`Program to Perform Hanawalt or Pink Search-
`
`Indexirzg Using Organics Entries in the ICDD
`PDF-4/Organics 2003 Database”, Adv. X-Ray
`Analysis, v. 47 (2004) pp166-173
`
`RS 1032 - 000001
`
`
`
`-D
`
`Copyright ©JCPDS - International Centre for Diffraction Data 2004, Advances in X-ray Analysis, Volume 47.
`
`n
`
`A PLUG-IN PROGRAM TO PERFORM HANAWALT OR FINK SEARCH-
`
`INDEXING USING ORGANICS ENTRIES IN THE ICDD PDF—
`
`4!ORGANlCS 2003 DATABASE
`
`J. Faber, C. A. Weth and J. Bridge*
`
`international Centre for Diffraction Da ta (ICDD)
`Newrown Square, PA i.90?3, USA
`* West Chester University, Department of Computer Science,
`West Chester. PA 19380
`
`ABSTRACT
`
`In an attempt to fill a gap between fully automatic searchfmatch programs and purely manual
`methods based on paper products, a relational database plug-in has been developed that functions
`as a PC-based Searchflnclex program for extracting information from PDF-4 powder diffraction
`databases. The plug-in provides an adjustable search window and match window to account for
`experimental errors. Both Hanawalt [1,2] and Fink [3] search methods are incorporated. In this
`paper, we report search-indexing results obtained with the new PDF-4 plug-in applied to a new
`relational database, the PDF-4/Organics 2003. This database has 24,385 experimental entries
`and 122,816 calculated patterns derived from the Cambridge Crystallographic Database (CSD).
`We introduce a Goodness of Match (GOM) parameter to describe the relative agreement
`between the experimental input data and selected reference patterns from the PDF-4-fOrganics
`2003. The relevance of the GOM is illustrated in several example problems. Multiphase
`samples can be treated on a phase—by—phase basis.
`
`INTRODUCTION
`
`The International Centre for Diffraction Data (ICDD) has been the primary reference for X—ray
`Powder Diffi'action (XRPD) data for over 50 years. The primary information in the PDF is the
`collection of d-I data pairs, where the d-spacing (d) is determined from the Bragg angle of
`diffraction, and the peak intensity (I) is obtained experimentally under the best possible
`conditions for a phase-pure material. These data provide a data mining [4-5] capability as well
`as “f'u1gerprint" of the compound because the d-spacings are fixed by the geometry of the crystal
`and the intensities are dependent on the contents of the unit cell. Hence, d-I data may be used for
`identification of unknown materials by locating matching d-I data in the PDF with the d-I pairs
`obtained from the unknown specimen. Identification is the most common use of the PDF, but
`the presence of considerable supporting information for each entry in the PDF allows further
`characterization of the specimen. Examination of the crystal data, Miller indices, intensity
`values, scale factors, physical property data and the comprehensive literature reference data
`provide extraordinarily useful information concerning the specimen under study. For
`pharmaceutical R&D, XRPD and the PDF have been used for example as an indispensable tool
`in phase identification (both qualitative and quantitative), in the identification of unknowns,
`evolution of polymorphism and solvate structures, and crystallinity determinations. The impact
`of the PDF as a reference pattern database has been used in patent disclosures and as such has
`immediate impact for pharmaceutical R&D.
`
`‘brnx4.l1::DflIh\f ll nfinmcrw I)--1
`wwv;.ittld.g:_ntI1
`
`RS 1032 - 000002
`
`
`
`Copyright ©JCPDS - International Centre for Diffraction Data 2004, Advances in X-ray Analysis, Volume 47.
`
`The PDF has exhibited recent dramatic growth in entry population over the past 5 years.
`Historically, the PDF-2 has been a flat file database that contains powder patterns of inorganic
`compounds. However, in late 1998, the ICDD reached an agreement with the Cambridge
`Crystallographic Data Center (CCDC) that allows for the calculation of x-ray powder patterns
`from the structural information in the Cambridge Structural Database (CSD). The resultant
`explosion in the organic population fi'orn 25,000 to 150,000 entries in 2003 is a direct result of
`this agreement. A completely new relation database (RDB) was used to house the new PDF-
`4/Organics 2003. The principal classes of database compounds are organic and organo-metallic.
`The properties of new PDF-4 databases are illustrated in Table 1.
`
`We could anticipate that some search-indexing problems may arise using the PDF-4/Organics
`database:
`
`a Only organic entries are present in the PDF-4/Organics 2003 database. However, note
`that 1,117 inorganic compounds are present in the PDF-4/Organics 2004. Inorganic
`excipients are particularly relevant for pharmaceutical R&D.
`There could be larger uncertainties in the lattice parameters since single crystal
`experiments are often not optimized for high-resolution d-spacing determinations. The
`focus is on integrated intensities. Only 623 calculated pattern entries fiorn the CSD
`indicate that cell constants were obtained using powder diffraction methods.
`Organic entries are often done at low temperature. Comparison between low-temperature
`reference data and room temperature powder diffraction data does not account for
`thermal expansion in the reference pattem_
`Organic powder diffraction patterns often contain substantial preferred orientation
`effects. However, as we shall see, we have implemented rotation operators that permute
`the strongest lines in the pattern (in both Hanawalt and Fink analyses), which has the
`effect of taking preferred orientation into account. Severe preferred orientation effects
`cannot be overcome since this would completely distort these strong line/long line
`methods. We will discuss this issue in more detail.
`
`The focus of this paper is to present applications that demonstrate the power of a PDF-
`4/Organics 2003. We shall demonstrate this analytic power by illustrating results obtained from
`phase identification and search-indexing, using Hanawalt and Fink methods. A preliminary
`report for PDF-4/Full File 2002 (a predominantly inorganic database) has been given [6].
`
`RS 1032 - 000003
`
`
`
`Copyright ©JCPDS - International Centre for Diffraction Data 2004, Advances in X-ray Analysis, Volume 47.
`
`4'!
`
`l’|)l~'--4.1"
`(")rg-.mit:s I’-.l|1J3
`
`l’l)l-‘--la''
`l-‘till File ?.tII}3
`25,509
`133,370‘
`1,931
`131,439—
`_EE!Ifl
`4,508
`l,l92
`E Tl
`
`|’|)l>'—-I,-"‘-'-‘
`(')rg:mIt's Zllll-1
`
`Piments
`I/Ic
`
`dd
`125.34
`147.201
`
`218.194
`
`157,048
`
`Table 1. Selected entry counts of the PDF-4 databases. Please note that I-‘DF-4/Organics 2004 will be released in
`November, 2003. Please note that because entries can be listed in both the inorganic and organic collection, the total
`number of distinct entries is obtained from the organic and only inorganic rows in the Table.
`
`PDF-4/ORGANICS 2003
`
`1
`
`'
`
`P”
`
`lxsa 2‘-"9«\;:H| can Ernl C}u.n2'ur;-
`liui LE3‘-i'W3 Rx ml“
`Flilfll CIIIQ’
`In... in em:
`fiulnl
`ll-II
`ihil
`TI
`
`f--1 on»: ma. am.
`
`“>15-r.
`
`..
`
`-.a-u.a- 54...-:a..-at,»-,.._..
`
`,._._. -...,;,,m,
`
`“
`
`Figure 1. Example data from the PDF-4/Organics 2003 for Citric Acid. Note the 2D structure display and the on-
`the-fly digitized pattern.
`
`RS 1032 - 000004
`
`
`
`Copyright ©ICPDS - International Centre for Diffraction Data 2004, Advances in X-ray Analysis, Volume 47.
`
`N
`
`The PDF-4 database contains interplanar spacings (d) and relative intensities (1). However, other
`usefiil data such as synthesis, physical properties and crystallographic data are also stored in the
`database. With this new format, we will provide a broader range of analyses, for example,
`improved quantitative analyses, filll pattern display, bibliographic cross referencing, etc. The
`PDF-4 uses relational database technology that provides pliable access to the database to carry
`out data mining studies and enhances the pursuit of conventional materials characterization using
`ditlfraction techniques (see Faber et al. [5]).
`In addition to better access to some of the RDB
`fields, users can also build search criteria by combining individual search conditions using
`Boolean operators. The availability of logical operators for combining the search condition is
`very useful in arriving at the desired information from the database
`
`The CSD database is being used to calculate entries in the PDF. Thus, to derive d-spacing and
`peak intensity data requires the synthesis of full diffiaction patterns, i.e., we use the structural
`data in the CSD database and then add instrumental resolution information. In addition to the
`peak intensities, |F(hkl)!2 , the square of the structure factor magnitudes will also be calculated.
`Thus, calculated powder patterns are obtained for all CSD entries in the PDF-4fOrganics 2003
`RDB. For example, we can calculate (on—the-fly) a selected profile function to describe
`paracrystallinity, or particle size andfor strain effects. PDF data for an ideally random crystal
`distribution in the absence of preferred orientation may also be obtained. In the fixture, preferred
`orientation models will be developed. The main focus is to provide tools that can be used for
`materials design.
`
`The CSD contains bitmap control integers that can be used to project out specific categories of
`entries in the CSD. Ofparticular interest for pharmaceuticals is the drug activity flag. There are
`approximately 250,000 entries in the CSD and of these, approximately 8000 have the drug
`activity flag set. The PDF-4/Organics 2003 contains calculated patterns for 4, 292 of these
`entries. The process of calculating PDF data is an ongoing task; we will calculate powder
`patterns for all entries in the CSD when the ICDD editorial review has been successfiilly
`completed.
`
`SEARCH-INDEXING USING THE PDF-4/ORGANICS RDB:
`HANAWALT AND FINK SEARCIUMATCH PROCEDURES
`
`Most of the commercial software packages for qualitative phase identification have been
`designed to implement fully automatic search/match sequences [6—l7]. On the other hand,
`traditional methods of searchfmatch (based on d-spacings, intensities and chemistry) are mainly
`manual techniques using paper-based search/indices. Manual techniques were first discussed by
`Hanawalt and these persist for a variety of reasons. The search-indexing plug-in discussed here
`follows a traditional path to act as a replacement for paper search manuals published by the
`ICDD. An advantage to this approach is that Hanawalt and Fink methods can be followed in
`great detail as search—indexing proceeds. The educational benefit of this approach is also
`realized.
`
`RS 1032 - 000005
`
`
`
`Copyright ©JCPDS - lntemational Centre for Diffraction Data 2004, Advances in X-ray Analysis, Volume 47.
`
`0
`
`Traditional methods for searchfmatch in powder diffraction are based upon combinations ofd-
`spacing, intensities and chemistry. Table 2 lists the types of search indices currently being used.
`
`Table 2. Types of Data Search Indexes
`
`ndex I Search Parameters
`Alhabetic
`Chemist
`, chem. formula framents
`Perrnuted chemical formula framents
`anawalt
`d,I airs, sorted in decreasin intensit
`3 stronest lines
`
`Fink
`
`d,l airs, sorted in decreasin d—s ace
`
`8 Ion est lines (lonest of the stronest)
`
`The Hanawalt search method has been implemented for many years at the ICDD. The method
`involves sorting the patterns in the PDF according to the d-spacing value of the strongest line.
`This list is broken into discrete d-space intervals defined as Hanawalt groups. A small overlap in
`d—intervals is employed to reduce the probability of missing powder pattern entries due to
`uncertainty in the d-space accuracy. Each Hanawalt group is sorted in order of decreasing d-
`spacing of the second most intense diffiaction line. Subsequent lines are listed in order of
`decreasing intensity. The analysis rests on the three most intense lines, but the eight most intense
`lines are listed. Considerable redundancy exists in this method because patterns appear twice for
`the (1,2) and (2,1) pairs when I;/I; > 0.75 and lgfll > 0.75. Patterns appear three times (1,2),
`(2,1), and (3,1) when 13/11 > 0.75 and 14/1. < 0.75. The rationale for multiple entries is to
`minimize problems of preferred orientation, especially when these affect the three strongest
`lines. In summary, the Hanawalt method relies on the d~spaces for the three strongest lines;
`fiirther confirmation of a search hit is taken liom matches on the eight strongest lines.
`
`The Fink method was designed as an index based on the eight strongest d—spaces in the
`experimental pattern, but these are ordered in decreasing d-spacing. In short, the Fink method
`considers the 8 longest of the strongest diffraction lines. Creating permutations of these is not
`practicable for large databases as the corresponding paper manuals become enormous. However,
`as we shall see, permuting the Hanawalt three strongest lines or the Fink eight longest lines is
`straightforward using computer methods. For both the Hanawalt and Fink methods, the problem
`is that the associated paper manuals have grown cumbersome and difficult to use. In addition,
`the integration of elemental composition and other important ancillary information is not easily
`accomplished with these methods. Filtering criteria need to be “remembered” while carrying out
`the search/indexing process. We have developed a “plug-in" for the PDF-4 databases that
`implements the Hanawalt/Fink strategy, including chemistry, subfile and quality-mark filters.
`
`SEARCH-INDEX PLUG-IN
`
`The basic idea of the plug-in is to provide d,I pairs as input to the program. The d—spaces are in
`A and the P5 are peak intensity values from the x-ray powder diffraction experiment. As
`additional input, P is a phase parameter associated with each d,I pair; if P=1, the peak is included
`in the analyses, otherwise the peak is ignored. As we shall see, this is quite helpful in multiphase
`problems. Also, contaminant peaks can be easily excluded in the analysis by adjusting P. The
`principal input is from an ASCII file that contains the d,I pairs. However, the plug-in can also
`accommodate 2 I9 ,1 pairs if the first ASCII record also contains the wavelength. The uncertainty
`in (1, Ad , can be obtained by taking the derivative of Bragg’s Law:
`
`RS 1032 - 000006
`
`
`
`Copyright ©JCPDS - International Centre for Diffraction Data 2004, Advances in X-ray Analysis, Volume 47.
`
`0
`
`Ad=d . cot6l as.
`
`Eq. 1
`
`In the case of the Hanawalt method, the search window, 5W = A{29sw ) defines the Hanawalt
`group. The angular dependence of the search window and match windows are defmed by Eq. 1.
`The match window, MW = A(2l9Mw) defines the PDF entry lines that match with the
`
`experimental data- Up to eight strongest experimental lines appear as d] — d8 just above the
`match list box in Figure 2. The best match is assigned to PDF # 000161157. For this match, at
`least 4 of the 8 strong lines are matched as discussed below.
`
`sun can
`1u.m.or«-lam-:u-Inna:-urn
`a
`ml?
`9
`:3!!!
`tan an un
`"I"
`-
`'-in
`;u
`in:
`:m¢-
`I ‘ll
`im 3.: V-
`HI
`‘JN
`
`E
`
`um:
`um:
`1'1!
`2 Eli
`~11!
`'I..'|.
`um rim .-ea
`794-
`‘JR n|.\\
`Hi!
`1 no
`3 n‘~'-
`I'M - re:
`tfii
`A5}!
`I‘;
`Fl’-I
`3'‘ KIM!” KI
`
`_
`
`'
`
`K
`:
`-Ieqinifin ITETI
`xfm 3
`‘s L‘:
`:mE‘| T"
`_
`
`nu run: up
`szntvutu
`I'\'i.UHM in
`My-t':35! 5-
`Iiitdfiii 9
`
`n
`
`;
`
`H..liIafltIIfl'IC
`
`Figure 2. Hanawalt method applied to an over-the-counter medication. The drug is Alka-Seltzer Plus, normally
`ingested afier dissolution in water. The tablets were ground and standard XRD experiments were performed. A
`peak-listing program was used to define d-spaces and peak intensities for all Bragg lines detected.
`
`Match window hits are selected and an algorithm is used to obtain the hit that best matches the
`experimental data. This best match is obtained by calculating the GOM, defined by
`
`GOM=l000-Z[l—|6d|/.S‘W]’,
`
`Eq. 2
`
`where 6d is taken from the difference between the d-spacing for the unknown and the d-spacing
`for the candidate reference pattern, the sum is taken over the experimental lines and their
`corresponding match lines in the selected PDF entry, and SW is the search window defined from
`Eq. 1. Notice that a perfect match between experiment and the PDF for 8 lines would yield
`GOM=8000. Also, GOM values <1000 are not significant since this corresponds to the
`identification of only a Hanawalt Group. GOM -C2000 means that no single reasonable entry has
`been identified within the Hanawalt Group. For the analysis presented in Figure 1, GOM=43 90.
`
`RS 1032 - 000007
`
`
`
`Copyright ©.lCPDS - International Centre for Diffraction Data 2004, Advances in X-ray Analysis, Volume 47.
`
`Q
`
`The Intensity Scale Factor seen in Figure 2 is calculated based on all matched lines in the
`
`analysis.
`
`_-3] .9; .9 u[-at. g|m[gfg|yg_g33.lr- it i
`fljdflh ' 1 uId' Illfi
`l‘
`tlwwtliall mus
`Ixli-II'l.
`ant.-tr an-in-sum!
`I.'\-52>"-P
`-Iwgqanpnr -ran-an
`IIIIEZF?
`..,._.._...,...,_,.
`.
`,~
`-
`k-urn‘-flI'm'::r”a->5”.
`:l:?:4;:‘*
`3.3»-¢,,*,-R-‘,1
`Ijraru-,1
`mac;-mum.»
`uumur
`uunynnu
`."‘
`"
`
`_.
`
`I
`
`7'.
`
`-‘
`“
`’
`.
`r"t7““r“"t"*r”a':—"£"
`..-.‘E“5=’_|"‘-'.::....rgT~i'*"“"'R"
`____A “
`5
`.ao-H5:
`i
`I
`lit» [35
`l
`‘E
`,,_,_,_,_.””
`an:-:1 av.
`3
`I
`g___
`.
`1!
`-
`am. rum. 3:
`"MW?! 9
`3''
`nun
`‘*l-'*‘‘“’
`=’-‘
`‘
`'"4f:“é'l‘iW 3'3." "
`I"
`Tm’ WI" ' iii} ‘ &
`\'rI'h‘|:‘l| L
`6
`V
`F‘;-"—f—"
`ollhiriit
`I
`'
`‘
`'
`'é}.’I'33I2i
`E
`ltlthbfaillll
`Wfiflfll III
`Misfire: I:I
`
`4%,!
`‘nut
`'~==‘-gja
`
`J,“
`is;
`
`R
`we
`I
`'
`EF.'..”t‘«~!k“ix” 1
`
`f‘
`
`. Erljf
`eznv a";““tzfa%.'t.—m|rrr-3-1:
`,,...s.w,m..
`V"?
`
`Figure 3. Two additional phases were identified in this analysis: a second and third search were performed. The
`GOM=5l-10 for the second phase (not shown) and GOM=4247 for the third phase.
`
`The GOM=4247 (Figure 3) indicates that nearly 4 of the 8 strongest lines have been matched.
`In this case, all apart fi'om one weak line (d=l4.092, l=2) have been account for. Notice also in
`Figure 2 that relative scaling factors are indicated for each detected phase. However, these are
`only approximate since RIRs (reference intensity ratios) are not used in this analysis. Obtaining
`serni—quantitative analyses using reference intensity ratios is under development. One of the
`powerful features of the plug-in is that the data grid on the right side of Figs. 2-3 can be filled by
`any selection fi'om the hit list. The data grid lists all experimental lines and all lines fi'om the
`selected entry in the hit list. Thus, a detailed comparison between the match hit and the input
`data can be seen at a glance.
`
`Enhanced search indexing is realized by allowing for permutation of the order of the
`experimental lines, thus removing the need for redundant (and complex) indexing of reference
`data from the database. In the Hanawalt search method we have implemented rotation operators
`to pennute the order of the three strongest lines. This is indicated by the rotation counter:
`“Rotation: l of 3”, seen in Figs. 2 and 3. The rotation operator is also available for the Fink
`method (not shown), however, in this case since we are ordering 8 strongest of the longest d-
`space lines, the range is 1-8. The utility of these rotation operators is that they alter the order of
`the strongest lines and hence allow moderate degrees of preferred orientation to be taken into
`account. It has been our experience that when intensities are problematic, the Fink method tends
`to help resolve the component phases. We have implemented a third method, designated "long
`8" that completely ignores the peak intensities. The utility of this third method in currently
`under evaluation.
`
`RS 1032 - 000008
`
`
`
`Copyright ©JCPDS - International Centre for Diffraction Data 2004, Advances in X-ray Analysis, Volume 47.
`
`SUMMARY
`
`We have illustrated several examples of the complementary use of XRPD techniques coupled
`with a new organic database, the PDF—4fOrganics 2003 RDB. An example, Citric Acid, was
`used to show some of the power features available in this new PDF-4f0rganics. In particular,
`calculated on-the-fly powder patterns were generated and 2D structures are available. The
`search-indexing example, Alka-Seltzer Plus, was used to show search- indexing results using
`Hanawalt and Fink methods for phase ID. In this case, we were able to identify the three most
`abundant components in the tablet. We feel that the importance of the PDF-4/Organics 2003
`RDB will grow as its use becomes commonplace in the pharmaceutical community.
`
`REFERENCES
`
`[1] Hanawalt, J. D. and Rinn, H. W., Ind. Eng. Chem. Anal. 8, 244 (1936); Hanawalt, J. D.,
`Advances in X-Ray Analysis A}, 63-73 (1976).
`[2] Hanawalt, J. D., Cryst. in North America, Apparatus and Methods, American
`Crystallographic Association, Chapter 2, 1983, pp.2l5-219.
`[3] Bigelow, W. and Smith, J.V., ASTM Spec Tech Pub]. STP 372, 54-89 (1965).
`[4] Faber, J., Kabekkodu, S.N. & Jenkins, R. (2001), International Conference on Materials
`for Advanced Technologies, Singapore, unpublished; Kabekkodu, S.N., Faber, J. and
`Fawcett, T., Acta Cryst., Vol. B58, 333-337 (2002).
`[5] Faber, J. and Fawcett, T., Acta Cryst., Vol. B58, 325-332 (2002).
`[6] Faber, J., Weth, C. A. and Jenkins, R. (2001), Materials Science Forum Vol. 378-381,
`106-111 (2001).
`[7] Johnson, G. G., Jr., and Vand, V., Ind. Eng. Chem., Vol. 59, 19 (1965).
`[8] Nichols, M. C., Lawrence Livermore Lab. Report UCRL—70078 (1966).
`[9] Frevel, L. K., Adams, C. E. and Ruhberg, L. R., J. Appl. Crystallogn, Vol. 9, 300-305
`(1976).
`[10] Marquardt, R. G., J. Appl. Crystallogn, Vol. 12, 629-634 (1979).
`[11] Snyder, R. L., Advances in X-Ray Analysis Vol. 24, 83-90 (1980).
`[12] Jobst, B. A., Goebel, H. E., ibid Q, 273-282 (1981).
`[13] Parrish, W., Ayers, G. L., Huang, T. C., ibid Vol. 25, 221-229 (1981).
`[14] Jenkins, R., Hahm, Y., Pearlrnan, S. and Schreiner, W. N., ibid Vol. 23, 279-285
`(1979).
`[I5] Goehner, R. P. and Garbauskas, M. F., X—Ray Spectrom. Vol. 13, 172-179 (1984).
`[16] Toby, B. H., Powder Diffi-action Vol. 5, 2-7 (1990).
`[17] Caussin, P., Nusinovici J. and Beard, D. W., Advances in X-ray Analysis, Vol. 31, 423-430
`(1987); ibid Vol. 32, 531-538 (1988); Nusinovici J. and Winter, M. J., ibid Vol. 37, 59-66
`(1993).
`
`RS 1032 - 000009