials Research Laboratory, The Pennsylvania State University, University Park, PA 16802 and R. A. Young, School of

Physics, Georgia Institute of Technology, Atlanta, GA, 30332.

ABSTRACT

DB2dI is a DOS-executable program for extraction peak-height d&I data from Rietveld analysis results for submission to the ICDD Powder Diffraction File, PDF-2. A simple listing of the calculated d and intensity values from the output of a Rietveld refinement is not sufficient. The PDF-2 is a searchable database of d values for discernible reflections and of intensity data based on peak heights alone whose primary application is phase identification. Further, as a practical matter, the search software should return only a reasonable number of “hits” realistically related to the resolution of the laboratory X-ray powder diffraction instruments normally used. Therefore a single d value is needed for each cluster of reflections not individually resolved by the instrument.

DB2dI has been developed to extract peak-height d’s and I’s for submission to the Powder Diffraction File from the results files obtained from a Rietveld refinement when using DBWS-9411 or later versions. The algorithm uses the calculated trace to locate each resolvable peak and determines its position and height. It then determines all the contributing reflections to each peak and retains for the final listing the hkl’s of the two strongest contributors. The routine obtains other necessary information from the results file and by interrogating the user. Once all the necessary information is available, DB2dI prepares an output file in AIDS format for direct submission to the International Centre for Diffraction Data.

INTRODUCTION

The Powder Diffraction File, PDF, is the primary source for X-ray powder diffraction d-1 data for the identification of crystalline phases. It was first published in 1941 containing 1000 diffraction patterns and will contain 115,344 patterns with the 1998 release. Historically, the data were based on visual interpretation of line positions and darkening on films and on peak-height position in diffractometer traces. For overlapped peaks, these readings lead to weighted averages for the d-1 values that were reported. Because the experimental data used for phase identification were read in the same manner, the peak-height information proved very satisfactory for the procedures used.

The PDF is still based on peak-height data. In 1998, the International Centre for Diffraction Data, ICDD, issued the usual yearly update containing 2,500 new patterns and with an additional 43,275 patterns calculated from the Inorganic Crystal Structure Database, ICSD. Except for a few entries in the update, all these patterns were based on peak-height positions and intensities. The calculated patterns from the ICSD were all prepared using POWD12+ (Smith and Johnson, 1998) that produced d-1 output lists

Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42 223Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42 223ISSN 1097-0002

This document was presented at the Denver X-ray Conference (DXC) on Applications of X-ray Analysis. Sponsored by the International Centre for Diffraction Data (ICDD). This document is provided by ICDD in cooperation with the authors and presenters of the DXC for the express purpose of educating the scientific community. All copyrights for the document are retained by ICDD. Usage is restricted for the purposes of education and scientific research. DXC Website – www.dxcicdd.com

ICDD Website - www.icdd.com

ISSN 1097-0002

based on the peak-height using essentially the same algorithm described in this paper. In order to maintain compatibility with the present PDF procedures, this algorithm has been adapted to read the data produced following a Rietveld refinement when using DBWS9411 (Young et al., 1995) and later versions. The algorithm could be adapted to other refinement programs also.

There is considerable debate within and without the ICDD on whether the peak- height representation has served its usefulness and should be replaced with fully resolved position and integrated intensity lists. Synchrotron and Guinier measurements are so well resolved now that there is a use for such data. However, the majority of users of the PDF acquire their data on laboratory-based, computer-controlled diffractometers whose resolution is still considerable less than the synchrotron or Guinier devices. Peak lists are prepared from the raw data that locate the peak positions and heights based on algorithms that still correspond to the classical approach, and the need for peak-height representations is still present. In the early development of POWD, the resolved, integrated position and intensity tables proved unsatisfactory for identification purposes which is why the trace simulation and peak position and intensity output was developed.

Problem of d-1 extraction from Rietveld results

Rietveld studies are based on powder diffraction data that was acquired with considerable care and often represents the best experimental data that anyone has obtained on the compound under study. This situation is true when the specimen used is a single phase or when there is very little second phase. There is no question that these data must be archived, and diffractionists are encouraged to submit these raw diffraction traces to ICDD for such purposes. In addition, a derived d-1 list should accompany the submission. There are several ways a d-1 list could be derived. 1. A peak-finding algorithm could be used to examine the original raw data in the same manner that is done on most APD systems. 2. The peak ositions could be calculated from the refined unit cell and the intensities based on plF” i!Y ‘I2 where p is the multiplicity. 3. The peak positions and integrated intensities could be obtained from the refined output data. 4. The peak- height d-1 list could be calculated from the refined crystal structure using a program such as POWD. 5. The peak positions and peak-height intensities could be located on the calculated trace that best fit the experimental data. This latter option represents profile fitting with restraints of a crystal structure and should yield the d-1 list most closely applicable to the PDF.

The POWD peak-finding algorithm

POWD was first developed in the early 196Os, and the peak finding algorithm was added in the late 1960s. It was modified and perfected in the 1970s when the ICDD first accepted calculated patterns in the PDF. Since that time, there have been few changes in the algorithm. POWD uses the crystal structure information to first calculate a set of theoretical 20’s (d’s), integrated I’s, and hkl’s for all non-zero reflections. The trace, which would be obtained from a diffractometer, is then generated by distributing each reflection, including the ai and a2 components, over the appropriate range on the 28

Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42 224Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42 224ISSN 1097-0002

scale, using a selected profile whose area is proportional the integrated intensity. The resolution of the peaks is controlled by the full-width at half maximum, FWHM, function. This trace is then examined to determine the locations and intensities of all resolved peaks. Because the trace is a smooth function of 20 and calculated in selected steps (0.02,0.01”28, etc.), a peak is defined as any point whose intensity value is higher than both the adjacent points on either side. Shoulders where there is no valley are not detected as peaks.

Once the peaks are located, the positions (in step units) and peak heights are tabulated. POWD then re-examines the full list of possible reflections to determine which reflections contribute to each peak. A reflection is considered a contributor if its intensity is non-zero and its theoretical position is within one FWHM of the peak position. (In this paper, the term “peak” will imply the location of a maximum in the trace and “reflection” will imply the theoretical position for each allowed hkl based on the unit cell.) If there is only one contributor, the hkl is assigned to the peak list. If there are two contributors, both hkl’s are assigned to the list. If there are more than two contributors, only the hkl’s of the two strongest contributors are retained for the list, and a plus sign, “+“, is added to the two retained. Once the d, I, hkl table is completed, POWD then prepares a tile in AIDS format for the PDF.

The algorithm for DB2dI follows the same concept except that the calculated trace as stored in the PLOTINFO file (reamed PLOTINFO.DAT for DB2dI) is the starting point. The cell parameters are obtained from the *.OUT file.

Preparing the AIDS file requires additional input from the user because the information is not contained in the *.OUT and PLOTINFO.DAT files. This information is requested in a dialogue that reminds the user how to supply the input. The AIDS format is very specific and requires most input to be in a very exacting form and sequence. DB2dI is designed to assist the user to meet the requirements with a minimum of effort.

Using DB2DI

DB2dI is to be used following a successful refinement of a single-phase using DBWS9411 or a later version where the formats of the output data files are matched. The PLOTINFO (renamed to PLOTINFO.DAT) file and the last *.OUT file must be checked and saved. After DB2dI has processed these files and extracted all the information available, it then requests the user to verify some of the information such as the formula, chemical name, and space group to confirm that they are properly written for the AIDS output. The crystal system and mineral or common name are confirmed or added. The user must then supply the reference information, identify the wavelength used, and supply the user name and affiliation

DB2dI is not designed to operate on output files that contain information for more than one phase. If the phase of interest is dominant in the mixture, the refinement should be run to include that phase only before applying DB2dI. The final fit of the data for one

Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42 225Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42 225ISSN 1097-0002

phase should be compared to the multi phase refinement to confirm that the differences are small. Examining the graphics usually would be sufficient especially if the impurity peaks not fit are no more than one percent of the maximum peaks of the main phase.

The user should now send the initial raw data file along with the AIDS file and any other documentation such as the chemical analysis and related manuscripts to the Editor, Powder Diffraction File, International Centre for Diffraction Data, 12 Campus Boulevard, Newtown Square, PA 19073-3273, USA.

~101 Profile-5, SrZCa2Pt06. Dee 28, 1996 150

50/26 10.7.9

: I”““,“,‘“,““‘,“’

69.00 71.00 72.00 73.00

Figure 1. A section of the Rietveld plot from the study of Sr2Ca2PtOe covering the range 68.3’20 to 74.4’28.

An example

To illustrate the operation of DB2dI, a refinement of Sr2Ca2PtOb is used (Wong- ng et al., 1998). Figure 1 shows a section from the final Rietveld plot illustrating the range 68.3”28 to 74.4’28 where there are overlapped peaks. The tick marks indicate the positions for reflections in the range (long tics are al positions and short ticks are a2 positions). In the range 71.6”28 to 72.4’28, there is a peak formed by two contributing reflections [(146) Iintc84.5 and (416) IintE74.51 exactly superposed at 71.835’28 (because of the R-3c space group) that acts as a singlet peak [Ipk=60, relative to Ioi2=999]. This

Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42 226Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42 226ISSN 1097-0002

peak is reported as an unresolved peak with both Miller indices. It is quite evident that a peak-finding program would have difficulty with the raw data profile locating the best peak position. The calculated trace, however, is not difficult to interpret with respect to the peak location. This example illustrates the reason for using the calculated trace for the DB2dI analysis.

The range 72.6’28 to 73.5’28 in Figure 2 shows a peak that is a result of two non- superposed contributing reflections. This peak is interpreted by DB2dI as one peak with the position corresponding to the maximum as the peak location and the peak height normalized to a relative scale whose strongest value is 999 as the intensity for both contributors. For this example, the 128 peak is at 72.727’28 [Ii,,=75.0], and the 425 peak is at 73.030”28 [Ii,t=12.6]. The combined peak is located at 72.740’28 [Irk=291 and is also reported with both Miller indices. It is obvious that the higher intensity of 128 dominates the peak location.

For more complicated peak clusters such as that illustrated in the range 70.2’28 to 71.5”28 in Figure 2, DB2dI will determine the location from the calculated trace. It will also determine all the contributing reflections and select the two strongest contributors to list for the Miller indices. In this example, there are 4 contributing reflections, 70.681’28 [(250) Iintz31.0 and (250) Iint=31.0], 70.696’28 [(137) IintE21.91, and 70.810’28 [(342) Ii”t=90.0]. The peak is reported at 70.800’28 [Ipk=61] with indices (342) and (250) and a plus sign to indicate additional contributors.

Program availability

DB2dI will be distributed as part of DBWS-9807 and is also available for users of DBWS-9411 by contacting R. A. Young directly (young@physics.gatech.edu).

References

Smith, D. K. and Johnson, G. G, Jr. (1998) “POWD 12+“, (FORTRAN version available from the authors by supplying two TK70 DECtapes to the PSU authors at the address given in the header.)

Young, R. A., Sakthivel, A., Moss, T. S., and Paiva-Santos, C. 0. (1995) “DBWS9411 - an upgrade of the DBWS*.* programs for Rietveld refinement with PC and mainframe computers”, J. Appl. Cryst., 28, 366-367.

Wong-ng, W., Kaduk, J. A., and Young, R. A. (1998) “An Investigation of (Sr4-&,CQ+s)PtOb Using X-ray Rietveld Refinement”, Powd. Diff., (in press).

Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42 227Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42 227ISSN 1097-0002