<< Back to Issues and Controversy: Measurement of Crystalline Silica

Evaluation of QXRPD as a Tool for Crystalline Silica Analysis

Compared with the alternative techniques for analyzing crystalline silica in bulk and respirable samples, XRD is unique in its sensitivity to the specific crystalline phase or phases that may be present in the sample. For identification applications to detect quartz and cristobalite and to determine the interfering minerals, diffraction techniques are fast and easy to apply. For quantification applications, diffraction methods have proved as accurate as any of the other methods available.

Recent comparisons with infrared by Pickard (1985) indicates that the performances of both methods in terms of detection limits and precision is equivalent and acceptable for occupational hygiene surveys at the present exposure limits. Anderson (1975) came to the conclusion that the X-ray diffraction method was unacceptable because there were too many interferences. Other comparisons by Swallow (1978, 1980) and Groff (1980) (quoted in Chung, 1982) using round-robin tests on samples prepared by one laboratory and analyzed by 61 other labs indicate precisions that are generally unacceptable for all methods. Huggens et al. (1985) describe another interagency test. No one method stands out as superior to any other method. Precisions, /x in %, are reported from 28 to 50 % for respirable quartz and 22 to 24 % for cristobalite in a bulk sample. These figures echo earlier reports by Freedman et al. (1974), Peters (1976) and Nagelschmidt (1956). Chung (1982) summarizes the situation as follows: "[XRD] needs validated procedures to attain precision; it needs certified primary standards to achieve accuracy." NIOSH Method 7500 is now a validated procedure, and NIST SRM's 1878 and 1879 are certified primary standards. However, this procedure and these standards are for only one specific analysis
  that is for the determination of respirable crystalline silica collected in a specific manner.

The objection to the NIOSH Method 7500 is its time-consuming preparation which precludes fast and economic analyses. Many laboratories have proposed and tested alternate procedures which are more economical and which compare favorable to Method 7500. It is apparent that further efforts are needed to locate a more generally acceptable method. Until then, Method 7500 is the only method with any legal significance. The precision and accuracy limits are related to the crystallite statistics discussed below. It may be that the methods are already at the limit.

Specific Problems in QXRPD Analysis

Regardless of the relative accuracy of the XRD methods, its sensitivity to the specific silica phase requires its use in many situations. This section will consider some of the experimental problems that have been recognized in recent years.

Problems Specific to Filter Techniques The many studies involving membrane filters as the collecting and sample supporting method have identified several problems. Different filters have different efficiencies at collecting and retaining particles (Mark, 1974). The organic filters, PVC and MCE, have one side which is smooth and one side which is rougher. The rough side retains particles better, but the smaller particles may be drawn into the interior of the filter and partially masked from the X-ray beam during the analysis. The smooth side is reported to lead to more reproducible results when used for calibration mounts (Edmonds et al. 1977). Because of the heavy element, chlorine, the PVC filters have a significant background. MCE filters have considerably lower backgrounds. Silver filters have low backgrounds because of their crystalline nature, but they also exhibit masking effects. The great advantage of silver is that it produces a diffraction peak that does not interfere with quartz or cristobalite. The attenuation of this peak can be used to measure the absorption coefficient of the sample.

Some preferred orientation of quartz has been reported by Edmonds et al. (1977). Quartz does have a tendency to fracture subparallel to (101) and may have a cleavage parallel to (100) (Frondel, 1962). The presence of flat surfaces on crushed grains would tend to orient the particles where the substrate is very smooth. This effect is more prominent with large particles, > 15 m, and with freshly crushed quartz than with small particles and precipitated quartz. The effect is strongest with small amounts of quartz on a smooth filter.

The particle size distribution has proved to be very critical because of the amorphous layer on the quartz particles. Lorberau et al. (1991) have illustrated the need to match the size distribution in the calibrating material and the analyte. Even the SRM 1878 needs to be sieved to remove the >10 m particles. For samples prepared by dispersing small amounts of respirable-sized calibrant as a thin film on a filter and then determining both the sample weight and the intensity response, a significant fraction of amorphous component will strongly diminish the intensity response and alter the calibration curve.

Problems with Bulk Samples The problems of processing bulk samples containing quartz is essentially the same as processing bulk samples in general, and there are many discussions of the difficulties encountered. The two major problems are preparing a sample that is representative of the bulk material and eliminating preferred orientation of the particles if they have a tendency to orient. The first is common to all analytical procedures. An aliquot must be obtained that is equal in composition to the material being analyzed whether it is a bin of raw materials or a small block of a product. This aliquot must then be pulverized to < 5 m particle sizes and split a into smaller aliquot for the diffraction analysis. The preparation must not allow any particle separation because of different physical properties of the different phases.

Eliminating orientation may be more difficult. There are several suggestions in the literature. Smith and Barrett (1979), Jenkins et al. (1986) and Bish and Reynolds (1989) have reviewed sample preparation for diffraction applications in general. Calvert et al. (1982) have compared several methods of orientation reduction, some simple, some complex. The side-rifted method of loading sample holders is shown to be the simplest effective method that works well in many cases. In severe cases spheroidizing will be necessary. Smith et al. (1979a and b) describe the spheroidizing process in detail which is patterned after a ceramic process called "spray drying". The concept is to create clusters of particles in a spherical shape and then to pack the spheres randomly into a cavity mount. This technique is effective regardless of the shape of the individual particles.

The potential accuracy obtainable from a bulk sample is related to the effective number of particles in the X-ray beam. As seen in the section on crystallite statistics, the number of particles is controlled by the absorption coefficient. The discussion considered a pure quartz sample, but if the quartz is mixed with iron oxide, the effective number of particles is considerably reduced, and the accuracy is also affected. Under ideal conditions using modern diffractometers with high resolution, the maximum achievable accuracy is around 2% absolute.

Detection limits which are usually quoted are for conditions which are ideal, i.e. when there are no interferences for the strongest peak of quartz or cristobalite. Fortunately, this situation does exist for many analyses. Unfortunately, the most common contaminants such as mica and the clay minerals do interfere with the (101) quartz peak and degrade the detection limit significantly as well as affect the quantification accuracy for most analyses. It is difficult to determine a detection limit where interferences exist.

The amorphous surface phase on quartz particles is probably of little significance in most bulk samples. The main reason is that it is really difficult to reduce the particle size to the range where the surface layer becomes dominant. Also, freshly broken p
articles may not develop the layer immediately, and the diffraction experiment may be completed before the layer forms.

Comments on Standardized Procedures for Respirable Silica

The standard NIOSH Method 7500 is the current legally accepted procedure for determining respirable quartz and cristobalite, and all other methods which are used must be compared with this method. What is the ultimate accuracy of the method? One of the factors in setting the limit of accuracy is the crystal statistics. The analysis for the bulk sample showed that the effective volume of the sample is around 20 mm3. For a respirable dust sample of 2 mg distributed evenly over a filter, the effective volume of sample is 0.75 mm3. This volume is only 4% of the bulk volume, and the crystallite statistics are affected accordingly. Assuming a 2 m average particle size, the number of effective particles in the sample is 1.8 x 108 and the number in diffraction is around 4000. At the 2.3 Sigma level, the accuracy is around 4% for this number of particles. This figure is an absolute accuracy regardless of the percentage of quartz in the sample. As the percentage of quartz decreases, the relative accuracy increases significantly.

With this estimate of absolute accuracy, it is apparent that the X-ray diffraction method is already near its limit with the present diffractometers. Several of the alternative procedures proposed to increase the economics of running many samples agree with accuracies close to this 4% figure. How then can accuracy be improved? The obvious direction is to increase the effective number of crystallites in diffraction. Use of a larger sample leads to problems in particle retention and masking which requires absorption corrections. It should be more effective to increase the range of diffraction to include more crystallites. Most diffractometers are set up and aligned for maximum resolution. If a broad-focus X-ray tube were used with coarse sollar slits and focal slits there would be a wider angular range for crystallites to diffract. If the sample is also rocked a few degrees while it is spinning, the angular range would be increased considerably. Because quantitative measurements are based on intensity, the loss in resolution would be inconsequential, and the quantification accuracy could be improved considerably.

Bulk Sample Analysis

Standardized procedures for quantifying crystalline silica in bulk samples do not exist in the sense of having a single recognized and accepted procedure which has legal status. However, many analytical service laboratories have established procedures specific to their local needs depending on the types of samples to be processed. The biggest single problem is the variability of the matrix in which the silica forms occur and the interfering effects of the matrix compounds. If there are no interferences, quantification down to the detection limit is possible by direct peak integration. If interferences are present, then the procedure must be tailored to this interference. Two main approaches may be used, either a correction is applied for the interfering phase of the interfering phase is removed from the sample prior to the diffraction analysis.

Correction procedures for interfering phases are difficult to apply to diffraction data unless a considerable amount of data is acquired. The interfering phase concentration must be determined by diffraction effects in parts of the diffraction trace that are free of their own interferences and then the contribution of this concentration to the silica peaks must be subtracted to obtain the silica quantification. Propagation of errors in the individual abundances makes this approach undesirable. The use of whole-pattern fitting methods alleviates some of the difficulties of pattern overlap but not all. Profile fitting of specific peaks also shows some promise but cannot resolve peaks that are directly superimposed.

Pre-concentration methods applied to eliminate the non-silica phases in a sample shows considerable promise. Physical separations are not effective, but chemical methods are. Quartz is sufficiently inert for many phases to be removed in its presence by chemically methods. In fact, the chemical methods for silica analysis rely on the quantitative removal of non-silica phases. For diffraction analysis, only the interfering phases need to be removed, and if 90% of the bulk sample can be eliminated, the silica concentration is increased an order of magnitude. The detection of 0.1% silica becomes the detection of 1% silica which considerably simplifies the analysis. If the silica phase is indeed inert, the concentration factor is simply the weight ratio of the bulk sample pre- and post-treatment. No determination of phases in the matrix is required. Because the samples are bulk with an adequate quantity of material available for an infinitely thick sample, any of the three methods: absorption correction, internal standard or external standard may be employed.

There is considerable doubt that a "standard procedure" for bulk analysis is feasible or desirable. Because of the variability of the matrixes, no single procedure is applicable to all situations. Because no single procedure is feasible, the establishment of any procedure as a "standard" could present legal difficulties in those situations where the "standard' procedure is not appropriate as it might preclude the acceptance of a more accurate appropriate procedure. Whenever legal status of an analysis is required, it would be more suitable to establish a set of criteria for creating and measuring calibration samples and for simulating tests of comparable bulk samples than to define and approve a single "standard" analytical procedure.

The Cristobalite Problem

The quantification of cristobalite requires considerably more study before quantification can be done with confidence. Studies must determine the applicability of diffraction techniques to all situations where cristobalite is encountered in both respirable dust and in bulk samples. It is already established that cristobalite may vary from the well-crystallized form encountered when the temperature of formation is high (~1500C) to very poorly crystalline forms when the formation temperature is low (~50C). There are many physical differences which may occur that alter the diffraction pattern, either affecting the shape of the diffraction peak or its angular position. The crystal structure of cristobalite is actually very complex, and there are several polymorphs which may occur. Usually only the low temperature form is encountered, but slight trace of an alkali element may stabilize one of the high-temperature forms. The structure transitions on cooling may lead to considerable twinning and domain structures which decrease the effective crystallite size. Low temperature growth may lead to very small coherent domains also. Structural defects mostly stacking faults lead to considerable tridymite character. This latter phenomena is common in the devitrification of opal.

Several approaches are available for the quantification of cristobalite. If the full range of the characteristic diffraction peaks are included in the integration, the area which is determined is theoretically proportional to the amount of cristobalite present. However, the possibility of amorphous surface layers as in quartz is very real, and the effect of stacking faults and twinning needs to be evaluated. One possibility exists that may be universally applicable. Heating the sample to 1500C does convert even the most poorly crystalline silica to well-crystallized cristobalite, so a heat treatment may result in a phase consistent enough for quantification. As long as there is no reaction with other matrix phases, the conversion should be quantitative.

In bulk samples which are the result of firing processes where cristobalite is one of the products, the physical state of the cristobalite may prevent any special treatments. If the cristobalite crystallites are incorporated in ceramic products at grain boundaries or within the other grains, the cristobalite cannot be isolated for additional treatments. Fortunately, the cristobalite is usually well crystallized, but it might be in very tiny crystallite sizes producing peak broadening. X-ray diffraction will detect this cristobalite where other methods may not be as sensitive to it. It would be difficult to establish calibration samples for this situation.

A serious interference problem in the determination of cristobalite is due to the similarity of the diffraction patterns of opal-C, opal-C and cristobalite as illustrated in Figure 3. Even the peak position of the amorphous band in opal-A occurs at the same diffraction angle. Consequently, whenever there is the potential for both phases to be present in the sample being monitored, cristobalite will be overestimated when using any of the standard methods based on the integration of the characteristic peaks. Distinguishing the two phases is not easy, but shape of the diffraction peak is indicative. Where the peak width at half height is larger than usual for the diffractometer in use, opal should be suspected. All opal is paracrystalline, and the diffraction peaks, if any, always show broadening due to the effective crystallite size. Very little crystobalite should show crystallite-size broadening effects in the diffraction pattern. Thus, whenever peak broadening is encountered in the diffraction analysis, it should be reported, and the phase should probably be interpreted as opal. An analysis for water in the sample would support the opal designation. Profile-fitting or pattern-fitting methods of diffraction analysis will help resolve this problem.

The Importance of Tridymite and Other Silicas

There is little to no evidence that tridymite is a significant phase in industrial environments. Based on the SiO2 phase diagram, it is logical to worry that tridymite might be present in silica products, but the evidence is that tridymite is difficult to form and may not be a stable phase without the presence of an alkali metal to stabilize it. Regardless, there is little justification at this time to spend any effort establishing procedures for its quantification.

The same may be said for the other forms of silica. Coesite and stishovite are so rare in nature that it is a challenge just to concentrate enough to detect in a X-ray pattern. The other crystalline forms would have to be a product of a reaction. Where the material is being produced for manufacture, e.g. a clathrasil, then the material should be monitored. Otherwise, there is no need to set up a general procedure for its quantification. Opal on the other hand, may be a more serious problem that is under estimated. It may be abundant in some environments without recognition, as its amorphous character makes it difficult to detect. More studies should be made on opal, both on how to detect it and on its potential toxicity.

The technique reported for quantifying opal, NIOSH Method 7501, is a viable approach to the detection and quantification of amorphous silicas. The procedure is based on the same method of sample treatment as is recommended for preparing standard samples of cristobalite from a silica precursor (Chung, 1978). Firing at 1500C quantitatively converts any silica form to cristobalite, even fine-grained quartz. Unfortunately, clay particles andther dust components could react with the silica to form compounds other than cristobalite and reduce the amount of cristobalite formed. The conversion steps including ashing the filter and firing in a platinum crucible may prevent reactions by keeping the particles separated. The major problem is quantitative recovery and transfer of the particles to a substrate for the diffraction analysis.

The Need for Standard Silica Samples

In order to achieve accuracy in any analytical procedure, proper calibration standards are required. At present, 1991, there are only two certified standards available, both supplied by the NIST/OSRD as quartz, SRM-1878, and cristobalite, SRM-1879. The certification for both is phase and chemical composition and average particle size. Altree-Williams et al. (1981a) have described the preparation of quartz standards. Chung (1982) has suggested Min-U-Sil as a standard for quartz and preparation procedures for cristobalite and tridymite. Min-U-Sil is now available from from many manufacturers and varies considerably from batch to batch. Before it could be used, a large quantity of a single sample would have to be set aside, homogenized and characterized. A cristobalite prepared by the Chung method is available from The Gem Dugout, State College, PA. Unfortunately, none of these materials are technically certified. Obviously, a program should be initiated somewhere to provide additional certified standards. Unfortunately, the expense of certification is high, adding significantly to the cost of the material. Only a few laboratories would have the legal status for the certification to be acceptable in court.

Proper certification involves careful physical measurements on aliquots of a sample of sufficient size to provide reference material for many years of supply. The sample should be examined for phase composition, chemical purity, particle size and size distribution, effective crystallite size and crystallite perfection, uniformity of the large mass of material for each parameter, long term stability of the sample, and the feasibility of supplying small aliquots to users at a reasonable cost. For a project such as certification, the cost must be amortized through the sales of the certified samples. Unfortunately, the high cost of each aliquot prevents many users from purchasing the samples and forces the unit cost to be set even higher. It is a "Catch 22" situation.

Samples of cristobalite are much more difficult to obtain than samples of quartz because there is no commercial product available nor is it common in nature. There is a small layered deposit in Eastern Oregon which is 96% SiO2 and only shows a cristobalite-like pattern by diffraction; however, the percent cristobalite (or opal) has not been determined. This natural material could be homogenized and characterized as a reference sample. Alternatively, a large batch of synthetic cristobalite could be prepared by a method similar to that described by Chung (1982) then homogenized and characterized. The high temperature of the synthesis, > 1500C, precludes its routine synthesis in readily available laboratory furnaces. Also, because it needs to be fired in a platinum container, only small batches may be processed at any one time. There are ways to fire larger batches at temperatures up to 1600C, but they would require special arrangements with ceramic processors and might "contaminate" a production line with silica.

Tridymite would have to be synthesized because there is no adequate natural source. First, it would have to be established that tridymite was indeed required. Its synthesis temperature, 1100C, is a more reasonable laboratory temperature, but it would still have to be prepared in small batches.

It is feasible for individual laboratories to synthesize small batches, but not to carry out the full characterization. Although the properties would be reasonably consistent to other batches synthesized in the same manner, such a product would not have the legal status of a certified standard. Particle size could vary considerably depending on how the material was treated after recovery from the crucible. Proper crushing could be tailored to provide specific size and size distributions.

Calibration Curves and Correction Factors

The most common procedure for calibrating the intensity response for the filter methods is to prepare a series of filters with uniformly deposited, weighed amounts of the crystalline silica phase. The intensity response is then measured and plotted versus weight of silica on the filter. As long as the response is linear, there is no particle interference, and no correction is necessary unless masking occurs in a highly absorbing filter such as silver. A departure from linearity usually implies particle masking which then requires an absorption correction. However, theoretical absorption corrections assume a continuous thin layer of sample which is not appropriate to thin particle dispersions which are discontinuous layers. Applying a correction for the absorption effect is difficult with discontinuous samples until the samples are many layers thick to average out the individual particle variations. Thus, it is wise to avoid samples with more than 2 mg on a filter which is where the particles tend to overlap significantly. However, where the absorption is measured directly on the analysis sample by transmission, the averaging effect of the X-ray beam compensates somewhat for the discontinuous distribution. In spite of the inherent difficulties, Casciani and Ripanucci (1984) do report nomograms for the determination of quartz content where the intensity response is not linear.

For bulk samples, absorption corrections are appropriate, but they are more difficult to measure. Because of the thickness of the samples, transmission methods on the same sample are impossible, and a second preparation must be made for the attenuation measurement. Most diffractionists prefer to avoid the absorption correction and to employ either the internal-standard or external-standard method which are feasible for bulk samples.

Detection Limits

The minimum detection limit, MDL, of crystalline silica in very small samples is as important as its quantification at any level. Most health regulations define a personal exposure level, PEL, or a maximum concentration above which action must be taken. What is critical is that the MDL of the analytical procedure be well below the PEL.

Tolerance limits are quoted in mg•m-3 for atmospheres and weight percent for bulk samples. X-ray diffraction measures weight of a sample on a filter or weight percent of a bulk sample. The weight measurement must be converted to weight per unit volume by correcting for the amount of air sampled. Typical samplers filter 1.7 1•min-1 which amounts to about 800 l in an 8 hour working shift. Thus, for respirable silica at the PEL, the amount would be 25 g of quartz. Most procedures are reporting MDL values of 3 - 5 g which is within the PEL. If the total mass of the sample is 2 mg, the MDL is below 1% with an accuracy around 4%.

For bulk samples the MDL is reported to be around 0.03 weight percent (Emig and Smith, 1989) without interferences using present equipment. Another report by Schreiner (1990) claims a detection limit of 0.01%. This value is insufficient to assure the PTL of 0.1 weight percent for bulk products because the accuracy of 2% is outside the acceptance range. Bulk samples may also be treated as respirable samples by depositing on a filter substrate, but Blount (1989) has shown that the MDL is insufficient to be meaningful.