ANALYTICAL PROCEEDINGS, JANUARY 1985, VOL 22 3 Research and Development Topics in Analytical Chemistry The following are summaries of sixteen of the papers and posters presented at a Meeting of the Analytical Division held on June 26th and 27th, 1984, in the University of Manchester Institute of Science and Technology. Some Aspects of the Pyrolysis Gas Chromatography of Quaternary Phosphonium Compounds Sarah J. Abraham and W. J. Criddle Department of Applied Chemistry, Redwood Building, UWIST, P.O. Box 13, Cardiff, CFI 3XF It is well accepted that quaternary ammonium halides of the general type R4N+X- decompose thermally to yield an amine, R3N, and the corresponding alkyl halide, RX.1-3 Early studies by Collie4-5 and others suggested that the thermal decomposition of quaternary phosphonium com- pounds follows a different path, forming a tertiary phosphine and an alkene when the anion is a halide: Et,P+Cl- + C2H4 + PEt3.HCI 2Me4P+Cl- + C2H4 + 2PMe3.HCl Ingold and co-workers later revised this6 to the expected tertiary phosphine and alkyl halide for phosphonium halides.The success of pyrolysis gas chromatography as an analytical tool for the determination of quaternary ammonium com- pounds7-9 suggested that the method would also be applicable to quaternary phosphonium compounds. In this paper we report the results of studies undertaken on the fundamental aspects of the pyrolysis of some simple phosphonium halides. Experimental The compounds studied were simple, readily obtainable or easily synthesised quaternary phosphonium halides.The cations used were (C2H5)4P+, (CH3)3P+Ph, (CH3)2P+Ph2, al C 0 P v) 2? P 8 a L a 1 Time Fig. 1. Pyrolysis of trimethylphenylphosphonium iodide by direct injection in aqueous solution. Injection port temperature, 350 "C CH3P+Ph3 and Ph4P+, while the anions used were chloride, bromide or iodide, these being easily interchangeable using ion exchange procedures. The pyrolysis procedures employed were direct injection (aqueous solution, 10-3 M) and the quartz tubekoil pyroprobe (CDS 190). Gas-chromatographic separations were carried out by using a Perkin-Elmer Sigma 2B gas chromatograph and data correlated on a Sigma 15 data station. The internal standards used were anilinium chloride and 4-toluidine oxalate. For sample preparation in oxygen-free atmospheres, a glove box flushed with white spot nitrogen was employed.Results and Discussion It will be immediately obvious from Table 1 that the pyrolysis of quaternary phosphonium compounds is considerably more complex than the corresponding nitrogen-based compounds. In addition, unless rigorous experimental conditions are adhered to, artifact formation can result. For example, in initial experiments using both the direct injection and pyro- probe techniques, the appropriate phosphine oxides were commonly formed, this effect being more serious for the chloride than for the iodides. Table 1. Products from the pyrolysis of typical quaternary phosphonium compounds Compound Products first X = halide observed at/"C Ph4PX Ph3P PhX PhPh PhH 225 MePh3PX MePh2P Ph,P Ph,PO MeX PhX 250 Observed products at 600 "C (probe) PhCH3 PhPh PhH + unknown Me2Ph2PX Me2PhP MePh2P MePh2P0 MeX 300 PhX PhH PhCH3 PhPh Me,PhPX Me,PhP Me2PhP0 Me3P MeX PhX 350 PhH PhPh PhCH3 However, oxygen from external sources, i.e., dissolved oxygen in the instance of direct injection pyrolysis, contributes to this effect.For pyroprobe pyrolyses, it Seems probable that the equilibrium can account for the appearance of phosphine oxides. This may be explained on the basis that the compound RX, if volatile, is partially lost during the sample work-up, and the free phosphine combines with atmospheric oxygen to give the R4P+X- C R3P + RX4 ANALYTICAL PROCEEDINGS, JANUARY 1985, VOL 22 a, !z 0 II v) F E 8 L a, n Time 0 (D, i Time C) r- x ? Time Fig. 2. Pyrolysis of trimethylphenylphosphonium iodide aqueous solutions de-aerated with nitrogen.( a ) , Flame ionisation detector; ( b ) . nitrogen - phosphorus detector in the nitrogen - phosphorus mode; (c), nitrogen - phosphorus detector in the phosphorus mode. Internal standard, anilinium chloride (aniline retention time -7.7 min) . corresponding oxide. Support for this explanation may be found in that the chloride of, say, the trimethylphenylphos- phonium salt, gave rise to more oxide than the corresponding iodide. This is as expected because methyl chloride (a gas at room temperature) will be readily lost during sample prepara- tion. Preparation of all samples under nitrogen eliminates the formation of phosphine oxides for all of the halides examined. It should also be noted that, with pyroprobe pyrolyses, it is essential to allow all of the air that enters during sample insertion to be eliminated by carrier gas purging for some minutes before firing.If pyrolysis is initiated immediately on equilibration (or even by direct injection of a non-deaerated solution), extremely complex and essentially valueless chro- matograms are obtained (Figs. 1 and 2). Turning now to the nature of the genuine pyrolysis products, those indicated in Table 1 suggest that, in part at least, a free radical process is involved. There is no reason to suppose that in the pyrolysis of quaternary phosphonium halides, the electron transfer - free radical mechanism proposed initially by Criddle and Thomas") for the analogous nitrogen compounds does not hold, e.g., Ph,P+X- [Ph4P] + X- Ph3P + Ph t 2Ph + PhPh biphenyl Ph + X+ PhX halobenzene see Table 1 Ph + H+ PhH benzene In conclusion, it would seem that, from an analytical standpoint, pyrolysis gas chromatography is considerably more difficult to use as a general procedure than for the correspond- ing nitrogen compounds, but in carefully selected instances could be a valuable method for both the qualitative and quantitative determination of quaternary phosphonium com- pounds.1. 2. 3. 4. 5 . 6. 7. 8. 9. 10. References Hoffman, A. W., Justus Liebig Ann. Chem.. 1851, 78. 253. von Meyer. E . , Chem. Zentralbl.. 1909, 1800. Emde, H., Arch. Pharm., 1910. 247, 351. Collie, N., Philos. Mag.. 1887, 24, 27. Collie, N., J . Chem. Soc., 1888, 713. Fenton. G. F.. Hey, L., and Ingold, C. K., J .Chem. Soc., 1933, 989. Martens, M. A., and Heydrickx, A . , J . Pharm. Belg., 1974,29, 449. Choi, P., Criddle, W. J . , and Thomas, J . , Analyst, 1979, 104, 451. Christofides, A . , and Criddle, W. J . , J . Anal. Appl. Pyrol.. 1982,4, 211. Criddle, W. J . , and Thomas. J . , J . Anal. Appl. Pyrol., 1980,2, 361. Metals as Labels in lmmunoassay Nichola J. Wilmott, James N. Miller and Julian F. Tyson Chemistry Department, University of Technology, Loughborough, Leicestershire, LEI 1 3TU During the last decade a considerable research effort has been directed towards the development of non-isotopic labels for immunoassay; in principle, metals should make good labels. Suitably chosen metals should fulfil the requirements for a non-isotopic label, i. e . , non-toxicity, low cost, ease of introduc- tion into sample molecules, an absence or low level in serum and a low limit of detection.Two distinct approaches have been made in the development of this type of immunoassay and these fall under the headings Metalloimmunoassay (MIA) and Sol-particle immunoassay (SPIA) . Metalloimmunoassay The idea of metalloimmunoassay was first introduced by Cais,' who proposed the use of metal ions in the form of an organometallic or co-ordination complex as the label. The work carried out by Cais et al. in this area has included the labelling of steroids with sandwich compounds, such as ferrocene1.2 and complexes containing cymatrene,3 as well as direct introduction into the antigen molecule in the instance of mercuration of steroid oestrogens.4 Generally these labels have been found to be relatively insensitive in immunoassay terms, no doubt partly because the metals used (iron, manganese, etc.) occur at high concentrations in biological samples.Sol-particle Immunoassay This approach to the use of metals as labels was introduced by Leuvering et al.5 and involves labelling with colloidal particles. Leuvering et ul. developed a method for the determination of human chorionic gonadotropin (HCG) by colloidal gold labelling and a simultaneous assay for HCG and human placental lactogen by labelling with colloidal gold and colloidal silver. Analysis of the components was carried out by using carbon furnace electrothermal atomisation atomic-absorption spectrometry and, in general, the detection limits compared favourably with radioimmunoassay procedures.Experimental Terbium - Transferrin Complex as a Label in Immunoassay The approach used in our laboratory to the development of an MIA procedure was to produce a fluorescent chelate complex of terbium with transferrin as the label. The complex was produced in 0.1 M Tris - hydrochloric acid buffer at pH 8.5 by addition of terbium to a solution of transferrin in the molar ratio 2 : 1 .6 The complex had an excitation maximum at 294 nm, an emission maximum at 548nm and was detectable down to lo-' M (measurements were made on a Perkin-Elmer LS5 spectrometer: slits 5.0 and 5.0, delay gate 0.05ms, gate time 5.0 ms).ANALYTICAL PROCEEDINGS. JANUARY 1985. VOL 22 5 Gentamicin (an aminoglycosidic antibiotic) was linked to the complex via a carbodiimide reaction, an addition of glycine being made to limit the number of gentamicin molecules bound to each transferrin m ~ l e c u l e .~ The subsequent terbium - transferrin - glycine - gentamicin complex was then tested for reaction with anti-gentamicin antibody: a double antibody technique was used whereby the gentamicin-containing com- plex was reacted with anti-gentamicin and subsequently with antibody to the anti-gentamicin to cause precipitation. The antibody titration results showed typical antigen - antibody reaction between the anti-gentamicin and the gentamicin complex. Attempts to use the gentamicin complex in an immunoassay procedure have so far proved relatively unsuccessful, no doubt partially due to the problems associated with kinetics and a macromolecular label.It is also possible that scattered light is affecting the fluorescence measurements, and at present experiments are being carried out on diluted solutions. We also intend to carry out the terbium determination by plasma emission spectrometry. Colloidal Gold as a Label in Immunoassay Colloidal gold was prepared by the reduction of chloroauric acid by sodium citrate, according to the method of Frens.8 Colloidal particles of average particle diameter 20 nm were used in the experiments. The adsorption of colloidal gold on to albumin (dialysed in order to remove electrolytes) was carried out by simple adsorption ~ aggregation of the labelled albumin being preven- ted by the addition of polyethylene glycol ( M , 20000).The exact amount of albumin required to stabilise the colloidal gold was determined by the construction of an adsorption isotherm4; varying concentrations of albumin (contained in 1 ml of water) were added to 5 ml of colloidal gold and allowed to stand for 1 min, 1 ml of 10% NaCl then being added and, after a period of 5 min. the degree of flocculation measured by the absorbance of the solution at 580 nm (the absorbance of the solution increasing for flocculated colloids which appear blue). The optimum pH of the colloidal solution was determined in a similar manner by using a concentration of albumin known to stabilise the colloidal gold. Having determined the optimum conditions for the prepara- tion of colloidal gold labelled albumin the labelled colloidal gold was used in an antibody titration.The double antibody technique was again used and results obtained showed good antibody recognition. All of the analyses carried out to date have involved spectrophotometric analysis. However, analysis of the labelled albumin by carbon furnace atomic-absorption spectrometry with electrothermal atomisation has been carried out and suggests that extremely small amounts can be detected by this means. Conclusion Generally, the methods of MIA have been found to be less sensitive than those of SPIA, mainly due to the number of metal ions bound per molecule, i . e . , one or a number less than 10 in the case of MIA compared with the many thousands per colloid particle. However, the number of possible complexes available for MIA is vast and does open up a number of areas and subsequently several different detection methods.But perhaps the most interesting factor concerning SPIA is the possibility of multi-component analysis which, with the use of such methods as plasma emission spectrometry and neutron activation analysis, may lead to simultaneous assays for two or more components. The authors thank the Trustees of the Analytical Trust Fund of the Royal Society of Chemistry for the award of an SAC Research S tuden tship. References 1. 2. 3. 4. 5. 6. 7. 8. 9. Cais, M., Nature (London), 1977, 270, 534. Cais, M., U.S. Pat., 4 205 952, 1980. Cais, M.. Bull. SOC. Chim. Belg., 1981, 90, 27. Cais, M., Actual. Chim., 1979 (7), 14. Leuvering, J. H. W., Thal, P. J . H. M., Van der Waart, M., and Schuurs.A. H. W. M., J . Zmmunoassay, 1980, 1, 77. Luk, C. K., Biochemistry, 1971, 10, 214. Wilmott. N. J . , Miller, J . N., and Tyson, J. F., Analyst, 1984, 109, 343. Frens, G., Nature (London), Phys. Sci., 1973, 241, 20. Geoghegan, W. D., and Ackerman, G. A , , J. Histochem. Cytochem., 1977, 25, 1187. The Simultaneous Determination of Chloride, Bromide and Iodide by HPLC Philip E. Moss and W. 1. Stephen Department of Chemistry, University of Birmingham, P. 0. Box 363, Birmingham, B 15 2TT A new technique is proposed for the simultaneous determi- nation of chloride, bromide and iodide in aqueous solution. The method depends on the ease with which an inorganic halide ion can be quantitatively converted to a covalently bound arylmercury(I1) halide, which is then measured by high-performance liquid chromatography (HPLC).Several methods are available for this conversion. e.g., the Peters' reaction,' a Grignard-type synthesis2 and an aryl- boronic acid technique.3 The last technique was adopted because it proceeds rapidly almost to completion at room temperature. Experimental Preparations of Phenylmercury(I1) Halides Phenylmercury( 11) chloride, bromide and iodide were pre- pared by addition of the corresponding alkali metal halide (potassium bromide, etc.) to a solution containing phenylboric acid (benzeneboronic acid) and a mercury(I1) salt (nitrate or perchlorate) made acidic to about pH 1.0 with the correspond- ing acid. The phenylmercury( 11) halide produced was extrac- ted into chloroform, the chloroform removed by evaporation and the compound purified by recrystallisation from ethanol.A series of substituted phenylmercury(I1) halides was produced by the same technique using substituted phenylboric acids. Those so far studied are 4-bromophenylboric acid (4-bromobenzeneboronic acid), 2,4-dichlorophenylboric acid (2,4-dichlorobenzeneboronic acid) and 3,5-bis(trifluoro- methy1)phenylboric acid [3,5-bis(trifluoromethyl)benzene- boronic acid]. Measurement of Phenylmercury(I1) Halides The phenylmercury(I1) halides (substituted and unsubstituted) were dissolved in a suitable solvent (ethanol or ether) and measured by HPLC under conditions of isocratic elution. The system used consisted of two 5 p ~ ODS columns linked together to provide a 35 cm long active column (4.6 mm i.d.). The injected volumes were of 5 1.11.Detection was by ultraviolet absorption at 220 nm [230 nm for 4-bromophenylmercury(II)6 ANALYTICAL PROCEEDINGS, JANUARY 1985, VOL 22 halides]; peak height and/or peak area measurements were used. Interferences Both cationic and anionic interferences were studied at ten-fold concentrations compared with the respective halide. Results Phenylboric Acid Technique The three phenylmercury(I1) halides were prepared and isolated as described above. Preliminary tests showed that phenyImercury(I1) chloride, bromide and iodide were best separated with a methanol - water (4 : 1) eluent at a flow-rate of 1 ml min-1, the order of elution being from chloride to iodide. When 1,2-dichloroethane was used, the extraction was quantitative for all three halides over the concentration range 0-100pgml-1.Recoveries of about 100% of the added amounts were possible for chloride and bromide, but only 70% for iodide. Substituted Phenylmercury(I1) Halides Several substituted derivatives were prepared as outlined above and examined under the same conditions as those used for the unsubstituted compounds. The chromatographic behaviour of these compounds differed markedly from the unsubstituted phenylmercury(I1) halides (see Discussion, below). 4-Bromophenylboric Acid Technique When the three 4-bromophenylmercury( 11) halides obtained from the 4-bromophenylboric acid were analysed under identical conditions to the unsubstituted compounds much better separations were obtained compared with the unsubsti- tuted compounds.When dichloromethane was used, the extraction of the 4-bromophenylmercury( 11) halides was quantitative over the range 0-100 pg ml-1 halide concentration with recoveries of about 100%. Discussion Phenylboric Acid Technique This can be used for the simultaneous determination of chloride, bromide and iodide but it has several drawbacks. The need to use dichloroethane as solvent leads to loss recoveries of iodide, possibly because of the lower solubility of phenylmer- cury(I1) iodide in dichloroethane. The relatively low A,,,. values (around 220 nm) for the phenylmercury( 11) halides lower the sensitivity and give higher noise than might be hoped for. The major drawback, however, is that phenylmercury(I1) chloride requires conditioning on the column prior to use, about 20 injections being sufficient to reach conditions of reproducible retention times and sensitivities.This may be caused by the adsorption of the phenylmercury(I1) chloride on to the silica column packing, it being the most polar of the three compounds. The problem can also be overcome by the addition of a small amount (about 1 pg ml-1) of the phenylmercury(I1) chloride to the eluting solvent. Substituted Phenylmercury(I1) Halides The substituted phenylmercury(I1) halides show a definite pattern in their chromatographic behaviour. The chlorides seem to be relatively unaffected by substitution and still elute first from the column. The bromides and iodides, however, show increasing retention times (and resolutions) as the electron-withdrawing power of the substituent increases, the order being (from least effective to most effective) 4-bromo- phenylmercury halides, 3,5-bis(trifuloromethyl)phenyl- mercury halides and 2,4-dichlorophenylmercury halides.The position of substitution also appears to be important, with the 2- and 4-positions having the greatest effect. A secondary effect is that as the electron withdrawing power increases the amounts of the corresponding substituted diphenylmercury by-products also increases. With the 33- bis(trifluoromethy1) substituents, the corresponding iodide has yet to be produced, probably because of the increased electron-withdrawing power weakening the already fairly weak mercury iodide bond. Because of this problem, the 4-bromo compounds were chosen for study. They also have a further advantage in giving a second absorption maximum at 230 nm.4-Bromophenylboric Acid Technique Although the conditioning problem also occurs with the 4-bromophenylmercury( 11) chloride compound, it can be overcome in the same way as for the unsubstituted chloride and under optimised conditions of eluent flow (1.2 ml min-1) it is a routine matter to determine simultaneously chloride, bromide and iodide in an aqueous solution with recoveries of about 100%. Interferences There are few interferences in this method. Cations that form sparingly soluble salts with halide ions ( e . g . , Ag+, Hg22+, Pb2+, etc.) interfere but at concentrations below the relevant solubility product the effect is small. Anionic interferences are cyanide and thiocyanate, which form the corresponding arylmercury compound, but acetate, phosphate, nitrate and perchlorate do not interfere. Further Studies An examination of the behaviour of the corresponding tolylmercury(I1) halides prepared from the three tolylboric acids (tolueneboronic acids) may provide interesting informa- tion on the chromatographic separation, which may lead to some improvement to the present technique.The possibility also exists for a modified version of the method to be used for the determination of low concentrations of mercury( 11) ions in aqueous media. References 1. Peters, W . , Ber. Dtsch. Chem. Ges., 1905, 38(3), 2565. 2. TsuTsui, M., “Characterisation of Organometallic Com- pounds,” Part 11, Wiley-Interscience, New York, 1971, p. 696. 3. Yazdi, A . S . , Ph.D. Thesis, University of Birmingham, 1982.Applications of Immobilized Enzymes in Flow Injection Analysis M. Masoom and Alan Townshend Chemistry Department, University of Hull, Hull, HU6 7RX Immobilized enzymes have received much attention in the past decade.’ They have a number of advantages over soluble enzymes when used in analytical systems, including decreased cost, greater stability and greater convenience of use. Their application is growing rapidly in clinical analysis, food analysis and for process control. The use of flow systems in conjunction with immobilized enzymes has been particularly effective. In this study flow injection analysis (FIA) is combined with immobilized enzyme technology and a flow-through ampero- metric detector. The cell consists of two parallel, platinumANALYTICAL PROCEEDINGS, JANUARY 1985, VOL 22 Phosphate buffer pH 6.8 7 1 *W - lnvertase electrodes, with a potential difference of 0.6 V, which detects hydrogen peroxide down to 1 x 1 0 - 6 ~ .The system first studied was the determination of glucose, by using glucose oxidase (GOD): Glucose + 02- Gluconic acid + H202 GOD was immobilized on controlled porosity glass (CPG) by cross-linking with glutaraldehyde: OEt OEt I I I 1 CPG -OH + EtO-Si-(CH2)3NH2 -.+ CPG-O--Si-(CH2)3NH2 OEt (A) OEt OEt H I I I (A) + OHC-(CH2)3-CH0 + CPG-0-5 -tCH2),-N=C(CH2)3-CH0 (B) OEt OEt H 1 I I (B) + H2N-Enz -* CPG -O-Si-(CH213-N=C-(CH2)3-CH=N-Enz OEt The immobilized enzyme was packed into a glass column (25 X 2.5mm) with an outer, thermostatted water jacket. Blood sera were analysed for glucose at a rate of 300 samples h-I simply by injecting them into a pH 7 phosphate buffer stream and passing them through the enzyme column.Fig. 1 shows typical recorder outputs for glucose calibration and two control blood sera. The results for the sera agreed well with the expected values. The manifold is so simple that it can easily be 5 rnin H 20 20 Fig. 1. Typical recorder output for standard glucose solutions followed by two control sera and standard glucose solutions automated and would be extremely suitable in a dinical laboratory for routine or emergency use, or both. Sucrose has been determined in a similar way, by incorpora- tion of an immobilized invertase - CPG column, prior to the GOD column: Sucrose D-Fructose + a-D-Glucose The glucose produced is then determined after passage through the GOD column, as before.The substrate for GOD, however, is P-D-glucose, therefore mutarotase is needed for the conversion of a- to 6-D-glucose. Various methods for the immobilization of mutarotase all gave unstable preparations, but immobilization of invertase and mutarotase together on CPG gave a stable product. The immobilization reactions were as shown in scheme 1. Fig. 2 shows the simple manifold used for sucrose determin- ation. Determination of Sucrose and Glucose In many food production processes, combinations of sugars, especially glucose and sucrose, are important, making their parallel measurement necessary for effective process control. Various enzyme electrodes2.3 have been suggested for this purpose, but such methods suffer from the disadvantages of poor reproducibility and long response and recovery times.The present system has been applied to the analysis of mixtures of sucrose and glucose. As shown in Fig. 2, a by-pass around the invertase column controlled by two 2-way keys enables the flow to pass either through both columns (for determination of sucrose and glucose) or through the GOD column (for determination of glucose only). The two sugars were determined within 35 s, both in standard mixtures (Table 1) and in soft drinks, by two sequential sample injections. The method is simple, fast, and reliable. a l n l n - z z $ s F CI L a - - al r: lu z u Time + Fig. 3. Typical output for the simultaneous determination of sucrose and glucose in a soft drink The system can be refined to determine sucrose and glucose in a single sample injection by splitting the sample so that it simultaneously flows through the invertase and GOD columns, and just the GOD column (Fig.2). The portion of sample not passing through the invertase column is delayed so that it reaches the GOD column after the other portion. Thus, two8 ANALYTICAL PROCEEDINGS. JANUARY 1985, VOL 22 OEt H O OEt H O H I I I NH2 + OHC- (cH2)3- CHO + CPG- 0 -Si -(CH2)3-N-C I CPG-O-Si-(CH2)3 -N-C b E t OEt (A) OEt H H I N=C-(CH2)3-CH =N-Enz I I (A) + H2N-Enz -+ CPG-0 -Si-(CHp),-N-C OEt Scheme 1. from sucrose and glucose, the second from glucose only, as can Such systems are currently being applied to many other Table 1. Analysis Of mixtures Of sucrose and glucose.Glucose, be Seen from Fig. 3. Again, fast, accurate results are obtained, 0.5 mM; signal, 0.433 PA. Current peak height given in FA Sucrose/ Combined Glucose Sucrose Pure sucrose oxidases* mM signal signal in mixture signal 1 .0 0.887 0.443 0.443 0.420 References 2.0 0.980 0.420 0.560 0.560 1. Carr, P. W., and Bowers, L. D., “Immobilized Enzymes in 3.0 1.260 0.420 0.841 0.794 Analytical and Clinical Chemistry,” John Wiley, New York, 4.0 1.401 0.420 0.981 0.981 1980. 5 .0 1.587 0.420 1.168 1.121 2. Pfeiffer, D., Scheller, F., Janchen, M.. and Bertermann. K.. Biochimie, 1980. 62, 587. Scheller, F., and Renneberg. R., Anal. Chirn. Acra, 1983, 152. 265. 3. signals are obtained in succession from the detector, the first Studies of Tin Oxide Semiconductors as Novel Gas-chromatographic Detectors S.J. Rowley, L. Ebdon and M. M. Rhead Department of Environmental Sciences, Plymouth Polytechnic, Drake Circus, Plymouth, PL4 8AA and D. A. Leathard Department of Chemistry, Sheffield City Polytechnic, Pond Street, Sheffield, S 7 7 WB The effect of gases upon the conductivity of certain semicon- ductors was first noted in the early 1950s.’ This phenomenon has since been developed in the manufacture of sensors in gas alarm systems.2 Although a certain selectivity to different gases can be induced by controlling the composition of the sensor during manufacture, or varying the operating conditions, this is insufficient to be of any real analytical value. However, combination of the high sensitivity of semiconductor sensors with the selectivity of gas chromatography (GC) can provide a practical analytical technique with many applications.There are several semiconducting materials which react with gases; the most popular are the oxides of zinc and tin, because of their sensitivity, and certain organic semiconductors for their selectivity. This paper is concerned only with tin(1V) oxide (SnO,) devices. Instrumentation There are several commercially available sensors designed for environmental monitoring and alarm systems. Some of these are readily convertible to GC detectors. A typical device is shown in Fig. 1. The wires connecting the electrodes and heater to their relevant circuits also serve as supports suspending the bead from pins set into a ceramic base. As a GC detector the device must be positioned in the post-column gas stream.To accomplish this a housing consist- ing of a rectangular block has been fabricated in stainless steel. A longitudinal hole is drilled through the housing and tapped at its ends to take the coupling that will connect it to the chromatographic column. A hole is drilked into the side of the block, enabling the sensor to be inserted in such a way that the bead protrudes into the gas flow. All fittings are gas tight. In order to complete the instrumentation a stable power supply is required for the integral sensor heater and precise amplification coupled with an ordinary chart recorder or integrator for data output. The column, oven, carrier gas control and sample injection port are conventional. Although no fuel gas is required, as with the flame ionisation detector (FID), the semiconductor detector does need a supply of oxygen.The device can actually be run in an oxygen-free atmosphere, in fact the response may be enhanced in this mode, but the sensor will eventually be reduced and degrade. There is probably an optimum level of oxygen for maximum response and sensor life, but as the sensors were originally designed to run in air, this provides a cheap and convenient carrier gas. Unfortunately, air limits column packings to those which are not damaged by oxidation, mainly solid absorbents It is, however, possible to use other types of packings, and an inert carrier gas, by the addition of oxygen as a post-column make up, either continuously or between runs. e H ntered tin oxide Alumina t u b e Fig.1. Diagram showing the structure of a typical Figaro gas sensorANALYTICAL PROCEEDINGS, JANUARY 1985, VOL 22 9 Operation Semiconductor detectors are temperature dependent, there- fore all variables affecting the surface temperature of the device must be controlled and optimised. There are four such variables, sensor heater voltage, detector flow, carrier gas flow and column temperature. Sensor heater voltage (V,) has the most effect upon the surface temperature and thus determines the over-all sensitiv- ity and selectivity to particular compounds. The surface temperature is proportional to VH as is, generally, the sensitivity. For example, when VH = 3 V the TGS 813 detector exhibits equal sensitivity to methane and propane, at V , = 4 V response to propane is greater than to methane, and at VH = 6 V this order is revzrsed. Hence, it is possible to tune the system to a certain extent according to the application.The other three variables affect the sensor temperature by thermal mass transport. Generally a hotter gas at a low flow-rate will result in a higher surface temperature, and consequently response, than a cool gas at .a high flow-rate. Column temperature and carrier gas flow are also the major variables controlling the chromatographic separation and peak shape. Detector flow can be controlled by the provision of a controlled split, or make up, between the column and the detector. All four variables are interdependent in their effect upon the sensor operation. Optimisation has been achieved by using the variable step size simplex technique.3 Peak height response was chosen as the criterion of merit, rejecting vertices not giving adequate component separation.4 4 2 co -1 Fig.2. Chromatogram showing exhaust gas analysis. Conditions: activated charcoal 2 m x 4 mm at 97 “C; carrier gas, air at 100 ml min l . Sensor, TGS 815, VII = 6 V Fig. 3. Chromatogram showing separation of hydrogen sulphide and methane. Conditions: 2 m x 4mm, activated alumina at 170°C. Carrier gas, air at 40ml min-I. Sensor. TGS 816, VIT = 6 V Applications Figs. 2, 3 and 4 are chromatograms demonstrating some of the features and advantages of semiconductor detectors. The first shows the analysis of petrol engine exhaust gases, providing useful information for both tuning and pollution control.The ability to analyse these three compounds, hydrogen, carbon monoxide and methane, on one column and with a single detector in one run, illustrates one of the main advantages- versatility. The detection limit for hydrogen of 40 p.p.b. clearly demonstrates the second advantage, namely sensitivity. Fig. 4. Chromatogram showing separation of nitrogen and argon. Conditions: 2 m x 4mm, molecular sieve 5A at 90°C; carrier gas, oxygen at 30mlmin-1. Sensor, TGS 813, VH = 6.23 V This particular separation was popular in early applications of the semiconductor detector, e . g . , determination of these compounds as components of “firedamp” in mine air analy~is,~ and analysis of these gases in fjord sediments. The robust nature of the device and the low cost of the carrier gas make it ideal for portable instrumentation.The second chromatogram (Fig. 3) shows the separation of hydrogen sulphide and methane, demonstrating resistance to poisoning, an advantage of this detector over other solid-state devices (i. e. , catalytic bead). The sensor manufacturer claims sensitivity to a wide variety of organic and inorganic compounds.6 The range of organic compounds includes alkanes, alkenes, alcohols, ketones, amines, acetates and freons. The semiconductor detector obviously has the potential to replace the FID in many applications. The following gases can also be detected: hydrogen, carbon monoxide, carbon dioxide, sulphur dioxide, hydrogen sulphide and ammonia. Fig. 4 shows the analysis of a mixture of argon and nitrogen.Sensitivity to these compounds is very low compared with the reducing gases, but indicates a possible thermal conductivity mechanism of response in addition to, but at a lower level than, the accepted reduction - valence bond promotion mechanism. The sensor response to other gases is known to be complex and current theories7 consider it likely that four or more mechanisms are involved simultaneously. Conclusions The semiconductor detector for GC is a versatile, sensitive, safe and robust device. It is a relatively low cost item with low running costs. In many applications it can replace the FID or thermal conductivity detector (TCD) and multi-detector arrangements. The robust nature of the detector and its non-reliance on bottled gas, when operated with an air compressor, make the device suitable for field operation.There are many applications in the industrial, mining, environ-10 ANALYTICAL PROCEEDINGS, JANUARY 1985, VOL 22 mental and medical fields as a portable or ambulatory 3. instrument.8 4. In the laboratory the potential of semiconductor detectors in gas analysis and replacement of multidetector systems are 5 * diverse and numerous. References 6. Brattain, W. H., and Bardeen, J., Bell Syst. Tech. J . , 1953,32, 7. 1. 8. Watson, J. , and Tanner, D., Radio Electron. Eng., 1974,44,85. 1. 2. Nelder, J. A., and Mead, R., Comput. J., 1965, 7 , 308. Ebdon, L., Ward, R. W., and Leathard, D. A., Analyst, 1982, 107, 129. Leathard, D. A., Wynne, A., and Ebdon, L., RSC Interna- tional Conference on “The Detection and Measurement of Hazardous Substances in the Atmosphere,” City University, London, 1982.Manufacturers literature, Figaro Engineering, Japan. Morrison, S. R., Sensors Actuators, 1982, 2, 329. Christman, N. T., and Hamilton, L. H., J. Chromatogr., Biomed. Appl., 1982, 229,259. The Effect of Materials of Clinical Interest on Calcium lon-selective Electrode Response Sajedah A. H. Khatil, G. J. Moody, G. de Oliveira Net0 and J. D. R. Thomas Applied Chemistry Department, Redwood Building, UWlST, P.O. Box 13, Cardiff, CF1 3XF The use of potentiometry in clinical chemistry has been widened by advances in ion-selective electrode technology. For calcium the recent advances in electrode design’ have greatly improved the scope of such measurements. However, the accurate use of calcium ion-selective electrodes is affected by proteins.2 Also, marry studies have been directed to determin- ing the clinical circumstances in which direct potentiometry with ion-selective electrodes gives different values and also determining which conditions lead to more accurate values for clinical use.3 The purpose of the present work has been the study of interferences by biochemical type materials on calcium ion- selective electrodes based on calcium bis{di-[4-( 1,1,3,3- tetramethylbutyl)phenyl]phosphate} in conjunction with each of dioctylphenylphosphonate (DOPP) (electrode I), tripentyl phosphate (TPP) (electrode 11) and trioctyl phosphate (TOP) (electrode 111) as plasticising solvent mediator.Experiments have also been carried out on electrodes made from mem- branes obtained from a commercial supplier (electrode IV).This study is merited since, on the whole, there has been little study of interference effects of materials of biochemicaI interest on calcium ion-selective electrodes to match those of the rather unusual interference on calcium ion-selective electrodes by anionic surfactants.4.5 Experimental Poly(viny1 chloride) (PVC) matrix membrane electrodes with inner solutions of 10-1 M calcium chloride were assembled by previously described procedures637 and calibrated against a saturated calomel reference electrode with serially diluted calcium chloride solutions (10-1-10-5 M range) in water or 0.15 M sodium chloride solution. The electrodes (I, 11, 111 and IV listed in the introduction) were tested for interference from various biochemical materials by adding 0.05-cm3 aliquots of a solution (5 x 10-2 M) of the interfering component under study to a calcium chloride containing solution (25 cm3).The calcium chloride solutions were of 10-2 or 10-3 M concentration, made up in either water or 0.15 M aqueous sodium chloride. E.m.f. readings were noted for each aliquot added until the background calcium chloride solution contained at least a 10-3 M concentration of interfer- ent. Because of their limited solubility in water, cholic acid, lecithin, cholesterol and vitamin D2 were dissolved in ethanol and propan-1-01, although cholesterol is only slightly soluble in ethanol. The effect of alcohol alone on each electrode was also studied. Table 1. E.m.f.changes for cells with calcium ISEs corresponding to various components added to test solutions containing calcium AE caused by added components to solutionslmv Component of Concentration 10-3 M or 0.5 cm3 alcohol added to 25 cm3 solution SDS DBSS Cholic acid (in propanol) Cholic acid (in ethanol) Cholesterol (in propanol) Cholesterol (in ethanol) Lecithin (in proganol) Lecithin (in ethanol) Vitamin D2 (in propanol) Vitamin D2 (in ethanol) Urea Glucose Imidazole Ethanol Propan-1-01 (I) Dioctylphenylphosphonate (11) Tripentylphosphate (111) Trioctylphosphate (IV) Philips membrane electrodes (DOPP) electrodes (TPP) electrodes (TOP) electrodes 1 , CaCl, in CaCl, in CaCl, in CaCl, in CaC12 in CaC12 in CaC12 in CaC12 in water 0.15 M NaCl water 0.15 M NaCl water 0.15 M NaCl water 0.15 M NaCl -------- 10-2 M 10-3 M 10-2 M 10-3 M 10-2 M 10-3 M 10-2 M 10-3 M 10-2 M 10-3 M 10-2 M 10-3 M 10-2 M 10-3 M 10- 2 M 10-3 M -55.0 -65.8 -85.0 -85.7 -63.0 -32.0 -40.0 -15.0 -9.1 -3.3 -12.5 -9.7 -35.9 -17.0 -41.5 -24.0 -10.4 -6.7 -24.3 -8.8 -26.0 -16.9 -10.0 -12.7 -17.1 -16.3 -13.1 -17.4 -9.6 -11.6 -6.8 -11.3 -3.0 -11.2 -3.1 -2.3 -21.0 -8.8 -12.6 -10.1 -3.1 -5.1 +4.1 +2.5 -2.8 -3.2 -0.6 -3.5 -8.6 -3.7 -6.4 -6.4 -10.0 -9.2 -7.3 -10.9 -2.4 -13.7 -5.6 -6.5 -5.4 -5.3 -10.5 -9.8 -3.5 -8.5 -5.3 -5.3 -10.3 -6.2 -11.2 -10.9 -2.4 -3.8 -3.0 -2.8 -5.0 -3.5 -6.5 -7.0 0.0 +0.1 -0.9 +0.2 +0.3 -0.8 +0.8 -0.1 0.0 +0.2 -0‘2 -0.1 0.0 0.0 -1.1 +0.4 -1.0 +2.2 0.0 -5.0 -2.2 +1.0 +0.3 -1.0 -2.8 -2.7 -4.8 -3.8 -2.6 -3.8 -2.5 -3.5 -6.7 -13.0 -9.1 -8.2 -10.5 -15.6 -8.0 -10.2 -9.0 -2.0 -4.0 -0.8 -2.0 +1.1 -5.0 + 1.0 -4.1 +2.0 0.0 0.0 +1.2 +3.7 +3.4 - 17.3 -2.0 -0.42 -2.6 +1.6 +1.6 +0.4 +1.6 -5.6 -7.8 -0.1 -0.3 +0.6 +4.8 +5.1 -2.8 -7.2 -102 -3.8 -1.5 -55.9 -2.2 -2.8 -2.4 -0.6 +1.2 -7.1 +1.7 +0.2 -3.5 -0.2 +1.3 -1.0 +0.2 +0.2 -0.6 -0.2 -7.6 -3.6 -0.2 +0.2 -1.7 -0.3 -0.1 -4.5 -3.8 -1.3 -1.7 +2.1 +3.3 -0.3 +3.2 +2.2 -2.3 +0.5 +1.9 +0.3 +1.0 +2.8 +1.0 -73.5 -26.5 -25.1 -93.3 -39.8 -59.8 -0.5 -1.2 -1.6 +3.3 -1.6 +0.4 -1.7 -2.2 -0.3 -0.6 0.0 -1.8 -2.6 -2.2 +0.6 -4.0 -1.5 -3.1 +0.9 -0.1 -0.8 0.0 -0.5 -0.5 +1.2 +0.7 -1.1 +0.2 +0.5 +0.1 -2.0 -2.0 -3.0 +0.8 +0.4 +3.3 +2.0 +5.7 +1.7ANALYTICAL PROCEEDINGS, JANUARY 1985, VOL 22 11 Results and Discussion E.m.f.changes of less than 0.5mV were observed for up to 10-3 M of starch, sucrose, uric acid, creatinine and bilirubin for all four electrodes.Table 1 summarises the e.m.f. changes corresponding to the presence of added interferent up to a concentration of 1 0 - 3 ~ ~ or of 0.5cm3 of alcohol added to 25 cm3 of the background calcium chloride solution. As was reported earlier,5 and confirmed here, electrode compositions represented by electrode I11 with a calcium bis{ di-[4-( 1,1,3,3-tetramethylbutyl)phenyl]phosphate} sensor and trioctyl phosphate solvent mediator are much superior to other membrane systems in resisting interference by anionic surfactants, such as sodium dodecylsulphate (SDS) and dodecylbenzene sodium sulphonate (DBSS) (see Table 1). Similar efficacy i s also shown by electrode 111 in the data given in Table 1 for the biochemical interferents from among those electrodes based on the organophosphate sensor (electrodes I, I1 and 111).For the biochemical interferents, note must be taken of the effect of ethanol and propan-1-01 interference on electrode response (Table 1), which will be superimposed on the interferences of the biochemical solute. Propan-1-01 seriously affects electrodes I and 11, while the effect of added alcohol in other instances is less than 5 mV for 0.5 cm3 of added alcohol. Different from the effect shown by other electrodes, the e.m.f. of electrode I11 increases when alcohol is added, and this may be a contributory factor in the apparently small interferences shown by the biochemical components recorded in Table 1. The data for electrode 1V (Philips membrane) show this electrode to be as little affected by added biochemical components as that of electrode 111.This is contrary to the experience with electrode IV for anionic detergent surfactant (see Table 1 and reference 5). The nature of the observed interferences for biochemical materials may be due to calcium binding sites or to surfactant properties of at least some of the materials, e.g. , lecithin. More generally, lipoproteins that contain cholesterol and phospho- lipids are surface active agents, so that the kind of interferences noted here are of interest in relation to the role of calcium ion-selective electrodes and, indeed, ion-selective electrodes in general in the clinical field. Conclusion Similarly to the observations with anionic surfactants,S the replacement of trioctyl phosphate (electrode 111) for dioctyl- phenylphosphonate (electrode I) solvent mediator in the PVC calcium ion-selective electrode based on bis{di-[4-( 1,1,3,3- te trame t h y 1 buty1)phenyllphuspha te} sensor can reduce inter- ferences from components likely to occur in body fluids, but there must be regard to the effect of alcohol used for dissolving interferents.Electrodes made from commercial membranes (electrode IV) behave similarly to electrode I11 with respect to relative freedom from interference by the biochemical materials studied. However, before making firm recommenda- tions on the relative merits of these electrodes, further studies have to be made on the effect of inorganic ions, such as potassium, sodium, magnesium and zinc, on their responses.The authors thank the Foundation of Technical Institutes, Baghdad, Iraq, for paid leave of absence and a studentship (granted to S. A. H. K.) and the Conselho Nacional de Desenvolvimento Cientifica e Technologico, Brazil, and the British Council, London, for financial support (to G. de. 0. N.) References 1. 2. Moody, G. J . , and Thomas, J. D. R., Ion-Sel. Electrode Rev., 1979,1,3. Payne, R. B., “International Symposium on Electroanalysis in Biomedical, Environmental and Industrial Sciences, UWIST, Cardiff, 5-8 April 1983,” paper 13. Ladenson, J. H.,Anal. Proc., 1983,20,554. Craggs, A., Moody, G. J . , Thomas, J. D. R., and Birch, B. J., Analyst, 1980,105,426. Frend, A. J., Moody, G. J., Thomas, J. D. R., and Birch, B. J., Analyst, 1983,108,1072. Moody, G.J . , Oke, R. B . , andThomas, J. D. R.,Analysr, 1970, 95, 910. Craggs, A., Moody, G. J., andThomas, J. D. R., J . Chem. Educ., 1974,51,541. 3. 4. 5. 6. 7. Piezoelectric Quartz Crystal Detection of Ammonia Colin S. 1. Lai, G. J. Moody and J. D. R. Thomas Applied Chemistry Department, Redwood Building, UWIST, P.O. Box 13, Cardiff, CFI 3XF Coated piezoelectric crystal detectors have developed into a highly sensitive technique for the detection of traces of atmospheric pollutants, 1 The detection of a specific component is observed as a change in the resonant frequency of the crystal as a result of selective sorption by the coating. Sauerbrey2 derived an expression relating the mass of metal films deposited on quartz crystals to the change in frequency. For commercially available crystals the form of this equation is AF = -2.3 x 1OhFAm/A, where AFis the change in frequency (Hz), F i s the initial frequency of the quartz plate (MHz), Am is the mass sorbed (8) and A is the area of the coating (cm2).From this it can be predicted that a 9MHz crystal with an approximately 0.5 cm2 coating would have a mass sensitivity of up to about 400Hzpg-1 to yield a detection limit of about Application of the piezoelectric quartz crystal detection method is decribed here for ammonia, using pyroxidine (vitamin Bb) hydrochloride as the detecting coating material. 10-’2g. Ammonia Detection with Pyridoxine Hydrochloride The role of Antarox CO 880, a nonylphenoxypolyethoxylate with 30 ethoxylate units, as a support matrix for a pyridoxine (vitamin B6) hydrochloride coating on a piezoelectric crystal for detecting ammonia has previously been described.3 It was shown that the Antarox CO 880 was an effective agent for prolonging the useful life of the piezoelectric ammonia detector.3 Thus, the frequency change corresponding to 30p.p.m.of ammonia for a piezoelectric crystal coated with pyridoxine hydrochloride alone dropped from about 300 Hz on day 2 to less than 50 Hz on day 6, while the frequency change for a piezoelectric crystal coated with pyridoxine hydrochloride supported by Antarox CO 880 remained at about 200Hz for over 50 days for samples containing the same level of ammonia.3 A similar pattern has been observed in the present study, the sample size being 1 cm3 in each instance. For larger samples (5cm3) it is possible to follow the frequency changes for more dilute samples using the apparatus previously described3 with a 9 MHz resonant frequency AT cut quartz crystal with a gold electrode.Again, the Antarox CO 880 matrix permits a long useful life for the piezoelectric detector. Typical results are summarised in Table 1 for p.p.b. (pg dm-3) levels of ammonia, the blank values corresponding to background moisture.12 ANALYTICAL PROCEEDINGS. JANUARY 1985, VOL 22 Table 1. Frequency changes (AF) in hertz corresponding to responses to ammonia-containing samples of a piezoelectric quartz crystal coated with pyridoxine hydrochloride and Antarox CO 880. Slopes of log AF versi~s log[NH3] graphs range from 0.156 to 0.206 Ammonia gas, p.p.b Period/d Blank 463 56 6.7 0.80 0.0% 0.012 1 28 325 211 137 86 59 43 2 20 362 233 125 93 61 41 3 18 299 189 125 94 60 44 20 232 169 123 82 52 42 I q 5 8 24 252 274 113 83 51 - 10 24 211 152 121 79 56 -:I Interferences to Ammonia Detection Tnterferences from other gases in the assay of ammonia are listed in Table 2 for Antarox CO 880 - pyridoxine hydro- chloride coatings.Table 2. Frequency changes (AF) for various interfering gases and controls for piezoelectric crystal coatings of Antarox - pyridoxine hydrochloride A FtHz for 5 cm3 samples passed over crystal coatings Gas NH3 NH3 SO? HCI coz HZS NO? Room air Dried room air Triethylarnine Triet h ylarnine Gas concen- tration, p.p.m. 0.463 32 101 75 109 1480 116 - - 3.6 36.4 Pyridoxine hydro- Pyridoxine chloride hydro- on Antarox chloride CO 880 199 320 21 44 47 35 45 1496 19 32 27 43 10 16 0 0 21 1 383 - - Antarox CO 880 45 34 32 2674 30 41 11 0 12 41 - The data indicate that, except for hydrogen chloride on the Antarox CO 800 containing coatings, the extent of interference from the various gases is low.The high response to hydrogen chloride by the Antarox CO 880 containing coatings is attributed to the formation of hydrogen bonds with the ethoxylate oxygen atoms. The fact that the frequency returns to the base line value indicates that this interaction is reversible. However, there would be interaction between hydrogen chloride (or acidic component) and any ammonia in a sample stream. The large frequency change brought about by hydrogen chloride for the Antarox CO 880 coated crystal suggests that Antarox CO 880 could be a suitable piezoelectric sensor for this gas and is now under investigation. The large frequency change brought about by hydrogen chloride for the Antarox CO 880 coated crystal suggests that Antarox CO 880 could be a suitable piezoelectric sensor for this gas and is now under investigation.Application of the Sauerbrey Equation The analytical utilisation of coated piezoelectric quartz crystal detectors has been based on the assumption that Sauerbrey’s equation (quoted above) is valid, that is, that the mass increase caused by sorption is analogous to the concentration of the sample in the flowing gas stream and is proportional to the decrease of the resonance frequency. 1 However, analysis of previously obtained data by Beitnes and SchroderJ shows that the sensitivity of piezoelectric crystal detectors for flowing gas streams does not obey the Sauerbrey equation.Thus, despite the considerably poorer sensitivities that should follow from incomplete sorption on the crystal coating, mixing effects with carrier gas, etc., the observed decreases in frequency are often greater than the value corresponding to Sauerbrey’s equation. Conversely, observed decreases in frequency relate. according to the Sauerbrey equation, to mass changes that are greater than can be sorbed under the prevailing conditions. and in some instances greater than the amount of the component sought in the sample (see Table 3). Table 3. Comparison of ammonia present in 5cm-7 sample according to dilution and ammonia calculated from the frequency change by the Sauerbrey equation to be sorbed on the piezoelectric crystal coating.Slope of log AF versus log[NH3] graph = 0.194 for 300-0.3 p.p.b., and 0.355 for 3000 and 300 p.p.b. Concentration of NH, in sample, p.p.b. 3848 463 56 6.7 0 80 Ammonia calculated to be (for A = 0.549 Ammonia sorbed on coating present in 5 cm3 A F/Hz sample/pg cm2)/pg 289 19.2 0.85 136 2.30 0.40 84 0.28 0.25 56 0,033 0.16 39 0,004 0.11 Even greater anomalies than exist between columns 3 and 4 of Table 3 occur for higher values of AF, e.g.. for the 0.01 p.p.b. ammonia sample of Hlavay and Guilbault’ with a AF of 386Hz, the ammonia calculated to be sorbed by the pyridoxine hydrochloride coating is about 1 pg compared with the 0.000 05 pg deemed to be present in the 5 cm-7 sample used.The log A F versus log AC graphs (where Cis concentration) should have a slope of unity. Some previously reported double logarithmic plots for the piezoelectric detection of ammonia5 gave slopes of 0.0615 for a coating of L-glutamic hydrochloride and 0.0978 for a coating of pyridoxine hydrochloride. In the present study, the slopes are also considerably less than unity, being between 0.156 and 0.206 for the data in Table 1 and 0.194 for those in Table 3 lying in the p.p.b. range and 0.355 at the higher end of the range. The low slopes cannot readily be explained, although the fall-off in the response slope with time implies that the quality of the coating has a role. Data such as the above have led Beitnes and Schroder4 to investigate the systematic errors in the syringe dilution method because of the likelihood of sorptions on the syringe walls between dilutions.However, the magnitude of the anomalies between columns 3 and 4 in Table 3 and the other data cited above, and the ease with which blank values can be obtained from syringes that have been used for syringe dilution, suggest that other factors are involved. Alternative dilution methods, such as bottle dilution, also give sensitivities that are better than predicted.4 Clearly the above observations call for further study and Beitnes and Schroder are already looking at the nature of the coating and the calibration methods in order to account for such deviations from Sauerbrey’s fundamental equation. Conclusion The use of a nonylphenoxypolyethoxylate (Antarox CO 880) as a support polymer prolongs the life of pyroxidine hydro- chloride as a sensitive sorbent coating during the piezoelectric crystal detection of ammonia.However, the matrix system is a source of possible interference from hydrogen chloride gas, the high level of interference suggesting a role for the poly-ANALY‘I’ICAL PROCEEDINGS. IANUARY 1985. VOL 22 13 ethoxylate as a selective piezoelectric crystal coating for hydrogen chloride. The extreme sensitivity of the piezoelectric crystal detection of ammonia to below the p.p.b. range (into the parts per 10’’ range) and its nature are inconsistent with the Sauerbrey equation, which normally applies to straightforward depo- 1. 2. 3 , 4. sitions on piezoelectric transducers. This merits the further studies now in progress.J 5 .References Guilbault. C;. G., lon-Sel. Electrode Rev., 1980. 2, 3. Sauerhrey. G . Z . . 2. I’hys.. 1059. 155, 206. Moody. 6. J . . Thoma\. J . D. R . , and Yarmo. M. A., Anal. Chini. Acta. 1983, 155. 225. Beitnes, H . . and Schroder. K . , Anal. (’him. Acta, 1984. 158,57. Hlavay, J . . and Guilbault. C . G.. ,4nal. Chem.. 1078. 50, 1044. Solvent Extraction Studies of Metal - Polyalkoxylate Complexes in Relation to Electrochemical Response Philip H. V. Alexander, Gwilym J. Moody and J. D. R. Thomas Applied Chemistry Department, Redwood Building, UWiST, P.O. Box 13, Cardiff, CF1 3XF The tetraphenylborates of metal ion - polyalkoxylate com- plexes are well established as ion-selective electrode sensors for cations such as barium and calcium.when the sensors are incorporated with a suitable solvent mediator in a liquid membrane or a PVC matrix membrane. The PVC matrix membrane electrodes have also found use in the measurement of polyalkoxylates in solution and for the determination of the critical micelle concentrations (cmc) of polyalkoxylate non- ionic surfactants. 1.2 The mechanism of the potentiometric response of the electrodes towards polyalkoxylate has not been characterised, but it is assumed to involve barium. Also, it is known that the tetraphenylborate of the barium complex with the nonylphe- noxypolyethoxylate, which has just 4 ethoxylate units (Antarox CO 430), is a more effective sensor for alkoxylates in solution than is the tetraphenylborate of the longer chain analogue, Antarox CO 880, which has 30 ethoxylate units and which is better at selectively sensing barium ions.’.’ These observations suggest that the electrode response varies accord- ing to the stability of the metal - polyalkoxylate complex concerned.An earlier study3 has shown that solvent extraction proce- dures are helpful in determining the relative order of stability of these metal ion - polyalkoxylate complexes. Here. the cationic complex formed in the aqueous phase is extracted into the organic phase, together with an easily polarisable coloured anion, such as picric acid. The amount of complex thus extracted is determined spectrophotometrically . In the present study, the bulk extraction coefficients of several divalent cation - polyalkoxylate systems have been determined and related to the response of PVC matrix membrane electrodes prepared from the tetraphenylborate salts of certain of these complexes.Solvent Extraction Studies The polyalkoxylate systems studied are listed in Table 1, together with their principal features. In addition to Antarox CO 430 and Antarox CO 880, which had previously been used in electrodes, these included Antarox CO 730 to exemplify a polyethoxylate of intermediate chain length, PEG 1500 as a close analogue to Antarox CO 880 but lacking the hydrophobic grouping, and two propoxylates, Glucam P10 and Glucam P20. The bulk extraction constants into dichloromethane with respect to barium, magnesium and zinc ions were determined for each polyalkoxylate, according to the principles and method previously described’ for picrate as the coloured polarisable anion, except that in this study dipicrylamine was the coloured anion.Dipicryalamine was selected because, except for Antarox CO 880 and Antarox CO 730, little or no extraction was observed with picrate. It has previously been shown that dipicrylamine extracts crown ether - metal com- plexes about one hundred times more powerfully than picric acid.4 Also, the spectrophotometric absorption coefficients in the aqueous (at 429nm) and organic (at 421 nm) phases are much greater for this anion. Table 2 summarises the principal solvent extraction data and the bulk extraction constants, Ki, the units of K , being related to the stoicheiometry of the relevant metal ion - polyalkoxylate complex. It is difficult to compare the different groups of Ki data because of the differences in units.Also, much higher concentrations of alkoxylate have had to be employed for Antarox CO 430, Glucan P 10 and Glucam P 20 in order to obtain sufficient extraction of the dipicrylamine. Nevertheless, it can be seen that the metal ions complex strongly with Antarox 430 and Antarox CO 730 when compared with Glucam P 10, although the low complexation of zinc ions with Antarox CO 730 is anomalous. Barium forms the strongest complex for each polyalkoxylate except for the Glucam P 10 systems, where the magnesium complex is the strongest. Zinc forms the weakest complexes of each group except for Antarox CO 430 and Glucam P 20. Potentiometric Studies PVC matrix membrane electrodes based on sensors made from the tetraphenylborates of the various metal polyalkoxylates in 2-nitrophenyl phenyl ether as solvent mediator confirm that the best barium ion-selective electrodes are based on Antarox CO 430, as previously described.1.2 The electrodes recommen- ded for use in ethoxylate analysis in aqueous solutions are those based on a “liquid ion-exchanger,” consisting of a Table 1. Features of polyalkoxylates studied Commercial name Antarox CO 430 Antarox UO 740 Antarox CO 880 PEG 1500 Glucam I’ 10 Glucam P 20 Nature o f alkoxylate and number of units Ethoxylate (4) Ethoxylate (15) Ethoxylate (30) Ethoxylate (34) Pro pox y la t e (10) P r opox y I at e (20) Nature of hydrophobe Nonylphenoxy Nonylphenoxy Nonylphenoxy None Methylglucoside Methylglucoside Average relative molecular mass 396 880 1540 1500 775 1355 Ba2 + : AOU ratio 4 1s 12 10.5 10 814 ANALYTICAL PROCEEDINGS, JANUARY 1985.VOL 22 Table 2. Extraction of dipicrylamine into dichloromethane by polyalkoxylates in the presence of barium, magnesium and zinc ions with bulk extraction constant ( K , ) data. [M?+] = mol dm-3. Equal volumes (10 cm-i) of water and dichloromethane Alkoxylate Antarox CO 730 PEG 1500 Antarox CO 880 Antarox CO 430 Glucam P 10 Glucam P 20 [Alkoxylate]i 0.703 1.38 x 106 1.38 X 10' 0.55 0.600 3.61 0.70 0.27 0.636 2.29 0.58 0.04 7.58 1.94 x 10' 6.68 x 10' 10-5 mol dm-' K, 1.59 X 103 1.12 47.2 184 7.86 2.07 9.46 7.08 8.05 Dipicrylamine Units of K , extracted, Yo moi- I dm? 92.2 61.9 7.9 51.9 40.2 32.4 40.6 30.3 17.90 mol- 1 dm' 64.2 53.8 61.9 mol- dm' 20.6 28.8 16.4 mol-0 1 dml 2 62.2 59.4 60.1 mol- (1 -3 dm0 'I mol- (1 1 dml 2 Dipicrylamine extracted [alkoxylateji 106% mol- I dm' 13.1 8.8 1.1 8.7 6.7 5.4 6.4 3.8 2.8 0.85 0.71 0.82 1.8 2.6 1.5 3.0 2.9 2.9 - saturated solution of the tetraphenylborate of barium - Antarox CO 430 complex in 2-nitrophenyl ether.Table 3 illustrates the responses to barium, magnesium and zinc ions of PVC matrix membrane electrodes that have been prepared using the tetraphenylborates of selected metal polypropoxylates in complexes involving Glucam P 10 and P 20 as sensors with 2-nitrophenyl phenyl ether as the solvent mediator. In no instance was a good response obtained for either magnesium or zinc, despite the indications from the K, data of Table 2 that complexes involving Glucam P 10 and Glucam P 20 might show potentiometric sensing properties for these ions.A barium ion response was evident in all instances, although these did not approach the quality of response characteristic of the established system5 based on the tetra- phenylborate of the barium complex with Antarox CO 880. Glucam P 10 and Glucam P 20 polypropoxylates cause a less than lOmV change in the potential over three decades of polypropoxylate concentration for electrodes based on sensors of either barium - Antarox CO 430 or barium - Antarox CO 880 complexes compared with the -30mV e.m.f. per decade change given by the Antarox polyethoxylates near the cmc region. However, electrodes employing polypropoxylate - barium complex sensors gave e.m.f. changes of -10mV per decade for Glucam P 10 and -25 mV per decade for Glucam P 20, but this is still dwarfed by the much bigger response (-120mV per decade) given by these electrodes towards Antarox CO 880.Correlation of Potentiometric and Solvent Extraction Studies Antarox CO 880 forms the basis of a very good barium ion-selective electrode. This accords well with the Ki value for the barium complex and with the good extraction of the complex into the organic phase, as depicted by the dipicrylam- ine extraction. Although this electrode system gives e.m.f. changes for alkoxylates in aqueous solutions, it is inferior to the Antarox CO 430 electrode system for this purpose. The better performance of the barium - Antarox CO 430 system as an electrochemical sensor for polyethoxylates may be due, in part, to the relatively poor affinity exhibited by the organic phase for this complex, as depicted by the dipicrylamine data in the last column of Table 2.Also, just four ethoxylate units are complexed to the barium cation, compared with the larger number of ethoxylate units complexed for other members of the Antarox series, which promote the transfer of the polyethoxylate being sensed towards the barium ions in the membrane. The high extractions of the dipicrylamine of the Antarox 730 and of the PEG 1500 systems do not lead to good electrode qualities for polyethoxylates, although the barium complex with Antarox CO 730 is a good potentiometric sensor towards barium ions. For the Glucam polypropoxylates, the somewhat higher extractability of the dipicrylamine of these systems compared Table 3. Responses of electrodes with metal - polypropoxylate sensors to selected divalent ions. No response times are quoted as stable responses are usually unobtainable Electrode sensor type Ba2+ (BaCL) MgZ + (MgC1,) ZnZ+ (ZnS04) (Glucam P 20)0.4 - Ba - TPB2 (Glucam P 20)" - Zn - TPB2 (Glucam P 20)(,,4 - Mg - TPBz Glucam P 10 - Ba - TPB, Poor, 5 x 10-4-10-1 -20 mV decade-' Linear response z= j x 10-5-10- 1 ~ 2 0 - 2 2 mV decade- Linear response = 19 mV decade - I Poor, linear response 5 x 10-"10 I = 18 mV decade- 1 =5 x 10-s-15 x 10-2 Minimal Minimal Minimal Poor Poor, non-linear, Minimal erratic Minimal Virtually noneANALYTICAL PROCEEDINGS, JANUARY 1985. VOL 22 15 with Antarox CO 430 systems is matched by the greater response given by the Glucam electrode system towards the polypropoxylate noted above. Also, the response of the Glucam electrode system towards Antarox ethoxylates is much greater in magnitude than is observed for the Antarox electrode systems, but the reproducibility and linearity of the AE versus log [Antarox] relation is poor. It is noted that Glucam P 10 and P 20 have fewer alkoxylate units per barium ion in the sensing complex than is the case for the barium complex with Antarox CO 880. Conclusion There are some inter-relations between the potentiometric electrode responses given by sensors based on the tetraphenyl- borates of metal complexes with polyalkoxylates and the solvent extraction parameters of the dipicrylaminates of these complexes. However, this does not adequately explain the superiority of electrodes containing the barium complex with Antarox 430 for sensing polyethoxylates, nor of the good barium ion-selective electrodes derived from the complex of barium with Antarox CO 880. More complete answers to these phenomena may lie in solvent extraction studies with tetra- phenylborates, rather than picrates or dipicryIaminates, and with 2-nitrophenyl phenyl ether soIvent rather than with dichloromethane, although this aromatic ether solvent is poor as a background spectrophotometric medium. The authors thank the Science and Engineering Research Council €or a studentship (to P.H.V.A.) under the Coopera- tive Awards in Engineering Scheme in conjunction with Unilever Research, Port Sunlight Laboratory. References 1. 2. 3 . 4. 5 . Jones, D. L., Moody. G. J.. and Thomas, J . D. R.. Analyst, 1981, 106, 439. Jones, D. L., Moody, G. J . , Thomas, J . D. R . , and Birch, B. J . , Analyst, 1981, 106, 974. Jaber, A. M. Y., Moody, G. J.. and Thomas, J . D. R., J . Inorg. Nucl. Chem., 1977, 39, 1689. Jawaid, M., and Ingman, F., Talanta, 1978, 25, 91. Jaber, A. M. Y.. Moody. G. J., and Thomas. J. D. R.. Analvst, 1976, 101. 179.