You are on page 1of 10

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/200045463

Accuracy in Rietveld Quantitative Phase Analysis


of Portland Cements

Article in Journal of Applied Crystallography · March 2003


DOI: 10.1107/S002188980301375X

CITATIONS READS

77 273

2 authors:

Angeles G De la Torre Miguel A. G. Aranda


University of Malaga ALBA synchrotron Light Source - Barcelona, Sp…
72 PUBLICATIONS 1,791 CITATIONS 283 PUBLICATIONS 6,923 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

MINECO GRANT 2015-2017 Optimisation and advanced characterisation of ye´elemite based


econcements View project

MINECO GRANT 2015-2017: Optimisation and advanced characterisation of ye'elimite based ecocements
View project

All content following this page was uploaded by Miguel A. G. Aranda on 01 June 2014.

The user has requested enhancement of the downloaded file.


electronic reprint
Journal of
Applied
Crystallography
ISSN 0021-8898

Accuracy in Rietveld quantitative phase analysis of Portland cements


A. G. De la Torre and M. A. G. Aranda

Copyright © International Union of Crystallography

Author(s) of this paper may load this reprint on their own web site provided that this cover page is retained. Republication of this article or its
storage in electronic databases or the like is not permitted without prior permission in writing from the IUCr.

J. Appl. Cryst. (2003). 36, 1169–1176 De la Torre and Aranda  Portland cements
research papers
Journal of
Applied
Accuracy in Rietveld quantitative phase analysis of
Crystallography Portland cements
ISSN 0021-8898

A. G. De la Torre and M. A. G. Aranda*


Received 19 February 2003
Accepted 16 June 2003
Departamento de QuõÂmica InorgaÂnica, CristalografõÂa y MineralogõÂa, Universidad de MaÂlaga, 29071
MaÂlaga, Spain. Correspondence e-mail: g_aranda@uma.es

The polymorphs that constitute most Portland cements have been synthesized:
tricalcium silicate, dicalcium silicate, aluminate, ferrite, gypsum, bassanite and
calcite. They have been used to prepare arti®cial mixtures, i.e. white Portland
clinker, grey Portland clinker and two types of grey Portland cements.
Quantitative mineralogical analyses of these mixtures have been obtained by
laboratory X-ray powder diffraction ( = 1.54 A Ê ) and the Rietveld method. To
assess the accuracy of these analyses, high-energy synchrotron X-ray powder
data ( = 0.40 A Ê ) for the same mixtures have also been studied. Furthermore,
synchrotron X-ray powder data were collected for binary mixtures of the
polymorphs and a corundum standard. This was done to determine the presence
of impurity crystalline phases in the synthesized samples and to check the
presence of non-negligible amorphous phase contents. The errors in the
synchrotron X-ray analyses are quite low (usually smaller than 1 wt%). The
relative errors in the laboratory X-ray analyses are of the order of 2% for the
main phases and increase to approximately 5±10% for the low-content
components. These errors are acceptable in the factory environment and the
# 2003 International Union of Crystallography routine application of this methodology in the cement industry is being
Printed in Great Britain ± all rights reserved implemented.

1. Introduction In the cement industry, QPA using X-ray powder diffrac-


X-ray powder diffractometry has been used as an analytical tometry and the Rietveld method is the most auspicious
technique for a long time (Klug & Alexander, 1974; Zevin & alternative to control on-line the production process (Schmidt
Kimmel, 1995). In recent years, there has been great interest in & Kern, 2001). Much effort to develop prototype systems has
the application of the Rietveld method (Rietveld, 1969) to been exerted recently (Manias et al., 2000; Scarlett et al., 2001),
obtain quantitative phase analysis (QPA) of simple mixtures allowing the mineralogical compositions of the clinker and
(Hill & Howard, 1987; Bish & Howard, 1988; Madsen et al., cement streams to be followed on-line in a factory. To date,
2001) using X-ray powder diffraction. This methodology is X-ray ¯uorescence (XRF) provides the elemental composition
standardless but the crystal structures for every crystalline used to infer the mineralogical composition of these materials
phase in the sample must be known as the process consists of by means of Bogue calculations (Bogue, 1929). Problems with
the comparison between the measured and calculated powder this methodology are well known (Taylor, 1997) and mainly
diffraction patterns. QPA by the Rietveld method gives phase arise from the lack of thermodynamical equilibrium in the kiln
fractions normalized to 100% crystalline phases, while the and because it does not take into account the elemental
amorphous/non-diffracting/non-de®ned crystalline content is substitutions in a given phase, which is usually important in
not usually accounted for. these phases. Nevertheless, the Bogue method is still in
Inherent bene®ts of QPA using the Rietveld method have worldwide use to estimate the main phase fractions as part of
encouraged its application to complex mixtures (Scarlett et al., quality control.
2002), and to industrial materials such as Portland cements To implement the Rietveld methodology for these mixtures,
(De la Torre, Cabeza et al., 2001; Taylor et al., 2000; Neubauer it is essential to have the crystallographic description for every
& Sieber, 1996), aluminous cements (Guirado et al., 2000), phase that may be present. White Portland clinkers have three
archaeological ceramics (Kockelmann & Kirfel, 2001), phases and grey Portland clinkers have an additional fourth
mineralized Portland cements (Pajares et al., 2002) and phase. Cement nomenclature will be used hereinafter, i.e. C =
calcium sulphoaluminate cements (Schmidt & PoÈllmann, CaO, S = SiO2, A = Al2O3, F = Fe2O3, S = SO3, C = CO2 and H =
2000). The Rietveld methodology has also been used to H2O. The main phase, alite (C3S or Ca3SiO5), has seven
determine indirectly the amorphous content in a given crys- polymorphs (Taylor & Aldridge, 1993), but the complex
talline sample by adding a suitable standard (De la Torre, monoclinic superstructure is usually present in Portland
Bruque & Aranda, 2001). cements because of the stabilization effect of the Al3+ and

J. Appl. Cryst. (2003). 36, 1169±1176 De la Torre and Aranda  Portland cements 1169
electronic reprint
research papers
Mg2+ elemental substitution (De la Torre et al., 2002). The sample volume yields a better determination of the true
second phase, belite (C2S or Ca2SiO4), has ®ve polymorphs, crystalline contents of the arti®cial mixtures. This work is a
but in the industrial samples, the phase is usually present step forward in the application of the Rietveld method to
(Mumme et al., 1996). Aluminate (C3A or Ca3Al2O6) may ordinary Portland cements.
crystallize in four polymorphs depending upon the alkaline
content (Na, K); the cubic and orthorhombic phases are
commonly present in cements (Goetz-Neunhoeffer & 2. Experimental section
Neubauer, 1997). In grey Portland clinkers, ferrite (C4AF or
Ca4Al2Fe2O10) is also present and it crystallizes in an orthor- 2.1. Specimen and sample preparation
hombic structure that can be primitive or body-centred 2.1.1. Synthesis of pure phases. The appropriate poly-
depending upon the Fe/Al ratio. The body-centred polymorph morphs of the main constituents of Portland clinkers (mono-
is usually found in clinkers (Guirado et al., 1996). In ®nal grey clinic C3S, monoclinic C2S, cubic C3A and orthorhombic
Portland cements, three additional phases are usually found: C4AF) were synthesized. Polycrystalline CaCO3 was used as
gypsum (CSH2 or CaSO4.2H2O), bassanite (CSH0.5 or received (99.95% from Alfa). Polycrystalline CaSO4.2H2O
CaSO4.0.5H2O) and calcite (CC or CaCO3). Bassanite is not was prepared by carefully grinding a piece of a high-quality
added but it is formed in the cement mills as a result of the gypsum single crystal. Bassanite was synthesized by grinding
partial dehydration of gypsum at the milling temperature. 3 g of gypsum single crystal. The powder was heated at 365 K
The application of Rietveld methodology to laboratory for 24 h and then to 373 K for another 24 h.
powder diffraction data of these systems has some known C3S-alite. Adequate amounts of CaCO3, Mg(OH)2.4Mg-
handicaps that can be partially overcome (De la Torre et al., CO3.5H2O (99% from Aldrich), SiO2 (99.7% from ABCR)
2002) to obtain relatively reliable QPA. Although the preci- and -Al2O3 (99.997% from Alfa) were weighed to obtain 6 g
sion of the laboratory studies can be estimated, the accuracy of of (Ca2.93Mg0.07)O3(SiO2)0.98(Al2O3)0.01. This Mg/Ca and Si/Al
such analyses is not known and it is dif®cult to measure. elemental substitution stabilizes the monoclinic alite structure
However, knowledge of the accuracy of these analyses is very at room temperature. The mixture was ground in an agate ball
important if this methodology is to be transferred to industry. mill at 200 r.p.m. for 30 min and heated at 1273 K in a Pt
To evaluate this, the use of commercial samples is not crucible for 6 h. The resulting solid was ground in an agate ball
adequate since the `true' mineralogical compositions are not mill at 200 r.p.m. for 30 min, pelletized, heated at 1723 K for
known and it is very dif®cult to measure them by other 6 h (10 K minÿ1 heating rate) and cooled by turning off the
techniques to an acceptable low error level. furnace (5 h to 423 K). Finally, the mixture was ground in an
In order to evaluate the accuracy of Rietveld QPA of agate mortar, pelletized, and heated at 1773 K for 6 h twice, as
Portland cements, arti®cial samples have been prepared by previously described.
mixing suitable synthesized phases in the appropriate C2S-belite. Dicalcium silicate has several polymorphs but
proportions. Hence, an expected mineralogical phase fraction, the elemental substitution and the thermal cooling in Portland
our `true mineralogical composition', is available and we can clinkers stabilize the form. -Ca2SiO4 is stable above 953 K
determine the accuracy of the procedure. Laboratory X-ray and it must be (meta)stabilized at room temperature. The
powder diffraction patterns may have several non-random stable polymorph at room temperature is the phase
errors, such as poor sample averaging, preferred orientation or (Nettleship et al., 1992), and the ! transformation on
texture, optical aberrations which change with 2, etc. There- cooling has to be avoided. It is known that this transformation
fore, high-energy parallel synchrotron X-ray powder diffrac- is precluded if the C2S particle size is small and the cooling
tion data have been collected for all samples as these data are rate is high. A mixture of the stoichiometric proportions of
almost free from these errors, providing the best analyses that CaCO3 and SiO2 was prepared to obtain 5 g of -Ca2SiO4. The
can be obtained. Furthermore, this may be slightly compli- mixture was heated at 1273 K for 4 h, cooled and ground in an
cated by the presence of non-diffracting regions in the agate ball mill at 100 r.p.m. for 1 h. The powdered sample was
prepared polymorphs. In order to check this, all `pure' arti®- pelletized and heated in a platinum crucible at 1473 K for 1 h
cial phases were mixed with a crystalline standard, -Al2O3, to (20 K minÿ1 heating rate) and quenched in air. The sample
obtain a set of corrected crystalline phase fractions. Related underwent this treatment three times.
studies where QPA Rietveld analyses using laboratory data C3A-aluminate. A mixture of stoichiometric amounts of
were compared with the results of optical microscopy have CaCO3 and -Al2O3 was prepared to obtain 3 g of cubic
been reported (Neubauer et al., 1997; Neubauer, 1998; FuÈll- Ca3Al2O6. The sample was ground in an agate mortar and
mann et al., 2001; Walenta et al., 2001; Kern, 2001). heated at 1273 K for 6 h. The product was pelletized and
The main aim of this work is to evaluate the accuracy of heated at 1723 K with a heating rate of 10 K minÿ1 for 6 h and
Rietveld QPA for Portland clinkers and cements using cooled by turning off the furnace.
laboratory X-ray powder diffraction data. The use of C4AF-ferrite. 5 g of orthorhombic Ca4Al2Fe2O10 was
synchrotron radiation has several advantages: (i) the high- synthesized by mixing stoichiometric proportions of CaCO3,
resolution synchrotron data allows one to distinguish the -Al2O3 and Fe2O3 (99.95% from Alfa). The mixture was
appropriate polymorph and to know if the structural heated at 1273 K during 4 h and the resulting product was
descriptions for the phases are adequate; (ii) the large tested ground in an agate mortar and pelletized. The pellets were

1170 De la Torre and Aranda  Portland cements J. Appl. Cryst. (2003). 36, 1169±1176
electronic reprint
research papers
Table 1 number of particles, 40, were studied for each phase.
Weight fractions for binary mixtures and synchrotron X-ray powder Gypsum and bassanite were not studied as they are not stable
diffraction results, along with RF and Rwp values.
in vacuum under the electron beam. Linear absorption coef-
Binary Weighed SXRPD RF Rwp Ê and  = 1.54 A
®cients, , for  = 0.40 A Ê X-rays are also given
mixture (Wi) (%) (Ri) (wt%) (%) (%) fc
in Table 2.
C3S 49.85 47.75 (6) 5.41 10.8 0.9194 Four mixtures, imitating commercial Portland samples, were
-Al2O3 50.15 52.25 (7) 2.20 prepared. The simplest mixture is the arti®cial white Portland
C2S 49.55 50.29 (4) 2.57 7.6 1.0300
-Al2O3 50.45 49.71 (5) 2.22 clinker (AWC), which contains three phases. The arti®cial grey
C3A² 50.14 49.68 (4) 4.72 7.5 0.9818 Portland clinker (AGC) contains four phases. Two arti®cial
-Al2O3 49.86 50.32 (5) 2.89 Portland cements (labelled AC1 and AC2) were also prepared,
C3A³ 50.14 49.30 (3) 3.87 6.9 0.9670
-Al2O3 49.86 50.70 (3) 2.05 which contain six and seven crystalline phases, respectively.
C4AF 50.02 47.91 (4) 3.59 7.9 0.9190 The weighed fractions are given in Tables 3±6.
-Al2O3 49.98 52.09 (4) 2.95
Gypsum 50.00 50.26 (4) 2.67 6.0 1.0105
-Al2O3 50.00 49.74 (4) 1.64 2.2. X-ray data collection
Bassanite 49.87 49.24 (6) 3.72 8.0 0.9751
-Al2O3 50.13 50.76 (6) 1.53 Synchrotron X-ray powder diffraction (SXRPD) patterns
CaCO3 51.03 52.58 (4) 3.21 6.5 1.0641 have been collected on the ID31 diffractometer of ESRF,
-Al2O3 48.97 47.42 (4) 1.98 European Synchrotron Radiation Facility (Grenoble, France),
using a short penetrating wavelength  = 0.400269 A Ê
² Measured in a capillary of 2 mm. ³ Measured in a capillary of 1 mm.
(30.97 keV) selected with a double-crystal Si (111) mono-
heated at 1623 K for 12 h (10 K minÿ1 heating rate) and chromator and calibrated with NIST Si (a = 5.43094 A Ê ). The
cooled by turning off the furnace. The synthesis temperature standard Debye±Scherrer con®guration was used. Samples
was well below the melting temperature. were loaded in borosilicate glass capillaries (diameter 2 mm)
The crystalline standard used to prepare the binary and rotated during data collection. The overall measuring time
mixtures was -Al2O3. This compound was synthesized from was 100 min in order to obtain very good statistics over the
-Al2O3. The solid was ground in an agate ball mill at 200 angular range 2.5±30 (in 2) (9.15±0.77 A Ê ). The data from the
r.p.m. for 30 min and heated at 1373 K for 4 h in a Pt crucible. multi-analyser Si(111) stage coupled with the nine scintillation
The oxide was cooled by turning off the furnace and ground in detectors were normalized and summed to 0.002 step size
an agate mortar for 5 min. A second thermal treatment was with local software to produce the ®nal raw data.
carried out at 1473 K for 6 h. The resulting solid was ground in Laboratory X-ray powder diffraction, LXRPD, patterns
an agate mortar and sieved (<0.125 mm) prior to being were recorded on a Siemens D5000 /2 diffractometer (¯at
weighed. The crystallinity of this compound has been studied re¯ection mode) by using Cu K 1,2 radiation (1.542 A Ê ) with a
in previous work (De la Torre, Bruque & Aranda, 2001). secondary curved graphite monochromator and a goniometer
2.1.2. Artificial mixtures. Seven binary mixtures of the radius of 220 mm. The diffractometer optic used to collect the
`pure' cement phases with the -Al2O3 standard were clinker samples comprised a ®xed aperture slit of 2 mm, one
prepared and analysed by synchrotron X-ray powder diffrac- scattered-radiation slit of 2 mm after the sample, followed by a
tion. In order to obtain good particle dispersion, each mixture system of secondary Soller slits and the detector slit of 0.2 mm.
(800 mg) was ground in an agate mortar for 30 min. The The diffractometer optic used to collect the cement samples
weighed fractions of each binary mixture are shown in Table 1. (with gypsum and bassanite) comprised a ®xed aperture slit of
The average particle size, d, for each pure cement phase is 1 mm, one scattered-radiation slit of 1 mm after the sample,
shown in Table 2. The sizes were estimated from a scanning followed by a system of secondary Soller slits and the detector
electron microscope (SEM) characterization using a Jeol SM slit of 0.1 mm. The X-ray tube operating conditions were
840 electron microscope with gold-metallized samples. A large 40 kV and 30 mA for clinkers and 40 kV and 40 mA for

Table 2
Mean particle size, linear absorption coef®cients, unit-cell parameters and reference for the structural description of the studied phases.
C3S C2S C3A C4AF CSH2 CSH0.5 CC A

d (mm) 8 2 15 15 ± ± 10 0.3
 = 0.4 AÊ (cmÿ1) 6.0 5.7 5.1 10.6 2.5 3.6 3.7 1.9
 = 1.54 AÊ (cmÿ1) 314 303 267 495 139 195 197 120
Space group Cm P21/n Pa3 Ibm2 I2/c I2 
R3c 
R3c
Ê)
a (A 33.1043 (2) 5.51124 (3) 15.26783 (2) 5.56501 (2) 5.67798 (4) 12.0153 (2) 4.99117 (1) 4.76011 (2)
b (AÊ) 7.03783 (4) 6.75397 (3) 15.26783 14.51501 (5) 15.2090 (1) 6.9445 (2) ± ±
Ê)
c (A 18.5263 (1) 9.31335 (5) 15.26783 5.34628 (2) 6.52670 (5) 12.6818 (1) 17.06772 (4) 12.99457 (6)
( ) 94.141 (1) 94.574 (1) ± ± 118.478 (1) 89.824 (2) ± ±
V (A Ê 3) 4305.0 (6) 345.565 (4) 3559.03 (1) 431.853 (3) 495.425 (5) 1058.17 (2) 368.222 (1) 254.992 (2)
Reference De la Torre et al. Mumme et al. Mondal & Colville & De la Torre et al. Bezou et al. Maslen et al. This work
(2002) (1995) Jeffery (1975) Geller (1971)² (unpublished) (1995) (1995)

² Occupation factors in the text.

J. Appl. Cryst. (2003). 36, 1169±1176 De la Torre and Aranda  Portland cements 1171
electronic reprint
research papers
cements. All samples were loaded in a methacrylate holder by ening, which has to be taken into account. For very high-
gently sample-front pressing and they were rotated during resolution synchrotron data, the approach based on multi-
data collection at 15 r.p.m. to obtain better powder averaging. dimensional distribution of lattice metrics reported by
The 2 range was 10±70 , in 0.03 steps, counting between 15 Stephens (1999) and included in GSAS is encouraged. Some
and 25 s per step. The data for the four arti®cial mixtures were phases showed preferred orientation in the LXRPD patterns.
collected twice; the samples were reground (5 min) and For simple texture effects, the March±Dollase correction
reloaded in the holder. (Dollase, 1986) was used (gypsum along [010], bassanite along
[100], calcite along [104] and C4AF along [020]). A minor
correction for gypsum and bassanite was also used in the
2.3. X-ray data analysis SXRPD re®nements. The texture effect of C3S, for front-
The powder patterns were re®ned by the Rietveld method loaded samples, in the LXRPD patterns is complex. These
with the GSAS suite of programs (Larson & Von Dreele, 1994) texture effects increase as the C3S particle sizes are larger, and
by using a pseudo-Voigt peak shape function (Thompson et al., are exacerbated for patterns collected from cement pellets.
1987) with the asymmetry correction of Finger et al. (1994) There is no doubt that the spherical-harmonic correction (Von
included. Some phases present anisotropic line shape broad- Dreele, 1997) gives much better ®ts and results for this phase

Table 3
Comparison of the weight fractions for arti®cial white Portland clinker.
Weighed and corrected weighed mass fractions are compared with the QPA from SXRPD and LXRPD analyses by the Rietveld method. Rwp values are also given.
Weighed (%) Weighedc (%) SXRPD (wt%) LXRPD1 (wt%) LXRPD2 (wt%)

C3S 79.98 78.33 79.53 (5) 78.3 (1) 79.0 (1)


C2S 15.01 16.47 15.10 (13) 17.3 (4) 16.2 (4)
C3A 5.01 5.20 5.37 (6) 4.4 (1) 4.8 (1)

Rwp (%) ± ± 12.5 7.5 7.7

Table 4
Comparison of the weight fractions for arti®cial grey Portland clinker as in Table 3.
Weighed (%) Weighedc (%) SXRPD (wt%) LXRPD1 (wt%) LXRPD2 (wt%)

C3S 60.00 58.25 60.15 (8) 60.5 (2) 62.6 (1)


C2S 20.00 21.75 20.24 (10) 22.8 (2) 20.0 (2)
C3A 10.00 10.29 9.59 (6) 8.5 (1) 9.4 (1)
C4AF 10.00 9.70 10.02 (7) 8.2 (2) 8.0 (2)

Rwp (%) ± ± 11.2 6.5 6.9

Table 5
Comparison of the weight fractions for arti®cial grey Portland cement-1 as in Table 3.
Weighed (%) Weighedc (%) SXRPD (wt%) LXRPD1 (wt%) LXRPD2 (wt%)

C3S 62.00 60.35 61.64 (7) 61.1 (2) 60.9 (2)


C2S 15.00 16.36 15.16 (10) 17.7 (5) 17.2 (5)
C3A 5.00 5.16 5.52 (5) 5.1 (1) 5.3 (1)
C4AF 10.00 9.73 9.49 (6) 7.6 (1) 8.0 (1)
Gypsum 4.00 4.28 4.29 (6) 3.9 (1) 4.1 (2)
Bassanite 4.00 4.13 3.90 (6) 4.6 (2) 4.5 (2)

Rwp (%) ± ± 10.6 10.3 12.6

Table 6
Comparison of the weight fractions for arti®cial grey Portland cement-2 as in Table 3.
Weighed (%) Weighedc (%) SXRPD1 (wt%) SXRPD2 (wt%) LXRPD1 (wt%) LXRPD2 (wt%)

C3S 65.00 63.05 64.06 (6) 63.78 (6) 62.7 (2) 62.2 (2)
C2S 10.00 10.87 9.66 (11) 9.69 (11) 11.5 (3) 10.8 (3)
C3A 5.00 5.14 5.36 (5) 5.50 (4) 5.1 (2) 5.2 (1)
C4AF 5.00 4.85 4.76 (5) 4.94 (5) 4.4 (2) 5.1 (2)
Gypsum 5.00 5.33 5.49 (7) 5.54 (7) 4.9 (1) 4.9 (2)
Bassanite 5.00 5.14 4.96 (6) 4.85 (6) 5.5 (2) 5.6 (2)
CaCO3 5.00 5.61 5.71 (6) 5.70 (6) 5.9 (1) 6.2 (2)

Rwp (%) ± ± 11.1 11.1 11.2 12.2

1172 De la Torre and Aranda  Portland cements J. Appl. Cryst. (2003). 36, 1169±1176
electronic reprint
research papers
in LXRPD patterns collected in ¯at geometry. A value of 1 for The re®ned overall parameters were background coef®-
the texture index represents an ideal random powder whereas cients, cell parameters, zero-shift error, peak shape para-
1 would stand for a single crystal. meters, preferred orientation (when appropriate) and phase
The references for the crystal structures used to calculate fractions. Brindley correction was not applied. The peak shape
powder patterns are given in Table 2. When anisotropic parameters (GW and LY) were re®ned freely for each phase in
vibration temperature factors are reported, these were the synchrotron patterns. However, the strong peak over-
converted to the corresponding isotropic values and intro- lapping in the laboratory X-ray patterns did not allow free
duced in the Rietveld analyses. The structure of -Al2O3 used re®nements of these parameters. For alite, we could re®ne
in all binary Rietveld re®nements has Al at (0, 0, 0.35227) and both GW and LY; for the remaining phases GW was set to
O at (0.69396, 0, 1/4) (Maslen et al., 1993). The thermal 5.0 (0.01 )2 and only LY was re®ned.
coef®cients were Uiso = 0.004 A Ê 2 for both atoms.
There are several reports (e.g. Scho®eld et al., 1996)
describing the crystal structure of gypsum. However, the ®ne 3. Results
structural details of such a simple inorganic solid,
CaSO4.2H2O, are not well known. This is due to problems with 3.1. Binary mixtures
the single crystals, which present spots with irregular shapes The ®rst step of this work was to characterize each `pure'
(for twin-free crystals). The positional parameters are not single phase with SXRPD data of binary mixtures with -
greatly affected but the thermal parameters vary enormously Al2O3. Quantitative phase results for these analyses are given
in the reported structures. We are carrying out a structural and in Table 1. From the results of these analyses, we can calculate
microstructural study of gypsum with ultra-high resolution a correction factor, fc, that takes into account the deviation of
synchrotron powder diffraction which will be reported else- the SXRPD Rietveld re®ned phase fractions from the weighed
where. ones. The unit-cell parameters for the studied phases are
reported in Table 2 for the sake of comparison with the
corresponding values for commercial Portland clinkers and
cements.
The SXRPD pattern of the alite + Al2O3 sample showed
that alite was a pure crystalline phase and belite was not
observed. Three synthesized samples were not completely
pure phase. (i) Diffraction peaks from -C2S were detected in
the SXRPD pattern of the -C2S + -Al2O3 binary mixture.
The Rietveld analysis of the three phases gave: 50.18 (4)% for
-C2S, 49.68 (5)% for -Al2O3 and 0.14 (2)% for -C2S (wt%).
The crystal structure description for -C2S was that reported
by Mumme et al. (1995). (ii) Anhydrite, CaSO4, was detected
as an impurity in the bassanite sample. This phase was
described with the crystal structure published by Kirfel & Will
(1980). The Rietveld result for the CaSO4.0.5H2O + -Al2O3
binary mixture was 48.86 (6)% for bassanite, 50.45 (5)% for
corundum and 0.69 (2)% for anhydrite (wt%). (iii) C3A was
detected in the C4AF + -Al2O3 binary mixture. The re®ne-
ment with the three phases converged to: 47.83 (4)% for
C4AF, 52.03 (4)% for -Al2O3 and 0.14 (1)% for C3A (wt%).
However, the impurity content was always smaller than 1%
and becomes very dilute in the prepared arti®cial mixtures.
The Rietveld analyses of the SXRPD patterns of the arti®cial
mixtures did not evidence these impurities. Hence, we report
the analysis of these three samples as two-phase mixtures in
Table 1.
The crystal structure of C4AF was re®ned starting from that
reported (Colville & Geller, 1971). This structure has two sites
for Al and Fe atoms, one tetrahedral and the other octahedral.
It is known that the Fe/Al ratios in these two sites depend
Figure 1 upon the synthesis, mainly the cooling rate. The Fe/Al ratios
(a) Low-angle region of an LXRPD2 Rietveld plot ( = 1.54 A Ê ) for
were constrained to full occupancy and to maintain the
arti®cial Portland cement-2 with the main diffraction peaks labelled. The
phases are: C3S, C2S, C4AF, C3A, CSH0.5, CSH2 and CC. (b)
nominal stoichiometry, overall Fe/Al = 1. Fe content
Intermediate-angle region of the LXRPD2 Rietveld plot ( = 1.54 A Ê) converged to 70.2 (2)% at the octahedral site [0, 0, 0.0023 (2)]
for arti®cial Portland cement-2 as in (a). and, therefore, to 29.8% in the tetrahedral site [0.9257 (2), 1/4,

J. Appl. Cryst. (2003). 36, 1169±1176 De la Torre and Aranda  Portland cements 1173
electronic reprint
research papers
0.9534 (1)]. Isotropic thermal parameters were also re®ned. weighed fractions that would account for deviations if strong
This re®ned structure was used for the structural description amorphous content is present in some `pure' crystalline phase.
of C4AF in the arti®cial mixtures. We must note that weighed fractions included all the mass but
SXRPD data for the C3A mixture were collected twice. the QPA with the Rietveld method only accounts for the
Firstly, the sample was loaded in a capillary of 2 mm and crystalline fractions. It must be emphasized that the poly-
secondly in a 1 mm capillary. Rietveld results for both analyses morphs used in this study are those present in the commercial
are given in Table 1. The main diffraction peak for C3A Portland clinkers and cements. Thus, the results of this study
created problems in the ®tting because it had a smaller with arti®cial mixtures can be directly exported to the analyses
experimental intensity than that calculated from the structure. of commercial samples.
All other peaks were ®tted nicely, so secondary extinction was
the likely cause of this intensity problem. We recorded an
additional pattern with a 1 mm capillary to rule out the 4.1. Binary mixtures
saturation problem in the detectors. The recorded intensity of It should be noted that the errors stated in Table 1 are those
the main peak presented the same problem. The extinction only arising from the counting statistics of the patterns. The
correction implemented in GSAS improved the ®t but the correction factor, fc, for each phase is inferred using the
problem was still present. Therefore, this region (8.46±8.53 RpWs/RsWp ratio, where Wp and Rp stand for the weighed
2) was removed from all the patterns where C3A was present. fraction and the Rietveld result for the characterized phase,
We also have problems with the main peak of CaCO3 (7.39± respectively, and Ws and Rs stand for the same values for the
7.71 2), again arising from secondary extinction. However, standard. In an ideal case, the fc value must be 1.00. The fc
the discrepancy for the main peak of this compound was value departs from 1.0 because (i) it may cause problems in
smaller. This region was also removed from the re®nement of the samples (minor crystalline phase not de®ned, amorphous
the binary mixture pattern. Secondary extinction was not and non-coherently diffracting regions, etc.), (ii) the particle
observed in any LXRPD pattern. statistics (powder averaging) are not ideal, (iii) problems
The Rwp values for all re®nements are given in Tables 1 and describing the peak shape, and (iv) inadequacies in the
3±6. We also report RF for each phase in Table 1 in order to structural descriptions of some phases. The fc values reported
illustrate the quality of the ®tting for each phase. It should be in Table 1 do not deviate very importantly from 1.0, which
noted that this R factor depends on the agreement between indicates that the samples do not have large amounts of
the recorded pattern and the structural description for that amorphous phases and that the structural descriptions for
phase. those phases are adequate. This is also independently shown in
Table 1 by the low value of the RF factors for these phases.
3.2. Artificial mixtures The SXRPD pattern for the C3A binary mixture was
collected twice. The results are quite close but not identical.
The four arti®cial Portland mixtures have been analysed by The relative error is smaller than 1% (absolute error lower
the Rietveld method using both SXRPD and LXRPD data. than 0.5%) indicating a high precision in the SXRPD analysis.
The results for the AWC sample are given in Table 3. They The fc value used for C3A was the average, 0.9744. fc values
include the weighed fractions, the corrected crystalline were used to obtain the corrected crystalline mass fractions for
weighed fractions (using fc), the re®ned fractions from the arti®cial mixtures, Weighedc which are given in Tables 3±6. It
analysis of the SXRPD pattern and the re®ned fractions from must be noted that C3A powder diffraction peaks are extre-
the study of both LXRPD patterns. mely sharp with a full width at half-maximum (FWHM) of
The results for the AGC sample, with four phases, are given 0.005 , while the FWHM for -Al2O3 was 0.009 . Owing to the
in Table 4. Similar results for the AC1 sample, with six phases, short wavelength used, microabsorption is not an important
are given in Table 5. The results for the AC2 sample, with problem in the SXRPD analyses (see Table 2). However,
seven phases (®ve being at low contents, 5 wt%) are given in secondary extinction was observed for C3A and CC in the
Table 6. SXRPD data for the AC2 sample were measured and SXRPD patterns (high-energy X-rays) but not in the LXRPD
analysed twice in order to evaluate the precision in the QPA of patterns (low-energy X-rays). This effect was likely caused by
these type of data. The Rietveld plots of the LXRPD2 data for the combination of high-energy X-rays, large particle sizes and
AC2 are shown in Fig. 1. the mosaic block size of the powder grains for the two phases.

4. Discussion 4.2. Artificial mixtures


The objective of this work is to evaluate the accuracy of QPA We will focus now on the core part of this work with the
of Portland cement mixtures by using LXRPD data analysed results shown Tables 3±6. Comparing the weighed fractions
with the Rietveld method. The standard mixtures were and the corrected weighed fractions, we inferred that the non-
analysed with high-energy parallel synchrotron X-rays in diffracting regions in the `pure' phases are not very important
order to obtain the best possible QPA results using the Riet- since these values do not differ signi®cantly. Furthermore, the
veld method. Furthermore, by analysing the binary mixtures QPA obtained from the Rietveld re®nement of the SXRPD
with corundum, we can obtain a set of corrected crystalline patterns agree fairly well with these values. It should be noted

1174 De la Torre and Aranda  Portland cements J. Appl. Cryst. (2003). 36, 1169±1176
electronic reprint
research papers
that the two main drawbacks of QPA using the Rietveld improves the particles averaging in these `relatively small'
re®nement and powder diffraction (microabsorption and poor probing volumes.
powder averaging of the different phases) are overcome by AWC is the simplest mixture; the results are shown in Table
high-energy synchrotron powder diffraction working in 3. The QPA results for the AGC sample are shown in Table 4.
Debye±Scherrer geometry with wide capillaries. The values are satisfactory as the relative errors from the
The QPA of SXRPD data of Portland cements is both LXRPD data are of the order of 2% for the main phases and
precise and accurate. The precision in the SXRPD data of the order of 8% for the low-content phases. It should be
analyses is very high, as shown in Table 6 (AC2 sample) where noted that C4AF is invariably underestimated in all LXRPD
two analyses with data recorded from two different capillaries analyses (see Tables 4±6). This is due to microabsorption
are presented. This is a very complex case with seven phases, effect in this phase. The linear absorption coef®cient of the
®ve of them being at the 5 wt% level. The relative errors AWC mixture for Cu K 1,2 is approximately 320 cmÿ1 and the
between the two analyses (precision) in the two main phases  value for C4AF is larger at 495 cmÿ1. The microabsorption
are smaller than 0.5%, and in the low-content phases they are effect is more pronounced in our arti®cial mixtures as the
smaller than 3% (absolute errors always lower than 0.3%). average particle size of C4AF is large, 15 mm (see Table 2).
Furthermore, the SXRPD data analyses are also considered to Furthermore, this phase showed a small preferred orientation
be accurate because the determined weight fractions agree effect, which was corrected as described in the experimental
fairly well with the Weighed and Weighedc values; see Tables section. However, these problems are not important in
3±6. It should be noted that the QPA values from the commercial samples, in which the particle size of this phase is
synchrotron studies are systematically slightly closer to the small (<1±2 mm) as it is the last crystallizing fraction on
weighed fractions than to the corrected values. The relative cooling in the kiln.
errors compared with the Weighed and Weighedc values The QPA results for the AC1 sample are shown in Table 5.
(accuracy) are of the order of 1% for the main phases and of The results are very satisfactory since the values obtained
the order of 3±4% for the low-content phases. These results from the LXRPD analyses are quite close to those expected.
are outstanding and they clearly show that the high-energy The main difference is the above-described underestimation
SXRPD method is the more appropriate method to char- of C4AF due to microabsorption. It must be noted that the
acterize/certify standard mixtures for QPA using powder relative errors in the low-content phases are not very high,
diffraction. being smaller than 10%.
However, SXRPD is an expensive technique to char- The QPA results for the AC2 sample are shown in Table 6.
acterize/certify standard mixtures and routine QPA using This is the most complex case with seven crystalline phases.
powder diffraction has to be carried out in-house with Composite cements may contain mineral additions such as
laboratory X-ray devices. Tables 3±6 present a comparison of limestone. Quanti®cation of this phase is not an easy task since
LXRPD data, the SXRPD results and Weighed and Weighedc the most intense peaks are overlapped with those arising from
values. We have repeated each analysis to evaluate the tricalcium silicate. However, its quanti®cation is possible and,
precision and accuracy of the QPA using LXRPD data of furthermore, the relative errors in the low-content phases are
Portland cements mixtures. At this point, it is important to not very high, being of the order of 10±15% for these phases.
estimate and compare the active sample volumes in both the The quality of the re®nements is good, as evidenced in the
SXRPD and LXRPD techniques. For SXRPD patterns difference curves of Fig. 1, which are quite ¯at.
recorded in transmission, the calculation of the irradiated It should be noted that the texture effect for C3S in these
volume is easy. Taking into account the beam size (4 mm arti®cial mixtures is not pronounced where alite particles have
horizontal and 2 mm vertical), the capillary radius of 1 mm, been grown in the solid state. This is much more important for
and for an approximate capillary compaction coef®cient of commercial samples. This is likely due to the difference in the
50%, the irradiated volume was 6.3 mm3. For LXRPD particle sizes for the different phases. In the studied samples,
patterns recorded in re¯ection, the calculation of the irra- the particle sizes are similar; however, for commercial samples
diated volume is approximate. Furthermore, it changes with the particle sizes for C3S are much larger than those of the
the slits used in the clinker and cements studies. For a 2, 2, other minor phases and, consequently, become more orien-
0.2 mm set of slits (clinkers), the beam size at 20 (2) was tated (De la Torre & Aranda, 2003). Furthermore, alite of
10 mm  16 mm wide. For an average linear absorption commercial Portland cements grows in the liquid phase, which
coef®cient of 300 cmÿ1 and an attenuation of 80%, the leads to well de®ned crystals. The alite particles with large and
resulting surface-perpendicular penetration depth at this clean surfaces exhibit strong preferred orientation.
angle was 5 mm. Thus, the calculated active sample volume was
0.7 mm3. For a 1, 1, 0.1 mm set of slits (cements), the beam
size at 20 (2) was 6 mm  16 mm wide. In this case, the 5. Conclusions
calculated active sample volume was 0.4 mm3. Hence, the We have shown that quanti®cation of Portland cement
irradiated sample volume in SXRPD is an order of magnitude mixtures is possible with laboratory powder diffraction data
larger than those in LXRPD, which leads to better particle and the Rietveld method. Firstly, the structural descriptions of
statistics and to a higher precision and accuracy in the QPA. the phases are adequate. Secondly, good quanti®cations of the
Rotation of the holder during LXRPD data collection Portland cement phases can be obtained with relative errors of

J. Appl. Cryst. (2003). 36, 1169±1176 De la Torre and Aranda  Portland cements 1175
electronic reprint
research papers
the order of 2±3% for the main phases and 5±10% for the low- Madsen, I. C., Scarlett, N. V. Y., Cranswick, L. M. D. & Lwin. T.
content phases. Finally, high-energy synchrotron powder (2001). J. Appl. Cryst. 34, 409±426.
Manias, C., Retallack, D. & Madsen, I. C. (2000). World Cement,
diffraction is highly appropriate to characterize standard
February, pp. 78±81.
mixtures for QPA using powder diffraction. Maslen, E. N., Streltsov, V. A., Streltsova, N. R. & Ishizawa, N. (1995).
Acta Cryst. B51, 929±939.
Maslen, E. N., Streltsov, V. A., Streltsova, N. R., Ishizawa, N. & Satow,
We thank the ESRF, Grenoble, for providing synchrotron Y. (1993). Acta Cryst. B49, 973±980.
X-ray beam time on ID31. Mondal, P. & Jeffery, J. W. (1975). Acta Cryst. B31, 689±697.
Mumme, W., Cranswick L. & Chakoumakos B. (1996). Neues Jahrb
Mineral.-Abhandlungen, 170, 171±188.
References Mumme, W. G., Hill, R. J., Bushnell-Wye, G. & Segnit, E. R. (1995).
Bezou, C., Nonat, A., Mutin, J. C., Christensen, A. N. & Lehmann, M. Neues Jahrb Mineral.-Abhandlungen, 169, 35±68.
S. (1995). J. Solid State Chem. 117, 165±176. Nettleship, I., Slavick, K. G., Kim, Y. J. & Kriven W. M. (1992). J. Am.
Bish, D. L. & Howard, S. A. (1988). J. Appl. Cryst. 21, 86±91. Ceram. Soc. 75, 2400±2406.
Bogue, R. H. (1929). Ind. Eng. Chem. (Anal. Ed.) 1, 192. Neubauer, J. (1998). Proceedings of the 20th International Conference
Colville, A. A. & Geller, S. (1971). Acta Cryst. B27, 2311±2315. on Cement Microscopy, Guadalajara, pp. 103±119.
De la Torre, A. G. & Aranda, M. A. G. (2003). Proceedings of the 11th Neubauer, J., PoÈllmann, H. & Meyer, H. W. (1997). Proceedings of the
International Congress on the Chemistry of Cement, pp. 135±144. 10th International Congress on the Chemistry of Cement, pp. 7±18,
Durban, South Africa. ISBN 0-958-40858-0. 3v007.
De la Torre, A. G., Bruque, S. & Aranda, M. A. G. (2001). J. Appl. Neubauer, J. & Sieber, R. (1996). Mater. Sci. Forum, 228±231, 807±
Cryst. 34, 196±202. 812.
De la Torre, A. G., Bruque, S., Campo, J. & Aranda, M. A. G. (2002). Pajares, I., De la Torre, A. G., MartõÂnez-RamõÂrez, S., Puertas, F.,
Cement Concrete Res. 32, 1347±1356. Blanco-Varela, M. T. & Aranda, M. A. G. (2002). Powder Diffr. 17,
De la Torre, A. G., Cabeza, A., Calvente, A., Bruque, S. & Aranda, M. 281±286.
A. G. (2001). Anal. Chem. 73, 151±156. Rietveld, H. M. (1969). J. Appl. Cryst. 2, 65±71.
Dollase, W. A. (1986). J. Appl. Cryst. 19, 267±272. Scarlett, N. V. Y., Madsen, I. C., Cranswick, L. M. D., Lwin. T.,
Finger, L. W., Cox, D. E. & Jephcoat, A. P. (1994). J. Appl. Cryst. 27, Groleau, E., Stephenson, G., Aylmore, M. & Agron-Olshina, N.
892±900. (2002). J. Appl. Cryst. 35, 383±400.
FuÈllmann, T., Walenta, G., PoÈllmann, H., Gimenez, M., Lauzon, C., Scarlett, N. V. Y., Madsen, I. C., Manias, C. & Retallack, D. (2001).
Hagopian-Babikian, S., Dalrymple, T. & Noon, P. (2001). Int. Powder Diffr. 16, 71±80.
Cement Rev. 2, 41±43. Schmidt, R. & Kern, A. (2001). World Cement, February, pp. 2±8.
Goetz-Neunhoeffer, F. & Neubauer J. (1997). Proceedings of the 10th Schmidt, R. & PoÈllmann, H. (2000). Mater. Sci. Forum, 321±324,
International Congress on the Chemistry of Cement, Gothenburg, 1022±1027.
edited by H. Justnes. Scho®eld, P. F., Knight, K. S. & Stretton, I. C. (1996). Am. Mineral. 81,
Guirado, F., GalõÂ, S. & ChinchoÂn, S. (1996). World Cement Res. Dev. 847±851.
pp. 73±76. Stephens, P. W. (1999). J. Appl. Cryst. 32, 281±289.
Guirado, F., GalõÂ, S. & ChinchoÂn, S. (2000). Cement Concrete Res. 30, Taylor, H. F. W. (1997). Cement Chemistry. London: Thomas Telford.
1023±1029. Taylor, J. C., Hinczak, I. & Matulis, C. E. (2000). Powder Diffr. 15, 7±
Hill, R. J. & Howard, C. J. (1987). J. Appl. Cryst. 20, 467±474. 18.
Kern, A. (2001). Accuracy in Powder Diffraction III. Gaithersburg: Taylor, J. C. & Aldridge, L. P. (1993). Powder Diffr. 3, 138±144.
NIST. Thompson, P., Cox, D. E. & Hasting, J. B. (1987). J. Appl. Cryst. 20,
Kirfel, A. & Will, G. (1980). Acta Cryst. B36, 2881±2890. 79±83.
Klug, H. & Alexander, L. E. (1974). X-ray Diffraction Procedures. Von Dreele, R. B. (1997). J. Appl. Cryst. 30, 517±525.
New York: John Wiley. Walenta, G., FuÈlmann, T., GimeÂnez, M., Leroy, I., Friedle, R., Schnedl,
Kockelmann, W. & Kirfel, A. (2001). J. Archaeol. Sci. 28, 213±222. G., Hartung, D., Staupendahl, G., Lauzon, C. & Decay, D. (2001).
Larson, A. C. & Von Dreele, R. B. (1994). GSAS ± General Structural International Cement Review, June, pp. 51±54.
Analysis System, Los Alamos National Laboratory Report No. LA- Zevin, L. S. & Kimmel G. (1995). Quantitative X-ray Diffractometry.
UR-86-748. New York: Springer-Verlag.

1176 De la Torre and Aranda  Portland cements J. Appl. Cryst. (2003). 36, 1169±1176
View publication stats
electronic reprint

You might also like