ARTICLE pubs.acs.org/ac
Coulometric Detection of Irreversible Electrochemical Reactions Occurring at Pt Microelectrodes Used for Neural Stimulation Silke Musa,*,†,‡ Danielle R. Rand,† Carmen Bartic,†,‡ Wolfgang Eberle,† Bart Nuttin,§ and Gustaaf Borghs†,‡ † ‡
Imec, Leuven, Belgium Department of Physics and Astronomy and §Department of Neuroscience, KU Leuven, Belgium
bS Supporting Information ABSTRACT: The electrochemistry of 50 μm diameter Pt electrodes used for neural stimulation was studied in vitro by reciprocal derivative chronopotentiometry. This differential method provides well-defined electrochemical signatures of the various polarization phenomena that occur at Pt microelectrodes and are generally obscured in voltage transients. In combination with a novel in situ coulometric approach, irreversible H2 and O2 evolution, Pt dissolution and reduction of dissolved O2 were detected. Measurements were performed with biphasic, charge-balanced, cathodic-first and anodic-first current pulses at charge densities ranging from 0.07 to 1.41 mC/cm2 (real surface area) in phosphate buffered saline (PBS) with and without bovine serum albumin (BSA). The extent to which O2 reduction occurs under the different stimulation conditions was compared in O2-saturated and deoxygenated PBS. Adsorption of BSA inhibited Pt dissolution as well as Pt oxidation and oxide reduction by blocking reactive sites on the electrode surface. This inhibitory effect promoted the onset of irreversible H2 and O2 evolution, which occurred at lower charge densities than those in PBS. Reduction of dissolved O2 on Pt electrodes accounted for 1934% of the total injected charge in O2-saturated PBS, while a contribution of 0.412% was estimated for in vivo stimulation. These result may prove important for the interpretation of histological damage induced by neural stimulation and therefore help define safer operational limits.
D
efining and understanding the operational limits of electrical stimulation of neural tissue are regarded as key requirements for the further development of safe neuroprosthetic devices.14 For most in vivo applications, charge-balanced, biphasic current pulses are used to avoid DC current flow and thus ensure the physicochemical integrity of the electrode material and surrounding tissue.510 Nonetheless, irreversible electrochemical reactions such as water hydrolysis and accompanying local pH changes, metal dissolution, and O2 reduction can occur, causing damage to neural tissue and electrode material.1120 It is therefore important to determine the type and extent of irreversible processes contributing to charge injection during electrical neural stimulation. Electrode overvoltages measured in response to applied biphasic current pulses have been used to define safe stimulation ranges in vitro.16,2124 Stimulation was deemed safe when the electrode overvoltages remained within the limits of water hydrolysis. These limits were assumed to correspond to the overvoltages for H2 and O2 evolution obtained with cyclic voltammetry (CV). This assumption, however, does not take into account that CV sweep rates are generally less than 0.1 V/s, while the electrode potential during current stimulation can change at rates higher than 1 kV/s. Such high rates can be accompanied by appreciable overvoltages depending on the r 2011 American Chemical Society
nature and time constants of the underlying electrode processes. In fact, it was shown that overpotentials for gas evolution under current stimulation exceed those observed with CV.25 Hence, larger electrode overvoltages are tolerable with current stimulation before the onset of gas evolution. Another technique, the so-called pulseclamp technique introduced by Bonner and Mortimer,26 involves applying a monophasic cathodic current pulse to an electrode. At the end of the pulse, the electrode potential is clamped back to its prepulse value.17,2729 The charge flowing within the 20 ms after switching to voltage clamp is the reversibly recoverable charge; any further flow of charge is considered irreversible or lost. Unfortunately, with this method only monophasic current pulses can be studied. Among the various electrode materials, such as iridium oxide, titanium nitride, stainless steel, Au, tungsten, conducting polymers, and others, that are currently used for neural implants, Pt is by far the most studied and best documented. While Pt is biocompatible and inert under open-circuit conditions, it can undergo corrosion during stimulation, releasing trace quantities Received: November 23, 2010 Accepted: April 18, 2011 Published: May 05, 2011 4012
dx.doi.org/10.1021/ac103037u | Anal. Chem. 2011, 83, 4012–4022
Analytical Chemistry of dissolved material that may be toxic to the surrounding tissue. Trace analysis for Pt dissolution in response to prolonged stimulation has been carried out using spectrophotometry1215,30 and mass spectrometry.31 While most of these experiments were performed in inorganic physiological media, two studies investigated the effect of protein adsorption on Pt dissolution.30,31 One such study demonstrated that stimulation-induced Pt dissolution is reduced in the presence of albumin, an abundant protein in the blood serum and cerebrospinal fluid (CSF). This effect was attributed to the strong interaction between albumin and the Pt surface, a process that is competitively inhibiting Cl adsorption onto the electrode. Supporting results for this corrosion inhibition were also found in vivo where the dissolution rate of Pt electrodes decreased with stimulation time.15 Because spectrophotometry is a destructive ex situ technique, it is not possible to correlate Pt dissolution with histological tissue damage. In order to allow such correlation, in situ analytical methods are needed to detect Pt dissolution during stimulation. While some attention has been devoted to Pt dissolution in the presence of organic material, there is a lack of similar information regarding H2 and O2 evolution. The irreversible reduction of dissolved O2 leading to reactive products has been considered another possible cause for tissue injury. An in vitro study using the pulseclamp method on Au wire electrodes concluded that O2 reduction accounts for up to 30% of the injected charge in O2 saturated buffered saline solution.17 Recent work by Cogan et al.20 using CV and voltage transient measurements under comparable conditions suggests a contribution of only 7% using Pt microelectrodes. The present investigation describes the use of CV and reciprocal derivative chronopotentiometry (RDC) to study Pt electrodes in vitro under various stimulation conditions in inorganic phosphate buffered saline (PBS) and in PBS supplemented with bovine serum albumin (PBS-BSA). In order to determine the extent of O2 reduction to the total injected charge, PBS was used either as-prepared or deoxygenated with N2 gas. Well-defined electrochemical signatures of the various interfacial processes occurring during pulsing were derived from the RDC curves. The influence of varying the pulse amplitude and pulse width as well as pulse polarity were investigated. In addition, a novel coulometric approach based on RDC is presented for the in situ detection, of charge losses due to irreversible gas evolution, Pt dissolution, and O2 reduction. The presented analytical approach offers the possibility to study individual electrode processes, which is not possible with the methods described earlier. It also provides practical guidelines to identify safe stimulation ranges under realistic stimulation conditions.
’ MATERIALS Reagents and Solutions. Phosphate buffered saline (0.150 M NaCl, 0.016 M Na2HPO4, 0.004 M KH2PO4) and PBS-BSA (1.5% w/v) were used in all experiments. The concentration of BSA was higher than that found in the CSF (0.02% w/v)32 and slightly lower than that found in blood serum (3.45.4% w/v)33 to account for the unavoidable contact of neural implants with blood during surgery. Comparison of in vivo and in vitro CV measurements using this BSA concentration suggest, however, that the local protein concentration at the electrode in vivo may be larger.34 The pH of all buffers was adjusted to 7.4 by small additions of KOH. The PBS was used either as-prepared, i.e. exposed to ambient O2 conditions, or deoxygenated by bubbling
ARTICLE
with N2 gas (PBSdeox) for at least 1 h prior to measurements. Bubbling was maintained throughout the experiments. All chemicals were analytical reagent grade and were used as delivered. Instrumentation and Software. All electrochemical measurements were conducted in a glass beaker using a three-electrode configuration placed inside a Faraday cage. A commercial double-junction Ag|AgCl reference electrode was used together with a large-area Pt counter electrode. Measurements were made using an Autolab PSTAT302N potentiostat with integrated frequency response analyzer controlled by the NOVA software (Ecochemie, The Netherlands). Chronopotentiometry required an additional AD converter module (ADC10M, Ecochemie, The Netherlands) for high speed sampling up to 10 M samples per second. The sampling frequency used in chronopotentiometric measurements was 250 kHz. Data postprocessing was performed in OriginPro 8.5. Electrode Fabrication. Microfabricated Si chips with arrays of Pt electrodes served as working electrodes. Processing was done on 8-in Si wafers. Fabrication details have been previously published.34 A 1 μm thick layer of Parylene C was used as insulation material. The 50 μm diameter electrodes were opened by reactive ion etching. After dicing, the chips were wirebonded onto custom PCBs and sealed with epoxy (353ND-T, Epo-Tek). Prior to performing experiments, the chips were thoroughly rinsed in acetone, isopropyl alcohol, and distilled water.
’ METHODS Cyclic Voltammetry. Cyclic voltammetry was performed in PBS, PBSdeox, and PBS-BSA with Pt electrodes of 50 μm diameter to identify the various electrode reactions. Cycling occurred between 0.8 and 1.1 V vs Ag|AgCl at 0.1 V/s. Measurements in PBS-BSA were performed only after electrodes had been immersed for at least 3 h. All measurements were performed at 25 °C. Potential cycling of the working electrode was performed in 0.5 M H2SO4 until a reproducible cyclic voltammogram was obtained. The real surface area was then determined by measuring the charge for the H adsorption peaks and dividing this by the known charge density for monolayer coverage of H adsorbed on Pt (210 μC/cm2).35 Reciprocal Derivative Chronopotentiometry (RDC). Principle. Chronopotentiometry involves applying a constant current between the electrode under study and a counter electrode while measuring the working electrode potential, E, vs time, t, with respect to a reference electrode (Figure 1a). In the case of a biphasic current pulse, an ohmic voltage drop (iR drop) related to wire and solution resistances appears at each current step. At short times, the voltage transient is dominated by the charging of the double layer capacitance. At higher potentials, charge is also consumed by Faradaic reactions causing depletion and/or accumulation of reactants at the electrode surface. These local concentration changes can be observed as inflections in the E(t) transient (indicated by * in Figure 1a). Capacitive contributions from the electrochemical double layer, oxide films and adsorbates lead to steep potential changes between Faradaic reactions. Electrochemical analysis has used RDC for theoretical and experimental studies of various electrode processes.3643 With RDC, the reciprocal time derivative of the electrode potential, dt/dE, is plotted versus E (Figure 1b). Differentiation is used to enhance resolution and correct for unwanted background signals such as capacitive charging. Faradaic reactions give rise to welldefined peaks in the RDC signal, while capacitive effects are characterized by a flattening of the dt/dE(E) curve. Numerous 4013
dx.doi.org/10.1021/ac103037u |Anal. Chem. 2011, 83, 4012–4022
Analytical Chemistry
ARTICLE
studies have investigated the transient behavior of Pt electrodes under current pulsing conditions providing substantial knowledge on the nature of the electrode processes giving rise to the various features of the E(t) transient.10,44 This knowledge can easily be applied to help interpret the RDC graphs. Assignment of the RDC features to specific electrode processes is also facilitated by comparing with CV curves. Both techniques provide qualitatively similar electrochemical signatures of the electrode processes. Reversible (Nernstian) electrode processes in RDC are characterized by a linear relation between the square root of the peak amplitude and the reactant concentration:45 yP 1=2 µ
n3=2 D1=2 c i
ð1Þ
where n is the number of electrons per mole of reaction, D is the diffusion coefficient, c* is the bulk concentration of reactant, and i is the current density. This is an almost equivalent expression as for the current peak height, IP, in CV: IP µ n3=2 D1=2 cv1=2
ð2Þ
where 1/i in eq 1 has been replaced by the square root of the scan rate v1/2. This shows that RDC can be treated in the same manner as CV under galvanostatic conditions.45
Coulometric Detection of Irreversible Faradaic Reactions Using RDC. The dt/dE(E) response presents a behavior qualitatively similar to the voltammetric i(E) response but with higher sensitivity and resolution.45 Combined with coulometric analysis, we propose a new approach to study charge imbalances due to irreversible Faradaic reactions. In order to find the charge densities at which irreversible reactions occur, cathodic and anodic charges densities, qC and qA, respectively, are determined between two corresponding cathodic and anodic states (CCAS) of the voltage transient (Figure 2a). These states are determined in the dt/dE(E) curve (Figure 2b) and may represent either corresponding redox peaks or corresponding capacitive minima. The cathodic and anodic dt/dE(E) minima in Figure 2b, marked by C and A, qualitatively correspond to the same points in the i(E) voltammogram in Figure 2c representing the double layer region of a Pt microelectrode. The presented method of charge calculation is conceptually also used with CV measurements to, e.g. determine the amount of charge involved in adsorption phenomena. In fact, the method described earlier to determine the surface area of Pt using the charge involved in H adsorption is one such example. Here, one determines the charge between point C in Figure 2c and the voltage limit for H2 evolution. The equivalent charge for H desorption is correspondingly determined between the limit for H2 evolution and point A. For purely capacitive electrodes, finding CCAS may be difficult due to the lack of redox reactions that give rise to peaks/minima in the RDC (and CV) curves. The graphically determined consumed charge densities, qC and qA, consist of capacitive and Faradaic contributions: qC=A ¼ qdl, C=A þ qF, C=A
Figure 1. (a) Typical voltage transient, E(t), of a Pt electrode measured in response to a biphasic, symmetric current pulse. (b) Corresponding reciprocal derivative of the voltage transient. Here we plot E(dt/dE) for better comparison to the E(t) transient in a. Indicated are the graph segments related to the iR drop and capacitive polarization. Faradaic reactions are marked with * in both curves.
ð3Þ
with dl and F indicating double layer and Faradaic contributions, respectively. For a perfectly symmetric system, i.e. reversible Faradaic reactions, qF,C = qF,A, and equal capacitive contributions, qdl,C = qdl,A, we obtain qC = qA. However, because dE/dt changes throughout the E(t) transient, there is always a varying capacitive current, idl = Cdl(dE/dt), which can differ for the cathodic and anodic phase. Hence, even for reversible reactions there can be a mismatch between corresponding qC and qA values due to unequal capacitive charging. For irreversible Faradaic reactions during cathodic-first (CF) pulsing, such as H2 gas evolution (Figure 2a), charge is lost since H2 molecules diffuse away from the electrode. In the subsequent anodic phase, this
Figure 2. (a) Voltage transient, E(t), measured in response to a cathodic-first current pulse with applied charge density q0. Indicated are the charge densities, qC and qA, consumed between two corresponding cathodic and anodic states (CCAS). The location of these states, C and A, is obtained from the dt/dE(E) plot in b which qualitatively correspond to the marked points in the CV curve in c. The plot shows irreversible H2 evolution at the end of the cathodic phase, causing qC > qA. 4014
dx.doi.org/10.1021/ac103037u |Anal. Chem. 2011, 83, 4012–4022
Analytical Chemistry
ARTICLE
charge is recovered by reactions outside the regime of qA. The consequence is a charge imbalance with qF,C > qF,A. Eventually, this Faradaic imbalance will exceed the capacitive mismatch causing qC to increase at a faster rate than qA. Similar considerations also apply for anodic-first (AF) pulsing. Here, qA would become larger than qC at the onset of irreversible reactions. Experimental RDC Protocols. Experiments were performed in PBS, PBSdeox, and PBS-BSA using charge-balanced, biphasic CF and AF current pulses. The influence of varying the pulse width and pulse amplitude was investigated. For all chronopotentiometric experiments, 50 successive current pulses were applied to the electrode to allow the system to stabilize and reach steady state. For data analysis, the 50th pulse was examined. All experiments were performed with successively increasing charge densities. Electrodes were conditioned by potential cycling in PBS and PBSdeox prior to applying the current pulses in order to establish a clean electrode surface and to minimize possible memory effects due to previous charge densities. Settings for the CV were as follows: 10 cycles between 0.9 and 1.1 V at a scan rate of 2 V/s. The initial scan direction was either anodic or cathodic depending on whether AF or CF pulses were applied afterward. When pulsing with a constant pulse amplitude (CPA), symmetric current pulses with a peak-to-peak amplitude of 20 μA were used. The phase width varied between 200 and 4000 μs, corresponding to charge densities, q0, ranging from 0.07 to 1.41 mC/cm2. When pulsing with a constant pulse width (CPW), symmetric current pulses with a phase width of 400 μs and varying amplitudes 5100 μA were used. The charge densities also ranged from 0.07 to 1.41 mC/cm2. For all experiments, the interpulse interval was 9.1 ms, resulting in a pulsing frequency ranging from 83 Hz for the longest to 105 Hz for the shortest pulses. No electrode preconditioning was performed in PBS-BSA because this would have led to desorption of the protein film. Otherwise, the stimulation sequences were the same as in PBS.
’ RESULTS AND DISCUSSION Cyclic Voltammetry. Figure 3 shows typical cyclic voltammograms obtained in PBS, PBSdeox, and PBS-BSA. The underlying electrode processes were assigned to the redox reactions described in eqs 410.
Pt þ Hþ þ e h Pt Hads
ðH adsorption=desorptionÞ ð4Þ
2 2H2 PO 4 þ 2e h H2 v þ 2HPO4 ðphosphate deprotonation=protonationÞ
Pt þ 2H2 O h PtO2 þ 4Hþ þ 4e
ð5Þ
ðPt oxide formation=reductionÞ
ð6Þ O2 þ 2Hþ þ 2e f H2 O2 2H2 O þ 2e h H2 v þ 2OH
ðO2 reductionÞ
ð7Þ
ðcathodic water hydrolysisÞ ð8Þ
2H2 O h O2 v þ 4Hþ þ 4e
ðanodic water hydrolysisÞ ð9Þ
For all three curves, H adsorption and desorption (eq 4) occurred between 0.3 and 0.6 V with less well-defined peaks
Figure 3. Cyclic voltammograms of Pt electrodes measured in PBS, PBSdeox, and PBS-BSA. The inset shows the details of the anodic region involving Pt and protein oxidation.
in PBS-BSA. Beginning at 0.7 V, a reduction shoulder was observed in PBS related to the deprotonation of phosphate anions (eq 5).46,47 The current density related to this reduction wave was much larger in PBSdeox most likely because the agitation of the electrolyte increased the rate of phosphate ion diffusion to the electrode. Phosphate deprotonation was also slightly enhanced in PBS-BSA as compared to PBS. This may be related to the net negative charge of the adsorbed BSA which forces the dissociation of H 2PO4. The acid dissociation constant of phosphate anions (pKa = 7.2) lies close to the experimental pH value. Phosphate anions therefore represent weak acids prone to dissociation in the presence of suitable proton acceptors (bases). While albumin itself is a weak acid at physiological conditions, it may undergo deprotonation (reduction) at negative electrode voltages and can therefore act as a base. A broad Pt oxidation region (eq 6) was seen in PBS and PBSdeox between 0.25 and 0.9 V. The corresponding Pt oxide reduction peak occurred at =0 V in PBS and PBSdeox. In contrast, surface electrochemical reactions including Pt oxidation and oxide reduction were inhibited in PBS-BSA due to the adsorbed protein film.48 This was evidenced by the consistent suppression of both the Pt oxidation shoulder at 0.4 V (Figure 3, inset) and the oxide reduction peak. The voltammogram recorded in PBSBSA showed an increased oxidative current in the anodic sweep from 0.5 V onward (Figure 3, inset), indicating the oxidation of amino acids.4951 The onset of H2 (eq 8) and O2 (eq 9) evolution were observed at 0.6 and 0.95 V, respectively, for all curves. The current step between 0 and 0.2 V is due to the irreversible reduction of dissolved O2 (eq 7), which was still present in PBSdeox to a certain extent. Deoxygenation decreased this current step and hence the amount of dissolved O2 by approximately 70%. Also, agitation of the PBSdeox may have promoted O2 diffusion, thus leading to a larger current density due to reduction of O2 than would have been observed in a quiescent solution. The O2 concentration in PBS can be estimated using Henry’s law, c = pg/kH, relating the solubility, c, of a gas to its partial pressure, pg, where kH is an empirical constant (kH = 781 atm(mol/L)1 for O2 dissolved in distilled water52). At ambient temperature (25 °C) and pressure (1 atm), the O2 partial pressure is pO2 = 0.21 atm taking into accounting the partial 4015
dx.doi.org/10.1021/ac103037u |Anal. Chem. 2011, 83, 4012–4022
Analytical Chemistry
ARTICLE
Figure 4. Series of dt/dE(E) plots obtained with constant pulse amplitude stimulation in PBS and PBS-BSA: (a and b) cathodic-first stimulation; (c and d) anodic-first stimulation. Only a selection of all applied (absolute) charge densities, q0, is shown. Graphs are stacked by a constant offset. Regions of Pt-specific reactivity are indicated by horizontal arrows. Long vertical arrows point in the direction of increasing applied charge densities, q0. The short vertical arrow in d marks the peak identified as O2 reduction.
fraction of O2 in air. We therefore obtain c = 269 μM. A reduction by 70% in PBSdeox hence corresponds to c = 81 μM. The O2 partial pressure in the rat’s brain is very heterogeneous and covers a range 4.6 10372.4 103 atm53 resulting in O2 concentrations between 5.9 and 93 μM. The charge for H adsorption obtained from CV in H2SO4 was used to calculate the real surface area of the Pt electrodes. The roughness factor, i.e. the ratio between real and geometric surface area, was 1.44 ( 0.07. All further area-specific values are based on this roughness factor. Reciprocal Derivative Chronopotentiometry (RDC). Graphical Analysis of RDC Plots: PBS versus PBS-BSA. Figure 4 shows two exemplary RDC data sets obtained with CPA stimulation in PBS and PBS-BSA. In each case, only a selection of all investigated charge densities is plotted. Throughout the discussion, charge density values q0, qC, and qA are referred to in terms of their absolute values. The polarities are implicit by the annotations CF (negative) and AF (positive) for q0 and the subscripts, C (negative) and A (positive) for qC and qA. For better clarity, the iR drop is not shown. The observations for CPA stimulation also apply to a large extent to CPW stimulation (Figure S-1 of the Supporting Information). Differences are mentioned where required. Similar to CV (Figure 3), Pt-specific reaction peaks were identified in the RDC graphs. Anodic features include H2 oxidation, phosphate protonation and H desorption (H2O/PHP/HD),
Pt oxidation and BSA oxidation (PO/AO), and Pt dissolution (eq 10) and O2 evolution (PD/O2E). Pt þ 4Cl h ½PtCl4 2 þ 2e ðPt dissolution=reduction of dissolution productsÞ
ð10Þ
The cathodic phase revealed a narrow reductive peak for AF stimulation in PBS which was assigned to the reduction of Pt dissolution products (R-PD). Results supporting this conclusion come from voltammetric measurements of Pt microelectrodes performed in PBS solution using an extended voltage range.47 Furthermore, a broad reduction band with multiple shoulders was assigned to the reduction of dissolved O2 and Pt oxide (O2R/OR). Hydrogen adsorption (HA), phosphate deprotonation, and H2 evolution (PHD/H2E) occurred at the end of the cathodic phase. Unlike CV, no distinct feature in the cathodic RDC curves could be linked to phosphate deprotonation. Because the product of phosphate deprotonation is H2 (eq 5), both H2E and PHD were grouped together. Additional experiments are needed to further examine the extent of PHD to the overall H2 evolution. A pronounced flat region at the beginning of the anodic phase for CF stimulation was related to the (dis-)charging of the double layer capacitance. CF Stimulation. In PBS-BSA, peaks for H adsorption and desorption appeared at lower charge densities (
