Use of X-Ray Powder Diffraction Rietveld Pattern-Fitting for Characterising

Preferred Orientation in Gibbsites

By Li Deyu* and Brian H. O'Connor**
Department of Applied Physics, Curtin University of Technology, GPO U1987, Perth, Western Australia. 6001
Gerald I.D. Roach and John B. Cornell
Research and Development Department, Alcoa of Australia Limited, Kwinana, Western Australia. 6167

Abstract The March Model of Preferred Orientation

A study has been conducted with gibbsite specimens, on the The model biases the Bragg peak "k" component intensities
use of Rietveld X-ray powder diffraction (XRPD) pattern- of a calculated powder pattern for a cylindrically-symmetric
fitting for quantitating preferred orientation in powders. sample comprising either disk- or rod-shaped crystallites ac-
This study has shown that an earlier formula gives results cording to the function,
which correlate closely with an empirical measure of mor-
Pk=(r2cosV r sin a k ) *
1 . 9 > '
phology proposed recently for gibbsite powders, viz., the ra- (2)
tio of the XRPD intensities for the (002) line and the (110, where r is a fractional measure of crystallite random orienta-
200) doublet lines. A method is proposed on the basis of this tion in the material (assuming r is independent of particle
correlation for the correction for preferred orientation of size in a nonuniform size distribution), and ock is the angle
line intensities in gibbsite powder patterns. The correction between the preferred orientation direction and the recipro-
method appears to have excellent potential for XRPD quan- cal lattice vector direction for Bragg peak "k". Common sam-
tification of gibbsite levels in mixtures, and could have gen- ple preparation/mounting methods usually produce the
eral application for coping with preferred orientation effects requisite cylindrical symmetry, and the condition will defi-
in the quantitation of other phases. nitely occur if a sample spinner is used to rotate the sample
about its cylindrical axis. In this study promising results were
Introduction obtained without use of sample spinning.
Gibbsite, A1(OH)3, is the main Al-bearing constituent of The percentage of crystallites displaying preferred or-
bauxite ore. The gibbsite is extracted by the Bayer process ientation according to the March model is taken to be
which involves dissolution in a caustic soda solution and p=(l-r)xl00% (3)
then re-precipitation as gibbsite. The precipitate can have a
variety of morphological forms for which there are marked Function Pk has the useful property that its integral over
differences in the relative intensities of their respective X-ray space is unity. Therefore Rietveld scale factors obtained with
powder diffraction (XRPD) patterns. These differences have the March model should be more reliable than those ob-
been postulated by Roach and Cornell (1988) to be caused tained assuming random orientation. Accordingly scale fac-
principally by preferred orientation. The differences were tors derived with the March model should be potentially use-
used by Roach and Cornell as a means of characterising ful for assessing the degree of crystallinity (or amorphous
gibbsite morphology according to an empirical index, desig- content) in single-phase specimens and in conducting quan-
nated the morphology ratio (MR), which is determined from titative phase analysis of mixtures containing gibbsite.
the integrated intensities of the (002) line and the (110,200) Parameter r may not exactly represent the fraction of
doublet: randomly oriented crystallites in a powder specimen if parti-
cle sizes are nonuniform as typically applies for mineralog-
MR=I(002)/I(110,200) (1) ical samples. Therefore, in general, r is an index of the ex-
tent of preferred orientation rather than an exact
The ratio was considered to be useful both in identifying the representation of the fraction of randomly-oriented
source of the gibbsite in bauxite samples and in process stud- crystallites.
ies. The present study has examined the reported intensity
differences using the March (1932) model of preferred or-
ientation as proposed by Dollase (1986) for use in Rietveld Experimental
pattern-fitting. The March model is described in the follow- Specimens
ing section. The gibbsite specimens were supplied by the Alcoa of Aus-
tralia Ltd, Research and Development Laboratory, Kwinana,
The specific objectives of the study were,
Western Australia. The materials were selected to cover a va-
(i) to clarify the cause of intensity anomalies in the
riety of morphology types. The specimens were:
XRPD patterns for different gibbsites;
(ii) to determine the degree of preferred orientation in • GIB-A, a coarse grained gibbsite scale with a well de-
gibbsites according to the March model and to estab- fined basal growth plane, taken from the wall of a thick-
lish the relationship between the March values and ener tank;
the MR indices with a view to correcting gibbsite in- • GIB-B, a hydrate (gibbsite) product with particle size
tensities according to the MR value for a given specified in the range 75-90 fim;
pattern. • GIB-C and GIB-F, hydrate products with particle sizes in
the range 75-106 /im;
• GIB-G, a finely agglomerated gibbsite precipitated from
Bayer liquor which had been left standing for approxi-
mately one year;
* Permanent Address: Shanghai Institute of Ceramics, Chinese Academy of Sciences, • A pair of specimens, P-l and P-2 representing two size
Shanghai, China.
** To whom all correspondence should be addressed fractions of a composite plant sample formed under
one set of conditions; the size fractions were <75 /im
and 75-106 flm, respectively.

79 Powder Diffraction, Vol. 5, No. 2, June 1990

Figure 1. SEM micrographs for the GIB-A coarse-grained scale specimens, Figure 2. SEM micrographs for the GIB-B hydrate product, (a) GIB-B as re-
(a) GIB-A5, aftermicronizing for 5 minutes, (b) GIB-AIO after micronizing ceived, (b) GIB-B also as received, showing the morphology of an individual
for 10 minutes, (c) GIB-A15 after micronizing for 15 minutes. grain, (c) GIB-B5 after micronizing for 5 minutes.

Powder Diffraction, Vol. 5, No. 2, June 1990 80

 GIB-A was so coarse that the material had to be milled Table 1.
in order to produce powder samples suitable for XRPD XRPD Data Acquisition Conditions
analysis. Three samples of GIB-A (GIB-A5, GIB-A10 and
GIB-A15) were prepared by milling the material for 5, 10 Radiation: Cu-anode tube (40 kV, 30 mA) unfikered
and 15 minutes, respectively, in order to generate addi- Optics: Bragg/Brentano geometry
tional data sets and to gain preliminary information on the 1° divergence for incident beam, 0.05° receiving slit
influence of milling on preferred orientation. Additional Diffracted-beam graphite monochromotor set for CuKa
samples of GIB-B, P-l, and P-2 (GIB-B5, P-l(5) and P-2(5), Detection: Nal scintillator with PHA
Goniometer range, 20 = 15 - 90°
respectively) were produced by wet milling of 3 g of mate- Step size, 0.04°
rial in 10 ml of acetone for 5 minutes. Sample milling was Counting time, 1 second/step.
performed with a McCrone micronizing device. GIB-C, GIB- Polarization correction factor = 0.80
F and GIB-G were examined as-received.
Scanning electron microscopy (SEM) was used to char-
acterise the specimens — see Figures 1 and 2 for examples. The nonstructural parameter types and the general
GIB-A, as shown in Figure l(a)-(c), after milling for 5, 10 Rietveld strategy were essentially those described by
and 15 minutes, comprised highly oriented small plates O'Connor and Raven (1988) with the addition of preferred
(predominantly <10 fim) which would be expected to pro- orientation parameter r for refinements with the March
duce diffraction patterns displaying substantial preferred or- model. The profile function employed was the pseudo-Voigt
ientation. GIB-B, see Figures 2 (a) and (b), comprised spher- type. Each pattern was Rietveld-analysed in two steps, first as-
ical agglomerates of small (ca.10 /fin) platy crystallites for suming random orientation (designated the non-March
which preferred orientation effects should be minimal. The model) and subsequently with the March model. The direc-
influence of grinding is shown in Figure 2(c) for GIB-B5 tion of preferred orientation was taken to be <001>.
which indicates the way in which, for this sample, milling
breaks up the spherical agglomerates and results in the for- Results and Discussion
mation of grains which favour preferred orientation. Effect of March Model on Quality of Pattern-Fitting
The results obtained for the R-factors, scale factors and the
X-ray Powder Diffraction Data preferred orientation factors are given in Table 2.
Samples were manally pressed into standard Siemens D500 The effect of neglecting preferred orientation is evi-
sample holders. The powdered material was front-loaded by dent from the relatively high R-factors obtained in the non-
lightly pressing with a glass slide. Step-scanned patterns March refinements for the highly oriented samples. The Rp,
were measured with a Siemens D500 Bragg/Brentano Rwp and RB indices for the samples showing substantial pre-
XRPD instrument fitted with a graphite diffracted-beam ferred orientation [GIB-A5, GIB-A10, GIB-A15, GIB-B5, P-l
curved monochromator. The instrument settings are given (5) and P-2(5)] range from 18.0-22.5%, 25.7-32.4% and
in Table 1. 10.5-14.4%, respectively, for the non-March refinements.
Whereas the RP values should approximate the correspond-
Rietveld Pattern-Fitting ing 'expected' R-factors (Rexp), the range of R values for
Calculations were performed with a VAX1185 mainframe the six patterns is 7.4-10.0% which points to the substantial
using the LHPM4 version of the Rietveld software described shortcoming in the random orientation model for these
by Hill and Howard (1986). samples. The data sets for these same samples produced dra-
The structural model and initial structural parameters matic improvements in R-factors when refined with the
were taken from Megaw (1973) as follows: space group March model. Profile factors Rp and R™ were reduced to
P21/n(14); Z=8 formula units A1(OH)3; cell parameters: values in the range 12.3-13.9% and 17.6-20.3%, the former
being considerably closer to the range of R values. The RB
al positions ± (xyz; 1/2-x, i/2+y, i/2-z): values for these oriented samples fell within the range 3.5-
5.2% which is close to values obtained for the subset of sam-
X y z ples displaying much less preferred orientation.
AI(1) 0.177 0.520 -0.005 The samples with relatively small degrees of preferred
Al(2) 0.333 0.020 -0.005 orientation (GIB-B, GIB-C, GIB-F, GIB-G, P-l and P-2) gave
O(l) 0.182 0.202 -0.110 indices Rp, R^ and RB of 11.0-12.4%, 15.7-17.3% and 4.4-
O(2) 0.682 0.687 -0.110 5.9%, respectively, after refinements assuming random or-
O(3) 0.515 0.132 -0.110 ientation, and produced only marginal changes in indices
O(4) -0.015 0.632 -0.110 with the March model. The extent of agreement between
C(5) 0.298 0.702 -0.100 the measured and calculated patterns for the random orien-
O(6) 0.838 0.172 -0.100 tation and March model refinements may also be judged
from the example in Figure 3 which shows the dramatic im-
provement in pattern matching when the March model is
used for highly oriented samples.

81 Powder Diffraction, Vol. 5, No. 2, June 1990

Table 2.
Rietveld Refinement Results for R-factors, Scale Factors and Preferred Orientation Factors

Refinements with zero preferred orien- Refinements with preferred orientation using
tation (non-March model) the March model
Expected
Sample R-factor* Rp** R»p**: P%
#GIB-A5 8.2 19.3 27.0 11.5 1.67(3) 13.1 19.0 4.3 1.07(2) 29.5 (4)
#GIB-A10 9.3 18.8 27.9 11.7 1.64(3) 12.5 18.7 3.5 0.97(1) 34.7 (4)
#GIB-A15 7.9 18.0 26.5 10.5 1.74(4) 12.3 18.1 3.6 1.01(2) 35.1 (4)
GIB-B 9.3 12.2 17.2 5.9 1.02(1) 12.0 17.1 5.4 1.04(1) -3.2 (4)
#GIB-B5 7.4 22.5 32.4 14.4 1.89(4) 13.9 20.3 5.2 1.06(2) 39.1 (3)
GIB-C 10.3 11.0 15.7 4.4 0.82(1) 11.1 16.3 4.3 0.81(1) 0.4 (4)
GIB-F 10.4 11.9 17.3 4.5 0.80(1) 11.7 17.4 4.0 0.81(1) -3.1 (4)
GIB-G 9.2 12.4 16.8 4.6 1.00(1) 12.3 16.8 4.7 1.01(1) -1.4 (4)

P-l 9.3 11.9 17.0 5.2 1.03(1) 12.0 17.7 4.9 1.02(1) 0.74(4)
#P-1(5) 10.0 18.9 25.7 12.2 1.44(2) 13.4 18.4 4.7 1.03(1) 24.2 (4)
P-2 9.2 12.4 17.1 5.5 1.04(1) 12.4 17.7 4.6 1.02(1) 2.94(4)
#P-2(5) 9.9 19.4 27.0 13.9 1.44(2) 12.4 17.6 4.9 1.04(1) 25.6 (3)

# Samples showing pronounced preferred orientation.

J (y, o -y i c ) 2
*Rexp=[(N-P)/Iwiyir , **Rp = ***Rw P =
2Wiyio2
where N = number of observations (profile points)
P = number of least squares parameters.
2000'

1000-

2000

38,400 CALCULATED (MARCH)

i/i
w 1000 -
c

2000.
37,800 CALCULATED (NON-MARCH)

1000-

A
15 25 35 45 55 65 75 85
Figure 3. Measured X-ray powder diffraction
powder patterns for GIB-B5, and corresponding
Rietveld-calculated patterns after refinements
with the March and non-March models. The 'dif-
£8 MEASURED-CALCULATED (MARCH)
(f) C
ference plots' at the foot of the diagram show the C CD
intensity difference between the measured and cal- Si W
culated pattern in each case. b MEASURED-CALCULATED (NON-MARCH)

Ponder Diffraction, Vol. 5, No. 2, June 1990 82

Influence of Preferred Orientation Model on Structural Parameters Correlation Between Preferred Orientation and Morphology Factors
Table 3 summarises the deviations from initial values of The morphology and preferred orientation factors, MR and
Megaw (1973) atomic coordinates obtained after refine- p, are listed in Table 4 and the relationship between the two
ments with the various data sets assuming random orienta- data sets is shown in Figure 4 which demonstrates that the
tion and using the March model. It can be seen that pre- relationship between MR and p is nonlinear over the ob-
ferred orientation has a more pronounced influence on the served range of MR values. The occurrence of negative val-
atomic coordinates set if the refinement model does not al- ues for p indicates that the MR values may be biased by addi-
low for the effect. The average values of deviations of atomic tional factors such as extinction. The origin (s) of these
coordinates y and z from initial values for samples of high negative values should have further study.
preferred orientation are three and eight times, respectively,
greater than those for samples of low orientation. The maxi-
mum shift in fractional coordinates for the non-March re-
finements of highly oriented specimens was that observed Table 4.
for the y-coordinate of atom O 4 , 0.118, which corresponds Measures of Preferred Orientation (p) from Rietveld Refinements
to a shift of 0.60 A along the b-direction. Clearly unreason- and Corresponding Morphological Ratios (MR).

Sample Degree of preferred Morphological

Orientation % (p) Ratic' ( M R )
Table 3. Measured Calculated
Dependence of Structural Parameters on Preferred Orientation GIB-A5 29.5(4) 10.2 8.9
Model GIB-A10 34.7(4) 14.3 12.5
GIB-A15 35.1(4) 17.1 12.9
The parameter shifts represent the differences between the final GIB-B -3.2(4) 1.8 1.6
positional parameters for the March and non-March models for the GIB-B5 39.1(3) 17.3 17.0
data sets examined and the parameters quoted by Megaw. GIB-C 0.4(4) 2.1 1.9
GIB-F -3.1(4) 1.7 1.6
Low-Orientation GIB-G -1.4(4) 1.8 1.7
Parameter High-Orientation
Shifts Specimens Specimens
(Fractional Non- P-1 0.7(4) 2.1 1.9
Non-
Coordinates) March March P-1(5) 24.2(4) 7.0 6.4
March March
P-2 2.9(4) 2.5 2.1
Mean Magnitude P-2 (5) 25.6(4) 7.4 7.0
0.020
0.033
0.013
0.015
0.004
0.013
0.017
0.010
0.011
0.015
0.005
40 - • Measured MR
Calculated MR
GIB-B5
^ *
0.023 •

Maximum Value
yS GIB -A10 GIB-A15
0.045(0^ 0.047(O6) 0.036(0!) 0.023(O2)
0.118(O4) 0.050(O2) 0.068 (O4) 0.036(O2>O6)
0.055(0,) 0.021 (O4) 0.023(O4) 0.016(O5) 30 -

/O • GIB-A5

/
P-2(5)
able interatomic distances result if the random orientation
model is assumed for highly oriented samples. For the sam-
ples showing minimal random orientation the deviations in
20 -

fractional coordinates are more reasonable and appear to

be acceptable in terms of obtaining reasonable interatomic
distances.
It is therefore evident that crystal structure refinements
6 / /

are likely to produce serious bias in at least some positions if I

preferred orientation effects are substantial. Use of the Q_ 10 -

March model may overcome the problem, but further inves-

tigation will be necessary before the March model can be
used with confidence in structural work. A possibility of par-
/
ticular concern is that bias in atomic positions caused by pre- • P-2
ferred orientation may thereby influence nonstructural Riet-
o ^GIB-C i I I
veld parameters which correlate with the positional 10 15

-t
5
parameters. GIB-G
GIB-B,GIB-F Morphological Ratio MR
-4

Figure 4. Plot of preferred orientation parameter p versus morphological ra-

tio MRfor the specimen suite. The continuous line represents the calculated
jNVfR relation as described in the text.

83 Powder Diffraction, Vol. 5, No. 2. June 1990

 The shape of the MR-p graph for measured MR data is Use of Morphological Ratio and Preferred Orientation Parameter to
consistent with the theoretical MR-p relation calculated from Correct Intensities
Equation (1) using the March definition of r in Equation (2). The promising nature of the MR-p correlations obtained in
For this exercise the integrated intensities 1(002), 1(110) and the present study indicates that the following steps may be
1(200) were set at 100, 36 and 18, respectively, according to used to correct integrated intensities in gibbsite samples for
the calculated intensity data published by the International preferred orientation.
Centre for Diffraction Data as PDF pattern 29-41 (data
source: D. Smith et al, 1977). The theoretical form of the (i) measurement of MR for a pattern of a given sample,
MR-p; relation is shown in Figure 4 for comparison with the (ii) conversion of MR to p and r using Figure 4,
values of MR observed from the Rietveld refinements. While (iii) use of Equation (2) to determine correction factor Pk
there is some significant systematic difference between the for any peak of interest.
theoretical and observed plots, the agreement adds consid- The effectiveness of the correction is shown in Table 5
erable support to the proposition that the substantial inten- which compares the integrated intensities of the more in-
sity bias in some of the measured patterns is predominantly tense lines for three patterns covering the range of MR val-
due to preferred orientation. ues observed -GIB-B, P-l (5) and GIB-B5. The table presents
data in the angular range 20 = 0 to 55° (dmin=1.67A). Two in-
tensity estimates are given for the various lines of each pat-
tern - as measured and corrected for preferred orientation
using the strategy outlined above.

Table 5.
Effect of preferred Orientation Correction on Integrated Intensities

The first line of intensity data for each hk/ gives the measured intensities. Intensity values in the sec-
ond line of data (enclosed in parentheses) have been corrected for preferred orientation. Sample
GIB-B has essentially no preferred orientation (MR = 1.8), P-l (5) has intermediate preferred orienta-
tion (MR = 24.2) and GIB-B5 has the highest preferred orientation (MR = 39.1) for the specimen set.

Intensities Intensities
hkZ GIB-B P-l(5) GIB-B5 hk/ GIB-B P-l(5) GIB-Bf
002 41408 91767 191115 -303 97 299 539
(41391) (46216) (308) (640)
1 10 16007 9954 7963 312 4518 2439 947
(14801) (16163) (2874) (1394)
200 7756 3758 3309 -1 14 351 775 607
(5566) (6681) (516) (343)
202 2308 2419 2151 -3 13 6761 5662 4660
(2491) (2556) (6231) (6143)
1 12 4194 3987 3793 -4 0 2 119 209 450
(3979) (4294) (281) (795)
1 12 3426 3352 3347 023 2426 2547 2019
(3187) (3510) (2667) (2461)
202 1471 1251 1374 410 36 84 420
(1170) (1404) (125) (849)
31 1 3410 2200 2076 222 806 714 703
(3183) (4057) (897) (1125)
021 2858 3026 1684 313 2974 3007 2146
(4270) (3178) (3034) (2470)
004 1812 1433 7680 -3 14 5406 5512 4860
(646) (1857) (5177) (4998)
31 1 8062 6721 4436 412 104 472 373
(9218) (8044) (596) (601)
312 1960 1733 1366 024 2539 2983 2537
(2234) (2268) (2674) (2420)
022 1617 1253 959 314 5108 5091 3867
(1546) (1493) (4427) (3508)

Powder Diffraction, Vol. 5, No. 2, June 1990 84

 The results indicate that the strategy should be of con- References
siderable value in correcting integrated intensities for pre- Bish, D.L. & Howard, S.A. (1988). Quantitative phase analysis using
ferred orientation in conventional XRPD quantitative analy- the Rietveld method./. Appl. Crystallogr. 21, 86-91.
Chung, F.H. (1974). Quantitative interpretation of X-ray diffrac-
sis such as methods described by Klug and Alexander (1974) tion patterns of mixtures I. Matrix-flushing method of quantita-
and Chung (1974). An attractive aspect of the proposed tive multicomponent analysis./. Crystallogr. 7,519-525.
strategy is the avoidance of Rietveld calculations by measur- Dollase, W.A. (1986). Correction of intensities for preferred orien-
ing MR. Currently the authors are developing the strategy tation in powder diffractometry: application of the March
for routine quantitaiton of gibbsite in mixtures. model./ Appl. Crystallogr. 19, 267-272.
Hill, R.J. & Howard, CJ. (1986). A computer program for Rietveld
analysis of fixed wavelength X-ray and neutron powder diffrac-
Potential Use of Rietveld Scale Factors for Quantitative Phase tion patterns: version LHPM1. Australian Atomic Energy Com-
Analysis mission Research Establishment.
Hill and Howard (1987), Bish and Howard (1988) and Hill, RJ. & Howard, CJ. (1987). Quantitative phase analysis from
neutron powder diffraction data using the Rietveld method. /.
O'Connor and Raven (1988) have described similar meth- Appl. Crystallogr. 20, 4&JA1A.
ods for quantitative XRPD analysis of mixtures using Riet- Klug, H.P. & Alexander, L.E. (1974). X-ray Diffraction Procedures.
veld scale factors - symbol Sk in Equation (3) of O'Connor 2nd ed., 531-562. New York: J. Wiley & Sons.
& Raven. We note from the S-values in Table 2 that there is March, A. (1932). Mathematische theorie der regelung nach der
a wide discrepancy between results for highly oriented korngestalt bei affiner deformation. Z. Kristallogr., Kristallgeom.,
KristaUphys., Kristallchem. 81, 285-297.
specimens and those with minimal orientation (range = Megaw, H.D. (1973). Crystal Structure: A Working Approach, 354-356.
0.80-1.89). Use of the March model substantially alters the Philadelphia: W. B. Saunders Co.
former values resulting in values falling within the range O'Connor, B.H. & Raven, M.D. (1988). Application of the Rietveld
0.81-1.07. refinement procedure in assaying powdered mixtures. Pow. Diff.
3, 2-6.
Roach, G.I.D. & Cornell, J.B. (1988). Morphology analysis by X-ray
Conclusions diffraction. Proceedings of AXAA88 Conference, Perth 1988,
(i) Rietveld pattern-fitting with the March model has 319-328.
been shown to be a promising method for character-
ising preferred orientation in gibbsite.
(ii) The R-factors for samples with a high degree of pre-
ferred orientation are reduced substantially to values
close to those for samples of random orientation af-
ter refinements using the March model. This result
supports the proposition that systematic differences
observed in the XRPD patterns of various gibbsite are
principally caused by preferred orientation.
(iii) A graphical relationship has been established be-
tween the morphology ratio MR and the preferred or-
ientation factor p. This has considerable potential as
a means of correcting for preferred orientation in
conventional quantitative XRPD.
(iv) The Rietveld scale factors vary substantially from sam-
ple to sample for non-March refinements. The results
are therefore promising in terms of the potential use
of Rietveld/March scale factors for XRPD assay.

Received August 6, 1989

RevisedJanuary 2, 1990 AcceptedJanuary 6, 1990

85 Powder Diffraction, Vol. 5, No. 2, June 1990