Introduction

Chemical vapour generation (CVG) by aqueous tetrahydroborate(III) derivatization has been employed for the trace element determination and speciation of many elements of the periodic table. These include As, Sb, Bi, Ge, Sn, Pb, Hg, Cd, Zn, transition and noble metals, e.g. Cu, Ag, Au, Ni, Pt, Pd, Ir, Rh and Ru [1, 2]. The mechanism of CVG has been the focus of many recent investigations and discussions [3–7]. The results reported reveal that tetrahydroborate(III) (THB) derivatization proceeds through complex intermediates formed by the interaction between the analyte containing substrate and the hydroboron species, which are formed in the reaction media by hydrolysis of THB [5, 6]. The same reaction scheme is considered to be valid for both classical hydride-forming elements and for transition and noble metals [6, 7]. In the case of the hydride-forming elements, it is possible to clarify many mechanistic details owing to the stability of the final products, i.e., AsH3, SbH3, BiH3, GeH4 and SnH4. For example, it is now understood that hydrogen is directly transferred from boron to the analyte atom [6]. In the case of transition and noble metals, the investigations are limited by the instability of the volatile derivates, the nature of which still remain to be identified [7].

CVG based on the simple acid–THB reaction system is dramatically influenced by the composition of the reaction media, which can modify the reactivity of the analyte with THB. In general, CVG of classical hydride-forming elements is negatively affected by the presence of some transition and noble metals that may produce severe interferences [1]. In order to avoid interference effects, the use of additives is largely employed in an effort to modify the reaction conditions [1, 3]. Numerous inorganic and organic acids as well as ligand/donor species combined with different reaction conditions (pH, amount of THB, CVG apparatus design) have been adopted in order to achieve better control of interferences and/or to perform selective generation of volatile species from a particular oxidation state or chemical species of the analyte [1, 3].

The mechanism of action of foreign species, either already present in the sample matrix or purposely added to it, is not well known, with the exception of the mechanism of interference by some transition and noble metals [1]. The mechanism of action of the ligand/donor species in the THB–analyte reaction system has received little attention despite numerous positive effects and useful applications in CVG [3]. A good example of this is the case of L-cysteine (Cys), which has been recognized to modify the THB–analyte reaction system through the formation of both THB–Cys (RS–BH3 −, thioloboranes) [8] and analyte–Cys complexes (for example As(SR)3) [9].

In the present study, the generation of bismuthane in the presence of several additives which exibit an ability to improve the generation efficiency of BiH3 at elevated Bi(III) concentration [10] is reported. The effect of these additives is to dramatically improve the linear dynamic ranges of analytical calibration graphs through the elimination of the formation of elemental Bi, which acts as an interferent similar to that previously reported for hydrogen telluride generation [11]. The studies were conducted under nonanalytical conditions, wherein the use of high Bi(III) concentrations (0.2–10 μg ml−1) permits the occurrence of chemical reactions and reaction pathways to be observed which otherwise remain hidden under ideal analytical conditions, i.e. trace amounts of analyte and high reductant-to-analyte ratios. In the present case, the most recent findings on the mechanisms of CVG have been combined with various diagnostic tools to generate new evidence for the role of additives in CVG.

Experimental

GC–MS measurements

Screw-cap reaction vials (borosilicate glass) fitted with PTFE/silicone septa (12 ml, Pierce Chemical Co., Rockford, IL, USA) were used for CVG in combination with GC–MS measurements. To generate bismuthane, 2 ml of acidified sample were placed in the vial and 1 ml of 0.2 M NaBH4 was added after purging the headspace of the vial with 2 ml min−1 N2 for 30 s. A gas-tight Hamilton syringe (5 ml) was employed for sampling headspace gases from the reaction vial headspace; sampled gases (2–3 ml) were injected directly into the GC–MS.

A Hewlett-Packard (Palo Alto, CA, USA) 6890 GC operated in the splitless mode and equipped with a Hewlett-Packard 5973 mass-selective detector was fitted with a 30 m × 0.25 mm i.d. capillary GC column (1 μm Valcobond VB-1). The GC was operated under the following conditions: injector temperature: 150 °C; oven temperature program: 35 °C, hold for 10 min, heated to 150 °C at 30 °C min−1; transfer line temperature 150 °C. The carrier gas was He flowing at a rate of 1.2 ml min−1.

In the case of derivatization with NaBD4, the same reaction conditions were employed and mass spectra were deconvoluted according to a recently developed procedure [12, 13] in order to obtain the relative abundances of BiHnD3–n isotopologues.

CVG–AAS apparatus

A continuous flow (CF) hydride generator was coupled to an atomic absorption spectrometric detection system (Model 503 atomic absorption spectrophotometer, Perkin-Elmer, Wellesley, MA, USA; CF–CVG–AAS). The atomizer was an Ar-H2 mini-flame supported on a 6.5 mm i.d. quartz tube [14].

Atomic absorption experiments were performed with a Bi electrodeless discharge lamp (Perkin Elmer EDL System II) source operated at the manufacturer’s recommended current. Absorbance measurements for bismuth were performed at the less sensitive atomic absorption line at 306.8 nm using a spectral bandpass of 0.7 nm.

The CF–CVG system was operated with a peristaltic pump (Pump head MS/CA4–12 from Ismatec, Glattbrugg, Switzerland, fitted on a Masterflex L drive H-7519–25 from Cole-Parmer Instrument Company, Vernon Hills, IL, USA). Ismatec Tygon microtubing of appropriate diameters were used to propel NaBH4 and waste solutions. The flow rates for the sample, NaBH4 and additive solutions were 4, 2 and 4 ml min−1, respectively.

The chemifold configurations employed to realize the different reagent mixing sequences are reported in Fig. 1 and are described below.

Fig. 1
figure 1

Schemes for chemifold set-up for continuous-flow HG of bismuth. Chemifold 1 (classical analytical set-up) L2=0, 15, 100, 500 μl; L3=500 μl. Chemifold 2 (for different mixing sequences of reagents) L1=0, 4, 50, 500 μl; L2=100, μl; L3=500 μl. See Table 1 for experimental parameters

Full size image

Chemifold 1 (see Fig. 1) was a typical analytical set-up: the acidified sample and the NaBH4 solution were merged at the first T-junction and allowed to react in L2 (15, 50, 500 μl loop volume, Teflon PFA, 0.5–0.8 mm i.d.); the outgoing reaction mixture was merged with argon stripping gas at the second T-junction, and the stripping of gaseous products continued in L3 (500 μl stripping loop, Teflon PFA, 0.5–0.8 mm i.d.); the mixture was then directed to the gas–liquid separator (borosilicate glass, 60 mm long, 10 mm i.d.), where the liquid was delivered to waste and the gaseous portion merged with the H2 gas flow, which was then directed to the atomizer. Ar and H2 flow rates were 150 ml min−1 and 175 ml min−1, respectively. In order to realize chemifold 1 with L2=0 μl (see Fig. 1), the sample, NaBH4 and Ar were simultaneously mixed in a X-junction.

Chemifold 2 (see Fig. 1) was employed to investigate the effects of different reagent mixing sequences and reaction times. Reagents 1 and 2 were first mixed in the T-junction and allowed to react in a mixing loop L1 (0, 4, 50, 500 μl; Teflon PFA, 0.5–0.8 mm i.d) followed by the delayed addition of Reagent 3. In the case where L1=0, the three reagents were simultaneously mixed in an X-junction (see Fig. 1) and the resulting reaction mixture was directly delivered to L2. After L2, the fate of the reaction mixture was the same as for Chemifold 1. Three different mixing sequences, namely A, B and C, could be realized, as detailed in Table 1.

Table 1 Mixing sequences employing Chemifold 2 and CF–CVG–AAS detectiona
Full size table

All mixing T-junctions and X-junctions were fabricated from Ismatec (Kel-F, 0.8 mm i.d.).

Chemicals

Solutions of NaBH4 (0.2 or 0.02 M) were prepared from pellets (Alfa Aesar, Word Hill, MA, USA); those of (0.2 M) NaBD4 were prepared from the powder (99% D, Cambridge Isotope Laboratories, MA, USA). The materials were dissolved in H2O and stabilized with the addition of 0.1 M NaOH.

Standard solution of 1000 μg ml−1 of Bi(III) was prepared from bismuth metal and dissolved in an appropriate concentration of HCl or HNO3. Working solutions for CVG using both a continuous flow reactor and septum-sealed vials were prepared by serial dilution of the stock.

All other reagents were of analytical grade.

Results and discussion

GC–MS experiments

The enhancement in efficiency of generation of bismuthane at high Bi(III) concentration (up to 10 μg ml−1) caused by the presence of additives such as KMnO4, K3[Fe(CN)6], Fe(III) and Mo(VI) at mM concentrations has recently been observed during attempts to increase the concentration of the hydride in the headspace of vials for GC–MS experiments dedicated to the study of H–D exchange [10].

In the absence of additives, the prevailing reaction appears to be the formation of elemental bismuth, as inferred by the appearance of an easily visible, finely dispersed black precipitate. In the presence of additives, formation of the black precipitate was not observed. Because no influence of the additive was found at lower analyte concentrations, calibration graphs for CVG of Bi obtained by GC–MS were linear up to 0.6 μg ml−1 (batch CVG in the reaction vial) and they were characterized by serious curvature and rollover above this concentration threshold. In continuing this work, it has been found that thiocyanate (as KSCN) was also able to produce a similar effect.

For some additives, it was evident that a reaction occurred upon addition of THB or TDB to the Bi(III)–additive acid solution. Permanganate was rapidly decoloured, while Mo(VI) solution changed from colourless to pale yellow, then colourless again or to light brown, depending on the Mo(VI) concentration.

CVG experiments were performed using NaBD4 (TDB) combined with GC–MS and mass spectral deconvolution in order to determine the relative abundance of the BiHnD3–n isotopologues and the total amount fraction of deuterium incorporated into the final hydride. In both the presence and absence of the abovementioned five additives (5 × 10−4–10−2 M), the reaction products were >96% BiD3 with only 1–2% BiHD2, while the amount fraction of deuterium incorporated was >98%. This indicates that the strong, positive perturbation produced by additives on the THB(TDB)–Bi(III) reaction system does not affect one of the basic features of hydride generation by THB derivatization: the direct transfer of hydrogen from boron to the analyte atom [6].

CF–CVG–AAS: calibration graphs for Bi

The shapes of the calibration graphs were investigated in the range 0.1–10 μg ml−1 Bi(III) in both nitric and hydrochloric acid media (see Fig. 2). Curvature and rollover are observed in both acid media, but they are more pronounced in nitric acid. Curvature and rollover increase with the residence time of the generated bismuthane in the reaction mixture: severe effects can be observed starting from residence times of about 0.5 s (L2=100 μl) and they increase dramatically at residence times of about 5 s (L2=500 μl). This is in agreement with the effect observed in batch CVG experiments (GC–MS detection), where the higher residence time of bismuthane in the solution limited the linear range of response to less than 0.6 μg ml−1 Bi(III). The concentration of reductant was also important: for example, in the case of the 1.0M HCl reaction medium (L2=100 μl), the linearity and the rollover point were detected at 0.5 and 0.1 μg ml−1 Bi(III) for 0.02M THB, and at 1.0 and 2.0 μg ml−1 Bi(III) for 0.2M THB. In all of the experiments a concentration of 0.2 M THB was used.

Fig. 2
figure 2

Calibration curves for Bi obtained by CF–CVG–AAS in nitric and hydrochloric acid reaction media using chemifold 1 (see Fig. 1) and 0.2 M NaBH4 with different reaction loop volumes L2=0, 15, 100 and 500 μl. Reported uncertainty is ±SD, n = 3 replicates

Full size image

The shape of the signal obtained by CF–CVG–AAS is progressively distorted with increasing concentrations of Bi(III) beyond the linear range: above the rollover concentration a double-peak shape is obtained. Formation of double peaks in the rollover region of calibration curves has been previously observed and investigated for the generation of H2Te [11] and during the atomization of AsH3 and H2Se [15]. The formation of double peaks is correlated with an interference process caused by a heterogeneous-phase reaction in which the gaseous hydride is captured on the surface of finely dispersed particles. The particles can originate from either the reduction of concomitant species or that of the analyte itself. In the case of H2Te generation, the production of elemental tellurium at high Te(IV) concentrations served as an interferent by capturing H2Te [11].

Effect of additives on the CF–CVG–AAS response from Bi

The effect of additive concentration was investigated beforehand at 5.0 μg ml−1 Bi(III) using Chemifold 1 with L2 = 100 (Fig. 1). The absorbance signal from Bi was enhanced 1.6–1.8-fold in 1M HCl and 3.6–4.1-fold in 1M HNO3, except for KSCN in HCl media where only a 10% enhancement was observed. The additive concentration did not significantly affect the signal enhancement factor over the following concentration ranges: 5 × 10−4 to 2 × 10−3 M for KMnO4 and K3[Fe(CN)6 , 2 × 10−3 to 8 × 10−3 M for Fe(III) and 10−3 to 10−2 M for KSCN, both in 1M HNO3 and 1M HCl. For Mo(VI), no significant effect was observed in 1M HCl in the range of 5 × 10−4 to 2 × 10−3 M and in 1M HNO3 in the range of 5 × 10−4 to 1 × 10−3M, while 15% signal depression was observed for 2 × 10−3 M Mo(VI) in 1M HNO3. The presence of additives (10−3M KMnO4, 10−3M K3[Fe(CN)6], 4 × 10−3 M Fe(III) and 10−3 M Mo(VI), both in HCl and HNO3) completely eliminates rollover in the calibration curves for bismuth, while hardly affecting the slope of the linear region. The only partial exception occurs for KSCN (2.5 × 10−3 M), which was a minor effect in HCl media. Examples relevant to KMnO4 and KSCN are reported in Fig. 3. A summary of the characteristics of the calibration curves for bismuth obtained in the presence and absence of additives is reported in Table 2.

Fig. 3
figure 3

Calibration curves for Bi obtained by CF–CVG–AAS in the presence of 10−3 M KMnO4 and in the presence of 2.5 × 10−3 M KSCN. 0.2 M NaBH4; reaction loop volume L2=100 μl. Reported uncertainty is ±SD, n = 3 replicates

Full size image
Table 2 Effect of additives on calibration curves for bismuth
Full size table

Mechanism of action of additives

The effect of the mixing sequence of the reagents, Bi(III), THB and additives was investigated using CF–CVG–AAS and Chemifold 2 (see Fig. 1). Reaction conditions for the three mixing sequences, A, B and C, are reported in Table 1. Comparison of the absorbance signal, S, obtained with a given mixing sequence, to that of the absorbance signal, S 0, obtained with the same analytical configuration (Chemifold 1, Fig. 1) employed in the study of the calibration curves (Figs. 2 and 3 and Table 2) was undertaken. The ratio S/S 0 indicates the extent to which a given mixing sequence is able to reproduce the enhancement effect obtained with the analytical configuration. Results are summarized in Fig. 4.

Fig. 4
figure 4

Effect of mixing sequence and reagent reaction time on the generation of BiH3 obtained using CF–CVG–AAS and chemifold 2 (Fig. 1) with reaction loop L1=4, 50, 500 μl. Curve A, delayed addition of Bi(III) to THB–additive mixture. Curve B, delayed addition of THB to additive–Bi(III) mixture. Curve C, delayed addition of additive to THB–Bi(III) mixture. Reported uncertainty is ±SD, n = 3 replicates

Full size image

At first sight, Fig. 4 indicates that the five additives generally behave in different ways in HCl and in HNO3 reaction media. This is in agreement with the much more pronounced ligand properties of chloride with respect to those of nitrate ions, which implies that chlorides can exert some influence over the formation of both Bi(III) and borane/hydroboron complexes. Many other differences become evident from an analysis of the different mixing sequences.

Mixing sequence A

Additive and THB are allowed to react in reaction loop L1 and the addition of Bi(III) is delayed by 0.04, 0.5 and 5 s. During this time, several reactions can occur: (i) acid-catalyzed hydrolysis of THB; (ii) additive-catalyzed hydrolysis of THB; (iii) redox reaction between the additive and THB, and (iv) formation of borane/hydroboron–additive complexes. The results indicate that, even after 5 s reaction time, the THB–Fe(III) and THB–[Fe(CN)6]3−additive reaction solution retains its full ability to produce the same enhancement effect observed with the analytical configuration. The opposite behaviour is seen with the \({\text{THB}} - {\text{MnO}}^{ - }_{{\text{4}}} \) system wherein the reactive species producing the signal enhancement seems to disappear after prolonged reaction between THB and permanganate ion. By considering the reported kinetics for permanganate [16] (\({\text{d}}{{\left[ {{\text{MnO}}^{ - }_{{\text{4}}} } \right]}} \mathord{\left/ {\vphantom {{{\left[ {{\text{MnO}}^{ - }_{{\text{4}}} } \right]}} {{\text{d}}t}}} \right. \kern-\nulldelimiterspace} {{\text{d}}t} = - k{\left[ {{\text{MnO}}^{ - }_{{\text{4}}} } \right]}{\left[ {{\text{BH}}^{ - }_{{\text{4}}} } \right]},<$> <$>\) k = 2 × 103 l mol−1 s−1 at 20 °C, pH-independent) and for hexacyanoferrate(III) (d[(Fe(CN)6)3−]/dt=−k[THB][H+], hexacyanoferrate(III)-independent, k = 5 × 106 l mol−1 s−1 at 25 °C) [17], it can be deduced that both additives are quickly reduced in reaction loop L1. Just after 0.04 s reaction time, the concentrations of these additives in their higher oxidation state should be several orders of magnitude below the Bi(III) concentration. This precludes any mechanisms based on the oxidative action of the two additives with Bi(III) in reaction loop L2 (see Fig. 1). Most probably, the decrease in signal observed for permanganate is due to formation of interfering species in reaction loop L1 (see Fig. 1).

The same considerations are valid for Fe(III) (in both HCl and HNO3 media) and Mo(VI) in nitric acid media. In particular, in nitric acid media S/S 0 tends to increase with reaction time between THB and the additive, and this also precludes any involvement of the Fe(III) and Mo(VI) oxidation states in a subsequent interaction with Bi(III) in reaction loop L2. With respect to thiocyanate, its interaction with THB can only produce complexes with borane/hydroboron species which are formed during hydrolysis. This appears to be very effective in HCl media, where the enhancement effect is even higher than the one obtained with the analytical configuration (S/S 0 >1).

Mixing sequence B

Additive and Bi(III) are allowed to react in reaction loop L1 and the addition of THB is delayed by 0.04, 0.5 and 5 s. This mixing sequence should provide information about the interaction between Bi(III) and the additive, principally complex formation, considering the high oxidation potential required to oxidize Bi(III) to Bi(V). With the exception of only thiocyanate, all of the additives gave the same response as the analytical configuration (S/S 0 was almost equal to unity in most cases). This means that, in general, the additive should be present in the reaction solution before the addition of THB (as in the case of the analytical configuration), but the possible role played by additive–Bi(III) complexes in the enhancement of the signal is not revealed. Indeed, curves obtained using configuration B could be explained on the basis of several hypotheses: the complexes play a role in signal enhancement but are formed more rapidly than the shortest reaction time used (0.04 s), or they do not play any significant role, or they are not formed at all. In the case of thiocyanate, formation of a Bi(III)–thiocyanate complex seems to play a role in signal enhancement, and the controlled reaction time between Bi(III) and thiocyanate results in superior signal enhancement than that obtained using the analytical configuration.

Mixing sequence C

Bi(III) and THB are allowed to react in reaction loop L1 and the addition of the additive is delayed by 0.04, 0.5 and 5 s. This mixing sequence should provide information on the ability of the additive to generate a signal enhancement after hydride generation has begun. This occurred only for Fe(III) in HNO3 media and hexacyanoferrate (III) in HCl media. In both cases, the enhancement was observed after the Bi(III) and THB solution had interacted for the longest reaction time tested. In the case of hexacyanoferrate(III) in HCl media, the signal enhancement is even greater than that obtained using the analytical configuration. This effect, which is difficult to explain, is similar to those observed during the generation of hydride-forming elements using milder amine boranes (tert-Bu amine– and dimethylamine–borane), where hydride evolution is slow. The addition of mM amounts of Fe(III) produced immediate formation, although not quantitative, of the hydride (D’Ulivo A, Mester Z, Sturgeon RE, unpublished results). It seems, therefore, that Fe(III) compounds possess the ability to catalyze the formation of some hydrides by acting on reaction intermediates.

Mechanism of formation of BiH3 and Bi0

The hypothesis that additives may act via the reoxidation of Bi0 to Bi(III) seems unlikely in light of the above discussion of the results of different mixing sequences. Furthermore, the oxidation pattern is out of consideration for the case of thiocyanate. This fact has interesting implications for the mechanism of formation of both bismuthane and elemental bismuth.

Two possible reaction pathways are reported in Fig. 5. In the first case (Fig. 5a), elemental bismuth is formed by decomposition of the hydride. This fact is in agreement with recently reported evidence [10] indicating that the thermally unstable bismuthane disappears much faster from the gas phase when it is in contact with the aqueous phase. Capture in the liquid phase is irreversible, resulting in decomposition of BiH3 upon interaction with the solvent to form nonvolatile Bi species. In the second case (Fig. 5b), both BiH3 and elemental bismuth are formed via the same reaction, i.e. through formation of reaction intermediates which can evolve towards the generation of BiH3 or Bi0, depending on the analyte/THB ratio. If the formation of BiH3 and Bi0 took place according to the reaction scheme outlined in Fig. 5a, the only possibility of avoiding formation of Bi0 is the reoxidation of Bi0 to Bi(III) by the additive. This reaction scheme, while compatible with the observations for permanganate, cannot explain the elimination of formation of Bi0 under nonoxidizing conditions by additives such as hexacyanoferrate(III) and thiocyanate. In general, the observed effects of additives are in agreement with the mechanism of formation of BiH3 and Bi0 illustrated in Fig. 5b. The reaction pathway in Fig. 5b represents a more general reaction model which is able to explain the behaviour of all of the additives (oxidant and nonoxidant).

Fig. 5a,b
figure 5

Reaction schemes for the formation of BiH3 and Bi0 by aqueous phase derivatization of Bi(III) with tetrahydroborate(III)

Full size image

Conclusions

The generation of bismuthane in the presence of several additives represents a particular reaction system in which the analyte and the interferent are the same element, but the element is in two different oxidation states, Bi(III) and Bi0. The additives enhancing the formation of bismuthane and decreasing the formation of Bi0 at higher Bi/THB ratios serve simultaneously as both reaction enhancers and masking agents. In the case of bismuth, the mechanism of action of additives is intimately related to the mechanism of hydride formation. Additives play a role in the formation of bismuthane reaction intermediates during derivatization of Bi(III) by THB in aqueous solution. They are able to drive the reaction towards the selective formation of the hydride, avoiding the formation of other by-products such as elemental bismuth.

The results of the present work are likely to be of more general relevance considering that the additives investigated have found applications in the CVG of many hydride-forming elements, as reported in the following examples. Fe(III) reduces the interferences generated by some transition metals in CVG of both H2Se and H2Te [18, 19]; [(Fe(CN)6)]3− and permanganate enhances the generation efficiency of plumbane [1, 20]; thiocyanate improves both the generation efficiency of H2Se and the control of interference [21], and Mo(VI) enhances the generation of AsH3 under particular reaction conditions [22]. In the case of liquid phase interferences among the hydride-forming elements, the formation of elemental bismuth was found to be the source of severe interference in the generation of SbH3 from Sb(III) [23] and of H2Se from Se(IV) [24]. In the case of selenium, the formation of elemental bismuth was the source of severe memory interference effects [24]. The use of the additives studied in the present work could have a beneficial effect on the control of these interferences.

The use of CVG apparatus based on different mixing sequences and controlled reaction times of the reagents represents a simple approach which is useful for diagnostic purposes and for providing a better understanding of the mechanisms operating in CVG, including the formation of volatile species, mechanisms of interference and the role played by the additives in the control of hydride generation efficiency and the removal of interferences. In addition to such diagnostic information, there are practical analytical advantages of such an apparatus that can be used to improve both the generation efficiency of volatile species and the control of interferences.