Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Technical Paper The Measurement of Silica Author: S. P. Ellis In-Line Instrumentation Seminar, November 17 - 19, 1993, Clearwater, FL Introduction With the advent of higher and higher pressure boilers and steam generators in nuclear power plants, carryover of impurities has posed a greater risk of corrosion to turbine blades and other downstream piping and equipment. The turbine-generator manufacturers have reacted to this risk by establishing ever-lowering specifications for concentrations of impurities in the feedwater. For many of the impurities, the specifications are at or near the limit of detection of many of the available analytical methods used for monitoring. One such specification is for reactive silica. The term, silica, will be used to denote dissolved, or reactive, silica throughout this paper. The most popular method for the measurement of silica is the heteropoly blue (also called molybdate or molybdenum blue) colorimetric method. This method for silica grew up, so to speak, with the power industry. It is used to detect silica breakthrough from strong base anion resin beds. It is also used to quantitatively determine silica in feedwater, as well as steam and steam condensate samples. Determination of the silica break from a resin bed does not require an analytical test having a low limit of detection. Further, when low pressure boilers are in question, easily detectable quantities of silica (10 - 20 ppb, or even higher) are not only tolerated, but are often considered to be desirable. Nevertheless, for higher pressure systems, silica at concentrations above 5 ppb is considered to be intolerable. All owners and operators of ion exchange deionization equipment are highly interested in the accuracy and precision of methods used for measuring silica. Failure of the equipment to produce water in conformance with prescribed specifications is, of course, quite costly. This is even more true for mobile water treatment companies, since transportation costs are often very high. Thus, this research was approached from the standpoint of trying to define the variability in the silica measurement. In so doing, it might be possible to work out acceptance criteria which would allow equipment to operate, at least for a period of time, without imposing contractual penalties. In some cases this is already being done in some form. At the same time, the dilemma of the power producer is also recognized. The status of the warranty for the turbine-generator is not trivial by any means, it remains absolutely necessary to do all that one can to ensure that conformance to the vendor’s specifications is maintained. Therefore, the power producer must use some measure to demonstrate the quality of the feedwater and the makeup water to the boiler, even if there is some question concerning precision and bias of the test. Numerous laboratory instruments using 50 mm or 100 mm cells have been used to obtain silica measurements at or below 5 ppb. Further, in-line instrument manufacturers report obtaining values as low as 0.5 ppb in their marketing literature. At least one manufacturer reports values of 3 ppb or less with both laboratory and in-line equipment using cell pathlengths of 1-inch1. Error in in-line silica measurements were a part of the EPRI RP 2712 study2. The results of that study indicated a sizeable error in both precision and bias. This paper describes the initial attempts to determine the degree of error involved in the measurement of silica, both by laboratory and inline Find a contact near you by visiting www.ge.com/water and clicking on “Contact Us”. * Trademark of General Electric Company; may be registered in one or more countries. ©2009, General Electric Company. All rights reserved. TP1055EN.doc Jun-09 instrumentation. This paper is largely introductory, and somewhat tutorial. The results of laboratory studies which were designed to quantify error in the measurements at low concentrations are presented. Results of comparisons of various methods of determining blank values, as well as some data concerning the effects of the amount of reagent, and elapsed times prior to reading the absorbance after addition of all reagents are also given. The absorption spectrum for the silica-molybdenum blue complex is shown in Figure 1. The most important feature of the spectrum is the intense, broad absorption band centered at about 815 nm. Because of this broad band, lower resolution spectrophotometers and even colorimeters may be used successfully. That is, an instrument with very low slit widths is unnecessary. Experimental Silica measurements were made using a Shimadzu Model UV-1201 UV/VIS spectrophotometer, and a Bausch & Lomb Spectronic 20 spectrophotometer (now owned and marketed by Milton-Roy). The latter instrument was factory-modified by replacing the standard focusing lens with a red lens and installing a nonstandard detector to make it more sensitive at 815 nm, the wavelength used for measurement. Additional electronic modifications were installed by Update Instruments to provide more stability. Reagents used in the test include an acidic ammonium molybdate solution, an oxalic acid solution and an amino acid solution. All of these solutions were obtained from Hach Chemical Co. in Loveland CO. A 1 ppm working standard solution of silica was also obtained from Hach. Solutions for analysis were generated by diluting the stock standard solution with deionized water to produce solutions whose concentrations ranged from zero to 100 ppb silica added. Each analytical solution was prepared 10 times and each of these solutions was analyzed once. A standard curve was prepared to compare the results of the tests and help illustrate the error in the measurements. Background Before presenting the analytical results, it is worthwhile to revisit some of the spectroscopic considerations which go into the colorimetric analysis of silica. These considerations include the absorption spectrum and the relationship between concentration and response (Absorbance). Sources of errors in measurements, the relationship of those errors to detection limits, and the ability of the instrument to detect small (0.5 to 2 ppb) differences in concentration are also generically discussed. Page 2 Figure 1: Absorption Spectrum of Heteropoly Blue Complex The absorption of light may be used to quantify the amount of an analyte in solution. A set of sol utions containing known amounts of the analyte in question are prepared and analyzed at the appropriate wavelength. The relationship between the concentration of an analyte in solution and the absorbance of the solution is given by the BeerLambert equation: A = abc where, A = absorbance a = absorption coefficient b = cell pathlength c = concentration For a given analyte, the absorption coefficient is constant. Therefore, the slope of the curve, which is theoretically a straight line, is dependent entirely upon the cell pathlength. The effect of this relationship is shown in Figure 2. TP1055EN Figure 2: A vs. Conc. Figure 3: A vs. Conc. with Error While any standard curve will theoretically produce results, the errors, which are associated with all measurements, may prevent determinations at low concentrations or discrimination between two similar concentrations. If errors in the measurement of the analyte are considered, a curve such as that shown in Figure 3 may be produced. It should be noted that the error bars are sufficiently high in this case to make adjacent points statistically indistinguishable. Indeed, sets of points even further away from any given point may be indistinguishable from one another. Obviously, this type of situation can overcome any theoretically derived limit of detection or ability of the test to discriminate between two very similar concentrations of analyte. These values may be contrasted with absorbance values calculated from the efficiency data of modern detectors. A good estimate of that efficiency is that they can detect increments of about 0.1% transmittance4,5. Thus, using a value of 99.9%T, and converting that value to absorbance (A = log (100/%T)), one obtains an absorbance of about 0.0004 in a 10 mm cell. This value translates to 0.001 for a 1-inch cell and 0.004 for a 100 mm cell. It will be noted that these values are quite close to those calculated for 1 ppb silica solutions from actual experimental data. It can, then, be assumed that the quality of the detector is theoretically sufficient to measure silica at the 1 ppb level of concentration. Actual absorbance values for the concentrations of silica may now be considered. Although spectral efficiency data were not located, the absorbance at low ppb concentrations can be calculated from available data, if conformance to the Beer-Lambert equation is assumed. The values of expected absorbances of 1 ppb solutions using 1-inch and 100 mm path cells from two independent sources are presented in Table 1. The sources of error in optical instrumental analysis arise from three main sources; chemical, spectroscopic/optical and electronic. In practice, the latter two categories of error are not so easily separated. The electronics that power a lamp, for example, can be made more stable which, in turn, makes the lamp more stable. Historically, however, the instrumental sources of error are commonly related to lamp instability, the sample itself, detector instability, optics and reagent impurities. Table 1: Expected Absorbances of a 1ppb Silica Solution Most of the time, the operator must use the instrument as it is purchased. He may not have the ability or the time to modify and improve the instrument. The operator must, therefore, spend his time on activities that he can control, such as TP1055EN Page 3 reagent purity, sample integrity and careful conduct of the analytical procedure. He must also know the limitations of the instrumentation, which he operates, in order to make appropriate judgements concerning data quality. Results and Discussion In an effort to quantify the error associated with the points used to construct a standard curve for a laboratory instrument, it was decided to first try to incorporate as much of the error that a single operator could make. Therefore, ten standards at each concentration were prepared and each solution was measured once to attain some degree of statistical significance. The mean values of these standards, measured in both a 1-inch cylindrical cuvette and a 100 mm rectangular cell, were plotted versus absorbance. The results are shown in Figure 4. minations of single solutions at each concentration should diminish the errors significantly. Another approach could be to standardize at a high concentration at which the standard deviation would be expected to be lower. This action would minimize the error at the upper end of the curve. However, a good standardization would still rely on a low error about the zero, which is not necessarily true. In order to study the error associated with the blank (or Zero) measurement, several methods for the colorimetric determination of silica were investigated. These methods came from power plants, ASTM, Standard Methods and Hach Chemical Co. All, of course, were similar. But almost all had small deviations from one another; particularly in the handling of the blanks. Five methods of handling the blank were investigated. 1. Reagents 1, 2 and 3 were added to deionized water, in order, with 10 min. and 2 min. reaction times allowed between reagent additions. Five minutes was allowed between the addition of reagent 3 and measurement. 2. Reagent 2 was added to a 1:1 solution of concentrated hydrochloric acid in deionized water. No molybdate or amino acid reducing reagents were added. 3. Reagents 1, 2 and 3 were added to deionized water, but not in order. The reagents were added in quick succession, starting with reagent 2, then 1 and then 3. 4. Reagents 1 and 2 were added to deionized water in quick succession, and allowed to react for 10 min. before measurement. 5. Reagents 1 and 2 were added together and allowed to react for 2 min. Then reagent 3 was added and the solution was allowed to react an additional 5 min. Finally, deionized water was added and the measurement was taken. Figure 4: SiO2 Standard Curves Clearly, the errors shown in this study are quite large in comparison to those which could be obtained by “normal” standardization techniques. Simply using the usual approach of multiple deterPage 4 In each case, the spectrophotometer was zeroed using deionized water, and the measurement made. Measurements were made on ten different solutions of each blanking method, and each solution was analyzed once. The results of these tests are presented in Table 2. TP1055EN Table 2: Comparison of Blank Methods Some of the silica methods which were considered warned against waiting too long to measure the absorbance after the color had developed. Therefore, a quick test was run to see if time delays were a problem. A 50 ppb silica standard was prepared and analyzed using 10 min. and 2 min. reaction times between reagent additions. The test solution was measured immediately after mixing with the final reagent, and at several intervals thereafter. The results of this test are shown in Table 3. Table 3. Time Effects on Absorbance Conclusions Laboratory measurements of silica are, as all measurements, beset by errors. If these errors are not recognized and dealt with, results can be misleading, at best. We unfortunately have a tendency to believe a pretty digital readout or a rock steady meter when we should remain skeptical. Results of the testing described above lead to the following conclusions. 1. Modern instruments are theoretically capable of determining 1 ppb silica in a cell with as short a pathlength as 10 mm. 2. Errors in measurement raise this limit to as high as 20 ppb in a 1-inch pathlength cell. 3. The use of a 100 mm pathlength cell lowers the limit of detection to about 3 ppb, even though the magnitudes of the errors associated with the development of the standard curves are similar to those for the standard curve using a 1-inch pathlength cell. TP1055EN 4. Modifications in the way the standard curves are developed should decrease the magnitudes of the errors. This should also decrease the limit of detection somewhat. 5. The errors associated with the zero, or blank, determination are of about the same magnitude as the other standard concentrations used in this study. Therefore, they cannot be ignored. 6. The time between the addition of the last reagent and reading the absorbance does not appear to be important from about 5 minutes to at least 45 minutes. The experiments described in this paper were designed to introduce a rather high amount of error into the determination of silica. Operator error is also high in manual methods compared to welloperating automatic systems. While it may be assumed that in-line instruments have less error associated with a measurement, error still exists. Such error may be sufficient to disallow the measurement of 1 ppb or sub-ppb concentrations of silica. In future work, the emphasis will be shifted to the study of response and error associated with in-line instruments. Of particular importance to us is the ability of the instruments to respond to small increments in concentration and the ability to discriminate between close concentrations. References 1. K. R. Doerr, D. Harp and L. Liou, Low Level Silica Verification for Analyzer Users, Ultrapure Water, July/August, 20-25, 1991. 2. J. K. Rice, D. M. Sopocy, R. B. Dooley, Quantification of Continuous Instrument Error, 1990 International Conference on Waterborne Trace Substances, August 1990, Baltimore, MD. 3. A. B. Carlson and C. V. Banks, Spectrophotometric Determination of Silicon, Analytical Chemistry, 24(3), 472-477, 1952. 4. R. E. Hanson and M. V. Buell, Photoelectric Technique for Spectrophotometry Between 194 and 225 Millimicrons, Analytical Chemistry, 31(5), 878-81, 1959. 5. J. Salpeter, Personal Communication, October 1993. Page 5