Iron-Based Flow Batteries: Improving Lifetime and Performance

IRON-BASED FLOW BATTERIES: IMPROVING

LIFETIME AND PERFORMANCE

STEVEN SELVERSTON

Submitted in partial fulﬁllment of the requirements

for the degree of Doctor of Philosophy

Department of Chemical and Biomolecular Engineering

CASE WESTERN RESERVE UNIVERSITY

August, 2017 CASE WESTERN RESERVE UNIVERSITY

SCHOOL OF GRADUATE STUDIES

We hereby approve the dissertation of Steven Selverston

candidate for the degree of Doctor of Philosophy*.

Committee Chair Dr. Robert Savinell

Committee Member Dr. Jesse Wainright

Committee Member Dr. Rohan Akolkar

Committee Member Dr. Gary Wnek

Date of Defense

May 18, 2017

*We also certify that written approval has been obtained

for any proprietary material contained therein. Contents

List of Tables iii

List of Figures v

Acknowledgments ix

Abstract x

1 Introduction 1

2 Literature Review 5 2.1 Flow Batteries ...... 6 2.2 All-Iron Hybrid Flow Batteries ...... 8 2.3 Electrolytes ...... 10 2.4 Iron Plating Electrodes ...... 17 2.5 Rebalancing ...... 20 2.6 System Modeling ...... 21 2.7 Zinc-Iron Electroplating ...... 24 2.8 Accelerated Lifetime Testing ...... 28

3 Dissertation Research 30 3.1 Electrolyte Rebalancing ...... 31 3.2 System Modeling ...... 32

i CONTENTS

3.3 Zinc-Iron Chloride Flow Batteries ...... 32

4 Model for Sealed Flow Batteries 34 4.1 Introduction ...... 35 4.2 Materials & Methods ...... 36 4.3 Model Development ...... 38 4.4 Results & Discussion ...... 44 4.5 Conclusions ...... 49

5 In-Tank Recombination 51 5.1 Introduction ...... 52 5.2 Materials & Methods ...... 55 5.3 Results & Discussion ...... 58 5.4 Conclusions ...... 63

6 Zinc-Iron Chloride Flow Batteries 65 6.1 Introduction ...... 66 6.2 Materials & Methods ...... 71 6.3 Results & Discussion ...... 72 6.4 Conclusions ...... 83

7 Conclusions and Recommendations for Future Research 84

A Supporting Information 89 A.1 Chemicals Used ...... 90 A.2 In-Tank Hydrogen-Ferric Ion Recombination ...... 90 A.3 Zinc-Iron Chloride Flow Batteries ...... 110 A.4 System Model ...... 121

References 142

ii List of Tables

2.1 Commercial status of ﬂow battery startup companies ...... 7 2.2 Complexation reactions in iron chloride electrolytes ...... 10 2.3 Equilibrium constants for acid ferrous chloride ...... 15

2.4 Measured and calculated pH of FeCl2-HCl-H2O solutions ...... 16 2.5 Studies of iron plating in flow batteries ...... 17 2.6 Effects of anions on iron kinetics ...... 18 2.7 Proposed mechanisms for anomalous codeposition ...... 26 2.8 Compositions and efficiencies of acidic Zn-Fe alloy plating baths ... 27 2.9 Selected flow battery lifetime studies ...... 29

4.1 All-iron hybrid battery reactions ...... 36

5.1 Electrode reactions in all-iron ﬂow batteries...... 52

A.1 Information regarding the chemicals used in the hydrogen-ferric ion recombination, all-iron battery and zinc-iron battery tests...... 90 A.2 Conditions for all-iron battery tests...... 93 A.3 Sample of raw pressure vs time data for measurement of recombination rate ...... 98 A.4 Example calculation of hydrogen oxidation rate from pressure data . 99 A.5 Secondary current distribution ...... 104 A.6 Measured conductivity of zinc-iron chloride electrolytes ...... 112

iii LIST OF TABLES

A.7 Example of pressure-based hydrogen generation rate measurements . 120 A.8 Initial concentrations and diﬀusivities used in the simulation...... 122 A.9 Baseline parameter values used ...... 122 A.10 Measurement of hydrogen generation rate ...... 123 A.11 Example of an iterative simulation setup ...... 135 A.12 Example program output, Part I ...... 137 A.13 Example program output, Part II ...... 138 A.14 Example program output, Part III ...... 139 A.15 Example program output, Part IV ...... 140

iv List of Figures

1.1 The Duck Chart interpretation of the energy storage challenge in Cal- ifornia...... 3 1.2 Concentrated solar power plant with 5 MWh vanadium redox ﬂow battery developed by Sumitomo Electric Industries ...... 4

2.1 Tree diagram of iron ﬂow batteries ...... 8 2.2 Schematic of the all-iron ﬂow battery ...... 9 2.3 Iron phase diagram and formation of hydroxides ...... 12

2.4 pH of ferrous chloride (FeCl2) ...... 13

2.5 pH of ferric chloride (FeCl3) ...... 13 2.6 Comparison between between measured and calculated pH values using different methods ...... 16 2.7 Iron corrosion rate ...... 19 2.8 Pourbaix diagram interpretation of the all-iron flow battery ...... 20 2.9 NASA iron-chromium flow battery...... 22 2.10 Rebalancing cells for Fe-Cr and All-V batteries...... 22 2.11 Relative positions of zinc and iron couples on the hydrogen scale. .. 24 2.12 Illustration of hydroxide suppression mechanism ...... 25

4.1 Illustration of electrode reactions and ion migration ...... 39

v LIST OF FIGURES

4.2 Example of pressure measurements during battery charging and discharging ...... 44 4.3 Eﬀect of pH on hydrogen generation rate ...... 45 4.4 Measured versus simulated gas pressure in the reservoir headspace .. 46 4.5 Variation of headspace pressure with time during continuous charge- discharge battery cycling ...... 46 4.6 Simulated iron concentrations in positive electrolyte with and without rebalancing ...... 47 4.7 Simulated hydrogen generation and consumption currents ...... 48 4.8 Eﬀects of separator porosity and thickness on steady-state pressure . 49

5.1 Schematic of a capillary-action galvanic reactor (CGR) ...... 56 5.2 Schematics of pressurized vessels for CGR testing ...... 57 5.3 Schematic of a sealed recombinant ﬂow battery ...... 58

5.4 CGR potential as function of pH2 ...... 59

5.5 CGR potential versus time during continuous H2 oxidation...... 59 5.6 Impedance and polarization measurements of a CGR ...... 60 5.7 Pressure and hydrogen oxidation rate measurements ...... 61 5.8 Pressure measurements during battery cycling ...... 62

6.1 Schematic of a zinc-iron chloride flow battery with mixed electrolytes 68 6.2 Other zinc-iron flow battery designs...... 70 6.3 Cyclic voltammogram interpretation of anomalous codeposition in the zinc-iron system ...... 73 6.4 Effect of electrolyte composition on plating and stripping processes . 74 6.5 Effect of bulk pH on the plating and stripping processes ...... 75 6.6 Effect of rotation rate and negative scan limit on deposition and stripping from mixed Zn-Fe electrolytes ...... 77

vi LIST OF FIGURES

6.7 Eﬀect of rotation rate and scan limit on deposition and stripping processes ...... 77 6.8 Eﬀect of zinc chloride on the Fe2+/3+ reaction ...... 78 6.9 Battery voltage during a charge-discharge cycle at 25 mA cm−2 ... 79 6.10 Comparison between all-iron and zinc-iron voltage during charging and discharging ...... 80 6.11 Cell potential and performance during continuous charge-discharge battery testing ...... 80 6.12 Cell potential during a 30-day battery cycling test ...... 82

A.1 Sealed reservoir schematic and photo ...... 94 A.2 Pressure and recombination rate measurements...... 95 A.3 CGR pressure, rate and mixed potential versus time...... 95 A.4 Typical 30-cm2 cell hardware ...... 100 A.5 Photograph of a complete battery system ...... 100 A.6 A reactor array after battery testing ...... 101 A.7 Appearance of electrolytes after 10 days of battery testing ...... 101 A.8 Electrochemical characterization of CGR as function of hydrogen partial pressure ...... 102 A.9 Comparison of battery pressure proﬁles with and without rebalancing 102 A.10 Daramic separator after battery cycling ...... 103 A.11 Secondary current distribution in the CGR ...... 105 A.12 Simulated potential and current in the CGR as function of height above surface ...... 105 A.13 Schematic of the H-Cell used for cyclic voltammetry ...... 110 A.14 Eﬀect of total metal ion concentration on voltammograms for mixed zinc-iron chloride electrolytes on a titanium substrate...... 111

vii LIST OF FIGURES

A.15 Effect of supporting electrolyte concentration on deposition and stripping behavior of mixed zinc-iron chloride electrolytes...... 111 A.16 Sketch of glass conductivity measurement cell (l/A = 200 cm−1). ... 112 A.17 Zinc-iron chloride conductivity ...... 113 A.18 Schematic of cell parts when using ribs...... 114 A.19 Photograph of an assembled 6.25 cm2 cell...... 115 A.20 Bulk deposit morphology at 100 mAh cm−2 on graphite surface ... 115 A.21 Bulk deposit morphology at 100 mAh cm−2 when using pulse-plating at 25 mAh cm−2 (duty cycle = 0.5) ...... 116 A.22 Effect of Zn/Fe molar ratio on morphology using DC plating from quiescent solutions ...... 116 A.23 Effect of plating waveform on deposit morphology from quiescent solutions ...... 117 A.24 Morphology at 50 mAh cm−2 deposited on titanium using DC waveform from stirred solution using quaternary ammonium additive ... 117 A.25 XPS analysis of surface composition ...... 118 A.26 Pressure measurements and hydrogen current densities during Zn-Fe battery cycling ...... 119 A.27 Screenshot of the OMEdit development environment ...... 121

viii Acknowledgments

This work is dedicated to my parents, Susan and Allen Selverston. I would like to thank my research advisors, Prof. Savinell and Prof. Wainright, for their guidance and support. I am grateful for the help, suggestions and foundational work done by previous and current EEEL members including Ertan Agar, Lauren Anderson, Mirko Antloga, Andrea Bourke, Bryan Erb, Ismailia Escalante-Garc´ıa,Krista Hawthorne, Xinyou Ke, Nathaniel Hoyt, Lichao Liu, James Mellentine, Mallory Miller, Joseph Murphy, Enoch Nagelli, Tyler Petek, Jason Pickering, Nicholas Sinclair and Elizabeth Stricker. This work would not have been possible without funding provided by the US Department of Energy, Oﬃce of Electricity and by Paciﬁc Northwest National Laboratory (PNNL).

ix Iron-Based Flow Batteries: Improving Lifetime and Performance

Abstract by STEVEN SELVERSTON

For grid-scale energy storage applications, iron-based hybrid flow batteries have advantages of safety, sustainability and low-cost. Still, several challenges such as device lifetime and efficiency have limited their development. In this work, a new type of hydrogen-ferric ion recombination reactor based on catalyzed, three-dimensional felt is proposed in order to maintain chemical balance in the electrolytes and hence improve the battery stability and lifetime. Cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS) were used to identify a diffusion-limited hydrogen oxidation current near 0.3 psig of hydrogen partial pressure and show that the perfor-

mance can be improved with increasing hydrogen pressure up to about PH2 = 10 psig. Also, pressure-based measurements showed that high rates of hydrogen recombination (greater than 20 mA cm−2 based on the geometric area and greater than 100 mA cm−2 based on the cross-sectional area) were possible using a floating, membrane-less reactor design. A flow battery model that incorporates the hydrogen evolution side-reactions and chemical rebalancing was developed using a system of differential and algebraic equations (DAE). A good agreement between simulated and measured pressure profiles was obtained for an all-iron flow battery operating at 100 mA cm−2. Effects of separator porosity and thickness were simulated, showing how increased thickness and reduced porosity can cause higher pH in the negative electrolyte and hence reduced hydrogen generation. Lastly, a new hybrid flow battery based on mixed, lightly acidic electrolytes was investigated. By using the anomalous codeposition (ACD) phenomenon, it was possible to electrodeposit nearly pure zinc from mixed ZnCl2-FeCl2 electrolytes. The cell

x Abstract was shown to provide 17 % higher voltaic efficiency and 40 % higher power density compared to an all-iron battery operating under the same conditions. A zinc-iron chloride flow battery was tested for 30 days and 175 cycles at i = 25 mA cm−2 and 50 mAh cm−2 charge loading (two-hour charges) without the use of dendrite suppression additives, flow-fields or temperature control. The average coulombic, voltaic and energy efficiencies were 87 %, 82 % and 71 %, respectively. Therefore, with further development, zinc-iron chloride flow batteries represent a promising new approach for grid-scale energy storage.

xi Chapter 1

Introduction

1 CHAPTER 1. INTRODUCTION

There is a growing interest in using solar and wind power, with global installed capacities increasing at rates of 60 % and 20 % per year, respectively [1]. California has nearly 14,000 MW of installed solar power, representing about 6 % of the total generation. Texas has more than 17,000 MW of installed wind power (enough for about 17 million homes) [2]. One challenge of using solar and wind power, though, is integrating them into the electrical grid. Energy storage technologies like pumped hydro (PHS), compressed air (CAES) and batteries (BES) can help solve this problem by smoothing the power output. Unfortunately, grid-scale energy storage systems are still too expensive for widespread utilization. A good example of a potential application of grid-scale energy storage is in the state of California, where its usage will be required in order to avoid curtailment of new solar power installations [3]. The scenario is illustrated using a net load curve or “Duck Curve” (see Figure 1.1), which was developed by the California Association of Independent System Operators (CAISO) [4]. Energy storage technologies can be used to transfer the excess energy generated during the middle of the day to the evening hours, where there is a surge in demand that otherwise necessitates the use of so-called “peaker plants” to provide the power. Several recent reviews describe the diﬀerent energy storage technologies that can be used at the grid scale [5–10]. Pumped hydroelectric storage (PHS) constitutes the vast majority of grid-scale energy storage in place (about 40 units in the United States and 300 units are in operating worldwide), but its growth is limited by the requirement of having mountainous terrain [7,11]. Furthermore, construction of PHS systems requires large investments of > $100 M and as much as 10 years of construction [7]. Compressed energy storage systems (CAES) represent another technology being investigated for large energy storage applications, and in principal they can be applied both above-ground or underground. However, there are currently only two CAES units in operation worldwide [8,12]. As with PHS, large-scale CAES that

2 CHAPTER 1. INTRODUCTION

energy deﬁcit 30

20 batteries

net load (GW) 10 solar surplus

12 24 time of day (h) Figure 1.1: The duck chart interpretation of the energy storage challenge in California, adapted from Ref. [4].

require underground formations (e.g., salt caverns) are limited by the natural avail- ability of such locations [5, 8]. Still, PHS and CAES are considered to be the only commercially-available technologies for bulk energy storage [11]. Due to their flexi- bility and versatility, batteries can play an important role in the future of grid-scale energy storage [13]. Many companies and researchers are developing batteries for grid-scale enery storage. The familiar lead-acid batteries used in cars have the advantage of being a mature technology as well as being easily-recyclable [14]. However, their limited lifetimes have hindered their widespread adoption. Currently, lithium-ion batteries are the technology of choice for new grid-scale installations because of their longer lifetimes and higher coulombic efficiencies [15]. However, even optimistic models predict lithium ion batteries will still cost $125-250 kWh−1 by the year 2030 [16], whereas for bulk energy storage the target cost is about $100 kWh−1 [17]. Furthermore, the recyling infrastructure for lithium-ion batteries is almost non-existent, despite its clear neces- sity [18], and lithium ion batteries have ongoing challenges with fires and explosions. Therefore, flow batteries, which are inherently designed for large-scale applications,

3 CHAPTER 1. INTRODUCTION

are being investigated as alternatives [1,19–25] . For example, a 5 MWh all-vanadium flow battery by Sumitomo Electric (see Figure 1.2) has been operational in Japan since 2012 [26]. Several important challenges, however, have prevented widespread adoption of flow battery technologies. The five major challenges to widespread utilization of energy storage technologies noted in a recent review were cost, reliability, safety, equitable regulatory environments and industry acceptance [1]. This thesis will help address the technical aspects of these challenges.

Figure 1.2: Concentrated solar power plant with 5 MWh vanadium redox ﬂow battery developed by Sumitomo Electric Industries [26].

4 Chapter 2

Literature Review

5 CHAPTER 2. LITERATURE REVIEW

2.1 Flow Batteries

Flow batteries are considered to be promising for applications that require long- duration discharges on the order of three hours or more, but they are currently too expensive and too unreliable for widespread utilization [1]. Several reviews have discussed the different types of flow battery chemistries, their advantages and dis- advantages in comparison to other energy storage technologies, and the challenges that hinder their development [1,19–25,27,28]. The first major experimental work on flow batteries was carried out in the mid-to-late 1970’s by NASA, who developed the iron-chromium system over a period of about eight years at the Lewis Research Cen- ter [29]. The iron-chromium battery operated according to the overall reaction shown in Equation 2.1. This work led to the several patents, reports and demonstrations of > 500 cycles at 80 mA cm−2 in a bismuth-catalyzed system [30].

Fe2+ + Cr3+ ⇌ Fe3+ + Cr2+ (2.1)

In 1977, Lim and others studied zinc bromine ﬂow battery (tested at i = 20 mA/cm2) with electrolyte circulation [31]. The overall reaction for zinc-bromine batteries is shown in Equation 2.2. Flow batteries of this type, wherein a solid metal phase is formed during charging, are referred to as hybrid ﬂow batteries.

2+ − 0 Zn + 2Br ⇌ Zn + Br2 (2.2)

The major development of other flow battery technologies continued in the 1980’s, which saw development of newer flow battery chemistries such as all-iron [32] and all- vanadium flow batteries, as well as continued work on iron-chromium [33, 34] and zinc-bromine batteries [35–40]. The vanadium flow battery chemistry, which operates based on the overall cell reaction shown in Equation 2.3, was developed by the Skyllas-

6 CHAPTER 2. LITERATURE REVIEW

Kazacos group in the University of New South Wales in Australia [41–44] and it remains the most widely studied and developed flow battery variant today. Currently, all-vanadium flow batteries are in operation with power sizes ranging from 50 kW to 1 MW, according to a 2013 DOE/EPRI report [45]. The main challenge of using vanadium flow batteries is its high cost. Another important disadvantage of using vanadium is its toxicity both to plants and animals [46, 47]. The commercial status of various flow battery chemistries is shown in Table 2.1.

2+ 3+ ⇌ + 2+ + VO + V + H2O VO2 + V + 2H (2.3)

Table 2.1: Commercial status of ﬂow-battery startup companies, adapted from Ref. [48].

Company Funding Chemistry Stage Notes Ashlawn ARPA-E Vanadium Early 1-MW/8MWh plant Cellstrom Gildemeister Vanadium Shipping 10-kW/100 kWh Enervault $25 M Iron-chromium Demo Bankrupt EnStorage >20 M Hydrogen-bromine Early Israel EnSync NYSE Zinc-bromine Early Formerly ZBB ESS ARPA-E, VC All-iron Early $100/kWh target Imergy >100 M Vanadium Shipping SunEdison Lockheed N/A CNT electrodes Early Sun Catalytix H2 N/A Vanadium Early Korea ITN DOE Vanadium Early Grants Primus $60 M Zinc-bromine Shipping Kazakhstan PO Prudent N/A Vanadium Shipping Using Sumitomo RedFlow ASX Zinc-bromine Shipping Australia Rongke N/A Vanadium Shipping Stacks to UET REDT >$7M Vanadium Demo Ireland Sumitomo N/A Vanadium Shipping Market Leader? UniEnergy $20 M Vanadium Shipping PNNL, 1MW/4MWh Vanadis N/A Vanadium Early Rongke/UET ViZn $30 M Zinc-ferricyanide Shipping $200/kWh target WattJoule N/A Vanadium Early PNNL,strategic

Of the many diﬀerent materials that can be used to store energy in batteries, iron has the advantages of being among the safest and least-expensive. As illustrated in

7 CHAPTER 2. LITERATURE REVIEW

Figure 2.1, iron can be used at either the negative or positive electrode (or both), and in acidic or alkaline chemistries. iron-based ﬂow batteries

acid alkaline hybrid

FeCl-FeCl FeOH-FeCN FeCN-FeTEOA 1.21 V 1.4 V 1.34 V Hruska et al., “all-iron” Arotech Gong et al. ESS, Inc.

FeSO -H FeCl-CrCl 4 2 0.77 V 1.18 V Pupkevich et al. NASA, Enervault

FeCN-ZnOH FeCN-Quinone 1.8 V 1.2 V ViZn, Inc. Lin et al.

FeCl-VCl FeCl-ZnCl 1.35 V 1.53 V Wang et al. Selverston et al.

Figure 2.1: A Tree diagram of iron-based ﬂow batteries. Shaded circles represent batteries being investigated in this research.

2.2 All-Iron Hybrid Flow Batteries

The all-iron battery, which is is illustrated in Figure 2.2, uses the Fe0/Fe2+ couple on the negative side and the Fe2+/Fe3+ couple on the positive side [32]. The overall reaction in an all-iron ﬂow battery is given by Equation 2.4.

chg 3Fe2+ ⇌ Fe0 + 2Fe3+ E0 = 1.21 V (2.4) dis

Its advantages include safety, low-cost, moderate electrolyte pH and material abundance. In terms of the negative (plating) electrode, iron has another important advantage that the deposits tend to be smooth without the need for leveling

8 CHAPTER 2. LITERATURE REVIEW additives or pulse waveforms. If the all-iron system can be perfected, it can have an important impact in the energy storage community by enabling the intermittent power sources with cheap, safe energy storage [49, 50]. It may also help solve energy storage problems in developing countries. Several challenges, however, have limited the all-iron battery to the laboratory or demonstration scales. One of the most frequently-cited challenges of using all-iron batteries is the ten- dency to generate undesired hydrogen from the negative electrode. This is associated not only with a loss of energy eﬃciency, but also with the establishment of an electrolyte imbalance (sometimes called a state-of-charge or SoC imbalance) between the negative and positive bodies of electrolyte. Essentially, the negative electrolyte loses protons while the positive electrolyte gains a proportional amount of ferric ions. For this reason, some recent studies have focused on methods for mitigating hydrogen evolution [51–53].

separator

2+ * 0 Fe ) Fe Fe2+ )* Fe3+

FeCl2, FeCl2 storage FeCl3 storage

porous redox iron plate electrode

Figure 2.2: Schematic of an all-iron hybrid ﬂow battery, adapted from Ref. [32].

9 CHAPTER 2. LITERATURE REVIEW

2.3 Electrolytes

Iron battery electrolytes can be challenging to model because of high ionic strengths (I > 5 M) and the wide range of iron complexes that can form in such solutions. A typical iron-chloride electrolyte may have over 15 diﬀerent species in solution (e.g., see Table 2.2), some of which interact with each other. Finally, since battery operation is inherently dynamical, the thermodynamic properties (e.g., pH, speciation, etc.) can only be calculated for a given state-of-charge (SoC).

Table 2.2: Some of reactions that can occur in iron chloride solutions [54,55].

Reaction log Keq 3+ − Fe + 3Cl ⇌ FeCl3 1.13 3+ − ⇌ − Fe + 4Cl 4FeCl4 -0.79 2+ + ⇌ 2− Fe - 4H + 4H2O Fe(OH)4 -46 Fe3+ + Cl− ⇌ FeCl2+ 1.4 3+ − ⇌ + Fe + 2Cl FeCl2 2.1 3+ + 2+ Fe + H2O ⇌ H + Fe(OH) -2.19 3+ ⇌ + + Fe + 2H2O 2H + Fe(OH)2 -5.67 3+ ⇌ + 4+ 2Fe + 2H2O 2H Fe2OH2 -2.95 Fe2+ + Cl− ⇌ FeCl+ 0.14 2+ + + Fe + H2O ⇌ H + Fe(OH) -9.5 3+ ⇌ + 5+ 3Fe + 4H2O 4H + Fe3(OH)4 -6.3

The pH of iron chloride solutions has been summarized in the Dechema corrosion

handbook [56]. An interesting observation is that the pH of an oxygenated FeCl2 solution is considerably lower than that of an oxygen-free solution (see Figure 2.4 (a)),

and that is because the ferrous ions rapidly oxidize to form FeCl3, which is a much

stronger acid. It has also been reported that the pH of FeCl2 solutions is aﬀected by the concentration of chloride in the solution, as shown in Figure 2.4 (b). In solutions with pH < 4, the rate of oxidation of ferrous ions by air can be estimated [57] by

10 CHAPTER 2. LITERATURE REVIEW using Equation 2.5.

dcFe2+ −5 − = 6 · 10 c 2+ (2.5) dt Fe

One of the challenges in developing hybrid iron flow batteries is preventing the precipitation of iron hydroxide precipitates from forming in the electrolyte. Part of the challenge stems from the ambiguous nature of the precipitated species, which may be an iron oxyhydroxide (e.g., ferrihydrite, goethite, hematite, etc.) [58–61]. Forma- tion of such solid phases can lead to performance degradation during flow battery operation; several mechanisms are feasible, and are not mutually-exclusive. Perhaps the most straight-forward effect is by clogging of the pores of the battery separator. This would be expected to reduce conductivity and therefore increase ohmic overpotential. It is also possible that the precipitates might interfere with adsorption of iron ions at the electrode surface. The exact conditions (viz., the pH, SoC) under which the precipitate may form are not clear. One way is through the direct oxidation of ferrous ions from air; if oxygen is present, the precipitate can form via Equation 2.6, as described in Ref. [62].

1 1 Fe2+ + O + 2OH− + H O → Fe(OH) (s) (2.6) 4 2 2 2 3

Generally, formation of the solid hydroxide is normally thought to be the last step in a series of hydrolysis reactions :

3+ → 2+ + Fe(H2O)6 + H2O Fe(H2O)5(OH) + H3O (2.7) 2+ → + + Fe(H2O)5(OH) + H2O Fe(H2O)4(OH)2 + H3O (2.8) + → 0 + Fe(H2O)4(OH)2 + H2O Fe(H2O)3(OH)3 + H3O (2.9)

First order estimates suggest that such electrolytes form such precipitation above

11 CHAPTER 2. LITERATURE REVIEW

about pH=2, as shown in Figure 2.3 (a). An example showing how ferric ions can react to form solid hydroxides is illustrated in Figure 2.3 (b).

0 OH/Fe

Fe3+ 0 -4 solid oxyhydroxides

-8 4+ 2.0 0 Fe2(OH)

Fe(OH)3 2 ion activity / log M

-12 Fe(III)oxyhydroxysalts − 2.7-2.8 4 sulfate-form nitrate-form Fe Fe(OH) Fe Fe Fe(OH) Fe(OH) Schwertmannite

3+ 3 2 -16 OH OH + 2-line ferrihydrite 2.9-3.0 4+ 2 5+ 2 2+ 4 6-line ferrihydrite

2 6 10 3.0 pH hematite goethite

(a) (b)

Figure 2.3: Phase diagram for iron(III) in water, adapted from Ref. [63], and (b) mechanism for hydroxide formation, adapted from [59].

The pH of iron chloride electrolytes is important to consider because it aﬀects electrode kinetics, the hydrogen evolution reaction and iron complexation. Ferrous

chloride (FeCl2), is a weak acid. However, even small amounts of oxygen from the

air can react to form ferric chloride (FeCl3), which is a fairly strong acid, as shown in Figure 2.4 (a). Also, it has been reported that increasing concentration of chloride ions can increase the activity of protons in solution, as shown in Figure 2.4 (b). Ferric chloride, which is generated in the positive electrolyte during battery charging, is a considerably stronger acid. As shown in Figure 2.5, reported pH values for ferric chloride solutions are almost as low as the theoretical pH for hydrochloric acid. Therefore, it can be expected that in an all-iron ﬂow battery, the positive electrolyte pH tends to be lower than the negative electrolyte pH.

12 CHAPTER 2. LITERATURE REVIEW

1e+01 sat'd

1e+00 3.4

3.2

-1 1e-01 3

1e-02 2.8 ]/ mol L 2 2.6 pH

[FeCl 1e-03 2.4

2.2 1e-04 with O2 without O2 2

1.8 1e-05 0 1 2 3 4 5 6 0 1 2 3 4 5 6 7 8 9 10 [Cl-]/ mol L-1 pH (a) (b)

Figure 2.4: (a) Natural pH of ferrous chloride tetrahydrate with and without oxygen in solution and (b) eﬀect of chloride ion concentration on measured pH value of FeCl2, adapted from [56].

1.5

0.5 pH

(iii) 0 (i) (ii) -0.5

-1 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 -1 [FeCl3] / mol L

Figure 2.5: Measured pH of FeCl3 solutions: (i) from Ref. [49], (ii) from Ref. [56] at 295 K, (iii) Ref. [56] at 323 K. The dashed line represents the theoretical pH for the equivalent concentration in hydrochloric acid calculated as pH=-log10[HCl].

Aqueous battery electrolytes always use concentrated solutions of metal salts and supporting electrolytes, so understanding the activities of the dominant species is critical for understanding and predictability of electrolyte properties. Perhaps the most important activity to be able to calculate and measure in the iron ﬂow battery

13 CHAPTER 2. LITERATURE REVIEW is that of the hydronium ion; this is because pH is the strongest factor governing both oxyhydroxide precipiation and coulombic efficiency of the battery. Despite the importance of pH, its measurement in concentrated electrolytes (brines) is non-trivial [64–66]. The pH can be measured spectrophotometrically [67], but this requires considerable hardware and analysis. The internationally-accepted standard method is by potential measurements between a reversible hydrogen electrode (RHE) and a silver/silver-chloride wire in a junctionless apparatus known as Harned’s cell [68]. Even using the Harned cell, the calculated pH depends on the arbitrary assignment of the chloride ion activity, which can be calculated using a variety of activity coefficient models. An even bigger challenge is the interference cause by the Fe(II)/Fe(III) redox couple at the platinum electrode. Surprisingly, there are few references that investigate pH, whether measured or calculated, in iron chloride electrolytes. When there is no oxygen present in solution, the pH is in the range of 4-5 at high concentrations (ca. 1-3 M) of FeCl2. However, even small amounts of oxygen in the electrolyte have a large effect by oxidizing Fe2+ into Fe3+, which acts as strong acid through hydrolysis. Therefore, for practical purposes, the measured pH for ferrous chloride solutions is always lower than what would be expected from the simple FeCl2 equilibrium calculations. For a given starting composition, calculating the pH and other electrolyte properties requires iterative solution to the nonlinear system of equations (chemical equilibria, activities, mass balances and charge balances), as carried out by M.S. Lee [54, 55, 69] for metallurgical applications. The solution is found using a Newton-

Raphson iteration technique. For example, the FeCl2-H2O-HCl system can be described using an iron mass balance,

2+ + 0 + [FeCl2]t = [Fe ] + [FeCl ] + [FeCl2] + [FeOH ] + [Fe(OH)2], (2.10)

14 CHAPTER 2. LITERATURE REVIEW

a chloride mass balance,

− + 0 2[FeCl2]t + [HCl] = [Cl ] + [FeCl ] + 2[FeCl2] , (2.11)

a charge balance,

[H+] + 2Fe2+ + [FeCl+] + [FeOH+] = [Cl−] + [OH−], (2.12)

an activity coeﬃcient model (Bromley equation [70]), where I is the ionic strength and B is an ion interaction coeﬃcient,   0.51z2 I0.5 ∑ (0.06 + 0.6 · B˙ ) (|z | + |z |)2 − M  ( )MX  · M X log γM = 0.5 + 2 [X], (2.13) 1 + I 5I 4 X 1 + |zM zX |

and lastly the chemical equilibria, shown in Table 2.3. All together, this system includes the 17 equations required to solve for the 17 variables, which are 8 species concentrations and 9 activity coeﬃcients [69].

Table 2.3: Equilibrium constants for acid ferrous chloride, from Ref. [69].

reaction log K + − H + OH ⇌ H2O 14 Fe2+ + Cl− ⇌ FeCl+ 1.9 2+ − ⇌ 0 Fe + 2Cl FeCl2 0.068 Fe2+ + OH− ⇌ FeOH+ 7.23 2+ − Fe + 2OH ⇌ Fe(OH)2 8.33

Software used by the geochemistry community is relatively mature and well- validated, and can provide many advantages compared to writing custom programs. PHREEQC was used here because it is powerful and easy to use [71]. Although there is no ability to use the Bromley activity coefficient model, PHREEQC does include an activity coefficient model for high ionic strength electrolytes based on specific-ion in-

15 CHAPTER 2. LITERATURE REVIEW

teraction theory [72]. We have found that by using the SIT activity coeﬃcient model along with the equilibrium constants provided by Ref. [69], it was possible to get a reasonably close agreement between calculated and measured pH values for various electrolyte compositions (see Table 2.4 and Figure 2.6).

Table 2.4: Measured and calculated pH of FeCl2-HCl-H2O solutions, adapted from Ref. [69]. The last column shows calculated values using this work.

n FeCl2/m HCl/m pH (meas) pH calc [69] pH calc (this work) 1 0.1 0.2 0.83 0.82 0.54 2 0.21 0.33 0.6 0.59 0.27 3 0.31 0.41 0.49 0.49 0.13 4 0.45 0.52 0.33 0.37 -0.01 5 0.52 0.63 0.22 0.28 -0.08 6 0.63 0.73 0.15 0.2 -0.15 7 0.75 0.85 0.01 0.1 -0.22 8 0.98 1.09 -0.15 -0.05 -0.33 9 1.05 1.16 -0.21 -0.1 -0.35 10 1.23 1.34 -0.35 -0.19 -0.41 11 1.48 1.59 -0.42 -0.33 -0.48 12 1.63 1.73 -0.49 -0.39 -0.51 13 1.74 1.86 -0.6 -0.45 -0.54 14 1.96 2.03 -0.67 -0.52 -0.57 15 2.02 2.14 -0.78 -0.57 -0.59

0.5 M.S. Lee 2004

0 pH (calculated) -0.5 PHREEQC

perfect fit line -1 -1 -0.5 0 0.5 1 pH (measured)

Figure 2.6: Comparison between between measured and calculated pH values using diﬀerent methods (data from Table 2.4)

16 CHAPTER 2. LITERATURE REVIEW

2.4 Iron Plating Electrodes

The negative electrode of the all-iron hybrid ﬂow battery is the Fe(II/0) couple (see

Equation 2.14), which has a standard potential of -0.44 VSHE.

Fe2+ + 2e ⇌ Fe0 E0 = −0.44 V (2.14)

A review of iron electrodeposition can be found in Ref. [73]. Because of its low cost and useful mechanical and magnetic properties, it has been electroplated in a wide range of applications, typically from either sulfate or chloride baths. However, most studies have focused on its use as a coating rather than as a battery electrode, so not all early studies are useful for energy storage applications. Iron electrodeposition for ﬂow batteries was investigated in 1981 by Hruska and Savinell [32] and more recently by Hawthorne et al. [74, 75] and Manohar et al. [53], as summarized in Table 2.5 .

Most often, NH4Cl is used as the support because its high conductivity and solubility.

Table 2.5: Studies of iron plating in ﬂow batteries

Year Author Electrolyte Ref

1981 Hruska and Savinell FeCl2+NH4Cl [32] 2014 Hawthorne et al. FeSO4, FeCl2, NaCl, NH4Cl [74] 2015 Hawthorne et al. FeSO4, FeCl2, NaCl, NH4Cl [75] 2016 Manohar et al. FeCl2, NH4Cl, various additives [53]

Iron electrode kinetics and corrosion properties were studied from both chloride- and sulfate-based electrolyte baths in the pH range of 1.2-4.9 [76]. It was found that the anodic (dissolution) kinetics were slower than that for deposition, the average i0 being on the order of 1·10−7 A cm−2 for anodic polarization compared to 6·10−6 A cm−2 for cathodic polarization at pH = 2.0. However, this trend appeared to reverse with increasing pH. For example, at pH = 4, the cathodic i0 was greater than the

17 CHAPTER 2. LITERATURE REVIEW

anodic i0. As shown in Table 2.6, the exchange current density of the iron electrode was found to be strongly inﬂuenced by the anions present; this was attributed to relative adsorption eﬀects. That is, iron exchange current densities were higher in

2− the presence of weakly adsorbed anions (e.g., SO4 ), and lower in the presence of strongly adsorbed anions (e.g., Cl−).

Table 2.6: Eﬀect of anions on iron exchange current densities at pH=3.0, from Ref. [76]

8 2 salts i0 ∗ 10 (A/cm )

FeClO4+NaClO4 50 FeSO4+Na2SO4 30 FeCl2+KCl 8 FeAc2+NaAc 1 Fe(NO3)2+ KNO3 0.02

One of the sources of coulombic eﬃciency losses is the undesired corrosion of plated iron metal. There are two ways this can occur. At low pH, iron corrosion occurs mainly due to the reduction of protons in solution (acid corrosion), according to the reaction shown in Equation 2.15. Corrosion rates as a function of pH are shown

in Figure 2.7 for 0.5 M FeSO4 + 0.5 M Na2SO4 [76], as well as for iron chloride in

NH4Cl [32]. In both studies, the corrosion rate showed a logarithmic dependence on the measured solution pH.

0 + 2+ Fe + 2H → 2Fe + H2 (2.15)

Iron can also undergo the so-called comproportionation reaction, whereby metallic iron is oxidized by ferric ions in solution as shown in Equation 2.16. Eﬃciency losses from this mechanism are associated with undesired crossover of ferric ions from the positive side of the cell through the separator and into the negative electrolyte.

Fe0 + 2Fe3+ → 3Fe2+ (2.16)

18 CHAPTER 2. LITERATURE REVIEW

100 (i)

10 -2

1 (ii)

0.1

iron corrosion rate / mA cm 0.01

0.001 -1 0 1 2 3 4 5 pH

Figure 2.7: Corrosion rate of iron as a function of pH in (i) in 2 M NH4Cl, from Ref. [32] and (ii) 0.5 M FeSO4+ 0.5 M Na2SO4, from Ref. [76]. Oxygen content in the electrolytes was not reported.

Another challenge of using all-iron batteries is the generation of hydrogen on the negative electrode via Equation 2.17. This is distinguished from iron corrosion in that it is part of the battery charging circuit and is not accompanied by dissolution of iron metal. The nature of the diﬀerence between the hydrogen side reaction during charging and iron corrosion during discharging is illustrated using the Pourbaix diagrams, as shown in Figure 2.8. Even when the majority of electrons go to the desired electrodeposition reaction, some of the current always goes to hydrogen generation. The hydrogen is not only a problem because of the decrease in coulombic eﬃciency, but it also causes the battery performance to degrade over time due to the changes in the electrolyte chemistry. A chemical imbalance forms, wherein the positive electrolyte becomes over-concentrated in ferric ions and depleted in ferrous ions. Simultaneously, the loss of protons from solution leads to an increase in pH, which leads to formation of iron hydroxides. The hydroxides precipitate out of solution and cause a variety of problems such as increased viscosity, electrode fouling, and a decrease in the number of iron ions available

19 CHAPTER 2. LITERATURE REVIEW

3+ Fe (aq) O2 + 4H + + 4e * + ) 2H 2 O 0.77 +

Fe(OH)3 (s) SHE

charging 2+

discharging Fe (aq) E/V 2H + + 2e )* H2 -0.44 - Fe(OH)2 (s) - Fe0 (s)

2 7

Figure 2.8: Pourbaix diagrams showing approximate electrode potentials of the all- iron hybrid ﬂow battery during (a) charging and (b) discharging modes of operation.

for the desired reactions.

+ 0 2H + 2e → H2 E = 0.0 V (at pH=0) (2.17)

2.5 Rebalancing

Electrolyte rebalancing was implemented in NASA’s iron chromium flow batteries (see Figure 2.9). In fact, the rebalance cell was considered to be “probably the most important system feature of the NASA redox energy storage system” [29]. Through- out the six years of NASA’s flow battery development, a large portion of their efforts, both in their reports and in their patents, was dedicated to this area. Their rebalancing fuel cell, which is illustrated in Figure 2.10 (a), was generally able to carry out the rebalancing. However, numerous problems eventually caused NASA to abandon the fuel cell design in favor a new approach wherein the hydrogen-consuming electrode

20 CHAPTER 2. LITERATURE REVIEW

was replaced with a chlorine-generating electrode, as shown in Figure 2.9 (b) [77]. That is, whereas the original rebalancing reaction was given by Equation 5.1, the new reaction was given by Equation 2.18. This introduced many new challenges, though, related to the production of chlorine gas. In fact, for a sealed system, the chlorine approach would require yet another rebalancing cell to recombine hydrogen with the chlorine. It is also noteworthy that the chlorine-based rebalance scheme only restores balance to the iron, but not the hydrogen.

3+ − 2+ 2Fe + 2Cl → 2Fe + Cl2 (2.18)

A passive rebalancing scheme similar to the work described here, as shown in Figure 2.10 (b), was described in 2013 by Whitehead et al. for vanadium flow batteries [78]. However, there are several important differences. For example, the design studied in the present work is based on three-dimensional porous felt rather than two-dimensional paper and does not require the use of a membrane. Also, this design extends vertically into the headspace, which has advantages including greater current per cross-sectional area of electrolyte. Lastly, the reactor is used for a different reaction (in the iron, rather than the vanadium battery electrolyte).

2.6 System Modeling

There is a large body of literature that involves flow battery modeling, but most of it has focused on all-vanadium flow batteries. For example, the Walsh group at Southampton has developed a variety of models for vanadium flow batteries with an emphasis on predicting the charge-discharge polarization [80–84]. They were success- fully able to match predicted to simulated battery behavior under a wide range of conditions. These studies focused on short-term behavior on the order of one and ten charge-discharge cycles.

21 CHAPTER 2. LITERATURE REVIEW

KOH

Cr Fe vent pos neg N2

- + + - Cl

rebalance cell redox cell rebalance cell H2 redox cell (a) (b)

Figure 2.9: (a) Original NASA iron-chromium system with external rebalancing system using hydrogen tank, adapted from [29], and (b) second-generation rebalancing system using chloride electrolysis, adapted from [30].

membrane electrolyte reservoir gasket

H2 electrode iron

outlet H2 inlet

catalyzed carbon paper microporous polypropylene O-ring

iron H2 outlet inlet positive electrolyte graphite endplate felt iron H2 electrode diﬀuser current collector (a) (b)

Figure 2.10: NASA single rebalance cell, from Ref. [79], and (b) a passive design by Whitehead et al. for vanadium batteries [78].

Crossover and capacity fade have been investigated for vanadium batteries by the Kumbur group at Drexel [85–88]. These multiphysics models, typically implemented in Comsol, were used mainly to identify operating conditions to minimize the unwanted eﬀects of species crossover. In their recent paper, it was shown that a zero-dimensional transient model was appropriate for obtaining suﬃciently-accurate

22 CHAPTER 2. LITERATURE REVIEW results with minimal required computational time [88]. The Skyllas-Kazakos group of the University of New South Wales has described system-level simulations of heat and mass transport in all-vanadium flow batteries [89–92]. The purposes of these simulations included predicting capacity fade through crossover as well as understanding electrolyte temperatures as functions of, for example, electrolyte flow rate. The studies described in reference [90,91] included the hydrogen side reaction, non-isothermal effects and unequal charging and discharging currents, also incorporating rest periods for 15 kWh battery systems. However, they did not consider long term equilibrium behavior or crossover losses. For instance, a constant loss of hydrogen would be expected to change the electrolyte properties (e.g., the pH) over time. Recently, a study by the United Technologies Research Center described a system- level transport model that compared the effects of various transport mechanisms on coulombic efficiency losses in both all-vanadium and hydrogen-bromine systems [93]. There have been fewer studies of other flow battery types. To the best of the author’s knowledge, there have been no studies describing system modeling of all-iron batteries. Most vanadium flow batteries use concentrated acid solutions along with proton exchange membranes (PEM) such as Nafion, and this leads to physical phenomena that don’t necessarily apply to other electrolytes like iron chloride. In vanadium systems, for example, ionic current is carried almost entirely by protons, but in iron chloride battery electrolytes charge is carried by chloride, potassium, protons and iron ions. The flux of protons through PEMs is known to cause an important electro- osmotic flux that has significant effects that are not as important when using microporous separators. Although not investigated here, this type of model can also be used to study other design aspects such as unequal charging/discharging currents on equilibrium and crossover losses, optimal starting electrolyte composition and optimal reservoir sizes. Also, no system models including the hydrogen evolution side reac-

23 CHAPTER 2. LITERATURE REVIEW tion and rebalancing have been described in the literature. Therefore, in this work, a model is developed for the sealed recombinant battery, including side reaction and hydrogen-ferric ion recombination.

2.7 Zinc-Iron Electroplating

The term “anomalous codeposition” has been credited to a 1963 textbook by Brenner, who deﬁned it as one of ﬁve categories of codeposition “characterized by the anomaly that the less noble metal deposits preferentially [94].” In mixed zinc-iron electrolytes, for example, iron is more noble than zinc and therefore might be expected to plate out predominantly (see Figure 2.11). Although ACD has not been exploited in the battery literature, the phenomenon has been widely reported by researchers in the alloy- plating community [95–97]. The most commonly-cited mechanism for explaining the behavior is called the hydroxide suppression mechanism (HSM), which was proposed by Dahms and Croll in 1965 for mixed Fe-Ni systems [98]. A related hydroxide oscillation model was proposed by Yan et al. [99].

-0.77 -0.44 0 0.77

2+ 0 2+ 0 + 2+ 3+ Zn /Zn Fe /Fe H /H2 Fe /Fe

Figure 2.11: Relative positions of zinc and iron couples on the hydrogen scale.

According to the HSM, the anomalous behavior is caused by a hydroxide ﬁlm

(e.g., Fe(OH)2 or Zn(OH)2) formed at the deposition surface due to locally-high pH from hydrogen evolution [100–104]. An illustration of this process is shown in Figure 2.12. The hydroxide layer is thought to increase the overpotential for reduction of the more noble metal (e.g., Ni) more so that it does for reduction of the less noble metal (e.g., Zn). This has been explained by considering that the more noble metal has to diﬀuse through a Zn(OH)2 ﬁlm before it can be reduced whereas the zinc reduction

24 CHAPTER 2. LITERATURE REVIEW

occurs be direct reduction of the hydroxide ﬁlm as per Equation 2.19, as described in Ref. [99].

− Zn(OH)2 + 2e → Zn + 2OH (2.19)

4 Zn(OH)2

pH at electrode surface 2

− Fe 10

8 2+ cathode [OH Zn − 6 (Zn) ] 2+ Zn 4 H+ 2 Ni −

OH deposition rate / mA cm 2+ Fe -600 -700 -800 electrode potential / mV vs NHE (a) (b)

Figure 2.12: Illustration of the hydroxide suppression mechanism for preferential electrodeposition of zinc from mixed Zn-Co electrolytes, adapted from Ref. [99]. Cor- relation between calculated surface pH and anomalous deposition in Fe-Ni system, adapted from Ref. [98].

Several authors, though, have reported results inconsistent with the HSM [100,102, 105,106]. For example, Eliaz et al. [106] found no evidence of zinc hydroxide ﬁlms from XPS and XRD experiments during deposition of Zn-Ni, a similar anomalous system. There are a wide variety of proposed mechanisms, including those based on mixed- potential theory [96], Fermi levels [107], proton adsorption [100], ratio of adsorbed hydroxide intermediates (e.g., Fe(OH)+, Ni(OH)+)[108], relative atomic radii sizes [109], exchange current density arguments [106] and underpotential deposition (UPD) eﬀects [110]. Despite the large number of investigations, there is still no universally-accepted mechanism for anomalous codeposition [102,111]. Furthermore, the mechanism may be unique to each system; Nakano et al. concluded that the important inhibitor

25 CHAPTER 2. LITERATURE REVIEW

in the Zn-Ni system was Zn(OH)2, whereas in the Fe-Ni system the inhibitor was

FeOHad [108]. The Zn-Fe system has been studied experimentally in sulfate [95] and chloride [107,109,112] baths.

Table 2.7: Proposed mechanisms for anomalous codeposition.

year pH system media mechanism reference 1965 * Zn-Ni - HSM [98] − 1981 2-10 Zn-Co SO4 HSM [113] − − 1993 1-5 Zn-Ni/Zn-Co/Zn-Fe SO4 /Cl modiﬁed HSM [112] 1994 * A-B * mixed-potential theory [96] − 2+ 1995 2-5 Zn-Co Cl Znad [105] − 1996 2 Zn-Co SO4 HSM [99] − 2+ 1999 3 Zn-Fe Cl Znad [114] 2000 4.8 Zn-Ni Cl− UPD [110] 2001 3-5 Zn-Fe Cl− Fermi levels [107] − − + 2002 1.5-5 Zn-Fe Cl ,SO4 Had [100] − 2004 1-3 Fe-Ni SO4 HSM [108] − 2004 1-3 Fe-Ni SO4 FeOHad [108] 2006 4.5 Zn-Fe Cl− atomic radius [109] 2007 3.5 Zn-Co Cl− ∆i0 [115] 2010 3.5 Zn-Ni/Zn-Co Cl− ∆i0 [106] − + 2013 5.0 Zn-Ni Cl Znad [102]

The present work uses acidic chloride baths mainly because chloride electrolytes are known to be more conductive than sulfate baths [116]. Furthermore, it was found that current efficiency was generally higher for the chloride system, which more strongly inhibited iron co-deposition, particularly at lower current densities [112]. Anomalous deposition was reportedly favored in chloride media by Diaz et al. [100], and advantages of chloride were also noted by Gomez et al. [114]. Electrodeposition from mixed solutions of zinc chloride and iron chloride has been described in several studies [100,101,112,114,116–118]. Plating efficiencies from mixed electrolytes without additives were measured at pH=3 as functions of current density and chloride concentration by Fukushima et al. [112], who found reported higher current efficiency

26 CHAPTER 2. LITERATURE REVIEW

Table 2.8: Compositions and eﬃciencies of acidic Zn-Fe alloy plating baths

year ref zinc salt iron salt support pH eﬃciency

1993 [112] 0.5 M ZnCl2 0.5 M FeCl2 4 M NaCl 3 90-100 1998 [117] 0.8 M ZnCl2 0.8 M FeCl2 200 g/l NH4Cl 2-3 90-98 1999 [114] ZnCl2 FeCl2 NaCl 3 N/A 2001 [107] ZnCl2 FeCl2 KCl 3-5.5 N/A 2002 [100] ZnSO4 FeSO4 none 1.5, 3.0, 5.0 40-100 2002 [119] ZnSO4 FeSO4 none 2.5 N/A 2006 [109] ZnCl2 FeSO4 KCl 4.5 N/A 2007 [120] ZnSO4 FeSO4 Na3C6H5O7 4 2010 [101] 0.37 M ZnCl2 0.04 M FeCl2 NH4Cl+KCl 3.5 90-95 2013 [118] ZnCl2 FeCl2 Boric Acid 2 “very high” 2013 [121] ZnSO4 FeSO4 Na3C6H5O7 3-6 N/A 2013 [103] ZnSO4 FeSO4 Na2SO4 2.5 ≤ 90 %

for chloride- than for sulfate-based baths. All the previous literature on electrodeposition from mixed zinc-iron electrolytes relates to the development of coatings with improved corrosion resistance. However, the requirements for possible use in battery applications are considerably diﬀerent. Since zinc has better thermodynamic and kinetic properties than iron as a negative electrode, it may be desirable to minimize the amount of iron in the deposit rather than trying to incorporate more of it. Also, unlike in plating applications, it is important to have suﬃciently high concentrations of each salt in the battery electrolyte for energy density reasons. Finally, additives have to be chosen more carefully since they have more stringent stability requirements, and should not interfere with the separator or accelerate corrosion during battery discharging.

27 CHAPTER 2. LITERATURE REVIEW

2.8 Accelerated Lifetime Testing

Considering the nature of the applications for flow batteries, lifetime performance (i.e., the rate of performance degradation) is one of the most important aspects of flow battery engineering. Unless the batteries can maintain good performance and efficiency with low maintenance for long periods of time, flow batteries cannot compete with other energy storage technologies. Studying lifetime and degradation of flow batteries, however, is challenging since it is time-consuming; it can take one month to run a 100-cycle test with representative SoC swings. For this reason, there have not been many lifetime studies that show flow batteries working for more than about 50 cycles. Only a small number of flow battery studies have included rebalancing methods (see Table 2.9). The only publication describing extensive lifetime testing (i.e., charge discharge cycling) data of flow battery systems was published by Sandia in 1999 [122]. This report showed hundreds of cycles completed on dozens of zinc-bromine hybrid flow battery stacks, some of which lasted over 1000 cycles. The vast majority of flow battery literature shows only between 1-100 cycles, with only a few papers describing systems that were cycled between 100-1000 times. However, there have been no papers describing extensive cycling data for all-iron flow batteries.

28 CHAPTER 2. LITERATURE REVIEW

Table 2.9: Selected ﬂow battery lifetime studies

Year System # Cycles Rebalancing Ref 1981 All-Iron 57 acid-dosing [32] 1985 Iron-Chromium >500 Fe-Cl [30] 1987 Iron-Chromium 200 no [123] 1992 Zinc-Bromine 200 no [124] 1999 Zinc-Bromine 290 no [125] 1999 Zinc-Bromine 1200 no [122] 2000 All-Vanadium 500 no [126] 2008 Copper-Lead Acid 400 no [127] 2012 All-Vanadium 120 yes [128] 2012 All-Vanadium 40 yes [129] 2012 Iron-Vanadium 100 no [130] 2013 Zinc-Bromine 70 no [131] 2013 Zinc-Bromine 100 no [132] 2016 All-Iron 50 no [53]

29 Chapter 3

Dissertation Research

30 CHAPTER 3. DISSERTATION RESEARCH

3.1 Electrolyte Rebalancing

The hydrogen evolution side-reaction leads not only to a decrease in coulombic efficiency, but also a bulk electrolyte imbalance (or state-of-charge imbalance) [32]. Therefore, several recent studies have described efforts to minimize the amount of hydrogen generated by through the use of different additives [53, 75]. While the additives were proven to help, none completely stopped the hydrogen. Until a method is found to completely suppress the hydrogen side reaction, electrolyte rebalancing systems will need to be employed in order for electrolytes to be usable in the batteries for long periods of time. Many different types of such rebalancing schemes have been proposed [78, 79, 133–135]. One promising approach that has not been widely investigated is to operate flow batteries as sealed, recombinant systems, i.e., as closed systems. Although it may not be possible to operate a flow battery with “perfect” seals, it is possible to greatly improve the electrolyte balance by using internal recombination, which has the advantage of being self-regulating- in principal, this approach can maintain perfect chemical balance without the need for pH measurement, acid-dosing or control systems. Furthermore, it may be possible to carry out the necessary rebalancing processes within the positive electrolyte reservoir so that the system footprint is not changed. The overall rebalancing reaction is shown in Equation 3.1, where the positive standard cell potential indicates a spontaneous reaction. Chapter 5 describes a three-dimensional felt-based reactor, or capillary-action galvanic reactor (CGR) designed to carry out this reaction effectively. Hydrogen can react in the reservoir head space and ferric ions can be reduced in the electrolyte. Pro- tons and electrons, liberated by hydrogen oxidation, can move vertically downwards (in the in-plane direction) through the electrolyte contained in the felt.

3+ + 2+ 0 2Fe + H2 → 2H + 2Fe E = 0.77 V (3.1)

31 CHAPTER 3. DISSERTATION RESEARCH

3.2 System Modeling

In order to help design flow battery systems, it would be helpful to have system models that can predict how the electrolytes will change over time. However, such models have not been widely described, particularly for iron-based flow batteries. Furthermore, related flow battery models typically neglect side reactions and the associated electrolyte rebalancing. However, the side-reactions have important effects, and without rebalancing the battery systems cannot achieve steady-state operation. Therefore, Chapter 4 describes the nature of the chemical imbalance problem and develops a dynamical flow battery model that incorporates the hydrogen evolution side reaction as well as the rebalancing reaction. The model is built using a system of differential-algebraic equations (DAE), which is simulated using free and open- source software. The model is used to help understand the nature of the electrolyte imbalance problem and the effect of rebalancing, as well as the effects of changing some of the tunable parameters.

3.3 Zinc-Iron Chloride Flow Batteries

All-iron flow batteries show great promise as a sustainable energy storage technology, but their progress has been limited largely by the sluggish kinetics of the negative iron plating electrode [32]. Similarly, zinc-bromine flow batteries have enjoyed extensive research and development, but the challenge of bromine toxicity has been an important drawback. Therefore, a new approach explored in this work is the development of zinc-iron chloride flow batteries, which could combine the negative zinc electrode of

the Zn-Br2 battery and the the positive iron electrode of the all-iron battery. It would be particularly advantageous if such a zinc-iron battery could operate using common, or mixed electrolytes-this would have important implications in terms of cost (e.g., the ability to use porous separators rather than ion-exchange membranes), and long

32 CHAPTER 3. DISSERTATION RESEARCH

device lifetime due to the crossover tolerance. Although zinc and iron have been used in flow batteries before, all previous methods required separated (un-mixed) electrolytes and expensive ion-exchange membranes and/or ferricyanide [17,36,136,137]. In Chapter 6 of this thesis, studies of a zinc-iron battery with common, or mixed electrolytes are carried out. Three-electrode studies, as well as flow battery tests, are performed in order to understand the electrodeposition from mixed electrolytes as well as the battery performance. It is shown that the zinc is preferentially deposited according to an anomalous codeposition (ACD) mechanism, wherein iron electrodeposition is strongly inhibited. These processes are investigated using steady-state polarization, cyclic voltammetry, and rotating-disc electrodes. A complete flow battery is tested at current densities of 20-40 mA cm−2 and loadings of 12.5-100 mAh cm−2.

33 Chapter 4

Model for Sealed Flow Batteries

34 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES

4.1 Introduction

Iron is an attractive element to use in energy storage applications because of its safety, sustainability and low-cost. Therefore, there has been renewed interest in the development of all-iron hybrid flow batteries [49, 51, 53, 74, 75, 138, 139]. All-iron batteries operate using the three available redox states of iron, Fe0, Fe2+ and Fe3+ (see Table 4.1). As with some other aqueous flow batteries, they can experience significant rates of hydrogen generation (ca. 1-10 % of the nominal operating current density). This hydrogen evolution represents a loss of protons from the electrolyte and it leads to a chemical imbalance with each charge-discharge cycle. However, the imbalance can be addressed by using an appropriate rebalancing (or “recombination”) reactor, which may be a fuel cell, packed-bed, or other type of reactor [29,30,78,79,135,140,141]. For example, a rebalancing fuel cell was developed by NASA for use in the iron-chromium redox flow battery [29]. It facilitated the spontaneous reaction between hydrogen and ferric (Fe3+) ions, given by Equation 5.1, and was considered to be one of the most important features of the flow battery system [29]. Equation 5.1, a spontaneous, hydrogen-ferric ion recombination reaction, has a standard cell potential of 770 mV.

3+ 2+ + H2 + 2Fe → 2Fe + 2H (4.1)

Whereas in the NASA system, the hydrogen was normally supplied from an external tank (known as “external rebalancing”), it may also possible to operate the battery using “internal” rebalancing, whereby the battery operates as a sealed and self-regulating system [29]. In principal, the chemical balance of the electrolytes can be completely restored by using this method. While there has been considerable development of different types of rebalancing reactors, there is a gap in the literature in terms of understanding and modeling their effects in sealed flow batteries. Therefore, this study focuses on measurements and modeling the electrolyte dynamics in sealed

35 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES

Table 4.1: All-iron hybrid battery reactions

electrode reaction E0/V negative Fe2+ + 2e ⇌ Fe0 -0.44 positive Fe3+ + e ⇌ Fe2+ 0.77 cell Fe0 + 2Fe2+ → 3Fe2+ 1.21

iron flow batteries when using internal rebalancing. In particular, this work considers using internal rebalancing with floating reactors that carry out Equation 4.1, as illustrated in Ref. [140]. A semi-empirical model has been developed by using a system of differential and algebraic equations (DAE) [142], which were solved numerically.

4.2 Materials & Methods

−3 2 For battery charge-discharge cycling tests, a single-cell battery (Ageo = 3·10 m ) was used with ﬂow provided by peristaltic pumps ﬂowing at 25 mL min−1. The negative

electrolyte (V = 125 mL) consisted of 2.25 M FeCl2 and 2.0 M KCl, and the positive

electrolyte (V = 250 mL) contained 1.5 M FeCl2, 2.0 M KCl, 0.5 M FeCl3 and 250 mM HCl. These compositions were chosen to approximate the electrolytes at 50 % state- of-charge (SoC), where additional HCl was added both to improve the kinetics of the Fe2+/3+ redox couple and to help prevent formation of solid precipitates. The battery was charged and discharged at 1000 A m−2, where the each charge was carried out for one hour. Discharges were carried out until the cell reached a cutoff voltage of 0 V. The separator was made of microporous polyethylene (Daramic, thickness = 175 µm) coated with poly-vinyl alcohol (PVA, thickness = 25 µm on each side); the PVA coating helped mitigate hydraulic crossover [143]. The electrodes were made using porous graphitic felts (SGL) bonded to graphite plates as described in Ref. [74]. Both electrodes operated in a flow-through configuration. In the case of a sealed flow battery without recombination, the hydrogen gen-

36 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES

eration rates during charging and discharging were calculated based on the rate of pressure increase over time inside a known volume. In a system with internal rebalancing, a capillary-action galvanic reactor (CGR) was placed in the positive electrolyte reservoir to carry out hydrogen-ferric ion recombination, and the reservoir headspace volumes were connected using a length of plastic tubing (l = 6”, OD = 3/8”) [140].

Electrolyte pH was adjusted by drop-wise addition of HCl (10 %w), and measured ⃝ using a Pinpoint R combination electrode (American Marine, Inc.). The primary criterion for reaching steady-state operation was that the rate of hydrogen generation at the negative electrode became equal to the rate of consumption ¯ by the recombination reactor as per Equation 4.2, where IH2,gen is the average rate ¯ of hydrogen generation at the negative electrodes and IH2,cons is the average rate of consumption by the rebalancing cell.

|¯ | |¯ | IH2,gen = IH2,cons (4.2)

Satisfying Equation 4.2 during continuous battery operation suggests a steady state has been reached in terms of both the hydrogen gas pressure and the bulk chemical balance. In sealed systems, the pressure stabilization is important not only for maintaining chemical balance, but also for safety and for minimizing leaks; in both cases it would be ideal to operate the battery at a low hydrogen pressure. For flow batteries that operate using a single-flow configuration, where there is no separator, satisfying Equation 4.2 is sufficient for ensuring complete electrolyte balance. Since all-iron flow batteries usually operate using separated cell compartments and electrolyte reservoirs, however, another step is involved to balance the proton concentration. That is, until Equation 4.2 is satisfied, the proton concentration decreases in the negative electrolyte due to hydrogen gas generation and simultaneously increases in the positive electrolyte via Equation 5.1. Therefore, the secondary criterion for

37 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES

reaching steady-state is that the rate of protons leaving the negative electrolyte as hydrogen gas generation must be equal to the rate of protons entering it from the positive electrolyte by migration and diﬀusion, as per Equation 4.3, where z = 1 is the charge on a proton, F is Faraday’s constant and NH+ is the ﬂux of protons into the negative electrolyte through the separator and A is the geometric cell area. The required size of a catalyzed recombination reactor depends on the rate of hydrogen generation in the battery and the rebalancing reactor performance; in practice, the hydrogen generation and consumption rates can be measured experimentally, as discussed in the following sections.

¯ + IH2,gen = 2zF NH A (4.3)

4.3 Model Development

The expressions given above provide some basic guidelines for operating sealed iron ﬂow battery systems, however they assume constant values for quantities such as pressure, pH and hydrogen generation. During more-realistic battery operation (e.g., charge-discharge cycling), these quantities are functions of time. For example, both the hydrogen generation rate and the consumption rate may be functions of several

3+ variables (e.g., pH, pH2 , cFe ), which themselves are functions of time during battery operation. The cell reactions and migration directions are illustrated in Figure 4.1. The bodies of electrolyte were approximated as batch reactors in that the species’ concentrations were calculated only as functions of time and not of spatial position (viz., well-mixed electrolyte). Unlike with batch reactors, however, the ﬂow batteries periodically reverse in direction, and also they have side reactions and rebalancing reactions. Since the model solved for concentrations using a batch-reactor approach, quantities were converted to chemical reaction rates (mol m−3s−1). For example,

38 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES

current density was converted into a reaction rate using r = iA/nF V , and the species flux (e.g., diffusion and migration fluxes) were converted using r = NA/V . In these equations, V is the volume of electrolyte.

to reservoir to reservoir separator to reservoir separator to reservoir

Fe2+ −→ 3+ Fe3+ → Fe2+ ←− Fe2+ Fe −→ + ←− Fe3+ Na −→ + Fe2+ → Fe0 ←− Na+ + 0 2+ H −→ + − − Fe → Fe ←− Cl− ←− H+ − + Cl −→ 0 3+ 2+ 3+ 2+ 2H → H2 Fe +2Fe → 3Fe Fe → Fe Fe2+ → Fe3+

+ 0 2+ 2H +Fe → H2+Fe

from reservoir from reservoir from reservoir from reservoir (a) (b)

Figure 4.1: Assumed electrode reactions and direction of ion migration in an all-iron hybrid ﬂow battery during (a) charging and (b) discharging.

At the negative electrode, the desired reaction during charging is the reduction of ferrous (Fe2+) ions according to Equation 4.4.

Fe2+ + 2e− ⇌ Fe0 (4.4)

However, the iron plating is always accompanied by the hydrogen side-reaction shown in Equation 4.5 [144].

+ − 2H + 2e → H2(g) (4.5)

Another side reaction that can occur during charging is the reduction of ferric ions (Fe3+), which can cross the separator into the negative electrolyte and subsequently reduce to ferrous (Fe2+) ions at the negative electrode according to Equation 4.6.

Fe3+ + e− → Fe2+ (4.6)

39 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES

A reasonable approximation [32] is that this reaction rate will be governed by the limiting-current for Fe3+ in the negative electrolyte, which can be estimated using

nF Dc 3+ (t) i (t) = Fe ,n , (4.7) 4.6n δ

where n = 1 is the number of electrons transferred in the Fe(II/III) reaction, F is Faraday’s constant (96,485 C/mol e−), D is the diﬀusivity of Fe3+ ions and δ = 10 µm is the estimated diﬀusion boundary layer thickness based on separate experiments. According to the Pourbaix diagram, iron is electrochemically protected during battery charging, so neither corrosion by protons nor by ferric ions can occur. During discharging, when the metal is more positive than its equilibrium potential, both forms of corrosion can take place concurrently along with the desired iron oxidation reaction. The hydrogen evolution by acid corrosion occurs according to

+ 0 2+ 2H + Fe → H2 + Fe , (4.8)

and iron corrosion by ferric ions according to Equation 4.9, which is the same as the overall battery discharging reaction.

Fe0 + 2Fe3+ → 3Fe2+ (4.9)

Unlike the side-reactions that occur during battery charging, the corrosion reactions that occur during discharging are isolated from the current ﬂowing through the external circuit. Therefore, the rate of Equation 4.4 is given by   − −  [i i4.5(t) i4.6(t)]Acell (charging) (nF Vneg) r4.4(t) =  (4.10)  −  iAcell (discharging) (nF Vneg)

The hydrogen gas pressure is determined by the competition between its genera-

40 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES

tion rate and its consumption rate, or:

dp [(r + r )V − r V ] RT H2 = 4.5 4.8 n 5.1 p , (4.11) dt (Vhs)

where Vhs = 450 mL is the combined headspace volume of the electrolyte reservoirs. Diffusivities of the various species can play an important role in terms of the rate of crossover and mixing. In the electrolyte, the effective diffusivity was estimated using

0 3/2 Di = Di ϵ , (4.12)

0 where Di is the effective diffusivity for component i, Di is the nominal diffusivity and ϵ is the membrane porosity. Equation 4.12 is the Bruggeman equation for diffusion through porous materials [85]. The effective mobilities were estimated based on the tabulated diffusivities using the Nernst-Einstein relation,

D u = i , (4.13) i RT

where ui is the mobility for component i, R is the universal gas constant and T is the temperature. The rate of change of concentration of a given species due to migration of a given component across the separator was then estimated using

z u F c A∇Φ r = i i i , (4.14) i,mig V

where ri,mig is the migration rate and ∇Φ is the potential gradient across the separator. The concentration from either the negative or positive electrolyte was chosen as appropriate based on the charge of the species and whether the battery is charging or discharging. The potential gradient across the separator was estimated using i ∇Φ = , (4.15) κ0ϵ3/2

41 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES

where κ0 = 15 S m−1 is the bulk electrolyte conductivity and ϵ = 0.5 is the porosity. Measurement of hydrogen generation can be challenging due to the dynamic nature of the battery operation, the varying potential of the negative electrode and the sensitive dependence on pH. An empirical curve used to estimate the fraction of current diverted to the hydrogen side reaction during charging is given by    − · 116e 1.15 pHneg(t) (charging) i4.5(t) = (4.16)  0 (discharging)

where pHn refers to the pH in the negative electrolyte and i is the charging current (A m−2). The hydrogen generation rate during charging was estimated from pressure measurements in sealed systems (see Figure 4.3 b). During discharging, hydrogen from the acidic corrosion reaction was assumed to follow a similar expression, given by   0 (charging) i4.8(t) =  (4.17)  − · 268e 2.25 pHneg(t) (discharging)

which is a ﬁt to published iron corrosion rate data [32]. The rate of consumption of hydrogen by the recombination reactor was approximated as

· r5.1(t) = α(1.62 pH2 )/2FVp, (4.18)

which is also based on ﬁtting published data [140], where α is a variable factor and

pH2 is the partial pressure of hydrogen in atm. Diﬀusion of each component through the microporous separator was assumed to occur according to Fick’s ﬁrst law, given in one dimension by

ri,diﬀ (t) = DiA∇ci(t)/Vn. (4.19)

42 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES

The charge balance, given by

∑ zici = 0, (4.20) i

was enforced by adjusting chloride transport to maintain local electroneutrality. Since ferric chloride has similar proton activity to hydrochloric acid, the pH was estimated using − pH = log10 [(cH+ + cFe3+ ) /1000] . (4.21)

The rate of change of the concentration of each component was estimated using the appropriate ODE. For the all-iron system with the side reactions considered here, these equations were for the negative electrolyte:

dcFe2+,n = 3r + r 2+ + r − r − r 2+ + r (4.22) dt 4.9 4.14,Fe 4.6 4.4 4.19,Fe 4.8 dc 0 Fe ,n = r − r − r (4.23) dt 4.4 4.9 4.8 dcFe3+,n = r 3+ − 2r + r 3+ − r (4.24) dt 4.19,Fe 4.9 4.14,Fe 4.6 dcH+,n = r 3+ − 2r + r + − 2r (4.25) dt 4.19,Fe 4.5 4.14,H 4.8 dcNa+,n = r + − r + (4.26) dt 4.14,Na 4.19,Na dc − dc 2+ dc 3+ dc + dc + Cl ,n = (2 Fe ,n + 3 Fe ,n + 1 Na ,n + 1 H ,n ), (4.27) dt dt dt dt dt

43 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES

and for the positive electrolyte,

dcFe2+,p = (r 2+ − 2r − 2r − r 2+ − r + 2r ) (4.28) dt 4.19,Fe 4.4 4.5 4.14,Fe 4.6 5.1 dcFe3+,p = (2r + 2r − r 3+ − r 3+ + r − 2r ) (4.29) dt 4.4 4.5 4.19,Fe 4.14,Fe 4.6 5.1 dcH+,p = (2r − r + − r + ) (4.30) dt 5.1 4.19,H 4.14,H dc + dc + Na ,p = − Na ,n (4.31) dt dt dc − dc − Cl ,p = − Cl ,n (4.32) dt dt dc 0 Fe ,p = 0. (4.33) dt

4.4 Results & Discussion

An example of pressure measurements during battery cycling without using rebalancing is shown in Figure 4.2. For the all-iron battery under the conditions tested, the rate of hydrogen generation during charging was approximately two to ﬁve times larger than during discharging; this eﬀect can likely be attributed to the more negative potential of the electrode during charging compared to the more positive potential during discharging. As shown in Figures 4.3 (a) and (b), the rate of hydrogen generation is roughly an exponential function of pH, and this may be attributed to the

dependence the hydrogen electrode kinetics (i.e., the i0) on the solution pH [145].

1.6 15.1

15 1.4 14.9 1.2 14.8 1 14.7

0.8 14.6 / psia 2 E / V H

14.5 p 0.6 14.4 0.4 14.3 0.2 14.2

0 14.1 0 1 2 3 4 5 t / s Figure 4.2: Example of pressure measurements during battery charging and discharg- −4 2 −2 ing without recombination (Acell = 6.25 · 10 m , i = 250 A m ).

44 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES

10 16.4 charging discharging

pH = 0.38 16.2 8 (i) (ii) 8.4*exp(-1.15*x)

16 -2 6 0.47 15.8 p/psia

0.53 / mA cm

2 4 H 15.6 0.77 i 1.0 1.6 2 15.4

15.2 0 0 0.5 1 1.5 2 0 0.5 1 1.5 2 2.5 t/h pH (a) (b) Figure 4.3: (a) Effect of pH on pressure profiles during charging and discharging iron on a flat graphite plate at 250 A m−2, and (b) calculated hydrogen current densities during charging, where (i) used a flat plate at 250 A m−2 and (ii) used a porous felt at 500 A m−2.

A reasonably good agreement was found between the measured and simulated pressure responses during several cycles of continuous charge-discharge battery testing when using α = 1/4 and ϵ = 0.3, as shown in Figure 4.4. That these parameters were needed to match the measured pressure suggests both that the hydrogen was not being consumed as rapidly as might be expected, and that the separator pores may have eﬀectively been smaller than the nominal value, which was closer to ϵ = 0.5. Some decrease in porosity could be explained by the presence of the PVA coating, and by any solid precipitates that may have formed inside the pores. We speculate that relatively low value of α may have been related to high water vapor pressure, as indicated by drops of water condensate that were observed on the reservoir walls during battery operation. It is also possible that there were gas transport eﬀects, and those were neglected in the present model. Still, the repeatable pattern in the pressure responses, as well as the stable battery performance, provided strong evidence of successful recombination using the in-tank reactor. Cycling voltages and the pressure response during ten days of continuous charge- discharge battery testing at room-temperature in a sealed battery (i = 1000 A m−2) are shown in Figure 4.5 (a). It can be observed that there was an initial

45 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES

0.4 experiment α = 1/4, ε = 0.3 0.35

0.3

0.25

/ atm 0.2 2 H p 0.15

0.1

0.05

0 0 1 2 3 4 5 6 7 8 t/h Figure 4.4: Measured versus simulated gas pressure in the reservoir headspace when a recombination reactor was installed in the positive electrolyte.

pressure overshoot followed by decay to a steady-state operation. That is, the initial hydrogen generation rate was fast compared to the hydrogen consumption rate of the recombination reactor. Then, the hydrogen evolution led to increased pH, causing a reduction in the hydrogen generation rate. Eventually, the rate had decreased to a rate suﬃciently slow that the partially-submerged recombination reactor could consume hydrogen faster than the generation. It was also speculated that some of the pressure drop (ca. 0.1 atm) may have been due to residual oxygen in the headspace gas of the reservoir vessels, which could react with hydrogen to form water at the catalyst layer of the recombination reactor. The battery voltaic, coulombic and energy eﬃciencies were relatively stable during this time, as shown in Figure 4.5 (b).

20 1 (i) (ii) 18 16 0.8 coulombic 1.6 14 )

1.5 a voltaic 1.4 0.6 1.3 1.2

1.1 Efficiency 0.4 energy

1 Pressure (psi Cell Potential (V) averages (%) 0.9 CE: 90 0.8 0.2 VE: 55 0.7 EE: 50 0.6 0 0 2 4 6 8 10 0 20 40 60 80 100 t / d cycle number (a) (b) Figure 4.5: (a) Variation of headspace pressure with time during continuous charge- discharge battery cycling at 1000 A m−2 in a sealed battery with in-tank recombination. There were two pauses in the battery cycling, denoted as (i) and (ii) in the plot. (b) Calculated cycle eﬃciencies.

46 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES

The simulated eﬀects of operating the battery with- and without rebalancing on the ferrous/ferric balance in the positive electrolyte are illustrated in Figure 4.6. Operating without rebalancing, as shown in Figure 4.6 (a), leads to an increasing concentration of Fe3+ and corresponding decrease in Fe2+ in the positive electrolyte. In practice, this decrease in Fe2+ leads to reactant starvation and hence concentration polarization during charging. Simultaneously, the increase in Fe3+ promotes faster crossover into the negative cell, and may also lead to more rapid formation of Fe(III) hydroxides. When the rebalancing cell is used, as illustrated in Figure 4.6 (b), the concentrations of Fe2+ and Fe3+ in the positive electrolyte are kept under control.

1400 1400

1300 Fe(II) 1300 Fe(III) 1200 1200 Fe(II) Fe(III) 1100 1100 -3 -3

1000 1000

c / mol m 900 c / mol m 900

800 800

700 700

600 600 0 2 4 6 8 10 0 2 4 6 8 10 cycle number cycle number (a) (b) Figure 4.6: Simulated iron concentrations in the positive tank (a) without rebalancing and (b) with rebalancing. For clarity, the plot uses the average concentration for each cycle.

Simulated hydrogen generation and consumption rates are shown in Figure 4.7 (a). Initially, there is a relatively high rate of hydrogen generation due to the low pH. Also, since the recombination reactor performance depends on the hydrogen partial pressure, there is initially a slow rate of recombination. Over time, the pH in the negative electrolyte increases due to the loss of protons to hydrogen gas, and concurrently the recombination reactor performance increases due to increasing hydrogen partial pressure in the headspace. The result is that a steady state is achieved when the average rate of hydrogen generation from the negative battery electrodes is balanced by the rate of hydrogen consumption at the recombination reactor. The steady-state

47 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES

pressure depends on the performance (or alternatively, the size) of the recombination cell. This is illustrated in Figure 4.7 (b), which shows simulated pressure responses for different α values. Since the transport of ions is dominated by migration, parameters such as membrane thickness and porosity have relatively small effects on the steady state pressure profiles. Hence, the most important factors to consider are the hydrogen generation rate, which depends on various factors such as the pH, current density and temperature, as well as the hydrogen consumption rate, which depends on the performance of the recombination system.

0.3 0.5 generation consumption 0.25 0.4 0.25

0.2 0.33 0.3 / A 2 0.15 / atm 2 H I H

p 0.2 0.50 0.1

0.1 1.00 0.05

0 0 0 2 4 6 8 10 12 0 2 4 6 8 10 12 t/h t / h (a) (b) Figure 4.7: (a) Simulated hydrogen generation and consumption currents, and (b) eﬀect of recombination reactor area on simulated pressure response, where the label indicates the value of α in Equation 4.18.

Simulated effects of separator porosity and thickness are shown in Figures 4.8 (a) and (b). It is shown increasing porosity has a similar effect to decreasing thickness, both of which allow faster transport from positive to negative electrolyte, and hence a lower negative electrolyte pH, which leads to faster hydrogen generation. Therefore, in terms of hydrogen evolution and system pressure, it makes sense to decrease porosity and increase thickness. However, a trade-off has to be made since these changes would also correspond to increased separator resistance and reduced voltaic efficiency. A sample of program output for 16 seconds of simulation with ϵ = 0.5, α = 1, tsep = 225 µm is shown in Chapter A.

48 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES

0.35 0.35

0.3 0.3

0.25 0.25

0.2 0.3 0.2 100

/ atm 0.5 / atm 175 2 2

H 0.15 0.7 H 0.15 250 p p

0.1 0.1

0.05 0.05

0 0 0 5 10 15 20 0 5 10 15 20 t / h t / h (a) (b)

Figure 4.8: Effects of (a) separator porosity and (b) separator thickness on simulated pressure profiles in sealed iron flow batteries.

4.5 Conclusions

Due to hydrogen side-reactions, electrolyte imbalance in iron flow batteries can be a fast process. The imbalance, which occurs simultaneously for iron and hydrogen species, can be corrected using an appropriate rebalancing system. In the case of sealed systems with internal rebalancing, the balance can be fully restored so that in principal, steady-state operation can be achieved. Development of sealed flow batteries with internal reblanacing is thus an important step toward the ideal “maintenance- free” operation. In this study, the criteria for stability in such systems was discussed and a semi-empirical model was developed in order to help elucidate the nature of hydrogen side-reactions, chemical imbalance, and rebalancing processes. Hydrogen generation rates were measured using sealed, pressurizable vessels. A good agreement was obtained between experiments and models for hydrogen pressure in sealed recombinant systems, and a sealed iron flow battery demonstrated stable operation at room-temperature for over 10 days and 100 cycles. Since the positive electrolyte is more acidic than the negative, increasing separator porosity or decreasing thickness leads to a more acidic negative electrolyte and hence faster hydrogen generation and a higher equilibrium pressure. The pressure can also be decreased by increasing the performance of the recombination system. The model can be easily modified or ex-

49 CHAPTER 4. MODEL FOR SEALED FLOW BATTERIES tended for other systems with different hydrogen generation rates or recombination reactors, as well as for different types of aqueous flow batteries that have hydrogen side-reactions.

50 Chapter 5

In-Tank Recombination

51 CHAPTER 5. IN-TANK RECOMBINATION

5.1 Introduction

Iron is highly abundant, and it can be electroplated with high eﬃciency over a wide temperature range, at moderate pH, and with good morphology. Hence, all-iron ﬂow batteries can use safe, sustainable, and low-cost electrolytes. However, there have only been a few publications on their development [32, 51, 53, 74, 75]. Here, we consider one of the practical challenges in developing such batteries, which is to maintain the balance of active species in the electrolyte. We describe a sealed iron

3+ ﬂow battery system that incorporates H2-Fe recombination to provide the chemical balance needed for long-term stability.

Table 5.1: Electrode reactions in all-iron ﬂow batteries.

Electrode Reaction E0 (V) negative Fe2+ + 2e ⇌ Fe0 −0.44 positive Fe3+ + e ⇌ Fe2+ 0.77 overall 3Fe2+ ⇌ Fe0 + 2Fe3+ 1.21

The electrode reactions during charging and discharging of all-iron flow batteries are shown in Table 5.1. However, iron electrodeposition at the negative electrodes (during battery charging) is always accompanied by hydrogen evolution [144]. One of the main challenges in developing all-iron flow batteries has been due to the electrolyte imbalance that results from this undesired side reaction [13]. Over time, the hydrogen side reactions can cause the electrolyte to become imbalanced in two ways; firstly, the loss of protons from solution causes the negative electrolyte pH to rise, promoting

the formation of solid hydroxide precipitates abbreviated as Fe(OH)3 [146–148]. The presence of such materials in the electrolyte can reduce capacity and leads to a sludge that can hinder electrolyte ﬂow, lower the electrode area and damage the porous separator. Secondly, the positive electrolyte can become over-concentrated in ferric ions (Fe3+), rendering the battery unable to charge when no ferrous ions are available

52 CHAPTER 5. IN-TANK RECOMBINATION to be oxidized at the positive battery electrodes. The latter problem, was addressed

3+ by NASA for the iron-chromium system by using an H2-Fe rebalancing fuel cell [133]. However, the fuel cell recombination system was ultimately abandoned due to problems associated with ﬂooding and platinum dissolution in the strongly acidic electrolytes used [30]. Recently, it has been shown that certain complexing ligands such as glycine can be used to help reduce hydrogen evolution during iron electrodeposition [51]. However, it was not possible to completely eliminate it, and so it may be assumed that there will always be some hydrogen generation during battery charging. For long-term electrolyte balancing, it is necessary to bring all the excess hydrogen and ferric ions

3+ back into chemical balance through Equation 5.1, the H2-Fe recombination reaction.

3+ + 2+ 0 2Fe + H2 → 2H + 2Fe E = 0.77 V (5.1)

While researchers have previously used PEM fuel cells ( [79, 133]), electrolyzers ([30, 134]) trickle bed reactors [149], and other flow-through cells [150] to carry out such processes, those approaches often required additional pumps, externally-supplied reactants or complex control systems. In addition, many of these methods vented hydrogen to the atmosphere. Since iron battery electrolytes are buffered, accurate and continuous pH measurement is challenging and so control systems based on pH probes are challenging for electrolyte control. For practical applications, it would be desirable to minimize the complexity and cost associated with the recombination system. Furthermore, the flow battery system should be sealed to prevent hydrogen losses as well as the oxidation of ferrous ions (Fe2+) by the oxygen in air. In iron electrolytes, for example, the presence of oxygen can cause the formation of solid ferric hydroxide species according to Equation 5.2 [62].

1 1 Fe2+ + O + 2OH− + H O → Fe(OH) (s) (5.2) 4 2 2 2 3

53 CHAPTER 5. IN-TANK RECOMBINATION

An in-tank recombination reactor can have important advantages such as simplic- ity, low-cost, without increasing the overall system footprint. Recently, Whitehead

+ and Harrer investigated an in-tank H2-VO2 recombination reactor for vanadium flow battery applications utilizing a horizontal floating reactor based on catalyzed carbon paper [78]. The horizontal reactor geometry, however, is limited to the cross-sectional area of the electrolyte reservoir, which may not be adequate for all applications. Fur- thermore, thin horizontal reactors were reportedly more prone to flooding, and the carbon paper reactor also required membrane filtration to mitigate problems with poor catalyst adhesion. In order to help address some of these issues, this study

3+ describes a three-dimensional H2-Fe recombination reactor designed for sealed iron ﬂow battery systems but with broader possible applications as well. We present a reactor design wherein the reactions take place on a vertically-oriented galvanic cell (or capillary-action galvanic reactor, CGR) that uses carbon felt as shown in Figure

5.1. In this configuration, we consider the performance relative to Ageo, which is the catalyzed area for hydrogen oxidation, and relative to Acs, which is the cross sectional area of the felt at the waterline. Since the area at the waterline is limited by the geometry of the electrolyte reservoir, and since hydrogen generation rates can be high, sufficiently high recombination rates relative to the waterline area are required for an in-tank reactor. The reactor was designed to oxidize hydrogen in the gas phase while simultaneously reducing ferric ions in solution. The vertical orientation allows for faster reaction rates (especially in terms of the cross section area, Acs) than could be obtained using a horizontal geometry. In a horizontal reactor, Ageo = Acs, but with a vertical reactor it possible to have Ageo > Acs, analogous to using fins to increase area in heat transfer applications. Furthermore, the CGR design operates at low hydrogen partial pressures (0-50 mol% H2), and without the need for any membranes or separators. We report here the characterization of a CGR using electrochemical and

54 CHAPTER 5. IN-TANK RECOMBINATION pressure-based measurements, and demonstrate pressure control in a sealed all-iron ﬂow battery. The CGR can also be applicable to other aqueous ﬂow batteries that produce hydrogen as a by-product from side reactions.

5.2 Materials & Methods

A three-dimensional, vertically-oriented reactor was developed and characterized in sealed, pressurizable vessels (V = 212 mL, see Figure A.1). The ionic pathway was made through capillary wetting within the pores of the porous carbon felt. Hydrogen was oxidized on the upper half of the CGR at the available triple-point sites. Protons and electrons, the products of hydrogen oxidation, were transported downward (viz., in the“in-plane” direction) into the positive electrolyte, where aqueous ferric ions (Fe3+) were reduced to ferrous ions (Fe2+) on the uncatalyzed carbon felt surface. The reactor was fabricated using a carbon felt substrate (Cera Materials PAN-based graphitic felt, thickness = 1/8 in) that was pre-treated by heating in air at 400 ◦C for 24 hours. The upper half of the CGR was coated with a layer of platinized carbon particles (E-Tek, Inc., 40 wt% Pt on Vulcan XC-72), which had been painted onto the felt by hand as an ink that used PVDF as a binder. The ink solvent was n-

−2 methyl-2-pyrrolidone (NMP), and the platinum loading was 3.2 mgPt cm based on the geometric area of 3 cm2. The cross-sectional (horizontal) area was 0.5 cm2, so in this case Ageo/Acs ≈ 6.

3+ The H2-Fe recombination reaction on the reactor was then investigated in several diﬀerent approaches. The ﬁrst used a three-electrode apparatus (see Figure 5.2), which was used to measure hydrogen polarization and impedance spectra in an NaCl solution as a symmetric hydrogen cell. As hydrogen was oxidized in the gas-phase portion of the CGR, the platinum mesh counter electrode was used to carry out proton reduction; therefore, the total hydrogen content was maintained constant. In

55 CHAPTER 5. IN-TANK RECOMBINATION

carbon felt catalyst layer

2 Ageo = 3 cm + − H2 ! 2H + 2e 2 Acs = 0:5 cm

4 cm electrolyte

2Fe3+ + 2e− ! 2Fe2+

1.5 cm

0.3 cm

Figure 5.1: Schematic of a capillary-action galvanic reactor (CGR) design to oxidize hydrogen and reduce aqueous ferric ions. Ageo is the catalyzed area for hydrogen oxidation and Acs is the cross sectional area of the felt at the waterline. Floats (not shown) ensured that the catalyzed area remained above the liquid level.

Figure 5.2, the counter electrode was a platinum mesh and the reference electrode was Ag/AgCl (3.0 M NaCl). Cyclic voltammetry was used to investigate the hydrogen oxidation behavior at diﬀerent partial pressures of hydrogen ranging from 0.3 to 11.3 psig. The second approach used a one-electrode conﬁguration with a solution containing

1 M FeCl3, 1 M FeCl2, and 2 M NaCl (see Figure 5.2 (b)). In some cases, a reference electrode was also used so that the mixed potential of the CGR could be measured (see Figure A.3). The reaction rate in this conﬁguration was measured by monitoring the change gas pressure as hydrogen was consumed. The CGR was partially submerged in the electrolyte such that the lower half was in solution to carry out Fe3+ reduction and the upper half was in the reactor headspace to facilitate hydrogen oxidation. The headspace was initially purged with nitrogen, then hydrogen was pulse-injected into

the reservoir at a partial pressure of PH2 =4.5 psig. Following the hydrogen injection, the pressure was recorded over time until all of the hydrogen had been consumed

56 CHAPTER 5. IN-TANK RECOMBINATION

to DAQ P to DAQ P we ref ce we N2 vent N2 vent inlet inlet

CGR CGR N2 H2 N2 H2 Pt mesh

NaCl, HCl Fe2+, Fe3+

(a) (b)

Figure 5.2: (a) Schematic of the three-electrode apparatus used to characterize the upper half (viz., the hydrogen electrode) of the CGR in 2 M NaCl (pH = 1.4) at several diﬀerent partial pressures of hydrogen. The reference electrode was Ag/AgCl (3 M NaCl), and the counter electrode was a platinum mesh. (b) Schematic of the 3+ one-electrode apparatus used to obtain the overall H2-Fe recombination rate.

(Omega PX-302 digital transducer). After injecting hydrogen, the pressure began to drop due to the hydrogen being consumed by oxidation on the CGR catalyst layer. Simultaneously, ferric ions were reduced at the felt surface that extended down into the electrolyte. The recombination

−2 reaction rate, iH2 (A cm ) was calculated from the slope of the pressure drop using Equations 3 and 4, where n=2 is the number of electrons involved in the reaction,

−1 F is Faraday’s constant (96,485 C mol ), Vhs is the headspace volume (L), R is the universal gas constant (0.08206 L atm mol−1K−1), and T is the temperature (293 K).

nFVhs(dP/dt) igeo = − (5.3) AgeoRT

nFVhs(dP/dt) ics = − (5.4) AcsRT

A sealed all-iron ﬂow battery system (see Figure 5.3) was cycled both with and without a CGR installed in the positive electrolyte reservoir for comparison of the

57 CHAPTER 5. IN-TANK RECOMBINATION pressure profiles. In each cycling test, the battery was charged and discharged at 20 mA cm−2, where each charge duration was one hour and each discharge was carried out until reaching a cell cutoff voltage of 0 V. The battery electrodes consisted of flat graphite plates with an area of 6.25 cm2. A porous separator was used (Daramic SLI 175). The system operated at ambient temperature, and both electrolytes were pumped at 50 mL min−1 using a peristaltic pump. Both electrolytes contained 1 M

FeCl2 and 2 M NaCl, and the positive electrolyte also contained an initial 0.5 M FeCl3 spike. Each reservoir contained an electrolyte volume of 100 mL.

headspace connection

separator

2 3 H 1 H 2 − + 2 CGR 4 3+ H+ Fe

negative electrolyte positive electrolyte

Figure 5.3: Use of the CGR in a complete ﬂow battery system. A headspace connection tube (OD = 0.25 in, length ≈ 12”) provided a pathway for hydrogen to transport from the negative reservoir to the positive reservoir.

5.3 Results & Discussion

The three-electrode apparatus was used to measure open-circuit potential as well as carrying out cyclic voltammetry and electrochemical impedance spectroscopy experiments in sodium chloride solutions adjusted to pH=1.4 by drop-wise addition of hydrochloric acid. The Nernstian behavior (see Figure 5.4) of the open-circuit potential measurements conﬁrmed that the upper half of the CGR operated as expected

58 CHAPTER 5. IN-TANK RECOMBINATION with respect to the partial pressure of hydrogen. Longer-duration experiments, as shown in Figure 5.5, showed that the reactor was relatively resistant to degradation by flooding or loss of catalyst. In some experiments, the performance appeared to increase over time, perhaps due to improved wetting of the felt fibers. This effect applied both when carrying out the recombination reaction as well as during hydrogen oxidation in the three-electrode apparatus in the absence of ferric ion in the electrolyte.

40 ( ) SHE + 2 0 E = RT ln [H ] nF PH mV 2 E/ -40

-80 0 1 2 3 4 5

PH2 /psig Figure 5.4: Potential of the galvanic reactor as a function of hydrogen partial pressure

(PH2 ), as measured using the apparatus shown in Figure 2. The dashed line represents the expected response based on the Nernst equation.

) 32 AgCl

/ 30 Ag 28 (mV

E 26

22 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 t (h)

Figure 5.5: Potential of a CGR during constant-current hydrogen oxidation at 40 mA using the apparatus shown in Figure 5.2.

Results from voltammetry and impedance experiments are shown in Figures 5.6 (a) and (b). In all experiments, the gas composition contained 1 atmosphere of nitrogen plus some additional hydrogen ranging from 0.3-8.0 psig. It was observed

59 CHAPTER 5. IN-TANK RECOMBINATION

that the hydrogen partial pressure had a signiﬁcant eﬀect on the hydrogen electrode

polarization up to about PH2 =8 psig, beyond which the rate became limited by other factors such as ohmic losses or ferric ion concentration polarization. An apparent

limiting hydrogen current was observed at the lowest partial pressure of PH2 =0.3 psig, consistent with a low frequency mass transfer loop observed in the EIS response. It was also possible to estimate the limiting regimes using pressure measurements, as

shown in Figure A.2. An additional data set for a reactor showing results for pH2 = 0 can be found in Figure A.8 in the Appendix section.

2.5 60

0.2 50 11.3 psig H2 2 8.2 0.6 40 3.3 1.5 ) 1.9 2 1.9 cm

′′ 30 1 8.2 / 1.0 Z 100 kHz 0.6 (mA 20

0.5 1 Hz i 10 0 7 Hz 0.3 0 −0.5 1.5 2 2.5 3 3.5 4 −0.2 0 0.2 0.4 0.6 0.8 ′ Z E (VAg/AgCl) (a) (b)

Figure 5.6: (a) Impedance response as a function of hydrogen partial pressure (PH2 ) as measured using the apparatus shown in Figure 5.2. The low frequency response suggested the presence of a mass-transfer resistance loop that decreased with increas-

ing PH2 from 0.2 psig to 8.2 psig. (b)Voltammetric response in unstirred solution as

a function of hydrogen partial pressure (PH2 ) as measured using the apparatus shown in Figure 5.2.

Transient measurements of the hydrogen partial pressure, as made using the apparatus shown in Figure 5.2 (b), were used to determine the rate of the recombination reaction in a sealed vessel. It was assumed that any measured pressure drop (after hydrogen addition) was due to hydrogen consumption via Equation 5.1, i.e., that there were no signiﬁcant liquid or gas leaks, and this assumption was checked before each test by pressurizing with nitrogen; when only nitrogen was present, the pressure did not change over time.

60 CHAPTER 5. IN-TANK RECOMBINATION

When only nitrogen was present in the vessel, the pressure remained constant, as shown in Figure 5.7 (a). After introducing hydrogen, however, the pressure im- mediately began to drop, until presumably all the hydrogen had been consumed. In some experiments, a reference electrode was also incorporated in order to measure the CGR mixed potential (see Figure A.3). Equivalent recombination rates of up to

180 mA were observed at PH2 = 4.5 psig, which was much greater than the limiting

2 current of ferric ion diffusion based on the cross-sectional area (Acs=0.5 cm ). That is, assuming a horizontal reactor with A=0.5 cm2 and a limiting Fe3+ current density of 30 mA cm−2 (a reasonable value for a one-electron reaction occurring at a smooth electrode with significant electrolyte flow over the electrode surface), the maximum possible current would have only been 14 mA. Using Equations 5.3 and 5.4, the observed current of 180 mA corresponds to approximately 60 mA cm−2 based on the geometric catalyzed area and 375 mA cm−2 based on the cross-sectional (horizontal) area (see Figure 5.7). This current density is over an order of magnitude greater than that reported for the reactor described in Ref. [78], despite operation at a lower pres-

sure of hydrogen gas (less than 24 mol% H2 in this study). This result suggests that the reactions are distributed over the wider area available as a result of the vertical, three-dimensional orientation of the reactor.

7 70 400 6 N2 pressure test 60 350 5 50 300 2 2 − ) − g 4 H2 recomb 40 250 (psi mAcm 2

200 mAcm / H

3 30 / P cs geo i i 150 2 20 100 1 10 50 0 0 0 0 0.2 0.4 0.6 0.8 1 1.2 0 0.5 1 1.5 2 2.5 3 3.5 4

t (h) PH2 /psig (a) (b)

Figure 5.7: (a) Comparison of pressure responses after injecting nitrogen versus hydrogen. (b) Hydrogen oxidation current density relative to the catalyzed and cross- sectional areas as shown in Figure 5.1, as calculated using Equations 5.3 and 5.4.

61 CHAPTER 5. IN-TANK RECOMBINATION

3 without CGR 2.5

2 psig

/ 1.5 2

H with CGR P 1

0.5

0 0 10 20 30 40 50 60 70 t/h

Figure 5.8: Comparison of pressure proﬁles during continuous battery cycling at 20 mA cm−2 using the apparatus illustrated in Figure 5.3. The charge and discharge durations were each one hour.

When implemented in a complete iron flow battery system, the reactor appeared to effectively control the pressure, as shown in Figure 5.8. Similar results were observed in other tests as well (see Figure A.9). When the reactor was not used, the pressure increased with each cycle. When the reactor was installed in the positive reservoir, however, the pressure initially increased at the same rate but then began to decrease and level off as hydrogen diffused through the headspace connection and the recombination reaction took effect. Without the CGR installed, there was still some decline in the slope over time, and this can be attributed mainly to increasing electrolyte pH. The difference between the slopes with- and without the CGR can then be attributed to hydrogen consumption. It is also speculated that part of the pressure decline may be attributable to residual oxygen, which could react with hydrogen at platinum sites to make water. Similar results were obtained in a larger-scale battery test that used an electrode area of 30 cm2 at current densities of 100 mA cm−2 with one-hour charges (see Figure 4.5). The 30-cm2 cell hardware is shown in Figures A.4 and A.5, and the reactor is shown in Figure A.6. After about 10 days of testing, the electrolytes looked clear, as shown in Figure A.7. However, brown precipitates were typically observed on the negative side of the separator, as shown in Figure A.10. Considering the reaction rates demonstrated here for the CGR, it is feasible to

62 CHAPTER 5. IN-TANK RECOMBINATION implement in-tank recombination for systems like the all-iron battery that have 1-5% coulombic losses to hydrogen generation during battery charging. A CGR system operating as described here would require ≈ 40% of the available electrolyte surface area. In addition, with a reasonable reduction in the platinum loading to 0.2 mg cm−2, we estimate that the materials cost for catalyst and felt would be <$10 per kW of battery power in an all-iron system. However, further development is needed to reduce catalyst loading, investigate alternative catalysts such as tungsten carbide (WC), and measure degradation rates over longer periods of time. Also, it would be useful to validate a component-scale model. A preliminary current distribution model (see Figures A.11 and A.12) is being developed, and can be used to for reactor design.

5.4 Conclusions

3+ An H2-Fe recombination reactor was developed for a sealed all-iron flow battery system. Although hydrogen suppression additives or acid additions can be used to minimize the effects of hydrogen evolution for smaller periods of time, these approaches are insufficient to ensure long-term chemical balance in the battery electrolytes. To control pH and ferric ion concentration well, therefore, a hydrogen recombination strategy should be implemented. Since hydrogen and ferric ions become out of balance in proportion to one another, a chemical balance can be achieved in a sealed recombinant flow battery system as described in this work without the need for any external control systems. This study described a simple, in-tank galvanic reactor that can carry out the recombination reaction without the need for additional pumps, control systems or membranes. It was shown that simple pressurizable vessels could be used to characterize such reactors using both electrochemically as well as by measuring the pressure after pulse-injecting hydrogen gas. A recombination current of

63 CHAPTER 5. IN-TANK RECOMBINATION up to 60 mA cm−2 (based on the platinum catalyzed area) or 375 mA cm−2 (based on the horizontal area taken up by the reactor at the waterline) was observed with

light stirring in 1.0 M FeCl3 at PH2 =4.5 psig. The reactor appeared to be stable and resistant to flooding, and there was no evidence of any issues associated with loss of catalyst layer adhesion. Future studies will look in more detail at the wick- ing behavior, current distribution, non-Pt based electrocatalysts, and scale-up. With continued research and development, the CGR holds promise as an effective tool for rebalancing electrolytes and hence extending lifetimes in sealed flow batteries. The methods described here can also apply to vanadium, iron-chromium, zinc-bromine, and other battery chemistries.

64 Chapter 6

Zinc-Iron Chloride Flow Batteries

65 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES

6.1 Introduction

Zinc-based hybrid ﬂow batteries are being widely-developed due to the desirable electrochemical properties of zinc such as its fast kinetics, negative potential (E0 = −0.76

VSHE) and high overpotential for the hydrogen evolution reaction (HER). Many groups are developing zinc-bromine batteries, and they address challenges associated with bromine toxicity and the organic complexing agents used to reduce its vapor pressure [1, 23]. Other positive electrode couples using cerium, vanadium, nickel and iron are also being investigated [151–153]. Zinc-ferricyanide ﬂow batteries using alkaline electrolytes were developed in the late 1970s, but progress was reportedly hindered by high membrane costs and challenges with handling solid zinc oxide precipitates [154]. Currently, zinc-ferricyanide ﬂow batteries are also being developed by

3−/4− ViZn, Inc. [155]. Some of the challenges of using the Fe(CN)6 couple include its low solubility on the order of 0.2-0.5 mol L−1, and the possible generation of toxic gas if it mixes with acid [156–158]. More recently, a zinc-iron ﬂow battery based on deep eutectic solvents (DES) with an open-circuit potential of 1.02 V was described, but it operated at a low current density of 0.5 mA cm−2 and so it was concluded that high-temperature operation would be required in order to obtain useful power densities [159]. Of the possible reactions to use for a positive electrode, the aqueous Fe(II/III) redox couple is among the safest and cheapest, and it has high solubility and fast kinetics even on uncatalyzed carbon materials [74]. However, most studies of zinc-iron

3−/4− batteries have focused on the alkaline chemistry (using Fe(CN)6 at the positive electrode), and there are only a few reports of zinc-iron ﬂow batteries based on the acidic chemistry. A recent study combined an alkaline (2.4 M NaOH) negative electrode with an acidic (1 M HCl) positive electrode to achieve high performance, but this required the use of two ion-exchange membranes as well as a third electrolyte pump [17]. In 2016, an acidic zinc-iron sulfate battery employing an ion-exchange

66 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES membrane demonstrated 50 charge-discharge cycles at 30 mA cm−2 from a bath

−1 containing 1.5 mol L H2SO4 and acetate buffers [137]. During battery cycling, the authors observed a performance fade and attributed it to crossover of iron ions. Therefore, it was stated that it would not be possible to operate a battery with a mixed zinc-iron electrolytes because any Fe2+ or Fe3+ present in the negative electrolyte would be reduced in place of the zinc ions, and it was concluded that future work should focus on development of more selective membranes. In conventional (single-membrane) architectures, the ion-exchange membranes ($120-500 m−2), can account for 20-40 % of the flow battery cost, and their use has been called “the stumbling block” toward flow battery commercialization [160, 161]. Whereas previous studies have used such membranes to prevent electrolyte mixing, this work investigates battery operation using mixed electrolytes in order to simplify the battery hardware and to reduce capital costs. As with the all-iron flow battery, moderate amounts of electrolyte crossover in this configuration would not cause irreversible performance loss [32]. Furthermore, the battery could operate using microporous separators (normally based on polyethylene or polypropylene), which are less expensive than ion-exchange membranes by more than an order of magnitude [162]. It should also be noted that in lightly-acidic iron chloride solutions, the conductivity of proton exchange membranes can be much lower than in more acidic solutions due to their conversion to an iron form rather than the desired proton form.

It might be expected that an electrodeposit made from the mixed ZnCl2-FeCl2 electrolytes would be an alloy of the two metals, and that the incorporation of iron would significantly decrease both the voltaic efficiency and the coulombic efficiency; the reduction in voltaic efficiency would be expected because of the more positive potential of iron (E0 = −0.44 V vs SHE) compared to zinc (E0 = −0.76 V vs SHE), and the reduction in coulombic efficiency could be expected since iron is a relatively good catalyst for hydrogen evolution [36]. However, anomalous codeposition (ACD)

67 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES

phenomena allow for preferential zinc plating from electrolytes containing zinc mixed with iron-group metal ions [100, 121, 163]. This study shows that it’s possible to use the ACD to enable a pseudo-zinc negative electrode that operates from mixed

ZnCl2-FeCl2 electrolytes, as illustrated in Figure 6.1. The associated reactions for the negative electrode, positive electrode and overall cell reaction of such a system are given by Equations 6.1, 6.2 and 6.3, respectively.

porous − ( ) separator (+)

Figure 6.1: Schematic of a zinc-iron chloride ﬂow battery with mixed electrolytes.

chg Zn2+ + 2e− ⇌ Zn E0 = −0.76 V (6.1) dis dis Fe3+ + e− ⇌ Fe2+ E0 = +0.77 V (6.2) chg chg Zn2+ + 2Fe2+ ⇌ Zn + 2Fe3+ E0 = 1.53 V (6.3) dis

The anomalous behavior of electroplating from such mixed electrolytes has been attributed to the so-called hydroxide suppression mechanism (HSM), wherein the formation of a surface hydroxide layer (e.g., Zn(OH)2) impedes the transport of the more noble metal ions to the underlying substrate [98, 164]. However, many studies reported that the HSM was inconsistent with experimental data [97, 100–102, 105–

68 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES

107, 111, 114, 115, 165–171], so there is still no universally-accepted mechanism for anomalous codeposition. Although more research is needed to fully understand ACD, in this study we take advantage of the ACD phenomena for application in a flow battery. Electrodeposition from related mixed electrolytes has been studied for alloy plating applications, where there are different requirements and performance metrics than there are for battery applications. In alloy deposition, for example, it is desired to incorporate iron into the deposit because the alloy has superior corrosion resistance compared to pure zinc [101, 103, 111, 112, 172]. However, the coulombic efficiency during plating decreases with increasing iron incorporation into the alloy [112]. For zinc-iron battery applications, it would be ideal to completely inhibit iron deposition in order to maintain the zinc potential and hinder hydrogen evolution. Furthermore, for battery applications, there can be additional challenges due to the galvanic displacement during discharge via Equation 6.4. If the galvanic displacement of zinc by iron is slow, then the charge-discharge processes can be carried out with performance similar to that which could be obtained from a “pure” electrolyte containing only Zn2+ as the active metal cation in the negative electrolyte. This study investigated the plating and stripping processes of zinc, iron and their mixtures at pH=1 in order to avoid the formation of insoluble iron hydroxide precipitates, which can form between 2

Zn + Fe2+ → Fe + Zn2+ E0 = 0.32 V (6.4)

The Fe2+/3+ positive redox couple has been used in iron-chromium and all-iron ﬂow batteries [30,32,79], and the Zn0/2+ redox couple has been used in zinc-bromine ﬂow batteries [31, 35, 36]. However, zinc and iron are normally not used together as zinc-

69 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES

iron flow batteries in acidic media. Zinc-ferricyanide flow batteries were developed for alkaline media [36,136,154,173], as illustrated in Figure 6.2 (a), and they are also being pursued by the startup company ViZn, Inc. [155]. However, zinc-ferricyanide flow batteries have challenges associated with high membrane costs and ZnO solids handling [154]. A more recently-proposed flow battery, illustrated in Figure 6.2 (b), combined zinc and iron species uses a cell that is half-alkaline, half-acidic [17]. This type of cell demonstrated high performance, but at the cost of using two ion-exchange membranes and three pumps. Unless the membranes perfect separation between the three electrolytes, the double-membrane design is inherently unstable. Therefore, there is a gap in the literature in terms of making zinc-iron flow batteries that use acidic media and simple, inexpensive hardware.

H2-Fe(CN)6 recomb CEM AEM

+ − O2 Na Cl

Zn(OH)2− 3+ sump 4 2Fe

negative positive ZnO electrolyte electrolyte ferri/ferro slurry Zn 2Fe2+ storage 20 % solid 50 % solids HEX cell stack T =60◦ C middle ﬁlter electrolyte

(a) (b)

Figure 6.2: (a) Schematic of an alkaline zinc-ferricyanide battery including solids handling and hydrogen recombination, from [173], and (b) a zinc-iron ﬂow battery design using two membranes and three bodies of electrolyte, from Ref. [17].

An acidic zinc-iron sulfate battery using a more conventional cell architecture (viz., a single proton exchange membrane) was recently demonstrated by Xie et al. [137]. It was concluded that it would be infeasible to operate the battery using mixed electrolytes, and that better ion-exchange membranes should be developed. In terms of reducing cost and complexity, however, it would be ideal to operate zinc-iron batteries using a microporous separator, which can be made of inexpensive

70 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES polyethylene or polypropylene. Such an approach not only would reduce capital costs, but it would also create a crossover-tolerant battery chemistry. However, there is a gap in the literature in terms of whether or not such an arrangement could work with acceptable stability and eﬃciency. Therefore, a new approach being developed in the present research is to employ anomalous codeposition (ACD) in order to deposit zinc from solutions containing considerable amounts of iron (e.g., equimolar solutions of

ZnCl2 and FeCl2). If this can be done eﬃciently at reasonable current densities and suﬃciently-high concentrations of each metal species, then zinc-iron batteries using microporous separators may represent a feasible approach to consider going forward.

6.2 Materials & Methods

Using an H-Cell with a Luggin capillary (see Figure A.13), deposition and dissolution were examined using solutions containing iron chloride, zinc chloride, and their mixtures. The solution pH was adjusted by drop-wise addition of 10 %w HCl. Titanium bar (0.020”, grade 2A, ASTM B265, A = 1 cm2) and glassy carbon (A = 0.197 cm2) were used as substrates. Initial pre-treatment of the titanium consisted of washing in detergent with hot water followed by rinsing in acetone and isopropyl alcohol. Subsequently, the electrode was sanded by hand (120-grit, Norton T461) to remove

−1 surface oxides. Unless otherwise noted, 1 mol L NH4Cl was used as a supporting electrolyte. Cyclic voltammetry was carried out by scanning the potential from -0.2 to -1.3 V vs Ag/AgCl at scan rates, ν, ranging from 2-50 mV s−1 using a Solartron Modulab potentiostat. Effects of transport were studied using a rotator (Pine, Inc.) with a glassy carbon electrode (A = 0.197 cm2). A flow battery (A = 6.25 cm2) consisting of graphite plates as electrodes and a microporous separator (Daramic 175, thickness = 175 µm) was used, with the flow provided by peristaltic pumps at 50 mL min−1. A sketch of the hardware is shown in

71 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES

Figure A.18, and photo of the assembled cell is shown in Figure A.19. The positive electrode employed 2-mm thick carbon felt bonded to the graphite plate, as described elsewhere [74]. Teﬂon gaskets (2-mm thick) were used on each side of the separator. The electrolytes were prepared to simulate operation at 30 % SoC, assuming a starting

composition of 1.6 M ZnCl2 and 0.8 M FeCl2 in both compartments. Therefore, the

−1 −1 initial negative electrolyte contained 1.12 mol L ZnCl2 and 0.8 mol L FeCl2,

−1 −1 and the initial positive electrolyte contained 1.6 mol L ZnCl2, 0.56 mol L FeCl2

−1 −1 and 0.24 mol L FeCl3. Both electrolytes contained 2 mol L NH4Cl supporting

−1 electrolyte and 2 g L PEG8000 to moderate dendrite growth [174], and the volume of each electrolyte was 175 mL. Charging and discharging were carried out at 25 mA

−2 cm . Coulombic and voltaic eﬃciencies were estimated using CE = qdischarge/qcharge ¯ ¯ ¯ and VE = Edischarge/Echarge, where q represents charge in coulombs and E represents the average value of potential. Battery charging was carried out for one hour, and discharges were carried out until the cell reached a cutoﬀ voltage of 0 V to completely strip any deposited metal. This represented a SoC swing of about 1.5 % based on the Zn2+ in the negative electrolyte and 4.2 % based on the Fe2+ in the positive electrolyte. Temperature was maintained at 25 ◦C using an in-line shell-and-tube heat exchanger.

6.3 Results & Discussion

The chloride medium was chosen for this study because electrolytes containing chlo-

ride tend to promote greater current eﬃciency (lower H2 evolution rates) and conductivity than those containing sulfate [112,114,116,175]. A comparison between the voltammograms of the iron, zinc, and mixed zinc-iron systems is shown in Figure 6.3.

0 It is noteworthy that in the iron-only solution (E = −0.66 VAg/AgCl), there is about 360 mV of overpotential before the onset of signiﬁcant reduction currents, compared

72 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES

0 to about 80 mV for zinc-containing solutions (E = −0.98 VAg/AgCl). Due to the high overpotential for iron deposition, the reduction curves for iron and zinc occur within about 50 mV of one another despite a 300 mV diﬀerence in their standard reduction potentials.

1 i ii iiiiivv 0.5 M ZnCl2 0.6

0.5 M ZnCl2 + a 0.2 ,

0.5 M FeCl2 p −2 140 mA cm i/i −0.2 0.5 M FeCl 2 −0.6 −1 −1.2 −1 −0.8 −0.6 −0.4 −0.2 0 −1.2 −1 −0.8 −0.6 −0.4 E /VAg/AgCl E/VAg/AgCl (a) (b) Figure 6.3: (a) Deposition and dissolution at pH=1 and ν = 10 mV s−1 when scanning to -1.2 V versus Ag/AgCl from zinc and iron chloride electrolytes onto titanium substrate in quiescent solution, and (b) eﬀect of negative scan limit (ν = 10 mV s−1) on deposition and stripping from equimolar ZnCl2-FeCl2 solutions, where current is normalized to the peak stripping current; (i) is -1100 mV, (ii) is -1150 mV, (iii) is -1200 mV, (iv) is -1250 mV, and (v) is -1300 mV. All solutions contained NH4Cl supporting electrolyte adjusted to maintain 3 mol L−1 Cl−.

As can be observed in Figure 6.3 (a), reduction currents for the iron-only electrolyte started at approximately -1.0 V versus Ag/AgCl, and the stripping current occurred in the region of -0.6 to -0.2 V. When the equimolar zinc-iron solution was used, however, there was no signiﬁcant iron stripping peak despite a clear zinc stripping peak in the region of -1.0 to -0.8 V. Based on the polarization curves, normal alloy plating theory would expect that iron reduction should have also taken place concurrently, but this was not observed. After plating at low overpotentials (less- negative scan limits), iron stripping peaks were not observed, even when zinc plating and stripping took place. When scanning the potential to more negative limits, however, subsequent iron stripping peaks were present, as shown in Figure 6.3 (b). While the iron stripping peaks were nearly absent when scanning to -1100 mV, -1150 mV

73 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES

0.5 M ZnCl2

0.2 M FeCl2 + 0.8 M ZnCl2

0.5 M FeCl2 + 0.5 M ZnCl2 170 mAcm−2 0.5 M FeCl2

−1.3 −1.1 −0.9 −0.7 −0.5 −0.3 E/VAg/AgCl

Figure 6.4: Comparison of deposition and stripping processes for diﬀerence compositions of ZnCl2, FeCl2 and mixed electrolytes on a titanium bar with a negative scan limit of -1.3 V vs Ag/AgCl for the mixed electrolytes.

and -1200 mV, the scans at -1250 mV and -1300 mV displayed noticeable iron stripping peaks at ca. -0.6 V. These results were consistent with a previous study, which also reported increasing iron content with more negative scan limits in NaCl solution with Zn/Fe = 1/3 [114]. Since iron stripping peaks were observed under some conditions, which would be undesirable for a flow battery, further experiments were conducted in order to compare the effects of pH, Zn/Fe molar ratio and zinc transport on the deposition and stripping behavior. First, voltammograms were carried out from quiescent solution onto titanium substrate at pH = 1 and pH = 3 at three different Zn/Fe molar ratios. As shown in Figure 6.5, it was found that iron stripping peak currents increased at lower Zn/Fe molar ratios, but there was little difference between voltammograms from the pH = 1 and pH = 3 solutions. In addition to the formation of the iron stripping peak, more negative scan limits were also associated with the bifurcation of the peak in the zinc region into two separate peaks. This behavior has been attributed to the

74 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES

formation of a Zn+Fe solid solution η-phase, wherein the ﬁrst peak corresponds to the zinc in the η-phase, the second corresponds to the iron in the η-phase, and peak in the iron-stripping region (-0.45 to -0.6 V vs Ag/AgCl) corresponds to pure α-phase iron [114]. In order to further elucidate the roles of Zn/Fe molar ratios and zinc ion transport, additional experiments were carried out using a rotating disk electrode with a glassy carbon substrate.

150 150

100 Zn:Fe = pH=1 100 Zn:Fe = pH=3 1.5 1.5 2 2 − 50 1.0 − 50 1.0 0.5 0.5 0 0 i/mAcm i/mAcm

−50 −50

−100 −100 −1.2 −1.1 −1 −0.9 −0.8 −0.7 −0.6 −0.5 −0.4 −0.3 −1.2 −1.1 −1 −0.9 −0.8 −0.7 −0.6 −0.5 −0.4 −0.3 E/VAg/AgCl E/VAg/AgCl (a) (b) Figure 6.5: (a) Deposition and dissolution onto titanium substrate from quiescent solution for three Zn/Fe ratios at (a) pH=1 and (b) pH=3, where all solutions contained −1 −1 CMe2+ = 1 mol L and 3 mol L NH4Cl supporting electrolyte.

As shown in Figure 6.6 (a), reduction currents between 20-30 mA cm−2 were measured at rotation rates of 300< ω <1500 rpm without any observable iron stripping

peaks from an equimolar ZnCl2-FeCl2 solution with negative scan limit of -1150 mV vs Ag/AgCl. That is, the zinc appeared to be deposited preferentially via an ACD mechanism. Since zinc-based flow batteries often charge at 10-50 mA cm−2 [1], this result suggested that zinc-rich deposits can be made (viz., the ACD process can be uti- lized) from mixed electrolytes at useful current densities in flow battery applications. Although not presented, XPS analysis was also used to confirm the predominance of zinc in the deposit surfaces (see Figure A.25). In order to obtain more insight into the deposition and stripping, further tests were conducted at more negative scan limits (higher plating current densities), as shown Figures 4 (b), (c) and (d). Here, some iron stripping peaks were observed, particularly at the lowest rotation rate (ω = 300

75 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES

rpm). Similar to the other experiments from quiescent solution at low Zn/Fe molar ratios (Figure 6.5), the zinc stripping peak bifurcated into two separate peaks when plating at higher current densities. When the rotation rate was increased though, the associated stripping current went from showing three peaks (two in the zinc region and one in the iron region) to showing only one peak. That is, faster rotation rates were necessary in order to keep up with higher current densities while avoiding bifurcation of the zinc stripping peak as well as formation of iron stripping peaks. The behavior from the mixed solution was also compared with a zinc-only control solution, as shown in Figure 6.7 for a negative scan limit of -1350 mV vs Ag/AgCl. It is clear that at relatively higher plating current densities (i > 50 mA cm−2), faster rotation rates led to stripping peaks that looked closer to the those from the zinc-only solution. These results seemed to disagree with the HSM theory, whereby the anomalous behavior should be enhanced by increasing pH (or equivalently, by decreasing the electrode rotation rate so that a signiﬁcant pH gradient can be established). In Figure 6.6 (a), for example, there were clear zinc stripping peaks present but no iron stripping peaks, despite plating on a disk spinning at 1500 rpm from a solution with an acidic solution (pH=1). It has been reported that the critical pH for zinc hydroxide precipitation is approximately pH = 5.18 [102, 113]. However, it can be shown though Fick’s ﬁrst law that such a high surface pH is unlikely to exist on an electrode rotating at 1500 rpm, wherein the surface pH should be close to the bulk pH. Furthermore, HSM theory would predict a trend of decreasing anomalous behavior with increasing rotation rate (since increasing rotation rate is associated with a small pH gradient near the electrode surface), but the opposite trend was observed. This is consistent with several previous studies, which have noted similar discrepancies Chassaing1991, Roventi2015. In the present results, the anomalous deposition was found to be enhanced with increasing rotation rate, rather than being

76 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES

40 150 30 100 20 50 2 2 − 10 − 0 0 /mAcm −10 /mAcm i i −50 300 rpm −20 300 rpm 900 rpm 900 −100 1500 rpm −30 1500 −40 −150 −1.2 −1.1 −1 −0.9 −0.8 −0.7 −0.6 −0.5 −1.3 −1.2 −1.1 −1 −0.9 −0.8 −0.7 −0.6 −0.5 E/VAg/AgCl E/VAg/AgCl (a) (b) 300 400 250 300 200 150 200

2 100 2 − − 50 100 0 0 /mAcm − /mAcm i 50 300 rpm i 300 rpm −100 900 rpm −100 −150 1500 rpm 900 rpm −200 1500 rpm −200 −250 −300 −1.4 −1.2 −1 −0.8 −0.6 −0.4 −0.2 0 −1.6 −1.4 −1.2 −1 −0.8 −0.6 −0.4 −0.2 0 E/VAg/AgCl E/VAg/AgCl (c) (d)

Figure 6.6: Eﬀect of rotation rate and negative scan limit on deposition and stripping −1 −1 from electrolytes containing 0.5 mol L ZnCl2 + 0.5 mol L FeCl2 on glassy carbon. The scan rate was ν = 50 mV s−1 and the negative scan limits were (a) -1150 mV, (b) -1250 mV, (c) -1350 mV and (d) -1450 mV vs Ag/AgCl. The solution contained −1 1 mol L NH4Cl supporting electrolyte.

200 200

100 100 2 Zn 2 Zn − Zn-Fe − Zn-Fe 0 0 /mAcm /mAcm i −100 −i 100

−200 −200 −1.4 −1.2 −1 −0.8 −0.6 −0.4 −0.2 0 −1.4 −1.2 −1 −0.8 −0.6 −0.4 −0.2 0 E/VAg/AgCl E/VAg/AgCl (a) (b) Figure 6.7: Comparison of deposition and stripping behavior between 0.5 mol L−1 −1 −1 ZnCl2 and 0.5 mol L ZnCl2 + 0.5 mol L FeCl2 at (a) 300 rpm and (b) 1200 rpm on glassy carbon with ν = 50 mV s−1 and negative scan limit of -1350 mV. The mixed −1 −1 solution contained 1 mol L NH4Cl and the zinc-only solution contained 2 mol L NH4Cl in order to maintain constant chloride concentration.

77 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES hindered, consistent with the findings by Gómezet al. for both Zn-Co and Zn-Fe electroplating, which suggested that the anomalous behavior depended mainly on zinc ion transport [114, 166]. The influence of mass transport was also observed in experiments measuring the effects of total metal ion concentration as support concentration, as shown in Figure A.14 and A.15. In terms of the positive Fe(II)/Fe(III) electrode, only a slight depression of the electrode activity was apparent in a solution containing excess zinc (0.2 mol L−1

−1 −1 FeCl2, 0.2 mol L FeCl3, 0.8 mol L ZnCl2), and there was no significant difference in peak shifting with increasing scan rate between the two solutions (see Figure 6.8). In both cases, the peak current varied linearly (R2 > 0.99) with the square root of the scan rate, consistent with fast reactions controlled by mass transport. This result confirms that in the mixed system, the positive Fe(II/III) electrode can behave in the same way as it does in iron-chromium, iron-hydrogen, and all-iron batteries. 2

30 − (a) 60 (b) 40 20 mAcm / a , 20 p i

2 10 3 4 5 6 7 8 − − (a) ν1/2/mV1/2s 1/2 (b) 0 mAcm i/ −10

−20

−30 0 0.2 0.4 0.6 0.8 1

E/VAg/AgCl

Figure 6.8: Eﬀect of zinc chloride on Fe(II)/(III) redox reaction on graphite electrode −1 −1 with 1 mol L NH4Cl supporting electrolyte at pH=1 at ν = 10 mV s . Initial electrolyte (a) contained 0.2 mol L−1 Fe(II) and 0.2 mol L−1 Fe(III), and the modiﬁed electrolyte (b) also contained 0.8 mol L−1 Zn(II). Inset: scans performed at 10, 30, and 50 mV s−1 showed only a small decrease in the current peak.

A proof-of-concept battery operating at (T = 25 ◦C) was tested over 10 days and 127 cycles of continuous charge-discharge cycling at 25 mA cm−2, where each charge

78 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES

was carried out for one hour (charge loading = 25 mAh cm−2, a similar value of metal loading as described in Ref. [17]). The open-circuit potential of about 1.5 V suggested that (1) the deposit contained primarily zinc, (2) the potential of the positive electrode was not affected by the presence of zinc and (3) the displacement of zinc by iron was not sufficiently fast so as to bring the electrode to the iron potential, where an OCV of about 1.2 V would have been observed. However, it was fast enough to produce some hydrogen gas as evidenced by visual inspection of the flowing electrolyte. Although this reaction does lead to some self-discharge, the electrolytes can be drained from the stacks during any gaps between charging and discharging in order to minimize any self-discharging when the battery is not in use. Voltage profiles were relatively flat, as shown for the first charge-discharge (Cycle # 1 of Figure 6.11) in Figure 6.9. As illustrated in Figure 6.10, a significant improvement in the discharge voltage, as well as the voltaic efficiency was observed when compared to an all-iron battery operated using the same cell hardware.

1.7 1.6 1.5 1.6 1.4 1.5 i = 35 mA cm−2 1.3 −2 1.4 loading = 105 mAh cm 1.2 V.E. = 81.0 % 1.1 1.3 C.E. = 88.7 % 1 1.2 E.E. = 71.8 % cell potential/volts 0.9 cell potential/volts 1.1 0.8 0.7 1 0 0.5 1 1.5 0 1 2 3 4 5 time/h time/h

Figure 6.9: (a) Close-up view of the ﬁrst charge-discharge (Cycle # 1 in Figure 6.11) at 25 mA cm−2, operating at T = 25 ◦C. (b) Close-up view of one cycle from a similar test demonstrating higher loading and current density.

The cell potential of the battery during the first two days (25 cycles) is shown in Figure 6.11 (a), and the variation of efficiencies during 127 cycles over a 10-day period is shown in Figure 6.11 (b). The battery test was voluntarily stopped after 10 days, which was deemed to be sufficient for demonstrating operation of the mixed-electrolyte battery system. The performance was relatively stable, without

79 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES

1.8

1.5 Zinc-Iron Chloride 1.2 Discharge: 1.35 V V.E.: 83 % 0.9 All-Iron Chloride Discharge: 0.95 V 0.6 V.E.: 66 % cell potential / V 0.3

0 0 1 2 time / h

Figure 6.10: Comparison of charging and discharging proﬁles between all-iron chloride and zinc-iron chloride at i = 25 mA cm−2 using ﬂat graphite plate as the negative electrode.

any evidence of degradation.

1.7 100 1.6 charging 1.5 80 OCV 1.4 discharging 60 1.3 1.2 V.E. 40 C.E.

efficiency/% E.E. 1.1 cell potential/volts 1 20 0.9 0.8 0 0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25 2.5 0 10 20 30 40 50 60 70 80 90 100 110 120 time/d cycle number (a) (b) Figure 6.11: (a) Cell potential as a function of time during the first two days of testing, and (b) estimated voltaic, coulombic and energy efficiencies during over 10 days of continuous charge-discharge cycling at T = 25 ◦C.

Initially, there was a pressure imbalance due to the fact that the positive electrolyte flowed through a porous felt whereas the negative electrolyte did not, so there was a gradual transfer of fluid from positive to negative. This was corrected by increasing the negative back-pressure with a needle valve (denoted as (i) in Figure 6.12 b); this increased the stability and increased coulombic efficiency, presumably by decreasing the rate of crossover from the positive electrolyte. It was also noted that during cycles 20-50, voltaic efficiency decreased while coulombic efficiency increased

80 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES

(reaching a maximum of 95 %). Based on visual observation of the color of the flowing electrolyte, we speculate that this was due to the formation of some hydroxide precipitates, which formed a coating on the separator and hindered crossover of Fe3+ and protons from the positive electrolyte. During the 65th cycle, 2.5 g of citric acid (equivalent to about 35 mmol L−1) was added to the electrolytes in order to complex the iron from the hydroxides (denoted as (ii) in Figure 6.12 b); this caused an increase in voltaic efficiency and a corresponding decrease in coulombic efficiency. Both effects may be attributed to clearing of the separator pores. Based on these results, identification of the appropriate complexing ligands, as well as their optimum concentrations, is an important aspect of the development of zinc-iron chloride flow batteries. The average coulombic, voltaic and energy efficiencies were 85 %, 80 % and 68 %, respectively, similar to those from Ref. [137], but operating with mixed electrolytes and a microporous separator. Considering the fast kinetics of the zinc and iron redox couples, the main source of low voltaic efficiency was likely the ohmic resistance in the separator and the large (∼ 2.2 mm) gap between positive and negative electrodes, as well as concentration polarization at the positive electrode (the Fe(II)/Fe(III) reaction is only a single-electron reaction, and the Fe2+ concentration was relatively low). As with all-iron batteries, the conductivity can range from about 100-300 mS cm−1 (as shown in Figure A.17, measured using the cell shown in Figure A.16), which is less than highly-acidic or highly-alkaline chemistries, but may be high enough for practical application. Cyclic voltammograms (not shown) indicated that the bulk plating and stripping processes were not significantly affected by the presence of the PEG. In the future, performance can be improved by optimizing various parameters such as the temperature, the electrolyte composition, the distance between electrodes, the electrode structure and the flow rate. It will also be important to determine the range of Zn:Fe and associated SoC that can be used for efficient battery operation; however, as with most hybrid flow batteries, any limitations on

81 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES the amount of energy stored will be dictated primarily by the cell geometry and volume for zinc plating rather than by the SoC of the electrolyte. Results from a similar battery cycling test with larger SoC swings is shown in Figure 6.12.

1.8 100

charging 1.6 80 ocv 1.4 60 discharging averages: V.E. E/V VE: 82 % 1.2 40 C.E. E.E. CE: 87 % eﬃciency / % EE: 71 % 1 20

0.8 0 0 5 10 15 20 25 30 0 20 40 60 80 100 120 140 160 180 t/d cycle number (a) (b) Figure 6.12: (a) Cell potential and (b) calculated cycle eﬃciencies during a 30-day, 175-cycle zinc-iron chloride battery test at room temperature with i = 25 mA cm−2 and two-hour charges (50 mAh cm−2).

The coulombic efficiencies were lower than those in batteries using separated electrolytes and ion-selective membranes, but still high enough to obtain energy efficiencies of over 60 % without optimization. It may be possible to reduce relative effects

3+ of both Fe crossover as well as H2 evolution by increasing the discharge current density, which would polarize the negative electrode in the positive direction. Also, in order to increase the stability for extended battery testing, an electrolyte rebalancing system should be incorporated in order to compensate for losing protons to hydrogen gas [140]. Most importantly, despite many days of operation using mixed electrolytes, neither the presence of Fe2+ nor Fe3+ in the negative electrolyte appeared to cause any irreversible performance degradation, and it was clear that zinc, rather than iron, was the primary species being deposited and stripped. The system cost was estimated using a model developed by PNNL [176], modiﬁed as needed for the zinc-iron chloride architecture (e.g., ZnCl2 and FeCl2 electrolytes, Daramic separators, etc.). Assuming a discharge at 1.2 V and 50 mA cm−2, we estimate the zinc-iron

82 CHAPTER 6. ZINC-IRON CHLORIDE FLOW BATTERIES chloride battery system would cost about $100 kWh−1 for a system capable of a 5.5 h discharge. However, achieving this will require additional research and development in order to better understand the cell behavior and the deposit morphology at higher states of charge. Preliminary investigations showed (see Figures A.20 - A.24) that a wide range of deposit morphology could be obtained, depending on the electrolyte formulation and deposition waveform.

6.4 Conclusions

It was shown that anomalous deposition of zinc from mixed ZnCl2-FeCl2 electrolytes can be used to enable zinc-iron chloride batteries that are crossover-tolerant and can use microporous separators. The inhibition of iron deposition by zinc ions was observed to occur without apparent dependence on substrate material, and iron incorporation into the deposits was only observed in cases with relatively low Zn/Fe (or high deposition current densities) that are outside the range of what is necessary for battery applications. The presence of excess zinc chloride (Zn/Fe = 4) only slightly depressed the activity of the Fe(II/III) reaction on graphite. A proof-of-concept ﬂow battery demonstrated an OCV of approximately 1.5 V and good stability during continuous charge-discharge cycling at 25 ◦C. Based on the cell behavior, it was clear that the primary processes occurring at the negative electrode were zinc deposition and stripping, even when mixed with relatively high concentrations of Fe2+. With further development, zinc-iron chloride batteries could achieve an excellent balance between cost, safety, and performance for grid-scale energy storage applications.

83 Chapter 7

Conclusions and Recommendations for Future Research

84 CHAPTER 7. CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE RESEARCH This work helped advance the state-of-the-art of iron-based flow batteries in three ways. The first way was to implement an in-tank hydrogen-ferric ion recombination method. This was based on a novel reactor design, which was characterized electrochemically inside pressurizable vessels. The second way was to build a mathematical framework for understanding iron flow battery system dynamics, including side-reactions and rebalancing processes. Lastly, a new type of zinc-iron hybrid flow battery based on anomalous codeposition was demonstrated. The conclusions from these investigations are outlined below, and recommendations are made for future research.

Rebalancing

One of the main challenges of using iron-based flow batteries is correcting the chemical imbalance caused by the hydrogen evolution side reaction. A promising approach explored in this work was to use internal rebalancing, wherein no external hydrogen sources or control systems are used. This involved the development of sealed electrolyte reservoirs as well as reactors designed to facilitate hydrogen-ferric ion recombination. It was shown that capillary-action galvanic reactors (CGR) based on catalyzed carbon felt were able to carry out this reaction sufficiently fast so that they can be placed inside the positive electrolyte reservoir, simplifying the system design without increasing the required area or “footprint.” It was shown that such reactors can be characterized using one-, two- and three-electrode measurements in pressurizable vessels. By using a three-dimensional reactor geometry that extends vertically into the reservoir headspace, it was possible to obtain reaction rates about an order of magnitude higher (in terms of the reactor area at the waterline) than those reported for two-dimensional, paper-based reactors used in vanadium batteries. Also, a limiting current for hydrogen oxidation was identified when for a partial pressure

of PH2 =0.2-0.3 psi. The performance improved with increasing hydrogen pressure up

85 CHAPTER 7. CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE RESEARCH

to about PH2 =10 psi, beyond which the reaction became limited by ferric-ion transport and ohmic losses. However, in some cases the reactor could eventually become ﬂooded, either by excessive capillary wetting or by electrolyte splashing during battery operation. Therefore, future work should focus on improving hydrophobicity of the catalyst layer without limiting the capillary wetting required to establish the ionic pathway. It may be possible to borrow methods from other galvanic cells that have liquid on one side and gas on the other side, such as iron-hydrogen or nickel- hydrogen battery systems. Another direction could be to separate the hydrogen and ferric electrodes so that the current can be directly measured, however this will decrease performance and increase complexity. Finally, both the catalyst loading and catalyst material should be optimized. It may be possible to use a non-PGM material such as tungsten carbide to oxidize the hydrogen rather than platinum [177–179]. Alternatively, the platinum loading should be reduced to lower values on the order of 0.2 mg/cm−2 or less. With continued work in this area, it should be possible to develop sealed aqueous ﬂow battery systems with exceptional longevity and minimal maintenance.

System Modeling

A new system-scale, dynamic flow battery was developed in order to help elucidate the nature of the chemical imbalance and rebalancing processes in sealed flow batteries. Whereas most previous models neglected side-reactions, the presently-developed model incorporated hydrogen and Fe3+ side-reactions, as well as the rebalancing reactions necessary for the system to achieve steady-state operation. A good agreement was found between simulated and measured profiles. The steady-state pressure occurs when the generation rate is opposite to the consumption rate, so the pressure can be reduced either by raising the pH (to reduce the average hydrogen generation rate) or by improving the recombination reactor performance or area. The effects of separa-

86 CHAPTER 7. CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE RESEARCH tor porosity and thickness on steady state pressure are analogous to their effects on crossover; increasing porosity, or decreasing thickness, leads to faster diffusion from the positive to the negative electrolyte and hence a lower pH (the positive electrolyte pH is always lower than the negative). The lower pH then leads to faster hydrogen generation and a higher steady-state pressure. On the other hand, decreasing porosity or increasing separator thickness allows the negative pH to rise to a higher steady-state value (i.e., a larger pH gradient can be realized between negative and positive electrolyte), reducing the hydrogen generation rate and steady-state pressure. There is considerable room for future development of these methods and measurements. Specifically, the pH values in the negative and positive electrolytes should be measured continuously, using data acquisition with isolated inputs so that there can be no “cross talk” between the various sensors. It would also be helpful to continuously measure iron concentrations, by using a method based on cyclic voltammetry or spectrophotometry. Lastly, it was speculated that there may have been some oxygen contamination due to diffusion through the Norprene tubing used by the peristaltic pump. In the future, the system should be modified to use standard (e.g., centrifugal) pumps with only hard tubing.

Zinc-Iron Chloride Flow Batteries

Lastly, it was shown that anomalous phenomena occurring during electrodeposition

from mixed ZnCl2-FeCl2 electrolytes can be exploited to enable zinc-iron chloride batteries that operate with microporous separators and which do not experience in- herent capacity fade due to species crossover. While anomalous codeposition phenomena have been known for decades, this was the first study taking advantage of it for battery applications. This work helped fill in the gaps in terms of how the phenomena could be used for batteries, wherein the species concentrations, current densities, and other factors are different from coatings applications. It was shown that indeed,

87 CHAPTER 7. CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE RESEARCH iron electrodeposition is strongly inhibited in the presence of zinc ions, and this can be used eﬀectively in conditions useful for battery applications. However, there are still many questions and challenges that should be addressed in future work. The question of zinc morphology was only lightly considered in this work, and so future studies should identify the appropriate additives and/or pulse parameters required for smooth, adherent deposits required to achieve high metal loading greater than 100 mAh cm−2. Also, most hydrogen in zinc-iron chloride batteries is formed during battery discharging, as shown in Figure A.26. It is hypothesized that the corrosion during discharging is accelerated by the galvanic displacement of zinc by iron ions in the electrolyte. However, this has not been veriﬁed, and such displacement reactions have not been studied in the literature. Therefore, it will be important to investigate it further, e.g., by hydrogen generation rate on a zinc surface as a function of cFe2+ .

88 Appendix A

Supporting Information

89 APPENDIX A. SUPPORTING INFORMATION

A.1 Chemicals Used

Table A.1: Information regarding the chemicals used in the hydrogen-ferric ion recombination, all-iron battery and zinc-iron battery tests.

Substance Formula Supplier CAS Purity Notes Zinc Chloride ZnCl2 Sigma 7646-85-7 ≥ 98 % anhydrous Iron(II) Chloride FeCl2 · xH2O Alfa 23838-02-0 99 % crystalline Iron(III) Chloride FeCl3 · 6H2O Alfa 10025-77-1 ACS/97% lump Ammonium Chloride NH4Cl Alfa 12125-02-09 ≥ 95 % granular Sodium Chloride NaCl Alfa 7647-14-5 ≥ 99 % crystalline Citric Acid C6H8O7· H2O Sigma 5949-29-1 98 % crystalline L-Ascorbic Acid C6H8O6 Sigma 50-81-7 reagent crystalline

A.2 In-Tank Hydrogen-Ferric Ion Recombination

The cyclic voltammograms (CV) and electrochemical impedance spectroscopy (EIS) experiments were carried out using the following steps:

1. Prepare the electrolyte (2 M NaCl, pH = 1.4)

2. Nitrogen purge the electrolyte for 10 minutes

3. Add 100 mL of electrolyte to the reservoir

4. Connect the cap ﬂange

5. Tighten the ﬂange bolts using wrenches

6. Connect the potentiostat leads to the titanium leads on the cap ﬂange

7. Purge nitrogen through the vessel for 5 minutes

8. Pulse-inject hydrogen gas to desired pressure (1-10 psig)

9. Carry out CV and EIS

90 APPENDIX A. SUPPORTING INFORMATION

10. Purge nitrogen through the vessel for 5 minutes

91 APPENDIX A. SUPPORTING INFORMATION

Measurement of the hydrogen recombination rate on the CGR was carried out using the following steps:

1. Prepare the electrolyte (1 M FeCl2, 1 M FeCl3, 2 M NaCl)

2. Nitrogen purge the electrolyte for 10 minutes

3. Add 100 mL of electrolyte to the reservoir

4. Connect the cap ﬂange

5. Tighten the ﬂange bolts using wrenches

6. Connect the potentiostat leads to the titanium leads on the cap ﬂange

7. Purge nitrogen through the vessel for 5 minutes

8. Pulse-inject hydrogen gas to desired pressure (1-10 psig)

9. Record pressure as a function of time

10. Purge nitrogen through the vessel for 5 minutes

92 APPENDIX A. SUPPORTING INFORMATION

Battery tests were carried out using the following parameters:

Table A.2: Conditions for all-iron battery tests.

temperature room-temp current density 20 mA cm−2 positive electrode graphite plate negative electrode graphite plate separator Daramic SLI 175 ﬂow rate 50 mL/min electrolyte volume 100 mL electrode geometric area 6.25 cm2 negative electrolyte 1 M FeCl2, 2 M NaCl positive electrolyte 1 M FeCl2, 2 M NaCl, 0.5 M FeCl3 charge time one-hour discharge cutoﬀ 0 V

Much of the work presented in the thesis used a custom, pressurizable vessel design as shown in Figure A.1. These vessels were used both for CGR performance testing as well as for battery testing and measurement of hydrogen generation rates.

93 APPENDIX A. SUPPORTING INFORMATION

Bolt

2”

Cap Flange

Gasket

Female Flange

Nut

PVC tubing 4 ”

End Cap

(a) (b)

Figure A.1: Sealed reservoir (a) schematic and (b) photograph. Ports were installed in the cap ﬂange by drilling and tapping manually. Titanium rods (1/8” diameter) were used for making electrical connections.

Representative behavior from pressure tests is shown in Figure A.2, where the

recombination reaction was carried out in a solution containing 0.5 M FeCl3. Up to a ≈ hydrogen pressure of PH2 0.2 atm, the reaction rate was linearly proportional to the partial pressure. At higher partial pressures of hydrogen, however, the rate entered a mixed-control and then a limiting-current regime controlled by Fe3+ diﬀusion and ohmic losses. This analysis was slightly aﬀected by the changing concentration of Fe3+ during the reaction process (about 8 % drop in molarity for an experiment using

one atmosphere of H2). In some cases, an Ag/AgCl reference electrode was also installed in the vessel during performance testing. This two electrode setup (see Figure A.3) showed how the mixed potential of the CGR varied after pulse-injecting 3-5 psig of hydrogen into the sealed reservoir.

94 APPENDIX A. SUPPORTING INFORMATION

1.9 60 1.8 50 1.7 1.6 40 1.5 30 (atm) 1.4 (mA) mixed control ohmic control I P 1.3 20 1.2 10 H2 1.1 control 1 0 0 5 10 15 20 25 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

t (h) PH2 (atm)

(a) (b)

3+ Figure A.2: Measured (a) pressure and (b) rate behavior of the H2-Fe recombination reaction in a sealed vessel. Hydrogen pressure was controlling at low pressures, but 3+ at higher values of PH2 , the rate became controlled by Fe and ohmic losses.

3 ) g 2 (psi 2

H 1 P

200 pseudo steady - state (mA) 100 I

(mV) -200 3 / 2 Fe E -400 0 0.5 1 1.5 t (h)

Figure A.3: Pressure, reaction rate, and potential of a CGR (area of the catalyzed 2 area in the headspace, AHS = 6 cm ) as a function of time after pulse-injecting 3 psig of hydrogen using the setup illustrated in Figure A.1.

95 APPENDIX A. SUPPORTING INFORMATION

Battery testing procedure

The procedure for battery testing consisted of the following main steps:

1. Assemble the battery cell

2. Connect pumps to cell and electrolyte reservoirs

3. Flow with 1 M NaCl rinse solution for 30 minutes while checking for any ﬂow- related problems

4. Remove rinse solution and add the negative and positive electrolytes

5. Purse headspace gas with nitrogen for 10 minutes

6. Seal the reservoirs by securing all the compression ﬁttings on reservoir ports

7. Turn on all data logging equipment and begin the battery cycling test

96 APPENDIX A. SUPPORTING INFORMATION

Battery cycling setup

The following battery test “setup” is representative of most of the battery cycling tests:

1. Potentiostatic electrochemical impedance spectroscopy (PEIS) from 10 kHz to 10 Hz with 10 mV amplitude and measuring at 6 points per decade

2. Chronopotentiometry (constant-current) charging for 1-2 hours

3. Open-circuit voltage (OCV) measurement for 10 seconds

4. Repeat Step 1 (PEIS)

5. Chronopotentiometry (constant-current) discharging until reaching a cell cutoﬀ voltage of 0 V

6. Low-current (“exhaustive”) discharge at 5 mA cm−2 until reaching a cell cutoﬀ of 0 V

7. Repeat Step 3 (OCV)

8. Loop to Step 1

97 APPENDIX A. SUPPORTING INFORMATION

Table A.3: Sample of raw pressure transducer vs time data after pulse-injection hydrogen into a sealed reservoir containing a hydrogen-ferric ion recombination reactor.

number date time transducer voltage 1 16/01/03 13:46:09 2.74228 2 16/01/03 13:46:10 2.74189 3 16/01/03 13:46:11 2.74174 4 16/01/03 13:46:12 2.74144 5 16/01/03 13:46:13 2.74113 6 16/01/03 13:46:14 2.74083 7 16/01/03 13:46:15 2.74052 8 16/01/03 13:46:16 2.74037 9 16/01/03 13:46:17 2.74006 10 16/01/03 13:46:18 2.73976

98 APPENDIX A. SUPPORTING INFORMATION

Mathematica code for calculation of reaction rate from pressure data time = data[[All, 1]]; voltage = data[[All, 3]]; pressure = (voltage*11.81 - 13.26 - 14.7)/14.7; tavg = 30; pvt = MovingAverage[Transpose[{time, pressure}], tavg]; pvtf = Interpolation[pvt]; avgtime = MovingAverage[time, tavg]; avgp = MovingAverage[pressure, tavg]; Vhs = 0.113; Ageo = 3; avgi = Table[(-1*pvtf’[t]*Vhs*2*96485*1000)/(Ageo*0.08206*298.15), {t,First[avgtime], Last[avgtime]}]; avgivp = Transpose[{avgp, avgi}]; Export["output_data.dat", MovingAverage[avgivp, 5]]

Table A.4: Calculated hydrogen oxidation rate from pressure data.

−2 p/atm i/mA cmgeo 0.28 64 0.26 58 0.24 55 0.22 52 0.21 46 0.19 43 0.18 42 0.16 40 0.15 38 0.14 35 0.13 33 0.11 28 0.10 27 0.10 23 0.09 24 0.08 24 0.07 22 0.07 18 0.06 19 0.05 17

Flow battery tests were based on a simple cell as shown in Figure A.4. Each cell consisted of PVC end plates, copper current collectors, teﬂon gaskets, and graphite electrodes with bonded carbon felt. In some cases, the negative electrode was a ﬂat

99 APPENDIX A. SUPPORTING INFORMATION plate with plastic ribs used as spacers, as illustrated in Figure A.19.

Figure A.4: Typical 30-cm2 cell hardware.

Figure A.5: Photograph of a complete battery system.

100 APPENDIX A. SUPPORTING INFORMATION

Figure A.6: Floating CGR array.

Figure A.7: Appearance of electrolytes after 10 days of battery testing.

101

APPENDIX A. SUPPORTING INFORMATION

✶ ✵

✻

✵

✶

✶✹✵

✹ ✿✽

✽

✶ ✵

✷

✿ ✵

✷ ✺

✶ ✵✵

✻

✂

✄

✁

✂

✭

✽ ✵ ✵✿✼✹

❂ ✵ ♣✠✡

✁

❍ ❣

✟

✭

■

❩

✹

✵

✻

✹ ✵

✵✿✼ ✹ ✷

✵

✷

✷✿✺ ✵

✵

✡ ♣☛☞

❍ ✝

✠

✹✿

✽

✵

✲✵ ✵ ✵ ✵✹ ✵ ✵✽ ✶

✷ ✷ ✻

✷ ✹ ✵ ✷ ✹ ✷✵ ✷✷

✻ ✽ ✶ ✶ ✶ ✶ ✻ ✶ ✽

❊ ☎ ✆ ✮

✄ ✆✝✞

❆ ✝ ❂❆ ✝✞✟ ❘☎ (a) (b)

Figure A.8: Characterization of (a) hydrogen polarization and (b) electrochemical impedance response on a capillary-action galvanic reactor for diﬀerent values of pH2 .

1.8 22 battery off without rebalance 1.6 battery o 21

1.4 20 1.2 without recomb 19 1

/psig 18 2 p/psia H 0.8

P battery o with recomb 0.6 17 with rebalance (continuous cycling) 0.4 16 0.2 15 0 14 0 5 10 15 20 25 30 35 40 45 50 0 2 4 6 8 10 12 14 16 18 t/h t/h (a) (b)

Figure A.9: Comparison of pressure proﬁles with and without rebalancing (a) at 25 mA cm−2 and (b) 100 mA cm−2, including pressure after turning the battery oﬀ.

102 APPENDIX A. SUPPORTING INFORMATION

(a) (b)

Figure A.10: (a) Negative and (b) positive sides of a Daramic separator after battery cycling.

103 APPENDIX A. SUPPORTING INFORMATION

Current distribution

A preliminary model for the current distribution in CGRs was developed using gmsh for mesh generation, Elmer Multiphysics for simulation, and Paraview for post- processing. Below, results from a secondary current distribution, as well as the input simulation code, are shown.

Table A.5: Secondary current distribution

Governing Equations

∇2Φ = 0 (Laplace Equation) i = −κ∇Φ (Ohm’s Law)

Boundary Conditions

− i0,AnAF − 0 B.C. 1: i = RT (Φ ΦA) − i0,BnBF − 0 B.C. 2: i = RT (Φ ΦB)

Parameters

2 i0,A = 10 A/m nB = 1 2 0 i0,B = 10 A/m ΦA = 0 V 0 T = 300 K ΦB = 0.77 V

κ = 14 S/m A ⇒ H2 3+ nA = 2 B ⇒ Fe

104 APPENDIX A. SUPPORTING INFORMATION

1 0.9 0.8 0.7 0.6 12 mm

max 0.5

i/i 0.4 6 mm 0.3 0.2 3 mm 0.1 0 0 0.2 0.4 0.6 0.8 1 l/L (a) (b)

Figure A.11: Simulation of the secondary current distribution, assuming linear kinetics, and comparison of normalized hydrogen current (in the x-direction) as a function of reactor thickness. For this example, L=4 cm.

1 1

0.8 0.8

0.6 0.6 max max i/i ϕ/ϕ 0.4 0.4

0.2 0.2

0 0 -0.5 -0.3 -0.1 0.1 0.3 0.5 -0.5 -0.3 -0.1 0.1 0.3 0.5 l/L l/L (a) (b)

Figure A.12: Simulated potential and current as a function of vertical reaction length, from numerical simulation.

105 APPENDIX A. SUPPORTING INFORMATION

This ﬁle was used as the input for ﬁnding the current distribution, as shown in Figure A.11 and A.12. The software used was Elmer (CSC, IT Center for Science).

Header CHECK KEYWORDS Warn Mesh DB "." "." Include Path "" Results Directory "" End Simulation Max Output Level = 5 Coordinate System = Cartesian Coordinate Mapping(3) = 1 2 3 Simulation Type = Steady state Steady State Max Iterations = 1 Output Intervals = 1 Timestepping Method = BDF BDF Order = 1 Solver Input File = case.sif Post File = case.vtu End Constants Gravity(4) = 0 -1 0 9.82 Stefan Boltzmann = 5.67e-08 Permittivity of Vacuum = 8.8542e-12 Boltzmann Constant = 1.3807e-23 Unit Charge = 1.602e-19 End Body 1 Target Bodies(1) = 1 Name = "Body 1" Equation = 1 Material = 1 End Solver 1 Equation = Static Current Conduction Calculate Volume Current = True Variable = Potential Procedure = "StatCurrentSolve" "StatCurrentSolver" Exec Solver = Always Stabilize = True Bubbles = False Lumped Mass Matrix = False Optimize Bandwidth = True Steady State Convergence Tolerance = 1.0e-5 Nonlinear System Convergence Tolerance = 1.0e-7 Nonlinear System Max Iterations = 50 Nonlinear System Newton After Iterations = 3 Nonlinear System Newton After Tolerance = 1.0e-3 Nonlinear System Relaxation Factor = 1 Linear System Solver = Iterative Linear System Iterative Method = GMRES Linear System Max Iterations = 1000 Linear System Convergence Tolerance = 1.0e-10

106 APPENDIX A. SUPPORTING INFORMATION

BiCGstabl polynomial degree = 2 Linear System Preconditioning = Diagonal Linear System ILUT Tolerance = 1.0e-3 Linear System Abort Not Converged = False Linear System Residual Output = 1 Linear System Precondition Recompute = 1 End Equation 1 Name = "Equation 1" Active Solvers(1) = 1 End Material 1 Name = "Material 1" Electric Conductivity = 13 Porosity Model = Always saturated End Boundary Condition 1 Target Boundaries(2) = 2 4 Name = "h2_linear" Current Density = Variable Potential; Real MATC "-774*(tx-0.0)" End Boundary Condition 2 Target Boundaries(2) = 1 5 Name = "fe_linear" Current Density = Variable Potential; Real MATC "200*(1-exp(12*(tx-0.7)))" End

107 APPENDIX A. SUPPORTING INFORMATION

Pandas script for modifying the text ﬁle output from Biologic # this program takes cycling data in the three-column format: time \t voltage \t q # and then counts the cycle numbers based on finding the value 0 for the charge density, # which indicates a new cycle # it then prints an improved "cleaned" file that includes cycle numbers and unique keys datafile=open(’charge.txt’,’r’) t=[] v=[] q=[] for item in datafile: splitted= item.split() t.append(splitted[0]) v.append(splitted[1]) q.append(splitted[2]) # remove headers t=t[1:] v=v[1:] q=q[1:] cyclenum=0 # this is meant to be used with the > function to generate an output file print ’pointnum’ + ’\t’ + ’cyclenum’ + ’\t’ + ’time’ + ’\t’ + ’voltage’ + ’\t’+ ’chargedensity’ for item in range(len(t)): if float(q[item])==0: cyclenum+=1 #print "cyclecount = "+str(cyclecount) print str(item) + ’\t’ + str(cyclenum) + ’\t’ + t[item] + ’\t’ + v[item] + ’\t’+ q[item] datafile.close() # this program takes a ’cleaned’ data file that includes unique keys and cycle numbers, and then it # runs some simple stats using groupby to find, e.g., the mean charge or discharge voltage for each cycle import pandas as pd pd.set_option(’display.max_rows’, 200) df = pd.DataFrame.from_csv(’znfel37_part3_discharge_cleaned.tsv’ , sep="\t") print df print df.head() print df.describe() print df.groupby(’cyclenum’).mean()[’voltage’] print df.groupby(’cyclenum’).min()[’chargedensity’]

108 APPENDIX A. SUPPORTING INFORMATION

Pandas script for calculating cycle eﬃciency values # this program takes a ’cleaned’ data file that includes unique keys and cycle numbers, and then it # runs some simple stats using groupby to find, e.g., the mean charge or discharge voltage for each cycle import pandas as pd pd.set_option(’display.max_rows’, 200) df = pd.DataFrame.from_csv(’znfel40_charge_cleaned.tsv’, sep="\t") print df print df.head() print df.describe() print df.groupby(’cyclenum’).mean()[’voltage’] print df.groupby(’cyclenum’).min()[’chargedensity’]

109 APPENDIX A. SUPPORTING INFORMATION

A.3 Zinc-Iron Chloride Flow Batteries

Additional information about the zinc-iron chloride battery experiments can be found below. A sketch of the three-compartment glass cell is shown in Figure A.13. Eﬀects of the total metal ion concentration and supporting electrolyte concentration are shown in Figure A.14 and A.15. Conductivity was measured using a separate glass cell, which is illustrated in Figure A.16. Measured conductivity values for several

ZnCl2-FeCl2 electrolyte compositions are shown in Table A.6 and Figure A.17.

ref W.E. C.E.

frit

Figure A.13: Schematic of the H-Cell used for cyclic voltammetry.

110 APPENDIX A. SUPPORTING INFORMATION

250 200 CMe2+ = 3.0 M 150 CMe2+ = 2.0 M 100

2 C 2+ = 1.0 M

− Me 50 0 mAcm i/ −50 −100 −150 −200 −1.2 −1 −0.8 −0.6 −0.4 −0.2 0

E/VAg/AgCl

Figure A.14: Eﬀect of total metal ion concentration on voltammograms for mixed zinc-iron chloride electrolytes on a titanium substrate.

150 150 1 M KCl

100 3 M KCl 100 1 M NH4Cl

3 M NH4Cl 2 2

− 50 Sat’d KCl − 50 5 M NH4Cl

mAcm 0 0 /mAcm i i/ −50 −50

−100 −100 −1.2 −1.1 −1 −0.9 −0.8 −0.7 −0.6 −0.5 −0.4 −1.2 −1.1 −1 −0.9 −0.8 −0.7 −0.6 −0.5 −0.4 E/VAg/AgCl E/VAg/AgCl (a) (b)

Figure A.15: Eﬀect of supporting electrolyte concentration on deposition and stripping behavior of mixed zinc-iron chloride electrolytes.

111 APPENDIX A. SUPPORTING INFORMATION

leads

platinized Pt

Figure A.16: Sketch of glass conductivity measurement cell (l/A = 200 cm−1).

Table A.6: Measured conductivity at T = 30 5◦C obtained using the conductivity cell shown in Figure A.16.

−1 [ZnCl2/M] [FeCl2 · 4H2O/M] Support pH κ/mS cm 0 1 3 M NH4Cl 2 286 0.5 0.5 3 M NH4Cl 2 292 1 0 3 M NH4Cl 2 257 0.8 0.2 0 M 1 96 0.8 0.2 1 M NH4Cl 1 162 1 0 1 M NH4Cl 1 155 0 1 1 M NH4Cl 1 185 0.8 0.2 1 M KCl 1 159 0.8 0.2 3 M NH4Cl 1 277 0.8 0.2 3 M NaCl 1 204 0.8 0.2 1 M NaCl 1 136 0.8 0.2 1 M NH4Cl 1 161

112 APPENDIX A. SUPPORTING INFORMATION

300

250 NH4Cl KCl

1 NaCl

− 200

/mScm 150 κ

100

50 0 0.5 1 1.5 2 2.5 3 support concentration / M

Figure A.17: Measured conductivities of mixed zinc-iron chloride electrolytes as a function of diﬀerent supporting electrolytes.

113 APPENDIX A. SUPPORTING INFORMATION

Morphology and Composition

Typically, the deposits from the mixed electroytes were rough and dendritic. How- ever, several experiments confirmed it was possible to produce uniform deposits by modifying the electrical waveform, the flow rate, or the elelctrolyte additives. Below, some examples of morphology from both battery tests (on graphite substrate) and three-electrode studies (on titanium substrate) are shown. For the zinc-iron chloride battery tests, the negative side of the cell used Teflon ribs, as illustrated in Figure A.18. Pictures of the zinc deposits on graphite from the battery tests are shown in Figures A.20 and A.21.

Daramic current end plastic separator collector plate ribs

carbon graphite 1/8” tubing teﬂon felt plate gasket A=6.25 cm2 thickness = 2 mm

Figure A.18: Schematic of cell parts when using ribs.

114 APPENDIX A. SUPPORTING INFORMATION

Figure A.19: Photograph of an assembled 6.25 cm2 cell.

(a) (b)

Figure A.20: (a) Bulk deposit at 100 mAh cm−2 on graphite surface from mixed −2 ZnCl2-FeCl2 using pulse plating (0.1 ms on, 0.1 ms oﬀ, iavg = 25 mA cm ). (b) Corresponding separator with two PTFE ribs to stabilize separator.

It was observed in the present work that, when depositing at 20 mA cm−2 from quiescent solutions, the morphology appeared to become rougher and more dendritic with increasing Zn/Fe molar ratio, as might be expected. This can be observed in Figure A.22, which shows optical microscope pictures of the morphology for from three different Zn/Fe molar ratios of 0.5, 1.0 and 1.5. It was found that using pulse-plating had a beneficial effect on the deposit morphology, as illustrated in Figure A.23. Whereas the direct-current deposit contained larger dendritic growths, the deposit made using a pulse-waveform looked compara-

115 APPENDIX A. SUPPORTING INFORMATION

Figure A.21: Close-up optical microscope image of 100 mAh cm−2 deposit on graphite −2 using pulse plating (0.1 ms on, 0.1 ms oﬀ, iavg = 25 mA cm ).

0.5 1.0 1.5

Figure A.22: Eﬀect of zinc/iron ratio on plating morphology from quiescent solutions. Increasing the concentration of zinc promotes dendritic growth in the deposit, whereas increasing the iron concentration promotes a smoother deposit. tively smoother. Organic additives can likely produce similar results. For example, Figure A.24 shows a zinc deposit at 50 mAh cm−2 made using a DC waveform from a stirred solution in the presence of tetrapentylammonium chloride. The electrochemical studies of plating from the mixed electrolytes, including both cyclic voltammograms as well as steady-state plating and stripping, provided strong evidence that zinc was the predominant species being electroplated and stripped, even from equimolar solutions (e.g., 0.5 M ZnCl2, 0.5 M FeCl2). In many cases, the measurements from the mixed solutions looked nearly identical to those from the

ZnCl2-only solutions, suggesting that the anomalous suppression of iron electrodeposition was not only observed, but that it was a very strong eﬀect. This result was

116 APPENDIX A. SUPPORTING INFORMATION

DC Pulsing

Figure A.23: Usage of pulse plating to improve deposit morphology from quiescent solutions, without additives. DC Plating was carried out at 20 mA/cm2, 20 mAh/cm2. Pulsing was -40 mA/cm2 with 1 ms on, 1 ms oﬀ (50 % duty cycle). Electrolyte contained 0.5 M ZnCl2, 0.5 M FeCl2 and 3 M NH4Cl.

Figure A.24: Close-up optical microscope image of 50 mAh cm−2 of zinc deposited onto a titanium substrate at 25 mA cm−2 from stirred solution containing 0.5 mol L−1 −1 −1 ZnCl2, 0.5 mol L FeCl2 and 0.002 mol L tetrapentylammonium chloride using a three-electrode cell. also consistent with XPS analysis of the surfaces of several of the samples, for which it was challenging to identify the presence of any iron peaks. An example of this is shown in Figure A.25, wherein the only observed iron peak was found in the deposit made from an electrolyte with Zn/Fe molar ratio of 0.5, but not with Zn/Fe = 1.0 or Zn/Fe = 1.5.

117 APPENDIX A. SUPPORTING INFORMATION

140000 2300 120000 fe3p (Zn/Fe = 0.5) 1900 100000

counts 1500

80000 1100 80 70 60 50 40 30

counts 60000 binding energy / eV

40000

20000

0 1000 800 600 400 200 0 binding energy / eV

Figure A.25: XPS analysis of three samples deposited at pH=2. The sample with Zn/Fe=0.5 showed a pronounced Fe3p peak and an estimated 10 %at iron.

118 APPENDIX A. SUPPORTING INFORMATION

Hydrogen Evolution Rate

As shown in Figure A.26, it was found that the vast majority of the hydrogen con- tribution to coulombic efficiency loss occurred during battery discharging. This was most likely because of the galvanic displacement of zinc by iron, followed by acid corrosion of the iron. Therefore, one of the main areas that should be investigated further is the factors governing the displacement of zinc by iron in chloride media. Nevertheless, the efficiencies measured were easily good enough to warrant further development, and could likely be improved to the point of commercial applicability. As with all-iron flow batteries, electrolyte rebalancing via hydrogen-ferric ion recombination is necessary for electrolyte stability during extended or long-term battery operation.

1.8 18 1.8 20 1.6 1.6 17.5 1.4 1.4 15

1.2 17 1.2 2 − 1 1 16.5 10 E/V E/V

0.8 0.8 mA cm p/psia / 2

0.6 16 0.6 H i 0.4 0.4 5 15.5 0.2 0.2 0 15 0 0 0 1 2 3 4 5 6 0 1 2 3 4 5 6 t/h t/h (a) (b)

Figure A.26: (a) Pressure measurements in a sealed ﬂow battery system without recombination, (b) calculated hydrogen current densities during charging and discharging.

119 APPENDIX A. SUPPORTING INFORMATION

Table A.7: Example of calculations used to generate Figure A.26. time transducer voltage p/psia p/atm ∆p/atm ∆p/∆t mA cm−2 11:50 0.2003 15.105 1.028 11:55 0.2004 15.111 1.028 4.16E-04 1.39E-06 0.66 12:00 0.2003 15.105 1.028 -4.16E-04 -1.39E-06 -0.66 12:05 0.2004 15.111 1.028 4.16E-04 1.39E-06 0.66 12:10 0.2004 15.117 1.028 4.16E-04 1.39E-06 0.66 12:15 0.2004 15.117 1.028 0.00E+00 0.00E+00 0.00 12:20 0.2005 15.123 1.029 3.64E-04 1.21E-06 0.58 12:25 0.2007 15.135 1.030 8.32E-04 2.77E-06 1.33 12:30 0.2007 15.135 1.030 0.00E+00 0.00E+00 0.00 12:35 0.2007 15.140 1.030 3.64E-04 1.21E-06 0.58 12:40 0.2008 15.146 1.030 4.16E-04 1.39E-06 0.66 12:45 0.2007 15.140 1.030 -4.16E-04 -1.39E-06 -0.66 12:50 0.2007 15.140 1.030 0.00E+00 0.00E+00 0.00 12:55 0.2011 15.169 1.032 1.97E-03 6.58E-06 3.16 13:00 0.2016 15.205 1.034 2.39E-03 7.97E-06 3.82 13:05 0.2023 15.257 1.038 3.59E-03 1.20E-05 5.73 13:10 0.2028 15.298 1.041 2.75E-03 9.18E-06 4.40 13:15 0.2035 15.350 1.044 3.59E-03 1.20E-05 5.73 13:20 0.2042 15.403 1.048 3.59E-03 1.20E-05 5.73 13:25 0.2048 15.450 1.051 3.17E-03 1.06E-05 5.07 13:30 0.2051 15.473 1.053 1.56E-03 5.20E-06 2.49 13:35 0.2052 15.484 1.053 7.80E-04 2.60E-06 1.25 13:40 0.2053 15.490 1.054 4.16E-04 1.39E-06 0.66 13:45 0.2055 15.502 1.055 7.80E-04 2.60E-06 1.25 13:50 0.2055 15.508 1.055 4.16E-04 1.39E-06 0.66 13:55 0.2055 15.508 1.055 0.00E+00 0.00E+00 0.00 14:00 0.2056 15.514 1.055 4.16E-04 1.39E-06 0.66 14:05 0.2056 15.514 1.055 0.00E+00 0.00E+00 0.00 14:10 0.2056 15.514 1.055 0.00E+00 0.00E+00 0.00 14:15 0.2057 15.519 1.056 3.64E-04 1.21E-06 0.58 14:20 0.2056 15.514 1.055 -3.64E-04 -1.21E-06 -0.58

120 APPENDIX A. SUPPORTING INFORMATION

A.4 System Model

The following computer code was developed using the Modelica programming lan- guage for simulation of the electrolyte dynamics in all-iron hybrid flow batteries including side reactions and rebalancing. The initial electrolyte compositions used in the model were based on those used in experimental battery tests (see Table A.8). The simulation can be carried out most easily by loading the electrolyte class file and the core flow battery model within OMEdit (see Figure A.27). The software is freely available on for all platforms at https://openmodelica.org. Optionally, for more advanced simulation settings, additional programming can be implemented. In this work, custom scripts were prepared using Python and bash scripts in order to carry out the simulation iteratively.

Figure A.27: Screenshot of the OMEdit development environment used for carrying out simulations written in Modelica.

121 APPENDIX A. SUPPORTING INFORMATION

Table A.8: Initial concentrations and diﬀusivities used in the simulation.

−3 −3 9 2 −1 cneg / mol m cneg / mol m D ·10 /m s Fe2+ 2500 1500 0.72 Fe0 0 0 n/a Fe3+ 0 500 0.61 Na+ 2000 2000 1.33 Cl− 7010 6750 2.03 H+ 10 250 9.31

Table A.9: Baseline parameter values used

parameter value conductivity 150 mS cm−1 ﬂow rate 25 mL min−1 separator porosity 0.5 negative electrolyte volume 125 mL positive electrolyte volume 250 mL cell area 30 cm−2 current density 100 mA cm−2 headspace volume 450 mL charge time 1 h separator thickness 225 µm

122 APPENDIX A. SUPPORTING INFORMATION

Table A.10: Example of measurements and calculations used to determine eﬀect of pH on hydrogen generation rate.

2 run initial pH ﬁnal pH average pH ∆p (atm) iH2 (A/m ) 1 1.92 1.89 1.9 0.0076 21 2 1.69 1.74 1.715 0.0069 19 3 1.54 1.69 1.615 0.0071 20 4 1.33 1.54 1.44 0.0078 22 5 1.04 1.1 1.07 0.0096 27 6 0.92 1.04 0.98 0.011 31 7 0.86 0.92 0.89 0.017 46 8 0.47 0.59 0.53 0.020 55 9 -0.06 0.02 -0.02 0.035 97 10 0.1 0.18 0.14 0.043 119 11 0.18 0.31 0.245 0.037 102 12 0.31 0.35 0.33 0.033 92 13 0.285 0.35 0.285 0.042 117 14 0.35 0.52 0.435 0.028 77 15 0.52 0.67 0.595 0.024 68 16 0.67 0.79 0.73 0.031 85

123 APPENDIX A. SUPPORTING INFORMATION

The electrolyte class class electrolyte parameter Real F = 96485; parameter Real MW = 0.018; //assume constant for now parameter Real Cp = 4184 * 0.018; //assume constant for now parameter Real rho = 1000; //assume constant for now parameter Real z1 = 2; parameter Real z3 = 3; parameter Real z4 = 1; parameter Real z5 = -1; parameter Real z6 = 1; parameter Real T = 298; constant Real R = 8.3144598; constant Real ccf = 20.0; constant Real dcf = 1.0; // conductivity, diffusivity linear "correction factors" //constant Real c4 = 1000; //constant Real c5 = 1000; //assume constant concentration of NaCl supporting electrolyte. //diffusivities in m^2/s! //iron effective diffusivities from tyler’s crossover paper //proton diffusivity just a guess for now // 1 is fe2+ // 3 is fe3+ // 4 is na+ // 5 is cl- // 6 is h+ parameter Real epsilon = 0.5; //membrane porosity parameter Real Dfe2 = dcf*0.72e-9 * epsilon ^ (3 / 2); parameter Real Dfe3 = dcf*0.61e-9 * epsilon ^ (3 / 2); parameter Real Dfe3soln = dcf*0.61e-9; parameter Real Dh2 = dcf*9.312e-9 * epsilon ^ (3 / 2); parameter Real Dna = dcf*1.334e-9 * epsilon ^ (3 / 2); parameter Real Dcl = dcf*2.032e-9 * epsilon ^ (3 / 2); //effective mobilities Real u1 = abs(z1) * F * Dfe2 / (R * T); Real u3 = abs(z3) * F * Dfe3 / (R * T); Real u4 = abs(z4) * F * Dna / (R * T); Real u5 = abs(z5) * F * Dcl / (R * T); Real u6 = abs(z6) * F * Dh2 / (R * T); //Neman versions Real u1n = u1 / 2 / F; Real u3n = u3 / 3 / F; Real u5n = u5 / 1 / F; Real u4n = u4 / 1 / F; Real u6n = u6 / 1 / F; parameter String name; parameter Real Cofe2; parameter Real Cofe0; parameter Real Cofe3; parameter Real Cona; parameter Real Cocl; parameter Real Coprotons;

124 APPENDIX A. SUPPORTING INFORMATION

Real Cfe2(start = Cofe2); Real Cfe0(start = Cofe0); Real Cfe3(start = Cofe3); Real Cna(start = Cona); Real Ccl(start = Cocl); Real Cprotons(start = Coprotons); // planning to eliminate the term ’cprotons’ Real conductivity(unit = "1/(ohm*m)"); Real sumzuc; Real tfe2; Real tfe3; Real tna; Real tcl; Real tprotons; equation sumzuc = abs(z1) * u1 * Cfe2 + abs(z3) * u3 * Cfe3 + abs(z4) // * u4 * Cna + abs(z5) * u5 * Ccl + abs(z6) * u6 * Cprotons; conductivity = F * sumzuc * ccf; tfe2 = abs(z1) * u1 * Cfe2 / sumzuc; tfe3 = abs(z3) * u3 * Cfe3 / sumzuc; tna = abs(z4) * u4 * Cna / sumzuc; tcl = abs(z5) * u5 * Ccl / sumzuc; tprotons = abs(z6) * u6 * Cprotons / sumzuc; // equation for conductivity (Bard 2.3.10), in 1/(ohm*m). //should be on order of 10 // iron metal not included since it isn’t contributing to conductivity end electrolyte;

125 APPENDIX A. SUPPORTING INFORMATION

Core ﬂow battery model class rfb2 // NE means negative cell electrolyte // PE means positive cell electrolyte // 1 means Fe2+, 2 means Fe0, 3 means Fe3+, 4 means Na+, //and 5 means Cl- // total volumetric flow rate constant Real R = 8.314; parameter Real ncells(unit = "None") = 1; //number of cells in a stack parameter Real VTankNeg(unit = "m^3") = 0.000125; parameter Real VTankPos(unit = "m^3") = 0.000250; //where VTankNeg is the (-) reservoir and VCellNeg is the // (-) flow cell, in m3 parameter Real ACell(unit = "m^2") = 0.003; // area of cell 1 reaction in m2 parameter Real j(unit = "A/m^2") = 1000; //current density in A/m2; 1000 A/m2 is equivalent to //100 mA/cm2 parameter Real nab(unit = "emol/mol") = 2; // mol e- per mol of metal deposited parameter Real F(unit = "C/emol") = 96485; //faradays’ number, (C/mole-) parameter Real go(unit = "meters") = 0.001; // initial gap in the cell on side 1 (deposition side) (m) parameter Real MWfe(unit = "kg/mol") = 0.055847; parameter Real rhofe(unit = "kg/m^3") = 7874; parameter Real T(unit = "K") = 298.15; // constant temperature system for now parameter Real Vhs1(unit = "L") = 0.225; // head space volumes parameter Real Vhs2(unit = "L") = 0.225; parameter Real Keqfe2 = 5e-09; parameter Real Keqfe3 = 6.7e-03; parameter Real tmem(unit = "m") = 0.000225; // thickness of membrane (m). 175 um plus 25 um per side of //PVA = 225 um parameter Real chargetime(unit = "s") = 3600; parameter Real dischargetime(unit = "s") = chargetime; parameter Real flowonlytime(unit="s") = 1000; parameter Real BI = 20; // buffering index inverse parameter Real rp = 1; // recombination reactor performance factor, parameter Real df = 1; // DISCHARGE FACTOR (df). multiplying factor- ratio of //discharge to charge current //Real jhydrogen(unit = "A/m^2") = j*(0.5*exp(-1.24*pHneg)); // h2 generation from side reaction during charging. //Real jhydrogen(unit = "A/m^2") = j * 0.025 / pHneg ^ 3; //Real jhydrogen = 0; // jhydrogen is assumed to be the CHARGING current only

126 APPENDIX A. SUPPORTING INFORMATION

Real jhydrogen(unit = "A/m^2") = 116*exp(-1.15*pHnegv0); // based on FEL38/39 pressure measurements

Real gap1(start = go, unit = "m"); // electrode gap on negative side (side 1) in meters parameter Real gap2(unit = "m") = 0.001; // electrode gap on positive side. Real rfe2(unit = "mol/s"); Real rh2gen(unit = "mol/s"); // probably should not use the variable "r", but rather n-dot for these hydrogen terms Real rh2cons(unit = "mol/s"); // rate of hydrogen consumption by rebalancing reactor Real Ih2cons(unit = "A"); Real Ih2gen(unit = "A"); Real h2pressure(start = 0, unit = "Atm"); // hydrogen partial pressure in head space Real cycletimevar; // using multiple cycletime variables to make parametric plots with multiple variables more easily Real cyclestate(start=0); // state 0 is charge, state 1 is charge, state 2 is discharge, state 3 is rest Real jcorr(unit = "A/m^2"); Real rcorr(start = 0, unit = "mol/s"); // mol/(m^3*s) of iron lost to comproportionation (NOT acid corrosion) Real jacidcorr(unit = "A/m^2"); Real racidcorr(start = 0, unit = "mol/s"); Real rh2gencorr(start=0, unit="mol/s"); // Real time4 = 14310; Real jpm; // jpm is j plus or minus, meaning it changes signs so that it can be used for migration transport. the original j term is an absolute value. Real namig(unit = "mol/s"); Real namigv3(unit= "mol/s"); Real namigv4(unit = "mol/s"); Real clmig(unit = "mol/s", start = -1.0); Real clmigv2; Real clmigv4(unit = "mol/s"); Real fe2mig(unit = "mol/s"); Real fe2migv4(unit = "mol/s"); Real fe3mig(unit = "mol/s"); Real fe3migv4(unit = "mol/s"); Real protonmig(unit = "mol/s"); Real protonmigv1(unit = "mol/s"); // v1 is the original version based // v2 is for the upcoming migration equations that include the diffusion potential Real protonmigv4(unit="mol/s"); // non-averaged migration

127 APPENDIX A. SUPPORTING INFORMATION

version designed to prevent negative concentrations Real jfe3r(unit="A/m^2"); Real jfe3rv2(unit="A/m^2"); Real rfe3r(unit = "mol/s"); Real fe2diff(unit = "mol/s"); //fe2 diffusion out of negative cell Real fe3diff(unit = "mol/s"); Real protondiff(unit = "mol/s"); Real nadiff(unit = "mol/s"); Real cldiff(unit = "mol/s"); parameter Real tau = 1.3; Real Edrop1(unit = "Volts") = abs(jpm * (1 / NE.conductivity) * gap1); // ohmic voltage loss in negative electrode gap. see Goodridge and Scott p. 69. PER CELL // ohmic voltage loss in membrane. see Arora 2004 Real Rmem(unit = "Ohms") = 1 / kappa * (tau ^ 2 * tmem / NE.epsilon / ACell); // from Arora and Zhang (2004) Real Rmemv2(unit="Ohms") = 1/kappa/(NE.epsilon^(3/2)) *tmem/ACell; //Bruggeman-based resistance Real Edropmem(unit = "Volts") = abs(jpm) * Rmem * ACell; Real Edrop2(unit = "Volts") = abs(jpm * (1 / PE.conductivity) * gap2); // ohmic voltage loss in positive electrode gap. see Goodridge and Scott p. 69 Real Edrop(unit = "Volts") = Edrop1 + Edrop2 + Edropmem; //Real cldiffPC; Real negcharge(unit = "electrons") = VTankNeg * (2 * NE.Cfe2 + 3 * NE.Cfe3 + 1 * NE.Cna + 1 * NE.Cprotons - 1 * NE.Ccl); Real poscharge(unit = "electrons") = VTankPos * (2 * PE.Cfe2 + 3 * PE.Cfe3 + 1 * PE.Cna + 1 * PE.Cprotons - 1 * PE.Ccl); Real negchargev2(unit = "electrons") = VTankNeg * (2 * NE.Cfe2 + 3 * NE.Cfe3 + 1 * NE.Cna + 1 * Cprotonsneg - 1 * NE.Ccl); Real poschargev2(unit = "electrons") = VTankPos * (2 * PE.Cfe2 + 3 * PE.Cfe3 + 1 * PE.Cna + 1 * Cprotonspos - 1 * PE.Ccl); // Real kappa = (NE.conductivity + PE.conductivity) / 2; Real kappa = 15; // enforce constant bulk electrolyte conductivity value of 150 mS/cm for gradphiv1 Real gradphiv1(unit = "volts/m") = -jpm/(kappa*0.4); // if sep conductivity is 40 % of bulk, based on various lab measurements Real gradphiv2(unit = "volts/m") = -jpm *ACell* Rmemv2/tmem; Real cyclenum(start = 1); Real pHpos(start=0.25); // Real pHnegv0; // using the simple logarithm definitions Real pHposv0;

128 APPENDIX A. SUPPORTING INFORMATION

Real Cprotonsneg; Real Cprotonspos; Real totironneg; //total moles of iron in negative side of the system Real totironpos; //total moles of iron in positive side of the system Real totiron; Real totcl; Real totprot; Real totprotv2; // totprot based on pH back-calculation version // total iron variable to check mass balaNE on iron species (calculated value should remain constant) // time1 is charge duration. time2 is (charge time + discharge time). time 3 is cycle time(charge time + discharge time + rest time). electrolyte NE(name = "negativetankelyte", Cofe2 = 2500, Cofe0 = 0, Cofe3 = 000.0, Cona = 2000, Cocl = 7040, Coprotons =10); electrolyte PE(name = "positivetankelyte", Cofe2 = 1500, Cofe0 = 0, Cofe3 = 500.0, Cona = 2000, Cocl = 6750, Coprotons = 250); equation totironneg = (NE.Cfe0 + NE.Cfe2 + NE.Cfe3) * VTankNeg; totironpos = (PE.Cfe0 + PE.Cfe2 + PE.Cfe3) * VTankPos; totiron = totironneg + totironpos; totcl = NE.Ccl*VTankNeg + PE.Ccl*VTankPos; totprot = NE.Cprotons*VTankNeg + PE.Cprotons*VTankPos + 2*h2pressure*(Vhs1+Vhs2)/0.08206/T; totprotv2 = Cprotonsneg*VTankNeg + Cprotonspos*VTankPos + 2*h2pressure*(Vhs1+Vhs2)/0.08206/T; //pHnegv0 = -log10((NE.Cprotons) / 1000); //pHposv0 = -log10((PE.Cprotons) / 1000); pHnegv0 = -log10((NE.Cprotons+NE.Cfe3) / 1000); pHposv0 = -log10((PE.Cprotons+PE.Cfe3) / 1000); //pHneg = 14+log10((1e-14)/(NE.Cprotons/1000)); //pHpos = 14+log10((1e-14)/(PE.Cprotons/1000)); //jpm = if cycletimevar < time1 then j elseif cycletimevar >= time1 and cycletimevar < time2 then -1 * df * j else 0; jpm = if cyclestate<0.99 then j elseif cyclestate>0.99 and cyclestate <1.01 then -1 * df * j else 0; // factor of 2 from taking the average transferrencE number in each cell electrolyte. migration defined as positive flux into cell // migration terms // namig = -NE.z4 * NE.u4n * F * ((NE.Cna + PE.Cna) / 2) * gradphiv2 * ACell; namigv3 = -NE.z4*NE.Dna * F/(R*T)*((NE.Cna + PE.Cna) / 2) *gradphiv2*ACell; // based on encyclopedia of electrochem eqn for mobility, for checking; // gives same results; namigv4=if cyclestate<0.99 then -NE.z4 * NE.u4n * F *

129 APPENDIX A. SUPPORTING INFORMATION

PE.Cna * gradphiv2 * ACell elseif cyclestate>0.99 and cyclestate <1.01 then -NE.z4 * NE.u4n * F * NE.Cna * gradphiv2 * ACell else 0; fe2mig = -NE.z1 * NE.u1n * F * ((NE.Cfe2 + PE.Cfe2) / 2) * gradphiv2 * ACell; fe2migv4 = if cyclestate<0.99 then -NE.z1 * NE.u1n * F * PE.Cfe2 * gradphiv2 * ACell elseif cyclestate>0.99 and cyclestate <1.01 then -NE.z1 * NE.u1n * F * NE.Cfe2 * gradphiv2 * ACell else 0; fe3mig = -NE.z3 * NE.u3n * F * ((NE.Cfe3 + PE.Cfe3) / 2) * gradphiv2 * ACell; fe3migv4 = if cyclestate<0.99 then -NE.z3 * NE.u3n * F * PE.Cfe3 * gradphiv2 * ACell elseif cyclestate>0.99 and cyclestate <1.01 then -NE.z3 * NE.u3n * F * NE.Cfe3 * gradphiv2 * ACell else 0; protonmigv1 = -NE.z6 * NE.u6n * F * ((NE.Cprotons + PE.Cprotons) / 2) * gradphiv2 * ACell; //original version based on infinite-dilution protonmig = -NE.z6 * NE.u6n * F * ((Cprotonsneg + Cprotonspos) / 2) * gradphiv2 * ACell; // new version based on pH back-calculation protonmigv4 = if cyclestate<0.99 then -NE.z6 * NE.u6n * F * PE.Cprotons * gradphiv2 * ACell elseif cyclestate>0.99 and cyclestate <1.01 then -NE.z6 * NE.u6n * F * NE.Cprotons * gradphiv2 * ACell else 0; clmig = (2 * der(NE.Cfe2) + 3 * der(NE.Cfe3) + 1 * der(NE.Cna) + 1 * der(NE.Cprotons)) * ncells * VTankNeg + cldiff; //clmig = (2 * der(NE.Cfe2) + 3 * der(NE.Cfe3) + 1 * der(NE.Cna) + 1 * der(Cprotonsneg)) * ncells * VTankNeg + cldiff; clmigv2 = -NE.z5 * NE.u5n * F * ((NE.Ccl + PE.Ccl) / 2) * gradphiv2 * ACell; clmigv4 = if cyclestate<0.99 then -NE.z5 * NE.u5n * F * NE.Ccl * gradphiv2 * ACell elseif cyclestate>0.99 and cyclestate <1.01 then -NE.z5 * NE.u5n * F * PE.Ccl * gradphiv2 * ACell else 0; // //diffsion terms // fe2diff = NE.Dfe2 * ACell * (NE.Cfe2 - PE.Cfe2) / tmem; fe3diff = NE.Dfe3 * ACell * (PE.Cfe3 - NE.Cfe3) / tmem; protondiff = NE.Dh2 * ACell * (PE.Cprotons - NE.Cprotons) / tmem; //protondiff = NE.Dh2 * ACell * (Cprotonspos - Cprotonsneg) / tmem; cldiff = NE.Dcl * ACell * (NE.Ccl - PE.Ccl) / tmem; nadiff = NE.Dna * ACell * (NE.Cna - PE.Cna) / tmem; // // // factor of 1000 to convert from mol/(m^3) to mol/L //corrosion reaction rate is zero during charge //rfe2 is rate of the metal deposition reaction fe+2 + 2e- --> fe+0 //rfe2 = if cycletimevar < time1 then (j - jhydrogen - rfe3r) * ACell / (nab * F) elseif cycletimevar >= time1 and cycletimevar < time2 then -1 * df * j * ACell / (nab * F)

130 APPENDIX A. SUPPORTING INFORMATION

else 0; rfe2 = if cyclestate < 0.99 then (j - jhydrogen - jfe3r) * ACell / (nab * F) elseif cyclestate>0.99 and cyclestate <1.01 then -1 * df * j * ACell / (nab * F) else 0; // j*0.01 is the initial guess for corrosion term. for the first cycle, cannot have it be f(NE.Cfe3) or the model appears to be too nonlinear to integrate smoothly rh2gen = if cyclestate < 0.99 then jhydrogen * ACell / 96485 / 2 else 0; rh2gencorr = if cyclestate < 0.99 then 0 else racidcorr; // rate of hydrogen generation due to acid corrosion during discharge jfe3r= if cyclestate < 0.99 then j * NE.Cfe3 / (NE.Cfe3 + NE.Cfe2) else 0; // current (A)-form of fe3 reduction current during charging jfe3rv2= if cyclestate < 0.99 then (1*96485*NE.Dfe3soln*NE.Cfe3)/0.00001 else 0; // limiting-current version rfe3r = if cyclestate < 0.99 then jfe3rv2 * ACell / 96485 / 1 else 0; // mol/s form of the equation //rfe3r = 0; // temporarily set to zero //moles per second of hydrogen generation total. based on real data- see mathematica notebook jcorr = if noEvent(cyclestate > 0.99 and NE.Cfe0 > 5) then NE.Cfe3 / (NE.Cfe3 + NE.Cfe2) else 0; // jcorr is ’corrosion’ by fe3, not protons // noEvent works to prevent chattering, but very slow //jcorr = 0; // temporarily set to 0 rcorr = jcorr*ACell / nab / F; jacidcorr=if noEvent(cyclestate > 0.99 and NE.Cfe0 > 5) then 268*exp(-2.25*pHnegv0) else 0; // simple exponential fit to data from Hruska 1981, R^2~0.97 racidcorr = jacidcorr*ACell / nab / F; rh2cons = rp*(1.62*h2pressure)/2/F; //rh2consv2 based on linear fit to the data shown in J. Power Sources (2016) //rh2cons = 0; //for comparison with no recombination Ih2cons = rh2cons * 2 * F; Ih2gen = (rh2gen+rh2gencorr) * 2 * F; Cprotonsneg = 10^(-1*pHneg)*1000; // back-calculating protons from pH estimates. factor of 1000 to go from mol/L to mol/m3 Cprotonspos = 10^(-1*pHpos)*1000; der(pHneg) = -BI*der(NE.Cprotons/1000); // alternative versions of pH estimation based on semi-empirical phreeqc/buffer index theory der(pHpos) = -BI*der(PE.Cprotons/1000); der(cycletimevar) = 1; // a time variable der(NE.Cfe2) = (3 * rcorr + fe2migv4 + rfe3r - rfe2 - fe2diff + racidcorr) / (ncells * VTankNeg); //simple mass balances in the cell and tank. PER-CELL BASIS

131 APPENDIX A. SUPPORTING INFORMATION

der(NE.Cfe0) = (rfe2 - rcorr - racidcorr) / (ncells * VTankNeg); der(gap1) = -der(NE.Cfe0) * MWfe * VTankNeg / (rhofe * ACell); //der(VCellNeg) = ACell * der(gap1); //der(VCellNeg) = 0; der(NE.Cfe3) = (fe3diff - 2 * rcorr + fe3migv4 - rfe3r) / (ncells * VTankNeg); der(NE.Cprotons) = (protondiff - 2 * rh2gen + protonmigv4 - 2*racidcorr) / (ncells * VTankNeg); der(NE.Cna) = (namigv4 - nadiff) / VTankNeg; der(NE.Ccl) = (clmig - cldiff) / (ncells * VTankNeg); // // begin positive side cell equations // der(PE.Cfe2) = (fe2diff - 2 * rfe2 - 2 * rh2gen - fe2migv4 - rfe3r) / (ncells * VTankPos) + 2 * rh2cons / VTankPos; der(PE.Cfe3) = (2 * rfe2 + 2 * rh2gen - fe3diff - fe3migv4 + rfe3r) / (ncells * VTankPos) - 2 * rh2cons / VTankPos; der(PE.Cprotons) = ((-1 * protondiff) - protonmigv4) / (ncells * VTankPos) + 2 * rh2cons / VTankPos; der(PE.Cna) = (nadiff - namigv4) / (ncells * VTankPos); der(PE.Ccl) = (cldiff - clmig) / (ncells * VTankPos); der(PE.Cfe0) = 0; // der(h2pressure) = (rh2gen + rh2gencorr - rh2cons) * 0.08206 * T / (Vhs1 + Vhs2); // start in a charging mode. der(cyclestate) = 0; // cycletimevar gets reinitialized to arbitrary time values (e.g., 50000 )that must be higher than the actual possible amount of time when {cycletimevar > chargetime} then // at the end of the charge, switch to discharge mode //reverse curent direction, and remove term for side reaction current reinit(cycletimevar,50000); reinit(cyclestate,1); end when; // carry out discharging step, then switch back to charging when {cycletimevar > (50000+dischargetime) or NE.Cfe0<=0} then reinit(cyclestate, 0); reinit(cycletimevar,0); cyclenum=pre(cyclenum)+1; end when; annotation(experiment(StartTime = 0, StopTime =76000, Tolerance = 4e-5, Interval = 10)); end rfb2;

132 APPENDIX A. SUPPORTING INFORMATION

Python script for iterative simulations

The following Python script was used for running the simulation iteratively such that parameters (e.g., separator thickness) could be varied with each simulation run. For example, one iterative test modiﬁed the parameters in the following way:

### iteratively prepares and runs modelica simulation with different parameter values import subprocess reportfile = open(’simreport.dat’,’w’) reportfile.write(’simID’ + ’\t’ ’rp’ + ’\t’+ ’porosity’ + ’\t’ + ’thickness (m)’ + ’\n’) for l in [1,2,3,4]: for i in range (3): for k in range (3): simnum = l+1 simfile = open("simscriptv2.mos","w") simfile.write("cd(\"/tmp/omedit/\"); " + "\n") simfile.write("loadModel(Modelica);"+ "\n") simfile.write("loadFile(\"/tmp/omedit/electrolyte.mo\"); "+ "\n") simfile.write("loadFile(\"/tmp/omedit/rfb2.mo\"); "+ "\n") simfile.write("setParameterValue(electrolyte,epsilon, "+str((i*0.2)+0.3)+");" + "\n") # Change the membrane porosity #simfile.write("setParameterValue(electrolyte,epsilon," +str(0.3)+");" + "\n") simfile.write("setParameterValue(electrolyte,dcf," +str(1.0)+");" + "\n") simfile.write("setParameterValue(rfb2,BI,"+str(20)+");" + "\n") # BI INVERSE (beta ~ 0.05, betainv ~ 20) based on fel39 pressure data (see rh2gen spreadsheet) simfile.write("setParameterValue(rfb2,rp," +str(1/float(l))+");" + "\n") # Change the recombination performance/area simfile.write("setParameterValue(rfb2,tmem," +str(k*0.000075+0.0001)+");" + "\n") # change the membrane thickness from 175 to 225 um simfile.write("instantiateModel(rfb2); "+ "\n") simfile.write("simulate(rfb2, stopTime=864000, numberOfIntervals=864000, outputFile=\"/tmp/ omedit/rfb2_res.mat\"); ") simfile.close() # run the separate ’run_simulation’ script in the same directory subprocess.call(’./run_simulation’, shell=True) # copy and rename the DyMat output file appropriately based on the simulation settings

133 APPENDIX A. SUPPORTING INFORMATION

subprocess.call(’cp /tmp/omedit/rfb2_res.mat.gpd.02 /tmp/omedit/o’+str(l)+str(i)+str(k)+’.dat’, shell=True) # record info to simulation report file reportfile.write(str(l)+str(i)+str(k) + ’\t’ + str(1/float(l))+ ’\t’ + str((i*0.2)+0.3) + ’\t’ + str(k*0.000075+0.0001) + ’\n’) reportfile.close() subprocess.call(’cp ~/Documents/modelica_rfb/simreport.dat /tmp/omedit/’, shell=True) subprocess.call(’echo running pandas’, shell=True) subprocess.os.chdir(’/tmp/omedit’) subprocess.call(’python pandas_rfb.py’, shell=True)

134 APPENDIX A. SUPPORTING INFORMATION

Table A.11: Example of an iterative simulation report ﬁle documenting the parameters used in each simulation run.

simID α ϵ tsep(m) 100 1 0.3 0.0001 101 1 0.3 0.000175 102 1 0.3 0.00025 110 1 0.5 0.0001 111 1 0.5 0.000175 112 1 0.5 0.00025 120 1 0.7 0.0001 121 1 0.7 0.000175 122 1 0.7 0.00025 200 0.5 0.3 0.0001 201 0.5 0.3 0.000175 202 0.5 0.3 0.00025 210 0.5 0.5 0.0001 211 0.5 0.5 0.000175 212 0.5 0.5 0.00025 220 0.5 0.7 0.0001 221 0.5 0.7 0.000175 222 0.5 0.7 0.00025 300 0.333 0.3 0.0001 301 0.333 0.3 0.000175 302 0.333 0.3 0.00025 310 0.333 0.5 0.0001 311 0.333 0.5 0.000175 312 0.333 0.5 0.00025 320 0.333 0.7 0.0001 321 0.333 0.7 0.000175 322 0.333 0.7 0.00025 400 0.25 0.3 0.0001 401 0.25 0.3 0.000175 402 0.25 0.3 0.00025 410 0.25 0.5 0.0001 411 0.25 0.5 0.000175 412 0.25 0.5 0.00025 420 0.25 0.7 0.0001 421 0.25 0.7 0.000175 422 0.25 0.7 0.00025

135 APPENDIX A. SUPPORTING INFORMATION

Bash script for simulating in temporary directory # Make temporary OMEdit directory mkdir /tmp/omedit echo "copying files to /tmp" # Copy files to the temporary OMEdit directory cp electrolyte.mo /tmp/omedit/ cp rfb2.mo /tmp/omedit/ cp gpscript_multisim_v2 /tmp/omedit/ cp /home/steve/Documents/stevenotes/lifetime_tests/ fel23/fel23_pressure.txt /tmp/omedit/ cp pandas_rfb.py /tmp/omedit/ # Run the modelica mos script echo "running simscript.mos..." omc simscriptv2.mos # Run DyMatExport on the .mat file to export variables of interest # note: make sure the python DyMat package has been installed. last tested with DyMat 0.7, installed by direct download of .tar.gz echo "running DyMatExport to convert .mat file..." DyMatExport.py -e "PE.Cfe2, PE.Cfe3, h2pressure, pHnegv0, pHposv0, NE.Cprotons, PE.Cprotons, protonmigv4, protondiff, Ih2gen, Ih2cons, negcharge, poscharge, der(NE.Cprotons), der(PE.Cprotons), cyclenum" -f Gnuplot /tmp/omedit/rfb2_res.mat # Change into the OMEdit directory cd /tmp/omedit/ #python pandas_rfb.py # Change back to the main directory so simulation can be restarted echo "all done!" cd ~/Documents/modelica_rfb/

136 APPENDIX A. SUPPORTING INFORMATION

Example of program output

Table A.12: Example program output for ϵ = 0.5, α = 1, tsep = 225 µm

∇ −1 time (s) cycle # Φsep (V m ) pH2 (atm) IH2,gen (A) IH2,cons (A) 0 1 -188.562 0 0.0348901 0 1 1 -188.562 9.90E-06 0.0354 1.60E-05 2 1 -188.562 1.99E-05 0.0359022 3.23E-05 3 1 -188.562 3.01E-05 0.0363969 4.88E-05 4 1 -188.562 4.04E-05 0.0368844 6.55E-05 5 1 -188.562 5.08E-05 0.037365 8.24E-05 6 1 -188.562 6.14E-05 0.037839 9.95E-05 7 1 -188.562 7.21E-05 0.0383067 0.000116792 8 1 -188.562 8.29E-05 0.0387682 0.000134301 9 1 -188.562 9.38E-05 0.0392237 0.000152012 10 1 -188.562 0.000104892 0.0396736 0.000169926 11 1 -188.562 0.000116075 0.0401179 0.000188042 12 1 -188.562 0.000127364 0.0405569 0.00020633 13 1 -188.562 0.000138772 0.0409908 0.00022481 14 1 -188.562 0.000150299 0.0414196 0.000243484 15 1 -188.562 0.000161944 0.0418435 0.00026235 16 1 -188.562 0.000173709 0.0422628 0.000281409

137 APPENDIX A. SUPPORTING INFORMATION

Table A.13: Example program output, continued.

∑ ∑ pHneg pHpos zici (neg) zici (pos) 2 0.124939 0 0 2 0.124939 2.27E-16 -2.27E-16 2 0.124939 2.27E-16 -2.27E-16 1.98738 0.124953 1.14E-16 0 1.97513 0.124967 1.14E-16 2.27E-16 1.96323 0.124981 1.14E-16 4.55E-16 1.95166 0.124995 0 0 1.94041 0.125009 2.27E-16 0 1.92944 0.125023 2.27E-16 0 1.91876 0.125036 1.14E-16 0 1.90835 0.12505 2.27E-16 2.27E-16 1.89819 0.125064 0 0 1.88828 0.125077 1.14E-16 0 1.87859 0.12509 1.14E-16 2.27E-16 1.86913 0.125104 3.41E-16 4.55E-16 1.85987 0.125117 2.27E-16 9.09E-16 1.85083 0.12513 2.27E-16 1.36E-15 1.84197 0.125143 2.27E-16 1.82E-15 1.8333 0.125156 2.27E-16 2.50E-15

138 APPENDIX A. SUPPORTING INFORMATION

Table A.14: Example program output, continued. F represents molar ﬂow rates (mol s−1).

FFe2+,diff FFe2+,mig FFe3+,diff FFe3+,mig FNa+,diff 3.39E-06 1.68E-05 1.44E-06 7.13E-06 0 3.39E-06 1.68E-05 1.44E-06 7.13E-06 5.10E-17 3.39E-06 1.68E-05 1.44E-06 7.13E-06 5.10E-17 3.39E-06 1.68E-05 1.44E-06 7.13E-06 1.57E-09 3.40E-06 1.68E-05 1.44E-06 7.13E-06 3.14E-09 3.40E-06 1.68E-05 1.44E-06 7.13E-06 4.70E-09 3.40E-06 1.68E-05 1.44E-06 7.13E-06 6.27E-09 3.40E-06 1.68E-05 1.44E-06 7.13E-06 7.84E-09 3.40E-06 1.68E-05 1.44E-06 7.13E-06 9.40E-09 3.40E-06 1.68E-05 1.44E-06 7.14E-06 1.10E-08 3.40E-06 1.68E-05 1.44E-06 7.14E-06 1.25E-08 3.40E-06 1.68E-05 1.44E-06 7.14E-06 1.41E-08 3.40E-06 1.68E-05 1.44E-06 7.14E-06 1.57E-08 3.40E-06 1.68E-05 1.44E-06 7.14E-06 1.72E-08 3.40E-06 1.68E-05 1.44E-06 7.14E-06 1.88E-08 3.40E-06 1.68E-05 1.44E-06 7.14E-06 2.04E-08 3.40E-06 1.68E-05 1.44E-06 7.14E-06 2.19E-08 3.40E-06 1.68E-05 1.44E-06 7.15E-06 2.35E-08 3.40E-06 1.68E-05 1.44E-06 7.15E-06 2.51E-08

139 APPENDIX A. SUPPORTING INFORMATION

Table A.15: Example output, continued. F represents molar ﬂow rates (mol s−1).

FNa+,mig FH+,diﬀ FH+,mig FCl−,cb 2.08E-05 1.05E-05 1.81E-05 7.34E-05 2.08E-05 1.05E-05 1.81E-05 7.34E-05 2.08E-05 1.05E-05 1.81E-05 7.34E-05 2.08E-05 1.05E-05 1.81E-05 7.34E-05 2.08E-05 1.05E-05 1.81E-05 7.33E-05 2.08E-05 1.05E-05 1.81E-05 7.33E-05 2.08E-05 1.05E-05 1.81E-05 7.33E-05 2.08E-05 1.05E-05 1.81E-05 7.32E-05 2.08E-05 1.04E-05 1.81E-05 7.32E-05 2.08E-05 1.04E-05 1.81E-05 7.32E-05 2.08E-05 1.04E-05 1.81E-05 7.32E-05 2.08E-05 1.04E-05 1.81E-05 7.31E-05 2.08E-05 1.04E-05 1.80E-05 7.31E-05 2.08E-05 1.04E-05 1.80E-05 7.31E-05 2.08E-05 1.04E-05 1.80E-05 7.30E-05 2.08E-05 1.03E-05 1.80E-05 7.30E-05 2.08E-05 1.03E-05 1.80E-05 7.30E-05 2.08E-05 1.03E-05 1.80E-05 7.29E-05 2.08E-05 1.03E-05 1.80E-05 7.29E-05

140 APPENDIX A. SUPPORTING INFORMATION

Pandas script for getting cycle-averaged output import pandas as pd import subprocess # for getting average concentrations of feII/feIII in positive tank comparing with and without recombination # can also be used to get other info though filelist=[’o110.dat’, ’o111.dat’, ’o411.dat’, ’o112.dat’, ’o212.dat’, ’o101.dat’, ’o121.dat’] # list of files to analyse using pandas groupby for file in filelist: subprocess.call(’tail -n +19 ’+file+ ’ > filenoheader.dat’ , shell=True) # removes first lines so that pandas can have easy access, but this is only data and headers need to be added subsequently cleanfile = open(’cleanfile.txt’, ’w’) # creates empty textfile with 18 general headers cleanfile.write(’col1’ + ’\t’ + ’col2’ + ’\t’ + ’col3’ + ’\t’ + ’col4’ + ’\t’ + ’col5’ + ’\t’ + ’col6’ + ’\t’ + ’col7’ + ’\t’ + ’col8’ + ’\t’ + ’col9’ + ’\t’ + ’col10’ + ’\t’ + ’col11’ + ’\t ’ + ’col12’ + ’\t’ + ’col13’ + ’\t’ + ’col4’ + ’\t’ + ’col15’ + ’\t’ + ’col16’ + ’\t’ + ’col17’ + ’\t’ + ’col18’ + ’\n’) cleanfile.close() subprocess.call(’cat filenoheader.dat >> cleanfile.txt’, shell=True) # appends data file to header file for a clean set # that pandas can access # requires removing the initial rows pd.set_option(’display.max_rows’, 200) df = pd.DataFrame.from_csv(’cleanfile.txt’, sep="\t") # creates dataframe for the cleaned file # optional commands to get dataframe statistics #print df #print df.head() #print df.describe() newdf= df.groupby(’col17’).mean() # calculates average values of fe(II), fe(III) for each cycle # col17 is the cycle number, col2 is fe(II), col3 is fe(III) newdf.to_csv(file[:-4]+’_’+’groupby.dat’, sep=’\t’)

141 References

[1] G. L. Soloveichik, “Flow Batteries: Current Status and Trends,” Chemical Re- views, vol. 115, no. 20, pp. 11533–11558, 2015.

[2] AWEA, “U. S. Wind Industry Fourth Quarter 2015 Market Report,” The Amer- ican Wind Energy Association Data Services, 2015.

[3] P. Denholm, M. O’Connell, G. Brinkman, and J. Jorgenson, “Overgeneration from Solar Energy in California: A Field Guide to the Duck Chart,” Technical Report, 2015.

[4] CAISO, “What the duck curve tells us about managing a green grid,” CAISO Public Reports, 2016.

[5] Purdue University Discovery Park - State Utility Forecasting Group, “Utility Scale Energy Storage Systems: Beneﬁts, Applications, and Technologies,” Tech. Rep. June, State Utility Forecasting Group, 2013.

[6] D. O. Akinyele and R. K. Rayudu, “Review of energy storage technologies for sustainable power networks,” Sustainable Energy Technologies and Assess- ments, vol. 8, pp. 74–91, 2014.

[7] T. Kousksou, P. Bruel, A. Jamil, T. El Rhaﬁki, and Y. Zeraouli, “Energy storage: Applications and challenges,” Solar Energy Materials and Solar Cells, vol. 120, no. PART A, pp. 59–80, 2014.

142 REFERENCES

[8] X. Luo, J. Wang, M. Dooner, and J. Clarke, “Overview of current development in electrical energy storage technologies and the application potential in power system operation,” Applied Energy, vol. 137, pp. 511–536, 2015.

[9] T. M. I. Mahlia, T. J. Saktisahdan, A. Jannifar, M. H. Hasan, and H. S. C. Matseelar, “A review of available methods and development on energy storage; Technology update,” Renewable and Sustainable Energy Reviews, vol. 33, pp. 532–545, 2014.

[10] A. Poullikkas, “A comparative overview of large-scale battery systems for electricity storage,” Renewable and Sustainable Energy Reviews, vol. 27, pp. 778– 788, 2013.

[11] A. A. Akhil, G. Huﬀ, A. B. Currier, B. C. Kaun, D. M. Rastler, S. B. Chen, A. L. Cotter, D. T. Bradshaw, and W. D. Gauntlett, “DOE/EPRI Electricity Storage Handbook,” Tech. Rep. February, 2015.

[12] M. Beaudin, H. Zareipour, A. Schellenberglabe, and W. Rosehart, “Energy storage for mitigating the variability of renewable electricity sources: An updated review,” Energy for Sustainable Development, vol. 14, no. 4, pp. 302–314, 2010.

[13] T. V. Nguyen and R. F. Savinell, “Flow Batteries,” Electrochemical Society Interface, vol. 19, no. 3, pp. 54–56, 2010.

[14] K. R. Bullock, “Lead Acid Battery Systems and Technology for Sustainable Energy,” in Batteries for Sustainability, ch. 5, Springer, 2013.

[15] Z. Zhang and P. Ramadass, “Lithium-Ion Battery Systems and Technology,” in Batteries for Sustainability 2, ch. 10, Springer, 2013.

143 REFERENCES

[16] R. E. Ciez and J. F. Whitacre, “Comparison between cylindrical and prismatic lithium-ion cell costs using a process based cost model,” Journal of Power Sources, vol. 340, pp. 273–281, 2017.

[17] K. Gong, X. Ma, K. M. Conforti, K. J. Kuttler, J. B. Grunewald, K. L. Yeager, M. Z. Bazant, S. Gu, and Y. Yan, “A zinciron redox-ﬂow battery under $100 per kW h of system capital cost,” Energy Environ. Science, vol. 8, no. 10, pp. 2941–2945, 2015.

[18] A. Sonoc, J. Jeswiet, and V. K. Soo, “Opportunities to Improve Recycling of Automotive Lithium Ion Batteries,” Procedia CIRP, vol. 29, pp. 752–757, 2015.

[19] A. Z. Weber, M. M. Mench, J. P. Meyers, P. N. Ross, J. T. Gostick, and Q. Liu, “Redox ﬂow batteries: A review,” Journal of Applied Electrochemistry, vol. 41, pp. 1137–1164, sep 2011.

[20] M. Skyllas-Kazacos, M. H. Chakrabarti, S. a. Hajimolana, F. S. Mjalli, and M. Saleem, “Progress in Flow Battery Research and Development,” Journal of The Electrochemical Society, vol. 158, no. 8, p. R55, 2011.

[21] P. Leung, X. Li, C. P. de Le´on,L. Berlouis, C. T. J. Low, and F. C. Walsh, “Progress in redox ﬂow batteries, remaining challenges and their applications in energy storage,” RSC Advances, vol. 2, no. 27, pp. 10125–10156, 2012.

[22] W. Wang, Q. Luo, B. Li, X. Wei, L. Li, and Z. Yang, “Recent progress in redox ﬂow battery research and development,” Advanced Functional Materials, vol. 23, no. 8, pp. 970–986, 2013.

[23] B. R. Chalamala, T. Soundappan, G. R. Fisher, M. R. Anstey, V. V. Viswanathan, and M. L. Perry, “Redox ﬂow batteries: An engineering per- spective,” Proceedings of the IEEE, vol. 102, no. 6, pp. 1–24, 2014.

144 REFERENCES

[24] P. Alotto, M. Guarnieri, and F. Moro, “Redox ﬂow batteries for the storage of renewable energy: A review,” Renewable and Sustainable Energy Reviews, vol. 29, pp. 325–335, 2014.

[25] M. L. Perry and A. Z. Weber, “Advanced Redox-Flow Batteries: A Perspec- tive,” Journal of The Electrochemical Society, vol. 163, no. 1, pp. A5064–A5067, 2016.

[26] SEI, “Redox Flow Battery,” http://global-sei.com/products/redox/, 2017.

[27] A. Parasuraman, T. M. Lim, C. Menictas, and M. Skyllas-Kazacos, “Review of material research and development for vanadium redox ﬂow battery applications,” Electrochimica Acta, vol. 101, pp. 27–40, jul 2013.

[28] J. Winsberg, T. Hagemann, T. Janoschka, M. D. Hager, and U. S. Schu- bert, “Redox-Flow Batteries: From Metals to Organic Redox-Active Materials,” Angewandte Chemie - International Edition, pp. 686–711, 2017.

[29] L. H. Thaller, “Redox Flow Cell Energy Storage Systems,” Nasa Tm-79143, p. 14, 1979.

[30] R. F. Gahn, N. H. Hagedorn, and J. A. Johnson, “Cycling Performace of the Iron- Chromium Redox Energy Storage System Conservation and Renewable Energy,” tech. rep., NASA, 1985.

[31] H. S. Lim, A. M. Lackner, and R. C. Knechtli, “Zinc-Bromine Secondary Battery,” Journal of The Electrochemical Society, vol. 124, no. August 1977, pp. 1154–1157, 1978.

[32] L. W. Hruska and R. F. Savinell, “Investigation of Factors Aﬀecting Perfor- mance of the Iron Redox Battery,” Journal of The Electrochemical Society, vol. 128, no. 1, pp. 18–25, 1981.

145 REFERENCES

[33] P. S. Fedkiw and R. W. Watts, “A Mathematical Model for the Iron / Chromium Redox Battery,” Journal of The Electrochemical Society, vol. 131, no. 4, p. 702, 1984.

[34] D. A. Johnson and M. A. Reid, “Chemical and Electrochemical Behavior of the Cr(lll)/Cr(ll) Half Cell in the Iron-Chromium Redox Energy Storage System,” J. eletrochem. Soc., vol. 132, no. 5, pp. 1058–1062, 1985.

[35] D. J. Eustace, “Bromine Complexation in ZincBromine Circulating Batteries,” Journal of The Electrochemical Society, vol. 127, no. 3, pp. 528–532, 1980.

[36] J. McBreen, “Rechargeable zinc batteries,” Journal of Electroanalytical Chem- istry, vol. 168, no. 1-2, pp. 415–432, 1984.

[37] T. I. Evans and R. White, “A Mathematical Model of a Zinc/Bromine Flow Cell,” Journal of The Electrochemical Society, vol. 134, no. 4, p. 866, 1987.

[38] T. I. Evans and R. White, “A Review of Mathematical Modeling of the Zinc/Bromine Flow Cell and Battery,” Journal of The Electrochemical Soci- ety, vol. 134, no. 11, p. 2725, 1987.

[39] G. D. Simpson, “An Algebraic Model for a Zinc/Bromine Flow Cell,” Journal of The Electrochemical Society, vol. 136, no. 8, pp. 2137–2144, 1989.

[40] G. D. Simpson, “A Simple Model for a Zinc/Bromine Flow Cell and Associ- ated Storage Tanks,” Journal of The Electrochemical Society, vol. 137, no. 6, pp. 1843–1846, 1990.

[41] E. Sum, M. Rychcik, and M. Skyllas-kazacos, “Investigation of the V(V)/V(IV) system for use in the positive half-cell of a redox battery,” Journal of Power Sources, vol. 16, no. 2, pp. 85–95, 1985.

146 REFERENCES

[42] M. Skyllas - Kazacos and F. Grossmith, “Eﬃcient Vanadium Redox Flow Cell,” Journal of The Electrochemical Society, vol. 134, no. 12, pp. 2950–2953, 1987.

[43] M. Rychcik and M. Skyllas-Kazacos, “Characteristics of a new all-vanadium redox ﬂow battery,” Journal of Power Sources, vol. 22, no. 1, pp. 59–67, 1988.

[44] M. Skyllas-Kazacos, M. Rychick, and R. Robins, “All-Vanadium Redox Flow Battery,” U. S. Patent 4,786,567, 1988.

[45] G. Huﬀ, A. B. Currier, B. C. Kaun, D. M. Rastler, S. B. Chen, D. T. Brad- shaw, and W. D. Gauntlett, “DOE/EPRI 2013 electricity storage handbook in collaboration with NRECA,” Tech. Rep. July, EPRI, 2013.

[46] S. K. Ghosh, R. Saha, and B. Saha, “Toxicity of inorganic vanadium com- pounds,” Research on Chemical Intermediates, no. January 2014, pp. 1–25, 2014.

[47] M. Imtiaz, M. S. Rizwan, S. Xiong, H. Li, M. Ashraf, S. M. Shahzad, M. Shahzad, M. Rizwan, and S. Tu, “Vanadium, recent advancements and research prospects: A review,” Environment International, vol. 80, pp. 79–88, 2015.

[48] E. Wesoﬀ, “Long-duration energy storage: ﬂow battery industry and technology survey, part 2,” 2015.

[49] M. C. Tucker, A. Phillips, and A. Z. Weber, “All-Iron Redox Flow Battery Tailored for Oﬀ-Grid Portable Applications,” ChemSusChem, vol. 8, no. 23, pp. 3996–4004, 2015.

[50] M. C. Tucker, D. Lambelet, M. Oueslati, B. Williams, W. C. J. Wang, and A. Z. Weber, “Improved low-cost, non-hazardous, all-iron cell for the developing world,” Journal of Power Sources, vol. 332, pp. 111–117, 2016.

147 REFERENCES

[51] K. L. Hawthorne, J. S. Wainright, and R. F. Savinell, “Studies of Iron-Ligand Complexes for an All-Iron Flow Battery Application,” Journal of The Electro- chemical Society, vol. 161, no. 10, pp. A1662–A1671, 2014.

[52] K. L. Hawthorne, Iron-ligand electrokinetics towards an all-iron hybrid redox ﬂow battery. PhD thesis, Case Western Reserve University, 2014.

[53] A. K. Manohar, K. M. Kim, E. Plichta, M. Hendrickson, S. Rawlings, and S. R. Narayanan, “A High Eﬃciency Iron-Chloride Redox Flow Battery for Large- Scale Energy Storage,” Journal of The Electrochemical Society, vol. 163, no. 1, pp. A5118–A5125, 2016.

[54] J. D. Hunt, A. Kavner, E. A. Schauble, D. Snyder, and C. E. Manning, “Poly- merization of aqueous silica in H2O-K2O solutions at 25-200C and 1bar to 20kbar,” Chemical Geology, vol. 283, no. 3-4, pp. 161–170, 2011.

[55] M.-S. Lee, “Use of the bromley equation for the analysis of ionic equilibria in mixed ferric and ferrous chloride solutions at 25 C,” Metallurgical and Materials Transactions B, vol. 37, no. April, pp. 173–179, 2006.

[56] G. Kreysa, “The Iron Chlorides,” in Dechema Corrosion Handbook, 2008.

[57] B. Morgan and O. Lahav, “The eﬀect of pH on the kinetics of spontaneous Fe(II) oxidation by O2 in aqueous solution - basic principles and a simple heuristic description,” Chemosphere, vol. 68, no. 11, pp. 2080–2084, 2007.

[58] J. L. Jambor and J. E. Dutrizac, “Occurrence and Constitution of Natural and Synthetic Ferrihydrite, a Widespread Iron Oxyhydroxide.,” Chemical reviews, vol. 98, no. 7, pp. 2549–2586, 1998.

148 REFERENCES

[59] U. Schwertmann, J. Friedl, and H. Stanjek, “From Fe(III) Ions to Ferrihydrite and then to Hematite.,” Journal of colloid and interface science, vol. 209, no. 1, pp. 215–223, 1999.

[60] J. Majzlan, A. Navrotsky, and U. Schwertmann, “Thermodynamics of iron oxides: Part III. Enthalpies of formation and stability of ferrihydrite (Fe(OH)3), schwertmannite (FeO(OH)3/4(SO4)1/8), and e-Fe2O3,” Geochimica et Cos- mochimica Acta, vol. 68, no. 5, pp. 1049–1059, 2004.

[61] F. M. Michel, L. Ehm, S. M. Antao, P. L. Lee, P. J. Chupas, G. Liu, D. R. Strongin, M. A. A. Schoonen, B. L. Phillips, and J. B. Parise, “The Structure of Ferrihydrite, a Nanocrystaline Material,” Science, vol. 316, no. June, pp. 1726– 1729, 2007.

[62] W. Stumm and F. G. Lee, “Oxygenation of ferrous iron,” Industrial and engineering chemistry, vol. 53, no. 2, pp. 143–146, 1961.

[63] T. J. Grundl and D. L. Macalady, “Electrode measurement of redox potential in anaerobic ferric/ferrous chloride systems,” Journal of Contaminant Hydrology, vol. 5, no. 1, pp. 97–117, 1989.

[64] J. N. Butler, Ionic Equilibrium: Solubility and pH Calculations. Wiley- Interscience, 1998.

[65] J. E. Hawley and R. M. Pytkowicz, “Interpretation of pH measurements in concentrated electrolyte solutions,” Marine Chemistry, vol. 1, no. 3, pp. 245– 250, 1973.

[66] K. G. Knauss, T. J. Wolery, and K. J. Jackson, “A new approach to measuring pH in brines and other concentrated electrolytes,” Geochimica et Cosmochimica Acta, vol. 54, no. 5, pp. 1519–1523, 1990.

149 REFERENCES

[67] F. J. Millero, B. DiTrolio, A. F. Suarez, and G. Lando, “Spectroscopic measurements of the pH in NaCl brines,” Geochimica et Cosmochimica Acta, vol. 73, no. 11, pp. 3109–3114, 2009.

[68] H. Harned and G. Mannweiler, “The Thenyodynamics of Ionized Water in Sodium Chloride Solutions,” Journal of the American Chemical Society, vol. 57, p. 1873, 1935.

[69] M.-S. Lee, “Chemical equilibria in ferrous chloride acid solution,” Metals and Materials International, vol. 10, no. 4, pp. 387–392, 2004.

[70] L. A. Bromley, “Thermodynamic properties of strong electrolytes in aqueous solutions,” AIChE Journal, vol. 19, pp. 313–320, mar 1973.

[71] D. Parkhurst, “Hydrogeochemical Modeling with PHREEQC. User’s guide to PHREEQC (version 2),” no. version 2, 1999.

[72] E. a. Guggenheim and J. C. Turgeon, “Speciﬁc interaction of ions,” Transactions of the Faraday Society, vol. 51, p. 747, 1955.

[73] M. Izaki, “Iron and Iron Alloys,” in Modern Electroplating, pp. i–xviii, 2011.

[74] K. L. Hawthorne, J. S. Wainright, and R. F. Savinell, “Maximizing plating density and eﬃciency for a negative deposition reaction in a ﬂow battery,” Journal of Power Sources, vol. 269, pp. 216–224, dec 2014.

[75] K. L. Hawthorne, T. J. Petek, M. A. Miller, J. S. Wainright, and R. F. Savinell, “An Investigation into Factors Aﬀecting the Iron Plating Reaction for an All- Iron Flow Battery,” Journal of The Electrochemical Society, vol. 162, no. 1, pp. A108–A113, 2015.

150 REFERENCES

[76] J. Bockris and D. Drazic, “The kinetics of deposition and dissolution of iron: Eﬀect of alloying impurities,” Electrochimica Acta, vol. 7, no. 3, pp. 293–313, 1961.

[77] P. S. Fedkiw and R. Watts, “Membranes for Redox Flow Battery Applications,” Tech. Rep. October 1984, NASA, 1984.

[78] A. H. Whitehead and M. Harrer, “Investigation of a method to hinder charge imbalance in the vanadium redox ﬂow battery,” Journal of Power Sources, vol. 230, pp. 271–276, 2013.

[79] J. C. Acevedo and D. K. Stalnaker, “An electrochemical rebalance cell for Redox systems,” tech. rep., NASA, 1983.

[80] A. A. Shah, M. J. Watt-Smith, and F. C. Walsh, “A dynamic performance model for redox-ﬂow batteries involving soluble species,” Electrochimica Acta, vol. 53, no. 27, pp. 8087–8100, 2008.

[81] H. Al-Fetlawi, A. A. Shah, and F. C. Walsh, “Non-isothermal modelling of the all-vanadium redox ﬂow battery,” Electrochimica Acta, vol. 55, no. 1, pp. 78–89, 2009.

[82] A. A. Shah, H. Al-Fetlawi, and F. C. Walsh, “Dynamic modelling of hydrogen evolution eﬀects in the all-vanadium redox ﬂow battery,” Electrochimica Acta, vol. 55, no. 3, pp. 1125–1139, 2010.

[83] A. Shah, R. Tangirala, R. Singh, R. Wills, and F. Walsh, “A Dynamic Unit Cell Model for the All-Vanadium Flow Battery,” Journal of The Electrochemical Society, vol. 158, no. 6, p. A671, 2011.

[84] M. J. Watt-Smith, P. Ridley, R. G. A. Wills, A. A. Shah, and F. C. Walsh, “The importance of key operational variables and electrolyte monitoring to

151 REFERENCES

the performance of an all vanadium redox ﬂow battery,” Journal of Chemical Technology and Biotechnology, vol. 88, no. 1, pp. 126–138, 2013.

[85] K. W. Knehr, E. Agar, C. R. Dennison, A. R. Kalidindi, and E. C. Kumbur, “A Transient Vanadium Flow Battery Model Incorporating Vanadium Crossover and Water Transport through the Membrane,” Journal of The Electrochemical Society, vol. 159, no. 9, pp. 1446–1459, 2012.

[86] K. W. Knehr and E. C. Kumbur, “Role of convection and related eﬀects on species crossover and capacity loss in vanadium redox ﬂow batteries,” Electro- chemistry Communications, vol. 23, no. 1, pp. 76–79, 2012.

[87] E. Agar, K. W. Knehr, D. Chen, M. A. Hickner, and E. C. Kumbur, “Species transport mechanisms governing capacity loss in vanadium ﬂow batteries: Com- paring Naﬁon and sulfonated Radel membranes,” Electrochimica Acta, vol. 98, pp. 66–74, 2013.

[88] P. A. Boettcher, E. Agar, C. R. Dennison, and E. C. Kumbur, “Modeling of Ion Crossover in Vanadium Redox Flow Batteries: A Computationally-Eﬃcient Lumped Parameter Approach for Extended Cycling,” Journal of The Electro- chemical Society, vol. 163, no. 1, pp. A5244–A5252, 2016.

[89] M. Skyllas-Kazacos and L. Goh, “Modeling of vanadium ion diﬀusion across the ion exchange membrane in the vanadium redox battery,” Journal of Membrane Science, vol. 399-400, no. 0, pp. 43–48, 2012.

[90] A. Tang, S. Ting, J. Bao, and M. Skyllas-Kazacos, “Thermal modelling and simulation of the all-vanadium redox ﬂow battery,” Journal of Power Sources, vol. 203, pp. 165–176, 2012.

152 REFERENCES

[91] A. Tang, J. Bao, and M. Skyllas-Kazacos, “Thermal modelling of battery con- ﬁguration and self-discharge reactions in vanadium redox ﬂow battery,” Journal of Power Sources, vol. 216, pp. 489–501, 2012.

[92] A. Tang, J. Bao, and M. Skyllas-Kazacos, “Studies on pressure losses and ﬂow rate optimization in vanadium redox ﬂow battery,” Journal of Power Sources, vol. 248, pp. 154–162, 2014.

[93] R. M. Darling, A. Z. Weber, M. C. Tucker, and M. L. Perry, “The Inﬂuence of Electric Field on Crossover in Redox-Flow Batteries,” Journal of the Electro- chemical Society, vol. 163, no. 1, pp. A5014–A5022, 2015.

[94] A. Brenner, Electrodeposition of Alloys, vol.1 Academic Press. Academic Press, 1963.

[95] R. Winand, “Electrodeposition of metals and alloys-new results and perspec- tives,” Electrochimica Acta, vol. 39, no. 8-9, pp. 1091–1105, 1994.

[96] D. Landolt, “Electrochemical and materials science aspects of alloy deposition,” Electrochimica Acta, vol. 39, no. 8-9, pp. 1075–1090, 1994.

[97] Z. F. Lodhi, J. M. C. Mol, A. Hovestad, H. Terryn, and J. H. W. de Wit, “Elec- trodeposition of Zn-Co and Zn-Co-Fe alloys from acidic chloride electrolytes,” Surface and Coatings Technology, vol. 202, no. 1, pp. 84–90, 2007.

[98] H. Dahms and I. M. Croll, “The Anomalous Codeposition of Iron-Nickel Alloys,” Journal of The Electrochemical Society, vol. 112, no. 8, p. 771, 1965.

[99] H. Yan, “A Model for Nanolaminated Growth Patterns in Zn and Zn-Co Elec- trodeposits,” Journal of The Electrochemical Society, vol. 143, no. 5, p. 1577, 1996.

153 REFERENCES

[100] S. L. D´ıaz,O. R. Mattos, O. E. Barcia, and F. J. Fabri Miranda, “ZnFe anomalous electrodeposition: Stationaries and local pH measurements,” Electrochim- ica Acta, vol. 47, no. 25, pp. 4091–4100, 2002.

[101] A. C. Hegde, K. Venkatakrishna, and N. Eliaz, “Electrodeposition of Zn-Ni, Zn-Fe and Zn-Ni-Fe alloys,” Surface and Coatings Technology, vol. 205, no. 7, pp. 2031–2041, 2010.

[102] X. Qiao, H. Li, W. Zhao, and D. Li, “Eﬀects of deposition temperature on electrodeposition of zinc-nickel alloy coatings,” Electrochimica Acta, vol. 89, pp. 771–777, 2013.

[103] K. O. Nayana, T. V. Venkatesha, and K. G. Chandrappa, “Inﬂuence of additive on nanocrystalline, bright Zn-Fe alloy electrodeposition and its properties,” Surface and Coatings Technology, vol. 235, pp. 461–468, nov 2013.

[104] V. R. Rao and A. C. Hegde, “Synergistic eﬀect of gelatin and glycerol on electrodeposition of Zn-Ni alloy,” Metallurgical and Materials Transactions B: Pro- cess Metallurgy and Materials Processing Science, vol. 44, no. 5, pp. 1236–1242, 2013.

[105] E. Gómezand E. Vallés,“Electrodeposition of zinc cobalt alloys: inhibitory effect of zinc with convection and pH of solution,” Journal of Electroanalytical Chemistry, vol. 397, pp. 177–184, 1995.

[106] N. Eliaz, K. Venkatakrishna, and A. C. Hegde, “Electroplating and characterization of Zn-Ni, Zn-Co and Zn-Ni-Co alloys,” Surface and Coatings Technology, vol. 205, no. 7, pp. 1969–1978, 2010.

[107] Z. Zhang, W. H. Leng, H. B. Shao, J. Q. Zhang, J. M. Wang, and C. N. Cao, “Study on the behavior of Zn-Fe alloy electroplating,” Journal of Electroana- lytical Chemistry, vol. 516, no. 1-2, pp. 127–130, 2001.

154 REFERENCES

[108] H. Nakano, M. Matsuno, S. Oue, M. Yano, S. Kobayashi, and H. Fukushima, “Mechanism of anomalous type electrodeposition of Fe-Ni alloys from sulfate solutions,” Materials Transactions, vol. 45, no. 11, pp. 3130–3135, 2004.

[109] Z. N. Yang, Z. Zhang, and J. Q. Zhang, “Electrodeposition of decorative and protective Zn-Fe coating onto low-carbon steel substrate,” Surface and Coatings Technology, vol. 200, no. 16-17, pp. 4810–4815, 2006.

[110] G. Roventi, R. Fratesi, R. A. Della Guardia, and G. Barucca, “Normal and anomalous codeposition of Zn-Ni alloys from chloride bath,” Journal of Applied Electrochemistry, vol. 30, no. 2, pp. 173–179, 2000.

[111] G. Roventi, R. Cecchini, a. Fabrizi, and T. Bellezze, “Electrodeposition of nick- elzinc alloy coatings with high nickel content,” Surface and Coatings Technology, vol. 276, pp. 1–7, 2015.

[112] H. Fukushima, T. Akiyama, M. Yano, T. Ishikawa, and R. Kammel, “Elec- trodeposition Behavior of Zn-lron-group Metal Alloys from Sulfate and Chloride Baths,” ISIJ International, vol. 33, no. 1 993, p. 1009, 1993.

[113] K. Higashi, “Mechanism of the Electrodeposition of Zinc Alloys Containing a Small Amount of Cobalt,” Journal of The Electrochemical Society, vol. 128, no. 10, p. 2081, 1981.

[114] E. G´omez,E. Pelaez, and E. Vall´es,“Electrodeposition of zinc+iron alloys: I. Analysis of the initial stages of the anomalous codeposition,” Journal of Electroanalytical Chemistry, vol. 469, no. 2, pp. 139–149, 1999.

[115] Z. F. Lodhi, J. M. C. Mol, W. J. Hamer, H. A. Terryn, and J. H. W. De Wit, “Cathodic inhibition and anomalous electrodeposition of Zn-Co alloys,” Electrochimica Acta, vol. 52, no. 17, pp. 5444–5452, 2007.

155 REFERENCES

[116] G. D. Wilcox and D. R. Gabe, “Electrodeposited zinc alloy coatings,” Corrosion Science, vol. 35, no. 5-8, pp. 1251–1258, 1993.

[117] J. D. Jensen, D. R. Gabe, and G. D. Wilcox, “The practical realisation of zinc iron CMA coatings,” Surface and Coatings Technology, vol. 105, no. February, pp. 240–250, 1998.

[118] S. Amirat, R. Rehamnia, M. Bordes, and J. Creus, “Zn-Fe alloy electrodeposition from chloride bath: Inﬂuence of deposition parameters on coatings morphology and structure,” Materials and Corrosion, vol. 64, no. 4, pp. 328–334, 2013.

[119] C. Q. Yang, Y. C. Zhou, and Z. L. Long, “Electrodeposition and physico- chemical properties of Zn-Fe alloy coatings from sulfate solution,” Journal of Materials Science Letters, vol. 21, no. 21, pp. 1677–1680, 2002.

[120] I. H. Karahan, “A study on electrodeposited Zn1-x Fe x alloys,” Journal of Materials Science, vol. 42, no. 24, pp. 10160–10163, 2007.

[121] I. H. Karahan, “Eﬀects of pH value of the electrolyte and glycine additive on formation and properties of electrodeposited Zn-Fe coatings,” The Scientiﬁc World Journal, vol. 2013, 2013.

[122] N. Clark, P. Eidler, and P. Lex, “Development oﬀ Zinc / Bromine Batteries for Load-Leveling Applications : Phase 2 Final Report,” Tech. Rep. July, Sandia National Laboratories, 1999.

[123] M. Inoue, Y. Tsuzuki, and M. Shimada, “Carbon Fiber Electrode for Redox Flow Battery,” Carbon, vol. 134, no. 3, pp. 756–757, 1987.

[124] P. Butler, P. Eidler, and P. Grimes, “Zinc/Bromine Batteries.,” in Advanced Battery Systems, 1992.

156 REFERENCES

[125] P. Eidler, “Development of Zinc/Bromine Batteries for Load-Leveling Applica- tions: Phase 1 Final Report (Sandia Report SAND99-1853),” tech. rep., 1999.

[126] N. Tokuda, T. Kumamoto, T. Shigematsu, H. Deguchi, T. Ito, N. Yoshikawa, and T. Hara, “Development of a redox ﬂow battery system,” SEI Tech Rev, vol. 50, pp. 88–94, 1998.

[127] J. Pan, Y. Sun, J. Cheng, Y. Wen, Y. Yang, and P. Wan, “Study on a new single ﬂow acid Cu-PbO2 battery,” Electrochemistry Communications, vol. 10, pp. 1226–1229, sep 2008.

[128] V. Sprenkle, S. Kim, A. Crawford, Z. Nie, V. Viswanathan, D. Stephenson, Q. Luo, W. Wang, Z. G. Yang, L. Li, G. Xia, and I. Gyuk, “Next Generation Redox Flow Battery Prototype Development,” 2012.

[129] S. Corcuera and M. Skyllas-Kazacos, “State-of-Charge Monitoring and Elec- trolyte Rebalancing Methods for the Vanadium Redox Flow Battery,” European Chemical Bulletin, vol. 1, no. 12, pp. 511–519, 2012.

[130] W. Wang, Z. Nie, B. Chen, F. Chen, Q. Luo, X. Wei, G. G. Xia, M. Skyllas- Kazacos, L. Li, and Z. Yang, “A new Fe/V redox ﬂow battery using a sul- furic/chloric mixed-acid supporting electrolyte,” Advanced Energy Materials, vol. 2, no. 4, pp. 487–493, 2012.

[131] Q. Lai, H. Zhang, X. Li, L. Zhang, and Y. Cheng, “A novel single ﬂow zinc- bromine battery with improved energy density,” Journal of Power Sources, vol. 235, pp. 1–4, aug 2013.

[132] D. Rose and S. Ferreira, “Performance Testing of Zinc-Bromine Flow Batteries for Remote Telecom Sites,” The Battcon 2013 Stationary Battery Conference and Trade Show, pp. 1–11, 2013.

157 REFERENCES

[133] L. H. Thaller, “Electrochemical cell for rebalancing redox ﬂow system.,” U. S. Pat. Appl., vol. 4159366, no. US914260A0, p. 12 pp. Avail. NTIS., 1978.

[134] R. F. Gahn, “Method and Apparatus for Rebalancing a Redox Flow Cell,” 1986.

[135] A. Q. Pham and K. Chang, “Systems and Methods for Rebalancing Redox Flow Battery,” US Patent 8980454 B2, 2015.

[136] G. B. Adams, “Electrically rechargeable battery,” US Patent 4180623, 1979.

[137] Z. Xie, Q. Su, A. Shi, B. Yang, B. Liu, J. Chen, X. Zhou, D. Cai, and L. Yang, “High performance of zinc-ferrum redox ﬂow battery with Ac-/HAc buﬀer solution,” Journal of Energy Chemistry, vol. 25, no. 3, pp. 495–499, 2016.

[138] J. A. Mellentine, “Performance Characterization and Cost Assessment of an Iron Hybrid Flow Battery,” M. S. Thesis, 2011.

[139] J. Escudero-Gonzálezand P. Amparo López-Jiménez,“Iron redox battery as electrical energy storage system in the Spanish energetic framework,” Interna- tional Journal of Electrical Power and Energy Systems, vol. 61, pp. 421–428, 2014.

[140] S. Selverston, R. F. Savinell, and J. S. Wainright, “In-tank hydrogen-ferric ion recombination,” Journal of Power Sources, vol. 324, pp. 674–678, 2016.

[141] Y. Zeng, T. Zhao, X. Zhou, J. Zou, and Y. Ren, “A hydrogen-ferric ion rebalance cell operating at low hydrogen concentrations for capacity restoration of iron- chromium redox ﬂow batteries,” Journal of Power Sources, vol. 352, pp. 77–82, 2017.

[142] L. R. Petzold, “A description of DASSL: a Diferential/Algebraic System Solver,” Scientiﬁc Computing, pp. 65–68, 1983.

158 REFERENCES

[143] E. Nagelli, J. Pickering, Y. Zhu, G. Wnek, J. S. Wainright, and R. F. Savinell, “Reducing hydraulic permeability through microporous separators with polyvinyl alcohol coatings,” In Preparation, 2017.

[144] D. M. Draˇzi´c,“Iron and Its Electrochemistry in an Active State,” in Modern Aspects of Electrochemistry, pp. 69–192, Plenum, 1989.

[145] W. Sheng, Z. Zhuang, M. Gao, J. Zheng, J. G. Chen, and Y. Yan, “Correlating hydrogen oxidation and evolution activity on platinum at diﬀerent pH with measured hydrogen binding energy,” Nature Communications, vol. 6, p. 5848, 2015.

[146] T. Grundl and J. Delwiche, “Kinetics of ferric oxyhydroxide precipitation,” Journal of Contaminant Hydrology, vol. 14, no. 1, pp. 71–87, 1993.

[147] A. Stef´ansson,“Iron(III) hydrolysis and solubility at 25C,” Environmental Sci- ence and Technology, vol. 41, no. 17, pp. 6117–6123, 2007.

[148] J. Baumgartner and D. Faivre, “Iron solubility, colloids and their impact on iron (oxyhydr)oxide formation from solution,” Earth-Science Reviews, vol. 150, pp. 520–530, 2015.

[149] C. E. Evans and Y. Song, “Method and System for Rebalancing Electrolytes in a Redox Flow Battery System,” 2014.

[150] O. K. Chang and Q. Pham, “Rebalancing Electrolytes in Redox Flow Battery Systems,” 2014.

[151] L. Zhang, J. Cheng, Y. sheng Yang, Y. hua Wen, X. dong Wang, and G. ping Cao, “Study of zinc electrodes for single ﬂow zinc/nickel battery application,” Journal of Power Sources, vol. 179, no. 1, pp. 381–387, 2008.

159 REFERENCES

[152] P. K. Leung, C. Ponce De Le´on,and F. C. Walsh, “An undivided zinc-cerium redox ﬂow battery operating at room temperature (295 K),” Electrochemistry Communications, vol. 13, pp. 770–773, aug 2011.

[153] P. K. Leung, C. Ponce-De-Leon, C. T. J. Low, and F. C. Walsh, “Zinc deposition and dissolution in methanesulfonic acid onto a carbon composite electrode as the negative electrode reactions in a hybrid redox ﬂow battery,” Electrochimica Acta, vol. 56, no. 18, pp. 6536–6546, 2011.

[154] F. R. McLarnon and E. J. Cairns, “The Secondary Alkaline Zinc Electrode,” Journal of The Electrochemical Society, vol. 138, no. 2, p. 645, 1991.

[155] S. Mullendore, “What are ﬂow batteries and how can they improve resiliency, continuity and safety in urban installations?,” Power Point Presentation, 2016.

[156] A. J. Esswein, J. Goeltz, and D. Amadeo, “High solubility iron hexacyanides,” US Patent App 20140051003, vol. 1, no. 19, 2014.

[157] M. C. Tucker, V. Srinivasan, P. N. Ross, and A. Z. Weber, “Performance and cycling of the iron-ion/hydrogen redox ﬂow cell with various catholyte salts,” Journal of Applied Electrochemistry, vol. 43, pp. 637–644, may 2013.

[158] Safety Data Sheet for Potassium ferricyanide(III), “Safety Data Sheet for Potas- sium ferricyanide(III),” Www.Sigmaaldrich.Com, 2016.

[159] D. Lloyd, T. Vainikka, M. Ronkainen, and K. Kontturi, “Characterisation and application of the Fe(II)/Fe(III) redox reaction in an ionic liquid analogue,” Electrochimica Acta, vol. 109, pp. 843–851, 2013.

[160] H. Prifti, A. Parasuraman, S. Winardi, T. M. Lim, and M. Skyllas-Kazacos, “Membranes for redox ﬂow battery applications,” Membranes, vol. 2, no. 2, pp. 275–306, 2012.

160 REFERENCES

[161] T. Janoschka, N. Martin, U. Martin, C. Friebe, S. Morgenstern, H. Hiller, M. D. Hager, and U. S. Schubert, “An aqueous, polymer-based redox-ﬂow battery using non-corrosive, safe, and low-cost materials,” Nature, vol. 527, no. 7576, pp. 78–81, 2015.

[162] J. A. Mellentine, W. J. Culver, and R. F. Savinell, “Simulation and optimization of a ﬂow battery in an area regulation application,” Journal of Applied Electrochemistry, vol. 41, pp. 1167–1174, jun 2011.

[163] R. Bhat, B. K. Udaya, and A. Chitharanjan Hegde, “Optimization of deposition conditions for bright Zn-Fe coatings and its characterization,” Protection of Metals and Physical Chemistry of Surfaces, vol. 47, no. 5, pp. 645–653, 2011.

[164] H. Fukushima, T. Akiyama, J.-h. Lee, M. Yamaguchi, and K. Higashi, “Mech- anism of the electrodeposition of Zn with Iron-group metals from sulphate baths,” Transaction of the Japan Institute of Metals, vol. 24, no. 3, pp. 125–131, 1983.

[165] M. F. Mathias and T. W. Chapman, “The Composition of Electrodeposited Zinc-Nickel Alloy Coatings,” Journal of The Electrochemical Society, pp. 1408– 1416, 1987.

[166] E. Chassaing and R. Wiart, “Electrocrystallization mechanism of ZNi alloys in chloride electrolytes,” Electrochimica Acta, vol. 37, no. 3, pp. 545–553, 1992.

[167] F. J. Fabri Miranda, O. E. Barcia, O. R. Mattos, and R. Wiart, “Electrodepo- sition of Zn-Ni Alloys in Sulfate Electrolytes II. Reaction Modeling,” Journal of The Electrochemical Society, vol. 144, no. 10, pp. 3441–3449, 1997.

[168] F. J. Fabri Miranda, O. E. Barcia, O. R. Mattos, and R. Wiart, “Electrodeposi- tion of Zn-Ni Alloys in Sulfate Electrolytes I. Experimental Approach,” Journal of The Electrochemical Society, vol. 144, no. 10, pp. 3441–3449, 1997.

161 REFERENCES

[169] K. Y. Sasaki and J. B. Talbot, “Electrodeposition of Binary Iron-Group Alloys,” Journal of the Electrochemical Society, vol. 142, no. 3, pp. 775–782, 1995.

[170] N. Zech, E. J. Podlaha, and D. Landolt, “Rotating cylinder Hull cell study of anomalous codeposition of binary iron-group alloys,” Journal of Applied Elec- trochemistry, vol. 28, pp. 1251–1260, 1998.

[171] N. Zech, E. J. Podlaha, and D. Landolt, “Anomalous Codeposition of Iron Group Metals: I. Experimental Results,” Journal of The Electrochemical Soci- ety, vol. 146, no. 8, p. 2886, 1999.

[172] D. Crotty, “Zinc alloy plating for the automotive industry,” Metal Finishing, vol. 94, no. September, pp. 54–58, 1996.

[173] N. J. Magnani, R. P. Clark, J. W. Braithwaite, D. M. Bush, P. C. Butler, J. M. Freese, K. R. Grothaus, K. D. Murphy, and P. E. Shoemaker, “Exploratory Bat- tery Technology Development and Testing Report for 1985,” Sandia National Laboratories, no. June, 1987.

[174] S. J. Banik and R. Akolkar, “Suppressing Dendritic Growth during Alkaline Zinc Electrodeposition using Polyethylenimine Additive,” Electrochimica Acta, vol. 179, pp. 475–481, sep 2015.

[175] R. Winand, “Electrodeposition of Zinc and Zinc Alloys,” Modern Electroplating: Fifth Edition, pp. 285–307, 2011.

[176] V. Viswanathan, A. Crawford, D. Stephenson, S. Kim, W. Wang, B. Li, G. Cof- fey, E. Thomsen, G. Graﬀ, P. Balducci, M. Kintner-Meyer, and V. Spren- kle, “Cost and performance model for redox ﬂow batteries,” Journal of Power Sources, vol. 247, pp. 1040–1051, 2014.

162 REFERENCES

[177] H. Dietz, L. Dittmar, D. Ohms, and M. Radwan, “Noble metal-free catalysts for the hydrogen-oxygen recombination in seale d lead acid batteries using im- mobilized electrolytes,” pp. 175–186, 1992.

[178] I. Nikolov, G. Papazov, and V. Naidenov, “Activity and corrosion of tungsten carbide recombination electrodes during lead/acid battery operation,” Journal of Power Sources, vol. 40, no. 3, pp. 333–340, 1992.

[179] F. Harnisch, G. Sievers, and U. Schroder, “Tungsten carbide as electrocatalyst for the hydrogen evolution reaction in pH neutral electrolyte solutions,” Applied Catalysis B: Environmental, vol. 89, no. 3-4, pp. 455–458, 2009.

163