Indirect determination of trace amounts of fluoride in natural waters by ion chromatography: a comparison of on-line post-column fluorimetry and ICP-MS detectors María Montes Bayón, Ana Rodríguez Garcia,† J. Ignacio Garc�ýa Alonso* and Alfredo Sanz-Medel Department of Physical and Analytical Chemistry, University of Oviedo, c/Julián Clavería 8, 33006 Oviedo, Spain Received 10th September 1998, Accepted 20th November 1998 An alternative method for the determination of trace levels of fluoride in drinking and sea-water samples is presented. It is based on the formation of the aluminium monofluoride complex in excess of Al3+ and separation of the two species formed (AlF2+ and Al3+) in a small (5 cm long, CG2) ion exchange guard column.The final determination is accomplished by both ICP-MS specific detection and post column derivatisation with fluorimetric detection. Fundamental studies on the formation kinetics of the complex, ion chromatographic separation and optimum aluminium concentration were carried out using spectrofluorimetric detection by post-column reaction of the species with 8-hydroxyquinoline-5-sulfonic acid in a micellar medium of cetyltrimethylammonium bromide.Fluorimetric detection showed good detection limits, but interferences from cations such as Mg2+ and Zn2+ required the use of the longer CS2 ion exchange column. Iron interfered in relatively large amounts but adding EDTA to the sample solution eliminated the interference.A similar separation methodology was applied using ICP-MS detection for the indirect determination of fluoride, by monitoring aluminium at mass 27. In this case, a detection limit of 0.1 ng ml21 was obtained using 0.45 m HNO3 as eluent and no interference caused by high concentrations of iron was observed. The proposed method was applied to the determination of very low levels of fluoride in natural waters. Introduction During the last decade, the majority of fluoride determinations have been performed using techniques such as potentiometry with fluoride ion selective electrodes (ISE),1,2 ion-exchange chromatography with conductivity detection,3–4 spectrophotometry5 and most recently capillary electrophoresis.6,7 The use of ISEs has been the preferred technique for this determination, but their sensitivity is insufficient to measure fluoride at ng ml21 levels. A spectrophotometric method using SPADNS has been applied to drinking waters.5 In the case of ion chromatography, the weak binding affinity of fluoride to the ion exchangers used to perform the separation process causes its early elution from the column, too close to the so-called injection peak, containing non-retained compounds and also the sample solvent.Most interferences in fluoride determination come from the presence of high levels of iron or aluminium in the sample. In these cases distillation of fluoride as HF can be performed.8 Atomic spectrometric techniques have not been used so far for direct fluoride determinations.The high excitation and ionisation potentials presented by this halogen resulted in poor sensitivity for atomic emission spectrometric (AES) detection even using powerful spectrochemical sources such as heliumbased plasmas (e.g., He MIP or He ICP). In this respect, good detection limits have been achieved for other halogens such as chloride or bromide,9,10 but no results have been reported so far on fluoride determinations by AES.One interesting and recently developed alternative involves fluoride determination by electrospray mass spectrometry, with promising results.11 Several groups have investigated indirect fluoride determination. Marco et al.12 determined fluoride by measuring the molecular absorption of the AlF2+ complex in a graphite tube using a Pt hollow cathode lamp. The chromatographic separation of Al–fluoride species was first described by Bertsch and Anderson,13 who determined the stability constants of the several possible AlFx species.Later, Jones14 determined fluoride as AlF2+ after chromatographic separation from the excess of Al3+ using indirect fluorimetric detection with 8-hydroxyquinoline-5-sulfonic acid, obtaining detection limits in the low ppb range. Previous work in our laboratory15 showed that aluminium could be better detected in the presence of cationic micelles of cetyltrimethylammonium bromide (CTAB) and it was also observed that the aluminium monofluoride complex could be detected by this fluorimetric reaction after ion chromatographic separation from Al3+.16, 17 Here, the optimum conditions for the formation of the AlF2+ complex were studied using ion chromatography and post-column fluorimetric detection.Two types of elution conditions were evaluated, using K2SO4 14,16 and HNO3 as eluents. The analytical characteristics using fluorimetric detection were compared with ICP-MS as a specific detection method monitoring aluminium at m/z 27.Examples of the application of the proposed method are presented for the determination of low fluoride levels in drinking and sea-water samples. Experimental Instrumentation The chromatographic system used consisted of a Pharmacia (Uppsala, Sweden) Model P-500 medium pressure pump and an Model 5M PA inert valve from Pharmacia fitted with a 100 ml sample loop and a 5 cm long Dionex (Camberley, UK) Ion Pac † Present address: Ingenieros Asesores SA, Polígono Silvota, Llanera, Asturias, Spain.Analyst, 1999, 124, 27–31 27HPIC-CG2 ion exchange column. For the experiments using the K2SO4 as eluent, the column was immersed in a water-bath at 50 °C. The spectrofluorimetric detector was a Shimadzu (Kyoto, Japan) R-F5000 equipped with a 12 ml flow cell. The excitation and emission wavelengths were 390 and 500 nm, respectively, and the chromatograms obtained were recorded using a Shimadzu Chromatopac C-R3A integrator.The post-column reagent was pumped using a Scharlau (Barcelona, Spain) HP4 peristaltic pump. Al-specific ICP-MS detection was carried out using an HP 4500 instrument (Hewlett-Packard, Yokogawa Analytical Systems, Tokyo, Japan) fitted with a concentric nebuliser and a Peltier cooled (2 °C) spray chamber. All chromatograms were obtained monitoring aluminium at m/z 27 using time resolved analysis. Fig. 1 shows the instrumental set-up of the system using alternatively fluorimetric or ICP-MS detection.Neither detection system could be used on-line because of incompatibility of the eluents and post-column reaction. For fluorimetric detection, the eluent from the column was mixed with the postcolumn reagent using a T-piece and a 2 m 3 0.5 mm id PTFE reaction coil. For ICP-MS detection, the eluent leaving the column was connected directly to the nebuliser. Reagents All chemicals were of analytical-reagent grade unless stated otherwise and water obtained from a Milli-Q system (Millipore, Molsheim, France) was used to prepare stock standard solutions of all the reagents.Aluminium standard solution (1000 mg ml21) was obtained from Merck (Darmstadt, Germany). Stock standard solutions of F2 (1000 and 1 mg ml21) were prepared by dissolving solid NaF (Merck) in water. For fluorimetric detection, K2SO4 (Merck) was used to prepare the mobile phase. The post-column reagents, 8-hydroxyquinoline- 5-sulfonic acid (HQS) and CTAB, were obtained from Sigma Aldrich, (St.Louis, MO, USA). The pH of the postcolumn reagent was adjusted using acetic acid–sodium acetate buffer (Merck). To prevent interferences from other metals present in the sample such as Fe, a standard solution of EDTA (Sigma Aldrich) was also used. For ICP-MS determination, no post-column reagent was necessary and the mobile phase used was prepared from 65% HNO3 (Suprapur, Merck) and diluted with Milli-Q water.Preparation of AlF2+ complex Samples and standard solutions were adjusted to pH 3 with nitric acid and spiked with Al3+, at least a five-fold mass excess of Al to fluoride being required to ensure that only AlF2+ was formed. The samples were diluted by volume (fluorimetric detection) or mass (ICP-MS detection), transferred into 10 ml polypropylene test-tubes and immersed in a water-bath at 50 °C for 60 min. Th ensured quantitative formation of the AlF2+ complex. For natural water samples and fluorimetric detection, EDTA was added at 1.6 3 1025 m.Chromatographic separation For fluorimetric detection, the mobile phase was 0.1 m K2SO4 in water adjusted to pH 3 with nitric acid. A flow rate of 1 ml min21 was used. The CG2 HPLC column was immersed in a water-bath at 50 °C following the recommendations of Jones et al.18 For ICP-MS detection, the mobile phase was 0.45 m HNO3 at a flow rate of 0.5 ml min21 and the column was kept at room temperature. Spectrofluorimetric detection The post-column reagent contained HQS and CTAB at optimum concentrations of 1 3 1023 and 2 3 1023 m, respectively and the optimum pH of 6 was adjusted with 0.25 m acetic acid–sodium acetate buffer.The optimum flow rate was 0.43 ml min21. The excitation and emission wavelengths selected were 390 and 500 nm respectively, providing the highest analytical signals for AlF2+ determination. The excitation and emission slits were both 5 nm. ICP-MS detection The ICP-MS operating conditions are summarised in Table 1. The output from the column was fed directly to the inlet of the concentric nebuliser and the m/z value monitored was 27 using the time resolved analysis mode, a 0.5 s integration time and 1 point per mass unit.Results and discussion Kinetic studies The initial aim of this work was to obtain chromatographic conditions that allowed the separation and detection of the AlF2+ complex from the excess Al3+, to provide a sensitive method for indirect fluoride determination. First, some studies on the formation kinetics were performed and the determination Fig. 1 Instrumental set-up of the system using (a) fluorimetric and (b) ICP-MS detection. Table 1 Typical operating conditions Instrument Rf power Nebuliser Spray chamber Sampling depth Gas flow rates— External Intermediate Carrier Ion lens settings— Extract 1 Extract 2 Einzel 1, 3 Einzel 2 Omega bias Omega (+) Omega (2) QP focus Ion deflector Oxide level (CeO+/Ce+) Double charged level (Ce2+/Ce+) HP 4500 1300 W Meinhard Scott type, double pass, cooled at 2 °C 5.7 mm 15 l min21 1 l min21 1.17 l min21 2221 V 2106 V 2144V 39.3V 248V 6 V 27 V 8 V 39 V < 0.5% < 1% 28 Analyst, 1999, 124, 27–31of the complex was carried out using spectrofluorimetric detection.In aqueous acidic solution, aluminium ions are present as [Al(H2O)6]3+, which can react with F2 to form the AlF2+ complex. It has been demonstrated1 that in a highly acidic medium, F2 reacts with H+ to form HF, leading to a decrease in the rate of complexation of Al3+ with F2.At pH > 3.0, the hydrolysis reaction of Al3+ can take place with the formation of Al(OH)i (32i)+, which reduces free aluminium and therefore the concentration of the complex. The optimum pH for the complex formation seems to be between 2 and 4,1,14 and therefore in this work the pH selected was 2.6–3, where the complex AlF2+ proved to be stable. Under these conditions, several parameters were evaluated in order to determine the formation kinetics of the complex, such as temperature and excess of aluminium necessary to obtain the quantitative formation of the AlF2+.Fig. 2 shows the fluorimetric peak heights obtained for the AlF2+ complex as a function of solution temperature (15, 22 and 50 °C with solutions containing 100 ng ml21 fluoride and 1 mg ml21 aluminium) and complexation time. As can be appreciated, on heating the solution containing fluoride and aluminium at 50 °C, the formation of the complex can be considered quantitative after 50 min.At room temperature, more than 5 h are necessary to obtain stable signals for the complex. The slow reaction kinetics of the AlF2+ complex formed at room temperature could allow its separation from the excess of Al3+ without any decomposition or formation of alternative species during passage through the chromatographic column. In order to optimise the aluminium concentration to be added for complete formation of the complex, solutions containing 200 ng ml21 fluoride were tested.Increasing amounts of aluminium were added to each sample and the solutions were heated at 50 °C for 1 h. The results obtained are shown in Fig. 3 and were measured fluorimetrically as peak height. As can be observed, a plateau was reached when using an aluminium concentration of 1 mg ml21 or higher. In order to increase the linear range as much as possible, a concentration of 10 mg g21 of aluminium was used in subsequent studies using fluorimetric detection.Ion exchange separation: study of the mobile phase Fig. 4 shows the typical chromatograms obtained for the separation of AlF2+ and Al3+ using 0.1 m K2SO4 as eluent and increasing amounts of fluoride from 0 to 400 ng ml21. As can be observed, using 0.1 m K2SO4 the AlF2+ peak increases with increase in fluoride concentration and the excess Al3+ elutes after 2.5 min.In order to optimise the concentration of K2SO4 in the eluent, several concentrations were evaluated and Fig. 5 shows the standard representation of the logarithm of the capacity factor (log kA) versus the negative logarithm of potassium concentration (2log [K+]).19 As can be observed, the slopes of the lines are 1.90 and 3.12 for AlF2+ and Al3+, respectively, and therefore it seems clear that the charge of the compounds is +2 and +3, respectively, and the structure of the complex is the one proposed. Fig. 2 Fluorimetric peaks heights obtained for the AlF2+ complex as a function of solution temperature (D, 15; -, 22; and 2, 50 °C) containing 100 ng ml21 fluoride and 1 mg ml21 aluminium. Fig. 3 Optimisation of the aluminium concentration required for complete formation of the AlF2+ complex in a solution containing 200 ng ml21 fluoride. Fig. 4 Typical chromatograms obtained for the separation of AlF2+ and Al3+ using 0.1 m K2SO4 as eluent and increasing amounts of fluoride from 0 to 400 ng ml21.Fig. 5 Representation of the logarithm of the capacity factor (log kA) obtained both for AlF2+ and Al3+ using different K2SO4 eluents. The slope of the log–log plot provides the charge of the species.19 Analyst, 1999, 124, 27–31 29Because saline solutions, such as K2SO4, are not recommended in ICP-MS (possible clogging of the central channel of the torch and deposits on the cones), alternative mobile phases were tested. HNO3 was found to be a good eluent for the separation of AlF2+ and Al3+ and detection by ICP-MS.Different concentrations of nitric acid, from 0.15 to 0.75 m, were tested for the above mentioned separation. The conditions chosen for subsequent studies were 0.45 m nitric acid at a flow rate of 0.5 ml min21 and detection at m/z = 27. The chromatogram obtained under these conditions for 20 ng g21 F2 in the presence of 100 ng g21 of Al is shown in Fig. 6. As can be observed, two aluminium containing peaks are detected.The first peak could be ascribed to the AlF2+ complex as its peak height/area was found to be proportional to the concentration of fluoride in the sample. Analytical performance characteristics Analytical performance characteristics for both detection modes are summarised in Table 2. The linear dynamic range for fluoride determination depends on the excess of aluminium added to the sample. It was observed with both detection modes that, for a given aluminium concentration, the upper linear limit for fluoride determinations was about one fifth of the total aluminium concentration.For ICP-MS, aluminium concentrations higher than 500 ng g21 were not tested. Using fluorimetric detection, linear upper limits up to 2000 ng ml F2 were obtained (using 10 mg ml21 excess Al3+). The detection limits obtained were 0.6 ng ml21and 0.1 ng g21 for fluorimetry and ICP-MS, respectively, calculated as three times the standard deviation of the blank divided by the slope of the linear calibration graph between 0 and 5 ng g21.The detection limit using ICP-MS detection is one of the lowest ever reported for the determination of fluoride. Interference studies An exhaustive study of possible interferences from other anions and cations was carried out first using spectrofluorimetric detection. The results are summarised in Table 3. As can be observed, typical anions present in natural waters such as HCO32 and Cl2 do not interfere at the maximum concentration tested (200 mg ml21).Other anions such as H2PO42 and BO3 32 can be present at 100 and 200 mg ml21, respectively, without causing interference. The main interferences were observed from trace metals, which can be present in natural waters. Fe3+ at concentrations higher than 0.5 mg ml21 decreased considerably the peak height from the AlF2+ peak owing to competition with Al3+ for fluoride or quenching of the fluorimetric reaction. However, in the presence of 1.6 3 1025 m EDTA this interference was eliminated.Ca2+ and Sr2+ up to 100 and 50 mg ml21, respectively, did not show any interference effect. Cu2+ and Pb2+ at concentrations higher than those in Table 3 produced a small decrease in the fluorimetric signal of AlF2+, which can be ascribed to co-elution with AlF2+ and the formation of competing chelates with the post-column reagent. On the other hand, Zn2+ and Mg2+ formed fluorescent chelates with the postcolumn reagent and co-eluted with AlF2+ using the 5 cm CG2 column. The use of a 25 cm CS2 column and a lower eluent concentration of 0.05 m K2SO4 resulted in the separation of the AlF2+ peak from Mg2+ and Zn2+.However, in this case the retention time for Al3+ increased to 30 min, as shown in Fig. 7 for a real water sample. Using ICP-MS aluminium specific detection, no effect on detection from co-eluting divalent cations should be expected. The only possible interference could be with the formation and retention of AlF2+ complex. No effect of Fe3+ on the height or area of the AlF2+ peak was obtained for Fe3+ concentrations up to 100 ng g21 at the same aluminium concentration and 20 ng g21 of fluoride.Also, when monitoring at m/z 57, Fig. 6 Chromatogram obtained for 20 ng g21 of fluoride in the presence of 100 ng g21 of aluminium using ICP-MS detection. Eluent, 0.45 m nitric acid. Table 2 Comparative analytical performance characteristics of spectrofluorimetric and ICP-MS detection Spectrofluorimetric ICP-MS Analytical characteristics detection detection Detection limit Precision Linear range Regression coefficient (r) (n = 7 points) 0.6 ng ml21 2.3%a Up to 2000c ng ml21 0.9995 0.1 ng g21 4%b Up to 100d ng g21 0.9993 a For five injections of 20 ng ml21 fluoride.b For five injections of 20 ng g21 fluoride. c Using 10 mg g21 aluminium. d Using 500 ng g21 aluminium. Table 3 Effect of foreign ions on the determination of fluoride with spectrofluorimetric detection (200 ppb F2, 10 ppm Al3+) Maximum concentration allowed/mg ml21 Interference (recovery 100 ± 5%) Fe3+ Zn2+ Mg2+ Cu2+ Ca2+ Sr2+ Pb2+ HCO2 (as NaHCO3) Cl2 (as KCl) H2PO42 (as NH4H2PO4) BO3 32 (as Na3BO3) 0.5 (10, in the presence of 1.6 3 1025 m EDTA) Interferes 10 1.6 100a 50a 2 200a 200a 100a 200a a Maximum concentration tested.Fig. 7 Chromatogram obtained for a real water sample using the CS2 column (25 cm long) and 0.05 m K2SO4 as eluent with fluorimetric detection. 30 Analyst, 1999, 124, 27–31representing a minor iron isotope, no elution of Fe from the column could be detected owing to insufficient sensitivity. Wang et al.1 have reported that the formation of the Fe–fluoride complex is strongly pH dependent. The optimum pH for iron– fluoride complex formation is 1.5 but at pH 2.5 complex formation decreased significantly.1 Our results on pH effects agreed well with such observations. Determination of fluoride in fresh and sea-water samples Fluorimetric detection.Natural water samples may contain a large range of fluoride concentration from a few ng ml21 to several mg ml21. The concentrations of interfering compounds should be well below those indicated in Table 3 except, perhaps, for Fe, Mg and Zn in some samples. The chromatogram of a mineral water sample diluted 1 + 9 with Milli-Q water and containing 3.1 mg ml21 of fluoride determined by ISE is shown in Fig. 7; 0.05 m H2SO4 as eluent and the CS2 column were used.As can be observed, the AlF2+ peak is well separated from Mg2+ and Zn2+. The concentration found in this sample by reference to a calibration graph was 3.1 ± 0.1 mg ml21 (n = 5) and the recovery for spiked fluoride was 97%. Unfortunately, under these conditions the retention time for Al3+ was > 30 min, increasing dramatically the time required for each measurement as Al3+ must be eluted from the column before the next injection.ICP-MS detection. Under the optimum separation conditions using HNO3 for elution, the retention time for Al3+ is < 8 min, allowing a sampling rate of 6–7 h21, and the detection is much more selective. Therefore, the proposed ICP-MS method was applied to the determination of fluoride in natural and drinking waters from a variety of sources and with different saline concentrations. As no stable aqueous fluoride reference material was available, it was decided to compare the proposed methodology with the fluoride ion selective electrode (FISE) and spiking the samples with fluoride to calculate the recoveries.In order to minimise aluminium addition to the samples and contamination of the ICP-MS system, up to 200-fold dilution of some drinking and sea-water samples was necessary. The results obtained are summarised in Table 4. As some of the concentrations found were around 150 ng g21, and sometimes lower, the determination using FISE was adequate in only a few cases. As can be observed, the results obtained were in good agreement with the values found by FISE, when this determination was possible.In the other cases tested, recoveries of 100 ± 10% were obtained, showing the applicability of the proposed methodology to perform fluoride determinations at extremely low levels. Conclusions The formation of the AlF2+ complex in excess of Al3+ can be considered quantitative after thermal treatment of the sample for 1 h at 50 °C.Once formed, the complex is stable and can be separated from the excess of aluminium by cation exchange chromatography without decomposition even using highly acidic eluents (0.45 m HNO3). The detection of aluminium can be accomplished either by a post-column fluorimetric reaction, with interferences from other cations such as Fe3+, Zn2+ and Mg2+, or by ICP-MS. The latter detection method proved to be extremely sensitive, with a detection limit of 0.1 ng g21 of F2, and free from interferences from other cations and anions in natural water samples. In comparison with the fluoride ISE, the proposed indirect ICP-MS method is at least two orders of magnitude more sensitive, it is not affected by interferences from aluminium and iron (in fact, the formation of the AlF2+ complex is the basis of the method) and it can be applied without modification to a large range of water samples of different salinity.Acknowledgements We thank Hewlett-Packard for the loan of the HP 4500 instrument and the DGCYT (Madrid) for financial support through Project DG-94-PB-1331.References 1 H. Wang, Z. Zhang, A. Sun, D. Liu and R. Liu, Talanta, 1996, 43, 2067. 2 R. W. Kahama, J. J. M. Damen and J. M. ten Cate, Analyst, 1997, 122, 855. 3 T. A. Biemer, N. Asral and A. Sippy, J. Chromatogr. A, 1997, 771, 355. 4 J. M. Talmage and T. A. Biemer, J. Chromatogr. A, 1987, 410, 494. 5 S. A. Sen, K. Kesava Rao, M. A. Frizzell and G. Rao, Field Anal.Chem. Technol., 1998, 2(1), 51. 6 P. Wang, S. F. Y. Li and H. K. Lee, J. Chromatogr. A., 1997, 765, 353. 7 S. A. Shamsi and N. D. Danielson, Anal. Chem., 1995, 37, 1845. 8 Standard Methods for the Examination of Water and Waste Water, American Public Health Association, New York, 15th edn., 1980, p. 337. 9 F. Camuña, J. E. Sánchez-Uría and A. Sanz-Medel, Spectrochim. Acta, Part B, 1993, 48, 1115. 10 A. H. Mohammed, J. T. Creed, T. M. Davidson and J. A. Caruso, Appl. Spectrosc., 1989, 43, 1127. 11 D. A. Barnett and G. Horlick, J. Anal. At. Spectrom., 1997, 12, 497. 12 V. Marco, F. Carrillo, C. Pérez-Conde and C. Cámara, Anal. Chim. Acta, 1993, 283, 489. 13 P. M. Bertsch and M. A. Anderson, Anal. Chem., 1989, 339, 535. 14 P. Jones, Anal. Chim. Acta., 1992, 258, 123. 15 J. I. Garcia Alonso, A. López García, E. Blanco González and A. Sanz-Medel, Anal. Chim. Acta, 1989, 225, 339. 16 J. I. Garcia Alonso, A. Rodriguez Garcia and A. Sanz-Medel, paper presented at Euroanalysis VII, Vienna, 1990. 17 A. Rodriguez Garcia, Master’s Degree, Faculty of Chemistry, University of Oviedo (1991). 18 P. Jones, L. Ebdon and T. Williams, Analyst, 1988, 113, 641. 19 R. Rosset, H. Conde and A. Jordy, Manuel Pratique de Chromatographie en Phase Liquide, Masson, Paris, 2nd edn., 1982. Paper 8/07079B Table 4 Results obtained for the determination of fluoride in water samples using ICP-MS after dilution of the samples Water sample (dilution factor) Concentration found (n = 3) by ICP-MS/ ng g21 Concentration found by FISE/ng g21 Spiked amount/ ng g21 Recovery (%) Fontecelta (200) Font-Vella (10) Tap water (10) Sea-watera (100) 8050 ± 80 182 ± 2 161 ± 1 1030 ± 60 7700 — — 1080 4300 205 210 1080 104 97.8 90 97.5 a Collected at Gijon, Asturias, Spain. Analyst, 1999, 124, 27–31 31