Miniaturisation of a matrix separation/preconcentration procedure for inductively coupled plasma mass spectrometry using 8- hydroxyquinoline immobilised on a microporous silica frit Simon D. Lofthouse,a Gillian M. Greenwaya and Sharon C. Stephenb aUniversity of Hull, Cottingham Road, Hull, N. Humberside, UK HU6 7RX bAvecia Ltd, Blackley, Manchester, UK M9 8ZS Received 9th September 1999, Accepted 11th October 1999 A comparison has been performed on miniaturised matrix separation/preconcentration procedures using packed micro-columns of imminodiacetate (IDA) chelating reagents and an 8-hydroxyquinoline (8-HQ) micro-column. Commercially available IDA reagents, Prosep and Muromac, were packed into micro-columns and 8-HQ has been immobilised on a novel microporous silica structure. These have been successfully utilised for the determination of several trace elements in complex matrix samples.The miniaturised matrix separation/ preconcentration procedures have been developed to reduce sample analysis time. A microconcentric nebuliser in the ICP-MS permits the multielement analysis on a smaller volume of solution leading to a reduction in reagent consumption and a more efÆcient procedure.Preparation of the micro-columns is described along with optimisation of the procedures with respect to the variables buffer concentration, buffer pH, eluent acid concentration and reagent Øow rates. Sample analysis times are compared for both miniaturised systems.Analysis times of 3.0 min for the IDA column and 2.3 min for the 8-HQ column are reported. Calibrations showed good linearity with correlation coefÆcients of 0.999±0.9998 for IDA columns and 0.999±0.9997 for 8-HQ column for a range of analytes. Recoveries ranging from 91±102% for IDA columns and 96±105% for the 8-HQ column are reported for a range of elements. The method was validated by the analysis of estuarine (SLEW-1) and coastal (CASS-2) certiÆed reference materials.Good agreement between the certiÆed and reference values was obtained for the materials. Introduction ICP-MS has evolved into one of the most accurate, sensitive and reliable trace element measurement techniques. The determination of trace and ultratrace elements in a wide variety of matrices is now common practice. Nevertheless, early work with ICP-MS identiÆed major disadvantages including a low tolerance to dissolved solids (v0.2% m/v) and the formation of polyatomic interferences.1,2 High levels of dissolved solids lead to blockage of the interface sampling oriÆce and/or injector tube of the torch.Matrix elements can also combine with the plasma gas or solvent components leading to interferences degrading quality of analysis. Alongside these, easily ionisable matrix elements can cause suppression effects in the plasma because of an increase in electron density. Considerable effort has been invested in the alleviation of these problems using on-line matrix separation and preconcentration techniques.3,4 On-line methods can themselves introduce other problems, mainly that the matrix separation/ preconcentration step takes between Æve and ten minutes.During this stage, the ICP-MS is idle and waiting for the sample to reach the plasma, thus wasting expensive resources. Although in general automated processes are preferred, often an off-line batch preparation procedure may be more cost effective. If the matrix separation/preconcentration can be shortened then the on-line technique could become more efÆcient.Miniaturisation can help speed up the system, but problems of poor recoveries and low volumes of eluent need to be overcome. A microconcentric nebuliser has been reported5,6 which provides stable sample introduction into an ICP at Øow rates °30 ml min21. This allows multielement determinations on a ml fraction of a sample. If such a nebuliser is employed in the online ICP-MS system, the sample size can be reduced and consequently the matrix separation/preconcentration step can be miniaturised leading to a reduction in analysis time.Nelms et al.7 successfully utilised a preconcentration column containing 0.04 g of a commercially available reagent in which imminodiacetate (IDA) is immobilised on controlled pore glass (Prosep) in a Øow injection manifold. Our communication8 recently described the preparation and evaluation of four columns prepared containing 0.0045±0.025 g of chelating agent.The small matrix separation/preconcentration column enabled the reduction in time of the sample preparation procedure from approximately 5 to 3 min with no loss of accuracy or precision. This was possible because the microconcentric nebuliser in the ICP-MS system permits the analysis of a smaller volume of solution. However, problems remain with the efÆcient packing of the mini-columns for the miniaturised Øow injection systems.It is well documented that 8-hydroxyquinoline (8-HQ) when covalently bonded on a support can be used to chelate a large number of metals under deÆned pH conditions. Supports used for this process include controlled pore glass,9 polymers,10 and silica.11 These materials were chosen such that there is no change in volume with changing pH or sample composition. The chelating surface is very reactive and is conditioned rapidly between samples, allowing a faster sample throughput.Sturgeon et al.12 immobilised 8-HQ onto silica gel for the preconcentration of a range of metals from seawater prior to determination by GF-AAS. The procedure provided a simple, rapid and reliable technique for the separation of the elements from sea-water. However, the batch capacity of the material was low but considered adequate for trace element determination. The capacity value reported for Cu was 0.061 mmol g21. Shan and co-workers13 described a novel support for immobilisation of 8-HQ to preconcentrate rare earth elements (REE) in sea-water prior to ICP-MS detection.The poly- J. Anal. At. Spectrom., 1999, 14, 1839±1842 1839 This journal is # The Royal Society of Chemistry 1999acrylonitrile (PAN) hollow Æber membrane immobilised with 8-HQ offers a relatively short preparation time and preconcentration of REE over a wide pH range. The Æbre was packed into a 6066 mm column for the preconcentration procedure. Samples are not buffered before analysis, reducing contamination, though extreme sample pHs could cause recovery problems from the column.The capacity of the 8-HQ PAN Æber membrane was determined to be 0.0715 mmol g21. Excellent recoveries were presented and agreement with previously analysed water samples. However, there is no mention of the time taken for the preconcentration or analysis of a sample. Recently a new method for the production of microporous silica structures has been reported14 for use in micro-reactor technology and it is suggested that this would provide an ideal support for 8-hydroxyquinoline in micro columns. This paper compares two rapid miniaturised matrix separation systems for ICP-MS with a microconcentric nebuliser, using packed IDA micro-columns and a novel micro-column immobilised with 8-HQ, for the analysis of trace elements in high salt matrices.Results regarding analysis time and effectiveness of the matrix separation procedures are reported, together with validation of the processes using certiÆed reference materials.Experimental Instrumentation The ICP-MS instrument used was a VG Elemental Plasma- Quad IIz (VG Elemental, Winsford, Cheshire, UK). Table 1 shows the operating conditions and measurement parameters for the ICP-MS instrument. The microconcentric nebuliser was an MCN-100 M2 (Cetac Technologies, Omaha, NE, USA) which was supplied with an end cap that Æts directly onto the Scott double pass glass spray chamber.The sample was fed using special low Øow capillary tubing supplied with the nebuliser. The polyamide nebuliser capillary was 150 mm in diameter with a dead volume of 2 ml. IDA packed micro-columns8 were made from glass capillaries, 0.2 mm in diameter, of varying lengths depending on the amount of packing material required. This compares to previous work13 where IDA columns were at least 3 mm in diameter and 2.5 cm in length. Microporous silica structures (frits) were prepared in glass capillaries 0.5 mm in diameter and 5 mm in length.Reagents Column packing materials were imminodiacetate immobilised on controlled pore glass (PROSEP Chelating-1 Bioprocessing, Consett, Co Durham, UK) and imminodiacetate immobilised on a cross linked polymer resin, (Muromac A-1 50±100 mm mesh, Muromachi Chemicals, Tokyo, Japan). Varying amounts of chelating material were slurry packed into glass capillary columns. Cleaned quartz wool was used for the end frits.The reagents 8-hydroxyquinoline, 3-aminopropyltriethoxysilane and p-nitrobenzoyl chloride (Sigma, Poole, Dorset, UK), sodium dithionite, sodium nitrite (Fisher ScientiÆc, Loughborough, Leicestershire, UK) and concentrated hydrochloric acid (Aristar, Merck Ltd., Poole, Dorset, UK) were used in the immobilisation procedure. Ammonium acetate buffer was prepared from the solid and puriÆed by passing through a column of Chelex-100. Microporous frits were fabricated from 10% formamide (Avocado Research Chemicals Ltd., Heysham, Lancashire, UK) and potassium silicate (21% SiO2, 9% K2O, Prolabo, Manchester, UK).National Research Council of Canada certiÆed reference samples (Ottawa, Canada) CASS-2 and SLEW-1 were used as received. High purity deionised water (18 MV cm resistivity, Elgastat UHQ PS, Elga, High Wycombe, UK) and super purity nitric acid (Romil, Cambridge, UK) were used throughout. Elemental stock solutions (1000 mg ml21, SpectrosoL, Merck) were used in the preparation of calibration and spiked solutions.Microporous frit preparation The fabrication of the frit for immobilisation was taken from a recent communication by Christensen et al.14 Formamide (18 ml) was mixed for 30 s with 140 ml of potassium silicate. The resulting solution was positioned in a glass capillary using a slow speed peristaltic pump (Minipuls 3, Gilson, Villiers-le-Bel, France). The frit was then placed in an oven for 1 h at 60 �C. Finally, an extra washing step, with dilute sodium hydroxide, was included to remove excess reagents.Immobilisation procedure The procedure for immobilisation of 8-HQ was taken from Nelms et al.15 This procedure for the immobilisation of 8-HQ onto controlled pore glass needed to be modiÆed because our micro-column frit was already prepared so the reagents were pumped through the frit using a peristaltic pump. Firstly, hot 10% v/v nitric acid was pumped through the frit to prepare the microporous silica structure. The frit was then dried in an oven at 80 �C and subsequently silanised by reaction with 20% v/v 3- aminopropyltriethoxysilane in anhydrous toluene for 30 min at room temperature.Again the frit was oven dried and reacted with 20% m/v p-nitrobenzoylchloride in chloroform for 24 h at room temperature. The frit was oven dried at 50 �C and further treated with a 20% boiling solution of sodium dithionite for 60 min, reducing the nitro group to the amine.After being oven dried at 50 �C, hydrochloric acid (2 mol l21) was pumped into the frit and reacted with sodium nitrite (2% m/v in water, slow pumping) at 0 �C to yield the diazonium salt. The Ænal stage was to pump 8-hydroxyquinoline (4% m/v in absolute ethanol) through the frit. A deep red colour in the frit indicated that immobilisation was successful and the diazo compound had been formed. The frit was Ænally washed with hydrochloric acid (2 mol l21) and water and stored in a desiccator until use.Determination of exchange capacity The capacity of the immobilised 8-HQ frit was determined for Mn by both a batch and a dynamic method using ICP-OES detection. Batch determination. The 8-HQ frit was removed from the glass capillary that it was formed in, powdered and weighed. This step was to ensure maximum interaction with the solution to obtain an absolute capacity value. A 0.05 g portion of immobilised 8-HQ frit was added to a solution of Mn (20 mg ml21) prepared in ammonium acetate buffer (0.05 mol l21). The solution was left to equilibrate overnight Table 1 Instrumental conditions and measurement parameters for the VG PlasmaQuad IIz ICP mass spectrometer Rf forward power/W 1350 ReØected power/W 0 Coolant gas Øow rate/l min21 14 Auxiliary gas Øow rate/l min21 1.2 Nebuliser gas Øow rate/l min21 0.880 Spray chamber Glass, water cooled at 4 �C Data acquisition mode Peak jumping Points per peak 3 Dwell time/ms 10.24 Detector mode Pulse counting 1840 J.Anal. At. Spectrom., 1999, 14, 1839±1842with stirring and the concentration of the resulting Mn solution was compared with the initial concentration. The decrease in the concentration of the solution was then used to calculate the capacity.15 Dynamic determination. Aliquots (1000 ml) of increasing concentration of Mn standard solution were injected onto the 8-HQ frit and eluted with 250 ml of 2 mol l21 HNO3 into the ICP-OES instrument.Three repeat analyses were made at each concentration and the results were evaluated in terms of peak height. The concentration beyond which no further increase in the emission intensity of the eluted peak was obtained was deemed to represent the dynamic capacity. This value was then used to calculate the dynamic capacity for the 8-HQ frit.15 Matrix separation procedure Aliquots of 100 ml of sample were injected into the buffer stream (1.5 mol l21 ammonium acetate) and loaded onto the column.The column was then Øushed with 2 mol l21 nitric acid. The Ærst 50 ml of acid stream was collected and measured by ICP-MS. The column was washed with nitric acid then with buffer before the next sample was loaded. Results and discussion Micro-columns were prepared containing IDA chelating reagents. These columns were slurry packed using a very slow pump speed to enable more efÆcient packing. The Ærst column contained 0.005 g Prosep and the second 0.0045 g Muromac.Although the potential of micro-columns packed with IDA chelating material has been shown in our previous communication,8 it is often the case that the reduction in analysis time created by the miniaturisation is outweighed by the difÆculty in efÆciently packing these columns. Frits were Ærst assigned for end stops for columns and to hold materials in place in micro technologies.12 Having looked at its structure it was considered an ideal support for a chelating reagent that could easily be used in a miniaturised system.The immobilisation procedure for immobilising 8-HQ onto controlled pore glass (cpg) was taken and applied to the immobilisation onto the frit structure. Initially no immobilisation was observed onto the frit and it was concluded that the porosity of the frit was too low. Ways of improving the porosity of the frit were looked at and it was found that by making the frits with 50% potassium silicate solution (diluted 1z1 with water), the porosity of the frits was increased to around 80%. The immobilisation procedure was then reapplied to the new frit.After the Ænal step it was clear that the immobilisation was more effective though not along the whole length of the frit. The deep red colour indicative of successful 8-HQ immobilisation appeared only at one end of the frit and a few tracks along its length. Due to the difference between cpg and the frit structure it was decided to change the immobilisation procedure.Owing to the fact that the reagents were being pumped through the frit, the time dependent stages of the procedure were increased. By increasing these time intervals, the interaction time between reagents and frit would also increase. Alongside this we decided to increase the concentration of the reagents as we thought that there may be more immobilisation sites available in the frit than the cpg. By increasing the time and concentration of the reagents immobilisation would be more successful. Once the procedure had been optimised successful immobilisation of 8- HQ along the whole frit was achieved.The immobilisation procedure is described in the experimental section of this paper. Capacity valuDA reagents has been well documented over the years.7,16 Capacity values in the order of 0.14 mmol g21 for Prosep and 0.22 mmol g21 for Muromac have been reported for Mn. Capacity of the immobilised 8-HQ frit was evaluated for Mn.A batch capacity value of 0.136°0.010 mmol g21 and a dynamic capacity value of 0.102°0.004 mmol g21 were measured for Mn. These capacity values are consistent with earlier reported values for similar 8- HQ immobilised materials. Nelms et al.15 reported a batch capacity value of 0.086 mmol g21 for 8-HQ immobilised on cpg. Sturgeon et al.12 and Marshall and Mottola's work17 showed values of 0.061 mmol g21 and 0.185 mmol g21 respectively for Cu on silica gel immobilised 8-HQ.Finally, Shan and co-workers13 presented a capacity value of 0.0715 mmol g21 for Cu PAN Æber membranes. The capacities with respect to other elements for the 8-HQ frit have not been investigated because the recovery results show that the capacity of the material is sufÆcient for a wide range of elements. The matrix separation procedures were optimised with respect to ammonium acetate buffer concentration, pH and acid eluent concentration for both the IDA micro-columns and the 8-HQ frit.The proÆles of the optimisation curves did not differ greatly from previous work15 where 8-HQ is immobilised on cpg for preconcentration prior to analysis by ICP-MS. Fig. 1 shows the variation in cps for Mn on changing the pH of the ammonium acetate buffer for the three miniaturised columns.A pH of 6 was chosen for the analysis. Decreased pH values gave reduced elemental recoveries and higher pH values could lead to the retention of matrix elements. A compromise set of conditions were selected for the miniaturised procedures: 1.5 mol l21 ammonium acetate buffer at a pH of 6.0 and an acid eluent concentration of 2 mol l21.The matrix separation Øow rate was investigated for each micro-column in order to reduce the analysis time. An optimum of 1 ml min21 was possible for both the IDA micro-columns resulting in a sample analysis time of 3 min including column reconditioning. The optimum Øow rate for the frit immobilised 8-HQ was 1.5 ml min21 resulting in an analysis time of 2.3 min.Increasing the Øow rate, hence reducing the analysis time further, led to decreased recoveries due to incomplete chelation and elution. Column washing and reconditioning periods contribute signiÆcantly to the total sample analysis time for the procedure and experiments were carried out to reduce this. It was found that by reducing this time to less than 60 s for the micro IDA columns produced unacceptable recoveries from spiked solutions.This was probably due to a sample to sample carry over effect and improper reconditioning of the column. However, a reconditioning time of 45 s for the 8-HQ frit still gave acceptable recoveries from a spiked solution. This is probably due to the quicker chelating recovery time when using immobilised 8-HQ. The total sample analysis time for the 8- HQ frit was 2.3 min, fractionally less than the micro IDA columns because of the more efÆciently packed column and the reduced reconditioning time. Once the procedures were optimised and the sample analysis times reduced signiÆcantly to 3 and 2.3 min for the different micro-columns, recovery experiments, matrix separation effectiveness and the analysis of certiÆed reference materials were studied.Recovery experiments were performed on a 5 ng ml21 solution containing the elements V, Cr, Mn, Co, Ni, Cu, Zn, Fig. 1 Effect of ammonium acetate buffer pH on Mn signal for miniaturised procedures.J. Anal. At. Spectrom., 1999, 14, 1839±1842 1841Cd, and Pb. Recoveries in the range 96 to 105% were observed for the 8-HQ frit. Recoveries between 71±101% and 91±102% were obtained for the Prosep and the Muromac micro-columns respectively. To determine the effect of repeated use on the columns, fresh columns were prepared and compared against the aged columns. There were no signiÆcant differences in the recoveries from a 5 ng ml21 solution after the columns had been in use for at least 100 h.To establish the efÆciency of the matrix separation procedure the 63Cu : 65Cu isotope ratios were determined. Procedures have been described7,13 for evaluating the inØuence of residual matrix on the ICP-MS results, using the 63Cu : 65Cu isotope ratio. If Na is present in the plasma the 63Cu : 65Cu is anomalously high due to the polyatomic overlap of 40Ar23Na on 63Cu or low due to the possible interference of 33S16O2 z on 65Cu. Table 2 shows that the matrix separation procedure is successful for all three micro-columns as the measured isotope ratio for 63Cu to 65Cu are close to the expected natural ratio.This can be difÆcult to achieve with the direct analysis of salt waters by ICP-MS. The IDA micro-columns and the immobilised 8-HQ frit column were used for the analysis of two reference materials, CASS-2 and SLEW-1. Six standards across the concentration range 0±15 ng ml21 (Mn), 0±1 ng ml21 (Co, Ni, and Cd) and 0±5 ng ml21 (Cu and Zn) were used.Table 3 details the correlation coefÆcient of the standards and the limit of detection calculated as three times the standard deviation of the blank plus the blank for the 8-HQ frit. Similar data were obtained for the other two IDA micro-columns. The elemental concentrations for the reference materials CASS-2 and SLEW- 1 are displayed in Table 4 for the 8-HQ frit and Table 5 for the Muromac micro-column. There is good agreement between the measured and certiÆed values.Conclusion A new miniaturised method incorporating the chelating agent 8-HQ has successfully been utilised to preconcentrate a suite of metals and separation of matrix components prior to analysis by ICP-MS. The frit structure provided an ideal support for 8-HQ avoiding the problems of efÆciently packing microcolumns. The 8-HQ frit column has been compared to miniaturised IDA based columns for the preconcentration of a range of metals in sea-water. The IDA micro-columns enabled the reduction in time of the sample preparation procedure from approximately 5 to 3 min with no loss of accuracy or precision.With the 8-HQ frit incorporated in the miniaturised system the sample analysis time is reduced further to 2.3 min. These results are possible because the microconcentric nebuliser in the ICPMS system permits the analysis of a smaller volume of solution. Further work will use the miniaturised 8-HQ frit system with a commercial automated sample preparation system so that when one sample is being analysed, the column is being washed and the next sample is loading.This would further reduce the sample preparation time giving a sample analysis rate of at least 30 samples h21. Such an automated system will reduce the time of sample preparation enabling efÆcient and cost effective analysis of multiple samples. Acknowledgements SDL thanks Avecia Ltd and EPSRC for funding this research. References 1 D. Beachemin, TrAC, Trends Anal.Chem. (Pers. Ed.), 1991, 10, 71. 2 J. W. McLaren, At. Spectrosc., 1993, 14, 191. 3 M. J. Bloxham, S. J. Hill and P. J. Worsfold, J. Anal. At. Spectrom., 1994, 9, 935. 4 D. Beauchemin and S. S. Berman, Anal. Chem., 1989, 61, 1857. 5 F. Vanhaecke, M. Van Holderbeke, L. Moens and R. Dams, J. Anal. At. Spectrom., 1996, 11, 543. 6 S. D. Lofthouse, G. M. Greenway and S. C. Stephen, J. Anal. At. Spectrom., 1997, 12, 1373. 7 S. M. Nelms, G. M. Greenway and D. Koller, J.Anal. At. Spectrom., 1996, 11, 907. 8 S. D. Lofthouse, G. M. Greenway and S. C. Stephen, Anal. Commun., 1998, 35, 177. 9 D. Beauchemin and S. S. Berman, Anal. Chem., 1989, 61, 1857. 10 J. A. Resing and M. Mottl, J. Anal. Chem., 1992, 64, 2682. 11 B. K. Daih and H. Huang, Anal. Chim. Acta, 1992, 258, 245. 12 R. E. Sturgeon, S. S. Berman, S. N. Willie and J. A. H. Desaulniers, Anal. Chem., 1981, 53, 2337. 13 B. Wen, X. Shan and S. Xu, Analyst, 1999, 124, 621. 14 P. D. Christensen, S.W.P. Johnson, T. McCreedy, V. Skelton and N. G. Wilson, Anal. Commun., 1998, 35, 341. 15 S. M. Nelms, G. M. Greenway and R. C. Hutton, J. Anal. At. Spectrom., 1995, 10, 929. 16 Y. Sung, Z. Liu and S. Huang, Spectrochim. Acta, Part B, 1997, 52, 755. 17 M. A. Marshall and H. A. Mottola, Anal. Chem., 1985, 57, 729. Paper 9/07308F Table 2 Isotope ratio measurements from the three micro-columns. Values quoted with range °2s (n~5), except natural ratio Isotope ratio 63Cu : 65Cu Prosep mico-column Muromac micro-column 8-HQ frit Natural ratio 2.24 2.24 2.24 Spiked pure water 2.28°0.05 2.20°0.05 2.26°0.05 Spiked sea water 2.26°0.05 2.29°0.05 2.28°0.05 Table 3 Calibration results for the 8-HQ frit Mn Co Cu Zn Cd RSD at 5 ng ml21 (%, n~5) 2.8 3.2 2.5 4.1 3.8 Correlation coefÆcient, r 0.9990 0.9992 0.9995 0.9990 0.9997 LOD/ng ml21 0.15 0.01 0.04 0.20 0.008 Table 4 Analysis results for reference material CASS-2 and SLEW-1 for 8-HQ frit. Concentrations in ng ml21. Uncertainty expressed as 2s of the instrument response to each analyte (95% conÆdence limit, n~3) CASS-2 SLEW-1 Element Measured CertiÆed Measured CertiÆed Mn 1.95°0.10 1.99°0.15 12.0°0.9 13.1°0.8 Co 0.024°0.004 0.025°0.006 0.044°0.005 0.046°0.007 Ni 0.301°0.009 0.298°0.036 0.752°0.070 0.743°0.078 Cu 0.670°0.090 0.675°0.039 1.79°0.15 1.76°0.09 Zn 2.00°0.10 1.97°0.12 0.90°0.10 0.86°0.15 Cd 0.022°0.004 0.019°0.004 0.019°0.002 0.018°0.003 Table 5 Analysis results for reference material CASS-2 and SLEW-1 for Muromac micro-column. Concentrations in ng ml21. Uncertainty expressed as 2s of the instrument response to each analyte (95% conÆdence limit, n~3) CASS-2 SLEW-1 Element Measured CertiÆed Measured CertiÆed Mn 1.90°0.10 1.99°0.15 10.5°0.9 13.1°0.8 Co 0.028°0.004 0.025°0.006 0.050°0.005 0.046°0.007 Ni 0.277°0.009 0.298°0.036 0.740°0.070 0.743°0.078 Cu 0.699°0.090 0.675°0.039 1.81°0.15 1.76°0.09 Zn 1.90°0.10 1.97°0.12 0.79°0.10 0.86°0.15 Cd 0.020°0.004 0.019°0.004 0.020°0.002 0.018°0.003 1842 J. Anal. At. Spectrom., 1999, 14, 1839±1842