Determination of Co(II) by chemiluminescence after in situ electrochemical pre-separation on a flow-through mercury film electrode

Abstract

A novel method of electrochemical pre-separation of Co(II) before detection by chemiluminescence is reported together with the associated instrumentation. The Co(II) ions were selectively pre-separated on a mercury film electrode (MFE) by on-line reduction, then the accumulated metal was oxidised and selectively stripped back into the flowing solution as Co(II). These secondary ions were quantified as a result of their catalytic activity on the chemiluminescent reaction between luminol and hydrogen peroxide that was also induced on-line. The whole sequence was carried out in an automated flow-through system, in which the electrochemical pre-separation of metals was performed in either continuous flow or flow injection analysis (FIA) regimes. The scope of the method, both in terms of selectivity and sensitivity, has been demonstrated and the quantitative determination of Co(II) by the proposed method has been investigated. For a period of continuous flow pre-separation of 4 min, the calibration curve for Co(II) was linear up to a concentration of 100 μg l⁻¹, the relative standard deviation was 4% at the 20 μg l⁻¹ level and the limit of detection was 0.5 μg l⁻¹ (at the 3σ level). The method was applied to the determination of the cobalt content in a high purity iron sample.

---

Introduction

Chemiluminescence (CL) is an extremely sensitive analytical technique used for the determination of both organic and inorganic species. The high sensitivity of CL makes it a particularly attractive choice for the quantification of trace metals in various samples.^1–3 The utility of the method is based on the catalytic (or inhibitory) effect of many metal ions on a suitable CL reaction. The most widely exploited, but by no means the only, such reaction is the oxidation of luminol by hydrogen peroxide in an alkaline solution. Within a certain concentration range, the catalytic CL intensity is proportional to the concentration of the metal ion in the sample. However, CL is not a selective method since several metal ions catalyse (or inhibit) most chemiluminescent reactions to varying degrees. As a consequence of this drawback, an on-line separation step is required prior to the detection stage. Traditionally, various methods have been employed to achieve this separation including selective complexation,⁴ liquid–liquid extraction, membrane separations and, by far the commonest approach, ion exchange columns.³ Recently, a method has been reported that made use of an electrochemical pre-separation stage on a gold electrode for the determination of traces of Cu(II) by chemiluminescence.^5,6

In this paper, we report a novel method that makes use of electrochemical pre-separation on a flow-through electrochemical (EC) cell equipped with a MFE in order to achieve the required selectivity and sensitivity for trace analysis. The MFE combines the relative advantages of the solid and mercury drop electrodes and is ideal for flow-through applications.⁷ The principle of the method was that the target metal ion was pre-separated on-line by reduction on the MFE. Following this separation stage, the metal was oxidised and stripped back again into the flowing carrier as a cation to be subsequently detected by CL in an optical cell. The required instrumentation is described and the instrumental and chemical parameters to be considered are discussed. In particular, the selectivity was maximised by optimisation of the pre-separation and stripping potentials while the sensitivity could be altered by varying the pre-separation mode and the accumulation time. The method was accessed and validated for the detection of trace Co(II) in a reference material.

Experimental

Chemicals

All the chemicals were of analytical grade and ultrapure water (conductivity > 18 MΩ⁻¹ cm⁻¹) was used for the preparation of the solutions. A carrier carbonate buffer solution pH 10.2 was prepared daily by mixing the appropriate amounts of NaHCO₃ and NaOH. A stock solution of luminol (10 mmol l⁻¹ ) was prepared every week and kept in the dark; a 1 mmol l⁻¹ solution was prepared daily and kept in a dark bottle during the experiments. A 10 mmol l⁻¹ hydrogen peroxide solution was also prepared daily from 30% w/v hydrogen peroxide. The mercury plating solution was 1 mmol l⁻¹ Hg(II) in 0.1 mol l⁻¹ KNO₃–0.01 mol l⁻¹ HNO₃. Standard Co(II) standard solutions were prepared daily from a 1000 mg l⁻¹ atomic absorption standard solution by serial dilution. A solution of 1000 mg l⁻¹ in Cu(II) was prepared from CuSO₄ and more dilute solutions were prepared by serial dilution.

Equipment

The automated flow system is a modified version of the automated flow system described previously,⁷ in this case optimised for CL. Both FIA and continuous flow pre-separation could be carried out depending on the positioning of the EC flow cell. The experimental configuration is illustrated in Fig. 1. The valves, configured either for mixing, M, or switching, I, were micro-solenoid devices purchased from the Lee Company (Westbrook, CT, USA). The following parameters were software adjustable: flow rates of the solutions, number of sample injections, injection time, and delay time. For the FIA operation, the volume of the sample loop was 100 μl. The electrochemical cell (ECC in Fig. 1) was a thin layer cell, designed in-house with a flow channel thickness of 0.2 mm. The working electrode was a glassy carbon rod (3 mm in diameter) (Ringsdorf Carbon, Germany), the reference electrode was a home-made Ag/AgCl electrode positioned opposite the working electrode and the counter electrode was a glassy carbon rod positioned downstream, near the outlet of the cell. The home-made optical flow cell, positioned downstream of the EC cell, consisted of a rectangular channel of zig-zag configuration, 10 cm in length, machined in a block of Perspex. The potentiostat was a PARC 273 (EG&G, NJ, USA) controlled by a 386 PC through the PARC Model 270 Electrochemical Software. The optical detector was a Hamamatsu HC120 integrated photomultiplier (PMT)-amplifier-power supply module operating at a voltage of 250 V. The optical flow cell was positioned in front of the PMT window in a light-tight box constructed in house. The output of the detector module (i.e. the CL response) was fed simultaneously to a BBC Coertz Metrawatt SE 120 chart recorder and to a LABPC+ data acquisition (DAQ) card (National Instruments, Austin, TX, USA) that was interfaced to a 486 PC and was controlled by a data acquisition program written in LabVIEW 4.0 (National Instruments). The response could be viewed on the screen of the computer and also saved on disk. A PARC Model 303 HMDE has been used for the voltammetric experiments with Co(II).

Fig. 1 Configuration of the flow system for the detection of Co(II) by CL after EC pre-separation in the FIA mode. M and I indicate mixing valves and injector valves, respectively. ECC represents the electrochemical thin layer cell. In the continuous flow mode, the loop is replaced by the ECC.

Experimental procedure

0.1 g of the iron sample was dissolved in 2 ml of 4 mol l⁻¹ M HNO₃, 0.2 g of potassium tartrate was added (in order to complex Fe(III)), the pH was adjusted to 10.2 with NaOH and the solution was diluted to 100 ml with water.

The glassy carbon surface was pre-treated as described previously.⁸ Then, the mercury film was plated at −1.0 V by passing the mercury solution through the cell at 0.6 ml min⁻¹ (continuous flow mode) for 60 s or by making 5 discrete injections of the mercury solution (FIA mode).

The potential was then switched to the pre-separation potential (usually −1.4 V) and the pre-separation of Co(II) was initiated by passing the sample through the cell under continuous flow at 0.6 ml min⁻¹ or by injecting the sample in the carrier (FIA mode). In the FIA configuration, the sampling time required for the sample to wash and fill the injection loop was 15 s. The accumulated cobalt was stripped off in the flowing carbonate buffer carrier as Co(II) by stepping to a more anodic stripping potential (ranging from −0.8 to −1.2 V), mixed with the combined luminol–hydrogen peroxide solution and detected by CL. Finally, the MFE was regenerated electrochemically in the carbonate buffer at a potential ranging from +0.2 to −0.8 V. The flow rates of the carrier solution and of the luminol–hydrogen peroxide mixture were 1.5 and 1.3 ml min⁻¹, respectively.

Results and discussion

Study of the electrochemical pre-separation and stripping of Co(II)

In principle, a combination of EC preconcentation and stripping prior to the actual detection process can enhance both the sensitivity and the selectivity of the analysis of metals. The sensitivity is improved as a result of the enrichment of the electrode surface with the metal under investigation. The metal is first reduced and deposited on the electrode surface under cathodic potentiostatic conditions (i.e. electrolysis) for a defined time period and, then, the accumulated metal is stripped off the electrode and detected. Stripping can be accomplished via either voltammetric techniques (involving a scan of the working electrode potential with respect to time in the anodic direction) or potential-step techniques (based on a step change of the potential of the working electrode to an anodic value). The former method of stripping is commonly exploited in anodic stripping voltammetry (ASV)⁹ in which the potential scan also serves as the detection scheme. Although this method offers adequate selectivity and sensitivity for some metals, it is hampered by drawbacks (such as capacitive currents, the need for deoxygenation, insensitivity to irreversible redox reactions etc.). On the other hand, with the potential-step techniques, the selectivity towards a particular metal can be improved by judiciously selecting the pre-separation and stripping potentials, as is diagramatically illustrated in Fig. 2. The anodic limit, E_al, and cathodic limit, E_cl, correspond to the potentials at which oxidation and reduction, respectively, of the supporting electrolyte occur and define the useful potential range within which EC pre-separation is feasible. As long as a pre-separation potential, E_p, is applied to the working electrode, metal ions with redox potential E ⩾ E_p will be reduced and accumulated on the electrode while metal ions with redox potential E ⩽ E_p will be unaffected. As soon as the potential is stepped to a more positive stripping potential, E_s, the accumulated metals with redox potentials, E_i, where E_s > E_i > E_p will be oxidised and stripped back into the solution while those metals with redox potentials E_i > E_s will remain on the electrode. Thus, from the potentially useful range (E_al, E_cl) only the metals in the narrow range (E_s, E_p) will be available for detection. In this mode of stripping, amperometric, spectroscopic or chemiluminescent detection can be employed. For the present work, CL detection was selected due to its simplicity and sensitivity. Dynamic EC techniques were initially used to study the pre-separation behaviour of Co(II) on a mercury electrode. The EC accumulation and stripping of Co(II) was investigated on a hanging mercury drop electrode (HMDE). ASV was initially investigated but this technique did not convey any useful information. Since the oxidation of the amalgamated cobalt in carbonate buffer is irreversible, reoxidation of the species did not occur and no peak appeared in the anodic scan. Instead, another useful, but hardly used, EC technique, electrochemical enrichment (EE), specifically devised for irreversible redox couples has been employed.¹⁰ This latter method makes use of three stages: preconcentration of the metal by reduction at a potential more negative than the half-wave potential under convective transport; stepping of the potential to a value more positive than the half-wave potential in static solution so that the metal is oxidised and the resulting ions accumulate at the electrode/solution interface; and finally, a rapid scan to the cathodic direction to reduce the ions generated in the previous stage. The reduction peak generated in this final stage is a measure of the amount of accumulated metal. The sequence of operations for the EE of Co(II) is illustrated in Fig. 3(A). A typical reduction peak for Co(II) after EE is shown in Fig. 3(B). It was found that, in a carbonate buffer with pH 10.2, the cobalt reduction peak occurred at −1.34 V (Fig. 3(B)). The relationship between the current and the Co(II) concentration was linear, indicating a promising pre-separation behaviour.

Fig. 2 Schematic diagram of the selectivity afforded by EC pre-separation. The double arrow defines the detection window.

Fig. 3 (A) The potential–time profile employed in the technique of EE of Co(II) and (B) A typical voltammetric response for 10 mg l⁻¹ Co(II) after EE on a HMDE. Potentials as in Fig. 3(A); supporting electrolyte: carbonate buffer pH 10.2; preconcentration time 120 s; stripping time 10 s; scan rate 100 mV s⁻¹.

Choice of electrode and stability of the MFE

In a flow system, a MFE is preferable to a HMDE as the working electrode due to its better mechanical stability, wide scope for various cell configurations and large surface area.⁷ Compared to bare metal electrodes (like the gold ones employed in previous work^5,6), MFEs are easier to clean, have a larger hydrogen overpotential (so that they can be safely used at more cathodic potentials) and promote the reversibility of EC reactions. EC regeneration and cleaning of the electrode from residual metals was conducted electrochemically in the flowing carrier solution by holding the potential of the electrode at an anodic value. An important advantage of the MFE as opposed to the metal electrode reported earlier^5,6 is that replacement of the electrode surface after prolonged use is easier and more efficient and can be carried out by simply stripping the mercury film at + 0.8 V and plating a new film in situ. An important aspect of this work is that metals ions with irreversible reduction kinetics (e.g. Co(II), Ni(II), Mn(II)), undetectable by direct ASV, can be quantified using the proposed method. The stability of the mercury film was satisfactory under the conditions prevailing during the experiments. The same mercury film could be used for 2–3 h without loss of activity provided that the potential of the electrode was controlled to values more cathodic than +0.2 V. Thus, the MFE was considered as the most appropriate electrode for the particular application.

Linearity, limit of detection and comparison between FIA and continuous flow

Fig. 4(A) illustrates typical CL response traces obtained for Co(II) standards in the range 100–700 μg l⁻¹ after FIA EC pre-separation (3 injections). For quantitative work, it was advantageous to use peak heights rather than peak areas, especially for continuous flow pre-separation where tailing effects are more pronounced, for reasons that will be discussed later. The calibration parameters for both FIA and continuous flow pre-separations are shown in Table 1. As expected, continuous flow pre-separation offered higher sensitivity than FIA pre-separation as already discussed in conjunction with stripping voltammetry.⁷ The sensitivity and linear range were, also, dependent on the accumulation time: for longer pre-separation times the sensitivity increased but the linear range was narrower. This property could be exploited to good effect in order to tailor the sensitivity for different samples in which the concentration of Co(II) is highly variable. The loss of linearity at higher concentrations may be due to a combination of electrode overloading (saturation) in the stage of metal accumulation or light quenching during the CL detection. The limit of detection could be further improved by extending the pre-separation period.

Fig. 4 (A) CL response for Co(II) after pre-separation on a MFE in the FIA mode (3 injections). Traces (a), (b), (c), (d) represent Co(II) concentrations of 100, 200, 400 and 700 μg l⁻¹, respectively. Potentials as in Fig. 3(A); (B) CL signal of a 400 μg l⁻¹ Co(II) solution as a function of: (i) the pre-separation potential (■) when the stripping potential was set at −0.8 V, and; (ii) the stripping potential (◆) when the pre-separation was carried out at −1.4 V. Pre-separation in the FIA mode (3 injections).

Table 1 Calibration parameters for the determination of Co(II) using different pre-separation schemes and times

Pre-separation mode	Upper limit of linear range/μg l^−1	Sensitivity/ mV L μg^−1	Limit of detection/ μg l^−1) (at the 3σ level)
FIA (3 injections)	1 × 10^3	7 × 10^−3	5
Continuous flow (2 min)	3 × 10^2	2 × 10^−2	1.5
Continuous flow (4 min)	1 × 10^2	5 × 10^−2	0.5

---

Selection of potentials and selectivity

The selection of the pre-separation and stripping potentials is of great significance for the successful application of this method. As mentioned earlier, the pre-separation potential must be set to a value more cathodic than the redox half-wave potential of the metal of interest. The variation of the CL signal as a function of the pre-separation potential is illustrated in Fig. 4(B). Although more cathodic potentials resulted in higher CL signals, in the case of Co(II), a pre-separation potential of −1.4 V was selected as a precaution to avoid hydrogen evolution which can occur at more negative potentials. The selection of the stripping potential was equally critical. The stripping potential should be set to a value more anodic than the redox half-wave potential of the metal of interest. Fig. 4(B) illustrates the variation of the peak height as a function of the stripping potential and suggests that the more cathodic the stripping potential the lower the sensitivity of the CL detection. The possible explanation of this effect is that a substantial anodic overpotential is required to activate irreversible reactions (in this case the oxidation of elemental cobalt). The more anodic the stripping potential, the higher the anodic overpotential and the more efficient the oxidation of cobalt. Nevertheless, excessively anodic stripping potentials might compromise the selectivity of the method since the detection window expands and other coexisting metals could strip off the electrode and interfere with the analysis. To demonstrate the effect of the stripping potential on the selectivity, a solution containing Co(II) and Cu(II) was studied with the pre-separation potential set to −1.4 V. The presence of Cu(II) did not affect the CL peak heights when the stripping was carried out at −0.8 V since at this potential only Co oxidised; however, when the stripping potential was set to +0.2 V, oxidation of the pre-concentrated Cu occurred and the CL peak height increased by more than 100% due to the contribution of the Cu(II) ions to the total response. In cases where several metal ions are present in the sample, the detection window should be as narrow as possible to avoid interference from those metals that possess redox potentials similar to that of cobalt. It was found that, in real samples, it was more appropriate to use a stripping potential of −1.2 V for Co(II) (i.e. marginally more anodic than the cobalt reduction peak illustrated in Fig. 3(B)). This narrow detection window of E_s = −1.2 V, E_p = −1.4 V, combined with the fact that, among metal ions, Co(II) has the strongest catalytic activity on the oxidation of luminol by hydrogen peroxide, endows the method with satisfactory selectivity. Under these conditions, the following metal ions did not interfere in the analysis of Co(II) at a 1000-fold excess over Co(II): Zn(II), Cu(II), Cd(II), Fe(III), Pb(II), Cr(VI), Sn(IV), Ti(IV), Mg(II), Mn(II). Ni(II) only interfered at concentrations over 200-fold higher than those of Co(II).

Dispersion effects

The choice of the pre-separation mode has two serious implications in terms of both sensitivity and shape of the response. A comparison of the final CL signals was made for 60 μg l⁻¹ Co(II) in the direct CL detection (i.e. without pre-separation), CL detection after continuous flow pre-separation and CL detection after FIA pre-separation. In particular, the ratio of peak heights was 23∶8∶1 and the ratio of peak widths at half peak heights was 1∶5∶2 for direct CL detection, CL detection after continuous flow pre-separation for 4 min and CL detection after FIA pre-separation. An interesting property of the detection is that the peak heights after pre-separation are lower and wider than in direct CL detection. This is due to two reasons. (i) The low coulometric efficiency of the EC cell employed in this work that results in a low preconcentration efficiency. A calculation based on a model derived previously in this laboratory showed that less than 10% of the amount of Co(II) flowing through the cell is reduced and accumulated on the MFE. Under optimised conditions of flow rates and EC cell geometry, a coulometric efficiency of 50% would be feasible. (ii) Dispersion effects which are more pronounced in the pre-separation modes than in direct CL detection. The physical presence of the EC cell causes an increase in dispersion. In addition, the EC stripping process contributes to the dispersion as, upon application of the stripping potential, oxidation of the accumulated Co is neither instantaneous nor quantitative. Indeed, some of the oxidised Co diffuses deep into the mercury bulk and needs some finite time to diffuse back through the mercury film and out into the solution as an ion. These ions will trail the main bulk of the stripped metal and cause tailing of the CL peak. On the other hand, the dispersion in the continuous mode is higher than in the FIA mode, as indicated by the ratio of the peak widths at half peak heights and also made apparent by the pronounced tailing of the peaks . This effect is due to the fact that the EC cell is further from the detector in the continuous configuration than in the FIA configuration (Fig. 1), meaning that the dispersion is higher in the continuous flow configuration.

Analytical application

The proposed method was applied to the analysis of a high purity iron sample (149/3 BCR) with a certified cobalt content of 0.0073 ± 0.0005%. Since the cobalt concentration in the sample was relatively high, the pre-separation was in the FIA mode. The cobalt content calculated using our method was 0.0078 ± 0.0006% (n = 6) at the 3σ level. The experimentally determined mean value is higher than the certified value and this is probably due to interference from the presence of tartrate ions added to complex iron. Experiments with artificial solutions indicated that, at concentration ratios similar to that in the sample, tartrate causes a 5% error in the determination of cobalt. Nevertheless, taking into consideration the complexity and the composition of the sample, there is satisfactory statistical agreement between the certified and determined content.

Acknowledgements

The financial contribution of the Engineering and Physical Sciences Research Council through a ROPA grant to AE is gratefully acknowledged.

References

1. U. Isacsson and G. Wettermark, Anal. Chim. Acta, 1974, 68, 339.
2. D. B. Paul, Talanta, 1978, 25, 377.
3. K. Robards and P. J. Worsfold, Anal. Chim. Acta, 1992, 266, 147.
4. Q. Lin, A. Guiraum and R. Escobar, Anal. Chim. Acta, 1993, 283, 379.
5. C. X. Zhang, S. C. Zhang and Z. J. Zhang, Analyst, 1998, 123, 1383.
6. W. Qin, Z. J. Zhang and H. J. Liu, Anal. Chem., 1998, 70, 3579.
7. A. Economou, P. R. Fielden and A. J. Packham, Analyst, 1994, 119, 279.
8. A. Economou and P. R. Fielden, Analyst, 1993, 118, 47.
9. J. Wang, Stripping Analysis, VCH, Deerfield Beach, FL, 1985..
10. V. Stara and M. Kopanica, Electroanalysis, 1993, 5, 595.

---

Footnote

† Present address: Laboratory of Analytical Chemistry, Department of Chemistry, Aristotle University of Thessaloniki, Thessaloniki 540 06, Greece.

---