16 ANALYTICAL PROCEEDINGS. JANUARY 1089. VOL 26 Determination of Organotins in Fish and Sediments by Gas Chromatography with Flame Photometric Detection Isabel Martin-Landa, Fernando de Pablos and lain L. Marr Department of Chemistry, University of Aberdeen, Meston Walk, Old Aberdeen AB9 2UE Organotin compounds, particularly tributyltins (TBT), are used as the active components in antifouling paints for vessels and structures immersed in sea water. Because of their high toxicity they are the focus of increased concern as marine pollutants. The effective concentration of tin in the water on the antifouling paint surface approaches 1 pg ml-1, which is sufficient to prevent fouling organisms (algae or barnacles) from attaching themselves. However, very much lower concen- trations, v i z .. about 1 pg 1-1 in sea water, are lethal to the larvae of many non-target invertebrates. 1 Even lower concen- trations can affect the growth rate and reproduction of commercial shellfish.2 Because of these problems, the retail sale of TBT-containing paints was prohibited by the govern- ment in the Spring of 1987. and similar regulations have been introduced in the United States and France.3 TBT treated netting has been used on salmon farms. As farmed salmon are harvested after about 3 years, there are still fish in the farms that have been exposed to tributyltin. On the other hand, organotin compounds undergo degradation in the environment and in micro-organisms to give, eventually, non-toxic tin(1V). In addition, methylation reactions have also been reported,l.j leading to the possibility of obtaining various organotins.Depending on the pH and salinity, a certain amount of the organotins in water may be adsorbed on to particles and reach soils and sediments. presenting a hazard to marine species which live in, or lay eggs in, those sediments. I n all these types of samples in which organotins may be present. the concentra- tions of organotins are usually lower than 1 pg ml-1 and the matrices are complex. Hence, there is a need for highly sensitive methods for the determination of organotins in order to be able to monitor the levels of toxic organotins or to obtain information on their fate in the environment. Gas chromatography with flame photometric detection6.7 is a very good approach which provides the high sensitivity and good selectivity required for the determination of organotins in complex matrices such as fish tissue and sediments.The method consists of the following steps: ( 1 ) digestion of the sample; (2) extraction of the organotins from the acid digest; (3) chemical derivatisation to obtain derivatives that have higher vapour pressures; (4) separation by gas chromato- graphy; and ( 5 ) measurement of tin in the emission of the Sn-H species in the flame photometric detector (FPD). Several problems arose when we used this method. Taking into account concentrations of the organotins in the samples, the analyses were run for 5-25 g of sample and the final extract was made up to 1 ml. This means that several coextractives, mainly fats and oils, were present at a high concentration in the aliquots injected into the gas chromatograph, giving problems when their retention times were similar to those of the organotins.After overcoming these problems the method was applied to the determination of tributylin and its degradation products (dibutyltin, monobutyltin and the methyl derivatives) in salmon liver and sediments. Experimental Apparatus A Shimadzu GC8AFP gas chromatograph equipped with a flame photometric detector with an EM1 9601B photomulti- plier (range 300-800 nm. maximum at 380 nm), a 610-nm interference filter (Infrared Engineering) and a glass capillary column (12 m x 0.53 mm i.d. coated with a 3.0-um film of BP-1) was used. The injector and detector were held at 250 "C and the column was temperature programmed from 50 to 250°C at a heating rate of 10°C min-1.Nitrogen was used as the carrier gas at a flow-rate of 8.5 ml min-1 through the column and of 37 ml min-1 through the splitter. The detector was operated with a hydrogen-rich flame, the hydrogen and air flow-rates being 50 and 85 ml min-I. respectively. The detectorANALYTICAL PROCEEDINGS. JANUARY 1989. VOL 26 17 output was collected on a Pye Unicam PU4810 computing integrator. Aliquots of 5 yl were injected and peak heights were recorded. The amount of each organotin present was determined from suitable caiibration graphs using Me2PezSn as the internal standard. Reagents and Standards All reagents were of analytical-reagent grade or better. The purity and stability of the standards were checked by GC - FID and GC - FPD under the same operating conditions as those described above.All glassware was washed with aqua regia, doubly distilled water, ethanol and diethyl ether in that order. Standard stock solutions of the internal standard, Me2SnC12 (about 279 (1-g ml-1). were prepared in absolute ethanol. Procedure Weigh accurately between 5 and 25 g of wet salmon liver or sediment into a 250-ml separating funnel fitted with a PTFE tap. Add 40 ml of concentrated hydrochloric acid, shake for 2 h and leave to stand for a further 2 h . Then add 10 ml of hydrobromic acid (48%) and after 15 min extract the mixture with 100 ml of a 0.05% (mlv) solution of tropolone in pentane by shaking vigorously for 2 min. After 1 h separate the organic layer and dry it over anhydrous magnesium sulphate.Place the dried organic layer in a 250-ml round-bottomed flask together with a suitable amount of the internal standard (about 3 pg of Me2SnCI2). Wash the magnesium sulphate with 5 ml of diethyl ether and add this to the organic phase. To the dried organic extract add 20 ml of 1 M pentylmagnesium bromide in diethyl ether and reflux the mixture, with constant stirring. for 1 h . Destroy the excess of Grignard reagent by slowly adding 25 ml of 1 N sulphuric acid. Separate the organic phase and extract the aqueous phase with two 10-ml portions of diethyl ether. At this point add 100 ml of 3% sodium hydroxide solution to the organic phase, shake for 5 min and allow to stand overnight. 7 1 Fig. 1. Micro-evaporator Filter the organic phase into a 250-ml round-bottomed flask and concentrate it on a rotary evaporator under reduced pressure at room temperature to a volume of about 2 ml.Transfer this liquid into a micro-evaporator (Fig. 1). rinsing the walls of the flask with 5 ml of diethyl ether and concentrate to 1 ml under a stream of nitrogen. Use 5-p1 aliquots for the GC analysis. Results and Discussion Digestion and Extraction The use of concentrated hydrochloric acid for the digestion of samples containing organotin compounds has been reported already.8 Previous studies carried out in our laboratory have shown that no loss of the organotin compounds occurs during the treatment with concentrated hydrochloric acid; hence this acid was chosen for the digestion of the samples. The organotins were extracted from the acid digest into a tropolone solution in pentane.Tropolone was used to prevent the loss of highly polar organotins, i . 4 . . BuSn3+. Hydrobromic acid enhances the recovery of the butyltin species," probably because it aids the desorption of these species from the glass wall of the vessel. Derivatisation Procedure The n-pentyl derivatives were prepared, by means of a Grignard reaction, because some of the ethyl and propyl derivatives are sufficiently volatile to be lost during the routine concentration procedures. Derivatisation was carried out under reflux (about 40°C) to ensure quantitative reaction of the di- and monoalkyltins. If the magnesium used to prepare the Grignard reagent is contaminated with metallic tin, then tetraphenyltin will also be obtained after the derivatisation step.Clean-up The injection of an extract containing large amounts of fats and oils. 4.g.. from fish tissue or some types of sediments, into the gas chromatograph could lead to several problems such as background interference and serious damage to the column and detector. Despite the high selectivity of the flame photometric detector (FPD) towards tin-containing com- pounds, it was found that direct injection of an extract from fish tissue into the column gave chromatograms which contained a very large background interference in the range of retention times where the organotin compounds appeared [Fig. 2(ci)]. Ib) 0 5 10 15 20 0 5 10 15 20 Time min Fig. 2. ( a ) Chromatogram of a fith digest not treated with 3% NaOH solution ( h ) Chromatogram of a fish digest alter treatment w i t h 3% NaOH solution Also, the behaviour of the detector changed notably after the injection of such a contaminated sample, resulting in base-line drift.For these reasons, a clean-up procedure was necessary prior to injection of the samples. As the main components of the fish extract will be natural oils and fatty acids. their18 conversion into water-soluble substances by hydrolysis with sodium hydroxide was envisaged as being suitable. Treatment of the fish extract with a 3% (rnlv) solution of sodium hydroxide was found to be very effective in eliminating all potential interferences [Fig. 2(6)]. Concentration After the extraction and derivatisation steps the organotins were contained in a very large volume (-120 ml) of solution and at a low concentration and so a concentration step was necessary.Concentration of the samples by rotary evaporation in a Wheaton flask was attempted first in order to try and avoid the use of a number of different flasks with consequent losses of organotins. However, this technique was very difficult to perform because of continuous bumping of the solution through the system. A Kuderna - Danish concentrator was also tried but the process was very time consuming and was also rejected. Concentration to a volume of 1 ml was carried out, first in a rotary evaporator and then in a cylindrical flask which ended in a narrow calibrated tube (Fig. 1). Determination of Organotins in Salmon Liver To test the utility of the proposed method several salmon liver samples, obtained from a salmon farm and from experiments on the accumulation effects of tributyltin (carried out in the Marine Laboratory at Aberdeen), were analysed.The results are presented in Table 1, while Fig. 3 shows the chromatogram of one of the samples. 1 I JJ - I 5 k 1 1 0 5 10 15 20 Tirnelrnin Fig. 3. Chromatogram of a salmon liver sample. 1, MezSnPez (internal standard); 2. Bu,SnPe; 3, Bu,SnPe,; 4, BuSnPe,; and 5 , Pe,Sn Determination of Organotins in Sediment The same procedure was applied to the determination of organotins in a sediment. The results obtained were: Bus++, 2.17; Bu2SnZ+, 3.72; BujSn+, 0.64; Me2BuSn+, 12.92; and MeBu2Sn+, 11.06 p.p.m. The corresponding chromatogram is shown in Fig. 4. Conclusions The proposed method is suitable for either environmental analysis or laboratory studies on the degradation effects of small amounts of organotin compounds.From the analysis of salmon liver samples it is obvious that degradation of tributyl- tin by debutylation to give di- and monobutyltin occurs. However, it is difficult to conclude whether this degradation ANALYTICAL PROCEEDINGS, JANUARY 1989, VOL 26 actually takes place in the liver or in the water in which the salmon are living. Similar conclusions can be drawn from the results of the organotin analysis in sediments, the most notable feature of the analysis being the high concentration of methyl derivatives found in the sediment sample. The method used Table 1. Concentration of organotin species in salmon liver Organotin/pg g- 1 Sample BuSn3+ BuzSn'+ BuzSn + 1 0.54 2.01 0.38 2 0.24 1 .oo 0.28 3* 0.22 0.59 4.01 4* 0.04 0.05 0.13 5* 0.21 0.33 0.21 6 0.56 0.27 0.85 7 0.64 0.42 1.39 * Concentration experiments in tanks with: 1.0 (3); 0.1 (4); and 1.0 ( 5 ) pg g-1 of TBT in the water.~ here (GC - FPD) is more time consuming than other methods available for TBT ( e . g . , GF-AASX), but it provides consider- ably more information on the degradation and behaviour of these compounds in the environment. Lower values for the concentration of TBT in samples are usually obtained with GC - FPD than with GF - AAS as the latter is based on a final 0 4 8 12 16 20 24 Ti rne/rn i n Fig. 4. Chromatogram of sediment sample. 1. Me2BuSnPe; Me2SnPez (internal standard); 3, MeBu?SnPe; 4. Bu,SnPe; Bu,SnPe,; 6, BuSnPe,; and 7, Pe$n 2, 5 , measurement of the inorganic tin content.A similar result has been found by Short.3 In our view, contamination or poor selectivity during the extraction procedure can cause more serious errors in the GF - AAS method. 1. 2. 3. 4. 5. 6. 7. 8. 9. References Hall, L. W., and Pinkney. A . E., Crit. Rev. Toxicol., 1985, 14. 159. Alzieu, C.. Heral, M., Thiband, Y., Dardignac. M. J . , and Feuillet, M., Red. Trav. Inst. Pech. Marit., 1982. 45, 101. Short, J . W . , Bull. Environ. Contam. Toxicol., 1987, 39, 417. Fanchiang, Y. T., and Wood, J . M.. J . Am. Chem. SOC., 1981. 103, 5100. Hallas, L. E., Means, J . C., and Cooney, J . J . , Science, 1982, 215, 1505. Braman, R. S . , and Tompkins. M. A . , Anal. Chem., 1979, 51, 12. Maguire, R.J . , and Huneault, H . , J . Chromatogr., 1981, 209, 458. Mackie, J . C., Anal. Chim. Acta, 1987, 197, 303. Meinema, H. A . , Burger-Wiersma, T . , Verluis-de Haan, G., and Gevers, E. Ch., Environ. Sci. Technol.. 1978. 12, 288.ANALYTICAL PROCEEDINGS, JANUARY 1989, VOL 26 - Carrier stream Flow Injection Determination of Organosulphur Compounds with Chem ilum inescence Detection hotomultiplier - Chart tube recorder J. Steven Lancaster and Paul J. Worsfold School of Chemistry, University of Hull, Hull HU6 7RX Certain chemical reactions (usually oxidation reactions) yield products in an electronically excited state. which can return to the ground state by emitting a photon, as shown below: A + B + P* + P + light This process is known as chemiluminescence and its efficiency can often be increased by using a suitably fluorescent molecule.which has an excitation spectrum overlapping the emission spectrum of the donor molecule. This is shown below: A + B + S + P * + S - - + P + S " - S + l i g h t The intensity of the emission is proportional to the rate of the reaction; therefore. chemiluminescence can be used for quantitative analysis. Examples of chemiluminescence analysis include the cobalt catalysed oxidation of luminol the determination of mor- phine by its oxidation with permanganate3.' and the determina- tion of fluorescent molecules by the peroxyoxalate reacti0n.i Flow injection analysis (FIA) provides a useful experimental system for the study of chemiluminescence reactions because it permits rapid and reproducible mixing of the sample and reagents.The transient chemiluminescence emission can be detected by suitable control of the flow-rate and flow-cell volumeh and a sample throughput of >120 h-1 is typical. 1. Sensitivity. Extremely sensitive analyses are often possible because the absence of a source eliminates scattering and noise associated with the lamp. Also, there is little or no background emission. 2 . Wide dynamic range. I n chemiluminescence analysis, calib- ration data are often linear over several orders of magnitude, e . g . . the calibration data for morphine are linear from 10-4 to Preliminary work showed that certain organosulphur com- pounds exhibited chemiluminescent emission when mixed with sodium hypochlorite. This paper describes the optimisation of a flow injection procedure for the liquid phase chemilumines- cence determination of 2-(ethy1thio)phenol and structurally related molecules.The advantages of chemiluminescence include: 10-10 M . 3 Experimental Fig. 1. shows the flow injection system used for the determina- tion of 2-(ethy1thio)phenol. The carrier stream was acetone and the oxidant stream was 1 . O M sodium hypochlorite in 0.1 M hydrogen carbonate buffer at pH 11.0. Each stream was pumped at 1.0 ml min-1 by a peristaltic pump (Gilson Minipuls 2: Pl), interfaced to a microcomputer (BBC). Teflon tubing (0.8 mm i.d.) was used throughout the remainder of the manifold. Standards were injected into the acetone carrier stream using a rotary valve (Rheodyne 5020) with a 50-ul sample loop. The valve was operated by an electrically activated switching device (Anachem) interfaced to the micro- computer.2-(Ethy1thio)phenol standards in acetone covering the range 1 x 10-3-1 x 10-1 M were prepared. Standards were drawn into the sample loop by means of a second peristaltic pump (Ismatec Mini S840; P2) which was also interfaced to the microcomputer. The standard and oxidant were mixed at a PTFE T-piece and the chemiluminescence emission occurred in a 177-pI glass coil (1.5 mm i.d.). The distance between the injection valve and the T-piece was 25 cm and between the T-piece and the glass coil 2.5 cm. The light was detected by a photomultiplier (Thorn EM1 Type 9789 QB) and the output recorded on a chart recorder (Chessell BD 4040). The reaction conditions stated above were used for all experiments and all results are the means of five injections unless stated otherwise.Autosampler Injection , 0 uooo valve \ Reagent % 4 i ~ l o w cell stream Waste Peristaltic Pumps Fig. 1. Flow injection manifold Results and Discussion A series of experiments were carried out to establish the optimum conditions for the flow injection procedure. The effect of the oxidant concentration at pH 11.0 was investigated using a 5 x 10-2 M 2-(ethy1thio)phenol standard. Hydrogen peroxide and sodium hypochlorite were used as the oxidants and the results are shown in Table 1. The most intense emission occurred with 2 M sodium hypochlorite. although in subsequent experiments a concentration of 1 M was used to conserve the reagent. Table 1. Effect of oxidant concentration E rn i ss i o n i n t e nsi t y /m V Oxidant conce n t r a t i o ni hi 0.001 0.01 0 .1 0.2 0.5 1 .o 2.0 3 . 0 4.0 With hydrogen peroxide 0 0 1 . 1 2.0 3.0 9.5 2.3 2.2 - With h y poc h I o r i t e 2.0 2.6 18.9 29.5 31.5 37.2 28.4 - - The effect of pH was investigated using 1 M sodium hypochlorite and a 5 x 10-2 M 2-(ethy1thio)phenol standard. The pH was varied from 8.0 to 13.0 and the results obtained are shown in Table 2. The maximum intensity occurred at pH 1 1 .(I. The effect of using different solvents for the standard solutions was also investigated. Acetone was found to be the best solvent of those investigated and gave an emission intensity of 103.5 mV for a 5 x ~ O - ' M 2-(ethylthio)phenol standard; methanol. propan-2-01. ethanol and propan-1 -01 gave signals of 68.1, 38.8.38.6 and 19.4 mV.respectively. The20 Table 2. Effect of pH PH Emission intensitylmv 8.0 9.0 9.0 9.2 10.0 12.2 10.5 18.5 11 .o 23.9 11.5 19.8 12.0 14.8 12.5 13.9 13.0 13.6 reduction in intensity observed on going from acetone to propan-1-01 may be due to an increase in the degree of hydrogen bonding of the solvent which restricts the access of the oxidant. Several other compounds were investigated in an attempt to elucidate the structural features necessary for chemilumines- cence to occur. 2-(Ethy1thio)phenol gave the most intense emission with a relative intensity of 100. p-(Ethy1thio)benzoic acid and p-ethylphenol gave relative emission intensities of 3.8 and 0.6, respectively. 0- and m-Ethylphenol, ethylbenzene and phenol gave negligible emissions, indicating that chemilu- minescence is due to the oxidation of divalent sulphur.The emission intensity of 2-(ethy1thio)phenol is much greater than that of p-(ethy1thio)benzoic acid. This may be caused by intramolecular hydrogen bonding in 2-(ethylthio)phenol, which makes radiationless deactivation less probable. The calibration data for 2-(ethy1thio)phenol are given in Table 3. Using a log - log plot of concentration I-rersiis emission intensity the response is linear over the range 1 x 10-3-5 x 10-2 M ( r = 0.997). The theoretical detection limit is4 x 10-4 M (20 above the background signal). Future work will include the development of totally non- aqueous systems using oxidants that are effective in organic media. This should improve the precision and reduce the problems associated with reproducible mixing and precipita- tion of the analyte when an aqueous stream is mixed with an organic stream. ANALYTICAL PROCEEDINGS.JANUARY 1989. VOL 36 Conclusions Flow injection analysis is an ideal technique for monitoring chemiluminescence reactions. Certain organosulphur compounds exhibit solution phase chemiluminescence and their emission intensities are dependent on their structure. The chemiluminescence emission intensity is influenced by the nature of the solvent. The limit of detection for 2-(ethy1thio)phenol is 0.3 mM (20 nmol). Table 3. Calibration data for 2-(ethylthio)phenol Concentration/ Emission RSD. "!o M in tensi t y/mV ( n = 10) 0 0 0 1 x 10-3 0.2 22.8 5 x 10-3 2.3 2.9 1 x 1 0 2 7.6 6.2 2 x 10-2 22.1 2.6 5 x 10-2 150.5 1.6 1 x 10-1 144.3 3.4 The authors thank Thornton Research Centre, Shell Research Limited, for supporting this work.References 1. 2. 3. 4. 5. 6. Boyle, E. A., Handy. B . . and Vangeen. A.. And. Chetn., 1987. 59. 1499. Burguera. J . L.. Townshend. A . , and Greenfield. S.. A n d . Chim. A m . 1980, 114, 209. Abbott. R. W.. Townshend, A . . and Gill. R.. A t z d y r . 1986. 111,635. Abbott, R. W.. Townshend. A.. and Gill. R.. Atui!\.sr. 1987. 112. 397. Sigvardson. K . W.. and Birks. J . W.. And. Chem.. 1983. 55. 432. Rule, G.. and Seitz. W. R.. Clin. Chem.. 1979. 25. 1635. Arsenic Speciation by Hydride Generation Atomic Absorption Spectrometry and its Application to the Study of Biological Cycling in the Coastal Environment S.D. W. Comber and A. G. Howard Department of Chemistry, The University, Southampton, Hampshire The speciation of arsenic in natural waters is indicative of its diverse reaction chemistry and can be used to investigate a number of geochemical and biological processes. There are four commonly reported forms of arsenic in natural waters: ( i ) oxidised inorganic arsenic [ arsenic(V)], believed to be arsen- ate; (ii) reduced arsenic. probably arsenite [arsenic(III)]; (iii) monomethylated arsenic; and (iv) dimethylated arsenic. 1-4 The study of the behaviour of arsenic in the aquatic environment requires the development of methods that can be used to measure routinely arsenic speciation. Such a system must have sufficient sensitivity to permit measurements to be made of species that are present at concentrations below 100 ng I-' while maintaining relatively high precision.As this system is to be used as a survey technique, it must also be rapid and robust. These criteria are largely met by methods that are based on the trapping of arsenic hyrides prior to sequential release into an atomic absorption spectrometer. Such methods are based on the procedures first used by Braman er u1.5 and Andreae.h This paper describes a rapid hydride generation system which has been developed for the speciation of dissolved arsenic by atomic absorption spectrometry. The system is simple to operate, highly reliable in routine use and has been extensively employed in the study of biogeochemical processes governing the speciation of arsenic in estuarine and coastal environments.Experimental Reagents Standard solutions were prepared from arsenic(II1) oxide. the disodium salt of monomethylarsonic acid and the sodium salt of dimethylarsinic acid. A stock mixed standard solution contain- ing 100 mg 1-1 of arsenic as each individual species was prepared using doubly distilled water and this was diluted further to give a daily working standard of 10 pg 1 - 1 . Method The semi-continuous method used here is similar to that described by Howard and Arbab-Zavar7 but with refinementsANALYTICAL PROCEEDINGS. JANUARY 1989. VOL 26 21 that have led to significant improvements in resolution. reproducibility and sample turnover since the original publica- tion (Fig. 1). A peristaltic pump is used to mix the sample with equal volumes of hydrochloric acid ( 1 + 9) and then sodium tetrahydroborate(II1) (2% rnlV).The resulting arsine. monomethylarsine and dimethylarsine then pass into a custom- built gas - liquid separator. The gas stream is dried with sodium hydroxide pellets and trapped on hydrofluoric acid etched glass beads (about 40 mesh) in a U-tube immersed in liquid nitrogen (-196°C). Once all the sample has been taken up, time is allowed for the arsines to be generated and trapped, after which the liquid nitrogen is removed and the trap allowed to warm to room temperature. The arsines are eluted by the nitrogen carrier gas according to their volatility (arsine Peristaltic pump (2.5 rnl rnin-l, all channels) Electrically heated quartz NaOH pellets atorniser tube in path of AAS Nitrogen Sample rl I- 7 HF-ett!hed beads t Waste Fig.1. Semi-continuous hydride generation system for arsenic speciation (not to scale) followed by monomethyl- and then dimethylarsine) into an electrically-heated quartz T-tube placed in the light path of a Baird A5100 atomic absorption spectrometer. The output is recorded on a Tekman TE200 chart recorder (Fig. 2). 1 2 10 cm I 30s Time - Fig. 2. Typical output from hydride generation AAS for a 1-ml sample containing 1 ng of inorganic arsenic. ( 1 ) Inorganic arsenic: (2) monomethylarsenic (MMAS): and (3) dimethylarsenic (DMAS) Arsenite is measured by controlling the pH of the reaction, the hydrochloric acid being replaced by a sodium acetate - acetic acid buffer (0.1 M ) (pH 5.0). Under these conditions.arsenic(II1) forms arsine but not arsenate. I n a previous paper7 the effect of high concentrations of interferents on this method were discussed. In the analysis of estuarine waters. interferences are not generally significant as potentially troublesome dissolved trace elements are normally present at relatively low concentrations. As a precaution. however, standard additions analyses were run occasionally to confirm the absence of interferents. A summary of the system parameters is given in Table 1. Table 1. System parameters Flow-rates and settings Nitrogen carrier gas . . . . . . . . Hydrochloric acid ( 10o/o) . . . . . . Sodium tetrahydroborate( 111) (2% miV) Air . . . . . . . . . . . . . . Sample . . . . . . . . . . . . Sodiumacetate buffer(O.1 M) . .. . Sampling rate . . . . . . . . . . Trapping time . . . . . . . . . . Source . . . . . . . . . . . . Lamp current . . . . . . . . . . Linear range . . . . . . . . . . Sample size . . . . . . . . . . . . Wavelength . . . . . . . . . . . . Furnace temperature . . . . . . . . 160 ml min 1 2.5 ml min I 2.5 ml min- I 2.5 ml min - I 2.5 ml min ~ I 2.5 ml rnin I 2 0 h I 90 s Arsenic hollow-cat hode lamp 8mA 197.3 nm ca. 900 "C (k3 ng 0.25-2.0 ml Procedure The trap is frozen and the clock started: the probe is placed in the sample cup and allowed to take up all of the sample (typically between 250 ul and 2 ml). It is then transferred into a wash solution of distilled water. After 90 s the liquid nitrogen is removed from the trap and the arsines are allowed to vaporise into the atomiser, where they are monitored by the atomic absorption spectrometer. Intermittently the trap is dried with a heat gun to remove the condensed water that builds up after a period of time and which would otherwise cause excessive back-pressure.At the beginning and end of the analysis a calibration of between 0.25 and 1 ng is run with quality control standards being analysed every 4 or 5 samples to ensure accuracy. The over-all performance of the system is presented in Table 2. Table 2. Performance data Average peak height/ Detection limiti Sample* cm RSD. '"0 ng: Arsine . . . . . . 14.22 k 0.65 4.6 0.0 19 Monomethylarsenic . . 9.88 k 0.76 6.1 (1. (J4S Dimethylarsenic . . 6.3 -t 0.32 5.1 0.061 Blank(AsH3) . . . . 1.3 f 0.05 3.8 * Ten replicates. + Detection limit defined as three times the standard deviation otthe blank.Typical Application The technique was used in a number of studies of arsenic in estuarine. coastal and oceanic environments. In the example described here, the method has been used to investigate factors that influenced the speciation of arsenic in Southampton Water during 1987. Samples were collected regularly throughout the year. with particular attention being paid to the warmer months when biological activity was high. At the same time, bacteria and phytoplankton populations were monitored by members of the Oceanography Department. Only arsenic(V) was present in the water all year round. Arsenite, monomethyl- and dimethylarsenic were only present in significant amounts from mid-May to the end of October, the period during which biological activity reached its maxi- mum.This confirmed our previous observations regarding speciation seasonality.x-10 Dissolved monomethylarsenic and dimethylarsenic levels reached a broad maximum in July and August. The thermodynamically unstable arsenite peaked early in the summer (18th May). when it accounted for over22 ANALYTICAL PROCEEDINGS. JANUARY 1980, VOL 26 50% of the total inorganic arsenic present. This coincided with a bloom of filamentous algae. The results are illustrated in Fig. 3. 1.2 1 .o 0.8 c - 0.6 I I 2 0.4 0.2 0 R I \ " I m Y- Y u - E 3 7 87 26,8,87 19 10 87 Date Fig. 3. Arsenic speciation during the 1987 temporal survey at Calshot Buoy. Southampton Water. UK. (A) AsV; (B) As"'; (C) MMAS; and (D) DMAS This work is only a small part of an ongoing project designed to study the behaviour of arsenic in a number of estuarine and coastal systems in order to clarify the relationship between the speciation of dissolved arsenic and biological activity in the water column.The authors thank the National Environment Research Coun- cil and Dr. A. Morris of the Plymouth Marine Laboratory for their support of this work. 1 . 2. 3 . 4. 5 . 6. 7. 8. 9. 10. References Sanders. J . G., and Rapp. P.-V., Reun. Cons. In!. Esplor. Mer.. 1986. 186, 185. Sanders. J. G.. Mar. Chem., 1985, 17, 329. Andreae. M. O., and Froelich, P. N.. TellirJ, 1984, 36B. 101. Andreae. M. O., Limnol. Oceunogr., 1970. 24. 440. Braman, R. S., Johnson. D . L., Foreback, C. C.. Arnrnons, J . M., and Bricker.J . L.. And. Chem.. 1977, 49, 621. Andreae, M. 0.. Anul. Chem., 1977, 49. 820. Howard, A. G., and Arbab-Zavar. M. H . , Anulw. 1981. 106, 213. Howard, A. G.. Arbab-Zavar, M. H.. and Apte. S. C.. Mar. Chem., 1982. 1 1 . 393. Howard. A. G., Arbab-Zavar. M. H.. and Apte. S. C.. Estuarine Coustal Shelf Sci., 1984. 19. 493. Apte, S. C.. Howard, A. G., Morris. R. J . . and McCartney. M. J . , Mar. Cheni., 1986, 20. 119. Selectivity Studies in Supercritical Fluid Chromatography M. Marsin Sanagi and Roger M. Smith Department of Chemistry, Lough boroug h University of Tech no logy, Loug h boroug h, Leicestershire L E I ? 3TU There has been a considerable interest in supercritical fluid chromatography (SFC) in the last few years as an alternative instrumental analytical technique complementary to high-per- formance liquid chromatography (HPLC) and gas - liquid chromatography (GLC).' SFC has been gaining acceptance in a wide range of areas in both analytical and industrial chemistry, particular for petrochemicals and pharmaceuticals. 1 As with HPLC and GLC, the retention and selectivity of the solute in SFC depend on the conditions used for the separation, including the pressure. temperature and stationary phase composition of the mobile phase.However, in contrast to HPLC, the density of the mobile phase, which is determined by pressure and temperature, also plays an important role in dictating retention. This gives additional flexibility when optimising separations in SFC and offers an alternative selectivity to other techniques.Good instrumentation is a prerequisite for performing sound chromatographic separations. However, good instrumentation alone will not provide the required separation without a sensible choice of the operating conditions and, in this respect, some knowledge of the retention behaviour of the solute will provide a valuable guide to achieving such separations. I n both GLC and HPLC the use of relative measurements is well established as a method of reducing the effect of small changes in the operating conditions. These measurements can either be relative to a single standard or to a retention index scale obtained using a series of homologous standard com- pounds. In GLC an excellent retention index system for n-alkanes has been proposed by Kovats,2 based on a constant increment in the logarithm of the capacity factor ( k ' ) with the length of the carbon chain (log k' = aC,, + h).The retention index of the n-alkane standards is defined as the carbon number x 100 and the values for test compounds are calculated by interpolation between standards. A similar concept has also been used in reversed-phase HPLC. but usually alkyl aryl ketones3 or alkan-2-onesj have been used as the standard compounds as the n-alkanes are highly retained compared with most analytes and they are not detectable by the commonly used spectroscopic detectors. In contrast, the alkyl aryl ketones are readily detectable by ultraviolet detectors and cover a wide retention range of most compounds of medium polarity. and consequently they have been used as a means of drug identification.5 This paper describes an investigation into the retention behaviour and selectivities of different groups of compounds with changes in the operating parameters and provides a detailed study of the use of different sets of standards as the basis of retention indices in SFC.Experimental The supercritical fluid chromatograph used in this work has been described elsewhereh; it consists of a modified Jasco BIP-1 HPLC pump and a Pye Unicam 104 gas chromatograph. A 150 X 4.6 mm i.d. column packed with PLRP-S 5 pm polystyrene divinyl benzene (PS-DVB) (Polymer Labora- tories) was used. Acetone was used as the dead-volume marker. Retention indices based on either alkyl aryl ketones or n-alkanes were calculated by fitting the log k' versus carbon number x 100 data for the standard compounds to a regression line and then interpolating the log k' values for the test compounds. Results and Discussion Retention Index Scales The alkyl aryl ketones and n-alkane homologues were eluted in order of increasing relative molecular mass and so theANALYTICAL PROCEEDINGS.JANUARY 1989. VOL 26 separations followed a reversed-phase type of retention mechanism (Fig. 1). Almost linear relationships between log k' and the carbon number was found although there appeared to be some deviation for the first two members of each series. The slopes of the relationship of log k' versus carbon number x 100 for the alkyl aryl ketones and n-alkanes were similar but not identical, suggesting that the addition of a methylene group does not have the same effect on the retentions of the two homologous series.Because a flame ionisation detector cannot be used when a modifier is added to the mobile phase and spectroscopic detectors are needed. the n-alkanes are unsuit- able for general use as SFC retention index standards. The retention index scale based on the alkyl aryl ketone homolo- gous series has therefore been used to compare the selectivities of different column materials with changes in the operating parameters, by measuring the retention of alkylbenzenes and a set of model compounds with different functional groups. 1 1000 1500 2000 Carbon number x 100 Fig. 1. Plot of log k ' for n-alkanes and alkyl aryl ketones \'crsiis carbon number x 100. Conditions: column, PLRP-S; temperature, 60°C; and mobile phase.carbon dioxide. (A) Alkyl aryl ketones; and (B) n-alkanes Effect of Pressure In general, the retention of a compound in SFC decreases with an increase in pressure but the relationship is not linear (Fig. 2). This is because the solvating power, which is roughly proportional to the density, increases non-linearly with an increases in pressure.' The relative retention of compounds in the homologous series did not change as shown by the graphs of log k' versus pressure for both the alkyl aryl ketones and alkylbenzenes. J 2000 2200 2400 2600 2800 Mean column pressureAb Fig. 2. Variation of log k' for alkyl aryl ketones and alkylbenzenes as a function of pressure. Conditions: column, PLRP-S; temperature, 60°C; and mobile phase, carbon dioxide.(A) Propylbenzene; (B) butylbenzene; (C) acetophenone; (D) propiophenone; (E) butyrophe- none; (F) valerophenone; (G) hexanophenone; and (H) heptanophe- none The retention indices, based on alkyl aryl ketones, of the model compounds were determined and plotted as a function 23 of the pressure (Fig. 3). A range of different behaviours was observed. The retention indices for benzaldehyde and methyl benzoate remained almost constant as the pressure was increased which suggests that these compounds have a similar retention behaviour to alkyl aryl ketones. On the other hand, as the pressure was increased the more polar compounds such as benzoic acid, benzamide and p-cresol were retained to a greater extent relative to the alkyl aryl ketones. This effect can be attributed to the decreasing solubility of these polar compounds in the mobile phase. / 2000 r .- : I 500 2000 2200 2400 2600 2800 Mean column pressureilb i n - 2 Fig.3. Variation of retention indices. based on alkyl aryl ketones, ot model compounds with increasing pressure. Conditions: column, PLRP-S; temperature, 60 "C; and mobile phase, carbon dioxide. (A) Benzaldehyde; (B) methyl benzoate; (C) benzyl alcohol; (D) p-cresol; (E) N-propylaniline; (F) benzoic acid; (G) benzamide; and (H) benzylamine Effect of Tern pera tu re The retention behaviour of different groups of compounds with changes in the operating temperature was studied. At a constant pressure above the critical point, increasing the temperature effectively decreases the density of the mobile phase and hence the over-all elution strength and, as expected, the retention increased for all compounds (Table 1).It was also found that changes in temperature could bring about relative selectivity changes between compounds with different func- tionalities and that at some points the elution order of the compounds could be reversed, e . g . , for p-cresol and N-propyl- aniline. Table 1. Capacity factors of model compounds at different tempera- tures. Conditions: column, PLRP-S, 5 pm; mean column pressure, 2515 Ib in-?; flame ionisation detection; mobile phase, carbon dioxide Capacity factor Tempera t u re/"C (Eluent densitylg cm- 3 ) 40 Compound (0.81 1) Benzaldehyde . . . . 1.27 Methyl benzoate . . . . 1.45 Benzylalcohol . . . . 1.87 Nitrobenzene . . . . 1.95 p-Cresol .. . . . . 3.19 N-Propylaniline . . . . 3.02 Benzoic acid . . . . . . 5.54 Benzamide . . . . . . 7.78 Benzylamine . . . . . . 12.50 60 (0.673) 1.51 1.70 2.13 2.31 3.25 3.31 5.61 8.70 14.26 80 (0.5 1 9) 2.04 2.34 2.77 3.19 3.96 4.26 6.96 11.14 20.70 100 (0.406) 2.72 3.32 3.65 4.42 4.86 5.81 7.99 14.12 32.2324 c1 d 0 H u ANALYTICAL PROCEEDINGS. JANUARY 1989, VOL 26 Effect of Modifier One of the drawbacks of using carbon dioxide as the mobile phase is that because of its low polarity, it is difficult to elute polar compounds. The separations of these compounds are typically characterised by very broad, tailing peaks, which lead to poor reproducibility of the retention time and peak area and poor sensitivity.' The use of organic modifiers in the carbon dioxide mobile phase was therefore investigated.A series of separations was carried out on the PLRP-S column with different percentages of methanol over the range 0-14.6%. The addition of a low percentage of methanol was found to reduce drastically some of the retentions. The modified eluent generally gave better separations and peak shapes for the polar compounds, particularly for the more polar compounds, benzoic acid, benzamide, benzylamine, and the alcohols. These results confirm previous reports detailing the drastic effects on packed column SFC of the addition of small amounts of a modifier to the carbon dioxide mobile phase. 8." The retention indices of the model compounds were determined and the values plotted as a function of the methanol concentration in the mobile phase (Fig.4). Marked selectivity changes were observed when the amount of modifier in the mobile phase was varied. The retention indices of the 1500 tv X -0 .- : 1000 .- c C al c. a 500 0 4 a 1 2 16 Methanol, '10 VN Fig. 4. Variation of retention indices, based on alkyl aryl ketones. of selected compounds with methanol concentration. Conditions: col- umn. PLRP-S; mean column pressure, 2515 Ib in-'. (A) Toluene; ( B ) methyl benzoate; (C) benzyl alcohol; (D) p-cresol; (E) N-propylani- line; (F) benzoic acid; ( G ) benzamide; and (H) benzylamine more polar compounds, benzoic acid, benzamide and the aromatic alcohols and phenols. decreased with an increase in the modifier concentration. In contrast, those of the non-polar compounds, toluene and methyl benzoate, were virtually unchanged or increased only slightly.Surprisingly, the rela- tively polar amines, benzylamine and N-propylamine. showed little change on this polymer column. Effect of Stationary Phase Similar experiments were also carried out using ODs- and cyano-bonded silica columns. The most marked difference was in the behaviour of amines such as benzylamine, which was highly retained on ODs-silica. This was presumably due to silanol interactions as the retention of these compounds decreased markedly with the addition of small amounts of modifier to the mobile phase. The cyano-silica columns showed greater retention of the more polar compounds and less retention of the non-polar compounds compared with the corresponding separations performed on ODs-silica or poly- mer columns.Conclusions The retention and selectivity in SFC varies with the operating pressure, temperature, density and composition of the mobile phase and stationary phase. Any one or a combination of these factors can be used to vary the selectivity of the system in order to optimise the separation. Retention indices were used as a convenient means of recording the variation in the selectivity with changes in the operating parameters. We thank Polymer Laboratories Ltd. (UK) for the generous gift of the PLRP-S column and the Public Services Department and the University of Technology, Malaysia. for a studentship to M.M.S. 1. 2. 3 . 4. 5 . 6. 7. 8. 9. References Smith, R. M.. Editor. "Supercritical Fluid Chromatography." RSC Chromatography Monographs, Volume 1.Royal Society of Chemistry, London, 1988. Kovats, E . . Helv. Chim. Acta, 1958. 41. 1915. Smith. R. M.. J. Chromutogr.. 1982. 236. 313. Baker, J . K.. and Ma. C. Y . . J . Chromatogr.. 1979. 169. 107. Smith, R. M.. Ad\!. Chromatogr.. 1987. 26. 277. Sanagi. M. M.. and Smith. R. M.. Anal. Proc.. 1987. 24. 304. Schoenmakers. P. J., Rothfusz. P. E.. and Verhoeven. F. C. C . J . G.. J . Chromarogr.. 1987. 395. 91. Blilie. A. L.. and Greibrokk. T.. And. C'hem.. 1985. 57. 2239. Levy, J . M.. and Ritchey. W. M.. J. Chromatogr. Sci.. 1986.23. 242. Retention Prediction in RP-HPLC Using a Functional Group Database and Expert System (CRIPES) Christina M. Burr and Roger M. Smith Department of Chemistry, Loughborough University of Technology, Loughborough, Leicestershire LE 1 1 3TU The possibilities of using expert systems in analytical chemistry applications have recently been the subject of a number of publications.'-" It has generally been agreed that expert systems can provide a convenient and user-friendly interface for different situations. Although many methods of prediction have been proposed for reversed-phase high-performance liquid chromatography very few are based on the molecular structure of the analyte.The aim of this work has been to develop a method of calculating the retention time of a compound from its structure. This is expressed as its retention index based on the alkyl aryl ketone scale. 1 ° . l l The calculation uses the equation RI = PI + ZSIR + ZSIAr-X + ZS1R-X + E I I y z where PI = retention index of parent: SIR = substituent index due to saturated alkyl chain; SI,,-X = substituent index of aromatic substituents; SIR_?( = substituent index of aliphatic substituents; and I f y z = interaction indices due to interactions between substituents.Each term in the above equation is related to the eluent composition. in either methanol - bufferANALYTICAL PROCEEDINGS. JANUARY 1089. VOL 26 800 750 X U .E 700 or acetonitrile - buffer, by a quadratic equation: I = ax2 + hx + c (s = '/o organic modifier) The values of the coefficients for a wide range of substituents and interactions have been determined experimentally from the retention indicesl2.13 of a range of model mono- and di-substituted aromatic compounds, over the eluent ranges 4&8O0/o methanol and 30--80°/~ acetonitrile.These include 11 aromatic substituents on benzene and 10 aliphatic substituents on a1 kylbenzenes. including alcohols, halides. cyano groups and acid derivatives, but under the experimental conditions employed free carboxylic acids could not be examined. The data has been used to form a database which is interrogated using an expert system program. The program (CRIPES, Chromatographic Retention Index Prediction Expert System) was written as the knowledge base of a commercial expert system shell (VP-Expert. Paperback Soft- ware). This particular shell can do mathematical calculations and is also capable of communicating with compatible external spreadsheets. The regression equations can therefore be held in the external spreadsheet, enabling them to be updated without extensive modification of the knowledge base.CRIPES has been used to provide a "user-friendly" interface to the database and for the calculation of retention indices. The program will also calculate the resolution between pairs of compounds and can suggest optimum separation conditions for mixtures. The Database The basis of the prediction system is the database collection of the a , b and c coefficients of the quadratic equations for the retention of the parent compound benzene, the substituents and the interaction terms. The individual substituent indices have been calculated from the retention index increments calculated as the difference, at the same eluent composition. between the retention index of a mono-substituted compound and the parent compound benzene (Fig.1). Quadratic regres- sion lines were fitted to the change in experimental retention index increments, with each eluent composition, to obtain the coefficients of the substituent index equations. For all the substituents the quadratic curves were a very close fit to the experimental retention index increments. The coefficients were transferred in each instance to a spreadsheet (VP- Planner, Paperback Software). Interactions which occur between substituents, such as hydrogen bonding. steric interactions and electronic interac- - - - 11501 1 .- 5 900 .- c C 850 1100 c 700 1000 I I I - t - I 750 t I 25 tions, are accounted for by using interaction indices. These are derived from the difference between the measured retention index of a di-substituted compound and the calculated sum of the parent retention index and the substituent indices: For example, the hydrogen bonding in 2-hydroxyacetophe- none considerably reduces the polarity and increases the retention (Fig. 2) so the interaction index can be expressed as A large number of interaction indices have been determined for aromatic substituents in the ortho, meta and para positions to either hydroxyl or methyl groups.These represent com- pounds in which strong electronic interactions would be expected and those in which only weak electronic interactions would be expected, respectively. The interactions vary with eluent composition and the values can also be fitted to quadratic expressions whose coefficients have been entered into a spreadsheet.IZ = RI,,, - ( P I + Z S I ) IIOH+COCHJ = RIcxp - + SIAr-OH -t slAr-COC'ti,) 900 1 .- LT 550 ( g C 0 C " 500 1 I 1 I I 1 ivieLiv concenrrariori, -10 Fig. 2. Determination of interaction indices for H-bonding in 2-hydroxyacetophenone as the difference between the retention index calculated as Rl = PI + ZS/ and the experimental retention index. I, Experimental retention index; 0. calculated retention index from individual i ncrem e n t s . I l O i 1 <.( )c I ~. interact ion index User input name I and substituents present I I 1 Sum PI, SI and II coefficients 1 I 1 Calculate retention indices at 40-80% MeOH and 30-80% MeCN I I DisDlav RI and a m r o x k' 1 Fig. 3. database and calculation of predicted retention indices CRIPES-path of the program steps for interrogation of the Operation of The Expert System (CRIPES) The expert system provides a simple and convenient method for bringing together the coefficients from the spreadsheet and26 ANALYTICAL PROCEEDINGS.JANUARY 1989, VOL 26 Table 1. Example of calculation of retention index Compound name : Thymol Saturated aliphatic carbons : C3H7, CH3 Substituents : Aromatic OH Interaction terms : mOH-R. PhCH-R, secondary CH3 (Coefficients of equations I = ax? + bx + c (x = '/" modifier) Eluent modifier Methanol Acetonitrile Index term PI S I R -CH3 -C3H7 SIAr-Oll I I m OH-R IIPhCll-R ( 2, II\CC-CH3 Sum Calculated Retention Indices Methanol, YO 40 50 60 70 80 a -0.0121 -0.0250 0 0 0.0064 0 - 0.0065 - 0.0372 RLlC 1042 1039 1028 1009 984 b 3.879 0 0 - 1.061 0 0.448 -0.270 C 749 -151 100 300 24 - 24 - 16 2.996 982 R L , 1042 1035 1030 1001 965 a - 0.0 140 0.0243 0 0 0 0 0.0051 - 0.0052 b 2.600 -4.840 0 0 0.150 0 0 -2.090 Acetonitrile.o/o RIG,,, 40 1001 50 984 60 969 70 955 80 94 1 C 845 - 89 100 300 - 26 - 24 - 30 1076 R L , 1005 993 953 955 963 calculating the retention index. The program needs to extract the appropriate coefficients for the parent index, substituent indices and interaction indices for the analyte. By summing the coefficients, the retention index can be expressed as quadratic equations for both methanol and acetonitrile. This enables the calculation of the retention index at eluent compositions within the range 30-80% acetonitrile and 40- 80% methanol. CRIPES produces menus enabling the user to select which aromatic and aliphatic substituents are in the compound.The program will then prompt the user to input the number of each substituent and its position. Following this the program extracts the coefficients of the appropriate substituent interaction equations. The information on which substituents are present and their positions is then used to determine the interactions and to extract the appropriate coefficients for the interaction index equations (Fig. 3). The retention indices for a range of eluent compositions are then calculated and dis- played. By using the relationship between the retention index and the capacity factor log k' = a'RI + 6' it is possible to back-calculate to predict capacity factors from the retention indices. It is also possible for CRIPES to compare the equations for two compounds, calculate the maximum separation and hence suggest optimum separation conditions. An example of the calculation is given for thymol in Table 1.For this compound the substituents are an aromatic hydroxyl group (SIAr-OH), a methyl group and a branched alkyl chain (SIR) on an aromatic ring. In addition, interaction terms are required for the interaction between a meta-substituted hydroxyl and alkyl group (IIm0H-R) and for the substitution on the benzylic carbon group (IIPhCH-R). This has been found to have a different increment from subsequent saturated carbon atoms. Finally, a term is required to account for the differences between the retention behaviour of straight and branched chain isomers (llsec-CHj).14 There was good agreement between the calculated and experimental retention index. The calculations could be carried out by using a calculator; however, this would prove time consuming and could become RI = x2Za + xZb + Zc (x = o/o organic modifier) complicated when several interactions were occurring. It would also be necessary to remember continually all the rules for the interactions and isomer effects, whereas these opera- tions are automatically included in CRIPES. CRIPES has been tested by calculating the predicted retention indices of a number of poly-functional model compounds, which have been compared with the retention indices of these compounds determined experimentally at selected eluent compositions. For most compounds there is, however, good agreement between the experimental and calculated retention indices.However, the database so far only contains parameters for a limited number of interactions between groups and cannot therefore compensate for all the interactions that might occur in some complex compounds. We thank the following: The Science and Engineering Research Council for a research grant and for a studentship to C. M. B., and Phase Separations Ltd. for a gift of Spherisorb ODS-2. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. References Pierce, T. H., and Hohne, B. A.. "Artificial Intelligence Applications in Chemistry," American Chemical Society. Washington, DC, USA, 1986. Bridge, T. P.. Williams. M. H.. and Fell, A. F.. Anal. Proc.. 1988, 25. 43. Ayscough. P. B., Chinnick. S . J . , Dybowski, R., and Edwards.P.. Chem. Ind.. 1987. 15. 515. Frazer, J . W., Mikrochim. Acra, 1986. 11, 163. Kleywegt, G . J . , Lab. Microcomprtr.. 1987, 6. 74. Glajch. J . L., LCGC, 1988, 6, 30. Dessy, R . E . , Anal. Chem., 1984. 56, 1201A. Dessy, R . E., Anal. Chem., 1984. 56, 1312A. Bridge, T. P., Williams, M. H., and Fell, A. F . . Chem. Br.. 1987, 23, 1085. Smith, R. M., Adv. Chromarogr.. 1987. 26. 277. Smith, R . M., J . Chromarogr., 1982. 236. 313. Burr, C. M.. and Smith, R. M.. Anal. Proc.. 1988. 25. 46. Smith, R . M., and Burr. C. M.. J . Chrornarogr., submitted for publication. Smith, R. M.. J . Chromarogr.. 1981, 209, 1 .ANALYTICAL PROCEEDINGS, JANUARY 1989, VOL 26 27 The Determination of Quantum Efficiencies of Laser Dye Using the Thermal Lens Effect Jun Shen and Richard D.Snook Department of Instrumentation and Analytical Science, UMIST, Manchester M60 1 OD The absolute fluorescence quantum efficiency of fluorescence materials is an important parameter for both fundamental theory and analytical applications. It is important for the study of radiation processes in molecules. 1 for correlation of predic- ted luminescence lifetimes with the observed lifetimes2 and for making assignments of electronic transitions.' It is necessary for calculating thresholds for laser action4 and for judging the suitability of materials as wavelength shifters in optical pumping experiments or for use as energy donors.' With luminescence data, it is also used to evaluate the purity of a material. 1 Although the absolute quantum efficiency is of great importance, measuring this parameter precisely is never easy.In the past, many measurement techniques have been devel- oped, including radiation integration methods, fluorescence lifetime methods, calorimetry, etc. However, all these tech- niques require absolute calibration for the optical detectors. Luminescence standard samples whose fluorescence quantum efficiencies are exactly known are also required. The accuracy attainable with very careful experimental procedure is typically 5-10%. When substantial corrections for different conditions, such as solvents, are invoked or when instrumental sensitivity at widely differing wavelengths must be considered, total uncertainties can easily reach 25%. In addition, the number of the absolute primary standards is extremely small.Consequently, the measurement becomes quite complicated. The measurement of absolute quantum efficiency using the thermal lens effect was first pointed out by Hu and Whinnerys and applied to liquid solution by Brannon and Magde.6 Unlike conventional quantum efficiency measurements, this method is absolute and does not need any luminescence standard and significantly reduces uncertainties. It could, therefore, become a standard laboratory technique for measuring quantum efficiency. Sodium fluorescein is probably the second best characterised luminescence standard after quinine sulphate in 0.1 or 1.0 N sulphuric acid. However, the quantum efficiency values of sodium fluorescein are still not accurately determined, lying between 0.77 and 0.95.This project concerns, therefore, the determination of absolute fluorescence quantum efficiencies of the laser dye sodium fluorescein in 0.1 N aqueous sodium hydroxide and in ethanol, using the thermal lens effect. Development and Detection of the Thermal Lens Thermal lensing was first reported and discussed in detail by Gordon et al.7 and was used to measure the absorption coefficients of liquids. The most important advance in the thermal lens technique was introduced by Hu and Whinnery.' The single-beam thermal lens experimental set up is quite simple (Fig. 1). A liquid sample is placed in the beam of the laser and is heated by the absorbed laser power. The heat dissipated to the solution creates a temperature gradient which Laser Shutter Cell Pinhole I I ~ P M T Filter Lens -1- r\ ,I u1 u2 u 3 Fig.1. Single-beam thermal lens experimental set up. D,, focal length of lens; Dz. confocal length; D3. arbitrary produces a refractive index gradient, thus creating an effective optical lens. As most liquids have a positive coefficient of thermal expansion. the temperature coefficient of the index of refraction, dnldt, is negative, and the thermal lens is divergent. When a Gaussian profile laser beam is passed through the liquid at t = 0, an effective lens develops in the liquid according to7 x k w 2 F(t) = (1 + t,./2r) . . . . (1) Pth(dn/d r ) Here F(t) is the focal length of the thermal lens (cm); k is the thermal conductivity (W cm-1 K-1); w is the radius of beam in the liquid (cm); 1,- wzpcl4k and is the critical time, where p = density (g cm-3) and c = specific heat (J g-1 K-I); P t h , the thermal (heat) power, is the incident laser power (W) which is absorbed and converted to heat; dnldt is the refractive index change with temperature (K-I).For a single-beam thermal lens experiment, the thermal lens in the sample is detected by its effect on the propagation of the laser beam. The magnitude of the thermal lens signal can be monitored by measuring the laser intensity at the beam centre, which is inversely proportional to the beam area in the target pinhole. If the sample is placed at the confocal length, past the focus, the maximum de-focusing effect on the laser beam will be exhibited. Here, the confocal length b = n wC12/h where h is the laser wavelength (cm), w ( ~ is the beam waist radius (cm).The expression relating the intensity Z(t) as a function of time is given as5 z(t) = z ( ~ ) [ i - e/(i + t,/2r) + fe2(i + tJ2t)I-l . . (2) . . . . . . ( 3 ) Pth(dn/dT) h k e = 8 and thus the thermal (heat) power can be determined by measuring the initial intensity Z(o) and the intensity after steady state has been established I(=). That is . * (4) e = 1 - (1 + 21)1 I = [ I = [Z(o) - I( q l I ( = ) . . Determination of Absolute Quantum Efficiency by Thermal Lens The laser power, P , incident on a sample is equal to the sum of transmitted power P,, the fluorescence emission power P f , and the thermal (heat) power f t h (Fig. 2). P = f , + f f + f , h . . . . . . ( 5 ) A Fig. 2. power; Pf fluorescence emission power; Prh thermal (heat) power Power conservation.P is incident laser power; P, transmitter28 ANALYTICAL PROCEEDINGS. JANUARY 1989. VOL 26 The transmission defined as: So the absorption ratio T and absorption coefficient A are T = P,/P . . . . . . . * (6) A = l - T . . . . . . . . (7) AP=Pf+Pth . . . . . . (8) power is The definition of fluorescence efficiency, Qf, is Number of photons emitted Qf = Number of photons absorbed That is (9) Here, ( v v ) and ( A f ) are the average huorescence frequency and wavelength; v and h are incident laser frequency and wavelength. (A,) can be obtained from the fluorescence spectrum E(h) of the sample. (A,) = / E ( h ) d h / \ y d h . . . . (10) The ratio (hf)/h takes Jaccount '6f the Stokes shift. The absorption coefficient, A , may be measured with an ordinary spectrophotometer and the thermal (heat) power can be measured using the thermal lens effect.In addition to measuring the sample, a measurement of a non-fluorescence absorber, designated by a superscript r, is made. We have From ( 3 ) , (4), (9) and (11) we have Q - , - (A,) - [ I--- A r f r p t h ] . . . . h A P P:h If the dilute solutions are used in the experiment, and the same solvent is used for both sample and reference absorber, then L 3 Equation (13) assumes that the laser wavelength is the same and that the optical properties of the solution are dominated by the solute, and the thermal optical properties of the solution are solely determined by the solvent. Experiment Fig. 1 shows a single-beam lens experiment set up. The laser was an argon ion laser (CR-6 Supergraphite Ion Laser, Coherent) operating at 488 nm.The laser is focused by a lens with a focal length of 20 cm with a mechanical shutter placed at the focal plane. A sample cell is placed at a distance of about 23 cm from the lens. The exact position is optimised by seeking the location which maximises the thermal lens effect. This has been shown to occur at the confocal distance past the focus.5 A photomultiplier tube with a pinhole (aperture = 2 mm) is located 3.5 m from the sample and is positioned accurately at the centre of the laser beam. Although a relatively large pinhole, the aperture of 2 mm radifis was much smaller than the far-field radius of the beam (3 cm), thus averaging out high frequency spatial noise from surface imperfections in the optical components.An inconel coated neutral density filter was placed very close to the pinhole in order to minimise the effect of stray light on the PMT. The output of the PMT was d.c. coupled to a storage oscilloscope (Tektronix 466). The photometric absorption measurements were made with an ultraviolet - visible spectrophotometer (Lambda 5. Perkin- Elmer). The fluorescence spectra of the samples were obtained from a luminescence spectrometer (LS-5, Perkin-Elmer). Solutions were contained in 1-cm fused silica cells, and the same cells were used for both photometric and thermal lens measurements. The sample, sodium fluorescein, and the reference absorber, pararosaniline, were obtained from the Sigma Chemical Company Limited. According to the assay data obtained from Sigma, no impurities in the sodium fluorescein were detected.The solvents were 0.1 N aqueous sodium hydroxide and ethanol. We noticed that several experimental constraints are imposed on the method in the derivation of Equations 1 4 : firstly, the laser beam must have a TEMoo Gaussian profile; secondly, the thermal (heat) power must be low enough to avoid spherical aberrations and convection effects, which requires Pth C2.2 h k/(dnldt); thirdly, the sample cell should be accurately positioned at a confocal length past the focal plane; fourthly, the sample cell length should be long compared with the beam diameter to avoid end effects; fifthly, the sample cell should be short compared with the confocal distance to ensure uniform beam area throughout the sample; and sixthly, the aperture of the pinhole must be small with respect to the far-field beam area and accurately centred. Water is not a sensitive solvent for the effect because of its large thermal conductivity and small drzldr. However, its use as a solvent was investigated. About 20 mW laser power was used to do the experiment when using water as a solvent compared with 10 mW laser power to measure the quantum yield when using ethanol as a solvent. For ethanol a blank correction was required, because f t h is additive for both sample and refer- ence. 0 = 0 (solution) - 8 (solvent) . . . . (14) In the experiments the shutter was opened for about 1.5 s at intervals of about 3 min so that the sample could relax sufficiently in the intervals. Solutions were freshly prepared and the measurements were taken at room temperature. The results of the measurements are shown in Table 1. being the average of twelve signals each. Results By using the data in Table 1. together with the Stokes factor. we obtained the absolute quantum efficiencies of sodium fluorescein in 0.1 N sodium hydroxide solution and ethanol, Qt = 0.95 2 0.02 and Qf = 0.97 k 0.02, respectively. Literature values for sodium fluorescein in 0.1 N NaOH lie in the range 0.77-0.95. Demas and Crosby suggested8 that Q, = 0.90 is a compromise value which is accurate within 10% and probably within 5 % . Our result agrees with it perfectly and is repro- ducible. It is noticed that quantum efficiencies in ethanol are a little higher than those in 0.1 N NaOH, and that the average emission wavelength is a little lower than literature values. These differences are probably the results of trace impurities in the solvent. Table 1. Comparison of sodium hydroxide and ethanol solvents Dye Solvent concentration (A,)inrn BiP W:P A Ar Pr 0.1 N NaOH l.0Ox 10-6 521.0 (5.95 t 0.48) (2.11 t 0.08) 0.120 0. I87 0.95 t 0.02 Ethanol l.0Ox 10 515.1 (1.30 k 0.20) 0.517 t 0.006 0.014 0.067 0.97 k 0.02 x 10-4 x 10-2 x10ANALYTICAL PROCEEDINGS. JANUARY 1989. VOL 26 The experiments have shown that the technique using the thermal lens effect is accurate, convenient to carry out, and does not require standardisation. It is also simple. inexpensive, and should become a viable technique for the measurement of absolute quantum efficiencies. References 1. Parker. C. A. “Photolurninescence of Solutions.” Elsevier Publishing Company. New York. NY. 1968. 29 2. Strickler, S. J . , and Berg. R. A . , J. Chem. Phys., 1962. 37.814. 3. Lytle, F. E., and Hercules, D. M.. J. Am. Chem. SOC.. 1969.91. 253. 4. Soroking, P. P.. Lankard, J . R., Moruzzi, V. L . , and Harnrnond, E. C.. J. Chem. Phys.. 1968,48. 3726. 5. Hu, C., and Whinnery, J . R., Appl. Opt.. 1973. 12. 72. 6. Brannon, J . H.. and Magde. D.. J. Phys. Chem., 1978,82,705. 7 . Gordon, J . P., Leite. R. C . C., Moore. R . S., Porto, S . P. S . . and Whinnery. J . R.. J. Appl. Phys., 1965, 36, 3. 8. Denas, J . N.. and Crosby, G. A.. J. Phys. Chem.. 1971,75.991.