2 RESEARCH AND DEVELOPMENT TOPICS Anal. Pvoc. Research and Development Topics in Analytical Chemistry The following are summaries of six of the papers presented at the Research and Development Topics in Analytical Chemistry meeting of the Analytical Division held on July 4th and 5th, 1979, at the University of Edinburgh. A further paper, by Aziz-Alrahman and Headridge, appeared in full in the October 1979 issue of The Analyst (p. 944). Summaries of a further six papers will appear in a future issue of Analytical Proceedings. Use of a Piezoelectric Sensor as a Continuous Monitor of Atmos- pheric Pollutants S. Cooke and T. S. West and P. Watts The Macaulay Institute for Soil Research, Craigiebuckler, Aberdeen, A B9 2Q J Ministry of Defence, Porton Down, Wiltshire The decrease in frequency that is caused by deposition of a small mass of material on the surface of a piezoelectric crystal oscillator has been known qualitatively for some time and is used regularly in order to obtain fine control of oscillator frequencies.Mathematical deriva- tions of the relationship between frequency and added mass have been made by Lostis,l Sauerbrey2 and St~ckbridge.~ The Sauerbrey derivation is most frequently used : -Am x F A x d x t AF =January, 1980 RESEARCH AND DEVELOPMENT TOPICS 3 where AF is the frequency change, F the initial frequency, Am the added mass, A the area of the deposit, d the density of quartz and t the thickness of the crystal. This linear relationship between frequency change and added mass enables a crystaJ oscillator to be used as a microbalance device with a theoretical detection limit of 10-12 g.2 The quartz-crystal device has been used as a moisture detector by coating the crystal with a hygroscopic compound and relating the mass of adsorbed water to the atmospheric water ~ o n t e n t .~ Subsequently, piezoelectric crystals coated with various adsorbant materials have been reported as sensors for other gases, including sulphur d i ~ x i d e , ~ - ~ nitrogen dioxidelot1l and ammonia.12 Methods reported so far, however, are concerned with the analysis of discrete gas samples and show little or no advantage over conventional techniques that employ liquid absorbants to preconcentrate the sample before analysis, e.g., the West - Gaeke procedure for sulphur dioxide.13 It was decided, therefore, to investigate the potential of the piezoelectric sorption detector as a continuous monitor of trace atmospheric constituents.This moisture detector was later developed into a commercial instrument .5 Equipment The detector element was a 9-MHz, AT-cut, quartz crystal (10 x 10 x 0.18 mm) (Quartz Crystal Co. Ltd., New Malden, Surrey), vibrating in the thickness shear mode. The frequency of the crystal was monitored by use of a digital counter, the output of which was converted to analogue form and displayed on a chart recorder. A gas-mixing system was used in order to simulate the low analyte concentrations required to assess the performance of the detector. Calibrated gas mixtures (Rank Hilger Ltd.) containing approximately 50 p.p.m. (V/V) of analyte gas were diluted in two stages to the re- quired levels.The water content of the diluent gas was continuously variable and the final water content of the mixture was monitored by means of a capacitance hygrometer (Shaw Moisture Meters Ltd., Model SH). This monitoring was necessary as water was known to be involved in the reaction between sulphur dioxide and amines and was, therefore, likely to affect the detector response. The detector crystal was housed in a stainless-steel double- impinger cell similar to that used by Karmarkar and G~ilbault.~ This design has been shown to provide the best contact between the detector coating and the gas stream and hence to give the greatest s e n ~ i t i v i t y ~ , ~ ~ ; however, with this arrangement volatile coating materials tend to bleed much more quickly from the crystal surface than with other cell designs.14 Detector Performance The initial investigation of the sorption detector as a gas sensor was conducted using sulphur dioxide as the analyte gas.A brief examination of several possible coating materials was made and triethanolamine was found to be the most sensitive. The triethanolamine detector responded to a sample of sulphur dioxide with an initial steep decrease in frequency followed by a gradual approach to equilibrium. This signal was superimposed on a base line of steadily increasing frequency (typically 2-3 Hz min-l). The base line drift, caused by the bleed of coating material from the surface of the crystal, was not completely regular and proved to be the main factor in determining the system's detection limit.Detector sensitivity The mass of sulphur dioxide that is adsorbed on to the detector coating can easily be calcu- lated from the Sauerbrey equation. The rate at which sulphur dioxide is adsorbed can, therefore, be calculated from the slope of the frequency - time profile over the first 2-3 min of exposure (before equilibrium is approached). The rates of adsorption calculated for 1.2 p.p.m. (Y/V) of sulphur dioxide passing over the detector at 50 ml min-l demonstrated that a large proportion (80-90%) of the sulphur dioxide passing into the cell during this initial pre- equilibrium stage was adsorbed. These results were checked by observing the response of two matched detector cells placed in series in the gas line.The second detector was exposed only to the sulphur dioxide that was not retained on the first. A comparison of the initial frequency slope for the two detectors again showed that an adsorption efficiency of 80-90% was obtained. Thus, in terms of sensitivity alone there is only a limited scope for further improvement in detector performance.4 RESEARCH AND DEVELOPMENT TOPICS Anal. Proc. Variation of response with coating mass The detector response increases linearly with the mass of adsorbant material but, as the coating bleed rate and the detector response time also increase with coating mass, the best detection limit is generally obtained with a light coating. A coating of approximately 20 pg of triethanolamine (equivalent to a frequency decrease on coating of about 12 kHz) was deposited on each side of the crystal; smaller amounts of coating were difficult to apply repro- ducibly .Variation of response with detector temperature The detector temperature was varied by immersing the cell in a thermostatic water-bath. An increase in temperature was found to decrease the equilibrium response by a considerable amount; the response at 40 "C was less than a quarter of that at 20 "C but, on the other hand, the response time was greatly improved so that only 30 min were required for equilibrium at 40 "C compared with 70 min at 20 "C. Detector calibration graphs Calibration graphs were obtained over the range 0.01-140 p.p.m. (V/V) of sulphur dioxide at detector temperatures of 20 and 30 "C. The response in each instance was linear up to about 10% of the coating saturation level, after which negative deviations from linearity were observed.Coating saturation occurred after a response of approximately 5.5 kHz. Within the linear range, the detector showed greater sensitivity at 20 than 30°C for low sulphur dioxide concentrations, but the temperature effect was much less severe at concentrations near to saturation. The linear range was 0.1-0.4 p.p.m. (V/V) at 20 "C and 0.05-2.0 p.p.m. (V/V) at 30 "C; the response at the top of the range was about 500 Hz in each instance. Membrane shielded detector cells It was thought that the high coating bleed rate associated with the impinger cell might be alleviated by the introduction of a gas-permeable membrane between the impinger gas jet and the detector element.A membrane (Celanese Plastics Ltd., Celgard 2500) was, therefore, introduced into the cell, and although this reduced the bleed rate by a factor of ten, the detector response time was increased to well over 2 h. Conclusions The potential of the piezoelectric sorption detector as a continuous gas monitor has been clearly shown. The electronic system is capable of resolving &O.l Hz, which is equivalent to a sulphur dioxide concentration of 0.000 1 p.p.m. (V/V) at a detector temperature of 20 "C. This performance can only be realised, however, if a solution to the problem of coating bleed is found. The minimum measurable frequency change of a triethanolamine-coated crystal is limited by coating bleed to about 10 Hz (which is equivalent to 0.01 p.p.m.of sulphur dioxide at 20 "C). One solution to this problem might be to have the active adsorbant sites situated in a polymer matrix. Alternatively, a search might be made for efficient involatile compounds, such as some silver salts that have recently been shown to act as efficient adsorbants for ammonia gas.15 Although an uncoated AT-cut crystal is not particularly prone to changes in temperature, it is advisable to maintain the temperature within &1 "C during a series of measurements. The massive stainless-steel block device provides reasonable stability under normal laboratory conditions and its temperature can easily be thermostatically con trolled. The other major problem with the sorption detector is that the relatively slow response time results in considerable smoothing of a continuous concentration versus time graph.There are, however, several ways in which the response time might be decreased: (i), by raising the temperature; (ii), by decreasing the mass of the coating material; or (iii), by increasing the gas flow-r ate. These procedures all have potential disadvantages: (i) and (iii) increase the coating bleed rate and (i) and (ii) reduce the absolute detector sensitivity, although with (ii) this dis- advantage can be offset by a decreased bleed rate so that a decrease in coating mass should not adversely affect the detection limit. Careful optimisation of all experimental conditions withJanuary, 1980 RESEARCH AND DEVELOPMENT TOPICS 5 regard to the tendency of a particular coating material to bleed, and the relative importance of sensitivity and speed for a particular application should, however, enable some improvement to be made, particularly if more stable adsorbant materials can be found.It is in this area of coating materials that most further work should be concentrated. We are grateful to the Ministry of Defence for the award of a studentship to S.C. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. References Lostis, M. P., J . Phys. Radium, 1959, 20, 25 s. Sauerbrey, G., 2. Phys., 1959, 155, 206. Stockbridge, C. D., “Vacuum Microbalance Techniques,” Volume 5, Plenum Press, New York, 1963, King, W. H., Anal. Chem., 1964, 36, 1735. King, W. H., U.S. Pat., 1969, No. 3, 427,864. Hartigan, M. J., PhD Thesis, University of Rhode Island, 1971. Guilbault, G.G., and Lopez-Roman, A., Environ. Lett., 1971, 2, 35. Guilbault, G. G., and Lopez-Roman, A., Anal. Lett., 1972, 5 , 225. Karmarkar, K. H., and Guilbault, G. G., Anal. Chim. Acta, 1974, 71, 419. Street, D. C., and West, T. S., Unpublished results. Karmarkar, K. H., and Guilbault, G. G., Anal. Chim. Acta, 1975, 75, 111. Webber, L. M., and Guilbault, G. G., AnaE. Chem., 1976, 48, 2244. West, P. W., and Gaeke, G. C., Anal. Chem., 1956, 28, 1916. Cooke, S., PhD Thesis, University of Aberdeen, 1979. Glass, E. B.-I., Edmonds, T. E., and West, T. S., Work in progress. p. 193. Pollution Studies in the Clyde Sea Area Jane L. Smith- Briggs Depavtment of Chemistry, University of Glasgow, Glasgow, G12 8QQ The Clyde Sea Area extends northwards of a line along 55”15’N, that is, from Finnart Point to the Mull of Kintyre.It has a total area of 2 500 km2 and receives the wastes and effluents of 2.5 million residents. The main contributors to pollution of the Clyde Sea Area are domestic sewage, industrial effluents and the dumping of sewage sludge and dredge spoils. There are inputs of treated, untreated and partially treated sewage in the upper and lower estuaries and along the Ayrshire coast. Industrial effluent releases come from a variety of manufacturing processes, for example, nylon plant, distillery and oil refinery wastes. Some 1.4 x 106 tonnes per year of wet sludge from sewage and industry are dumped in deep water south of Garroch Head; also 3 x 105 tonnes of wet sediment are dredged annually from the Clyde navigation channel and dumped near the mouth of the Holy Loch.This research is designed to assess the effects of these pollutants on the natural sedimentary rbgimes. It consists of two parts, therefore : firstly, to investigate unpolluted sediments in terms of sedimentation rates and trace metal concentrations; and secondly, to study, in the same terms, areas considered to be environmentally endangered. The “polluted” areas chosen for study were Irvine Bay (adjacent to the ICI Ardeer plant), Garroch Head (which receives sewage sludge) and the mouth of the Holy Loch (receiving the dredge spoil). The Arran Deep and the south of Loch Long were selected as the natural sites. Recent sedimentation rates are derived by using both the 210Pb dating methodl and by observing the 13’Cs and 134Cs concentrations in sediment profiles.Levels of the latter species in the Clyde area are dominated by the output and subsequent transport of radiocaesium from the British Nuclear Fuels Ltd. fuel reprocessing plant at Windscale in Cumbria. Therefore, the first appearance of radiocaesium at depth in the sediment profile corresponds approximately to the horizon deposited immediately after the initial output from Windscale in 1958. Pre- industrial sedimentation rates are obtained by using 14C dating and palaeomagnetic methods. The trace metals under study are copper, nickel, zinc, iron, manganese, chromium, lead and aluminium, all except the last being measured by flame atomic-absorption spectroscopy. Aluminium is determined by neutron-activation analysis.6 RESEARCH AND DEVELOPMENT TOPICS Anal.PYOC. The sediment samples were retrieved by using three coring devices, gravity, box and craib corers. The cores were divided into 1-cm sections, using plastic implements to avoid con- tamination. The sediment was weighed while wet and then dried, at 80 “C, to a constant mass. It was subsequently homogenised with a mortar and pestle. Sediment samples of 3-5g are taken for the 210Pb and trace-metal determinations. For both procedures the sediment has to be totally dissolved. This dissolution is achieved by a series of strong acid leaches using hydrochloric, nitric and hydrofluoric acids. The leachates are combined and made up to a total of 500 ml. A spike of 208Po is added to the sediment initially so that an extraction and detection efficiency for 210Po can be assessed; 25ml of leachate are then taken for tcace metal analysis, 100 ml for 210Po and 250 ml for 226Ra analysis.The 210Pb dating method arises from the fact that, within the 238U natural decay series, a precursor of 210Pb is the gaseous nuclide 222Rn. Because of its mobility as a gas, some 222Rn escapes from continental rocks into the atmosphere, there decaying with its 3.8-d half-life to form 210Pb. The 210Pb in particulate form is subsequently rained out on to the land and surface waters where it decays with its 22 year half-life. In the sediment, therefore, there are two types of 2loPb including “supported” 210Pb, which is in secular equilibrium with the 238U activity of sedimentary minerals, and is decaying with the half-life of 238U (4.5 x 109 years). There is also, however, the “unsupported” or excess 210Pb component precipitated from the atmosphere.If we can measure the excess 2loPb in a sediment profile we can determine the rate of sediment accumulation from the rate of decrease of excess 210Pb activity. The excess 2loPb concentration can be obtained by determining the activity of supported 210Pb and the total activity of 210Pb present. The former is measured by determining the 226Ra concentration, and the latter by measuring the concentration of 21OPo.2 2lOPo is the granddaughter of 210Pb and, with a half-life of 148 d, will be in secular equili- brium with all of the 21°Pb present. The analytical procedure is based on the fact that under certain conditions of pH and temperature polonium spontaneously deposits on to silver.The 100-ml volume solution of dissolved sediment is reduced by heating to approximately 15 ml. Hydroxylammonium chloride is added, the pH adjusted to 2 and the solution heated to 85 “C. A silver disc (2.5 cm in diameter) is introduced, precautions being taken so that only one side is available to polonium, and the solution stirred at 85 “C for 4 h (see Fig. 1). The sections were sealed in polythene bags and frozen until used. This excess 210Pb decays with its own 22 year half-life. Magnetic stirrer Fig. 1. Polonium plating cell. All dimensions are in Subsequently, the disc is a-counted by using a surface barrier detector, amplifier and multi- As 208Po is deposited and counted with the same efficiency as The proximities of the centimetres.channel analyser assembly. 210Po, the concentration of 210Po in the sample can be determined.January, 1980 RESEARCH AND DEVELOPMENT TOPICS 7 2O*Po and 210Po cc energies (5.1 1 and 5.308 MeV, respectively) make high-resolution counting essential. It is therefore reasonably assumed to be in secular equilibrium with the “supported” 210Pb com- ponent. Radium-226 is measured by sealing 250 ml of dissolved sediment in an equilibration flask and allowing 222Rn to grow in, with its 3.8-d half-life to secular equilibrium, The radon is then extracted by flushing with helium and transferred to an a-scintillation detector (see Fig. 2).3 The detector is comprised of a hollow sphere coated with zinc sulphide. The a- particles emitted by 2z2Rn and its daughter interact with the zinc sulphide to produce photons, which are subsequently counted on a photomultiplier and scalar assembly.Radium-226 is a precursor to both 222Rn and 210Pb in the 238U natural decay series. Helium manometer 1 lid COP - acetone traps spiral trap Fig. 2. 222Rn emanation system. Flame atomic-absorption analysis is used to determine iron, manganese, copper, chromium, zinc, nickel and lead in the 25-ml aliquot of dissolved sediment. The equipment used is an Instruments Laboratory 151, with background correction via a deuterium lamp. The samples are measured along with standards prepared in the same matrix. Aluminium is measured by means of neutron-activation analysis. Approximately 0.2 g of dried sediment is placed inside a small plastic vial, which is subsequently marked and flame sealed. This vial, and a similar one containing a standard, are irradiated together for 60 s and counted immediately after for 2 min on an 80 cc Ge(Li) detector. The neutron-activated product, 28A1, has a half-life of 2.3 min, 27Al(n,y)28A1, with a y decay mode at 1.78 MeV.Because both sample and standard have identical geometries and have been exposed to the same neutron-flux, etc., the concentration of aluminium in the sample is derived directly from a comparison of the activities and the amount of aluminium in the standard. The analytical procedure for radiocaesium does not involve any chemical pre-treatment . A known amount of dried and ground sediment is placed in a smdl sample bottle and y- counted on a 100 cc Ge(Li) detector for a period of (usually) 800 min.Caesium-137 (t+ = 30.0 years) decays via 2/3- modes to form 137Bam (t+ = 2.55 min). Barium-137m decays by y-emission and, because it is in secular equilibrium with 137Cs, the activity of the observed y radiation is equal to the activity of lS7Cs. Caesium-134 (it = 2.05 years) has 3/3- and 9y decay modes. The efficiency of the system is determined by counting known radiocaesium concentrations. References 1. 2. 3. Koide, M., Soutar, A., and Goldberg, E. D., Earth Planetary Sci. Lett., 1972, 14, 442. Flynn, W. W., Anal. Chim. Ada, 1968, 43, 221. Ferrante, E. R., Gourski, E., and Boulenger, R. P., in Adams and Lowder, Editovs, “The Natural Radiation Environment,” Chicago University Press, Chicago, 1964, p.353.8 RESEARCH AND DEVELOPMENT TOPICS Anal. Proc. Concentration Procedures for Trace Element Analysis by Spark Source Mass Spectrometry Karen H. Welch and Allan M. Ure The Macaulay Institute for Soil Research, Craigiebuckler, Aberdeen, A B9 2Q J , Scotland Spark source mass spectrometry (SSMS) is a sensitive, multi-element technique which has been applied at The Macaulay Institute for Soil Research to the analysis of soils, rocks, plants and biological material~l-~ The technique used has been described in and employs an AEI MS702R mass spectrometer with a hydraulically pressed electrode; this electrode is prepared from a mixture of 25 mg of finely ground (particle size less than 150 pm) sample and 25 mg of pure aluminium powder (particle size less than 70 pm) containing iridium and indium as the internal standard elements.The mass spectrum obtained by sparking the sample electrode against a pure aluminium counter-electrode is recorded on an Ilford Q2 photographic plate from which element contents are found after evaluation by microdensitometry. Although SSMS has a uniformly high sensitivity for most elements, typically 0.01-1 p.p.m. with aluminium metal as the conducting material, some elements, such as the noble metals, have abundances in natural samples that are too low for direct analysis. For this reason alone pre-concentration procedures are necessary. These elements, together with copper and silver, because of their similar properties, form the first group of elements for which a common group concentration procedure is sought.For most of the elements in a second group, arsenic, bismuth, lead, antimony, selenium, tin, tellurium and thallium, which are of geochemical, environmental and biological interest, a group concentration procedure is required, not only on the grounds of sensitivity, but also to obtain freedom from the interference effects caused by matrix compounds. This is the case in the determination of selenium by SSMS in liver ash. Two spectra for simulated liver ash are shown in Fig. 1 (a) and (b) ; one was obtained before and one after the application of a cemen- tation concentration procedure discussed below. This concentration procedure is successful in removing elements forming molecular ionic species, which overlap the selenium (and arsenic) isotopic lines required for analysis.For SSMS analysis, the sample must be a conducting solid and the concentrate should, therefore, ideally take this form. This criterion restricts the method of concentration to procedures such as co-precipitation, electrochemical and cementation techniques or adsorption on activated carbon. Of these methods the application of a co-precipitation method is des- cribed and the development of a cementation procedure discussed. Electrochemical methods of deposition have not been attempted although they have considerable potential, especially for single element determination. Collection by activated carbon was not considered because of the interference effects arising in the spectra from carbon species.6 Concentration by Co-precipitation The particular co-precipit ation technique used was chosen because it was already established and in use for the trace-element analysis of soil extracts and plant material by optical-emission spectrographic methods and considerable experience in its use was available at the Macaulay Institute.The main objections to the method are firstly, that it is a relatively slow laboratory procedure, and secondly that it requires an ashing stage in order to remove the organic reagents before SSMS analysis and this may lead to the loss of volatile elements. I t does, however, provide high concentration factors (up to 400-fold) for a large group of elements of biological and geological importance, for many of which the method has been shown to be quantitative.'^^ The concentration procedure is outlined in Fig. 2.This technique was applied to acetic acid extracts9 of various soils and, in Table I, typical results obtained by SSMS analysis of the concentrates are compared with d.c.-arc spectro- graphic analysis of the concentrates and, for copper, zinc and cadmium, with direct analysis of the soil extracts by atomic-absorption spectrometry. As can be seen from Table I, good agreement was obtained between the SSMS analysis and the other techniques for this group of biologically important elements. Other elements determined by SSMS in concentrates from extracts of soils in acetic acid and likely to be collected quantitively by this procedure include arsenic, niobium, selenium, thorium, uranium and tungsten. The acetic-acid extractableJanuary, 1980 RESEARCH AND DEVELOPMENT TOPICS 9 'I 81 Fig.1. Interfering ions 40Ca3SC1+ a t m/z 75, 30K37C1+ and 41K36C1+ a t m/z 76, 40Ca37Cl+ and 42Ca36C1+ a t m/z 77, soK,+ at m/z 78 and 40Ca,+ and 3QK41K+ a t m/z 80, shown in the spec- trogram of a simulated liver ash ana- lysed directly Fig. l(a), are absent in the spectrogram from a concentrate [Fig. l(b)] prepared by a cementation procedure. The line a t m/z 81 in (a) and (b) is due to 27A13+. I I Evaporate r'l HCL solution Precipitate (pH 5.2) with quinolin-8-olI 30-40 mg ash Fig. 2. Concentration by co-precipi- tation with aluminium using quinolin-8- 01 - tannic acid - thionalide. Phosphate, alkalis and alkaline earths left in solution. TABLE I ACETIC ACID (0.5 N) EXTRACTABLE CONTENTS OF TOPSOIL BY SSMS* AND OTHER^ METHODS Micrograms per gram of oven-dried soil.Soil A Soil B w v 7 Element SSMS Other? SSMS Other? Cd co Cr c u Mo Ni Pb Sn V Zn 0.12 0.45 0.33 0.46 0.01 0.38 0.38 0.04 1.3 2.9 0.08 0.47 0.28 0.39 <0.03 0.38 <0.5 (0.5 1.3 2.0 0.03 0.08 0.29 0.13 (0.01 0.21 0.20 (0.02 1.1 1.4 0.05 0.07 0.19 0.20 <0.03 0.20 €0.3 ~ 0 . 5 0.94 1.2 *SSMS analysis of concentrate. TSpectrographic analysis of concentrate or atomic-absorption spectrometric analysis of soil extract.10 RESEARCH AND DEVELOPMENT TOPICS Anal. Proc. contents of these element in the concentrates, and the total contents in the soil, both deter- mined by SSMS, are shown in Table I1 for two soils. TABLE I1 TOTAL SOIL CONTENTS DETERMINED DIRECTLY AND ACETIC ACID EXTRACTABLE SOIL CONTENTS DETERMINED IN CONCENTRATES BY SSMS FOR SOME ELEMENTS THAT ARE LIKELY TO BE EFFECTIVELY COLLECTED BY THE QUINOLIN-S-OL - TANNIC ACID - THIONALIDE CONCENTRATION PROCEDURE Micrograms per gram of oven-dried soil. Soil A Soil B -- Total Extractable Total Extractable Element content content content content As 2.9 0.01 2.4 0.003 Nb 20 0.01 17 0.02 Se 0.28 0.004 0.10 0.003 Th 4.5 0.06 5.9 0.06 U 2.0 0.07 1.9 0.02 w 0.76 0.005 ~ 0 .5 <0.003 Development of a Method for Concentration by Cementation The general equation for a cementation reaction can be represented as follows: %A*+ + yB (solid) = XA (solid) + yBG+ where B is a metal less noble than A. Aluminium metal has been used industrially as a cementant in solution-purification processes.lO In the cementation, for example, of gold from solution by aluminium, the aluminium reduces the ionic goId to elemental gold, which is deposited on to the aluminium surface.In theory, aluminium should reduce all those elements which lie below it in the electrochemical series and interest lay in determining whether the aluminium powder used to make sample electrodes for SSMS could function as a cementant to collect groups of elements from solution. This would provide a simple procedure for direct collection on to the electrode material and thus minimise contamination problems arising from elaborate chemical pre-treatments. Various methods of mixing aluminium powder with solutions were investigated but in most of these methods it proved difficult to collect the aluminium powder after cementation.This problem was largely overcome by placing 50 mg of aluminium powder in a short, narrow PTFE column, and by drawing the sample solution through the column three times in suc- cession by gentle suction from a water-pump. Three passes provided optimum recovery. The apparatus is shown schematically in Fig. 3. Fig. 3. Schematic diagram of cementation apparatus : X, sample solution reservoir (100 ml); Y, cementation column (3 mm diameter, 6 mm length) of aluminium powder in a PTFE tube; 2, water-operated vacuum pump.January, 1980 RESEARCH AND DEVELOPMENT TOPICS 11 Most cementation procedures have been carried out in acidic media, generally in hydro- thloric acid. Acid strength is one of the factors that is known to affect the rate of the cemen- cation reactionll and its effect varies from element to element.For example, for palladium, reduction has been shown to be effected best in acidic solutions, whereas for platinum, better results are achieved in alkaline solutions.12 Because, however, of the high solubility of finely powdered (less than 70 pm) aluminium powder in hydrochloric acid at strengths greater than 0.1 N, and because a group concentration was envisaged, the acid strength was fixed at 0.06 N hydrochloric acid throughout. Recoveries of silver, gold, copper and lead as chlorides, collected by cementation from 0.06 N hydrochloric acid at room temperature, are shown in Table 111. Recoveries of 80-100y0 were obtained for silver, gold and copper with a relative standard deviation of & 1-16%, but for lead the recovery was much poorer, 37%, and the relative standard deviation very much greater, & 34%.Nitrate is known to interfere with cementation reactions,13 and as the standard solutions of silver and lead are more usually prepared from the nitrates and also as aqua regia digests were envisaged for soil and sewage sludge sample preparation, a brief investigation of this effect was undertaken. The effect of using solutions of the nitrates of silver, gold, copper and lead in 0.06 N hydrochloric acid was to reduce the recovery of all of the elements to less than loyo, an exception being gold, which was almost unchanged. The recoveries were calculated from atomic-absorption spectrometric analysis of solutions pre- pared from the aluminium powder concentrates and verified by analysis of the initial and residual solutions.TABLE I11 PERCENTAGE RECOVERIES ON ALUMINIUM AT ROOM TEMPERATURE FROM 0.06 N HYDRO- CHLORIC ACID CONTAINING 1 mg 1-1 OF COPPER, 2 mg 1-1 OF LEAD, 0.04 mg 1-1 OF SILVER AND 0.2 mg 1-1 OF GOLD AS CHLORIDES Recovery, yo 100 82 83 37 Ag Au c u Pb Relative standard deviation, yo & l +16 f l 0 f34 The recoveries of arsenic, antimony, selenium, tin, tellurium and thallium from solution by means of a similar procedure using aluminium powder as a cemetant were found to be negli- gible. If, however, the aluminium was amalgamated by drawing a solution of mercury (11) chloride (40 mg 1-1) through it prior to cementation the recoveries were substantial, particularly for selenium. These recoveries are shown in Table IV, which also shows those for osmium, palladium, platinum and rhodium.Preliminary investigations have indicated that recoveries of the elements palladium, platinum and rhodium can be appreciably increased by heating the sample solution in a modified column. Typical results obtained when the sample solution is heated to 60 "C prior to being drawn once through a column are also shown in, Table IV. TABLE IV CONTAINING 1 mg 1-1 OF ARSENIC, ANTIMONY, SELENIUM, TIN, TELLURIUM AND THALLIUM and 1 mg 1-1 OF OSMIUM, PALLADIUM, PLATINUM AND RHODIUM PERCENTAGE RECOVERIES ON AMALGAMATED ALUMINIUM FROM SOLUTIONS Solution temperature w Element 25 "C 60°C As Sh Se Sn Te T1 0 s Pd Pt Rh 20 30 100 30 0 15 26 20 50 50 77 20 65Aaal. Proc. A comparison between the recoveries of silver, gold, copper and lead on amalgamated and unamalgamated aluminium revealed that there was no significant difference between the two sets of results and in all further experiments the aluminium powder was amalgamated in the column before use.Having established that amalgamated aluminium appeared to be able to collect all except one of the elements of interest, tellurium, a few natural samples were studied. A European sewage sludge and two Scottish soil samples ( 5 g ) were digested three times with 25ml of aqua regia (6 N re-distilled hydrochloric acid - nitric acid, 3 + 1). After each addition the digest was evaporated to dryness on a steam-bath. In order to remove nitrate from the residue it was treated three times with 6 N hydrochloric acid, evaporating the solution to dryness after each addition.Then, 50 ml of 0.06 N hydrochloric acid were added and the solution filtered before being drawn through the amalgamated aluminium cementation column. An SSMS analysis had previously been carried out directly on the ashed sewage sludge and the results are compared with SSMS analysis of the aluminium cementation concentrate inTable V. 12 RESEARCH AND DEVELOPMENT TOPICS TABLE V TRACE ELEMENT CONTENTS OF A EUROPEAN SEWAGE SLUDGE Micrograms per gram of air-dried sludge. Element Analysis of ash c u 450 Se - As - Rh - Pd - 25 0.05 Pt Au 0.92 T1 0.04 Pb 180 Bi 20 - 2 Analysis of cementation concentrate 330 0.3 0 03 0.18 0.08 25 0 02 0 05 0.75 0.06 130 10 There was good agreement between the silver, gold, bismuth, copper, lead, platinum and thallium results.The elements arsenic, osmium, palladium, rhodium and selenium, which were observed and determined in the aluminium concentrate, were not detected in the ashed material. This may be due, in part, to the loss of volatile elements during the ashing stage but it also reflects the improved detection limits obtained following concentration. In the aluminium concentrates obtained from both soil samples arsenic, palladium , selenium and thallium were also detected, as is shown in Table VI. For those elements which could be determined by semi-quantitative spectrographic analysis of the soil ashes, reasonable agree- ment was found between these and SSMS analyses of the aluminium concentrates. The discrepancies may occur because amounts extracted by an aqua regia digest will approach, TABLE VI TRACE ELEMENT CONTENTS OF TWO SOIL SAMPLES Micrograms per gram of air-dried soil.Acid igneous Ultra-basic parent material parent material && SSMS* Spec.? SSMS* Spec.? Ag 0.01 <1 0.06 <1 Au 0.04 t l 0.05 <1 cu 3 10 350 300 Pb 5 9 1 4 * SSMS = Analysis of cementation concentrate from aqua regia digest by spark source mass spectro- t Spec. = Analysis of soil ash by semi-quantitative optical-emission spectrography. metry.January, 1980 RESEARCH AND DEVELOPMENT TOPICS 13 but will not always be quantitatively equal to, the total element content measured by spectrographic analysis of the solid soil material. One of the authors, K. H. Welch, thanks the trustees of the Analytical Chemistry Trust Fund of the Chemical Society for the award of an SAC Research Studentship.1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. References Ure, A. M., Bacon, J. R.. Berrow, M. L., and Watt, J. J., Geoderma, 1979, 22, 1. Ure, A. M., and Bacon, J. R., Geochim. Cosmochim. Acta, 1978, 42, 651. Ure, A. M., and Bacon, J. R., Proc. Anal. Div. Chem. Soc., 1976, 13, 124. Ure, A. M.. and Bacon, J. R., Analyst, 1978, 103, 807. Bacon, J. R., and Ure, A. M., Anal. Chim. Acta, 1979, 105, 163. Bacon, J. R., and Ure, A. M., “Advances in Mass Spectrometry,” Volume 8, in the press. Scott, R. O., and Mitchell, R. L., J . Soc. Chem. Ind., 1943, 62, 4. Mitchell, R. L., and Scott, R. O., J . Soc. Chem. Ind., 1947, 66, 330. Scott, R. O., Mitchell, R. L., Purves, D., and Voss, R.C., “Spectrochemical Methods for the Analysis of Soils, Plants and Other Agricultural Materials,” Consultative Committee for Development of Spectrochemical Work, Bulletin 2, 1971, The Macaulay Institute for Soil Research, 1971, p. 23. Bratt, G. C., and Newman, 0. M. G., Ras. Disclosure, 1976, 147, 66. Pandey, U. M., Chem. Era, 1976, 12, 357. Hawley, J. E., Wark, W. J., Lewis, C . R., and Ott, W. L., Can. Min. Metall. Bull., 1951, 474, 669. Svehla, G., Editor, “Comprehensive Analytical Chemistry,” Volume 111, Elsevier, Amsterdam, 1975, p. 242. Evaluation of Diskpin Samples for the Sampling and Analysis of Liquid Iron and Steel J. McCaig, A. Cunningham, J. Lindsay and J. Little and J. M. Ottaway Chemical Laboratory, Ravenscraig Works, British Steel Corporation, Motherwell, Lanarkshire Department of Pure and Applied Chemistry, University of Strathclyde, Cathedral Street, Glasgow, G1 1XL Technological advances have caused the speed and diversity of modern steelmaking processes to increase rapidly, necessitating the provision of a rapid, accurate analytical service that is able to provide data on the chemical composition of liquid melts at various stages of the refining process.The speed and dynamic nature of modern processes such as L.D. basic oxyge n steel- making has continued to demand even faster analysis from the analyst in order that process control can be carried out within the stringent time limits available. Sophisticated computer-controlled techniques such as optic a1 emission spectrometry and X-ray fluorescence spectrometry are used routinely in the British Steel Corporation Ravens- Craig Works to enable the two demands of speed and accuracy to be met.These analytical techniques have been developed to the extent where it is unlikely that their speed will be surpassed in the immediate future and thus it became apparent that any further savings in the time of analysis would only accrue by streamlining the slowest stage in the system, the sampling and sample preparation operation. The first step in doing this was to select a uniform sample form to replace, if possible, the four completely different forms of sample at present in use. Of the various commercial sampling devices available it was decided to select a paddle and pin shape (Diskpin) rather than cylindrical, as the former type fulfils a multi-purpose role by providing a disk section for spectrometry and a pin section for the rapid determination of ele- ments such as carbon, sulphur, nitrogen or acid-soluble aluminium.Sample preparation times for paddle and pin samples are much reduced, because time-consuming sectioning or drilling prior to analysis is unnecessary. The Diskpin sample had the added advantage of being available in both immersion and pneumatic modes for areas where immersion was impractical, e.g., the continuous casting mould. Initial sampling trials with the Diskpin sampling devices were used to define the appropriate technique for each process area and also to highlight areas where the sampling devices could be14 RESEARCH AND DEVELOPMENT TOPICS Anal.PYOC. improved. The three improved devices will be described only very briefly but consisted basically of a mild-steel disk-shaped mould with a silica pin entry channel encased in resin- bonded sand and held in a cardboard support tube. Metal caps protected the assembly from slag and the pneumatic sampler was hermetically sealed to allow suction of the liquid metal into the mould to occur when a partial vacuum was applied. The immersion samplers filled due to the action of ferrostatic pressure when immersed in liquid metal and the final sample shape was formed due to liquid metal chilling in the disk and entry channel to give the uniform sample shape shown in Fig. 1. Fig. 1. The Diskpin sample. The results of sampling trials carried out with these devices are indicated in Table I.TABLE I SUMMARY OF SAMPLING TRIALS Process area Technique Sampler Success rate, yo Steelmaking convertor . . . . .. .. Immersion 513-005 97 Steelmaking ladles . , .. . . .. Immersion 5 13-004 98 Ingot moulds . . .. , .. .. .. Pneumatic 5 13-002 97 Casting tundish . . .. .. .. .. Pneumatic A 1 3-002 9s Torpedo ladles . . .. . . . . .. Immersion 5 13-004 < 10 Blast furnace runner . . .. .. . . Pneumatic 5 13-002 40 The addition of titanium to the sample mould meant that a wide range of steel types could be sampled successfully even in highly oxidised conditions such as are found in “rimming” steels. Low success rates experienced when sampling liquid iron were due mainly to the sample mould failing to chill the liquid iron satisfactorily, which resulted in loss of metal and unsatis- factory samples.A comparison of the various sampling operations involved is shown in Table 11, with approxi- mate times for each operation. It can be seen that the Diskpin offered substantial savings in time over two of the established techniques and offered a practical alternative for sampling liquid steel. Further work is under way to define a suitable device for sampling liquid iron. Analytical Examination of Diskpin Samples A series of Diskpin samples drawn concurrently with the established cylindrical samples were analysed by emission spectrometry for the main elements of interest, viz., carbon, silicon, sulphur, phosphorus, manganese and aluminium. Good agreement was obtained as reflected by the values of the regression line for each ele- ment shown in Table 111.January, 1980 RESEARCH AND DEVELOPMENT TOPICS TABLE I1 COMPARISON OF SAMPLING TIMES 15 Operation Sampling .. .. .. .. . . Stripping . . .. . . .. .. Cooling . . .. .. .. . . Preparation time- Spectrometry . . . . . . .. Combustion . . . . .. .. Total time .. . . . . .. Sampling time/s r I Spoon and mould Immersion cylinder Diskpin A 120 10 10 10 15 20 30 150 60 60 120 340 60 120 355 60 20 170 TABLE I11 REGRESSION DATA FOR DISKPIN AND CYLINDRICAL SAMPLES Element Slope of regression line Intercept Carbon . . .. . . 1.01 9 x 10-5 Silicon . . .. .. 0.99 4 x 10-4 Sulphur .. .. 1.03 7 x 10-4 Phosphorus . . .. 1.01 4 x 10-4 Manganese . . .. 1.00 7 x 10-3 Aluminium . . .. 1.00 8 x The reproducibilities of replicate analyses by emission spectrometry of the Diskpin, cylinder and standard spectrometric samples were determined and the Diskpin sample exhibited marginally improved precision over the established sample for carbon, silicon, phosphorus, manganese and aluminium, although both sample types were significantly less reproducible than the standard samples, which are specially homogenised for spectrometric analysis.The Diskpin sample was examined for possible heterogeneity both across and through the sample section. This is of paramount importance in metallurgical analysis, where only a small area and mass of the sample is atomised for analysis and thus local imperfections caused by variation in metallurgical structure or inclusions can have a drastic influence on the final result if they occur within the focal region of the spark.This examination was carried out by analysing a group of Diskpin samples of different compositions at four discrete positions at each of twelve discrete levels through the section of the sample utilising emission spectrometry. Results obtained for carbon, silicon, sulphur, phosphorus, manganese, nickel, copper, chromium, molybdenum and aluminium were obtained from the Diskpin and standard samples. A statistical classification of the variances produced was used to demonstrate the presence of heterogeneity for these elements, which, although statistically significant, was considered small enough to render the Diskpin sample acceptable for the routine technical classification of steels. These results were also used to confirm the random nature of the heterogeneity, which sug- gested that no specific depth of preparation was required prior to spectrometric analysis being carried out.The pin section of the sample was also examined with a view to determining the suitability of the pin for the rapid determination of carbon, sulphur nitrogen and acid-soluble aluminium for process control purposes. This was carried out by comparing the analysis of the disk and pin sections of the samples selected using the techniques outlined in Table IV. TABLE IV COMPARISON OF ANALYSES OF DISK AND PIN SECTIONS OF DISKPIN SAMPLES Element Analytical method Slope of regression line Intercept Carbon .. . . Combustion - non-aqueous titration Sulphur . . . . Combustion - titration Nitrogen .. . . Carrier - gas fusion Aluminium . . . . Automated spectrophotometry 0.99 6 x 1.01 7 x 10-4 1 .oo 1 x 10-4 0.97 1 x 10-316 RESEARCH AND DEVELOPMENT TOPICS Anal. PYOC. Repli- cate analyses of discrete samples taken along the pin length for carbon, sulphur, nitrogen and aluminium gave reproducibilities approximating closely to the precision of the technique used for each element, and this confirmed the homogeneity of the pin section and confirmed its suitability for rapid control analysis of these elements. The suitability of the pin section for use in the method for the rapid determination of acid- soluble aluminium1 reported by Gale has been demonstrated and is now in routine use. The general applicability of the Diskpin device has been demonstrated for the routine analysis of liquid steel in the Ravenscrajg Works. Further work is continuing to develop a device suitable for sampling liquid iron and modified devices that give an improved representa- tion of the aluminium content of liquid melts.The authors thank the Ravenscraig Laboratory of the British Steel Corporation for facilities to carry out this work. Good agreement was obtained between the disk and pin sections for these elements. Reference 1. Gale, P., British Steel Corporation, Research Report CC/1/78, 1977. Some Applications of Inductively Coupled Plasma Emission Spectrometry to the Determination of Toxic Trace Elements G. F. Kirkbright, D. L. Millard and R. G. Snook Department of Chemistry, Imperial College of Science and Technology, London, S W7 2A Y The combination of the inductively coupled plasma (ICP) with multi-channel or even single- channel spectrometers has provided the analyst with a powerful optical emission spectro- metric technique for the determination of elements at trace, minor and major levels in many types of sample.The most important advantages that this combination of instrumentation has over other spectrometric techniques are the extended linear calibration ranges obtained (typically six orders of magnitude with respect to concentration) and the absence of inter- ferences in the ICP source itself. Solution samples are normally introduced into the plasma by the use of pneumatic nebulisa- tion and for most solution applications the pneumatic nebuliser is an ideal device for intro- duction of the sample into the ICP.There are, however, limitations to the use of pneumatic nebulisation. For example, the solution uptake rate may be altered if the viscosity of the sample solution changes from one sample to another, although this can be compensated for by correct and careful matching of standards. Another more difficult situation to deal with occurs when nebulising solutions with high dissolved solids contents, which frequently block the orifice of the nebuliser. More recently, ultrasonic nebulisation has been employed as a technique for sample introduction and its application frequently leads to superior detection limits than are obtainable with pneumatic nebulisation. It may be necessary, however, to employ desolvation procedures to minimise the adverse effects observed in the plasma owing to increased sample solution loading of the plasma.Even when employing desolvation pro- cedures certain undesirable effects may be observed if temperature stability of the desolvation apparatus is not maintained. A disadvantage of both nebulisation procedures is the relatively large volume of solution (1-2 cm3) required for each analysis, although Greenfield and Smith have devised a system for the nebulisation of microlitre volumes of so1ution.l In our laboratory we have investigated an alternative sample introduction system, which employs a resistively heated graphite rod to vaporise the analyte from small sample volumes (10-20 p1) directly into the injector of the ICP torch and hence into the axial injector channel of the ICP.The graphite rod is enclosed in a glass chamber (volume 1 dm3), which is connec- ted to the injector tube of the ICP torch with 0.5 m of plastic tubing. Volumes of 10 pl of solution can be applied to the graphite rod through a sampling port in the chamber using a micropipette and can be desolvated and subsequently vaporised into the ICP by applying pre-programmed heating cycles to the graphite rod. The vaporised analyte is carried to the plasma on a continuously flowing stream of argon that passes through the glassJanuary, 1980 RESEARCH AND DEVELOPMENT TOPICS 17 chamber. This procedure frequently leads to superior detection limits than are obtainable with pneumatic nebulisation as the analyte passes through the axial channel as a pulse of sample, giving rise to a transient analyte atomic-emission signal.This transient signal is detected at the photomultiplier of the single-channel spectrometer and recorded on a potentiometric chart recorder, enabling accurate peak-height measurements to be taken. Under the optimum conditions reported previously2 we have established that for many elements calibration graphs can be obtained over five orders of magnitude with respect to concentration of the analyte present and detection limits of 10-10 to 10-l2g can also be obtained using the graphite rod as a sample introduction device. The precision of these determinations is acceptable, with relative standard deviations of typically about 0.05. For most of the elements investigated, inter-element effects and interference effects have been absent. For some of the more volatile toxic elements investigated, however, we observed a marked increase in emission intensity in the presence of concomitant elements. The best example of this is the effect of selenium(V1) on the observed emission intensity of cadmium and arsenic.In the absence of selenium(V1) a calibration graph for cadmium plotted on double logarith- mic axes was rectilinear over a concentration range of only two orders of magnitude and showed a slope of 1.5; the detection limit for cadmium under these conditions was g. In the presence of 10 pg of selenium(VI), however, a graph on similar axes was rectilinear over four orders of magnitude; the resulting detection limit for cadmium was 10-l1g. As no similar effect was observed when cadmium solutions in the absence and presence of selenium were nebulised into the ICP it was thought that the effect is due to an inter-element effect in the vaporisation apparatus rather than an interference in the plasma source.Stabilisation of the volatile elements by selenium during pre-vaporisation heating cycles was ruled out as the effect was still observed if the analyte and selenium(V1) solutions were placed in different depressions in the graphite rod (thus precluding intimate mixing on the rod). Washing experiments were undertaken to assess the extent of possible deposition of the elements in question along the connecting tube from the vaporisation chamber to the ICP torch injector and it was found that in the absence of selenium appreciable deposition of cadmium occurs, resulting in poor transport efficiency to the plasma; in the presence of selenium, however, the deposition of cadmium is much reduced.Radiotracer experiments were then undertaken to quantify the effect of selenium on the transport efficiency of cadmium to the plasma. The distribution of 115Cdm (i4 = 47 d) in the vaporisation apparatus and connecting tube was determined, and also the amount of 1Wdm delivered to the injector tip of the plasma torch in the presence and absence of selenium- (VI). These results are summarised in Table I, and clearly indicate that the enhancement of observed emission intensity is a result of increased transport efficiency, by prevention of deposition of the volatile elements on the walls of the vaporisation chamber and connecting tube to the plasma.We believe from evidence presented by microscope slides of the vaporised material collected 1.0 cm above the graphite rod that the transport of analytes to the plasma is facilitated by the formation of dry condensed aggregates of material a few centimetres above the graphite rod after vaporisation. These dry aggregates are successfully transferred to the ICP torch. When considering the more volatile elements, however, there appears to be an insufficient path length between the graphite rod and the walls of the glass chamber to allow the analyte vapour phase to cool to form aggregates, and deposition therefore occurs. The TABLE I DISTRIBUTION OF CADMIUM-115111 ON MANIFOLD, CONNECTING TUBE AND INJECTOR TIP Distribution of Cd, yo* A Cd concentration/ Concomitant 7 pg cm-3 ion (10 p g ) Chamber Tubing Injector tip 0.1 - 44 17 9 1.0 - 40 22 18 10.0 - 38 18 38 0.1 Se(V1) 7 4 63 1.0 Se(V1) 9 8 64 10.0 Se(V1) 7 5 59 * Expressed as percentage of Cd applied to graphite rod.18 RESEARCH AND DEVELOPMENT TOPICS Anal.PYOC. addition of concomitant elements such as selenium(V1) assists the formation of aggregates and therefore enhances the transport efficiency of the volatile elements to the plasma. The use of the graphite rod vaporisation apparatus as a sample introduction device has permitted the determination of many trace elements in a variety of sample types. For example, ten elements have been determined successfully in uranium solutions using a sample volume of lop1 of solution applied to the graphite rod, and 10-fold superior detection limits than observed using pneumatic nebulisation have been obtained.Similarly, the direct determination of toxic elements such as lead, cadmium and zinc has been performed on bovine serum with no sample pre-treatment other than the addition of a surfactant (Triton X) to assist in “wetting” the graphite rod, 0.5% of sodium sulphide solution to stabilise volatile elements during drylng and 0.1% of selenium(V1) to prevent deposition of these elements in the connecting tube of the apparatus. These additives can be prepared as a single solution and simply added to sample solutions in situ on the graphite rod to provide a rapid method of analysis requiring only small sample volumes. References 1.2. Greenfield, S., and Smith, P. B., Anal. Chim. Acta, 1972, 59, 341. Gunn, A. M., Millard, D. L., and Kirkbright, G. F., Analyst, 1978, 103, 1066. Application of Differential Spectrophotometry to the Determination of Fluoride Using AFBS C. Jordan Department of Analytical Chemistry, The Queen’s University of Belfast, Belfast, B T9 5A G , Northern Ireland Fluoride can be determined with the use of a calibration line prepared using Deane’s pro- cedurel for AFBS. This procedure, an example of the only positive absorptiometric method for fluoride, is based on the colour of the ternary complex formed between AFBS, La and F- in aqueous solutions of fixed pH (around 4.7) with an AFBS to La ratio of 1 : 4. Solutions are prepared by adding various aliquots of standard sodium fluoride solution to a mixture of AFBS in a constant ionic strength buffer followed by addition of lanthanum(II1).After 20 min, the transmittance (T) of each solution (x) can be measured at 635 nm against an AFBS - La blank (s) in the ordinary way [giving T(x/s) readings] and the calibration line drawn by plotting -log T(x/s) against the fluoride concentration (C, or C,) : -log T(x/s) = A(x/s) = €C,Z The fluoride concentration of an unknown solution can be read from the calibration line if its absorbance, A (x/s), is known. The uncertainty in the value of this concentration, due to the uncertainty S , in the transmittance measurements, is usually quoted2 by reference to -0.434 s c - c - T(x/s)logT(x/s) ST which reaches a minimum of about 3 x S, at T(x/s) = 0.368.As-the scale reading error makes an important contribution to S , in the older type of spec- trophotometer, it is reasonable to expect that this type of error, and hence the photometric error, could be reduced by scale expansion. This scale expansion can be achieved by either electrical or “chemical” means. With chemical scale expansion, standard solutions of the analyte are used to set the ends of the %T scale of the spectrophotometer. The three possi- bilities arising from this differential approach are detailed by Reilley and Cra~ford.~ As other worker^^-^ have already pointed out that, in practice, the trace analysis and ulti- mate precision techniques do not necessarily lead to any gain in precision, it was decided that, of the two, only the more general, ultimate precision technique would be used.Thus, in the following, the ordinary technique for fluoride determinations with AFBS is compared with the transmittance ratio and ultimate precision techniques only.January, 1980 RESEARCH AND DEVELOPMENT TOPICS 19 If the 100yoT point is set with reference solution r in the beam (concentration Cr) and the OyoT point is set with reference solution R (concentration C,) then the reading on the expanded yoT scale, which we shall call the relative transmittance, Trel, is related to the ordinary transmittances of solutions x, r and R by If Beer's law holds for the sample, equations (1) and (3) give For the transmittance ratio technique, T(R/s) = 0 and equation (4) simplifies to -log Trel = -log T(x/r) = J C x + log T(r/s) (54 and the calibration graph will be linear (intercept # 0).precision calibration graphs of -log Trel against C, are non-linear. The trace analysis and ultimate Photometric Error Contribution The photometric error associated with the use of these differential methods can be deduced' by applying the law of propagation of errorsS to equation (4), giving We therefore expect [from equation (6a)l a decrease in the photometric error, for a given solution, as T(r/s) decreases and/or T(R/s) increases, i.e., as scale expansion is increased. (For the more general case, the original paper of Reilley and Crawford3 should be consulted.) As increased scale expansion is achieved, however, one finds that STrel tends to increase (Table I). TABLE I EFFECT OF SCALE EXPANSION ON THE STANDARD DEVIATION OF THE TRANSMITTANCE MEASUREMENTS, ST,,^ C , CR Scale ST'reI 'T'rel x t0.05, Technique x los/,, x 1 0 6 / ~ expansion x 103 d G Ordinary, n = 6 Transmittance ratio, 9% = 6 Ultimate precision, n = 3 0 Shutter 1.00 1.17 1.23 6 Shutter 1.67 2.05 2.15 7 1.82 2.02 2.12 8 2.01 2.06 2.16 9 2.18 1.94 2.04 7 13 4.41 7.2 17.9 8 12 6.80 11.6 28.8 was determined for each technique from n parallel %T measurements. emptied and re-filled and the spectrophotometer readjusted for each measurement. The cell was The net effect of changes in T(r/s) and T(R/s) on the photometric error [calculated using equations (6a and 6b)l is shown in Fig.1. Fig. 1 shows that, while there is a net reduction in the photometric error when the transmittance ratio technique is used, there is a net increase in the photometric error with the ultimate precision technique where is increasing more rapidly than T(r/s)-T(R/s) is decreasing.This can be traced to the poor medium-term stability of the SP500 instrument and the relatively long time required with the ultimate precision technique to place the sample solution in the spectrophotometer beam once the expanded scale has been set up.20 RESEARCH AND DEVELOPMENT TOPICS Anal. PYOC. Effect of Scatter In addition to considering the photometric error contribution, one should also make an allowance for the scatter of the points about the calibration line. This is easiest achieved by carrying out a linear regression analysis on the calibration points, the ultimate precision results being made amenable to analysis by converting Trel readings into T(x/s) values using equation (3) and treating the calibration line as if it had been obtained by the ordinary technique.We can then represent all the calibration lines by equation (5a), which can be rewritten as where a = intercept on the y-axis and b = E = slope of line. Application of the law of propagation of errors to this expression permits one to deduce that the maximum relative concentration error with 95% confidence limits, associated with a concentration determination from this line, is given by7 [tor.1-' x $I2 = [ (,,34sT}]2 + [ %I2 + S b dfi y-a IT(x/r) Y-U (7) where S, and s b are the standard deviations of the intercept and slope, respectively, and are evaluated during the linear regression analysis of the results.We see immediately that the first term of equation (7) makes allowance for the photometric error while the other two allow for the scatter of the points about the line. Linear regression analysis of the N experimental points also permits the calculation of the actual error, ACF, involved in calculating C, from a given experimental point Yo with the regression equation. Thus, where ACF is given by9 As with equation (7), the error decreases as N, n and b (or E ) increase. ACF is at a minimum if the middle of the calibration line is used (yo = 7). Fig. 2 shows how the actual relative con- centration error, ACIC, varies with the number of parallels (n) and with the concentrations of the reference solutions (Cr and C,) and that of the sample (C,).The variation of the maximum relative concentration error, Sc/C, with n, Cr, C, and C,,l0 assessed by equation (7), exactly parallels the results shown in Fig. 2 except that the values are approximately 2.5 times as large. We see that the so-called ultimate precision technique actually leads to a considerable loss in precision over the ordinary technique (especially as Cr and C, approach one another) whereas the transmittance ratio technique, despite the limited scale expansion possible here, leads to a net increase in precision as T(r/s) is decreased (Cr increased). In conclusion, it can be said that it would be a waste of time using the ultimate precision technique in an attempt to increase the precision of a spectrophotometric method unless the effect of instrumental drift on the transmittance measurements can be kept low either by using a new generation spectrophotometerwith good medium-term stability or, less expensively, by obtaining a thermostated compartment with three matched cells for the old type of spec- trophotometer. The latter alternative would reduce s, by minimising the time required to place the sample solution into the beam once the expanded scale was set up. Unless this modification is available, the old type of spectrophotometer such as the SP500 is best suited to the transmittance ratio technique, which can lead to useful decreases in both the photometric and overall error, especially for low-transmittance samples. It is probably for this reason, as well as that of simplicity and convenience, that the latter technique has been the most widelyJanuary, 1980 EQUIPMENT NEWS 21 (n=31 oc 0 h c W v “ u c 9 6 7 8 9 10 11 12 13 14 CF ( X I 06)/M Fig. 1. The photometric error, as assessed by (6 a,b), for the ordinary a, transmittance ratio 0 and ultimate precision - .-.--. techniques, for values of Cp ranging from 6 to 13 x 10-6 M. The curve numbers refer to the values ( x lo6) of C,/Ca used to set the ends of the transmittance scale. Those curves labelled with a prime correspond to n = 6; otherwise n = 3. 0 6 9 6’ 0’ 7’ 8’19’ ” 7 8 9 10 11 12 13 14 CF(X106)/M Fig. 2. The actual over-all error, as assessed by (8), for the ordinary a, transmittance ratio 0 and ultimate precision - . - . - . techniques, over the range CF = 9-13 x M. The curve numbers refer to the values ( x 106) of C,/C, used to set the ends of the transmittance scale. Those curves labelled with a prime correspond to n = 6; otherwise n = 3. used differential method to date, having been used for the determination of often relatively high concentrations of about 20 elements in a wide variety of media, with S,/C values as low as 0.1% claimedll (although this is for the photometric error contribution alone). 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. References Deane, S. F., and Leonard, M. A., Analyst, 1977, 102, 340. Willard, H. H., Merritt, L. L., and Dean, J. A., “Instrumental Methods of Analysis,” Fifth Edition, Reilley, C . N., and Crawford, C . M., Anal. Chem., 1955, 27, 716. Shigematsu, T., and Tabushi, M., Bull. Inst. Chem. Res. Kyoto Univ., 1958, 36, 127; Anal. Abstr., Bontnik, B., Chem. Anal. (Warsaw), 1964, 9, 717; Anal. Abstr., 1965, 12, 6357. Ingle, J. D., Anal. Chem., 1973, 45, 861. Jordan, C., MSc Thesis, Queen’s University of Belfast, 1978. Nalimov, V. V., “The Application of Mathematical Statistics to Chemical Analysis,” Pergamon Press, Davies, 0. L., and Goldsmith, P. L., Editors, ‘‘Statistical Methods in Research and Production,” Jordan, C., and Leonard, M. A., Microchem. J., to be published. Svehla, G., Talanta, 1966, 13, 641. Van Nostrand, New York, 1974, pp. 92-97. 1959, 6, 3332. Oxford, 1963, p. 34. Oliver and Boyd, Edinburgh, 1972, p. 206.