Simultaneous determination of arsenic, selenium, tin and mercury by non-dispersive atomic fluorescence spectrometry

Talmra, Vol. 32, No. 2, pp. 103-109, 1985 Printed in Great Britain. All rights reserved

Copyright 0

0039-9140/85 $3.00 + 0.00 1985 Pergamon Press Ltd

SIMULTANEOUS DETERMINATION OF ARSENIC, SELENIUM, TIN AND MERCURY BY NONDISPERSIVE ATOMIC FLUORESCENCE SPECTROMETRY* A. D’ULIVO, R. Fuoco and P. PAPOFF Istituto di Chimica Analitica Strumentale de1 C.N.R., Via Risorgimento, 35, 56100 Pisa, Italy (Received 12 March 1984. Revised 29 August 1984. Accepted 14 September 1984) Summary-A

procedure is described for simultaneous determination of arsenic, selenium, tin and mercury in aqueous solution by non-dispersive atomic-fluorescence spectrometry. Radiofrequency-excited EDLs, 100% modulated in the kHz region, were used for atom excitation. Sodium tetrahydroborate was used as reductant and a hydrogen-argon miniflame as atomizer. In the optimized procedure, which uses 1 ml of sample, the limits of detection (three times the standard deviation of the blank) were 0.04, 0.08, 0.1 and 0.1 ng/ml for arsenic, selenium, tin and mercury respectively. The linear dynamic range was greater than three decades’ for all analytes and the precision was better than 7% (typically 3%) for concentrations 2 1 ng/ml. Results for mutual interference effects are reported. Copper, nickel, lead and cobalt interfered only with selenium (5 ng/ml), when present in at least 200-fold weight ratio to it. Using 5 ml of sample improved the limits of detection for selenium and arsenic (0.01 and 0.02 ng/ml respectively), but at the expense of greater interference. Recovery from spiked natural water samples was better than 95% at the ng/ml level, except for selenium in sea-water, when the recovery was only 85%. Determination of the four elements, including standard-addition and background measurements, requires about 10 min.

Non-dispersive atomic-fluorescence spectrometry (NDAFS) has proved to be superior to dispersive atomic-fluorescence and atomic-absorption spectrometry on account of its potential for simultaneous multielement analysis and its lower detection limit and larger dynamic range.lm5 In a previous paper,6 describing a new instrument for simultaneous determination of elements by NDAFS, attention was mainly focused on the characteristics of the optical and electronic components and optimization of the signal to noise ratio (SNR). The present paper describes use of that instrument for simultaneous determination of arsenic, selenium, tin and mercury at sub-ng/ml levels, with sodium tetrahydroborate as reductant for generation of the volatile species measured. EXPERIMENTAL

Apparatus

Radiofrequency-excited electrodeless discharge lamps (RFEDL) (Perkin-Elmer) were used in the four-channel excitation system.6 Individual modulation of the RF power of each channel in the range 6-8 kHz allowed approximately 100% modulation of the light-beam, with a low (d.c.) background level. An argon-hydrogen entrained air miniflame’ was preferred because of its optical transparency and low noise as I

.

*Presented in part at the “ lo Convegno Incontri di Chimica Analitica dell’Ambiente”, Genoa, Italy, 23-24 May 1983.

well as its chemical inertia and the absence of memory effects (which often occur with electrothermal atomizers8). The flame was supported on a glass tube (4 mm bore) and obtained, unless otherwise specified, with flow-rates of 0.35 l./min for hydrogen and 1.O l./min for argon, the latter also serving as the carrier gas. Ball flowmeters (Sho-Rate 150, Brooks) were used to measure the flow-rates. The 70-ml reaction vessel was constructed as described by others.‘s9 Disposable micropipettes were used for adding the tetrahydroborate and for sample injection. Reagents

As(III), Se(N), Sn(IV) and Hg(I1) stock solutions (1000 pg/ml) were prepared from pure As,O,, SeO,, Sn and Hg. Working standards were prepared just before use. The sodium tetrahydroborate solution was stabilized with sodium hydroxide (0.1%) and filtered.‘O Erba RLE sodium tetrahydroborate (powder) was preferred because of its high purity and low cost. Merck “Suprapur” hydrochloric acid and sodium hydroxide were used. Procedure

One ml of 2% sodium tetrahydroborate solution was placed in the reaction vessel and a disposable micropipette containing the sample (0.2-1.0 ml) was placed in the injector port. Argon was then passed through the reaction vessel until stable base-lines were obtained. The sample was then injected and the signals were recorded. For a sample volume of 5 ml, the procedure was modified. The sample was placed in the reaction vessel and stirred (magnetic stirrer) with argon passing through the vessel, then 1 ml of sodium tetrahydroborate solution was added through the injector port. The connecting tubing and the glass burner were conditioned with 3M hydrochloric acid before measurements, whenever determination of selenium was required. 103

A. D’ULIVOet al. hydrogen miniflame is used, the enhancement of the signal caused at increasing acidity by the increased rate of hydrogen production and faster stripping of volatile species, is offset by the change in geometry and stoichiometry of the flame, which decreases the degree of atomic excitation. Anomalous behaviour was observed for selenium. If the glassware and tubing were not preconditioned, strong and reproducible signals were obtained only when the sample acid concentration was 1.2M hydrochloric acid. For instance, when successive l-ml samples in 0.6M hydrochloric acid medium were injected, a progressive decrease in signal peak height was observed (Fig. 2). We attribute this to the relatively high acidity of hydrogen selenide (K, = 2.2 x lo-‘), resulting in its reacting along the walls of the transfer line with basic impurities resulting from decomposition of the sodium tetrahydroborate and carried away by the transport gas.

1

2

Acidity

A

3

(Ml

Fig. 1. Effect of sample acidity (HCl) on the atomicfluorescence signals of arsenic, selenium, tin and mercury, each present at a concentration of 10 ng/ml. The arrows indicate the sample acidity proposed by Thompson and Thomerson? Conditions: l-ml sample, 1 ml of 2% NaBH, solution, observation height 15 mm, H, 0.35 I./mitt, argon 1.O l./min.

0

C D

E

RESULTS AND DISCUSSION E$ect of gas flow-rates

ii

!N

L

In agreement with the findings of Tsujii and Kuga,’ the hydrogen flow-rate was found to affect the SNR, depending on the acidity and volume of the solution injected into the reaction vessel. With 0.6M hydrochloric acid sample medium and different sample volumes, the best SNR was obtained with the following hydrogen flow-rates: 0.25 l./min for 0.2 and 0.5 ml samples, and 0.35 and 0.45 l./min for 1 and 5 ml samples respectively. An argon flow-rate of 1.0 + 0.2 l./min resulted in relatively large (7-9 set) reproducible peak widths at half peak-height. It was therefore possible to use a relatively high time-constant (10 set) at the output low-pass filter, with consequent gain in SNR. E#ect of acidity

Figure 1 shows, for each element, the effect of acidity on the signal peak heights. The arrows show the acid concentration giving maximal response, as obtained by Thompson and Thomerson9 by AAS with a quartz furnace atomizer, other conditions being the same. For the elements for which the comparison is possible, the highest signals are reached at lower acidities in our procedure. This can be attributed to the fact that when the argon-

Fig. 2. Effect of acid preconditioning on the signal from 1 ml of 2-ng/ml selenium solution. Signals A-E were obtained with non-preconditioned apparatus, and signals A,-E, with acid-preconditioned apparatus. Signals A and A, were obtained with 1.2M HCl sample medium. Signals B, C, D, E and B,, C, , D, , E, were obtained with 0.6M HCl medium, injected straight after A and A, respectively. Other conditions: 1 ml of 2% NaBH, solution, H, 0.35 I./mitt, argon 1.O l./min, observation height 15 mm.

Determination

Hg

t I

I

I

10

20

30

L

Burner

height

105

of As, Se, Sn and Hg by AFS

(mm)

Fig. 3. Effect of observation height on atomic-fluorescence signal of arsenic, selenium, tin and mercury, each present at IO-ng/ml concentration. Conditions: 1 ml of sample in 0.6M HCl, 1 ml of 2% NaBH, solution, H, 0.35 l./min, argon 1.0 l./min.

lation of C, in AFS, was practically devoid of meaning in this work, since the nature and size of the noise signal before and during the tetrahydroborate reaction were quite different. Table 1 shows that the C, values tended to decrease with increasing sample volume for all four elements, but with different patterns according to whether the impurities were mainly present in the acid or in the tetrahydroborate. In the case of selenium it is evident that the hydrochloric acid in the sample does not contribute to the size of C,, since the product VC, is fairly constant. The slope of the calibration graph would be expected to remain constant for constant yield of atomic events per unit volume of sample, i.e., m/V should be constant. This was found to be approximately the case for arsenic over the whole sample volume range and for selenium in the l-5 ml range, but for tin and mercury the yield decreased consistently with sample volume. Since the decrease in detection limit when the sample volume is increased from 1.0 to 5 ml is very significant only for selenium, the use of 1 ml of sample is suggested, especially as interferences are higher with 5 ml of sample than with 1 ml, as will appear later.

Preconditioning with 3M acid prevented this effect, and reproducible signals were obtained (Fig. 2) for several successive injections of sample. For multielement determination, a sample acidity of 0.6M hydrochloric acid provides maximal sensitivity for arsenic and selenium and only a l&20% loss in sensitivity for mercury and tin. Effect of observation

height

The optimum position for the optical path of the EDL light-beams is 15 mm above the burner top, which gives the highest sensitivity for arsenic and tin, and about 70 and 90% of maximum sensitivity for mercury and selenium respectively (Fig. 3). ESfect of amount of sodium tetrahydroborate Figure 4 shows that the highest fluorescence signals were obtained for arsenic, mercury and selenium with 1 ml of 2% sodium tetrahydroborate solution. The signal for tin was 90% of the maximal value (reached with 1 ml of 4% sodium tetrahydroborate solution). Effect of sample volume The influence of sample volume on some of the performance characteristics is summarized in Table 1. Changing the sample volume changes the pH reached by the solution after the addition of tetrahydroborate (the change being biggest for the smallest sample volume), and the procedure must be changed when a 5-ml sample is used. The blank signal, S,, is assumed to be due to the analyte present in the reagents. The detection limit C, is taken as three times the standard deviation of the concentration in the blank, C,. The peak-to-peak noise measured for the background near the signal, which is sometimes used for calcu-

in

d I

I

I

I

1

2

3

4

NaBH,

concentratw7

(%

w/v

1

Fig. 4. Effect of concentration of NaBH, solution on blank signal and on atomic-fluorescence signal of arsenic, selenium, tin and mercury, each present at IO-ng/ml concentration. Conditions: 1 ml of sample in 0.6M HCl, 1 ml of NaBH, solution, H, 0.35 l./min, argon 1.0 I./min, observation height 15 mm.

10.1 10.0 7.1 7.0

0.2 0.5 1.0 5.0

13 31 63 254

m

65 62 63 51

m]V

0.4 0.2 0.18 0.08

ca

RSD, C, ~-2.5 0.1, 22.2 0.06 28.1 0.04 11.0 0.02

-m/V 25 34 65 68

m 5 17 65 340

Selenium

0.5 0.3 0.1 0.03

C, 0.3, 0.2, 0.08 0.01

CL 6.7. 5.8 6.2 3.4

RSD, 19 37 49 60

fn 95 74 49 12

m/v

Tin

Table 1. Figures of merit: dependence on sample volume

0.9 0.6 0.6 1.0

.~c, 0.1, 0.1, 0.1, 0.1,

1.0 0.7 0.5 0.3

105 76 68 29

C,,

RSD, _---___ 7.4 2t 8.1 38 6.7 68 9.0 147

Mercury --m m/V C, 0.2, 0.1, 0.1, 0.08

-. CL

300 450 400 300

3.9 3.8 3.6 3.5

1.0 5.0 7.0 4.8 5.2

3.6 7.2 3.5 4.0

3.0

2.1 5.9 3.2 4.2

10.0

2.4 2.8 2.0 4.3

IO0

Precision at various levels, nsimtt --

___I

Conditions: 1 ml of sample in 0.6M HCI, I ml of 2% NaBH, solution, HZ 0.35 I./min, Ar 1.0 i./min, observation height 15 mm. *Calculated as log (CUpper,lm,t/CL). TReiative standard deviation on ten replicate measurements for each of the four levels.

Arsenic Selenium Tin Mercury

Analyte

Dynamic range ---l____-Upper limit nglml Decades*

Table 2. Dynamic range and precision in singte-element mode

Conditions: 1 ml of 2% NaBH, solution, observation height 15 mm, Ar I.0 l./min. For the different volumes of sample, see the relevant procedure in the text. RSD, = Relative standard deviation of blank signal (10 replicate measurements). m = Slope of calibration curves (arbitrary units x ml/rig). C, = Apparent concentration of blank signal (ng/ml). C, = Limit of detection (@ml), taken as three times the standard deviation of the blank.

RSD,

Sample volume, v, ml

Arsenic

Determination Table

3. Mutual

interelement

of As,

Se, Sn and Hg by AFS

107

effect on precision* Precision9

Analyte (1 ng/ml)

Precision?

Arsenic Selenium Tin Mercury

5.0 7.0 4.8 5.2

As

Se

Sn

Hg

8.2 6.1 7.3

5.1 6.2 4.8

4.1 5.8 6.2

5.0 5.8 5.5 -

*Conditions: as for Table 2. iRelative standard deviation of ten replicate in absence of interferent. §Relative standard deviation of ten replicate in presence of a 500 ng/ml concentration

measurements IO2

10

measurements of interferent.

Dynamic range and precision

Table 2 shows that the precision was between 2 and 7% for all four elements at the 1-ng/ml level and improved with concentration, except for mercury, for which it was almost constant. The precision was unaffected by the presence of the other three elements in 500-fold w/w ratio to the analyte (Table 3) though there were some effects on signal height (Table 4, see below).

lo3 Tin

2

30

S ‘‘0 c z

2010 0 -__P_____O__~_b__----_--__ -10

-

-20

-

00

E i

00

:

-40

-

interelement

10

0.05 pg/ml Analyte (5 ngiml)

As

Se

Arsenic? Arsenic5 Selenium? Seleniums Tint Tins Mercury? Mercurys

0 0 0 0 0 0

0 0+20 0 0 0 0

Sn 0

I

I

102

103

0 0 0 0

As

Se

0 0 0 0 O-20 O-20 -

0 0

0+10 0+45 0 0 0 0

0 0

*Expressed as per cent change with respect Variations less than twice the s.d. were not iSample volume 1 ml, 0.6M HCl. See text @ample volume 5 ml, 0.6M HCl. See text

( ng/ml

I lo4

1

concentration of the interferent and not on the concentration of analyte or the analyte/interferent concentration ratio (see Fig. 5). With 5 ml of sample, the effect of arsenic on the tin signal was the same as for 1 ml, but the effect of tin on the arsenic signal is evident at tin concentrations above 0.03 pg/ml. Mutual effects, not observed with 1 ml of sample, were found for the selenium-tin pair at interferent concentrations higher than 2 pg/ml and for the arsenic-selenium pair at interferent concentrations higher than 5 pgg/ml. Other interferences. Copper was considered as a typical example of an element which strongly interferes in the hydride generation of many elements.8*“J2 effect on signal peak height

in peak height,

0.5 pg/ml Hg

8

Fig. 5. (I). Interference effect of tin(IV) on arsenic signal for 1 ml of sample: As 25 ng/ml (0) and 250 ng/ml (0). (II). Interference effect of arsenic(II1) on tin signal for 1 ml of sample: Sn 25 ng/ml (0) and 250 ng/ml (0).

interference Change

%

(II)

Arsenic

Table 4. Mutual

z

-x)-

Interferences

In the preceding paper6 the instrumental cross-talk was measured separately from chemical and optical effects and was calculated to be at least 68 dB. Mutual interferences. Varying the sample volume but using a constant amount of sodium tetrahydroborate was found to affect the size of the interference effects, except for mercury (Table 4). Mercury is the only one of the four elements which does not form a hydride, so the interference probably arises in the hydride formation. With 1 ml of sample, mercury interferes with the selenium signal only when present at concentrations higher than 5 pg/ml. Arsenic and tin present a peculiar mutual effect: the arsenic peak is enhanced by tin concentrations > 0.5 pg/ml with maximal increase at tin > 2 pg/ml. Arsenic affects the tin response with a similar but negative effect. The mutual arsenic-tin effect depends mainly on the

lo4

(ng/ml)

%*

2.0 pg/ml

Sn

Hg

As

0 0 0 0

0 0 0 0 O-20 O-20 -

0 0

0 0

to signal for each analyte taken into account. for procedure. for procedure.

Se

Sn

0+30 -10 +45 0 0 0 0 0 0 0 0 in pure solution,

5.0 pg/ml Hg

As

0 0 0 0 0 -15 O-20 o-25-15 0 0

Se

Sn

Hg

0+30 -15 +35 0 -15 0 0 0 0 0

0 0

for four interferent

+15 -10

levels.

0 0 -

A.

108

D’ULIVO

II)

-80

I 1

10

Interference

0 -20 -

I

102 concsntmtim

lo3 I *g/ml J

‘,.

-40 -Eo -

Fig. 6. (I). Interference effects on signal for selenium (5 ng/ml). Effect of copper(H) and tin(N) as observed in present work: (0) copper, 1ml of sample; (0) copper, 5 ml of sample; (A) tin, 1 ml of sample; (A) tin, 5 ml of sample.

Curves a and h refer to the effect of copper and tin on selenium (10 ng/ml> as observed by Welz and Melcher’ (curve a) and Verlinden and Deelstrai2 (curve 6). (II). Interference effect of selenium(IV) on signal for arsenic (5 ng/ml) as observed in the present work with 1 ml (0) or 5 ml (a) of sample, and as observed by Welz and Melcher’ (curve a).

With 1 ml of sample, each analyte at the 5-ng/ml level, and the copper concentration varied in the range 0.005-5 pgg/ml, no interference with the signals of arsenic, tin and mercury was found. With selenium, copper did not interfere up to 1 pg/ml (Fig. 6). With 5 ml of sample and the same concentration of selenium, copper interfered at concentrations down to 0.2 pig/ml. The effect of nickel, cobalt and lead (tested in the 0.005-S pg/ml range) on the selenium signal for 1 ml of sample was then studied. It was found that nickel and cobalt exhibited the same behaviour as copper, whereas lead did not interfere.

The interference effects could be compared with the literature only for arsenic and selenium, for which data obtained under similar experimental conditions were available.8,” The difference ih pattern from the literature data was essentially due to differences in the time elapsing between mixing of the reactants and stripping of the volatile hydrides from the cell, and to variation of the sample volume with constant tetrahydroborate volume. The interference threshold shifts towards higher interferent concentrations as the sample volume is decreased. It is the sample volume which is the more likely to play the decisive role. Decreasing the sample volume should increase the

et al.

Determination of As, Se, Sn and Hg by AFS rapidity of mixing the sample and reagent solutions and give a higher reagent concentration, thus reducing the risk of incomplete analyte hydride formation on account of side-reactions with interferents. In this respect it is interesting that no mutual interference effects between arsenic, selenium and tin were reported in a study of their automatic continuous-flow hydride-generation, even at very high interferent concentrations.‘3.‘4 Application

to water samples

The natural concentration levels’s2’ of mercury and tin in uncontaminated waters being lower than the detection limits, only selenium and arsenic can be determined in such samples without preconcentration. For potable or polluted waters the method can indicate whether the four elements are above or below the allowable limits. Depending on the analyte, the simultaneous determination may give speciation information, or only the total concentration. Under the experimental conditions used, inorganic arsenic and methylarsenic acids are reduced to the corresponding arsines, which give the same analytical response;‘7,22 inorganic mercury, methylmercury, ethylmercury and phenylmercury are all reduced to mercury metal;23 tin(H) and tin(IV) give the same response,24 but some alkyltins found in natural waters cannot be quantitatively determined, owing to the high boiling point.of the alkylstannanes generated by tetrahydroborate reduction.20,2’ Only selenium(IV) reacts with tetrahydroborate to give hydrogen selenide, and is practically the only form present in natural water. 25*26 If the presence of Se(VI) or Se(H) is suspected (as in polluted water) the sample must be chemically pretreated to transform all the selenium species into Se(IV).27*28Total selenium and tin must therefore be determined by single-mode procedures.2’,27.28With the four-channel detector, the determination of arsenic, mercury, selenium and tin (including standard-addition and reagent background correction) requires about 10 min. This significant shortening of operation time allows replicate analysis of the same sample or analysis of a larger number of samples, with a consistent improvement of the statistical meaning of the analytical data. The accuracy of the method was tested by analysis of spiked samples. At the l-2 ng/ml level the recovery was better than 95% except for selenium in sea-water, for which it was only 85% (Table 5). The relative standard deviation for five replicate measurements was better than 10%. CONCLUSIONS

Arsenic, selenium, tin and mercury can be determined simultaneously at sub-ng/ml levels by nondispersive atomic-fluorescence spectrometry com-

109

bined with the vapour-generation technique with sodium tetrahydroborate as reductant. Sample volumes from 0.2 to 5 ml can be used. The lowest detection limit is obtained with 5 ml of sample for selenium and arsenic and either 1 or 5 ml of sample for mercury and tin. Except for determination of selenium and arsenic at concentrations near the detection limit, 1 ml of sample is to be preferred because the interferences are then minimal. The relative standard deviation is then less than 7% for each of the four analytes within their dynamic ranges and down to 1 ng/ml, regardless of the presence of high concentrations of the other three analytes. The linear dynamic ranges cover more than three orders of magnitude for all four analytes. Acknowledgements-This work was supported (Rome). One of the authors (P.P.) acknowledges from the M.P.I.

by C.N.R. assistance

REFERENCES

1. 2.

P. D. Warr, Talanta, 1970, 17, 543. T. J. Vickers, P. J. Slevin, V. I. Muscat and L. T. Farias,

Anal. Chem., 1972, 44, 930. 3. T. S. West, Analyst, 1974, 99, 886. 4. R. F. Browner, ibid., 1974, 99, 617. 5. N. Omenetto and J. D. Winefordner, Prog. Anal. Atom. Spectrosc., 1979, 2, 1. 6. A. D’Ulivo, C. Festa and P. Papoff, Talanta, 1983, 30, 907. 7. K. Tsujii and K. Kuga, Anal. Chim. Acta, 1978, 97, 51. 8. B. Welz and M. Melcher, ibid., 1981, 131, 17. 9. K. C. Thompson and D. R. Thomerson. Analvst. 1974. , 99, 595. 10. J. R. Knetchel and J. L. Fraser, ibid., 1978, 103, 104. 11. M. Verlinden and H. Deelstra. Z. Anal. Chem.. 1979.

296, 253. 12. A. E. Smith, Analyst, 1975, 100, 300. 13. F. D. Pierce and H. R. Brown, Anal. Chem., 1977, 49, 1417. 14. F. D. Pierce, T. C. Lamoreaux, H. R. Brown and R. S. Fraser, Appl. Spectrosc., 1976, 30, 38. 15. M. Dall’Aglio, War. Anal. Bull., 1982, 7, 163. and R. Van Grieken, Talanta, 1982,29, 16. H. Robberecht 823. 17. R. S. Braman, D. L. Johnson, C. C. Foreback, J. M. Ammons and J. L. Bricker, Anal. Chem., 1977,49,621. 18. A. Seritti, A. Petrosino, E. Morelli, R. Ferrara and C. Barghigiani, Environ. Technol. Lett., 1982, 3, 251. 19. J. Olafsson, Mar. Chem., 1982, 11, 129. 20. R. S. Braman and M. A. Thompkins, Anal. Chem., 1979, 51, 12. 21. V. F. Hodge, S. L. Seidel and E. D. Goldberg, ibid., 1979, 51, 1256. 22. J. S. Edmonds and K. A. Francesconi, ibid., 1976, 48, 2019. 23. J. Toffaletti and J. Savory, ibid., 1975, 13, 2091. 24. K. Tsujii and K. Kuga, Anal. Chim. Acta, 1978, 101, 199. 25. Y. K. Chau and J. P. Riley, ibid., 1965, 33, 36. 26. F. Leutwein, Handbook of Geochemistry, K. H. Wedephol (ed.), p. 34-H, 2,3. Springer, Berlin, 1972. 27. J. M. Rankin, Environ. Sci. Technol., 1973, 7, 823. 28. G. A. Cutter, Anal. Chim. Acta, 1978, 98, 59.