Damion K.
Corrigan
a,
Justin P.
Elliott
a,
Ewen O.
Blair
b,
Simon J.
Reeves
a,
Ilka
Schmüser
b,
Anthony J.
Walton
b and
Andrew R.
Mount
*a
aEaStCHEM, School of Chemistry, The University of Edinburgh, King's Buildings, Edinburgh, EH9 3FJ, UK. E-mail: a.mount@ed.ac.uk
bSMC, School of Engineering, The University of Edinburgh, King's Buildings, Edinburgh, EH9 3FF, UK
First published on 11th February 2016
Microelectrodes have a number of advantages over macroelectrodes for quantitative electroanalysis and monitoring, including reduced iR drop, a high signal-to-noise ratio and reduced sensitivity to convection. Their use in molten salts has been generally precluded by the combined materials challenges of stresses associated with thermal cycling and physical and corrosive chemical degradation at the relatively high temperatures involved. We have shown that microfabrication, employing high precision photolithographic patterning in combination with the controlled deposition of materials, can be used to successfully address these challenges. The resulting molten salt compatible microelectrodes (MSMs) enable prolonged quantitative microelectrode measurements in molten salts (MSs). This paper reports the fabrication of novel MSM disc electrodes, chosen because they have an established ambient analytical response. It includes a detailed set of electrochemical characterisation studies which demonstrate both their enhanced capability over macroelectrodes and over commercial glass pulled microelectrodes, and their ability to extract quantitative electroanalytical information from MS systems. MSM measurements are then used to demonstrate their potential for shedding new light on the fundamental properties of, and processes in, MSs, such as mass transport, charge transfer reaction rates and the selective plating/stripping and alloying reactions of liquid Bi and other metals; this will underpin the development of enhanced MS industrial processes, including pyrochemical spent nuclear fuel reprocessing.
As a result of its industrial importance, particularly in the nuclear sector, there has been a desire to develop analysis techniques suitable for MS systems, which address the challenges of withstanding the high temperatures and corrosive nature of the melts when they contain reactive species.9–11 Electrochemical methods have typically been practically limited to macroelectrode measurements12–20 and therefore suffer from a low signal-to-noise ratio, a time-dependent signal, iR distortion of the response and sensitivity to convection.21 These effects at best complicate and often preclude quantitative analysis. Optical spectroscopies suffer from the disadvantages of the broad emissions of many of the target analytes, difficulties in identifying multiple analytes and corrosion of sampling windows.10
When considering electroanalysis, microelectrodes can be fabricated using a number of methods and have well-established advantages under ambient conditions of high signal-to-noise ratios, and steady-state (time-invariant) and quantitative signals which are relatively insensitive to convection when compared to macroelectrodes.22,23 They have therefore been developed for ambient and near-ambient sensing applications24 and for use in the monitoring of industrial processes.25 However, there has been a significant technological barrier to the development of similar microelectrode sensors for reproducible and quantitative MS analysis and monitoring. For example, microelectrodes fabricated by drawing a wire encapsulated in glass and then exposing the end, have been employed with limited success, as this has been shown to result in initial and time-dependent variation of the response.26 The initial variability can be attributed to differences in the wire size and cross section due to variation in the wire drawing process, which as a result requires optical and/or electrochemical characterization of electrode size and shape. The time-dependent variation in the MS can be attributed to one or more of: the tendency for chemical attack of the glass, which compromises the electrode insulation; the differential thermal expansion of the microelectrode materials, which causes materials stress and cracking; and the chemical corrosion of the electrode, which can cause it to progressively recess with time. All of these effects cause a progressive change in effective electrode area and the diffusion regime, which precludes the effective deployment of microelectrodes in MS media for physicochemical analysis.
Previously, we have reported the development and characterisation of an MSM designed to overcome the above issues.21,27–29 In this study we have built on our previous work, employing our established photolithographic fabrication techniques to produce novel disc-shaped MSMs with precisely controlled dimensions and geometries. Disc microelectrodes have been chosen due to their well-established and quantitative ambient analytical response.29 We first establish whether the microelectrode response of the disc MSMs under ambient aqueous conditions is similar to other microelectrodes and then compare and contrast the performance of these devices in the LKE MS21 both with macroelectrodes and with the ambient microelectrode response. This paper has three main aims: (1) to clearly demonstrate the enhanced capabilities of the MSM in MS by benchmarking against a traditional glass pulled microelectrode, (2) to quantitate and explain the analytical response and extract physicochemical parameters in order to gain fundamental insight into the behaviour of the LKE molten salt system and (3) to investigate the potential for gaining fundamental insight into industrially relevant molten salt processes through MSM measurements. In this work we have concentrated on the characterisation of mass transport and charge transport in redox processes, along with the plating/stripping of metals and the formation of metal alloys. These are essential steps in the development of molten salt industrial processes, including pyrochemical spent nuclear fuel reprocessing, which we are developing through UK (the EPSRC REFINE programme) and EU (SACSESS programme) activities.
All experiments in 2.2 and 2.3 were performed using a PC-controlled Autolab PGSTAT12 potentiostat (EcoChemie, Utrecht, Netherlands) and results were analysed using GPES 4.9 and NOVA 1.11 software (EcoChemie), with iterative fitting carried out using Origin 2015 (OriginLab). EIS experiments were typically carried out by applying a 10 mV rms amplitude sinusoid superimposed on the dc potential, EDC, using logarithmically spaced frequencies between 100 kHz and 0.1 Hz.
The quantitative response of cone microelectrodes has been established as:30
iL = 4nFDcr(1 + qHp) | (1) |
However, over the course of ten minutes in LKE these devices were observed to progressively physically deteriorate, with the etching and removal of the insulating glass seen to be the most prominent failure mechanism (see Fig. 2B–D). Thus, the exposed electrode area clearly enlarged with time. Given this, it is not surprising that the cyclic voltammograms (CVs) observed for Zn2+ reduction in LKE showed a change in response with scan rate and time (Fig. 2G). The observed increasing capacitance, and increasing and non-steady-state currents, together indicates the absence of characteristic microelectrode behaviour and an increasing exposed electrode area attributable to progressive insulator compromise. In contrast an r = 10 μm Pt MSM employed for the same measurement showed no appreciable visible deterioration and produced a stable and characteristic microelectrode response, with the zinc reduction plating currents observed at E < −0.5 V being invariant with time or sweep rate (Fig. 2H). The resulting diffusion coefficient for Zn2+ in LKE of 2.85 × 10−5 cm2 s−1, obtained from the mass transport limited current for this reduction, and the established analytical response of a disc microelectrode:
iL = 4nFDcr | (2) |
In chloride melts, Eu(III) has been shown to react according to the following scheme:32
2EuCl3 ⇌ 2EuCl2 + Cl2 |
It is clear that both oxidation and reduction waves can be seen on the CV, centred around E = +0.4 V, which arise from Eu(II) and Eu(III) species respectively, which we attribute to the effects of the above reactions in LKE (actually, the proportion of Eu(III) and Eu(II) is likely to vary between experiments with the concentration of dissolved chlorine in, and/or the overall oxidising environment of, the LKE). Reassuringly, the CV has a wave-like form characteristic of a microelectrode response, along with the characteristic microelectrode sweep rate independence (Fig. 3B) (there is only a small additional capacitive current whose importance increases as expected with sweep rate, leading to a slightly increased separation between forward and backward scans). Previous work32 has shown the diffusion coefficients for Eu(III) and Eu(II) to be comparable. Assuming therefore a global diffusion coefficient, D, and taking iL as the difference in oxidation and reduction mass transport limited current and c = 19 mM, it is possible to calculate this global diffusion coefficient from eqn (2) as (5.0 ± 0.2) × 10−6 cm2 s−1 in LKE at 703 K; for any CV, the combined heights of the oxidation and reduction waves (Fig. 2A and B) or the steady-state chronoamperometric currents observed at long times when carrying out both mass transport limited oxidation and reduction (e.g.Fig. 2C) through eqn (1) can then be used to determine the concentrations of Eu(III) and Eu(II). Fig. 3D shows an Arrhenius plot of ln(D/T) vs. 1/T derived from the global diffusion coefficient data obtained using the MSMs at various temperatures (note that this plot assumes the validity of the Stokes–Einstein equation33, where k is the Boltzmann constant, rh is the effective radius of the diffusing species and η is the viscosity of the MS, which is assumed to vary with temperature according to , where η0 is a constant and Ea is the activation energy). The value of Ea obtained from the slope was (43 ± 5) kJ mol−1 which shows good agreement (within error) with that reported previously (44 kJ mol−1),12 establishing the principle of disc MSM LKE activation energy measurement.
In fact, it should also be possible to extract values for D and c from analysing the shape of the chronoamperometric transients for disc MSMs. If one avoids fitting the currents obtained in the first few tens of milliseconds, where double layer charging dominates the current, Xiong et al.34 have shown satisfactory fits to the equation developed by Shoup and Szabo:35
(3) |
Fig. 4A shows a comparable chronoamperometric transient recorded in aqueous media, when using an r = 25 μm Pt MSM to study 5 mM ferricyanide in 100 mM aqueous KCl. Consistent with Xiong et al., it is clear that a good fit of eqn (3) to the data is observed if one limits the fit to the data recorded between 0.05 s and 1 s. The adjusted R2 value was found to be 0.9988 and physically realistic values for ferrocyanide of c = 5.6 ± 0.1 mM and D = (4.2 ± 0.1) × 10−6 cm2 s−1 were obtained. Also included in the plot is a fit to eqn (4), the established equation for the transient current at a hemispherical electrode.34
(4) |
Fig. 4 Current–time transient and fits to eqn (3) and (4) for an r = 25 μm MSM Pt disc in (A) 100 mM aqueous KCl with 5 mM ferricyanide (E jumped from +0.34 to 0.14 V at t = 0) and (B) LKE with 19 mM added EuCl3 (E jumped from +0.5 to +0.2 V at t = 0). |
Fig. 4B shows the typical corresponding fits to the current–time transient recorded with EuCl3 in LKE. In this case the fit to the Shoup–Szabo eqn (3) was again reasonable, with D = (4.4 ± 0.3) × 10−6 cm2 s−1, which is consistent with the global value of D, c = 9.5 ± 0.4 mM and adj. R2 = 0.9694. The fit to eqn (4) again produced comparable parameters with c = 8.7 ± 0.5 mM and D = (4.3 ± 0.3) × 10−6 cm2 s−1 and was only marginally worse with adj. R2 = 0.9655.
Fig. 5A shows the typical EIS response for a r = 25 μm Pt disc MSM to an aqueous 5 mM potassium ferricyanide solution at 293 K. It is reassuring that the response fits well (with χ2 = 0.002) to the established modified Randles' equivalent circuit for the disc microelectrode36–38 (Fig. 5C), confirming its applicability to ambient aqueous measurements with disc MSMs as with other disc microelectrodes.
Fig. 5B shows the corresponding response of an MSM to the Eu(III/II) couple at its formal potential in LKE solution at 703 K. For these and all other EIS LKE disc MSM data, the best fit was achieved by replacing the Warburg element with a constant phase element (CPE), which is required to fit the gradient of greater than 45° observed at higher frequencies in the Nyquist plot (see inset to Fig. 5B). The fitted values for these data were: CDL = 1.66 ± 0.02 nF; RCT = 1590 ± 50 kΩ; Q0 = (1.12 ± 0.03) × 10−7 S sn; n = 0.695 ± 0.005; RNL = 2.42 ± 0.04 MΩ, with χ2 = 0.03. It is interesting that the equivalent data recorded in the same LKE solution using a bare tungsten macrowire electrode also requires the use of a constant phase element in place of the Warburg element in the Randles' circuit, and that the fitted value of n = 0.69 ± 0.03 is the same within experimental error as that observed for the MSM in LKE. This suggests a fundamental difference in the linear diffusional properties of ambient aqueous and LKE MS solution at these frequencies. Cuervo-Reyes et al.39 have recently interpreted a diffusional CPE as being due to subdiffusive transport and local trapping of charge carriers. It is tempting to infer that some of the strong electrostatic coupling and local mobile ion trapping postulated in this work in batteries and mobile glasses may also be present to some extent in LKE MS, and that this may also be the origin of the observed CPE.
Turning to the other measured EIS parameters, we have previously shown for square ambient aqueous microelectrodes that at the formal potential of a one-electron redox couple the minimum values of the two resistances RCT and RNL should be given by:29
(5) |
and
(6) |
(7) |
and
(8) |
It is clear from Fig. 6B that there is good agreement between the disc MSM data and eqn (7), and the fitted value of RminNL = 2.42 ± 0.02 MΩ also agrees well with the value of 2.41 ± 0.03 MΩ calculated from the data in Fig. 3A and B using eqn (5). This indicates that the established EIS MSM theory (eqn (5)–(8)) can be used to extract quantitative information from disc MSM measurements in LKE MS. Using the observed value of RminCT = 13.1 ± 0.2 kΩ a value of k0 = 0.26 cm s−1 can be extracted from eqn (6).29 This value is two orders of magnitude faster than that previously reported in the literature for this system (3.31 × 10−3 cm s−1 on a glassy carbon macroelectrode).40 This is consistent with previous ambient aqueous microelectrode measurements, which have demonstrated that reduction of the dimensions of the electrode to the micron scale allows for markedly more efficient mass transport and the measurement of more realistic charge transfer rates.41 Given this, it is interesting that the RCT data (Fig. 6A) shows the predicted minimum at EDC = E0, but that these data do not fit well to eqn (8). Although the errors on these data are relatively large, this may suggest some difference between charge transfer behaviour in LKE MS and in ambient aqueous solution.
Fig. 7A shows typical CVs of the behaviour of the bismuth(III/0) couple on Pt and W disc MSM electrodes. When cycling the W MSM electrodes, a simple plating wave (for Bi(III) reduction and deposition) and essentially single stripping peak (for Bi(0) oxidation and dissolution) were observed, centred near E = −0.06 V. This is consistent with the phase diagram, which indicates that W and Bi do not form alloys. Using eqn (1) and the observed mass transport limited reduction current, the diffusion coefficient for Bi3+ was determined to be (6.7 ± 0.02) × 10−6 cm2 s−1 at 698 K. In contrast, the Pt MSMs showed Bi(III)/(0) plating and associated stripping electrochemistry, centred slightly above 0 V, with three identifiable, overlapping stripping peaks. It is satisfactory that there is very close agreement for the values of iL found for the Pt and W MSM electrodes (a 2.1% variation in iL, which is within the error for the devices), which demonstrates that these differences are not due to changes in n or A, but rather to the effects of Pt/Bi alloying.
The Pt–Bi phase diagram44 indicates two thermodynamically stable solid alloy phases under these conditions (PtBi2 and PtBi), with a third state of intermediate composition (Pt2Bi3) which is thermodynamically stable above 843 K. We tentatively propose these three stripping peaks originate from stable PtBi2, metastable Pt2Bi3 and stable PtBi phases respectively; further quantitative experiments are underway to establish whether this is the case. Fig. 7B also shows that when a Pt disc MSM electrode is swept over a range of scan rates a fourth stripping peak is increasingly observed at the most cathodic potential at the slowest scan rates. This we attribute to stripping of the stable liquid metal alloy (predominantly Bi) phase expected from the phase diagram at high Bi concentrations. Additionally, Fig. 7C and D show the effect of progressively increasing the negative potential limit of the CV scan (which also increases the deposition time). In this case all peaks increase in size, consistent with increased Bi deposition and alloy formation, and unlike in Fig. 7B, the peaks shift progressively to more positive potentials, indicating the contribution of stripping kinetics at these increased coverages. That significant Bi is deposited was confirmed by visual inspection following prolonged plating (Fig. 8C and D). The maximum thickness of liquid Bi plated across the r = 25 μm Pt disc surface, at the slowest sweep rate, was calculated to be ∼9 nm. It is interesting that the reduction to bismuth metal takes place at a higher potential on the MSM than on a Pt wire macroelectrode (Fig. 8A), and distinct stripping peaks were not observed when using this macrowire, even though the calculated amount of Bi deposited per unit area was similar on each electrode. This indicates greater richness in Bi alloy characterisation information when using MSMs.
As with Eu(III)/(II) above, the values of iL observed for Bi(III) plating on the disc MSMs were used in eqn (1) to obtain Bi(III) diffusion coefficients, from which an Arrhenius plot was constructed (Fig. 8B). The resulting value of Ea = 39 ± 4 kJ mol−1 was in close agreement with that found for Eu.
Bismuth binary and ternary alloys that do not involve Pt (of potential importance when carrying out coextraction using Bi electrodes) can also be studied by using W disc MSMs. Fig. 9 shows typical CVs for a mixed Bi(III), Zn(II) system in LKE, which shows Bi(III)/(0) plating/stripping (near E = 0 V), Zn(II)/Zn(0) plating stripping (near E = −0.9 V) and Li(I)/(0) plating/stripping (at and below E = −1.8 V). The distinct peaks observed in Li(I)/(0) plating/stripping and Zn(0) stripping features are again attributed to alloy formation, and further indicate the richness of alloying information provided by these disc MSMs.
In addition to these established similarities to aqueous disc MSM responses, which demonstrate the potential of disc MSMs in enhanced MS sensing, analysis and process development, there are also some interesting differences. Whilst a minimum is observed at E0 for RCT in EIS, the shape of the variation with EDC is apparently not as predicted by ambient aqueous theory. There are hints from variation in the EIS equivalent circuit that there may be a contribution from subdiffusive transport in the LKE MS, as seen previously in glass melts and Li-ion battery materials. There is also the marked enhancement of characterisation information for metal alloy formation from LKE MS compared to macroelectrodes. These measurements provide real motivation for additional study to gain further physicochemical insight into the properties of LKE and other MS systems.
This journal is © The Royal Society of Chemistry 2016 |