- Email: [email protected]
Slurry electrodes for iron plating in an all-iron flow battery
Journal of Power Sources 294 (2015) 620e626
Contents lists available at ScienceDirect
Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour
Slurry electrodes for iron plating in an all-iron flow battery Tyler J. Petek*, Nathaniel C. Hoyt, Robert F. Savinell, Jesse S. Wainright Department of Chemical Engineering, Case Western Reserve University, AW Smith 116, 10900 Euclid Avenue, Cleveland, OH 44106, USA
h i g h l i g h t s Slurries are investigated as negative electrodes in all-iron flow batteries. Current distribution is modeled as a function of area and electrical conductivity. MWCNTs slurries contained >95% of battery charge (as opposed to current collector). Slurry electrodes can be effectively used to decouple all-iron flow batteries.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 10 March 2015 Received in revised form 23 May 2015 Accepted 9 June 2015 Available online xxx
Slurry electrodes are investigated in order to decouple the energy storage capacity from the power delivery capability in an all-iron flow battery. For the slurry electrode to perform effectively, the battery negative reaction must occur on the slurry particles at reasonably high current densities. Mathematical modeling is used to investigate the current distribution in a slurry electrode as a function of the slurry specific area and electrical conductivity in order to achieve >95% plating in the slurry electrode (not on the flat plate) at > 200 mA cm2. From the mathematical modeling, MWCNTs are selected to demonstrate slurry electrode performance. The experimental performance of all-iron batteries charged using the MWCNT slurry electrodes is found to improve while increasing the battery state-of-charge. Two possible mechanisms contributing to this effect are an increase in electronic conductivity of the slurry and an increase in plating kinetics. After cycling the battery, <5% of the total battery charge was observed to have plated onto the current collector. © 2015 Elsevier B.V. All rights reserved.
Keywords: Slurry electrodes Current distribution Analytical model Flow battery Electroplating
1. Introduction Redox flow batteries (RFBs) have seen a surge of interest over recent years for their ability to provide ancillary services to the electric grid which could lead to increased robustness, flexibility, and the ability to rely on intermittent renewable energy sources such as wind and solar [1e5]. Traditional RFBs, including the ironechromium [6], ironevanadium [7,8], bromide-polysulfide [9], and all-vanadium [10] systems, store energy chemically in electroactive species contained in electrolytes that are stored in external reservoirs. During operation, the electrolyte is pumped through the RFB where the chemical energy is directly converted to electrical energy. The architecture of RFBs allows the energy storage capacity (dependent on the amount of active species) to be scaled independently of the power rating (dependent on the area of the * Corresponding author. E-mail addresses: [email protected], [email protected] (T.J. Petek), [email protected] (N.C. Hoyt), [email protected] (R.F. Savinell), jesse. [email protected] (J.S. Wainright). http://dx.doi.org/10.1016/j.jpowsour.2015.06.050 0378-7753/© 2015 Elsevier B.V. All rights reserved.
RFB conversion device) leading to flexible operation and implementation. Despite the advantages, the cost [11], abundance, and toxicity of the active species have hindered the market penetration of these systems. An attractive alternative to the traditional RFB chemistries is the all-iron flow battery [12] because iron, the only active element, is low cost, abundant, and environmentally benign. Equations (1) and (2) describe the positive and negative reactions in the all-iron flow battery, respectively. On charge, ferrous ions (Fe2þ) are oxidized to ferric ions (Fe3þ) in the positive half-cell while ferrous ions are reduced to iron metal in the negative half-cell. The all-iron battery has a standard cell voltage of 1.21 V.
2Fe2þ #2Fe3þ þ 2e Fe2þ þ 2e #Fe0
Eo ¼ þ0:77V vs: RHE
Eo ¼ 0:44V vs: RHE
(1) (2)
RFBs typically employ electrodes made of porous carbon structures (e.g. felts, cloths, or papers) fixed in the electrochemical
T.J. Petek et al. / Journal of Power Sources 294 (2015) 620e626
device. When using these stationary electrodes in the all-iron flow battery, iron metal is plated into their structure and subsequently stored in the electrochemical cell. For a majority of grid applications, it is expected that an energy to power ratio of 3e8 h is required [2]. Because iron metal is stored in the electrochemical cell, the conventional all-iron battery is limited to <4 h of energy storage at reasonably high current densities [13]. Existing capital cost models suggest that current densities above 200 mA cm2 will be necessary to achieve acceptable stack costs for large scale adoption of flow batteries [11,14,15]. In order to have a battery with the advantages inherent to traditional flow battery architectures and the advantages of the all-iron chemistry, a different electrode structure must be employed that enables higher energy storage to power ratios. One such option is the slurry electrode. Slurry electrodes are made by suspending solid, electronically conductive particles in the ionically conducting electrolyte and have been studied for a variety of electrochemical applications [17e23]. During charge and discharge, the electrolyte is pumped through the electrochemical cell as in normal RFB operation. However, when using slurry electrodes, the solid particles are carried by the electrolyte into and out of the electrochemical cell. If the volume fraction of the solid particles is high enough, a continuous electrically conductive network forms [24,25]. This allows the redox reaction to occur on the surface of the slurry particles. When slurries are employed as the negative electrode of the all-iron flow battery, iron metal is plated onto the slurry particles. As the slurry is circulated out of the electrochemical cell back into the external reservoir, the iron metal is carried with the particles, enabling the energy storage capacity to be decoupled from the power rating. Additional advantages of slurry electrodes include their ability to have high surface areas, simple manufacturing/assembly, and ease of maintenance/recycling through filtering. Previous studies have investigated the ferrous/ferric redox couple on slurry electrodes [23]. This paper reports results of an investigation of the performance of slurry electrodes in the full all-iron flow battery while focusing on the plating of iron metal into the negative slurry electrode. 2. Analytical current distribution models Mathematical modeling of the slurry electrode in the negative half-cell was performed in order to permit the selection of viable candidate slurries. The selection of an appropriate particle to use as the flowable slurry electrode was performed with regards to two principal performance metrics. The first performance metric was related to the achievable current density. Cost models have shown that for the all-iron slurry battery to be practical that the battery needs to be able to support a current density of 200 mA cm2 at a voltaic efficiency of at least 70% [11,14,15]. Given the 1.2 V cell potential of the all-iron battery, this means that the desired current density must be achieved with at most 212 mV of total overpotential across the entire cell. The second performance metric is related to the amount of plating that occurs onto the current collector. If plating occurs onto the current collector and not onto the slurry particles, then the energy density and power density of the battery are no longer decoupled. As decoupling is a desirable trait for grid-scale energy storage, a criterion was established demanding that less than 5% of the total metal plating should occur onto the current collector. The slurry characteristics necessary to match the performance metrics were examined using standard one-dimensional macrohomogeneous porous electrode equations [26,27]. The governing equations are identical to those used by Fleischmann and Oldfield for fluidized bed electrodes [28]. The kinetics of the plating reaction were assumed to be linear, resulting in the following equations:
621
s
d2 F1 a0 zfio h ¼ 0 dx2
(3a)
k
d2 F2 þ a0 zfio h ¼ 0 dx2
(3b)
Here, s is the electronic conductivity of the slurry phase, k is the ionic conductivity of the solution phase, F1 is the potential of the slurry phase, F2 is the potential of the solution phase, a0 is the specific area of the slurry electrode, z is the number of electrons involved in the plating reaction, f¼FR1 T1 is related to Faraday's constant, i0 is the exchange current density, x is the spatial coordinate across the channel gap, h¼F1F2Uref is the overpotential, and Uref is the open-circuit potential of the electrode reaction relative to a suitable reference electrode such that the overpotential is zero at equilibrium. Effects such as mass transfer overpotentials or advectively-induced capacitive currents are ignored by this model, but it is believed to capture the dominant behavior associated with slurry electrodes in the presence of fast faradaic reactions. In order to account for the fact that reactions can occur on both the current collector plate and the slurry particles, the standard boundary conditions used by Fleischmann and Oldfield needed to be altered. The modified boundary conditions at x ¼ 0 (the current collector) and x ¼ d (the location of the separator) take the form:
dF1 ði xzfio ho Þ dF2 xzfio ho ; ¼ d ¼ s dx dx k dF1 dF2 id ¼ 0; ¼ dx dx k
at x ¼ 0
at x ¼ d
(4a)
(4b)
Here, id is the applied current density, x is the roughness factor of the current collector, and h0 is the overpotential at x ¼ 0. The total overpotential across the half-cell can be described as htot¼F1(0) F2(d)Uref. An additional factor accounting for the area of the current collector that is occluded by the presence of the slurry could also be included in the boundary conditions, but for low volume fraction slurries (like those in this paper), this effect is believed to be negligible. The slurry properties that affect the achievement of the above performance metrics are the specific interfacial area and the electronic conductivity. Using typical values for cell dimensions, iron plating reaction kinetics, and electrolyte properties (see Table 1), Equations (3a) and (3b) were solved in order to generate maps of the combinations of s and a0 that satisfy the metrics. These maps are shown in Fig. 1. Regions corresponding to the polarization metric (<100 mV of overpotential in the negative half cell at 200 mA cm2 or higher) and to the current collector plating metric are both included. Any viable slurry must lie within the bounds of both regions. As shown in Fig. 1, slurries made with MWCNTs are predicted to be able to achieve both performance metrics [23]. It is likely that other less expensive particles can also match the criteria, but, as a proof of concept, the MWCNT slurry was adopted to permit the
Table 1 Parameters used for slurry electrode model. Parameter
Value
d
0.1 cm 2 mole e 2þ mole Fe 10 cm2 cm2 1 mA cm2 200 mS cm1 100 mV
z
x i0
k htot
622
T.J. Petek et al. / Journal of Power Sources 294 (2015) 620e626
polycarbonate tube (3/16 in. ID) through which the slurry was run at various flow rates (ranging from 20 mL/min to 130 mL/min). Differential pressure transducers located 100 cm apart along the length of the tube were used to determine the pressure drop. The reference side of each pressure sensor was open to the atmosphere. During each test, the particles were continually mixed in the reservoir to ensure stable results. The length and diameter of the polycarbonate tube were carefully selected in order to achieve a similar range of shear rates to those that are encountered in the electrochemical flow cell (while still maintaining laminar flow within the tube). The tube viscometer was used in order to avoid the settling and shear-history issues that often encountered with, e.g., cone-and-plate rheometers. The Ostwaldede Waele (power law) model was adopted to describe the non-Newtonian behavior of the slurry over the range of shear rates considered, as has been done in the work of others [31]. This model describes the shear stress in terms of the shear rate, vu/vy, a flow consistency index, K, and a flow behavior index, n, as related through Equation (5). Fig. 1. Slurry electrode performance maps. The red shaded region represents the combinations of s and a0 that satisfy the requirement that the slurry electrode be able to support a 200 mA cm2 current density for 100 mV total overpotential across the negative half-cell. The blue shaded region represents the combinations of s and a0 that satisfy the requirement that the slurry electrode be able to plate less than 5% of the total metal onto the current collector. The purple region denotes satisfaction of both metrics. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
examination of the slurry battery performance. 3. Experimental 3.1. Materials The slurry electrodes in this study were made with MWCNTs purchased from Nanostructured and Amorphous Materials, Inc. (Houston, TX, USA). These MWCNTs have an OD of 50e80 nm, length of 10e20 mm (aspect ratio of 125e400), a reported surface area of 40 m2/g, and a carbon purity of 99.9%. For these studies, the MWCNTs were used as received. Before tests were performed, the particles were mixed by hand into the electrolyte and then pumped through the system for at least 30 min. All-iron flow batteries were studied with slurries serving as both the positive and the negative electrodes. In all cases presented herein, the slurry electrodes were composed of 4.8 vol% MWCNTs in the electrolyte. The electronic conductivity of these slurries was previously found to be about 85 mS cm1 while flowing [23]. The positive slurry was initially 0.5 M FeCl2, 0.5 M FeCl3, and 1.0 M NH4Cl while the negative slurry was initially 0.5 M FeCl2 and 1.0 M NH4Cl. The ionic conductivity of these electrolytes is about 150 mS/ cm. Adding ferric chloride to the positive electrolyte ensured that on discharge, the negative slurry would always be limiting. All salts used were reagent grade and were obtained from Fisher Scientific (USA). All solutions were made with deionized water (>18 MU-cm, Barnstead Milli-Q). When using ferrous salts, solutions were made under a continuous nitrogen blanket and experiments were performed with the electrolyte under a continuous nitrogen purge to mitigate the air oxidation of ferrous to ferric iron [29,30]. All tests were conducted at room temperature (20e25 C). 3.2. Rheological studies The slurries in these studies are intended for use as flowing electrodes. As such, the viscosity of the slurry electrodes are of interest and were investigated with an in-house tube viscometer. The tube viscometer consisted of a vertically-oriented clear
t¼K
n vu vy
(5)
The values for K and n can be determined from tube viscometry data through comparison to an analytical prediction of the pressure drop. The equation for the analytically predicted pressure drop of an Ostwaldede Waele fluid through a circular cylindrical tube is:
h n iDp1=n Q ¼p r ð3nþ1Þ=n 3n þ 1 2KL
(6)
Here, Q is the flow rate, Dp is the pressure drop, L is the length of the tube, and r is the tube radius [32]. Table 2 shows the values for K and n that were determined for various loadings of multi-walled carbon nanotubes in a 1 M solution of ferric chloride. Similar to the observations in previous studies [19e21], the slurries did exhibit shear thinning behavior with smaller flow behavior indices as the volume loading was increased. For flow in the electrochemical cell described below at the flow rates used throughout this experimental work (250 mL/min), the shear rate at the wall for the slurry with the highest volume loading (4.3 vol%) in Table 1 is 3750 s1. The effective viscosity at the wall for this slurry can therefore be calculated to be 10.54 cP. As this viscosity value is only roughly five times that of the bare electrolyte, the pumping costs for the flowable slurry electrode should not be prohibitively large. In addition, the additional pressure drop that arises in standard redox flow batteries from the presence of porous electrodes or serpentine/interdigitated flow fields is absent in slurry flow batteries. However, future work must be done to quantify the parasitic losses due to pumping more fully. It was observed that when the volume fraction of the solid particles was above some critical concentration, there was no longer enough liquid in the slurry to fully wet all of the particles. When the volume fraction of solids rose above this critical volume fraction, the slurry transitioned, quite quickly, from a shear
Table 2 Flow consistency index, K, and flow behavior index, n, for various weight percentages of MWCNTs in 1 M ferric chloride. Vol%
n
K (cP sn1)
0.5 1.9 2.8 4.3
0.99 0.64 0.49 0.40
1.92 59.6 351 1470
T.J. Petek et al. / Journal of Power Sources 294 (2015) 620e626
623
thinning fluid to a semi-solid gel. At this point, the slurry is no longer able to flow. This viscosity critical volume fraction is described in detail by Youssry et al. [21] For slurries made with the MWCNTs used in these studies, the viscosity critical volume fraction was found to be about 6 vol%. 3.3. Electrochemical studies The all-iron flow battery investigations were conducted in an inhouse channel cell through which the slurry electrodes flow during operation. This channel cell was defined by two CPVC, chlorinated polyvinyl chloride, flow fields containing straight rectangular channels with a width of 1 cm and length of 12 cm (in the direction of flow). The current collectors were 1 cm wide and 6.725 cm long (A ¼ 6.725 cm2) graphite plates that were centered in the middle of the CPVC channel. The gap between the current collector of each channel and the separator, d, was 0.2 cm. During operation, a peristaltic pump (ColeeParmer, USA) was used for pumping the slurry electrodes. In all experiments, the slurry was pumped vertically up through the electrochemical channel cell at a volumetric flow rate of 250 mL/min. The inlet and exit ports of the channel cell were at roughly 30 angles (with respect to the current collector face) to mitigate clogging. Current was collected from this cell through two brass plates that were pressed into the back of the graphite current collectors. The cell was held together with stainless steel endplates. Teflon sheets were used to insulate the endplates and Grafoil (Graftech, USA) was used to minimize contact resistances between the brass plates and the graphite plates. The cell assembly was compressed with 8 bolts that were tightened to 90 in-lbs of torque. A schematic of the cell (not to scale) is shown in Fig. 2. The reservoirs used to hold the slurry that was pumped through the channel cell were Nalgene bottles modified to be bottom drawn. Unless otherwise noted, all cell materials were purchased from McMaster-Carr (Cleveland, OH, USA). Daramic, LLC (Charlotte, NC) graciously donated Daramic 250 Flatsheet for use in these studies. Daramic 250 Flatsheet has a reported average pore size of 0.1 microns, average porosity of 58±7.5%, and a thickness of about 225 microns measured after use in experiments. The Daramic was soaked in methanol and rinsed in DI water before use so that air bubbles trapped inside the pores were removed; ensuring that the separator be completely wetted [8]. The ionic conductivity of these separators, kDaramic, has been estimated to be 40% of the conductivity of the electrolyte using the reported average porosity, e, of 58% and the Bruggeman approximation, Equation (7) [16].
kDaramic ¼ eð3=2Þ
Fig. 2. A schematic of the electrochemical channel cell (not to scale) used to characterize the slurry performance. The channel is defined by the flow frame and the electric field is defined by the inset current collectors. The slurry flows up vertically through the cell and the inlet/exit are at approximately 30 to vertical to help mitigate clogging. Not shown are the materials to compress the cell and the brass plates from which current is drawn.
(7)
All electrochemical studies were performed using a Solartron 1280B potentiostat. Electrochemical impedance spectroscopy (EIS) was performed at varying cell potentials with a sinusoidal perturbation of 10 mV amplitude over the frequency range 20 kHze0.2 Hz. The EIS experiments were controlled using ZPlot (Scribner Associates, USA). Cell polarization scans and an oscilloscope (Tektronix, USA) were used in congress to make sure that the response of the system under perturbation remained linear. Cyclic voltammetry with a scan rate of 20 mV/s was used to approximate the polarization of the all-iron flow battery and was controlled using CorrWare (Scribner). 4. Results and discussion Before the initial charging of the all-iron flow battery, EIS measurements were first conducted at the open circuit potential, shown in Fig. 3. The high frequency resistance of the battery should
Fig. 3. The Nyquist representation of the EIS response of the all-iron flow battery at different states of charge.
be the sum of the parallel ionic and electronic resistances of each slurry electrode along with the separator resistance [23]. For a battery with 2 mm channels on either side of Daramic 250 Flatsheet, the total battery high frequency resistance should be about 2.0 U cm2 based on the independent electrical and ionic conductivity measurements. As seen in Fig. 3, the initial EIS measurements are in good agreement with this prediction; confirming the conductivities of the slurry electrodes and separator. The EIS results at subsequent states-of-charge (SOC) will be discussed below. Fig. 4 shows the typical behavior of an all-iron flow battery while charging. The initial battery charge cycle was performed at
624
T.J. Petek et al. / Journal of Power Sources 294 (2015) 620e626
Fig. 4. The voltage profiles while charging an all-iron flow battery at constant current with both positive and negative slurry electrodes. The discontinuity in the 75 mA cm2 curve occurred when the tubing between the cell and reservoir was readjusted.
75 mA/cm2. The battery voltage rose with the SOC before leveling off around 1.5 V. At the end of the initial charge, just over 1.0 Ah of charge had been stored in the battery bringing the 500 mL negative slurry to just over 7% state-of-charge (SOC). At this point, a cyclic voltammogram was performed on the battery and the results are shown in Fig. 5. As the voltammogram shows, the open circuit potential of the all-iron flow battery was about 1.2 V. From the open circuit potential, the cell voltage was scanned first positive 500 mV and then to 500 mV below the open circuit potential before returning to the open circuit potential at a rate of 20 mV s1. The data shown is the second of three cycles performed. Each cycle showed good reproducibility compared to the previous. In these experiments, the positive electrode is the working electrode and the negative electrode is the counter and reference. Positive currents represent oxidation at the working electrode, i.e., charging the battery.
Fig. 5. The cyclic voltammetry of the all-iron all-slurry flow battery at 7% state-ofcharge after charging at 75 mA cm2. The voltammogram is scanned at 20 mV/s first positive from the open circuit potential and negative to 500 mV vs the OCV before scanning back to the OCV. The data shown in this figure is the second of three cycles; each showing good reproducibility.
After the voltammogram was obtained the charge was continued at 150 mA cm2 for 2 additional hours (constituting a total battery charge of just over 3.0 Ah). During this period, the cell charging voltage fell to 1.45 V after first going through a voltage maximum. This decrease in cell overpotentials at increased current density operation was corroborated by the EIS response of the battery, Fig. 3, decreasing as the SOC of the battery increases. EIS was performed at a discharge potential so that the results can be more easily compared at different SOCs. During discharge the dissolution of iron metal is directly studied and the effects of the variation of ferrous iron should be mitigated. Also, any additional side reactions that may occur during charging, discussed below, will not occur. There are two main mechanisms thought to contribute to the improved performance. As the SOC of the negative slurry rises, the amount of metal deposited in the slurry increases and may therefore increase the slurry electronic conductivity. Also, after the carbon nanotubes are seeded with iron metal, the kinetics of iron plating are enhanced because the kinetic hindrance associated with iron nucleation on carbon is no longer a factor [33]. The decrease in the overpotential on charge (whether by increased conductivity and/or improved kinetics) has an additional benefit. When charging at large potentials, the negative electrode can become negative enough for hydrogen evolution to occur [33]. When this happens, the electrolyte pH will rise, potentially causing ferric or ferrous iron to precipitate out of solution as their respective hydroxides [33]. The precipitated iron hydroxide can form a film on the slurry particles making discharge of the battery more difficult. Fig. 6 shows the voltage profiles while cycling the all-iron flow battery at ±75 mA cm2. Each half-cycle is 0.5 h; constituting swings of ±0.25 Ah. The average charging potential is 1.55 V while the average discharge potential is 0.8 V. The average voltaic efficiency is just above 50%. The voltaic efficiency could be improved by optimizing the separator and cell design. Over the twelve cycles shown, no appreciable capacity fade was observed (each half cycle lasted the 0.5 h) and the potential profiles were steady. These cycles were performed after the two initial charging profiles shown in Fig. 4 (constituting ±0.25 Ah cycles around 3.0 Ah). After cycling was concluded, the battery was disassembled at the end of a charge. A thin film (<0.02 cm) of iron was visible
Fig. 6. Voltage profiles while cycling the all-iron flow battery with MWCNT slurry electrodes at ±75 mA cm2. Each cycle is ±0.25 Ah and was performed after the initial charging profiles shown in Fig. 4.
T.J. Petek et al. / Journal of Power Sources 294 (2015) 620e626
on < 25% of the current collector surface area. If all 3.0 Ah put into the battery was plated onto the current collector, the resulting metal plate would have to be approximately 0.07 cm, or 33% of the channel thickness. The iron that had plated onto the current collector was dissolved into a known volume of HCl and the resulting concentration was determined with a platinum microelectrode [34]. From the dissolved iron concentration and the volume of HCl, the amount of iron deposited on the current collector was found to be <5% of the total charge put into the battery, confirming that most of the metal was plated onto the slurry particles. The cycles performed in Fig. 6 are relatively shallow and at low current densities in order to mitigate the effects of hydrogen evolution. Despite this, after the battery charging and cycling experiments were concluded, Figs. 4 and 6, the electrolyte pH was observed to have risen from its initial value of 2 (due to the mildly acidic nature of ferrous ions) to a pH of about 4; presumably due to hydrogen evolution. This shift in pH, if solely due to hydrogen evolution, would account for <0.25 Ah total over the >16 h of operation and 3.25 Ah of charge deposited. It is likely that the hydrogen evolution occurs with a similar current distribution as discussed in Section 2 above. As such, the majority of the hydrogen evolution is anticipated to occur on the slurry particles and should not have a significant effect on the distribution of iron in the system. Future work is required to gain a better understanding of how to mitigate the effects of hydrogen evolution and to better characterize the coulombic efficiency. While the cycles performed in these studies were relatively shallow, for a slurry to be truly viable for the all-iron battery it must maintain physical and electrochemical properties over wide SOC swings. Fig. 7 shows the expected change in the slurry particle physical properties if all of the iron in a 2 M ferrous chloride negative electrolyte was plated (100% SOC). As iron metal is plated on the particles, the particle density will increase. However, as the amount of particles in the slurry increase, the amount of iron plated onto each particle (assuming uniform distribution) decreases. This analysis suggests that particles capable of forming slurries with acceptable flowability at higher volume fractions are required in order to mitigate the effects of charging at high SOC. Also, achieving higher particle loading has an additional benefit of increased
specific area, described in Fig. 1. 5. Conclusion Slurry electrodes were investigated in order to decouple the energy storage capacity from the power delivery rate in an all-iron flow battery. For the slurry electrode to perform effectively, the battery negative reaction must occur on the slurry particles at reasonably high current densities. Mathematical modeling was used to investigate the current distribution in a slurry electrode as a function of the slurry specific area and electrical conductivity in order to achieve >95% plating in the slurry electrode (not on the flat plate) at > 200 mA cm2. From the mathematical modeling, MWCNTs were selected as one viable slurry electrode and were used in these studies to prove the feasibility of a slurry for the alliron battery. All-iron flow batteries were operated with MWCNT slurries as both the positive and negative electrodes. As the battery state-ofcharge increased, the voltaic efficiency of the battery also increased. Two mechanisms thought to contribute to the improved performance are an increase in the slurry electronic conductivity (due to metal deposition on the particles) and an increase in plating kinetics. After cycling the battery around 25% state-of-charge, <5% of the charge of the battery was observed as iron metal on the current collector. While the MWCNT slurry electrodes discussed in this work were shown to operate with current distributions that effectively decouple the system, future work must still be accomplished for the all-iron battery to be a truly viable energy storage device. The slurry electrode must be developed in parallel with improved electrolyte chemistries so that a more cost effective slurry can be utilized (and still meet the criteria for viable current distribution) that can support higher current densities and high states-of-charge with respect to iron deposition with acceptably high coulombic efficiencies. Acknowledgments The authors would like to thank Nicholas Sinclair, Jason Pickering, Clay Kacica, Ashley Yarus, Bryan Erb, Joseph Plazek, Matthew Madler, and Mirko Antloga for assistance in conducting experimental trials and maintaining laboratory equipment. The funding for this project is through ARPA-E, contract number DEAR0000352. Nomenclature a0 F id io K L n p Q R r T Uref vu/vy
Fig. 7. The physical properties of the slurry particles at 100% SOC plating density relative to properties at 0% SOC (assuming uniform distribution) when in 2 M negative electrolyte of an all-iron battery plotted versus the particle volume fraction. Each property is normalized by that of pure carbon.
625
F1 F2 z
d
Slurry specific area, [m2 m3] 1 Faraday's constant, ½96; 485 A s mole e 2 applied current density, [A cm ] exchange current density, [A cm2] Flow consistency index, [cP sn1] length of viscometer tube, [cm] flow behavior index pressure, [cP s1] flow rate, [cm3 s1] gas constant, [8.314 A V s K1 mol1] inner radius of tube viscometer, [cm] temperature, [K] open-circuit potential of electrode reaction, [V] shear rate, [s1] solid phase potential, [V] solution phase potential, [V] electron equivalence, [mole e/mole Fe] Electrode channel gap, [cm]
626
h k
x s t
e
T.J. Petek et al. / Journal of Power Sources 294 (2015) 620e626
overpotential, [V] Solution phase ionic conductivity, [S cm1] current collector roughness factor Solid phase electronic conductivity, [S cm1] shear stress, [cP s1] void fraction
References n, A. Frías-Ferrer, J. Gonza lez-García, D.A. Sza nto, F.C. Walsh, [1] C. Ponce de Leo J. Power Sources 160 (2006) 716. [2] J. Eyer, G. Corey, Energy Storage for the Electricity Grid: Benefits and Market Potential Assessment Guide, Sandia Report SAND2010e0815, 2010. [3] T. Nguyen, R.F. Savinell, ECS Interface 19 (2010) 54. [4] M. Skyllas-Kazacos, M.H. Chakrabarti, S.A. Hajimolana, F.S. Mjalli, M. Saleem, J. Electrochem. Soc. 158 (2011) R55. [5] A.Z. Weber, M.M. Mench, J.P. Meyers, P.N. Ross, J.T. Gostick, Q. Liu, J. Appl. Electrochem 41 (2011) 1137. [6] M. Lopez-Atalaya, G. Codina, J.R. Perez, J.L. Vazquez, A. Aldaz, J. Power Sources 39 (1992) 147. [7] W. Wang, S. Kim, B. Chen, Z. Nie, J. Zhang, G.-G. Xia, L. Li, Z. Yang, Energy Environ. Sci. 4 (2011) 4068. [8] X. Wei, L. Li, Q. Luo, Z. Nie, W. Wang, B. Li, G.-G. Xia, E. Miller, J. Chambers, Z. Yang, J. Power Sources 218 (2012) 39. [9] A. Price, S. Bartley, S. Male, G. Cooley, Power Eng. J. 13 (1999) 122. [10] M. Rychcik, M. Skyllas-Kazacos, J. Power Sources 22 (1988) 59. [11] V. Viswanathan, A. Crawford, D. Stephenson, S. Kim, W. Wang, B. Li, G. Coffey, E. Thomsen, G. Graff, P. Balducci, M. Kintner-Meyer, V. Sprenkle, J. Power Sources 247 (2014) 1040. [12] L.W. Hruska, R.F. Savinell, J. Electrochem. Soc. 128 (1981) 18. [13] K.L. Hawthorne, J.S. Wainright, R.F. Savinell, J. Power Sources 269 (2014) 216.
[14] D.P. Scamman, G.W. Reade, E.P.L. Roberts, J. Power Sources 189 (2009) 1231. [15] Vanadium redox flow batteries: an in-depth analysis, EPRI, Palo Alto, CA, 2007, p. 1014836. [16] S. Banisi, J.A. Finch, A.R. Laplante, Minerals Engineering 6 (1993) 269. [17] D.N. Baria, H.M. Hulburt, J. Electrochem. Soc. 120 (1973) 1333. [18] A.J. Appleby, M. Jacquier, J. Power Sources 1 (1976) 17. [19] M. Duduta, B. Ho, V.C. Wood, P. Limthongkul, V.E. Brunini, W.C. Carter, Y.M. Chiang, Adv. Energy Mater. 1 (2011) 511. [20] V. Presser, C.R. Dennison, J. Campos, K.W. Knehr, E.C. Kumbur, Y. Gogotsi, Adv. Energy Mater. 2 (2012) 895. [21] M. Youssry, L. Madec, P. Soudan, M. Cerbelaud, D. Guyomard, B. Lestriez, Phys. Chem. Chem. Phys. 15 (2013) 14476. [22] S. Jeon, H. Park, J. Yeo, S.C. Yang, C.H. Cho, M.H. Han, D.K. Kim, Energy Environ. Sci. 6 (2013) 1471. [23] T.J. Petek, N.C. Hoyt, J.S. Wainright, R.F. Savinell, J. Electrochem. Soc. 163 (2016) A5001. [24] C.G. Granqvist, O. Hunderi, Phys. Rev. B 18 (1978) 1554. [25] L. Gao, X. Zhou, Y. Ding, Chem. Phys. Lett. 434 (2007) 297. [26] J. Newman, C. Tobias, J. Electrochem. Soc. 109 (1962) 1183e1191. [27] J. Newman, W. Tiedemann, AIChe J. 21 (1975) 1. [28] M. Fleischmann, J.W. Oldfield, J. Electroanal. Chem. Interfacial Electrochem 29 (1971) 211. [29] P.C. Singer, W. Strumm, Sci. (N. Y.) 167 (1970) 1121. [30] N.F. Gray, Environ. Geol. 30 (1997) 62. [31] J.W. Campos, M. Beidaghi, K.B. Hatzell, C.R. Dennison, B. Musci, V. Presser, E.C. Kumbur, Y. Gogotsi, Electrochim. Acta 98 (2013) 123. [32] R.P. Chhabra, J.F. Richardson, Non-newtonian Flow in the Process Industries: Fundamentals and Applications, Butterworth-Heinemann, Oxford England, 1999. [33] K.L. Hawthorne, T.J. Petek, M.A. Miller, J.S. Wainright, R.F. Savinell, J. Electrochem. Soc. 162 (2015) A108. [34] R.A. Malmsten, C.P. Smith, H.S. White, J. Electroanal. Chem. 215 (1986) 223.
Contents lists available at ScienceDirect
Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour
Slurry electrodes for iron plating in an all-iron flow battery Tyler J. Petek*, Nathaniel C. Hoyt, Robert F. Savinell, Jesse S. Wainright Department of Chemical Engineering, Case Western Reserve University, AW Smith 116, 10900 Euclid Avenue, Cleveland, OH 44106, USA
h i g h l i g h t s Slurries are investigated as negative electrodes in all-iron flow batteries. Current distribution is modeled as a function of area and electrical conductivity. MWCNTs slurries contained >95% of battery charge (as opposed to current collector). Slurry electrodes can be effectively used to decouple all-iron flow batteries.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 10 March 2015 Received in revised form 23 May 2015 Accepted 9 June 2015 Available online xxx
Slurry electrodes are investigated in order to decouple the energy storage capacity from the power delivery capability in an all-iron flow battery. For the slurry electrode to perform effectively, the battery negative reaction must occur on the slurry particles at reasonably high current densities. Mathematical modeling is used to investigate the current distribution in a slurry electrode as a function of the slurry specific area and electrical conductivity in order to achieve >95% plating in the slurry electrode (not on the flat plate) at > 200 mA cm2. From the mathematical modeling, MWCNTs are selected to demonstrate slurry electrode performance. The experimental performance of all-iron batteries charged using the MWCNT slurry electrodes is found to improve while increasing the battery state-of-charge. Two possible mechanisms contributing to this effect are an increase in electronic conductivity of the slurry and an increase in plating kinetics. After cycling the battery, <5% of the total battery charge was observed to have plated onto the current collector. © 2015 Elsevier B.V. All rights reserved.
Keywords: Slurry electrodes Current distribution Analytical model Flow battery Electroplating
1. Introduction Redox flow batteries (RFBs) have seen a surge of interest over recent years for their ability to provide ancillary services to the electric grid which could lead to increased robustness, flexibility, and the ability to rely on intermittent renewable energy sources such as wind and solar [1e5]. Traditional RFBs, including the ironechromium [6], ironevanadium [7,8], bromide-polysulfide [9], and all-vanadium [10] systems, store energy chemically in electroactive species contained in electrolytes that are stored in external reservoirs. During operation, the electrolyte is pumped through the RFB where the chemical energy is directly converted to electrical energy. The architecture of RFBs allows the energy storage capacity (dependent on the amount of active species) to be scaled independently of the power rating (dependent on the area of the * Corresponding author. E-mail addresses: [email protected], [email protected] (T.J. Petek), [email protected] (N.C. Hoyt), [email protected] (R.F. Savinell), jesse. [email protected] (J.S. Wainright). http://dx.doi.org/10.1016/j.jpowsour.2015.06.050 0378-7753/© 2015 Elsevier B.V. All rights reserved.
RFB conversion device) leading to flexible operation and implementation. Despite the advantages, the cost [11], abundance, and toxicity of the active species have hindered the market penetration of these systems. An attractive alternative to the traditional RFB chemistries is the all-iron flow battery [12] because iron, the only active element, is low cost, abundant, and environmentally benign. Equations (1) and (2) describe the positive and negative reactions in the all-iron flow battery, respectively. On charge, ferrous ions (Fe2þ) are oxidized to ferric ions (Fe3þ) in the positive half-cell while ferrous ions are reduced to iron metal in the negative half-cell. The all-iron battery has a standard cell voltage of 1.21 V.
2Fe2þ #2Fe3þ þ 2e Fe2þ þ 2e #Fe0
Eo ¼ þ0:77V vs: RHE
Eo ¼ 0:44V vs: RHE
(1) (2)
RFBs typically employ electrodes made of porous carbon structures (e.g. felts, cloths, or papers) fixed in the electrochemical
T.J. Petek et al. / Journal of Power Sources 294 (2015) 620e626
device. When using these stationary electrodes in the all-iron flow battery, iron metal is plated into their structure and subsequently stored in the electrochemical cell. For a majority of grid applications, it is expected that an energy to power ratio of 3e8 h is required [2]. Because iron metal is stored in the electrochemical cell, the conventional all-iron battery is limited to <4 h of energy storage at reasonably high current densities [13]. Existing capital cost models suggest that current densities above 200 mA cm2 will be necessary to achieve acceptable stack costs for large scale adoption of flow batteries [11,14,15]. In order to have a battery with the advantages inherent to traditional flow battery architectures and the advantages of the all-iron chemistry, a different electrode structure must be employed that enables higher energy storage to power ratios. One such option is the slurry electrode. Slurry electrodes are made by suspending solid, electronically conductive particles in the ionically conducting electrolyte and have been studied for a variety of electrochemical applications [17e23]. During charge and discharge, the electrolyte is pumped through the electrochemical cell as in normal RFB operation. However, when using slurry electrodes, the solid particles are carried by the electrolyte into and out of the electrochemical cell. If the volume fraction of the solid particles is high enough, a continuous electrically conductive network forms [24,25]. This allows the redox reaction to occur on the surface of the slurry particles. When slurries are employed as the negative electrode of the all-iron flow battery, iron metal is plated onto the slurry particles. As the slurry is circulated out of the electrochemical cell back into the external reservoir, the iron metal is carried with the particles, enabling the energy storage capacity to be decoupled from the power rating. Additional advantages of slurry electrodes include their ability to have high surface areas, simple manufacturing/assembly, and ease of maintenance/recycling through filtering. Previous studies have investigated the ferrous/ferric redox couple on slurry electrodes [23]. This paper reports results of an investigation of the performance of slurry electrodes in the full all-iron flow battery while focusing on the plating of iron metal into the negative slurry electrode. 2. Analytical current distribution models Mathematical modeling of the slurry electrode in the negative half-cell was performed in order to permit the selection of viable candidate slurries. The selection of an appropriate particle to use as the flowable slurry electrode was performed with regards to two principal performance metrics. The first performance metric was related to the achievable current density. Cost models have shown that for the all-iron slurry battery to be practical that the battery needs to be able to support a current density of 200 mA cm2 at a voltaic efficiency of at least 70% [11,14,15]. Given the 1.2 V cell potential of the all-iron battery, this means that the desired current density must be achieved with at most 212 mV of total overpotential across the entire cell. The second performance metric is related to the amount of plating that occurs onto the current collector. If plating occurs onto the current collector and not onto the slurry particles, then the energy density and power density of the battery are no longer decoupled. As decoupling is a desirable trait for grid-scale energy storage, a criterion was established demanding that less than 5% of the total metal plating should occur onto the current collector. The slurry characteristics necessary to match the performance metrics were examined using standard one-dimensional macrohomogeneous porous electrode equations [26,27]. The governing equations are identical to those used by Fleischmann and Oldfield for fluidized bed electrodes [28]. The kinetics of the plating reaction were assumed to be linear, resulting in the following equations:
621
s
d2 F1 a0 zfio h ¼ 0 dx2
(3a)
k
d2 F2 þ a0 zfio h ¼ 0 dx2
(3b)
Here, s is the electronic conductivity of the slurry phase, k is the ionic conductivity of the solution phase, F1 is the potential of the slurry phase, F2 is the potential of the solution phase, a0 is the specific area of the slurry electrode, z is the number of electrons involved in the plating reaction, f¼FR1 T1 is related to Faraday's constant, i0 is the exchange current density, x is the spatial coordinate across the channel gap, h¼F1F2Uref is the overpotential, and Uref is the open-circuit potential of the electrode reaction relative to a suitable reference electrode such that the overpotential is zero at equilibrium. Effects such as mass transfer overpotentials or advectively-induced capacitive currents are ignored by this model, but it is believed to capture the dominant behavior associated with slurry electrodes in the presence of fast faradaic reactions. In order to account for the fact that reactions can occur on both the current collector plate and the slurry particles, the standard boundary conditions used by Fleischmann and Oldfield needed to be altered. The modified boundary conditions at x ¼ 0 (the current collector) and x ¼ d (the location of the separator) take the form:
dF1 ði xzfio ho Þ dF2 xzfio ho ; ¼ d ¼ s dx dx k dF1 dF2 id ¼ 0; ¼ dx dx k
at x ¼ 0
at x ¼ d
(4a)
(4b)
Here, id is the applied current density, x is the roughness factor of the current collector, and h0 is the overpotential at x ¼ 0. The total overpotential across the half-cell can be described as htot¼F1(0) F2(d)Uref. An additional factor accounting for the area of the current collector that is occluded by the presence of the slurry could also be included in the boundary conditions, but for low volume fraction slurries (like those in this paper), this effect is believed to be negligible. The slurry properties that affect the achievement of the above performance metrics are the specific interfacial area and the electronic conductivity. Using typical values for cell dimensions, iron plating reaction kinetics, and electrolyte properties (see Table 1), Equations (3a) and (3b) were solved in order to generate maps of the combinations of s and a0 that satisfy the metrics. These maps are shown in Fig. 1. Regions corresponding to the polarization metric (<100 mV of overpotential in the negative half cell at 200 mA cm2 or higher) and to the current collector plating metric are both included. Any viable slurry must lie within the bounds of both regions. As shown in Fig. 1, slurries made with MWCNTs are predicted to be able to achieve both performance metrics [23]. It is likely that other less expensive particles can also match the criteria, but, as a proof of concept, the MWCNT slurry was adopted to permit the
Table 1 Parameters used for slurry electrode model. Parameter
Value
d
0.1 cm 2 mole e 2þ mole Fe 10 cm2 cm2 1 mA cm2 200 mS cm1 100 mV
z
x i0
k htot
622
T.J. Petek et al. / Journal of Power Sources 294 (2015) 620e626
polycarbonate tube (3/16 in. ID) through which the slurry was run at various flow rates (ranging from 20 mL/min to 130 mL/min). Differential pressure transducers located 100 cm apart along the length of the tube were used to determine the pressure drop. The reference side of each pressure sensor was open to the atmosphere. During each test, the particles were continually mixed in the reservoir to ensure stable results. The length and diameter of the polycarbonate tube were carefully selected in order to achieve a similar range of shear rates to those that are encountered in the electrochemical flow cell (while still maintaining laminar flow within the tube). The tube viscometer was used in order to avoid the settling and shear-history issues that often encountered with, e.g., cone-and-plate rheometers. The Ostwaldede Waele (power law) model was adopted to describe the non-Newtonian behavior of the slurry over the range of shear rates considered, as has been done in the work of others [31]. This model describes the shear stress in terms of the shear rate, vu/vy, a flow consistency index, K, and a flow behavior index, n, as related through Equation (5). Fig. 1. Slurry electrode performance maps. The red shaded region represents the combinations of s and a0 that satisfy the requirement that the slurry electrode be able to support a 200 mA cm2 current density for 100 mV total overpotential across the negative half-cell. The blue shaded region represents the combinations of s and a0 that satisfy the requirement that the slurry electrode be able to plate less than 5% of the total metal onto the current collector. The purple region denotes satisfaction of both metrics. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
examination of the slurry battery performance. 3. Experimental 3.1. Materials The slurry electrodes in this study were made with MWCNTs purchased from Nanostructured and Amorphous Materials, Inc. (Houston, TX, USA). These MWCNTs have an OD of 50e80 nm, length of 10e20 mm (aspect ratio of 125e400), a reported surface area of 40 m2/g, and a carbon purity of 99.9%. For these studies, the MWCNTs were used as received. Before tests were performed, the particles were mixed by hand into the electrolyte and then pumped through the system for at least 30 min. All-iron flow batteries were studied with slurries serving as both the positive and the negative electrodes. In all cases presented herein, the slurry electrodes were composed of 4.8 vol% MWCNTs in the electrolyte. The electronic conductivity of these slurries was previously found to be about 85 mS cm1 while flowing [23]. The positive slurry was initially 0.5 M FeCl2, 0.5 M FeCl3, and 1.0 M NH4Cl while the negative slurry was initially 0.5 M FeCl2 and 1.0 M NH4Cl. The ionic conductivity of these electrolytes is about 150 mS/ cm. Adding ferric chloride to the positive electrolyte ensured that on discharge, the negative slurry would always be limiting. All salts used were reagent grade and were obtained from Fisher Scientific (USA). All solutions were made with deionized water (>18 MU-cm, Barnstead Milli-Q). When using ferrous salts, solutions were made under a continuous nitrogen blanket and experiments were performed with the electrolyte under a continuous nitrogen purge to mitigate the air oxidation of ferrous to ferric iron [29,30]. All tests were conducted at room temperature (20e25 C). 3.2. Rheological studies The slurries in these studies are intended for use as flowing electrodes. As such, the viscosity of the slurry electrodes are of interest and were investigated with an in-house tube viscometer. The tube viscometer consisted of a vertically-oriented clear
t¼K
n vu vy
(5)
The values for K and n can be determined from tube viscometry data through comparison to an analytical prediction of the pressure drop. The equation for the analytically predicted pressure drop of an Ostwaldede Waele fluid through a circular cylindrical tube is:
h n iDp1=n Q ¼p r ð3nþ1Þ=n 3n þ 1 2KL
(6)
Here, Q is the flow rate, Dp is the pressure drop, L is the length of the tube, and r is the tube radius [32]. Table 2 shows the values for K and n that were determined for various loadings of multi-walled carbon nanotubes in a 1 M solution of ferric chloride. Similar to the observations in previous studies [19e21], the slurries did exhibit shear thinning behavior with smaller flow behavior indices as the volume loading was increased. For flow in the electrochemical cell described below at the flow rates used throughout this experimental work (250 mL/min), the shear rate at the wall for the slurry with the highest volume loading (4.3 vol%) in Table 1 is 3750 s1. The effective viscosity at the wall for this slurry can therefore be calculated to be 10.54 cP. As this viscosity value is only roughly five times that of the bare electrolyte, the pumping costs for the flowable slurry electrode should not be prohibitively large. In addition, the additional pressure drop that arises in standard redox flow batteries from the presence of porous electrodes or serpentine/interdigitated flow fields is absent in slurry flow batteries. However, future work must be done to quantify the parasitic losses due to pumping more fully. It was observed that when the volume fraction of the solid particles was above some critical concentration, there was no longer enough liquid in the slurry to fully wet all of the particles. When the volume fraction of solids rose above this critical volume fraction, the slurry transitioned, quite quickly, from a shear
Table 2 Flow consistency index, K, and flow behavior index, n, for various weight percentages of MWCNTs in 1 M ferric chloride. Vol%
n
K (cP sn1)
0.5 1.9 2.8 4.3
0.99 0.64 0.49 0.40
1.92 59.6 351 1470
T.J. Petek et al. / Journal of Power Sources 294 (2015) 620e626
623
thinning fluid to a semi-solid gel. At this point, the slurry is no longer able to flow. This viscosity critical volume fraction is described in detail by Youssry et al. [21] For slurries made with the MWCNTs used in these studies, the viscosity critical volume fraction was found to be about 6 vol%. 3.3. Electrochemical studies The all-iron flow battery investigations were conducted in an inhouse channel cell through which the slurry electrodes flow during operation. This channel cell was defined by two CPVC, chlorinated polyvinyl chloride, flow fields containing straight rectangular channels with a width of 1 cm and length of 12 cm (in the direction of flow). The current collectors were 1 cm wide and 6.725 cm long (A ¼ 6.725 cm2) graphite plates that were centered in the middle of the CPVC channel. The gap between the current collector of each channel and the separator, d, was 0.2 cm. During operation, a peristaltic pump (ColeeParmer, USA) was used for pumping the slurry electrodes. In all experiments, the slurry was pumped vertically up through the electrochemical channel cell at a volumetric flow rate of 250 mL/min. The inlet and exit ports of the channel cell were at roughly 30 angles (with respect to the current collector face) to mitigate clogging. Current was collected from this cell through two brass plates that were pressed into the back of the graphite current collectors. The cell was held together with stainless steel endplates. Teflon sheets were used to insulate the endplates and Grafoil (Graftech, USA) was used to minimize contact resistances between the brass plates and the graphite plates. The cell assembly was compressed with 8 bolts that were tightened to 90 in-lbs of torque. A schematic of the cell (not to scale) is shown in Fig. 2. The reservoirs used to hold the slurry that was pumped through the channel cell were Nalgene bottles modified to be bottom drawn. Unless otherwise noted, all cell materials were purchased from McMaster-Carr (Cleveland, OH, USA). Daramic, LLC (Charlotte, NC) graciously donated Daramic 250 Flatsheet for use in these studies. Daramic 250 Flatsheet has a reported average pore size of 0.1 microns, average porosity of 58±7.5%, and a thickness of about 225 microns measured after use in experiments. The Daramic was soaked in methanol and rinsed in DI water before use so that air bubbles trapped inside the pores were removed; ensuring that the separator be completely wetted [8]. The ionic conductivity of these separators, kDaramic, has been estimated to be 40% of the conductivity of the electrolyte using the reported average porosity, e, of 58% and the Bruggeman approximation, Equation (7) [16].
kDaramic ¼ eð3=2Þ
Fig. 2. A schematic of the electrochemical channel cell (not to scale) used to characterize the slurry performance. The channel is defined by the flow frame and the electric field is defined by the inset current collectors. The slurry flows up vertically through the cell and the inlet/exit are at approximately 30 to vertical to help mitigate clogging. Not shown are the materials to compress the cell and the brass plates from which current is drawn.
(7)
All electrochemical studies were performed using a Solartron 1280B potentiostat. Electrochemical impedance spectroscopy (EIS) was performed at varying cell potentials with a sinusoidal perturbation of 10 mV amplitude over the frequency range 20 kHze0.2 Hz. The EIS experiments were controlled using ZPlot (Scribner Associates, USA). Cell polarization scans and an oscilloscope (Tektronix, USA) were used in congress to make sure that the response of the system under perturbation remained linear. Cyclic voltammetry with a scan rate of 20 mV/s was used to approximate the polarization of the all-iron flow battery and was controlled using CorrWare (Scribner). 4. Results and discussion Before the initial charging of the all-iron flow battery, EIS measurements were first conducted at the open circuit potential, shown in Fig. 3. The high frequency resistance of the battery should
Fig. 3. The Nyquist representation of the EIS response of the all-iron flow battery at different states of charge.
be the sum of the parallel ionic and electronic resistances of each slurry electrode along with the separator resistance [23]. For a battery with 2 mm channels on either side of Daramic 250 Flatsheet, the total battery high frequency resistance should be about 2.0 U cm2 based on the independent electrical and ionic conductivity measurements. As seen in Fig. 3, the initial EIS measurements are in good agreement with this prediction; confirming the conductivities of the slurry electrodes and separator. The EIS results at subsequent states-of-charge (SOC) will be discussed below. Fig. 4 shows the typical behavior of an all-iron flow battery while charging. The initial battery charge cycle was performed at
624
T.J. Petek et al. / Journal of Power Sources 294 (2015) 620e626
Fig. 4. The voltage profiles while charging an all-iron flow battery at constant current with both positive and negative slurry electrodes. The discontinuity in the 75 mA cm2 curve occurred when the tubing between the cell and reservoir was readjusted.
75 mA/cm2. The battery voltage rose with the SOC before leveling off around 1.5 V. At the end of the initial charge, just over 1.0 Ah of charge had been stored in the battery bringing the 500 mL negative slurry to just over 7% state-of-charge (SOC). At this point, a cyclic voltammogram was performed on the battery and the results are shown in Fig. 5. As the voltammogram shows, the open circuit potential of the all-iron flow battery was about 1.2 V. From the open circuit potential, the cell voltage was scanned first positive 500 mV and then to 500 mV below the open circuit potential before returning to the open circuit potential at a rate of 20 mV s1. The data shown is the second of three cycles performed. Each cycle showed good reproducibility compared to the previous. In these experiments, the positive electrode is the working electrode and the negative electrode is the counter and reference. Positive currents represent oxidation at the working electrode, i.e., charging the battery.
Fig. 5. The cyclic voltammetry of the all-iron all-slurry flow battery at 7% state-ofcharge after charging at 75 mA cm2. The voltammogram is scanned at 20 mV/s first positive from the open circuit potential and negative to 500 mV vs the OCV before scanning back to the OCV. The data shown in this figure is the second of three cycles; each showing good reproducibility.
After the voltammogram was obtained the charge was continued at 150 mA cm2 for 2 additional hours (constituting a total battery charge of just over 3.0 Ah). During this period, the cell charging voltage fell to 1.45 V after first going through a voltage maximum. This decrease in cell overpotentials at increased current density operation was corroborated by the EIS response of the battery, Fig. 3, decreasing as the SOC of the battery increases. EIS was performed at a discharge potential so that the results can be more easily compared at different SOCs. During discharge the dissolution of iron metal is directly studied and the effects of the variation of ferrous iron should be mitigated. Also, any additional side reactions that may occur during charging, discussed below, will not occur. There are two main mechanisms thought to contribute to the improved performance. As the SOC of the negative slurry rises, the amount of metal deposited in the slurry increases and may therefore increase the slurry electronic conductivity. Also, after the carbon nanotubes are seeded with iron metal, the kinetics of iron plating are enhanced because the kinetic hindrance associated with iron nucleation on carbon is no longer a factor [33]. The decrease in the overpotential on charge (whether by increased conductivity and/or improved kinetics) has an additional benefit. When charging at large potentials, the negative electrode can become negative enough for hydrogen evolution to occur [33]. When this happens, the electrolyte pH will rise, potentially causing ferric or ferrous iron to precipitate out of solution as their respective hydroxides [33]. The precipitated iron hydroxide can form a film on the slurry particles making discharge of the battery more difficult. Fig. 6 shows the voltage profiles while cycling the all-iron flow battery at ±75 mA cm2. Each half-cycle is 0.5 h; constituting swings of ±0.25 Ah. The average charging potential is 1.55 V while the average discharge potential is 0.8 V. The average voltaic efficiency is just above 50%. The voltaic efficiency could be improved by optimizing the separator and cell design. Over the twelve cycles shown, no appreciable capacity fade was observed (each half cycle lasted the 0.5 h) and the potential profiles were steady. These cycles were performed after the two initial charging profiles shown in Fig. 4 (constituting ±0.25 Ah cycles around 3.0 Ah). After cycling was concluded, the battery was disassembled at the end of a charge. A thin film (<0.02 cm) of iron was visible
Fig. 6. Voltage profiles while cycling the all-iron flow battery with MWCNT slurry electrodes at ±75 mA cm2. Each cycle is ±0.25 Ah and was performed after the initial charging profiles shown in Fig. 4.
T.J. Petek et al. / Journal of Power Sources 294 (2015) 620e626
on < 25% of the current collector surface area. If all 3.0 Ah put into the battery was plated onto the current collector, the resulting metal plate would have to be approximately 0.07 cm, or 33% of the channel thickness. The iron that had plated onto the current collector was dissolved into a known volume of HCl and the resulting concentration was determined with a platinum microelectrode [34]. From the dissolved iron concentration and the volume of HCl, the amount of iron deposited on the current collector was found to be <5% of the total charge put into the battery, confirming that most of the metal was plated onto the slurry particles. The cycles performed in Fig. 6 are relatively shallow and at low current densities in order to mitigate the effects of hydrogen evolution. Despite this, after the battery charging and cycling experiments were concluded, Figs. 4 and 6, the electrolyte pH was observed to have risen from its initial value of 2 (due to the mildly acidic nature of ferrous ions) to a pH of about 4; presumably due to hydrogen evolution. This shift in pH, if solely due to hydrogen evolution, would account for <0.25 Ah total over the >16 h of operation and 3.25 Ah of charge deposited. It is likely that the hydrogen evolution occurs with a similar current distribution as discussed in Section 2 above. As such, the majority of the hydrogen evolution is anticipated to occur on the slurry particles and should not have a significant effect on the distribution of iron in the system. Future work is required to gain a better understanding of how to mitigate the effects of hydrogen evolution and to better characterize the coulombic efficiency. While the cycles performed in these studies were relatively shallow, for a slurry to be truly viable for the all-iron battery it must maintain physical and electrochemical properties over wide SOC swings. Fig. 7 shows the expected change in the slurry particle physical properties if all of the iron in a 2 M ferrous chloride negative electrolyte was plated (100% SOC). As iron metal is plated on the particles, the particle density will increase. However, as the amount of particles in the slurry increase, the amount of iron plated onto each particle (assuming uniform distribution) decreases. This analysis suggests that particles capable of forming slurries with acceptable flowability at higher volume fractions are required in order to mitigate the effects of charging at high SOC. Also, achieving higher particle loading has an additional benefit of increased
specific area, described in Fig. 1. 5. Conclusion Slurry electrodes were investigated in order to decouple the energy storage capacity from the power delivery rate in an all-iron flow battery. For the slurry electrode to perform effectively, the battery negative reaction must occur on the slurry particles at reasonably high current densities. Mathematical modeling was used to investigate the current distribution in a slurry electrode as a function of the slurry specific area and electrical conductivity in order to achieve >95% plating in the slurry electrode (not on the flat plate) at > 200 mA cm2. From the mathematical modeling, MWCNTs were selected as one viable slurry electrode and were used in these studies to prove the feasibility of a slurry for the alliron battery. All-iron flow batteries were operated with MWCNT slurries as both the positive and negative electrodes. As the battery state-ofcharge increased, the voltaic efficiency of the battery also increased. Two mechanisms thought to contribute to the improved performance are an increase in the slurry electronic conductivity (due to metal deposition on the particles) and an increase in plating kinetics. After cycling the battery around 25% state-of-charge, <5% of the charge of the battery was observed as iron metal on the current collector. While the MWCNT slurry electrodes discussed in this work were shown to operate with current distributions that effectively decouple the system, future work must still be accomplished for the all-iron battery to be a truly viable energy storage device. The slurry electrode must be developed in parallel with improved electrolyte chemistries so that a more cost effective slurry can be utilized (and still meet the criteria for viable current distribution) that can support higher current densities and high states-of-charge with respect to iron deposition with acceptably high coulombic efficiencies. Acknowledgments The authors would like to thank Nicholas Sinclair, Jason Pickering, Clay Kacica, Ashley Yarus, Bryan Erb, Joseph Plazek, Matthew Madler, and Mirko Antloga for assistance in conducting experimental trials and maintaining laboratory equipment. The funding for this project is through ARPA-E, contract number DEAR0000352. Nomenclature a0 F id io K L n p Q R r T Uref vu/vy
Fig. 7. The physical properties of the slurry particles at 100% SOC plating density relative to properties at 0% SOC (assuming uniform distribution) when in 2 M negative electrolyte of an all-iron battery plotted versus the particle volume fraction. Each property is normalized by that of pure carbon.
625
F1 F2 z
d
Slurry specific area, [m2 m3] 1 Faraday's constant, ½96; 485 A s mole e 2 applied current density, [A cm ] exchange current density, [A cm2] Flow consistency index, [cP sn1] length of viscometer tube, [cm] flow behavior index pressure, [cP s1] flow rate, [cm3 s1] gas constant, [8.314 A V s K1 mol1] inner radius of tube viscometer, [cm] temperature, [K] open-circuit potential of electrode reaction, [V] shear rate, [s1] solid phase potential, [V] solution phase potential, [V] electron equivalence, [mole e/mole Fe] Electrode channel gap, [cm]
626
h k
x s t
e
T.J. Petek et al. / Journal of Power Sources 294 (2015) 620e626
overpotential, [V] Solution phase ionic conductivity, [S cm1] current collector roughness factor Solid phase electronic conductivity, [S cm1] shear stress, [cP s1] void fraction
References n, A. Frías-Ferrer, J. Gonza lez-García, D.A. Sza nto, F.C. Walsh, [1] C. Ponce de Leo J. Power Sources 160 (2006) 716. [2] J. Eyer, G. Corey, Energy Storage for the Electricity Grid: Benefits and Market Potential Assessment Guide, Sandia Report SAND2010e0815, 2010. [3] T. Nguyen, R.F. Savinell, ECS Interface 19 (2010) 54. [4] M. Skyllas-Kazacos, M.H. Chakrabarti, S.A. Hajimolana, F.S. Mjalli, M. Saleem, J. Electrochem. Soc. 158 (2011) R55. [5] A.Z. Weber, M.M. Mench, J.P. Meyers, P.N. Ross, J.T. Gostick, Q. Liu, J. Appl. Electrochem 41 (2011) 1137. [6] M. Lopez-Atalaya, G. Codina, J.R. Perez, J.L. Vazquez, A. Aldaz, J. Power Sources 39 (1992) 147. [7] W. Wang, S. Kim, B. Chen, Z. Nie, J. Zhang, G.-G. Xia, L. Li, Z. Yang, Energy Environ. Sci. 4 (2011) 4068. [8] X. Wei, L. Li, Q. Luo, Z. Nie, W. Wang, B. Li, G.-G. Xia, E. Miller, J. Chambers, Z. Yang, J. Power Sources 218 (2012) 39. [9] A. Price, S. Bartley, S. Male, G. Cooley, Power Eng. J. 13 (1999) 122. [10] M. Rychcik, M. Skyllas-Kazacos, J. Power Sources 22 (1988) 59. [11] V. Viswanathan, A. Crawford, D. Stephenson, S. Kim, W. Wang, B. Li, G. Coffey, E. Thomsen, G. Graff, P. Balducci, M. Kintner-Meyer, V. Sprenkle, J. Power Sources 247 (2014) 1040. [12] L.W. Hruska, R.F. Savinell, J. Electrochem. Soc. 128 (1981) 18. [13] K.L. Hawthorne, J.S. Wainright, R.F. Savinell, J. Power Sources 269 (2014) 216.
[14] D.P. Scamman, G.W. Reade, E.P.L. Roberts, J. Power Sources 189 (2009) 1231. [15] Vanadium redox flow batteries: an in-depth analysis, EPRI, Palo Alto, CA, 2007, p. 1014836. [16] S. Banisi, J.A. Finch, A.R. Laplante, Minerals Engineering 6 (1993) 269. [17] D.N. Baria, H.M. Hulburt, J. Electrochem. Soc. 120 (1973) 1333. [18] A.J. Appleby, M. Jacquier, J. Power Sources 1 (1976) 17. [19] M. Duduta, B. Ho, V.C. Wood, P. Limthongkul, V.E. Brunini, W.C. Carter, Y.M. Chiang, Adv. Energy Mater. 1 (2011) 511. [20] V. Presser, C.R. Dennison, J. Campos, K.W. Knehr, E.C. Kumbur, Y. Gogotsi, Adv. Energy Mater. 2 (2012) 895. [21] M. Youssry, L. Madec, P. Soudan, M. Cerbelaud, D. Guyomard, B. Lestriez, Phys. Chem. Chem. Phys. 15 (2013) 14476. [22] S. Jeon, H. Park, J. Yeo, S.C. Yang, C.H. Cho, M.H. Han, D.K. Kim, Energy Environ. Sci. 6 (2013) 1471. [23] T.J. Petek, N.C. Hoyt, J.S. Wainright, R.F. Savinell, J. Electrochem. Soc. 163 (2016) A5001. [24] C.G. Granqvist, O. Hunderi, Phys. Rev. B 18 (1978) 1554. [25] L. Gao, X. Zhou, Y. Ding, Chem. Phys. Lett. 434 (2007) 297. [26] J. Newman, C. Tobias, J. Electrochem. Soc. 109 (1962) 1183e1191. [27] J. Newman, W. Tiedemann, AIChe J. 21 (1975) 1. [28] M. Fleischmann, J.W. Oldfield, J. Electroanal. Chem. Interfacial Electrochem 29 (1971) 211. [29] P.C. Singer, W. Strumm, Sci. (N. Y.) 167 (1970) 1121. [30] N.F. Gray, Environ. Geol. 30 (1997) 62. [31] J.W. Campos, M. Beidaghi, K.B. Hatzell, C.R. Dennison, B. Musci, V. Presser, E.C. Kumbur, Y. Gogotsi, Electrochim. Acta 98 (2013) 123. [32] R.P. Chhabra, J.F. Richardson, Non-newtonian Flow in the Process Industries: Fundamentals and Applications, Butterworth-Heinemann, Oxford England, 1999. [33] K.L. Hawthorne, T.J. Petek, M.A. Miller, J.S. Wainright, R.F. Savinell, J. Electrochem. Soc. 162 (2015) A108. [34] R.A. Malmsten, C.P. Smith, H.S. White, J. Electroanal. Chem. 215 (1986) 223.