International
Tables for
Crystallography
Volume B
Reciprocal space
Edited by U. Shmueli

International Tables for Crystallography (2010). Vol. B, ch. 2.1, pp. 203-207   | 1 | 2 |

## Section 2.1.7. Non-ideal distributions: the correction-factor approach

U. Shmuelia* and A. J. C. Wilsonb

aSchool of Chemistry, Tel Aviv University, Tel Aviv 69 978, Israel, and bSt John's College, Cambridge, England
Correspondence e-mail:  ushmueli@post.tau.ac.il

### 2.1.7. Non-ideal distributions: the correction-factor approach

| top | pdf |

#### 2.1.7.1. Introduction

| top | pdf |

The probability density functions (p.d.f.'s) of the magnitude of the structure factor, presented in Section 2.1.5, are based on the central-limit theorem discussed above. In particular, the centric and acentric p.d.f.'s given by equations (2.1.5.11) and (2.1.5.8), respectively, are expected to account for the statistical properties of diffraction patterns obtained from crystals consisting of nearly equal atoms, which obey the fundamental assumptions of uniformity and independence of the atomic contributions and are not affected by noncrystallographic symmetry and dispersion. It is also assumed there that the number of atoms in the asymmetric unit is large. Distributions of structure-factor magnitudes which are based on the central-limit theorem, and thus obey the above assumptions, have been termed ideal', and the subjects of the following sections are those distributions for which some of the above assumptions/restrictions are not fulfilled; the latter distributions will be called non-ideal'.

We recall that the assumption of uniformity consists of the requirement that the fractional part of the scalar product be uniformly distributed over the [0, 1] interval, which holds well if are rationally independent (Hauptman & Karle, 1953), and permits one to regard the atomic contribution to the structure factor as a random variable. This is of course a necessary requirement for any statistical treatment. If, however, the atomic composition of the asymmetric unit is widely heterogeneous, the structure factor is then a sum of unequally distributed random variables and the Lindeberg–Lévy version of the central-limit theorem (cf. Section 2.1.4.4) cannot be expected to apply. Other versions of this theorem might still predict a normal p.d.f. of the sum, but at the expense of a correspondingly large number of terms/atoms. It is well known that atomic heterogeneity gives rise to severe deviations from ideal behaviour (e.g. Howells et al., 1950) and one of the aims of crystallographic statistics has been the introduction of a correct dependence on the atomic composition into the non-ideal p.d.f.'s [for a review of the early work on non-ideal distributions see Srinivasan & Parthasarathy (1976)]. A somewhat less well known fact is that the dependence of the p.d.f.'s of on space-group symmetry becomes more conspicuous as the composition becomes more heterogeneous (e.g. Shmueli, 1979; Shmueli & Wilson, 1981). Hence both the composition and the symmetry dependence of the intensity statistics are of interest. Other problems, which likewise give rise to non-ideal p.d.f.'s, are the presence of heavy atoms in (variable) special positions, heterogeneous structures with complete or partial noncrystallographic symmetry, and the presence of outstandingly heavy dispersive scatterers.

The need for theoretical representations of non-ideal p.d.f.'s is exemplified in Fig. 2.1.7.1, which shows the ideal centric and acentric p.d.f.'s together with a frequency histogram of values, recalculated for a centrosymmetric structure containing a platinum atom in the asymmetric unit of (Faggiani et al., 1980). Clearly, the deviation from the Gaussian p.d.f., predicted by the central-limit theorem, is here very large and a comparison with the possible ideal distributions can (in this case) lead to wrong conclusions.

 Figure 2.1.7.1 | top | pdf |Atomic heterogeneity and intensity statistics. The histogram appearing in this figure was constructed from values which were recalculated from atomic parameters published for the centrosymmetric structure of C6H18Cl2N4O4Pt (Faggiani et al., 1980). The space group of the crystal is , , i.e. all the atoms are located in general positions. The figure shows a comparison of the recalculated distribution of with the ideal centric [equation (2.1.5.11)] and acentric [equation (2.1.5.8)] p.d.f.'s, denoted by and 1, respectively.

Two general approaches have so far been employed in derivations of non-ideal p.d.f.'s which account for the above-mentioned problems: the correction-factor approach, to be dealt with in the following sections, and the more recently introduced Fourier method, to which Section 2.1.8 is dedicated. In what follows, we introduce briefly the mathematical background of the correction-factor approach, apply this formalism to centric and acentric non-ideal p.d.f.'s, and present the numerical values of the moments of the trigonometric structure factor which permit an approximate evaluation of such p.d.f.'s for all the three-dimensional space groups.

#### 2.1.7.2. Mathematical background

| top | pdf |

Suppose that is a p.d.f. which accurately describes the experimental distribution of the random variable x, where x is related to a sum of random variables and can be assumed to obey (to some approximation) an ideal p.d.f., say , based on the central-limit theorem. In the correction-factor approach we seek to represent as where are coefficients which depend on the cause of the deviation of from the central-limit theorem approximation and are suitably chosen functions of x. A choice of the set is deemed suitable, if only from a practical point of view, if it allows the convenient introduction of the cause of the above deviation of into the expansion coefficients . This requirement is satisfied – also from a theoretical point of view – by taking as a set of polynomials which are orthogonal with respect to the ideal p.d.f., taken as their weight function (e.g. Cramér, 1951). That is, the functions so chosen have to obey the relationshipwhere is the range of existence of all the functions involved. It can be readily shown that the coefficients are given by where the brackets in equation (2.1.7.3) denote averaging with respect to the unknown p.d.f. and is the coefficient of the nth power of x in the polynomial . The coefficients are thus directly related to the moments of the non-ideal distribution and the coefficients of the powers of x in the orthogonal polynomials. The latter coefficients can be obtained by the Gram–Schmidt procedure (e.g. Spiegel, 1974), or by direct use of the Szegö determinants (e.g. Cramér, 1951), for any weight function that has finite moments. However, the feasibility of the present approach depends on our ability to obtain the moments without the knowledge of the non-ideal p.d.f., .

#### 2.1.7.3. Application to centric and acentric distributions

| top | pdf |

We shall summarize here the non-ideal centric and acentric distributions of the magnitude of the normalized structure factor E (e.g. Shmueli & Wilson, 1981; Shmueli, 1982). We assume that (i) all the atoms are located in general positions and have rationally independent coordinates, (ii) all the scatterers are dispersionless, and (iii) there is no noncrystallographic symmetry. Arbitrary atomic composition and space-group symmetry are admitted. The appropriate weight functions and the corresponding orthogonal polynomials are where and are Hermite and Laguerre polynomials, respectively, as defined, for example, by Abramowitz & Stegun (1972). Equations (2.1.7.2), (2.1.7.3) and (2.1.7.4) suffice for the general formulation of the above non-ideal p.d.f.'s of . Their full derivation entails (i) the expression of a sufficient number of moments of in terms of absolute moments of the trigonometric structure factor (e.g. Shmueli & Wilson, 1981; Shmueli, 1982) and (ii) calculation of the latter moments for the various symmetries (Wilson, 1978b; Shmueli & Kaldor, 1981, 1983). The notation below is similar to that employed by Shmueli (1982).

These non-ideal p.d.f.'s of , for which the first five expansion terms are available, are given by and for centrosymmetric and noncentrosymmetric space groups, respectively, where and are the ideal centric and acentric p.d.f.'s [see (2.1.7.4)] and the unified form of the coefficients and , for k = 2, 3, 4 and 5, is (Shmueli, 1982), where U = 35 or 18, V = 210 or 100 and W = 3150 or 900 according as or is required, respectively, and the other quantities in equation (2.1.7.7) are given below. The composition-dependent terms in equations (2.1.7.7) are where m is the number of atoms in the asymmetric unit, are their scattering factors, and the symmetry dependence is expressed by the coefficients in equation (2.1.7.7), as follows: where according as the space group is centrosymmetric or noncentrosymmetric, respectively, and in equation (2.1.7.9) is given by where is the kth absolute moment of the trigonometric structure factor In equation (2.1.7.12), g is the number of general equivalent positions listed in IT A (2005) for the space group in question, times the multiplicity of the Bravais lattice, is the sth space-group operator and is an atomic position vector.

The cumulative distribution functions, obtained by integrating equations (2.1.7.5) and (2.1.7.6), are given by and for centrosymmetric and noncentrosymmetric space groups, respectively, where the coefficients are defined in equations (2.1.7.7)–(2.1.7.12). Note that the first term on the right-hand side of equation (2.1.7.13) and the first two terms on the right-hand side of equation (2.1.7.14) are just the cumulative distributions derived from the ideal centric and acentric p.d.f.'s in Section 2.1.5.6.

The moments were compiled for all the space groups by Wilson (1978b) for 1 and 2, and by Shmueli & Kaldor (1981, 1983) for 1, 2, 3 and 4. These results are presented in Table 2.1.7.1. Closed expressions for the normalized moments were obtained by Shmueli (1982) for the triclinic, monoclinic and orthorhombic space groups except and (see Table 2.1.7.2). The composition-dependent terms, , are most conveniently computed as weighted averages over the ranges of which were used in the construction of the Wilson plot for the computation of the values.

 Table 2.1.7.1| top | pdf | Some even absolute moments of the trigonometric structure factor
 The symbols p, q, r and s denote the second, fourth, sixth and eighth absolute moments of the trigonometric structure factor T [equation (2.1.7.12)], respectively, and the columns of the table contain (for some conciseness) and . The numbers in parentheses, appearing beside some space-group entries, refer to hkl subsets which are defined in the note at the end of the table. These subset references are identical with those given by Shmueli & Kaldor (1981, 1983). The symbols q, r and s are also equivalent to , and , respectively, where are the normalized absolute moments given by equation (2.1.7.11).
Space groups(s)pq
Point group: 1
P1 1 1 1 1
Point group:
2 6 10 17½
Point groups: 2, m
All P 2 6 10 17½
All C 4 48 160 560
Point group:
All P 4 36 100 306¼
All C 8 288 1600 9800
Point group: 222
All P 4 28 64 169¾
All C and I 8 224 1024 5432
F222 16 1792 16384 173824
Point group: mm2
All P 4 36 100 306¼
All A, C and I 8 288 1600 9800
Fmm2 16 2304 25600 313600
Fdd2 (1) 16 2304 25600 313600
Fdd2 (2) 16 1280 7168 43264
Point group: mmm
All P 8 216 1000 5359
All C and I 16 1728 16000 171500
Fmmm 32 13824 256000 5488000
Fddd (1) 32 13824 256000 5488000
Fddd (2) 32 7680 71680 757120
Point group: 4
4 36 100 306¼
(3) 4 36 100 306¼
(4) 4 20 28 42¼
8 288 1600 9800
(5) 8 288 1600 9800
(6) 8 160 448 1352
Point group:
4 28 64 169¾
8 224 1024 5432
Point group:
All P 8 216 1000 5359
16 1728 16000 171500
(7) 16 1728 16000 171500
(8) 16 960 4480 23660
Point group: 422
, , , 8 136 424 1682
, (3) 8 136 424 1682
, (4) 8 104 208 470
I422 16 1088 6784 53828
(7) 16 1088 6784 53828
(8) 16 832 3328 15044
Point group: 4mm
All P 8 168 640 2970
I4mm, I4cm 16 1344 10240 95060
(7) 16 1344 10240 95060
(8) 16 832 3328 15188
Point groups:
All P 8 136 424 1682
16 1088 6784 53828
(5) 16 1088 6784 53828
(6) 16 832 3328 15044
Point group: 4/mmm
All P 16 1008 6400 51985
, 32 8064 102400 1663550
(5) 32 8064 102400 1663550
(6) 32 4992 33280 265790
Point group: 3
All P and R 3 15 31 71
Point group:
All P and R 6 90 310 1242½
Point group: 32
All P and R 6 66 166 508½
Point group: 3m
P3m1, P31m, R3m 6 66 178 604½
P3c1, P31c (3), R3c (1) 6 66 178 604½
P3c1, P31c (4), R3c (2) 6 66 154 412½
Point group:
12 396 1780 10578¾
(3), (1) 12 396 1780 10578¾
(4), (2) 12 396 1540 7218¾
Point group: 6
P6 6 90 340 1522½
(9) 6 90 340 1522½
(10) 6 54 91 161½
(11) 6 54 97 193½
(12) 6 90 280 962½
(13) 6 90 340 1522½
(14) 6 54 97 193½
(3) 6 90 340 1522½
(4) 6 90 280 962½
Point group:
6 90 310 1242½
Point group:
12 540 3400 26643¾
(3) 12 540 3400 26643¾
(4) 12 540 2800 16843¾
Point group: 622
P622 12 324 1150 5506¼
(9) 12 324 1150 5506¼
(10) 12 252 577 1537¾
(11) 12 252 583 1601¾
(12) 12 324 1090 4746¼
(13) 12 324 1150 5506¼
(14) 12 252 583 1601¾
(3) 12 324 1150 5506¼
(4) 12 324 1090 4746¼
Point group: 6mm
P6mm 12 396 1930 12818¾
P6cc (3) 12 396 1930 12818¾
P6cc (4) 12 396 1450 6098¾
(3) 12 396 1930 12818¾
(4) 12 396 1630 8338¾
Point groups:
12 396 1780 10578¾
(3) 12 396 1780 10578¾
(4) 12 396 1540 7218¾
Point group: 6/mmm
P6/mmm 24 2376 19300 224328
P6/mcc (3) 24 2376 19300 224328
P6/mcc (4) 24 2376 14500 106728
P6/mcm, P6/mmc (3) 24 2376 19300 224328
, (4) 24 2376 16300 145928
Point group: 23
P23, 12 276 760 2695¼
I23, 24 2208 12160 86248
F23 48 17664 194560 2759936
Point group:
24 1800 9400 67703
48 14400 150400 2166500
96 115200 2406400 69328000
(1) 96 115200 2406400 69328000
(2) 96 96768 1484800 28183680
Point group: 432
24 1272 4648 25216
(15) 24 1272 4648 25216
(16) 24 1176 3568 13916
(17) 24 1080 2776 8664
(18) 24 984 2272 6580
I432 48 10176 74368 806940
(15) 48 10176 74368 806940
(17) 48 8640 44416 277276
F432 96 81408 1189888 25822080
(15) 96 81408 1189888 25822080
(18) 96 62976 581632 6738816
Point group:
24 1272 5128 32896
(1) 24 1272 5128 32896
(2) 24 1272 4168 17536
48 10176 82048 1052700
(15); (20) 48 10176 82048 1052700
(15); (21) 48 10176 66688 561180
(17) 48 8640 44416 277276
96 81408 1312768 33686400
(15) 96 81408 1312768 33686400
(18) 96 81408 1067008 17957760
Point group:
48 8784 72160 972717
(1) 48 8784 72160 972717
(2) 48 8784 56800 488877
96 70272 1154560 31126970
(15); (20) 96 70272 1154560 31126970
(15); (21) 96 51840 432640 4497850
(17) 96 70272 908800 15644090
192 562176 18472960 996063040
(1) 192 562176 18472960 996063040
(2) 192 562176 14540800 500610880
(1) 192 562176 18472960 996063040
(2) 192 414720 7782400 205432640
(1) 192 562176 18472960 996063040
(2) 192 414720 6799360 136619840

Note. hkl subsets: (1) ; (2) ; (3) ; (4) ; (5) ; (6) ; (7) ; (8) ; (9) ; (10) ; (11) ; (12) ; (13) ; (14) , ; (15) hkl all even; (16) only one index odd; (17) only one index even; (18) hkl all odd; (19) two indices odd; (20) ; (21) .
And the enantiomorphous space group.
 Table 2.1.7.2| top | pdf | Closed expressions for [equation (2.1.7.11)] for space groups of low symmetry
 The normalized moments are expressed in terms of , where and , which takes on the values 1, 2 or 4 according as the Bravais lattice is of type P, one of the types A, B, C or I, or type F, respectively. The expressions for are identical for all the space groups based on a given point group, except Fdd2 and Fddd. The expressions are valid for general reflections and under the restrictions given in the text.
Point group(s)Expression for
1 1
mmm
222

#### 2.1.7.4. Fourier versus Hermite approximations

| top | pdf |

As noted in Section 2.1.8.7 below, the Fourier representation of the probability distribution of is usually much better than the particular orthogonal-function representation discussed in Section 2.1.7.3. Many, perhaps most, non-ideal centric distributions look like slight distortions of the ideal (Gaussian) distribution and have no resemblance to a cosine function. The empirical observation thus seems paradoxical. The probable explanation has been pointed out by Wilson (1986b). A truncated Fourier series is a best approximation, in the least-squares sense, to the function represented. The particular orthogonal-function approach used in equation (2.1.7.5), on the other hand, is not a least-squares approximation to , but is a least-squares approximation to The usual expansions (often known as Gram–Charlier or Edgeworth) thus give great weight to fitting the distribution of the (compararively few) strong reflections, at the expense of a poor fit for the (much more numerous) weak-to-medium ones. Presumably, a similar situation exists for the representation of acentric distributions, but this has not been investigated in detail. Since the centric distributions often look nearly Gaussian, one is led to ask if there is an expansion in orthogonal functions that (i) has the leading term and (ii) is a least-squares (as well as an orthogonal-function)2 fit to . One does exist, based on the orthogonal functions where is the Gaussian distribution (Myller-Lebedeff, 1907). Unfortunately, no reasonably simple relationship between the coefficients and readily evaluated properties of has been found, and the Myller-Lebedeff expansion has not, as yet, been applied in crystallography. Although Stuart & Ord (1994, p. 112) dismiss it in a three-line footnote, it does have important applications in astronomy (van der Marel & Franx, 1993; Gerhard, 1993).

### References

International Tables for Crystallography (2005). Vol. A, Space-Group Symmetry, edited by Th. Hahn. Heidelberg: Springer.
Abramowitz, M. & Stegun, I. A. (1972). Handbook of Mathematical Functions. New York: Dover.
Cramér, H. (1951). Mathematical Methods of Statistics. Princeton University Press.
Faggiani, R., Lippert, B. & Lock, C. J. L. (1980). Heavy transition metal complexes of biologically important molecules. 4. Crystal and molecular structure of pentahydroxonium chloro(uracilato-N(1))(ethylene­diamine)platinum(II)chloride (H5O2)[PtCl­(NH2­CH2­CH2­NH2)­(C4­H5­N2O2)]Cl, and chloro(thyminato-N(1))­(ethyl­ene­di­amine)­platinum(II), PtCl(NH2CH2CH2NH2)­(C5H5­N2O2). Inorg. Chem. 19, 295–300.
Gerhard, O. E. (1993). Line-of-sight velocity profiles in spherical galaxies: breaking the degeneracy between anisotropy and mass. Mon. Not. R. Astron. Soc. 265, 213–230.
Hauptman, H. & Karle, J. (1953). Solution of the Phase Problem. I. The Centrosymmetric Crystal. Am. Crystallogr. Assoc. Monograph No. 3. Dayton, Ohio: Polycrystal Book Service.
Howells, E. R., Phillips, D. C. & Rogers, D. (1950). The probability distribution of X-ray intensities. II. Experimental investigation and the X-ray detection of centers of symmetry. Acta Cryst. 3, 210–214.
Marel, R. P. van der & Franx, M. (1993). A new method for the identification of non-Gaussian line profiles in elliptical galaxies. Astrophys. J. 407, 525–539.
Myller-Lebedeff, W. (1907). Die Theorie der Integralgleichungen in Anwendung auf einige Reihenentwicklungen. Math. Ann. 64, 388–416.
Shmueli, U. (1979). Symmetry- and composition-dependent cumulative distributions of the normalized structure amplitude for use in intensity statistics. Acta Cryst. A35, 282–286.
Shmueli, U. (1982). A study of generalized intensity statistics: extension of the theory and practical examples. Acta Cryst. A38, 362–371.
Shmueli, U. & Kaldor, U. (1981). Calculation of even moments of the trigonometric structure factor. Methods and results. Acta Cryst. A37, 76–80.
Shmueli, U. & Kaldor, U. (1983). Moments of the trigonometric structure factor. Acta Cryst. A39, 615–621.
Shmueli, U. & Wilson, A. J. C. (1981). Effects of space-group symmetry and atomic heterogeneity on intensity statistics. Acta Cryst. A37, 342–353.
Spiegel, M. R. (1974). Theory and Problems of Fourier Analysis. Schaum's Outline Series. New York: McGraw-Hill.
Srinivasan, R. & Parthasarathy, S. (1976). Some Statistical Applications in X-ray Crystallography. Oxford: Pergamon Press.
Stuart, A. & Ord, K. (1994). Kendall's Advanced Theory of Statistics, Vol. 1, Distribution Theory, 6th ed. London: Edward Arnold.
Wilson, A. J. C. (1978b). Variance of X-ray intensities: effect of dispersion and higher symmetries. Acta Cryst. A34, 986–994.
Wilson, A. J. C. (1986b). Fourier versus Hermite representations of probability distributions. Acta Cryst. A42, 81–83.