METHODS FOR SAMPLING AND
INORGANIC ANALYSIS OF COAL
U.S. Geological Survey
Bulletin 1823
Edited by D.W. Golightly and F.O. Simon
AUTOMATED SEMIQUANTITATIVE
DIRECT-CURRENT ARC SPECTROGRAPHIC
DETERMINATION OF 64 ELEMENTS IN COAL ASH
By A.F. Dorrzapf, Jr., C.J. Skeen, and W.B. Crandell
A semiquantitative, direct-current arc spectrographic method is routinely used to determine 64 elements in coal ash. The automated method is rapid and economical for evaluating both major and trace-element concentrations. The method, a listing of the spectral lines used in the analytical scheme, and data from analyses of National Bureau of Standards standard reference coal fly ashes are presented.
A semiquantitative, direct-current (dc) arc spectrographic method is routinely used to determine 64 elements in coal ash. The automated data acquisition and analysis system used in conjunction with this approach is an outgrowth of a scanning microphotometer concept (Helz, 1965; Helz and others, 1969) that ultimately was extended to its present form. The system maintains long-term consistency of results and provides archival storage capability on both photoplates and microfiche. The dc-arc spectrographic method is applicable in laboratories without the scanning microphotometer system.
This semiquantitative approach achieves analytical ranges and detectability comparable to those of visual estimation procedures (Myers and others, 1961). Standards are diluted with a synthetic silicate matrix to provide six evenly spaced logarithmic divisions (steps) per decade of concentration. When available, natural rock standards are preferred over synthetic standards and are diluted to correspond to the six steps: 1, 1.47, 2.15, 3.16, 4.64, and 6.81. Because the standards used generally do not match the approximate composition of the samples analyzed, the expected accuracy is limited to ±1 step, which corresponds to roughly +50 or -33 percent. The procedure is semiquantitative because the computer algorithm extrapolates concentrations based on prestored coefficients calculated from previously arced standards. Analytical curves are not established from spectra of standards on the same plate as the samples.
The operating conditions and the spectrograph are described in table 4 (Dorrzapf, 1973). Dilution of the sample with graphite increases the uniformity of the introduction of the sample into the arc. The spectrograph has been modified to include a two-position mask near the focal plane (Helz, 1973). In one position, the mask allows only the wavelength regions adjacent to the cadmium lines at 274.8 and 441.5 nm to be exposed. The
spectral lines from a cadmium Osram lamp serve as fiducial lines for wavelength calibration. In the other mask position, the cadmium "windows" are blocked, but the remainder of the spectral region from 230 to 470 nm is exposed. Thus, spectra from the cadmium lamp and the dc arc are coexposed without moving the photoplate. The Helz jet (Helz, 1964) was chosen rather than a Stallwood jet (Stallwood, 1954; Shaw and others, 1958) because the procedure for changing samples was simplified if there was no jet dome to be cleaned. An argon-oxygen atmosphere minimizes cyanogen-band formation and thus frees the wavelength region from approximately 350 to 420 nm for measurments of sensitive lines for Cr, Eu, Gd, K, Mo, Pr, Sc, Sr, Th, Tl, W, and Yb.
THE SCANNING MICROPHOTOMETER SYSTEM
A scanning microphotometer (Helz, 1973) provides the basis for all spectral measurements. The optical system for the microphotometer follows conventional practices for good resolution, high contrast, and low scattered or diffuse light detection. The fixed entrance slit consists of a precision ruled window (100m m wide) in a mirror. A 32-mm focal length Micro-TessarTM lens focuses an image one-tenth the slit size on the plane of the spectrum. The optics are designed to sample a portion of the spectrum 1 mm high by 7 m m wide (about a quarter of the width of a spectral line).
The optical system is mounted on a MooreR no. 3 measuring machine. Lateral motion of the milling machine bed, on which the photoplate rests, is accomplished by a precision lead screw. The signal produced by light passing through the photographic emulsion and detected by a solid-state detector is sampled by an analog-to-digital (A/D) converter that is gated by a rotating shaft encoder attached directly to the ring motor that drives the screw. A scanning speed of 70 s per spectrum provides a practical interval for data collection and minimizes vibrational effects on the optical system and thermal and mechanical wear on the screw. The photoelectric signal is sampled at 5m m increments of plate motion that correspond to wavelength changes of 0.0025 nm. Spectral lines can be located with a reproducibility of ±0.005 nm. During a scan of the spectrum from 230.0 to 470.0 nm, approximately 92,000 transmission measurements are made, corresponding to a sampling rate of 1.3 kHz.
THE MINICOMPUTER CONFIGURATION
The heart of the signal processing system is a Hewlett-PackardR 2100S minicomputer that has 32 kilobytes of random-access memory.
The operator of the microphotometer controls the computer via a model 2615A terminal when recording plates. The portion of computer memory that remains after installation of the real time executive operating system (RTE II) is divided into a small foreground (5 kilobytes) and a large background area (16 kilobytes). Programs are run concurrently in both areas. The data acquisition program runs in the foreground, while data reduction programs are operating in the background. The Hewlett-PackardR 91000A A/D converter uses a differential input when sampling voltages and is gated by a square-wave timing pulse from the shaft encoder. A specially written assembly language driver, which allows a sampling rate up to 40 kHz, controls the operation of the A/D converter. All 92,000 voltage measurements, which constitute a scan, are dumped into a file on the Hewlett-PackardR 7900A 5-megabyte disk. Although the resolution of the A/D converter is 1 part in 10,235, only 1 part in 1,000 is needed for this work.
Within 5 min of recording the last spectrum on a photoplate, the necessary number of two-page report forms listing the concentrations of 64 elements for as many as 10 samples are printed on a model 2607A line printer. Information about effective arc temperature and electron pressure during each arcing (Golightly and others, 1977) and the calculated total oxides (considering only major constituents) for each sample are printed on a separate page. The program next writes an ASCII magnetic tape that contains detailed data on plate emulsion calibration and on each spectral line. These data are processed on an International Business MachineR 370 computer to produce a second tape that is organized for input to the QuantorR 105 microfiche recorder. However, information can also be printed out on the line printer to provide the detailed analysis immediately.
Programs for data collection and interpretation have been written by Walthall (1974) to use on a mainframe computer and by Thomas (1979) to use on the minicomputer system just described.
The two cadmium fiducial lines coexposed on each spectrum, through specially masked windows, are used for wavelength calibration. The computer program searches each of these windows for the first line with a transmittance below some predetermined threshold. The lines thus located are accepted as the cadmium fiducial lines. The number of readings (transmission measurements) between these cadmium lines is compared with the expected number of readings; a discrepancy of more than ±1 reading is assumed to correspond to an expansion or contraction of the spectrum due to temperature variation or a change in the optics. The first task, then, is to calculate the estimated positions of iron and analyte lines relative to Fe 233.2 nm, using the dispersion of the spectrograph-microphotometer system in units of the number of percent transmission readings per nanometer. The expected positions of all the other iron or analyte lines are calculated by applying corrections to their estimated positions relative to Fe 233.2 nm.
Both single- and multiple-element standards are arced to provide the working relationships between concentration and spectral line intensity required for the analytical curve for each spectral line of each analyte element. For each analyte line, a series of programs produce a single page of output on which are tabulated intensities of the line in all the spectra on the photoplate. The tabulation includes the percent transmittances of the peak and background of the analyte line and neighboring lines within ±0.0175 nm, the relative intensity of the analytical line, and the concentration. The output also presents an evaluation of the suitability of the line for analyses that is based on the shape and intensity of the line and the number of readings away from the estimated position at which the line was located in each spectrum.
The computer algorithm calculates the coefficients of first- and second-degree polynomials for the analytical curve of the natural logarithm of intensity versus the natural logarithm of concentration. It evaluates the curve for range, goodness of fit, and slope, and suggests a working concentration range for the line, with a lower limit defined by a signal-to-noise ratio of 2.
The analyst evaluates the analytical usefulness of each line on the basis of profile, intensity, and location. The coefficients of the analytical curve are automatically entered into a data base. When the process is complete for each analyte line, the data base is defined as the wavelength table for the system. For each analyte line in wavelength order, the table contains the element symbol, wavelength, line priority, determination limits, and coefficients of the second-degree polynomial that are used to determine the concentration of the analyte in the sample. A list of the spectral lines and determination limits for each line are given in table 5.
SPECTRAL-INTERFERENCE CORRECTIONS
On the basis of elemental concentrations determined for standard reference materials, interferences can be identified and line priorities and concentration ranges adjusted to minimize systematic errors (Thomas, 1979).
Generalized interference treatments are of two types, subtraction or line switching. The subtraction routine makes corrections by subtracting an equivalent concentration contributed by the interfering line. This subtraction uses predetermined coefficients, relating the concentration of interference to apparent concentration of analyte.
The line-switching treatments are of three types. Two are used when the analyte line is hidden by an interference. One type gives no estimate of concentration. The computer then checks the concentration of the interference to be sure that it is high enough to interfere and checks the computerized evaluation of the profile of the line to be sure that the line is obscured. In the second type, if both of the previous conditions are met, the computer algorithm switches to an auxiliary line, which may give a concentration greater than the upper limit of the last interference-free line in the priority scheme. The third line-switching treatment is designed to guard against spurious answers from band lines, scratches, or ghosts. Here, one or two similar, more abundant, elements, such as platinum and palladium for the noble metals series, are monitored. For this example, if platinum or palladium is not reported, the algorithm switches to a less-sensitive line for ruthenium and reports less than its lower limit of determination.
A modified (Walthall, 1974) Churchill two-step procedure is used for emulsion calibration. The recorded spectrum is divided into ten 25-nm segments. Twenty-six iron lines are used to define the preliminary plate emulsion curve in each 25-nm segment. These iron lines and the exposure time for the two-step iron spectrum are selected so that the transmittance regions <20 percent, 20 to 80 percent, and >80 percent are approximately equally represented. The preliminary emulsion calibration curve (natural logarithm of percent transmittance of filtered step versus the natural logarithm of percent transmittance of unfiltered step) is represented by a quadratic equation. The 45° tangent to the curve is used as a starting place for finding the inflection point of the logarithm of relative intensity versus the logarithm of percent transmittance of the final plate emulsion curve. The actual inflection point is defined as the point from which one step of the filter factor on either side produces an equal change on the logarithm percent transmittance scale. The Newton method of successive approximation (Abramowitz, 1964) is used to find this point on the preliminary curve. The inflection point is then translocated to the final curve so that it falls on the line with a 45° slope that passes through the point where the relative intensity equals 106 and the percent transmittance equals 0.1 percent.
Two other computer programs detect possible analysis errors. Each photoplate contains spectra from one to three reference materials in addition to spectra from up to 23 samples and an iron two-step calibration spectrum. The reference malerials are selected from 20 well-characterized standards, such as U.S. Geological Survey (USGS) diabase W-2, USGS granite G-2, or National Bureau of Standards (NBS) 1633a coal ash, chosen to represent the variety of materials commonly analyzed. The analyst selects standards that are compositionally similar to the samples on the basis of the description of the samples provided by the scientist requesting the analyses. The computer program recognizes these standards and compares the concentrations determined for Si, Al, Fe, Ca, Mg, Ti, Co, Pb, and Zr with accepted analyte concentrations for these standards. These nine elements include the major rock constituents, have a wide range of volatility in the dc arc, and enable checks for loss of refractory elements that would occur if a molten bead is lost from the anode during the arcing process. If the concentration computed for any of these elements is more than one step away from the accepted concentration, a message is printed comparing the reported value with the accepted value. Another program computes and tabulates total oxides, effective temperature, and electron pressure for each sample. Total concentration of oxides is computed by converting concentrations of the rock-forming elements (Si, Al, Fe, Mg, Ca, Na, K, Ti, P, and Mn) to concentrations of the individual oxides and by summing the results. Ideally, the total concentration of oxides for a silicate rock equals 100 percent.
In the analyses of coal ash, for evaluation of both precision and accuracy of this method, NBS standard reference materials 1633 and 1633a (NBS, 1975, 1979) are routinely included as control samples in the analytical procedure. Table 6 lists the concentrations provided by NBS certificates, the mean concentrations determined in U.S. Geological Survey laboratories over an 11-month period, and the associated relative standard deviations. These data were accumulated during a period that coincides with the analyses of over 2,000 coal-ash samples.
The automated dc arc emission spectrographic analysis of coal ash is a rapid, economical method for evaluating both major and trace-element concentrations. Concentrations determined for coal ash are within the limits of precision and accuracy for which the method was designed. The high relative standard deviations for barium and zirconium indicate the heterogeneity documented for these reference materials (Filby, 1985). The system has the capability for the analysis of 4,000 samples, or 256,000 determinations, per year of effort by a single analyst.
Abramowitz, Milton, 1964, Elementary analytical methods, in Abramowitz, Milton, and Stegun, I.A., eds., Handbook of mathematical functions: National Bureau of Standards applied mathematics series 55, Washington, D.C., U.S. Government Printing Office, chap. 3, p. 18.
Dorrzapf, A.F., Jr., 1973, Spectrochemical computer analysis-- argon-oxygen D.C. arc method for silicate rocks: U.S. Geological Survey Journal of Research, v. 1, no. 5, p. 559-562.
Filby, R.H., Nguyen, Son, Grimm, C.A., Markowski, G.R., Ekambaram, Vanavan, Tanaka, Tsuyoshi, and Grossman, Lawrence, 1985, Evaluation of geochemical standard reference materials for microanalysis: Analytical Chemistry, v. 57, no. 2, p. 551-555.
Golightly, D.W., Dorrzapf, A.F., Jr., and Thomas, C.P., 1977, Sets of spectral lines for spectrographic thermometry and monometry in D.C. arcs of geologic materials: Spectrochimica Acta, v. 32B, p. 313-325.
Helz, A.W., 1964, A gas jet for D.C. arc spectroscopy: U.S. Geological Survey Professional Paper 475-C, p. D176-D178.
_______ 1965, The problem of automatic plate reading and computer interpretation for spectrochemical analysis: U.S. Geological Survey Professional Paper 525-B, p. B160.
_______ 1973, Spectrochemical computer analysis instrumentation: U.S. Geological Survey Journal of Research, v. 1, no. 4, p. 475-487.
Helz, A.W., Walthall, F.G., and Berman, Sol, 1969, Computer analysis of photographed optical emission spectra: Applied Spectroscopy, v. 23, no. 5, p. 508-518.
Myers, A.T., Havens, R.G., and Dunton, P.J., 1961, A spectrochemical method for the semiquantitative analysis of rocks, minerals, and ores: U.S. Geological Survey Bulletin 1084-1, p. 11-1229.
National Bureau of Standards (NBS), 1975, Certificates of analysis, standard reference material 1633: Washington, D.C., Office of Standard Reference Materials, U.S. Department of Commerce.
_______ 1979, Certificate of analysis, standard reference material 1633a: Washington, D.C., Office of Standard Reference Materials, U.S. Department of Commerce.
Shaw, D.M., Wickremasenghe, O., and Yip, C., 1958, A simple device for the spectrochemical analysis of minerals in an inert atmosphere using the Stallwood jet: Spectrochimica Acta, v. 13, no. 3, p. 197-201.
Stallwood, B.J., 1954, Air-cooled electrodes for the spectrochemical analysis of powders: Journal of the Optical Society of America, v. 44, no. 2, p. 171-176.
Thomas, C.P., 1979, A minicomputer-based emission spectrographic analysis system dependent on scanning microphotometry: Applied Spectroscopy, v. 33, no. 6, p. 604-612.
Walthall, F.G., 1974, Spectrochemical computer analysis--Program description: U.S. Geological Survey Journal of Research, v. 2, no. 1, p. 61-71.
[ Contents | Coal Quality CD ]
Created by the EERT WWW Staff.
[an error occurred while processing this directive]