Quantitative X-ray diffraction phase analysis of cements containing several main constituents
H. Vaupel, Clausthal-Zellerfeld/Germany SUMMARY
The application of X-ray diffraction analysis to the investigation of cements containing several main constituents was examined from the viewpoint of an external testing laboratory to which only the finished product, but not the starting materials, were available. This showed that accuracies of ± 0.1% can be achieved when determining gypsum and anhydrite, and of ± 0.2% with hemihydrate. With limestone the main interference peak of the calcite is obstructed by a strong clinker interference peak, so an accuracy of ± 0.2% can be achieved only if the original components are available. If the original components are not available then it is hardly possible to achieve an accuracy better than ± 2%. Blastfurnace slag is an amorphous component so it cannot be measured directly with an X-ray diffractometer. Crystallization of the glassy fraction in a slag cement does not lead to a usable result as reactions take place between the individual constituents in spite of the relatively low temperature of about 900°C. It is therefore necessary to develop methods for determining the other main constituents of the cements by X-ray diffraction analysis.
1. Introduction
Many cement works now use X-ray diffractometers, but mainly just for routine monitoring of the free lime content in the clinker. This does not make full use of the equipment so efforts are being made to find additional uses. One obvious solution is also to use it for monitoring other crystalline components, like limestone, gypsum, hemihydrate and anhydrite, which are only added in small quantities to the cement. To an increasing extent cement plant manufacturers and research laboratories are giving preference to the Rietveld analysis [1, 2]. The entire X-ray diffractogram is recorded and the phase composition of the clinker is calculated with the aid of the diffractograms of the clinker phases present. However, the mineral composition of the clinker depends primarily on the chemical composition of the raw meal. The optimum raw meal composition is stipulated, and then checked and optimized at regular intervals. The constancy of the chemical composition of the raw meal is monitored with the aid of X-ray fluorescence analysis, and the quality of the clinker burning is checked by measuring the free lime content. The mineral composition of the clinker is in fact of fundamental interest to the cement manufacturer, but continuous monitoring is not necessary. Checking the main constituents of the cement, such as cement clinker, blastfurnace slag, oil shale ash, trass, etc., is essential for cement manufacture. However, this information is provided automatically by the metering equipment for the cement mills, so separate monitoring with an X-ray diffractometer is not normally necessary.
occur in materials testing, so the familiar X-ray analysis is preferred to the Rietveld method for the situation in a materials testing laboratory.
3. Principle of the measurement
The quantity of a crystal phase in a phase mixture is measured by the intensity of the X-ray interference peaks which characterize the phase. It is represented by the area under the graphic plot of the interference peak. The height should only be used if the same sample material is tested repeatedly and is always prepared in the same way. This method is therefore generally only appropriate for production monitoring. There are two methods normally used for obtaining the area: pulse counting and profile analysis of the interference peak. With pulse counting all the pulses which are associated with the interference peak are counted by the goniometer and the count result is corrected by the one- or two-sided background measurement. A computer control system is recommended for the measurement, but is not essential. Computer-control systems permit quantitative evaluation of the interference peak of the recorded diffractogram at the conclusion of the measurement. If an interference peak which has to be measured overlaps partially with the interference peak of a foreign phase then pulse counting can only be used to a restricted extent or not at all. For the profile analysis the diffractogram, or sections of it, are processed in the computer. The computer then matches bell curves to the diffractogram in the required range and calculates the area under the bell curve and its height. In this way it is largely possible to separate overlapping interference peaks by calculation. The calculation steps can be followed on screen, and it is possible to make a corrective intervention in the calculation process. The pulse count method presupposes non-overlapping interference peaks. This requirement is not fulfilled by the individual phases of cement clinker. Rietveld analysis and, with certain preconditions, profile analysis of the interference peaks are suitable for phase analysis of cement clinker. Determination of the total clinker fraction without taking the phase composition into account is usually possible both with the pulse count method and with profile analysis.
2. Situation
The number of parameters to be monitored with an X-ray diffractometer in a cement works is generally limited. The possible applications are more extensive in building materials testing, where the laboratories have to examine not only cement but also other binders, including premixed mortar, hardened mortar and concretes. In such cases it is usually necessary to determine the total clinker fraction, and possibly also other constituents, without samples of the pure components being available for reference. Conventional X-ray diffraction analysis is easier to adapt to the special tasks associated with the variety of problems which
(Translation by Mr. Robin B. C. Baker)
4. Test conditions
A powder diffractometer with cobalt tube and fixed attenuators was available for the measurements described here. The samples were ground and then, in the same way as the sample material already available in ground form, carefully reground in portions of 0.5 g in an agate mortar in order to eliminate as far as possible any particle size effects during the measurement. To produce the preparations for diffraction analysis the sample was filled slightly in excess into the front of sample carriers which are closed at the rear, and then lightly compressed. The surplus was then removed with the edge of a glass plate. This avoids any preferred crystal orientation in the surface of the preparation. The resulting slight surface roughness of the powder bed has no disruptive effect on the test result. Excessive roughness indicates that the powder is not fine enough. In some cases measurements were also carried out on smoothly prepared surfaces. The diffractograms were recorded with a step width of 0.06 °(2 ) and a count duration of 1.5 s per step. The recordings, usually of five preparations of the same sample, were added together and then evaluated jointly with the Philips APD Software Version 3.5. This program allows simultaneous evaluation of a maximum of eight input interference peaks.
TABLE 1: Absolute and relative (in brackets) pulse outputs from gypsum in Portland cement Gipsgehalt [M.-%] Gypsum content [wt.–%] 1 100 7,61 Å 7,61 Å text. 7.61 Å 7.61 Å pref. cryst. orient. 12.1 (0.62) 1960 (100) 27.9 (0.77) 3610 4,27 Å 4.27 Å 9.3 (0.64) 1440 4,27 Å text. 4.27 Å pref. cryst. orient. 11.0 (0.73) 1510
100% hemihydrate. Nevertheless, only preparations without preferred crystal orientation should be used here. The second fairly strong hemihydrate interference peak at 3.47 Å is of only limited suitability for the determination. It is located at a distance of only 0.3° (Co tube) from the main anhydrite interference peak; its intensity at 1% by mass is 6.6 pulses, and at 100% is 465 pulses. This interference peak can also only be measured by profile analysis. It cannot be recorded for preparations with preferential crystal orientations. Overall it is hardly possible to achieve an accuracy of the test result better than ± 0.2%. 5.3 Anhydrite as an interground additive Only the main interference peak can be used for determining the anhydrite. An intensity of 30 pulses was measured at a content of 1% by mass, and of 3860 pulses at 100%. Any disruption by the adjacent hemihydrate interference peak can be ignored. On the other side of the anhydrite peak there is a clinker interference peak with 24 pulses at a distance of only 0.2° (Co tube), so once again only profile analysis is possible. The preferred crystal orientation, which here again should be avoided, increases the intensity considerably to 109 pulses for 1% by mass of anhydrite and to 12500 pulses for 100%. It is possible to maintain an accuracy of test result of ± 0.1%. 5.4 Limestone meal Limestone meal consists almost exclusively of calcite. In spite of careful comminution in an agate mortar up to incipient widening of the interference and great care to achieve a preparation which is free from preferential crystal orientation the intensities are very severely scattered peak. The scatter of the count results is considerably lower with precipitated calcium carbonate (chemical) and the intensities of the various interference peaks show some differences from limestone meal. Additions of natural chalk were not investigated. The main interference peak of calcite at 3.04 Å coincides completely with a strong clinker interference peak so its use requires considerable correction. Weak, less distorted, interference peaks at 2.285, 2.095,1.913 and 1.875 Å have relative intensities, measured on natural calcite prepared without any preferential crystal orientation, of 11.9, 10.9, 16.3 and 14.5 (according to X-ray index card 5-586: 18, 18, 17 and 17). The absolute count rates of the weak interference peaks change only slightly where there is a preferred crystal orientation, but the intensity of the main interference peak can rise to double the value. According to information in the literature [3] accuracies of detection of ± 0.1% were achieved when using the very weak interference peak at 2.095 Å under optimized conditions, and further improvement was not ruled out. A linear relationship between concentration and intensity was also established with this interference peak. In in-house investigations clinker meal was mixed in different proportions with coarsely crystalline limestone meal and very carefully ground and homogenized. The proportional pulse rate for the clinker calculated linearly for the clinker fraction was subtracted from total pulse rate of the interference peak at 3.04 Å measured on preparations
5. Test results
5.1 Dihydrate as an interground additive The sulfate content of Portland cements is usually about 3% SO3 by mass, a small part of which is already contained in the clinker and the remainder of which is added during the grinding in the form of dihydrate and/or anhydrite. The dihydrate is partially dewatered to hemihydrate under the influence of the increased temperature in the cement mills. X-ray diffraction determination of these components can be a problem if they are present in fairly large percentages because of their tendency to form a preferred crystal orientation during the sample preparation. In cement the percentages are small and can therefore be determined relatively accurately, but the low absolute count output is a disadvantage. The main interference peak at 7.61 Å is the most suitable for dihydrate determination, and possibly also the somewhat weaker interference peak at 4.27 Å. All the other interference peaks are weaker and become heavily distorted. The pulse count method can be used in this measurement, but it should be borne in mind that there is a C4AF interference peak from the clinker at a distance of 0.6° (Co tube) from the main interference peak. The count rate of the main dihydrate interference peak is doubled by measuring smooth surfaces with a preferred crystal orientation, but the secondary interference peak is only slightly increased. The pulse rates are summarized in Table 1. The reproducibility of the count result is so good with the low levels in preparations with and without preferred crystal orientation that an accuracy of the test result of ± 0.1% can be achieved. In spite of the good reproducibility it is essential in the present case to ensure that the preparations are free from any preferred crystal orientation, as preparations of uniform preferential crystal orientation cannot be produced with different dihydrates. 5.2 Hemihydrate as an interground additive The interference peak at 6.0 Å is the most suitable for determining hemihydrate. This peak can only be recorded with profile analysis as there is a clinker interference peak at a distance of only 0.2° (Co tube). The pulse output for 1% hemihydrate by mass is 8.5 (1190 pulses for 100% by mass), the adjacent clinker interference peak has an intensity of 27 pulses. Here again the count rate is doubled to 14.5 pulses with a smooth sample surface, and to 2600 pulses for
free from preferential crystal orientation. 271 pulses were measured on the in-house equipment for 100% clinker by mass. The calibration curve for calcite in mixtures with cement clinker is practically linear with 22.2 pulses/% by mass in the range investigated up to 65% by mass calcite. An intensity of 3460 pulses was measured for 100% calcite by mass so it can be assumed that the calibration curve is very strongly curved in the upper range. The absolute count rates also show that the accuracy of detection with this method cannot be better than ± 0.2% calcite by mass if the pure clinker used is available for comparison purposes. This is the case for plant production control, but generally not for an external test laboratory. A later work will examine how reliably the intensity of the coinciding clinker interference peak can be calculated from the other interference peaks. A check of the weak interference peaks at 2.285 and 2.095 Å for their usability showed that in both cases the measurement is distorted by very weak diffuse clinker interference peaks. Intensity measurement by profile analysis is therefore unsuitable. In both cases the profile analysis gives a signal for the height of the interference peak which also has to be corrected with the weak signal of the calcite-free clinker. However, the signal is not the same for different clinkers, so the analysis error can be considerably increased if the calcite-free clinker used in the cement under investigation is not available. It is not difficult to grind calcite to the stage of peak widening. The height of the calcite interference peak is therefore only usable if the same limestone is ground in the same way. This is true for a plant laboratory but not for a test laboratory. In a plant laboratory errors can be minimized down to ± 0.1% under optimum conditions [3] but in an external test laboratory errors of ± 2% occur in the most favourable case when using these interference peaks. The calcite interference peaks at 1.91 and 1.875 Å were also checked for usability for quantitative determination. However, in both cases the count rate was so low and the reproducibility so poor that it is hardly possible to achieve an accuracy of ± 5% calcite. There was a possibility that the interaction between the cobalt radiation used and the high calcium content of the binder had a detrimental effect on the accuracy of the analysis, so a further test series was carried out with copper radiation. However, the improvement was only very slight. 5.5 Blastfurnace slag Density separation methods and microscope methods for determining blastfurnace slag are described in the DIN EN 197 and BS 6699 cement standards. The optical method has certain advantages as it is simple, rapid and accurate. It is found in practice, however, that reliable results can only be achieved by experienced specialists. According to Sylla and Syberts [4] good reproducibility can be achieved with chemically selective dissolving methods provided the blastfurnace slag does not have an excessively high MgO content and contains no crystalline fractions. According to Erntroy [5] it is in no way sufficient just to determine the blastfurnace slag fraction in the cement. The melilite, which crystallizes during excessively slow cooling of the slag, does not contribute to the hardening and therefore causes a reduction in strength. Erntroy and Drissen [6]
TABLE 2: Comparison of the relative pulse rates measured on tempered blastfurnace slags (HS) Interferenz Interference peak 3.08 Å 2.86 Å HS A HS B HS C HS D HS E HS F
point out that microscopic assessment of the slag in a ground section is unsuitable as crystalline fractions could be assigned to the clinker. The X-ray diffraction method is therefore preferred. The blastfurnace slag used for producing slag cements contains predominantly glassy silicates and therefore cannot be determined directly by X-ray diffraction. It is, however, possible to determine the clinker fraction and the other crystalline constituents quantitatively by X-ray diffraction and evaluate the difference from 100% as the blastfurnace slag fraction. Another option could be to crystallize the blastfurnace slag by heating it to red heat and then determine it directly with a diffractometer. The following comments deal with these two options. The third option, namely direct measurement of the glass phase, will be examined in a later work. Six blastfurnace slags from different German steelworks were tempered for 3 h at different temperatures and the products then examined by X-ray diffraction. It was found that the intensity of the melilite interference peaks no longer increased above 900° C. However, the absolute pulse rates differed widely. Chemical proximate analyses were not carried out. The melilite pulse output for the tempered blastfurnace slag A was taken to be 90%, as in addition to 5.5% by mass of C2S very small quantities of merwinite were also found (no reference sample was available). This was taken as the reference point for the pulse rates of the other tempered blastfurnace slags. The interference peak at 3.08 Å and the main interference peak at 2.86 Å are suitable for quantitative analysis. The pulse rates emitted are summarized in Table 2 as relative numbers in %. These analysis results show that differences, which are not acceptable for quantitative analysis, occur in the melilite pulse outputs. This indicates that recrystallization of the glass fraction in different blastfurnace slags does not produce comparable pulse rates. In spite of this, slag cements were tempered in further tests. The slag cements were produced by mixing commercial Portland cements (PZ) with blastfurnace slag and by mixing cement clinker meal with blastfurnace slag and adding 5% by mass of natural anhydrite. The mixes were tempered for 3 h at 950° C. The levels of melilite determined by X-ray diffraction are summarized in Table 3. The data in % by mass relates to the pulse rates of the tempered blastfurnace slag fraction. Two examples of the pulse rates of mechanical mixtures of Portland cement and recrystallized blastfurnace slag are given for comparison. The conclusion drawn from these test results is that the pulse outputs in the two-material mix of cement clinker and crystalline blastfurnace slag deviate only slightly from linearity. The method discussed, of tempering (and thereby crystallizing) the slag cement and determining the content of melilite formed by X-ray diffraction and using it as a measure of the slag content does not provide reliable results. Apparently the cause is the reactions between clinker, blastfurnace slag and sulfates which take place even at relatively
TABLE 3: Jointly tempered mixtures of blastfurnace slag HS and Portland cement PZ Zusammen- PZ + HS C setzung Composition (PZ + HS) 0% HS 30% HS 50% HS – – 37.5 PZ + HS D KL + HS C KL + HS D
0.6 6.9 36.8 45.5 44.5 100
– – 33.4 – – 100
0 4.9 22.0 – 41.2 100
50% Melilith 47.5 50% melilite 90.0 90.0 87.9 87.8 90.6 85.9 82.7 87.0 74.7 74.9 90.0 71.3 70% HS 100% HS – 100
low temperatures. This means that the differential analysis mentioned above or some other type of direct analysis are the only methods which can be considered for X-ray diffraction determination of the blastfurnace slag content. 5.6 Pozzolana Trass, ash and similar main constituents of the cement are, with the exception of the cement clinker, partially or pre-
dominantly non-crystalline and therefore cannot be measured with an X-ray diffractometer. If these substances contain a homogeneously distributed crystalline phase then this can be used in a plant laboratory for quantifying the relevant main constituent, but not in an external testing laboratory which does not have this main constituent available for comparison purposes.