A novel approach to non-segmented flow analysis: Part 4. Aluminium in river waters

A rapid and precise method is developed for the determination of aluminium in water. The results demonstrate that the calibration range of the assay can be extended by a simple manipulation of the control program of the flow analyser. An RSD of 1.8% is achieved for injection of standards and the theoretical limit of detection is estimated at 0.33 ppm alum (equivalent to 18 ppb Al3+). The method is applied to monitoring of environmental samples.


Introduction
Aluminium is the third most abundant element on the Earth's surface. It was brought to the public's attention recently, when, in 1988, a significant quantity of aluminium sulphate (alum) was mistakenly discharged into the water supply system to the village of Camelford in Cornwall. Local residents soon complained of adverse discomfort and hair discolouration. Alum, at more appropriate levels, is added to natural waters to remove impurities from peaty soils and to reduce treatment necessary by chlorination. Although it has been extensively reported in the media that aluminium is the probable cause of Alzheimer's disease, a recent report by the Aluminium Association in the USA has concluded that the cause of Alzheimer's disease is not known and ordinary environmental exposure to aluminium is safe. It should be noted that only 4% of the total exposure of aluminium to humans comes from natural waters and over 90% of it comes from food. Aluminium is leached from the soil by acidic precipitation, the mechanism of ionic transport is studied by Driscoll and Schecher [2].
Instrumental methods which are most widely used for the determination of aluminium include atomic absorption spectroscopy, inductively coupled plasma emission spectroscopy and neutron activation analysis. Photometric techniques for its determination are widely available. However, they rely on the reaction of a complexing agent with A1+, the latter is susceptible to hydrolysis in aqueous solutions. Below pH 5, the A1 a+ ion predominates, this slowly hydrolyses as the pH increases. At neutral pH, it was originally thought that soluble aquospecies of aluminium did not exist and they could only be stabilized by complexation or the formation of insoluble hydroxides [3]. Bertsch [4] conducted a detailed study of the hydrolytic products of aluminium and concluded that around neutral pH there exists a number of species, particularly polynuclear aluminium complexes, which are metastable in aqueous solutions. The actual structure and solution chemistry of these species are too complicated to be included within the scope of this study. Therefore, the determination of aquo-, as opposed to hydroxyl-, species of aluminium should be performed at or below pH 5, whereby their concentrations increase exponentially with decreasing pH. Another peril to aluminium determinations is the existence of complexing ligands which compete with aquo-ligands for the central A1 atom. Species which form strong complexes with aluminium include hydroxide and fluoride.
The number of organic reagents given in the literature that are suitable for the determination of aluminium photometrically is staggering, Burger [5] listed no less than 21 such chromogenic reagents in his book. Of those listed, only a handful exhibit high selectivity for only one or two metals, including aluminium, examples are aluminon, chrome azurol S (CAS), eriochrome cyanin R (ECR) and stilbazo. Up to the early 1970s, the reagent that was most widely used for the determination of aluminium in Britain was 'aluminon'. It was favoured for the stability of its coloured product and its wide pH tolerance range, but it suffered from such disadvantages as non-linear calibrations and heating for full colour development. Aluminon, together with other commonly used organic reagents, were comprehensively reviewed by Dougan and Wilson [6]. In one of the papers being reviewed, Pakalns compared the use of CAS, ECR and aluminon and concluded that CAS was the best on the grounds of simplicity, selectivity and reliability. In spite of this, Dougan and Wilson themselves discarded the use of CAS due to its high blank values and opted for pyrocatechol violet (PCV). Zhu et al. [7] highlighted the importance of pH as most of the organic reagents possess an extremely narrow optimum pH range. Marczenko and Jarosz [8] compared the use of three organic reagents: ECR, CAS and PCV in conjunction with various cationic surfactants and demonstrated that the addition of a third component, i.e. the surfactant, could enhance the sensitivity of aluminium determinations. They also calculated the stoichiometry of the ternary complexes. Liu [9] studied the effect of adding ethanol to a ternary system and observed a three-fold increase in the absorbance value in the presence of between 25 and 40% ethanol compared with no ethanol at all. Sampson and Fleck 10] again compared the use of some organic reagents with cationic surfactants and they used a CAS/Cetylpyridinium chloride (CPC) system for the determination of aluminium in dialysis fluids and water. Many of these methods have been adapted to flow injection analysis (FIA) systems and applied to a variety of sample matrices. The Brazilian pioneers in FIA, Reis et al., determined aluminium in soil and plant digest [11].
Wyganowski et al. used a combined reagent of bromopyrogallol red (BPR), n-tetradecyltrimethylammonium bromide (TDTA) and ethanol in a hexamine buffer using a dual-line manifold for analysing river water [12]. Royset compared the performance of some organic 0142-0453/91 $3.00 (C) 1991 Taylor & Francis Ltd. reagents in the presence and absence of a surfactant in 14 FIA [13] and adapted a system based on PCV [14]. Bouzid and Macdonald [15] used the CAS/CPC/ethanol combination and achieved phenomenal sensitivity. A comparison chart is also provided for various ternary 12 complexes of aluminium. Henshaw et al. [16] used a system, not dissimilar to the one employed by Royset, for the determination of 'monomeric aluminium species' in natural waters. 10

Experimental
The aim of our experiments was to devise a method for the routine monitoring ofwater for aluminium in the form of alum. Water authorities discharge around 20 ppm of alum into water supplies at source to clarify the water.
Hence, a calibration of up to 20 ppm of alum is required, corresponding to 1" 13 ppm of A1.
The flow analyser has remained virtually unchanged from that previously described [17]. The combined reagent comprising CAS/CPC/ethanol was chosen as it has proven to be sensitive, selective, versatile (in terms of pH tolerance) and possess a good linear range. The stock reagent solutions were prepared by dissolving 1"08 g of CAS and 0"64 g of CPC in separate litre flasks. The stock masking agent comprised of 100 g ofhydroxylammonium chloride (HOAC) with 1"0 g of 1,10-phenanthroline monohydrate (PAL) per litre. The stock buffer solution contained 420"57 g of hexamine with 50ml of 37% hydrochloric acid per litre. Both the masking agent and the buffer solution were prepared only when required. The stock alum standard solution was prepared by dissolving 1"00 g of aluminium potassium sulphate dodecahydrate in litre. Working solutions of the standard were prepared upon dilution of the stock standard giving solutions containing 5, 10, 15 and 20 ppm of alum standards. All the solutions were prepared in de-ionized water which was filtered through a 0"45 micron filter and de-gassed using nitrogen. A combined reagent system was prepared by mixing the following stock solutions: 10 ml CAS + 20 ml CPC + 3 ml masking agent + 3 ml buffer and 30 ml ethanol. The constituents were added in the order stated and the mixture was allowed to cool before topping up to 100 ml with water.
This reagent system had a pH of 5"0 + 0" and was stable for at least 24 hours. A pH of 5"0 was selected to ensure that the species analysed would be predominantly monomeric A13+, it also fell conveniently into the optimum pH range for the CAS/CPC combination [8].
Such high concentrations for the dye and surfactant were required as we elected to operate in the region ofoptimum stoichiometry of the A1-CAS-CPC ternary complex. The minimum two-fold excess of CAS and four-fold excess of CPC over A13+ was far exceeded. Inevitable consequences were relatively high blank values caused by a substantial reagent peak. A simple method was devised to tackle this particular problem.
In a previous paper [17] the authors reported on the determination of a wide range ofanalyte concentration by incorporating an extra valve into the manifold to perform in-line dilution of the sample reagent mixture, which had the effect of substantially extending the calibration range available. Although this method was shown to be successful, the authors have now developed a technique which requires no modification to the flow pattern, and so may be used to record usable data without prior knowledge of the analyte concentration range within the sample. A simple modification to the software overcomes the problem of high blank values and extends the usable calibration range. Tyson 18] studied the phenomenon ofreagent dispersion in a single-line manifold and explained the occurrence of double peaks, generated when the reagent and the sample components became inter-dispersed. In the authors' experiments, the sample-reagent mixture peak at low sample concentrations appeared as a small peak along the leading edge of the flow profile, and is followed by a larger peak caused by absorbance of the higher concentration reagent mixture. At higher sample concentrations the situation was reversed as the sample-reagent peak is much larger than the reagent peak. The time between the sample injection and the appearance of the first peak, the sample-reagent peak, was the same regardless of the concentration of the sample, as would be expected from basic FIA principles. In these experiments the product absorbance was recorded at a point before the maximum of the sample-reagent mixture peak and this value was used for the analytical measurement. This time was called the 'peak time'. This procedure ensured that the analytical measurement was derived from the samplereagent mixture and not influenced by excess reagent absorbance.  At a flow rate of 5 ml min-1, the leading edge of the sample arrived at the flow cell at less than 5 s after the injection of the sample. In order to establish the peak time, a preliminary run was conducted using de-ionized water as the sample. Various peak times were tested down to a minimum of s before the maximum ofthe first peak. The influence of the peak time on some calibration characteristics is listed in table for a series of five alum standards ranging from 0 to 20 ppm.
Mean RSD 1"79% The theoretical limit of detection (twice the standard deviation of the blank divided by the slope of the calibration) is estimated to be 0"328 ppm alum, which is equivalent to 18 ppb of A13+. However, none of the real samples used in this study have had alum levels close to the limit of detection.
A peak time of 4 s was selected for subsequent experiments. Note that a smaller peak time could be selected where lower sensitivity was required. Other flow analyser parameters used for the determination of alum, in line with the terminology that has been previously described, Having established that the calibrations were reproducible, the precision for each standard was measured and listed in table 3. (AU/ppm) (AU) Interference study Potential interferents were investigated for their effect on 10 ppm standard solutions of alum. Commonly interfering species which are most likely to be found in potable water include: iron2+, copper2+, fluoride, orthophosphate and detergents. The concentration of these substances were prepared to match very closely to the concentrations normally encountered in potable water.
All the substances looked at caused negative interference on the standard. The results are summarized in table 4.
Low level interference implies that the presence of these substances caused less than 10% suppression of the 10 ppm signal. Moderate implies between 10 and 20% suppression. Anything above 20% suppression is classified as extensive interference.
Fluoride interference is widely reported in the literature for the determination of A1. Due to its ionic radius, the fluoride ion has a high affinity for the central A1 atom. The stability constant for the equilibrium reaction: The influence of fluoride extends from pH 4"5 to pH 6"5. Not surprisingly, the fluoride interference cannot be easily overcome. Iron interference was minimized in the conventional way. A pH of 5"0 is well within the operational range of PAL but the level of Fe + causing moderate interference has exceeded the amount normally

Analysis of water samples
Samples of water were collected along the River Stour in Kent at 12 different sites over a period of several months for the determination of alum. Week one was given the prefix '01' and the first site was given the suffix 'A'. For example, '03F' denotes a sample gathered from site F during the third week of collection. The samples were collected over a few months to observe periodic fluctuations; the purpose of amassing samples from different sites along the river was to investigate the effect of the surroundings on the level of alum. The sites are as follows:

K:
A: River mouth, eddy current along the bank, silty Samp. no. 19A 19B 19C 19D 19E 19F Alum, ppm 13"3 17"9 11"0 13"3 13"2 9"0 Samp. no. 19G 19H 191 19J 19K Alum, ppm 14"2 6"0 10" 10"3 8"9 appears to be a correlation between the level of alum measured and the tidal rise along different stretches of the river. For example, the highest concentration was found at site B where the tidal rise was 8 feet; the alum levels rose again from C to D and from F to G where the tidal levels also increased from the preceding site. It must be stressed, however, that this correlation could not be substantiated in all cases. Concentration at site I was lower than that ofsiteJ although the latter had a tidal rise of foot, perhaps because samples I and J were collected just before and just after a sewage works respectively. Despite the absence of solid correlation, an overall downward trend is observed indicating that the level of alum has been diluted gradually (or leached into the surrounding soil) along the river.
In existing FIA methodologies for the determination of aluminium, Wyganowski et al. 12] used a reagent in 60% ethanol to produce calibrations of 0-0"3 (0-5"31) and 0-0"1 ppm A1 (0-1"77 ppm alum). In his comparison of various organic reagents for the flow injection determination of A1, Royset [13] had not provided any linearity data. Royset revealed that the aluminon method was able to tolerate up to 500 ppm P without causing significant interference. Unfortunately, this method was unsuitable for A1 a+ concentrations ofless than 100 ppb. Fluoride can only be tolerated at levels of<0"2 ppm in all the methods.
The same author used pyrocatechol violet [14] and obtained linear calibration from 0"1-3 ppm A1 (1"77-53"1 ppm alum) upon 70 and 200 microlitre injections. The linear range was extended, presumably from 0" ppm of A1, to 10 ppm A1 (177 ppm alum) when 10 microlitre was injected. A 6-m coil and a 30 s reaction time was employed. Furthermore, the PCV mixed reagent was unstable for periods in excess of two hours.
Bouzid and Macdonald [15] abandoned this system on the basis of its narrow operational pH range, 6"0 + 0"1 pH unit. They instead opted for the CAS/CPC/ ethanol system. The buffer that they had used unusually comprised of hexamine/ammonia, in place of the commonly used hexamine/hydrochloric acid to keep the pH at around 6. A linear range of 0-0.4 ppm A1 (0-7.08 ppm alum) was obtained with a limit of detection of 5 ppb A1 (88"5 ppb alum). For the calibration from 0 to 0"2 ppm A1 (0-3"54ppm alum), a slope of 28"8AU/ppm was obtained, yielding a possible signal of>5'7 AU at the top of the calibration range. The  In their studies, a 0.07 ppm of fluoride spiked into a 0" ppm aluminium solution gave a recovery of 110% of aluminium whereas a 0"7 ppm fluoride spiked into a solution of the same concentration caused only 1% positive interference. This appears to suggest that fluoride has an inhibitory effect at lower concentrations but once this has been exceeded, the level of interference is reduced.

Conclusion
The present automatic flow analyser offers increased performance compared with other FIA instruments in terms of calibration range, precision and sample throughput for the determination ofaluminium. The sensitivity of the assay and its ability to cope with interferents is also as good as existing instruments. In these experiments the versatility of the prototype has been demonstrated by its success in analysing concentrations of aluminium up to exceptional levels and its ability to cope with intensely coloured reagents.