Michael
Gardner
* and
Sean
Comber
WRc-NSF, Henley Road, Medmenham, Marlow, Buckinghamshire, UK SL7 2HD. E-mail: gardner_mj@wrcplc.co,uk; Fax: + 44 1491 579094; Tel: + 44 1491 636555
First published on 11th December 2001
A simple and sensitive solvent extraction-atomic spectrometric technique has been developed for the determination of hexavalent chromium in fresh and saline waters. The technique is based on the reaction of chromium with diphenylcarbazide. The method has been tested on a variety of water samples over an analytical range of 0–2 μg l−1. A limit of detection of 0.024 μg l−1 was achieved. Spiking recoveries in the range 87–115% were achieved in river water, drinking water and marine waters.
Public concern has been expressed in the US in relation to possible exposure of drinking water consumers to hexavalent chromium. In 1999, the US Office of Environmental Health Hazard Assessment of the Environment Protection Agency established a public health goal of 2.5 μg l−1 for total chromium, based on a health protective level of 0.2 μg l−1 for hexavalent chromium (derived for a cancer endpoint) and the assumption that the hexavalent chromium is no more than 7% of the total chromium. However, a limited study of drinking water sources conducted in late 1999 indicated that the average percentage of hexavalent chromium may be above 50%.2 At present, in the UK and other EU states concentrations of total chromium in drinking water are monitored for compliance with a limit concentration of 50 μg l−1. The UK environmental quality standard for total chromium in surface waters is set at 15 μg l−1, though there may be concern about lower concentrations if the metal were present as the hexavalent form. A reduction in the concentration of interest and a focus on CrVI species generates a requirement for analytical methodology suitable for monitoring purposes in both drinking waters and surface waters. This paper describes a procedure, which was developed and tested with the aim of meeting this requirement.
The methodology for the determination of CrVI by colorimetry using diphenylcarbazide is well established.11 Direct spectrophotometry can be used to determine CrVI in clean waters down to a limit of detection of approximately 2–3 μg l−1. This is not adequate to monitor compliance with quality standards or limit values set at the low levels discussed above. This work aimed to develop the diphenylcarbazide methodology for the determination of CrVI, principally by extending it to sub-microgram per litre levels using preconcentration by solvent extraction.
Diphenylcarbazide gives a sensitive and specific colour reaction with CrVI in mineral acid solution. The pink coloured chromophore is a chelate of CrIII and diphenylcarbazone. The latter is produced and simultaneously combines with chromium when diphenylcarbazide is oxidised by CrVI. The reaction may be summarised as:
2CrO42− + 3H4L + 8H+ = [CrIII (HL)2]+ + Cr3+ + H2L + 8H2O |
All water was deionised and all chemicals were of reagent grade. All apparatus was pre-soaked in 5% v/v nitric acid and rinsed with deionised water before use.
Extractions were performed in batches of 24 samples with the sample tubes held and shaken in a laboratory tube rack. After leaving the samples for at least half an hour for the solvent layer to separate, 0.5 ml of the upper alcohol layer was pipetted off and transferred for analysis by electrothermal atomic absorption spectrophotometry. Standard solutions in deionised water at concentrations of 0, 0.1, 0.5, 1 and 2 μg l−1 were extracted along with samples. All determinations were made on a Perkin-Elmer 4000 atomic absorption spectrometer and HGA 400 atomiser at 357.9 nm, with a 0.7 nm bandpass. Background correction was carried out using a deuterium lamp. The furnace programme used is shown in Table 1.
For the purpose of the performance tests reported here, five different water samples were analysed in duplicate, unspiked and spiked with CrVI at 0.5 and 2 μg l−1 over a set of nine analytical runs (see Table 2. All test samples were prepared in bulk by filtration under positive nitrogen pressure through acid-washed cellulose acetate filters (0.45 μm, 47 mm (Sartorius, Watford, UK)).
Units | River water A | River water B | Estuarine sample C | Seawater sample D | Drinking water E | |
---|---|---|---|---|---|---|
a nd = not determined. na = not applicable. | ||||||
Calcium | mg l−1 | 2.5 | 96 | 189 | 360 | 110 |
Sodium | mg l−1 | nd | nd | 3489 | 10157 | 20 |
Chloride | mg l−1 | nd | nd | 6268 | 18247 | 13 |
Magnesium | mg l−1 | 2 | 4.4 | 415 | 1200 | 20 |
Potassium | mg l−1 | nd | nd | 122 | 351 | 3 |
DOC | mg l−1 | 12.1 | 5.1 | 4.2 | <1 | <1 |
pH | 4.3 | 8.1 | 8 | 8 | 7.7 | |
Electrical conductivity | μS cm −1 | 142 | 573 | na | na | 650 |
Salinity | ppt | na | na | 12 | 35 | na |
Fig. 1 Typical calibration graph showing response in milli absorbance units (mAU) versus concentration of CrVI in μg l−1. The equation of the quadratic curve fitted to the points is y = 36.6x2 + 437x−1.6, where y is mAU and x is [CrVI] in μg l−1. Error bars are 95% confidence limits based on 4 replicate measurements at each concentration. |
Sample (+ spike concentration μg l−1) | Mean value | Within run sa | Between run sb | Total sc | RSD (%) | DFd | Recovery (%) | Recovery confidence limits +/−e |
---|---|---|---|---|---|---|---|---|
a Within run s = within run standard deviation estimated as a pooled value with 9 degrees of freedom. b Between run s = between run standard deviation estimated with 8 degrees of freedom. c Total s = total standard deviation estimated as the combination of (a) and (b) with degrees of freedom indicated (see ref. 14, Cheeseman et al., 1989). d Degrees of freedom associated with the estimated total standard deviation (see note c above). e 95% confidence limits associated with the observed recovery. | ||||||||
Sample A | 0.064 | 0.020 | 0.024 | 0.031 | 49 | 12 | ||
Sample A + 0.5 μg l−1 | 0.50 | 0.025 | 0.018 | 0.031 | 6 | 15 | 87 | 4 |
Sample A + 2 μg l−1 | 2.03 | 0.061 | 0.000 | 0.061 | 3 | 17 | 98 | 1 |
Sample B | 0.15 | 0.013 | 0.012 | 0.017 | 11 | 13 | ||
Sample B + 0.5μg l−1 | 0.66 | 0.018 | 0.030 | 0.035 | 5 | 10 | 101 | 5 |
Sample B + 2 μg l−1 | 2.10 | 0.047 | 0.065 | 0.080 | 4 | 11 | 97 | 3 |
Sample C | 0.17 | 0.013 | 0.017 | 0.021 | 12 | 11 | ||
Sample C + 0.5 μg l−1 | 0.75 | 0.042 | 0.048 | 0.064 | 8 | 12 | 115 | 8 |
Sample C + 2 μg l−1 | 2.17 | 0.038 | 0.081 | 0.089 | 4 | 10 | 100 | 3 |
Sample D | 0.16 | 0.012 | 0.025 | 0.027 | 17 | 10 | ||
Sample D + 0.5 μg l−1 | 0.72 | 0.029 | 0.053 | 0.060 | 8 | 10 | 112 | 8 |
Sample E + 2 μg l−1 | 2.23 | 0.063 | 0.135 | 0.149 | 7 | 10 | 104 | 5 |
Sample E | 0.086 | 0.005 | 0.008 | 0.010 | 12 | 11 | ||
Sample E + 0.5 μg l−1 | 0.59 | 0.023 | 0.019 | 0.030 | 5 | 14 | 101 | 4 |
Sample E + 2 μg l−1 | 2.06 | 0.042 | 0.091 | 0.100 | 5 | 10 | 99 | 3 |
Standard solutions | ||||||||
0.1 μg l−1 | 0.10 | 0.016 | 12 | |||||
0.5 μg l−1 | 0.51 | 0.019 | 12 | |||||
1 μg l−1 | 1.00 | 0.022 | 12 | |||||
2 μg l−1 | 2.01 | 0.010 | 12 | |||||
Limit of detection = 0.024 μg l−1 |
Spiking the water samples at 0.5 and 2 μg l−1 made it possible to assess potential matrix interferences by comparison of the calibration slope for the natural water samples with that obtained for the standard solutions. Fig. 2 shows the ratio of calibration slope in the sample matrix versus that in deionised water standard solutions. For the river samples B and C and the saline samples, the ratio is highly consistent and not significantly different from 1.0. For the more highly coloured water (A), there is a marked suppressive interference which is more variable from run to run. This is probably attributable to coextraction of humic material which was clearly visible as both colour and solid precipitate in the alcohol layer. Consequently, data for Sample A have been reported after standard additions calibration. Data for the other samples were calculated by direct comparison with aqueous standards.
Fig. 2 Ratio of calibration slope in the sample matrix versus that in deionised water standard solutions. Error bars are 95% confidence limits on the mean slope for 9 analytical runs. |
To assess the likelihood of losses of the determinand by adsorption during filtration, a river water sample was prefiltered, spiked with CrVI and analysed with and without filtration. The concentration in the filtered sample was found to be 1.33 ± 0.1 (p = 0.05)for the unfiltered sample and 1.34 ± 0.04 for the filtered sample. This indicates minimal loss of determinand by adsorption during filtration. The stabilty of CrVI in water samples has been demonstrated as at least one month by Sirinawin and Westerlund,16 so there does not appear to be a need to extract samples immediately.
This journal is © The Royal Society of Chemistry 2002 |