Journal of The Electrochemical Society, 161 (3) A275-A284 (2014) 0013-4651/2014/161(3)/A275/10/$31.00 © The Electrochemical Society

A275

An In Situ Synchrotron Study of Zinc Anode Planarization by a Bismuth Additive Joshua W. Gallaway,a,∗,z Abhinav M. Gaikwad,a Benjamin Hertzberg,b,c Can K. Erdonmez,d Yu-chen Karen Chen-Wiegart,e Lev A. Sviridov,a Kenneth Evans-Lutterodt,e Jun Wang,e Sanjoy Banerjee,a and Daniel A. Steingartb,c,∗,z a The

CUNY Energy Institute at the City College of New York, Department of Chemical Engineering, New York, New York 10031, USA b Department of Mechanical and Aerospace Engineering, Princeton University, Princeton, New Jersey 08544, USA c Andlinger Center for Energy and the Environment, Princeton University, Princeton, New Jersey 08544, USA d Energy Storage Group, Brookhaven National Laboratory, Upton, New York 11973, USA e Photon Sciences Directorate, Brookhaven National Laboratory, Upton, New York 11973, USA Cyclic voltammetry of zinc plated from flowing alkaline zincate electrolyte with a bismuth additive showed a marked mass transport effect during metal layer deplating. This bismuth was added as Bi2 O3 and had a saturated concentration of 26 ppm bismuth. Using a small, transparent window flow cell the mechanism was studied in situ using synchrotron X-rays. X-ray microdiffraction revealed that the metal-electrolyte interface was bismuth rich, and bismuth behaved in a manner similar to a surfactant. Transmission X-ray microscopy revealed that in the presence of bismuth additive, 5 μm raised features on the metal layer were preferentially dissolved during deplating. However, macro-morphology experiments demonstrated that at 26 ppm a detrimental bismuth buildup occurred over many cycles. By reducing additive concentration to 3 ppm a metal layer was planarized compared to a no-additive control, while avoiding the bismuth buildup. These findings suggested that 3 ppm bismuth could be used to planarize zinc metal layers such as those in flow-assisted zinc batteries. However, concentration will need to be well-controlled. © 2013 The Electrochemical Society. [DOI: 10.1149/2.037403jes] All rights reserved. Manuscript submitted September 30, 2013; revised manuscript received November 13, 2013. Published December 20, 2013.

Zinc anode batteries are desirable for secondary energy storage due to the water-compatibility, natural abundance, safety, and high energy density of zinc.1 Zinc anodes are found in several forms, which vary widely in design. Examples include the anodes in the following: metallic zinc flow battery hybrids, pelletized zinc-air battery/fuel cell hybrids, zinc-manganese dioxide batteries, and flow-assisted alkaline nickel-zinc batteries.2–9 Systems in which a zinc anode is plated onto a current collector from electrolyte under forced convection have received attention for large scale applications, as in the flow-assisted nickel-zinc battery referenced above.4 A key challenge for these systems is high aspect ratio zinc “dendrites” which progressively form during charge-discharge cycling of the zinc layer. Unless mitigated, these structures cross the electrolyte gap and short the cell. The micro-morphology of these progressive dendrites is generally the mossy form of zinc. Thus they are not like crystalline dendrites or diffusion limited aggregation structures except in their aspect ratio. Strategies to prevent dendrites during metal plating are the subject of frequent investigations, such as varying current density, electrolyte flow rate, and current waveforms.10–12 These all aim to prevent initial formation of a high aspect ratio profile on the metal layer, as once that has occurred electric field effects will intensify dendrite formation. Additives or leveling agents also work in this way, suppressing deposition at asperity tips.13 Screening electrolyte additives for use in flow-assisted batteries, we observed that ppm quantities of bismuth oxide provoked a substantial mass transport effect during anodic dissolution of zinc. This lead us to speculate that small quantities of bismuth oxide could act as a “reverse” leveling agent, planarizing zinc layers during discharge of the layer. The action of bismuth on zinc plating has been discussed in recent years, however the experiments were performed in quiescent electrolyte.14,15 While practically relevant, this situation makes experimental results difficult to interpret as free convection currents dominate and are challenging to quantify. In this study, electrolyte under forced convection was used to control mass transport effects. In a flowing alkaline electrolyte system PbO and Na2 WO4 additives have recently been shown to inhibit spongy zinc growth of metallic zinc.16 By using nano-scale transmission X-ray microscopy in situ we found that bismuth oxide planarization of zinc occurs primarily during anodic dissolution, which was unrelated to the plating findings ∗ z

Electrochemical Society Active Member. E-mail: [email protected]; [email protected]

of Wen et al. We predict from the data of Wen et al that lead oxide would achieve the same effect but tungstate would not. Inorganic oxides are frequently used as additives in porous, pasted zinc anodes, where their mechanism is related but ultimately dissimilar to that in flow, as pore diffusion dominates in such systems.17,18 McBreen et al. have studied many additives in pasted electrodes, including bismuth oxide, finding that additives plate on the current collector and result in a substrate effect either due to suppression or enhancement.12,19–24 For example PbO and In(OH)3 result in substrates that increase anode polarizability, while CdO and Ga2 O3 reduce it. With Bi2 O3 specifically, the plated additive metal also results in a conductive matrix throughout the electrode.25 In the current study it was found that bismuth additive planarized zinc layers cycled in a flowing electrolyte. The mechanism of planarization occurred during discharge of the layer. Bismuth was concentrated at the metalelectrolyte interface, and thus resembled the action of a surfactant. Experimental Electrolyte preparation.— ACS grade potassium hydroxide (KOH) and zinc oxide (ZnO) were purchased from Alfa Aesar. Bismuth(III) oxide (Bi2 O3 ) at 99.999% purity was purchased from Sigma Aldrich. Alkaline zincate electrolyte was prepared by dissolving 8.9 M KOH in deionized water, followed by addition of 610 mM ZnO, stirred overnight. Bismuth additive was added as Bi2 O3 in excess at ∼80 mg per 100 mL, and this was stirred for 24 hours. Some Bi2 O3 remained in solid form, which was allowed to settle while the test solution was decanted from the top. Atomic absorption studies were performed to confirm that after 24 hours of dissolution, bismuth concentration achieved a steady state and was saturated. Lower concentrations of bismuth additive were prepared by combining virgin zincate electrolyte with bismuth-saturated solution in the desired ratio. The concentrations of these dilutions was determined using a Thermo M series atomic absorption spectrometer. Atomic absorption spectroscopy (AA) indicated the bismuth saturation concentration in 8.9 M KOH with 0.6 M ZnO was 25.6 ± 0.9 mg Bi/L based on five experiments. This resulted in a molar concentration of 1.2 × 10−4 M in the form BiO2 − , in general agreement with previously reported values in similar KOH electrolytes.26–29 Bismuthsaturated 8.9 M KOH without zinc oxide had a concentration of 28 mg Bi/L, slightly higher than the concentration with zincate present. Dilutions of bismuth saturated electrolyte resulted in linear bismuth

Downloaded on 2013-12-21 to IP 174.101.38.169 address. Redistribution subject to ECS license or copyright; see ecsdl.org/site/terms_use

A276

Journal of The Electrochemical Society, 161 (3) A275-A284 (2014)

Figure 1. Design of the transparent window flow cell used to monitor zinc plating and deplating in situ using synchrotron X-rays. Size of the flow cell was small enough to allow work in confined spaces. (a) Nickel foil electrodes and PDMS on a PET backing. The doctored PDMS layer was approximately 150 μm thick in the direction out of the page, making a flexible, thin cell. (b) A flow channel laser cut into the PDMS. (c) Acrylic cell compression rig added, showing window for X-ray transmission. Perimeter holes were for cap screws. Thickness of the cell at the window was approximately 200 μm.

concentrations, to within resolution of the AA. For example a 1:1 dilution of electrolyte with saturated bismuth resulted in 13 ppm bismuth. Throughout this work concentrations measured as ppm quantities of bismuth are interchangeably referred to as fractions of saturation and molarity of BiO2 − . For example 26 ppm bismuth = 100% of bismuth saturation = 1.2 × 10−4 M BiO2 − and 13 ppm bismuth = 50% of bismuth saturation = 6.0 × 10−5 M BiO2 − . Experimental flow cell construction.— To study the impact of bismuth oxide additive on zinc plating/deplating in forced convection, we used three flow cell designs. Each flow cell served a different purpose, either allowing the metal layer to be monitored by light (visible or X-ray) or minimizing ohmic resistance in the potential signal. For in situ X-ray studies a transparent window flow cell was designed to allow X-ray transmission through the flow channel at the working electrode interface. This allowed imaging the beginning stages of metal plating. Construction of the window cell is diagrammed in Figure 1. Nickel foil of 100 μm thickness (99.5%, Alfa Aesar) was cut into electrode flags 6 mm long. The edges of these electrodes that would contact the flow channel were polished using silicon carbide abrasive in decreasing grit size in the order 600, 800, 1200. Polishing the edges was accomplished by sandwiching the foil electrodes in a PTFE holder. The polished electrodes were attached to a 1 mil PET film. The flat edges of the electrodes were separated by a 1 mm gap. A liquid PDMS layer was doctored over the film and electrodes using a micron-adjustable film applicator (MTI Corporation, Richmond CA). After curing, the PDMS layer was approximately 150 μm thick, which was just enough to cover the electrodes. A 1 mm wide flow channel was cut between the electrodes using a VersaLaser VLS6.60 laser engraver (Universal Laser Systems; Scottsdale, AZ). Alignment of the laser was aided using a stereoscope. The flow channel had two landings, one each for flow inlet and outlet. After cutting out the flow channel, the electrode edges were skimmed with a steel probe to assure no residual PDMS remained. A second PET film capped the cell and enclosed the flow channel. The procedure above resulted in a thin, flexible cell. Light passing through the flow channel was only attenuated by 150 μm of electrolyte and 50 μm of PET film (2 films × ∼25 μm each). The choice of PET was for good X-ray transparency and resistance to the highly basic KOH electrolyte. After 24 hours of exposure to KOH the PET showed no detectable damage. After constructing the thin cell, it was compressed in an acrylic rig to seal the cell. Flow tubes for the electrolyte were connected to this rig, and they mated with the flow channel landings through holes punched in one of the PET films. A cutout window in the acrylic rig matched the cell electrode area, allowing a path for X-ray transmission. A transparent window flow cell is shown in the confined TXM apparatus in Figure 2. One electrode served as

Figure 2. Transparent window flow cell shown in the confined TXM apparatus at beamline X8C at the National Synchrotron Light Source. Flow tubing and test leads were connected to a syringe pump and potentiostat off camera. The cell could be rotated to focus on the working electrode via a visible light camera. Bottom image shows the setup from above.

the working electrode, while the other was a dual counter/reference electrode. A three-electrode microfluidic cell, which has been described in detail previously, was used for cyclic voltammetry (CV).30,31 This cell was originally designed to resolve fast electrolyte changes over the working electrode. However in this work we used it to minimize ohmic resistance, while simultaneously controlling working electrode hydrodynamics. This was important because we have previously reported that localized current density on dendrite tips can be high.32 While a rotating disk electrode effectively allows identification of kinetic, ohmic, and mass transport contributions to the WE overpotential, at high currents ohmic resistance can dominate. In the three-electrode microfluidic cell, both WE and CE were micro-electrodes (100 μm and 250 μm diameter respectively) with a separation of ∼400 μm. The WE diameter was large compared to the diffusion layer thickness, so the diffusion layer could be modeled as linear. The working channel had a rectangular cross section 500 μm wide and 180 μm tall and a parallel reference channel of identical size contained a Ag|AgCl reference electrode (RE) (BASi, model RE-6). The macro-morphology of electrodeposited metal layers was imaged in a lateral flow cell, which has also been reported previously.32 This cell was similar to the transparent window flow cell but was produced from custom-drawn printed circuit boards (PCBs) fabricated through Sunstone Circuits (Mulino, OR). While easier to produce,

Downloaded on 2013-12-21 to IP 174.101.38.169 address. Redistribution subject to ECS license or copyright; see ecsdl.org/site/terms_use

Journal of The Electrochemical Society, 161 (3) A275-A284 (2014) these cells could not be used for the in situ X-ray work because the board material (FR4) was not X-ray transparent. Two rectangular silver-coated copper electrodes served as working (WE) and counter/reference (CE/RE) These were 6 mm long, 30 μm high, and 1 mm apart. The PCBs were coated with a 200 μm layer of poly(dimethylsiloxane) (PDMS) using a micrometer adjustable film applicator (MTI Corporation, Richmond, CA). After curing for 1.5 h at 70◦ C, the PDMS had a cured thickness of 100 μm. A VersaLaser VLS6.60 laser engraver was used to cut a flow channel in the PDMS, flush with the electrode edges. Electrochemical data.— For three-electrode cell CV experiments, potential sweep rates of 2, 5, and 10 mV/s yielded comparable results, therefore 5 mV/s was used in all experiments. Measurements were recorded with a Metrohm μAutolab III. CV potentials were reported relative to a Ag|AgCl reference electrode. Electrolyte flow was controlled by a syringe pump (New Era Pump Systems, Farmingdale NY) in all flow cells. For plating and deplating experiments performed in the transparent window flow cell, galvanostatic current was controlled using a Metrohm μAutolab III. The WE active area was determined to be 0.0127 cm2 by matching previously reported potentials for a zincoxygen cell (see supplemental material for details). Metal plating was performed at 0.5 mA or 39 mA/cm2 for the time necessary to plate the desired layer thickness. Deplating was performed at 1.5 mA or 118 mA/cm2 . All window flow cell experiments were carried out at 10 mL/hr electrolyte flowrate, corresponding to a superficial velocity of 1.8 cm/s. Electrolyte flow direction was from cell top to bottom (see Figure 2), to allow for more well-developed boundary layers at the electrodes. WE potentials were relative to the CE/RE potential, which was generally either that of oxygen formation or zinc plating (see supplemental). For cycling experiments in a lateral flow cell, zinc was deposited on the WE at 35 mA/cm2 for 2107 s, followed by dissolution at 35 mA/cm2 for 600 s. This imbalance was to simulate layer buildup during real-world cycling conditions. Every time the current switched there was a 20 s open circuit rest period. This plating/deplating cycle was repeated 14 times, or until the cell shorted. For the lateral flow cell results reported, flowrate was 1 mL/hr, corresponding to a superficial velocity of 0.3 cm/s. The rigidity of the lateral flow cells allowed a lower flowrate than in transparent window flow cells. A Princeton Applied Research VersaSTAT 4 potentiostat was used to control WE current. Scanning electron micrographs of metal layers were collected ex situ with a Hitachi TM-3000 benchtop SEM. Metal layers were imaged in situ using an Omano microscope and a Flea2 digital camera (Point Gray Research, Inc., Richmond, BC) interfaced with Astro IIDC software (Aupperle Services and Contracting, Calgary, AB). Image analysis was performed using ImageJ (Wayne Rasband, NIH). X-ray Microdiffraction.— Synchrotron X-ray microdiffraction measurements were performed at the National Synchrotron Light Source beamline X13B at X-ray wavelength 0.65 Å. The beamline had an in vacuum undulator, described previously.33 There was a water cooled monochromator to select the X-ray wavelength, with highly polished Si (111) oriented surfaces, aligned in an artificial channel cut arrangement. A diamond window was the exit window from the beamline. The experiment was done in transmission with the sample at the beam spot, which was 2 μm vertical by 5 μm horizontal focused by a K-B mirrors system with a working distance of 1 cm and focal lengths of 5 cm and 10 cm in the horizontal and vertical respectively. The asymmetry of the focused X-ray beam size reflected the asymmetry of the electron beam inside the NSLS synchrotron ring. Beam size was measured with fluorescence from a nano-patterned Cr knife edge. A Princeton Instruments CCD area detector was used to acquire diffraction patterns. No flood field or spatial distortion corrections were performed, due to reasonably high fidelity of the fiber optic taper.

A277

Focusing the 2 × 5 μm beam on the working electrode interface in the transparent window flow cell allowed continual monitoring of the working electrode during metal deposition. By moving the cell laterally (horizontally) on a translational stage, the deposited metal layer could be scanned for material variations in the direction normal to the working electrode surface. The resolution in this direction was 5 μm, determined by the horizontal beam size. Diffraction data collected at the area detector was converted to angle dispersive data using a custom-made Python program. This was done by circular integration along the diffraction rings, in order to maximize the signal to noise ratio. The diffraction background caused by the cell was a fraction of the noise caused by the aqueous electrolyte, which resulted in an amorphous signal at very low angles. Transmission X-ray microscopy.— Transmission X-ray microscopy (TXM) experiments were performed at beamline X8C at the National Synchrotron Light Source. TXM is a full-field imaging technique with a Fresnel zone plate as the objective lens. A 28 nm spatial resolution has been demonstrated at X8C.34 This beamline therefore enabled in situ nano-scale imaging of the profiles of metal layers using the transparent window flow cell. X-ray projections of the metal layers through the flow cell were continuously recorded by a 2k × 2k CCD detector during the plating/deplating. Absorption contrast mode was used and therefore the contrast in the recorded images was determined by the attenuation coefficient and the thickness of the metal layers. X-ray energies both below and above the Zn absorption K-edge (9659 eV) were used: 9600 and 9700 eV respectively. The 2k × 2k CCD camera binned 4 × 4 pixels. A 42 × 42 μm field of view of the microscope was used, resulting in a pixel size of 82 nm. This binning enabled a short exposure time of 1 s to capture the dynamics of the plating and deplating reactions. As a trade-off, the features which could be resolved during the in situ TXM experiments were ∼100 nm. For each in situ imaging experiment, 8 background images were collected separately and averaged for pixel by pixel normalization. Results and Discussion Cyclic voltammetry of zinc with bismuth additive.— Figure 3 compares the cyclic voltammograms (CVs) of virgin zincate solution (8.9 M KOH and 0.6 M ZnO, which is solubilized as zincate ions) and the same solution with 1.2 × 10−4 M BiO2 − in the electrolyte, which was measured as 26 ppm bismuth and 100% of saturation. The working electrode (WE) was a platinum disk with a diameter of 100 μm. The WE was flush with the flow channel wall, and the electrolyte was flowing at 10 mL/hr or a superficial velocity of 3 cm/s. Using the Leveque solution, the average zincate diffusion layer thickness during deposition was estimated to be ∼5.5 μm. As the WE diameter was large compared to this value, it could be approximately treated as a flat plate with linear diffusion. Scans in Figure 3 were from −0.2 V to −1.6 V and returning, at 5 mV/s. With virgin zincate solution, zinc deposition (black arrow) was observed at −1.51 V, resulting in a large cathodic current. When the scan reversed zinc was deplated, resulting in a large anodic current until the zinc layer was exhausted and current fell to effectively zero. The Figure 3 inset shows a reduced current scale. As platinum was an effective hydrogen evolution surface, cathodic hydrogen generation was observed at potentials below −1.15 V (red arrow). These two cathodic reaction are those observed during charging of a zinc anode in a flow-assisted battery.4,35 These electrode potentials for zinc and hydrogen reduction observed (−1.31 V vs. SHE for zinc, −0.93 V vs. SHE for hydrogen) matched reported values from the literature.36,37 The presence of 26 ppm bismuth or 100% bismuth saturation had a pronounced impact on the CV results. Zinc deposition was inhibited, occurring at a lower potential and with lower overall current. Initially, deplating of the metal layer was kinetically more rapid in the presence of bismuth, judged by the Tafel slope. However, the overall anodic curve was more rounded. The bismuth metal electrode potential was observed at −0.75 V. A small bismuth deposition current

Downloaded on 2013-12-21 to IP 174.101.38.169 address. Redistribution subject to ECS license or copyright; see ecsdl.org/site/terms_use

A278

Journal of The Electrochemical Society, 161 (3) A275-A284 (2014)

Figure 3. Cyclic voltammograms of plating and deplating for zinc (black curve) and zinc with saturated bismuth additive (blue curve) on a platinum working electrode. Potential scan rate was 5 mV/s from −0.2 V to −1.60 V and back. Flowrate was 10 mL/hr. Inset shows a zoomed current density scale. Arrows show outbound reaction onsets. Blue arrow: bismuth plating. Red arrow: hydrogen formation. Black arrow: zinc plating.

of 1.3 mA/cm2 was observed at −1 V during outbound scans (blue arrow). Multiple peaks associated with bismuth deplating were observed during returning scans, in agreement with the work of Vivier and co-workers.26–28 Figure 4 shows the effect of varying electrolyte flowrate on metal plating and deplating. This revealed how mass transport affected the electrode reactions. Outbound scans highlighted in Figure 4a show that below currents of ∼100 mA/cm2 an order of magnitude change in virgin zincate flowrate (1 and 10 mL/hr) had little effect, with a Tafel slope of ∼51 mV. This agreed with our previous work on copper electrodes, and works by Cachet and Despi´c.32,38–40 The plating suppression caused by bismuth could be seen, along with sudden plating activation at −1.56 V. Flowrate had little effect on plating with bismuth except in the case of 1 mL/hr, which owing to low bismuth mass transport appeared to be an intermediate case between the conditions with and without bismuth. Bismuth also lowered the hydrogen formation current. The effect of bismuth mass transport on metal deplating was substantial, as shown in Figure 4b. Anodic deplating current increased with increasing electrolyte flowrate. Comparing the deplating current at −1.51 V for the 20 and 1 mL/hr cases finds a large difference of 122 vs. 10 mA/cm2 . Quantifying this effect using the anodic Tafel slope reveals a change from 14 mV at 20 mL/hr to 32 mV at 1 mL/hr. This effect was not caused by a difference in material inventory deposited on the electrode, as illustrated by the control data without bismuth, in which the kinetic region of the deplating wave was unaffected by flowrate, within experimental error. Without bismuth there was little to no effect of flowrate, and anodic Tafel slope was ∼16 mV. This demonstrated that anodic deplating was a strong function of mass transport when bismuth was present. This mass transport enhanced dissolution suggested that this bismuth species would act as a reverse leveling agent. Traditional leveling agents work by suppressing deposition current in areas of higher mass transport. As asperity tips and nascent dendrites are locations of increased mass transport, leveling agents suppress deposition there and encourage the growth of flat metal layers. In contrast, this bismuth species encouraged increased deplating in such locations. Therefore, provided a metal layer underwent both plating and deplating cycles, this small ppm concentration of bismuth would planarize a zinc layer, which is the effect desired in flow-assisted zinc anode batteries. A similar co-deposition effect has been reported as a method to control the composition of a binary alloy when produced by reversepulse plating of two metals.41,42 Roy and co-workers demonstrated

that during reverse-pulse plating of nickel and copper, copper was deposited at all stages of the pulse cycle, while nickel was alternately plated and deplated during the pulse on-time and reversal-time.43,44 In the work of Roy and co-workers the more noble metal, copper, was at low concentration causing it to plate at the mass transfer limit during all stages of the pulse cycle. In this work bismuth plating was under these circumstances. Thus the less noble metal (nickel in the case of Roy, zinc in this case) was selectively dissolved or displaced by the more noble metal during the reversal-time (copper in the case of Roy, bismuth in this case). The situation of a battery with a plated metal anode during cycling is essentially that of reverse-pulse plating with extremely long on and reversal times. Selective displacement of the less noble metal will be greatest at asperity tips, planarizing the metal layer. The CV results above were repeated in triplicate, and were also qualitatively the same on a copper WE. A few secondary results follow. Due to hydrogen suppression, bismuth increased the plating/deplating coulombic efficiency from 0.79 to 0.84. The presence of bismuth additive lowered the zinc electrode potential (EZn ) during returning scans. At 10 mL/hr, EZn = −1.507 ± 0.007 V with virgin solution. With bismuth additive this value lowered to EZn = −1.525 ± 0.005 V. This is also likely due to lessening of the Zn-H2 mixed potential. X-ray microdiffraction.— The suppression of current on outbound potential scans could be explained by a substrate effect, as bismuth had been plated before zinc, or by a surface blocking mechanism. However, the change during returning scans indicates that bismuth has a kinetic effect on zinc anodic dissolution. At these potentials, zinc was expected to deplate while the bismuth would remain at a plating potential, resulting in exchange plating and preferential corrosion of nascent dendrite tips. This was different than the blocking mechanisms seen by Wen et al., in which Na2 WO4 did not affect anodic kinetics, and PbO did only at high concentration.16 Wang et al. observed only a minor change in anodic Tafel slope in quiescent electrolyte, adding bismuth in the form BiCl3 .14 X-ray microdiffraction was used to elucidate the mechanism that resulted in these findings. This allowed rapid collection of diffraction data from crystalline materials on the electrode surface. The microdiffraction capabilities at beamline X13B at NSLS enabled a beam spot size of 2 × 5 μm. Using the transparent window flow cell, this beam was focused on the WE interface to observed material evolution during plating. The window flow cell was constructed with nickel

Downloaded on 2013-12-21 to IP 174.101.38.169 address. Redistribution subject to ECS license or copyright; see ecsdl.org/site/terms_use

Journal of The Electrochemical Society, 161 (3) A275-A284 (2014)

A279

39 mA/cm2 from zincate electrolyte with 100% saturated bismuth additive. As a metal layer plated, the beam signal was attenuated compared to the results in panel A. The zinc (101) reflection was the first deposition signal detected, 120 s after beginning deposition. At this time the theoretical metal layer thickness was 2 μm. Bismuth metal (012) was first detected at this location after 380 s. While studying additive mechanisms in porous, pasted zinc anodes in stagnant electrolyte, McBreen and Gannon observed that bismuth prevented zinc underpotential deposition and served as an inhibitor. This was due to a substrate effect, as bismuth metal plated over the current collector.25 As we observed zinc reflections as the first sign of metal deposition, we conclude that if there was a similar substrate effect in our flowing electrolyte system, this bismuth layer was below the X-ray diffraction detection limit. Plating was halted at 2170 s, when the theoretical metal layer thickness was 42 μm. (Cell potential data is reported in the supplemental section.) With the cell at open circuit, the metal layer thickness was scanned normal to the WE surface. The metal layer thickness was found to be ∼53 μm, suggesting this location was slightly raised compared to the metal layer average. Figure 6 shows diffraction data acquired from −7.5 μm (just past the WE interface) to 52.5 μm (the metal-electrolyte interface), with 0 μm defined as the location monitored in Figure 5. As expected nickel was only detected at the current collector. Zinc reflections faded near the electrolyte interface, while bismuth reflections dominated. Thus bismuth was located throughout the metal layer, but was the dominant component near the electrolyte. This suggested bismuth acted in a surfactant manner during codeposition. The primary bismuth reflection at 26.1 degrees was at a slightly lower angle than expected (27.1 degrees) and displayed some peak splitting. The possibility of zinc-bismuth alloy formation could explain these observations. The thermal equilibrium phase diagram of zinc and bismuth indicates no bismuth alloying in zinc and only 1% zinc alloying in bismuth. However, it is possible that electrodeposited layers display different behavior. The beam was again focused on the original nickel interface and the metal layer was galvanostatically deplated at 118 mA/cm2 . Timeresolved results are shown in Figure 7. While the zinc signal disappeared completely, bismuth was concentrated to such an extent that secondary bismuth reflections became pronounced. In the range 2θ = 35–40 the (002) and (100) zinc reflections gave way to a strong bismuth signal. This demonstrated that bismuth remained at the electrolyte interface as the metal layer deplated until the only electrode species were bismuth and nickel. This residual bismuth was only eliminated when the WE was scanned to more positive potentials, as apparent in Figure 3. Figure 4. Cyclic voltammograms of plating and deplating for zinc (black curves) and zinc with saturated bismuth additive (blue curves) on a platinum working electrode, with electrolyte flowrates as indicated. Potential scan rate was 5 mV/s from −0.2 V to −1.60 V and back. (a) Outbound (negative) scans showing the region of initial zinc deposition. (b) Returning (positive) scans showing metal deplating. Inset: Outbound and returning scans in the region of bismuth plating/deplating at 5 mL/hr flowrate.

foil electrodes, as nickel is a typical current collector in zinc anode batteries. When the microdiffraction beam was focused on the nickel WE interface with flowing zincate electrolyte, results were a diffraction pattern as shown in Figure 5A with characteristic nickel metal reflections. Raw results were transformed to angle dispersive results using a homemade Python script, as shown in the right panel. Further into the bulk of the WE nickel showed a strong preferred orientation, however at the interface the diffraction rings were smooth due to the polishing procedure used. Position of the beam along the electrode was at the approximate midpoint ∼3 mm from the electrode leading edge. Data collection time for each diffraction pattern was ∼20 s. Figure 5b shows the diffraction pattern at the initial nickel electrode interface after 400 s of galvanostatic metal deposition at

Transmission X-ray microscopy.— Transmission X-ray microscopy was used to confirm that metal layers were planarized during deplating when bismuth additive was present. TXM experiments were conducted in the same manner as the microdiffraction experiments. The WE interface was continually monitored via the X-ray beam near the midpoint of the length (∼3 mm from the leading edge). A metal layer was plated at 39 mA/cm2 . The metal layer was then deplated at 118 mA/cm2 . Electrolyte flowrate was 10 mL/hr. A metal layer with a theoretical thickness of 8 μm was plated from zincate electrolyte with 100% bismuth saturation. The profile of this metal layer is shown in the first panel of Figure 8. This TXM data was collected with a beam energy of 9700 eV, above the K-edge of zinc (9659 eV). For this reason the metal layer appeared opaque to the beam. A dashed gray line shows the position of the nickel electrode interface, which was to the right in the image. Thus the direction of metal layer growth was to the left, in the direction of the CE. Electrolyte flow was from top to bottom. It can be seen that the plated metal layer displayed a sinusoidal profile with an amplitude of ∼5 μm. Our hypothesis was that raised locations on the layer, being closer to the CE, would experience a higher growth rate due to the electric field and become nascent dendrites. However during deplating any raised nascent dendrite would undergo mass transport enhanced dissolution due to the bismuth additive. During deplating, this phenomenon was

Downloaded on 2013-12-21 to IP 174.101.38.169 address. Redistribution subject to ECS license or copyright; see ecsdl.org/site/terms_use

A280

Journal of The Electrochemical Society, 161 (3) A275-A284 (2014)

Figure 5. X-ray microdiffraction from a 2 × 5 μm spot focused on a nickel electrode interface in contact with flowing zincate electrolyte (1.8 cm/s) with 100% saturated bismuth additive. Raw diffraction results were converted to angle dispersive results, shown on the right. Shadow on the top left was from the fluid inlet tubing. (a) Initial nickel interface. (b) The interface during co-deposition of zinc and bismuth.

observed. White and blue arrows mark raised and recessed locations on the initial metal layer profile. In subsequent images taken during deplating these arrows remain in place for illustration. After deplating

Figure 6. A microdiffraction spatial scan of the metal layer plated in Figure 5b. The WE as maintained at open circuit potential and the cell was scanned to observe the material variation normal to the WE. Distances were with respect to the initial nickel interface. Bismuth dominated the diffraction signal at 52.5 μm, the metal layer interface with the electrolyte, shown by the arrow.

for 22 s, dissolution was confined ∼3 μm at the raised tip, while no dissolution was observed in the recessed location. After 40 s a second raised location at the bottom of the image was dissolved and the layer was more planar. After 50 s the entire layer had dissolved, revealing the nickel electrode interface.

Figure 7. Time resolved microdiffraction of the initial nickel interface location during metal layer deplating. The zinc signal vanished while that of bismuth intensified.

Downloaded on 2013-12-21 to IP 174.101.38.169 address. Redistribution subject to ECS license or copyright; see ecsdl.org/site/terms_use

Journal of The Electrochemical Society, 161 (3) A275-A284 (2014)

A281

Figure 8. Deplating of a Zn/Bi metal layer in 100% Bi saturated zincate electrolyte. Dashed gray line: position of the nickel current electrode. White arrow: Beginning position of a raised feature. Blue arrow: Beginning position of a recessed area.

Microdiffraction results suggested that the deplated nickel WE in the 50 s image of Figure 8 had a concentrated bismuth layer. This electrode was scanned to a positive potential and the characteristic bismuth stripping peaks were observed by CV (see supplemental). However, this residual bismuth layer was not detected by TXM. Note that the shadow in the 50 s image of Figure 8 was not a metal layer and was observed both before plating and after bismuth stripping. This suggested that the concentrated bismuth layer formed during deplating was physically smaller than the ∼82 nm detection limit of the TXM experiments imposed by the camera binning and short exposure time used.

A metal layer with a theoretical thickness of 9 μm was plated from zincate electrolyte with no bismuth, as shown in the first panel of Figure 9. This layer was smaller than expected, ∼4 μm, which was attributed to inefficiency due to hydrogen formation. This layer was more planar than the Zn-Bi layer in Figure 8, and did not have a sinusoidal profile. Figure 9 shows deplating at 118 mA/cm2 in zincate electrolyte with no bismuth. The white and blue arrows in Figure 9 emphasize locations on the metal layer profile that became raised and recessed during deplating. Thus deplating encouraged formation of a more irregular profile, rather than planarizing. The TXM images in Figure 9 were collected at a beam energy of 9600 eV, below the

Figure 9. Deplating of a Zn metal layer in zincate electrolyte with no bismuth. Dashed gray line: position of the nickel current electrode. White and blue arrows: Raised and recessed locations on the metal layer during deplating.

Downloaded on 2013-12-21 to IP 174.101.38.169 address. Redistribution subject to ECS license or copyright; see ecsdl.org/site/terms_use

A282

Journal of The Electrochemical Society, 161 (3) A275-A284 (2014)

Figure 10. Metal layers produced by repeated cycling at 35 mA/cm2 in a lateral flow cell with electrolyte flowing from right to left at 1 mL/hr. Top electrode was the WE, bottom electrode was the CE. Time and capacity for all plots was equivalent: 5 hr 22 min after the experiment began. (a) bismuthfree electrolyte. (b) 10% bismuth saturation. (c) 50% bismuth saturation. (d) 100% bismuth saturation. Gray arrow: planarized metal layer. Black arrow: macroscopic buildup of bismuth metal.

K-edge of zinc. This made the metal layer appear more transparent than that in Figure 8. However, when viewed at equivalent beam energies, these layers displayed similar transparency. Macro-morphology of continually cycled zinc layers.— Lateral flow cells were used to monitor the macro-morphology of metal layers continually cycled in zincate electrolyte with varying concentrations of bismuth additive. The lateral flow cells, which were rigid and based on PCBs, had parallel plate electrodes 6 mm in length separated by a 1 mm gap. This approximated the flow channel in a flow-assisted zinc anode cell. Metal layers were continually cycled on the lateral flow cell WE. The plating capacity was greater than the deplating capacity, to encourage layer growth in a manner similar to incomplete discharge observed in flow-assisted nickel-zinc batteries.4 Figure 10 shows metal layers electrodeposited in a PCB-based lateral flow cell with flowing zincate electrolyte containing 0, 3, 13, and 26 ppm bismuth additive. Figure 11 shows the corresponding cell potentials during the first plate/deplate cycles for the layers in Figure 10. Data for these first cycles were representative of later cycles, which were thus excluded from the figure for simplicity. The transition from plating (cathodic WE) to deplating (anodic WE) and back were indicated by dashed lines. The second cycle began at 2747 s, which was the total time of a cycle. The in situ microscope images shown in Figure 10 were at 19,320 s, just after completion of the 7th cycle. This time was chosen as it was the instant that the case with 26 ppm bismuth (100% of saturation) shorted the cell, shown in Figure 10d. In Figure 11 the cell potential during electrodeposition was approximately −2.3 V, defined by the zinc cathode and oxygen anode potentials. The potentials of these two electrodes were previously measured at this current density of 35 mA/cm2 using a reference electrode.32 At 35 mA/cm2 the zinc WE potential was −1.36 V vs. SHE; the oxygen CE potential was 0.9 V vs. SHE. This large oxygen-electrode overpotential was expected and consistent with previous results.32 When the current switched, both anode and cathode were defined by a zinc potential, giving ∼0 V. The chief effect of bismuth in the cell potential response was the potential shelf at −1.23 V during the second cycle, which decreased in duration with increasing bismuth concentration. This was also seen in subsequent cycles, and is attributed to silver oxidation, which was prevented by a blocking layer of bismuth on the CE.

Figure 11. Cell potential response for the first cycle of the experiments shown in Figure 10. Later cycles were similar and are excluded for simplicity. A cycle was defined as: 2107 s of metal plating on the WE at 35 mA/cm2 , a 20 s open circuit rest, 600 s of metal deplating on the WE at 35 mA/cm2 , and another 20 s rest. This sequence defined a cycle, which repeated continually. Current transitions are marked by dotted lines.

As can be seen in Figure 10, the metal layers consisted of gray zinc as expected. When bismuth additive was present a black phase was also apparent. In the cases of 100% and 50% bismuth saturation, buildup of this black material over many cycles resulted in large formations. The black phase was confirmed by ex situ XRD to be bismuth metal (see supplemental). This black material was not rigid and deformed at high electrolyte flowrates. However, it was conductive, and in the 100% bismuth saturation case shorted the cell at 19,320 s (5 hours, 22 minutes). The in situ microdiffraction results suggested a continual bismuth buildup, as during metal layer deplating bismuth was concentrated while zinc dissolved. At a reduced concentration of bismuth additive of 3 ppm (10% of saturation) there was no large macroscopic bismuth buildup, while a significantly more planar metal layer resulted than in the case without bismuth. A thin black layer at the interface in Figure 10b on the order of 10 μm was in agreement with microdiffraction results in Figures 6 and 7. These results showed that the layer was bismuth-rich at the electrolyte interface and bismuth was concentrated at the interface during deplating. The mass transport enhanced dissolution hypothesis suggested by the CV results in Figure 4, and observed by TXM in Figure 8, is consistent with bismuth concentration and exchange plating at the interface. The dense, planar metal layer in Figure 10b showed that at 3 ppm bismuth concentration a mass transport enhanced dissolution mechanism flattened the metal interface. Flatness of the metal layers in Figure 10 was quantified by the standard deviation of the layer thickness averaged along the entire 6 mm length of the WE, calculated by image analysis. The metal layer in Figure 10a had an average thickness of 246 ± 62 μm while that with 10% additive was 204 ± 30 μm, a decrease in standard deviation by half. This could be seen qualitatively by eye, as the bismuthfree metal layer was more diffuse than that with 10% additive. The cells with 50% and 100% additive had standard deviations of 110 and 190 μm respectively. This meant that in the 100% case 1.2 × 10−4 M BiO2 − lead to bismuth buildup and a surface amplitude of approximately 20% of the channel width. The cells in Figures 10a

Downloaded on 2013-12-21 to IP 174.101.38.169 address. Redistribution subject to ECS license or copyright; see ecsdl.org/site/terms_use

Journal of The Electrochemical Society, 161 (3) A275-A284 (2014)

A283

Figure 14. Scanning electron micrographs of a zinc metal layer cycled at 30 mA/cm2 with 10% bismuth additive. Bismuth was interspersed in the mossy zinc structure, similar to that in Figure 13. (a) 1000X, box shows location of B. (b) 6000X.

Figure 12. Metal layers in flowing electrolyte monitored in situ during layer cycling. Cycle times and bismuth additive concentrations of the electrolytes were as indicated. Panel C demonstrated the robust result that bismuth additive at 10% of saturation (3 ppm) resulted in a planarized metal layer.

and 10b maintained similar appearances for 14 cycles. The cell in Figure 10c shorted with the CE in the 11th cycle. Planarization of a cycled zinc layer by 3 ppm bismuth additive was repeatable and robust. Figure 12 compares metal layers cycled at

Figure 13. Scanning electron micrographs of a zinc metal layer cycled at 40 mA/cm2 with 100% bismuth additive. Bismuth was interspersed in the mossy zinc structure, and richer at the electrolyte and current collector interfaces. (a) 1000X, box shows location of B. (b) 10,000X. (c) 2000X. (d) 4000X.

several bismuth concentrations and times. Similar results for these bismuth concentrations were found when the cycling profile was changed, provided the layers underwent both plating and deplating. SEM images of a layer cycled at 40 mA/cm2 with 100% bismuth are shown in Figure 13. EDX confirmed that the smooth domains were bismuth. Mossy zinc with interspersed bismuth can be seen at two magnifications in panels A and B. Bismuth deposits were more concentrated near the metal layer interface with the electrolyte, which was in the upper left of panel A. A bismuth layer at the metal-current collector interface is shown in panel C. Such a bismuth layer was detected by microdiffraction, but was not observed by TXM. As the metal layer plated during the TXM experiments was only ∼5 μm and cycled a single time, this suggests that an interfacial layer of bismuth increases in thickness during cycling. Panel D shows a dendrite tip rich in bismuth. A metal layer cycled with 10% bismuth additive displayed similar microscale characteristics to one cycled at 100%. Figure 14 shows bismuth interspersed through mossy zinc cycled at 30 mA/cm2 with 10% bismuth. Bismuth domains filled a portion of the mossy zinc pore space in this layer. Conclusions In this work we studied cyclic voltammetry of zinc plated from flowing alkaline zincate electrolyte both with and without a bismuth additive. This bismuth was added as Bi2 O3 and had a saturated concentration of 26 ppm bismuth (= 100% of saturation, = 1.2 × 10−4 M BiO2 − ). This small concentration of bismuth caused a marked mass transport effect during metal layer deplating. We termed this mass transport enhanced dissolution. This suggested the additive would planarize metal layers undergoing repeated plating/deplating cycles, as in zinc anodes in flowing-electrolyte batteries. A small, transparent window flow cell was developed to study this mechanism in situ using synchrotron X-rays. X-ray microdiffraction revealed that the metalelectrolyte interface was bismuth rich, and the bismuth signal at this interface intensified during deplating. Transmission X-ray microscopy showed that in the presence of bismuth additive, 5 μm raised features on the metal layer were preferentially dissolved during deplating. Macro-morphology experiments demonstrated that a bismuth additive concentration of 3 ppm (= 10% of saturation, = 1.2 × 10−5 M BiO2 − ) resulted in a planar, compact zinc layer when deposited in a flow channel. Concentrations above 13 ppm (= 50% of saturation, = 6 × 10−5 M BiO2 − ) produced highly irregular layers due to macroscopic bismuth buildup, which lead to cell shorting. These findings suggested that a tightly controlled 3 ppm bismuth additive concentration could be used to planarize zinc metal layers such as those in flow-assisted zinc batteries. However, concentration would need to be well-controlled, as an increase in concentration to 13 ppm would result in bismuth buildup, which caused a cell short in Figure 10c. Above this concentration, corresponding to only 6 × 10−5 M of BiO2 − , bismuth could be viewed as a contaminant rather than a beneficial additive. The surface activity of bismuth on

Downloaded on 2013-12-21 to IP 174.101.38.169 address. Redistribution subject to ECS license or copyright; see ecsdl.org/site/terms_use

A284

Journal of The Electrochemical Society, 161 (3) A275-A284 (2014)

zinc caused it to have a great impact, either positive or negative, at comparatively low concentrations. Acknowledgments This work was supported by the Laboratory Directed Research and Development Program of Brookhaven National Laboratory (LDRDBNL) Under Contract No. DE-AC02-98CH 10866 with the U.S. Department of Energy. Use of the National Synchrotron Light Source, Brookhaven National Laboratory, was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. DE-AC02-98CH10886. Support for DAS and BH in part from NSF CMMI 1031208. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.

F. R. McLarnon and E. J. Cairns, J Electrochem Soc, 138, 645 (1991). S. Smedley, Ieee Aero El Sys Mag, 15, 19 (2000). S. I. Smedley and X. G. Zhang, J Power Sources, 165, 897 (2007). Y. Ito, M. Nyce, R. Plivelich, M. Klein, D. Steingart, and S. Banerjee, J Power Sources, 196, 2340 (2011). C. Ponce-de-Leon, P. K. Leung, C. T. J. Low, and F. C. Walsh, Electrochim Acta, 56, 6536 (2011). P. K. Leung, C. Ponce-de-Leon, C. T. J. Low, A. A. Shah, and F. C. Walsh, J Power Sources, 196, 5174 (2011). T. H. Yang, S. Venkatesan, C. H. Lien, J. L. Chang, and J. M. Zen, Electrochim Acta, 56, 6205 (2011). Z. H. Yang, D. Q. Zeng, S. W. Wang, X. Ni, D. J. Ai, and Q. Q. Zhang, Electrochim Acta, 56, 4075 (2011). M. Ghaemi, R. Amrollahi, F. Ataherian, and M. Z. Kassaee, J Power Sources, 117, 233 (2003). R. D. Naybour, J Electrochem Soc, 116, 520 (1969). D. T. Chin, R. Sethi, and J. Mcbreen, J Electrochem Soc, 129, 2677 (1982). J. McBreen and E. Gannon, J Electrochem Soc, 130, 1980 (1983). J. J. Kelly and A. C. West, J Electrochem Soc, 145, 3472 (1998). J. M. Wang, L. Zhang, C. Zhang, and J. Q. Zhang, J Power Sources, 102, 139 (2001). E. D. Woumfo and O. Vittori, Journal of Applied Electrochemistry, 21, 77 (1991).

16. Y. H. Wen, J. Cheng, L. Zhang, X. Yan, and Y. S. Yang, J Power Sources, 193, 890 (2009). 17. Y. F. Yuan, Y. Li, S. Tao, F. C. Ye, J. L. Yang, S. Y. Guo, and J. P. Tu, Electrochim Acta, 54, 6617 (2009). 18. Y. F. Yuan, J. P. Tu, H. M. Wu, B. Zhang, X. H. Huang, and X. B. Zhao, Electrochem Commun, 8, 653 (2006). 19. J. McBreen, J Electrochem Soc, 119, 1620 (1972). 20. J. McBreen, E. Gannon, and M. G. Chu, J Electrochem Soc, 127, C349 (1980). 21. M. G. Chu, J. Mcbreen, and G. Adzic, J Electrochem Soc, 128, 2281 (1981). 22. J. McBreen, M. G. Chu, and G. Adzic, J Electrochem Soc, 128, 2287 (1981). 23. J. McBreen and E. Gannon, Electrochim Acta, 26, 1439 (1981). 24. J. McBreen and E. Gannon, J Electrochem Soc, 130, C304 (1983). 25. J. McBreen and E. Gannon, J Electrochem Soc, 130, C303 (1983). 26. C. Cachet-Vivier, V. Vivier, A. Regis, G. Sagon, J. Y. Nedelec, and L. T. Yu, Electrochim Acta, 46, 907 (2001). 27. V. Vivier, C. Cachet-Vivier, B. L. Wu, C. S. Cha, J. Y. Nedelec, and L. T. Yu, Electrochem Solid St, 2, 385 (1999). 28. V. Vivier, C. Cachet-Vivier, S. Mezaille, B. L. Wu, C. S. Cha, J. Y. Nedelec, M. Fedoroff, D. Michel, and L. T. Yu, J Electrochem Soc, 147, 4252 (2000). 29. C. G. Castledine and B. E. Conway, Journal of Applied Electrochemistry, 25, 707 (1995). 30. J. W. Gallaway, M. J. Willey, and A. C. West, J Electrochem Soc, 156, D146 (2009). 31. M. J. Willey and A. C. West, Electrochem Solid St, 9, E17 (2006). 32. J. W. Gallaway, D. Desai, A. Gaikwad, C. Corredor, S. Banerjee, and D. Steingart, J Electrochem Soc, 157, A1279 (2010). 33. J. M. Ablett, L. E. Berman, C. C. Kao, G. Rakowsky, and D. Lynch, J Synchrotron Radiat, 11, 129 (2004). 34. J. Wang, Y. C. K. Chen, Q. X. Yuan, A. Tkachuk, C. Erdonmez, B. Hornberger, and M. Feser, Appl Phys Lett, 100, 143107 (2012). 35. Y. Ito, M. Nyce, R. Plivelich, M. Klein, and S. Banerjee, J Power Sources, 196, 6583 (2011). 36. T. P. Dirkse, J Electrochem Soc, 126, 1456 (1979). 37. M. Pourbaix, Atlas of electrochemical equilibria in aqueous solutions, Pergamon Press Ltd., Oxford (1966). 38. A. R. Despic, D. Jovanovic, and T. Rakic, Electrochim Acta, 21, 63 (1976). 39. C. Cachet, B. Saidani, and R. Wiart, J Electrochem Soc, 138, 678 (1991). 40. C. Cachet, B. Saidani, and R. Wiart, J Electrochem Soc, 139, 644 (1992). 41. K. M. Yin and R. E. White, Aiche J, 36, 187 (1990). 42. M. W. Verbrugge and C. W. Tobias, J Electrochem Soc, 132, 1298 (1985). 43. S. Roy and D. Landolt, J Electrochem Soc, 142, 3021 (1995). 44. S. Roy, M. Matlosz, and D. Landolt, J Electrochem Soc, 141, 1509 (1994).

Downloaded on 2013-12-21 to IP 174.101.38.169 address. Redistribution subject to ECS license or copyright; see ecsdl.org/site/terms_use