a benchmarking approach for routine determination of flow

42
doi.org/10.26434/chemrxiv.6203297.v1 A Benchmarking Approach for Routine Determination of Flow Battery Kinetics Tejal Sawant, James McKone Submitted date: 01/05/2018 Posted date: 01/05/2018 Licence: CC BY-NC-ND 4.0 Citation information: Sawant, Tejal; McKone, James (2018): A Benchmarking Approach for Routine Determination of Flow Battery Kinetics. ChemRxiv. Preprint. This work focusses on improved precision and reproducibility in the study of redox flow battery (RFB) kinetics. We measured the electron-transfer reaction rates of the Fe(III/II) redox couple at polycrystalline Pt and Au electrodes in aqueous HCl supporting electrolyte using rotating disk electrode voltammetry. We made considerable effort to implement a systematic electrode preparation protocol, which was necessary for reproducibility. We found the reaction to be quasi-reversible at both electrodes and Pt to be a slightly more effective catalyst than Au. We further discuss some of the benefits and challenges of applying classical electroanalysis to RFB device design. File list (2) download file view on ChemRxiv Sawant_RFB_Benchmarking.pdf (2.50 MiB) download file view on ChemRxiv Sawant_RFB_Benchmarking_SI.pdf (1.32 MiB)

Upload: others

Post on 04-May-2022

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: A Benchmarking Approach for Routine Determination of Flow

doi.org/10.26434/chemrxiv.6203297.v1

A Benchmarking Approach for Routine Determination of Flow BatteryKineticsTejal Sawant, James McKone

Submitted date: 01/05/2018 • Posted date: 01/05/2018Licence: CC BY-NC-ND 4.0Citation information: Sawant, Tejal; McKone, James (2018): A Benchmarking Approach for RoutineDetermination of Flow Battery Kinetics. ChemRxiv. Preprint.

This work focusses on improved precision and reproducibility in the study of redox flow battery (RFB) kinetics.We measured the electron-transfer reaction rates of the Fe(III/II) redox couple at polycrystalline Pt and Auelectrodes in aqueous HCl supporting electrolyte using rotating disk electrode voltammetry. We madeconsiderable effort to implement a systematic electrode preparation protocol, which was necessary forreproducibility. We found the reaction to be quasi-reversible at both electrodes and Pt to be a slightly moreeffective catalyst than Au. We further discuss some of the benefits and challenges of applying classicalelectroanalysis to RFB device design.

File list (2)

download fileview on ChemRxivSawant_RFB_Benchmarking.pdf (2.50 MiB)

download fileview on ChemRxivSawant_RFB_Benchmarking_SI.pdf (1.32 MiB)

Page 2: A Benchmarking Approach for Routine Determination of Flow

A Benchmarking Approach for Routine

Determination of Flow Battery Kinetics

Tejal V. Sawant and James R. McKone∗

Department of Chemical and Petroleum Engineering, Swanson School of Engineering, University

of Pittsburgh, Pittsburgh, PA 15261, USA

E-mail: [email protected]

Abstract

The redox flow battery (RFB) is a promising method for large-scale electrochemical energy

storage, but research progress has been hampered by conflicting reports of electron-transfer

kinetics even for well-established RFB chemistries. We report a method for benchmarking

the electron-transfer kinetics of RFB electrode-electrolyte combinations using rotating disk

electrode(RDE) voltammetry. Our approach was modeled after analogous methods for bench-

marking heterogeneous electrocatalysts and critically relies on consistent electrode surface

preparation to obtain reproducible results. We have applied this approach to aqueous Fe3+/2+

redox chemistry using Pt and Au electrodes in HCl(aq) as a well-behaved model of a Fe/Cr

RFB positive electrolyte chamber. Our measurements yielded exchange current densities of 3.7

±0.5 and 1.3 ±0.2 mA/cm2 for Pt and Au, respectively, in electrolytes containing 10 mM total

dissolved Fe. Both the variability and relative sluggishness of these rates are clear evidence

that inner-sphere (catalytic) processes are important even in the 1-electron redox chemistry of

Fe aquo complexes. Moreover, the approach can be easily adapted for studying electrocatalytic

properties of a range of RFB electrolytes at well-defined electrode surfaces.

1

Page 3: A Benchmarking Approach for Routine Determination of Flow

Introduction

Energy demand worldwide is massive, and our current energy system is widely seen as environ-

mentally unsustainable. Renewable energy technologies, particularly wind and solar electricity,

are highly attractive methods for decarbonization of the global energy supply.1–3 However, these

types of renewable resources are spatially and temporally intermittent, which would make it diffi-

cult to maintain a stable electric grid based predominantly on wind and solar power.4–7 To address

this challenge, grid-scale electrochemical energy storage can be used to provide large quantities of

renewable electricity when and where it is needed.8–10

The redox flow battery (RFB) is one promising method for grid-scale electrochemical

energy storage.11–14 RFBs combine elements of solid-state secondary batteries and fuel cells. Like

solid-state batteries, they operate via reversible redox reactions between two sets of electroactive

materials, such as transition metal complexes or redox-active organics. However, in RFBs the

electroactive materials are dissolved or suspended in liquid media to form positive and negative

electrolytes, which flow through a charge-discharge stack to interconvert between electrical and

chemical energy in a process analogous to the operation of a fuel cell. RFB electrolytes can be

stored in tanks of arbitrary size, which make these systems amenable to scalable implementation.

Moreover, the storage tanks and charge/discharge stacks can be sized independently, which pro-

vides flexibility in the relationship between energy and power in RFBs.

RFBs have been the subject of considerable attention in both laboratory studies and com-

mercial development, as exemplified by the extensive review literature.15–24 Early work at NASA

focused primarily on Fe and Cr salts dissolved in aqueous acidic solutions.25–27 The all-vanadium

RFB, which uses V3+/2+ and V5+/4+ redox couples as the respective negative and positive active

materials, has also been studied extensively.28–36 Considerable work is now focused on develop-

ment of new electrolytes; thus, there is a growing library of active materials of interest for aqueous

RFBs.37–43 There is also increasing work on nonaqueous RFB active materials, particularly elec-

troactive organics and organometallics, which offer the advantage of high cell voltage due to the

2

Page 4: A Benchmarking Approach for Routine Determination of Flow

extended stability window of organic solvents.44–48

A major challenge for the design and implementation of RFBs involves understanding

and improving the rate of interfacial electron transfer from the positive and negative electrodes to

the corresponding electrolyte species. For example, the aqueous Fe3+/2+ redox couple serves as the

positive electrolyte in Fe/Cr RFBs, and the electrochemical kinetics of Fe3+/2+ redox chemistry have

been measured using various methods, including RDE voltammetry,49–57 AC impedance,58 faradaic

rectification,59 and current/voltage pulse techniques.60,61 While it is potentially valuable to com-

pare results across multiple measurement techniques, these studies nevertheless report values of the

interfacial electron-transfer rate constant, k0, that vary over at least two orders of magnitude even

for relatively well-understood electrode materials like polycrystalline Pt and Au (reported values

are collected into a table in the Supporting Information). This variability is highly problematic for

RFB design, as illustrated in Figure 1 in which we have modeled the simulated charge-discharge

performance of a hypothetical Fe3+/2+ RFB electrolyte at a smooth electrode surface at the lower

and upper limits of reported intrinsic reaction rate constants at Au. Moreover, this level of vari-

Figure 1: Simulated performance of a Fe3+/2+ RFB positive electrolyte in terms of half-cell overpo-tential versus state of charge, wherein the heterogeneous electron transfer rate constant was variedbetween k0 = 10−2 and 10−5 cm/s, corresponding to the outer bounds of reported values at Auelectrodes. Complete simulation details are described in the Supporting Information.

ability is not restricted to iron redox chemistry; vanadium RFB electrolytes have also been the

subject of some debate over whether the positive V5+/4+ or negative V3+/2+ redox couple exhibits

3

Page 5: A Benchmarking Approach for Routine Determination of Flow

more sluggish electron-transfer kinetics.19,62–68

We are working toward generalizable methods for characterizing the interfacial chem-

istry of RFB electrode-electrolyte combinations in the interest of resolving ambiguity regarding

electrode kinetics and improving our ability to model and build efficient devices. Thus, we have

developed an RFB kinetics benchmarking protocol based on rotating disk electrode (RDE) voltam-

metry that is modeled after analogous techniques for standardized characterization of fuel cell and

water electrolysis catalysts.69–72 Our approach emphasizes rigorous electrode surface preparation,

which we have found to be extremely important to obtain reproducible kinetics of RFB-mimicking

electrolytes. We report here a detailed description of our benchmarking method as applied to the

redox chemistry of Fe3+/2+ using polycrystalline Pt and Au electrodes operated under conditions

intended to approximate those that exist in a functional RFB.

We found exchange current densities (j0) of Fe3+/2+ redox reaction to be 3.7 ±0.5 and

1.3 ±0.2 mA/cm2 on Pt and Au electrodes, respectively, in electrolytes containing 5 mM concen-

trations each of Fe3+ and Fe2+. Importantly, we also found that electrode surfaces needed to be

repeatedly refreshed using electrochemical cleaning methods to obtain reproducible results. We

were also able to apply our analytical approach to electrolytes with 1 M total Fe concentration to

resemble a practical Fe3+/2+ RFB positive electrolyte. Interestingly, even after compensating for

series resistance, we measured j0 = 38 ±4 mA/cm2 at polycrystalline Pt, which was 10-fold slower

than would be predicted from conventional kinetic models where j0 is proportional to reactant con-

centration. Thus, it is clear that, even for well-studied RFB chemistries, interfacial electron transfer

dynamics are complex and merit careful scrutiny using the tools of applied electroanalysis.

Experimental

Hydrochloric acid (Certified ACS plus), sulfuric acid (trace metal grade), FeCl2 (tetrahydrate salt,

98%) and FeCl3 (hexahydrate salt, 97%), were obtained from Fisher Scientific and were used

4

Page 6: A Benchmarking Approach for Routine Determination of Flow

without further purification. All solutions were prepared using deionized water with resistivity of

≥18.2 MΩ·cm (Millipore, Milli-Q Advantage A10). Commercial acidic and alkaline detergents

(Citranox and Alconox, respectively) were obtained from W.W. Grainger Inc. Alumina powders

and polishing pads (Micropad) were obtained from Pace Technologies. Graphite electrodes were

spectroscopic grade with a diameter 1/4” and porosity of 16.5 % and were obtained from Electron

Microscopy Sciences. Reference electrodes were Ag/AgCl gel-type electrodes with 3 N NaCl fill

solution and were obtained from Fisher Scientific. Ultra-high purity H2(g) (99.999%) and zero

grade N2(g) (99.998%) were obtained from Matheson.

The RDE apparatus was a Pine MSR rotator equipped with ChangeDisk electrodes that

were 5 mm in diameter. All electrochemical measurements were performed on a Gamry Interface

1000 E potentiostat. The electrochemical cell used was a 100 mL glass single chamber equipped

with a tight fitting Teflon cap into which holes were drilled to introduce electrodes and gas flow tub-

ing. Our benchmarking method was built around the use of RDE voltammetry to extract transport

and kinetic properties of RFB active materials. Figure 2 depicts the associated experimental setup

for these measurements where the only variable was the identity of the RDE working electrode.

A representative experimental procedure for executing RDE measurements with poly-

crystalline Pt and Au working electrodes is as follows. First, the glass electrochemical cell and the

separate reference electrode chamber were cleaned in 1 wt% aqueous Citranox solution by boiling

for 30 mins followed by rinsing 10 times with pure water. The same treatment was then repeated

using 1 wt% aqueous Alconox solution. Meanwhile, 50 mL of 0.5 M sulfuric acid solution was

prepared, and the equilibrium potential of the Ag/AgCl reference electrode was calibrated against

a reversible hydrogen electrode (RHE) with this solution using a clean Pt button electrode. A rep-

resentative result for this calibration was EAg/AgCl = 0.233 V vs. RHE, which is in close agreement

with the predicted value of 0.227 V vs. RHE (0.209 V vs. NHE).

Next, a stock solution of 150 mL of 0.5 M HCl was prepared. 50 mL of this solution was

added to a large polyethylene vial and appropriate quantities of FeCl2 · 4 H2O and FeCl3 · 6 H2O

5

Page 7: A Benchmarking Approach for Routine Determination of Flow

Figure 2: Schematic (left) and photograph (right) of the RDE experimental setup. Componentslabeled in the schematic are as follows: (a) glass cell, (b) teflon cap, (c) nitrogen purge tube, (d)vent, (e) RDE motor, (f) shaft, (g) 5mM FeCl2 and 5 mM FeCl3 in 0.5 M HCl, (h) Pt workingelectrode, (i) graphite counter electrode, (j) stopcock, (k) electrolyte bridge, (l) Ag/AgCl referenceelectrode and (m) reference electrode compartment.

were added to the vial to produce a solution of 5 mM FeCl2 and 5mM FeCl3. A few mL of this

solution was used to rinse the electrochemical cell, and the balance was then added to the cell for

data collection.

For the preparation of the working electrode, 3 separate pieces of polishing pad were

impregnated with 5, 1 and 0.05 µm alumina slurries in water, respectively. The working electrode

was first polished on 5 µm alumina slurry in a circular fashion with successive sets of 20 clockwise

and counter-clockwise rotations for a total of 1 minute. It was then rinsed and sonicated to remove

alumina residue by submersing in a water-filled 20 mL vial that was in turn placed in the sonicator

bath for 30 seconds. This polish–rinse–sonicate procedure was then repeated using the 1 and 0.05

µm alumina slurries. The electrode was finally rinsed thoroughly with water, attached to the elec-

trode rotator assembly, and lowered into the electrochemical cell without being allowed to dry. We

separately characterized the surface morphology and residual roughness of a Pt electrode prepared

in this manner; complete details and discussion are included in the Supporting Information.

6

Page 8: A Benchmarking Approach for Routine Determination of Flow

The graphite counter electrode was cleaned using gentle abrasion with a paper wipe

and placed in the same chamber as the working electrode. To connect the reference electrode,

a separate 20 mL vial was connected to the main cell via electrolyte bridge, which consisted of

a length of polymer tubing (Flexelene, Cole Parmer) coupled to a PVDF stopcock as shown in

Figure 2. To fill the vial with electrolyte from the main cell, first a syringe was used to draw an

initial quantity of solution from the main chamber into the tubing, after which the stopcock was

closed to hold the solution in place. Then the opposite end of the tubing was placed into the vial,

which was positioned at a lower height compared to the main cell to allow the flow of electrolyte by

gravity. Upon opening the stopcock, the electrolyte began flowing into the vial from the cell, then

the vial was raised and the liquid levels were allowed to equilibrate so that the reference chamber

contained ∼5 mL of electrolyte. The calibrated Ag/AgCl reference electrode was placed in the vial

to complete the cell setup. After cell assembly and prior to experimentation, the working electrode

chamber was sparged with N2(g) by bubbling through the electrolyte using Tygon tubing coupled

to a glass pipette as a sparge tube. The sparging proceeded for at least 5 minutes, which was

independently determined to be sufficient amount of time to remove atmospheric O2 as described

in the Supporting Information. Finally, the working electrode was cleaned electrochemically by

cycling in the potential range of -0.25 V to 1.25 V versus Ag/AgCl at 100 mV/s for 20 cycles.

Experimental data collection was performed as a repeating sequence of clean and mea-

sure steps. First, the working electrode was rotated at the desired rate and a cleaning step was

performed by cycling 5 times between -0.25 V to 1.25 V versus Ag/AgCl at 200 mV/s. Measure-

ments were then performed (while still rotating) by scanning the cell potential between 0.225 and

0.725 V versus Ag/AgCl. The experimental scan rate was chosen empirically from the range of

5–20 mV/s to obtain the lowest amount of hysteresis between the forward and reverse sweeps. The

voltage range was selected as the minimum needed to obtain steady-state anodic and cathodic lim-

iting currents. This sequence of cleaning and measurement steps was then repeated at each rotation

rate of interest. Rotation rates were varied in the order 100, 400, 900, 1600, 2500, 225, 625, 1225,

2025 and 100 rpm, and the experimental data were taken to be valid only if the first and last cycles

7

Page 9: A Benchmarking Approach for Routine Determination of Flow

were found to overlay.

After experimentation with Fe3+/2+ electrolyte was complete, background measurements

were completed by repeating the entire glassware cleaning protocol and replacing the electrolyte

with fresh 0.5 M HCl solution. The aforementioned N2 purge and electrode cleaning protocols

were repeated once more, and then background data were collected by running cyclic voltammetry

from 0.225 to 0.725 V versus Ag/AgCl without electrode rotation at the same scan rate as the

measurements taken in the presence of active species . These background data were then subtracted

from the experimental data prior to analysis.

We performed Koutecky-Levich (KL) analysis to find the transport properties (Diffu-

sivity, D) and non-linear Butler-Volmer fits to find exchange current densities (j0) and symmetry

factors (αox and αred) from the background-subtracted RDE data. We found current density as a

function of inverse square root of rotation rate, ω−0.5 for various overpotentials (η = E − Eeq).

Diffusivity was found using analysis which relies on the KL equation 1, which expresses the rela-

tionship between current density and rotation rate in an RDE experiment:

1

j=

1

jk+

1

0.620nFCD2/3ν−1/6ω−0.5 (1)

where j is current density, jk is kinetic current density, n is number of electrons transferred (1 in

the case of Fe3+/2+), F is Faraday’s constant (96485 C/mol), C is bulk reactant concentration, D

is diffusivity, ν is kinematic viscosity (taken as 0.01 cm2/s for dilute aqueous solutions at room

temperature), and ω is rotation rate.

From the intercepts of KL analysis, we extracted kinetic current density,jk as a function

of overpotential, η as the transport free polarization data. We used Butler-Volmer model to fit jk as

a function of η using non-linear least-square approach to find the exchange current density, j0, and

the symmetry factors, αox and αred according to Equation 2.

jk = j0[exp(αoxnFη

RT) − exp(

−αrednFη

RT)] (2)

8

Page 10: A Benchmarking Approach for Routine Determination of Flow

where j0 is exchange current density, αox and αred are symmetry factors for oxidation and reduction

respectively, R is the universal gas constant, T is temperature and η = E − Eeq where Eeq is the

equilibrium potential of the system. This exchange current density can be expressed as a function

of concentration, C and reaction rate constant, k0 according to Equation 3,

j0 = nFCk0 (3)

where k0 is the intrinsic reaction rate constant.

Results

A qualitative picture of the kinetics of Fe3+/2+ is readily apparent from voltammograms under qui-

escent conditions. Figure 3 depicts representative cyclic voltammetry (CV) data collected at 0

rpm and a scan rate of 200 mV/s for Pt and Au electrodes. Oxidative and reductive waves are

well-resolved in each case. The peak-to-peak separation values, which vary inversely with reac-

tion kinetics, were 0.08 and 0.14 V for Pt and Au electrodes, respectively, at this scan rate. The

electron-transfer can therefore be taken as quasi-reversible in both cases, with Pt exhibiting some-

what faster kinetics.73 More extensive scan rate dependence data for quiescent CV experiments are

included in the Supporting Information, and show the expected linear increase in peak current with

the square root of scan rate for a diffusional redox process.

The main focus of this work was on RDE voltammetry to further quantify the kinetics of

Fe3+/2+ redox chemistry. Figure 4 therefore compiles background-subtracted RDE current density

vs. applied potential (j–E) data for each electrode type over a range of rotation rates from 100

to 2500 rpm. Each of these datasets exhibits a clear and consistent open-circuit potential of 0.475

V vs. Ag/AgCl (0.684 V vs. NHE), which is close to the reported standard reduction potential

of 0.7 V vs. NHE for Fe3+/2+ redox couple in 1 M HCl electrolyte.73 Oxidative and reductive

current densities increased monotonically in magnitude with increasing rotation rate, as expected

9

Page 11: A Benchmarking Approach for Routine Determination of Flow

Figure 3: Representative current density versus applied potential data for (a) polycrystalline Pt and(b) polycrystalline Au in 5mM FeCl2 and 5 mM FeCl3 in 0.5 M HCl(aq) collected at 0 rpm and atscan rate of 200 mV/s.

Figure 4: Representative current density versus potential data for RDE voltammetry between 100and 2500 rpm using (a) polycrystalline Pt and (b) polycrystalline Au in 5mM FeCl2 and 5mMFeCl3 in 0.5 M HCl(aq) at a scan rate of 5 mV/s.

10

Page 12: A Benchmarking Approach for Routine Determination of Flow

for progressively diminished transport limitations. Steady-state limiting current densities were

also clearly obtained except at the highest rotation rates in the oxidizing direction, where slightly

sloping j–E behavior was instead observed.

Figure 5 depicts representative KL analysis performed on the RDE data, comprising plots

of inverse current vs. ω−0.5 for overpotentials from 0.02 to 0.14 V in the positive and negative

directions. These data exhibited clear linear trends at overpotentials exceeding 0.02 V in magni-

Figure 5: Koutecky-Levich analysis depicting inverse current versus inverse square root of rotationrate data for (a) iron oxidation at Pt, (b) iron reduction at Pt, (c) iron oxidation at Au and (d) ironreduction at Au for overpotentials from 20 to 140 mV in 5mM FeCl2 and 5 mM FeCl3 in 0.5 MHCl.

tude. KL data collected at smaller overpotentials exhibited greater spread, which is consistent with

11

Page 13: A Benchmarking Approach for Routine Determination of Flow

increased uncertainty due to measurement noise and background subtraction artifacts when the

observed current densities are low. The slopes of the best fit lines through the data also converged

to a single value at overpotentials exceeding 0.06 V, which is consistent with the transition from

primarily kinetic to primarily diffusion limitation. Thus, the slopes of the data collected at ≥100

mV overpotential were used to extract diffusivity values.

Figure 6 shows transport-free polarization data for Pt and Au electrodes extracted from

linear fits to the KL equation. These data are plotted as natural log of kinetic current density

versus overpotential for both the oxidative and reductive half reactions. These data qualitatively

agree with a Butler-Volmer description of electrode kinetics, as the data show a linear trend at

overpotentials exceeding ∼60 mV and deflect to smaller current densities at lower overpotentials;

this negative deflection is attributable to competition between forward and reverse reactions at

small overpotentials.

Interestingly, application of a conventional Butler-Volmer model where the symmetry

factors (αox and αred) were constrained to sum to 1 for the oxidative and reductive half reactions

resulted in rather poor fits, as shown in the Supporting Information. Thus, the fits depicted as black

lines in Figure 6 correspond to least squares regression where the α values were allowed to vary

independently between 0 and 1 for each half reaction. These fits also included only overpotentials

exceeding 40 mV in magnitude, due to the uncertainty implicit in KL analysis of low-overpotential

data where considerable spread was observed. This accounts for the slight systematic deviation of

the fit lines from the transport-free polarization data at low overpotential.

We performed these RDE experiments and corresponding analysis a total of 5 times for

each electrode type. Table 1 summarizes the extracted transport and kinetics properties of Fe3+/2+

redox couple on Pt and Au electrodes along with the associated experimental uncertainties. Error

bounds were taken at 1 standard deviation from the mean of all 5 replicates. We found that the

diffusivity values for Pt and Au were between 3–4 x 10−6 cm2/s, and were indistinguishable within

a single standard deviation. However, extracted exchange current densities were clearly different:

12

Page 14: A Benchmarking Approach for Routine Determination of Flow

Figure 6: Transport free polarization data fitted using Butler-Volmer model with non-linear leastsquare analysis for iron oxidation and reduction at (a)Pt and (b) Au in 5mM FeCl2 and 5 mM FeCl3in 0.5 M HCl.

Table 1: Transport and kinetics properties of iron oxidation and reduction at Pt and Au electrodes.Standard deviation values are reported on the basis of n=5 set of experiments

Electrode Diffusivity,Dox

(cm2/s)Diffusivity,Dred

(cm2/s)Exchange current density

j0 (mA/cm2) αox αred

Pt(n=5)4 x 10−6

(±0.1 x 10−6)3 x 10−6

(±0.2 x 10−6)3.7

(±0.5)0.33

(±0.08)0.35

(±0.09)

Au(n=5)4 x 10−6

(±0.5 x 10−6)3 x 10−6

(±0.2 x 10−6)1.3

(±0.2)0.25

(±0.05)0.45

(±0.05)

13

Page 15: A Benchmarking Approach for Routine Determination of Flow

3.7 ±0.5 mA/cm2 for polycrystalline Pt and 1.3 ±0.2 mA/cm2 for polycrystalline Au. Thus, the

Fe3+/2+ redox reaction was indeed faster at Pt than Au, but only by a factor of ∼3. Finally, when

symmetry factors were allowed to vary arbitrarily, they were found to be α < 0.5 in all cases. The

αox for Pt and Au was between 0.25 and 0.33, while αred was between 0.35 and 0.45.

Discussion

Our experimental approach was designed to balance between the goals of mimicking functional

flow battery behavior while also maintaining suitable analytical precision. Aqueous Fe chloride

salts and HCl supporting electrolyte were therefore chosen because they form the basis of the

Fe positive electrolyte in the Fe/Cr RFB, which we expected to be a well-behaved electrolyte for

validating a RFB electrode–electrolyte benchmarking protocol. The redox couple was added at

low concentration—10 mM total Fe—while the supporting electrolyte was maintained at 0.5 M

concentration. This diverges from electrolyte conditions in a functional flow battery to avoid the

confounding effects of overpotential attributable to uncompensated resistance, which arises from

high current densities in these systems. We also chose to study equimolar mixtures of Fe2+ and

Fe3+, which simplifies the experimental procedure by enabling analysis of oxidative and reductive

half-reactions in a single series of RDE voltammograms. Moreover, a mixture of oxidized and

reduced forms of the redox couple closely resembles the conditions of a working RFB cell, since

these batteries are rarely cycled to 0 or 100 % state of charge once in operation. Although Pt and

Au are not widely used in functional RFBs (due to their high cost), they were chosen for this study

because there already exist well-established preparation and cleaning protocols for these materials.

RDE voltammetry also approximates the hydrodynamic conditions in an RFB. The dif-

fusion boundary at a rotating electrode is in fact more homogeneous than in a flow cell, thereby

permitting straightforward use of KL analysis to extract kinetic and transport parameters. RDE

voltammetry has the advantage of being widely used in standard analytical protocols for character-

izing the kinetics of electrocatalysts.75,76 Thus, instrumentation and expertise is readily translatable

14

Page 16: A Benchmarking Approach for Routine Determination of Flow

between these two related fields. Indeed, many of the procedures we used were adapted directly

from prior work on benchmarking methods for electrolysis and fuel cell electrocatalysts.69–72 Sim-

ilar methods could also be deployed for nonaqueous RFB electrolytes, although great care would

be required to eliminate sources of contamination, particularly by air and water.

Although the cell and electrode preparation procedure we used here is onerous, it is con-

sistent with common practice in precision electroanalysis.74 In fact, we found that every step of

the process was necessary to successfully differentiate the catalytic activities of Pt and Au toward

Fe3+/2+. To illustrate, Figure 7 shows the results of RDE voltammetry experiments at a Pt electrode

where only one step—the electrochemical cleaning step implemented between collecting voltam-

mograms at different rotation rates—was omitted. In this case, we observed a clear progressive

decrease in the current response at a given applied potential over time, which is particularly evi-

dent when the rotation rates are varied as shown in Figure 7. We attribute the decreased current

to progressive electrode fouling, which was mitigated by electrochemically cleaning the electrode

between each RDE measurement. It is relevant to note that, while cleaning steps like those em-

ployed here are routinely implemented in the course of electroanalysis, it would not be plausible

to implement an analogous cleaning protocol in a functional RFB. Thus, it may not be reasonable

to expect that pre-treatment procedures, even at well-defined electrode surfaces, will give rise to

optimal electrode kinetics over extended battery operation.

Our experimental results broadly show that, although the Fe3+/2+ redox couple involves

the transfer of only one electron, the kinetics are clearly indicative of a multi-step reaction mecha-

nism where the electrode surface can play a catalytic role. Two readily apparent indications are that

(a) Pt and Au electrodes clearly show different electron-transfer kinetics, and (b) even at Pt, the re-

action kinetics are considerably slower than those routinely observed from outer-sphere, 1-electron

transfer reagents like ferrocene.77 Moreover, acceptable fits to transport-free polarization data did

not obey a classical BV model where the symmetry factors for the forward and reverse reactions

sum to one. This characteristic of the BV model is based on the assumption of a mechanism where

a single electron-transfer step governs the rate of the forward and reverse reactions; thus, α values

15

Page 17: A Benchmarking Approach for Routine Determination of Flow

Figure 7: (a) RDE current density versus potential data at different rotation rates at a scan rate of5mV/s for platinum electrode in 5 mM FeCl2 and 5 mM FeCl3 in 0.5 M HCl obtained by omittingcleaning steps between different rotation rates. (b) Current density versus rotation rate at a potentialof 375 mV versus Ag/AgCl with and without cleaning steps. Numbers represent the order in whichRDE measurements were performed.

that diverge from the expected relationship are also consistent with multi-step redox chemistry.

This is not surprising in light of the known complexities of Fe3+/2+ electron transfer self-exchange

processes in aqueous solutions, which have in fact been the subject of extensive scrutiny.78–83

Based on our observations, the incongruence in prior reports of RFB electrode kinetics

may arise from a combination of varying electrode treatment procedures along with the ques-

tionable use of interfacial electron transfer rate constant, k0 as the main kinetic descriptor. The

parameter k0 is rightly understood as an intrinsic electron-transfer rate, analogous to the kinetic

constant k in solution-phase reactions. However, unlike homogeneous reactions in which the re-

acting species are highly uniform in composition and structure, polycrystalline electrode surfaces

exhibit considerable surface site heterogeneity. Thus, for reactions in which the electrode surface

acts as a catalyst, each site likely exhibits a different characteristic k0 and these may vary widely.

This helps explain why consistent electrode surface preparation is so important to obtain repro-

ducible kinetics data. We further conjecture that consistent application of several different surface

preparation protocols may even give rise to consistent reaction kinetics that nonetheless diverge

from one another. Thus, we recommend against the use of k0 to describe RFB electron-transfer ki-

16

Page 18: A Benchmarking Approach for Routine Determination of Flow

netics except in the unusual circumstance where the electrode surface is structurally homogeneous.

We have instead reported j0 values, which clearly depend on extensive properties of the electrode

(number density of active sites) and electrolyte (reactant concentration). Nevertheless, for the sake

of comparison j0 is easily interconverted with “apparent” k0 via the relationship in Equation 3.

Many analogous transition metal complexes and even organic/organometallic species of

interest for RFBs are likely to benefit from careful studies of interfacial catalytic reactivity. This

further motivates the use of a consistent benchmarking approach, where it will be helpful to char-

acterize the electron-transfer kinetics of a wide range of RFB electrode-electrolyte combinations

under mutually similar conditions. Improved insights into RFB interfacial electron transfer dynam-

ics will depend critically on developing consistent preparation methods not just for model electrode

surfaces, like the ones used here, but also for technologically relevant RFB electrode materials like

graphitic carbon in its various forms.

It is important to assess the utility of our analytical approach for predicting the practical

performance of functional flow batteries. Thus, we have carried out RDE analysis of Pt electrode

in a more realistic Fe RFB positive electrolyte containing 1 M total Fe (0.5 M each of Fe3+ and

Fe2+) in 2 M HCl(aq). All analytical procedures were followed exactly as described in the Experi-

mental section, except for the change in electrolyte composition and the compensation of solution

resistance, which was measured to be 4 ohms via impedance spectroscopy at high frequency. Fig-

ure 8 shows the associated RDE and transport-free polarization data. We first observed a negative

shift in the equilibrium potential to 0.425 V vs Ag/AgCl in the concentrated Fe RFB electrolyte,

even though it contained the same 1:1 mixture of Fe3+ and Fe2+ as in the prior measurements. We

also found j0 = 38 ±4 mA/cm2, which is almost 10-fold lower than the value of 370 mA/cm2 that

would be predicted from the simple extrapolation of the low-concentration data based on a linear

relationship between j0 and reactant concentration. Nevertheless, our experimental method yielded

reproducible results even at higher concentrations if the electrode preparation and analysis steps

were performed carefully. These results therefore provide considerable impetus for further studies

aimed at understanding differences in thermodynamics and kinetics of RFB redox chemistry under

17

Page 19: A Benchmarking Approach for Routine Determination of Flow

analytical (low concentration) and practical (high concentration) conditions. We hypothesize that

these differences are mainly attributable to variable speciation of the redox-active Fe species as a

function of total Fe and chloride concentration.

Figure 8: (a) RDE current density versus potential data at different rotation rates at a scan rate of5mV/s with 4 ohms resistance compensation (Inset represents the same data without compensatingfor resistance), (b) Transport free polarization data fitted using Butler-Volmer model with leastsqaure analysis approach for iron oxidation and reduction at Pt in 500 mM FeCl2 and 500 mMFeCl3 in 2 M HCl. Open circuit potential of the system =0.425V.

Although our RDE-based analytical approach was found to work well for studying Fe3+/2+

kinetics at Pt and Au electrodes, it may not work equally well for all RFB electrodes and elec-

trolytes. Because consistent results depend on the electrode composition remaining unchanged

during the course of the experiment, it may not be possible to extract reproducible kinetics param-

eters from electrode-electrolyte combinations that require potential excursions outside the elec-

trode/solvent stability window to approach mass-transfer limited operation. This condition may

arise from the use of redox couples that exhibit extreme standard reduction potentials, electrodes

that exhibit slow electron-transfer kinetics, or both. An analogous issue also exists in functional

RFBs, where the necessary application of highly oxidizing and reducing potentials in a working

battery may irreversibly modify the electrode surface. As a result, it may not be possible to op-

timize and maintain active electrodes for RFB electrolytes with redox potentials that lie near the

18

Page 20: A Benchmarking Approach for Routine Determination of Flow

boundaries of the solvent window. This possibility further illustrates the importance of understand-

ing the interplay between surface chemistry and electrocatalysis in technologically relevant RFB

electrode materials.

Conclusions

The primary aim of this study was to validate a systematic approach for extracting electron-transfer

kinetics properties of RFB electrode-electrolyte pairs by using established RDE methods. Our

approach gives reproducible results even for battery electrolytes at low and high concentrations,

but only if cleaning steps are performed to maintain scrupulously clean surfaces. Our results also

clearly indicate that the electrode surface plays a catalytic role for Fe3+/2+ chemistry. Moreover,

the incompatibility of polarization data with a conventional Butler-Volmer model emphasizes the

presence of inner sphere chemistry.

Our approach is straightforward and builds on established electroanalytical techniques

for which instrumentation and expertise are widely available. It is generalizable across multiple

electrode-electrolyte combinations and yields precise results when suitable care is taken to pre-

pare electrode surfaces. This approach could be readily implemented as a routine characterization

technique for a library of candidate RFB electrolytes, but further work is warranted to determine

whether or not RDE approaches can predict kinetics and transport behavior of functional RFBs.

To this end, ongoing work in our lab is focused on further development of precise electroanalyt-

ical methods that more closely resemble functional RFB architectures. Overall, the results from

this study emphasize the importance of electrochemical catalysis in the design of RFBs and there-

fore provide a basis for continued convergence between research in batteries and fuel cell energy

storage.

19

Page 21: A Benchmarking Approach for Routine Determination of Flow

Acknowledgements

We gratefully acknowledge the Swanson School of Engineering at the University of Pittsburgh for

financial and material support of this work. We also acknowledge Rituja Patil, Eli Bostian, Yifan

Deng, Aayush Mantri, and Emily Siegel for their editorial feedback during the preparation of this

manuscript.

References

(1) Johansson, T. B., Kelly, H., Reddy, A. K. . & Willliams, R. . Renewable fuels and electricity

for a growing world economy: Defining and achieving the potential. Energy Stud. Rev. 4, 201

- 212 (1992).

(2) Dincer, I. Renewable energy and sustainable development: a crucial review. Renew. Sustain.

Energy Rev. 4, 157 - 175 (2000).

(3) Luo, X., Wang, J., Dooner, M. & Clarke, J. Overview of current development in electrical

energy storage technologies and the application potential in power system operation. Appl.

Energy 137, 511 - 536 (2015).

(4) Hill, C. A., Such, M. C., Chen, D., Gonzalez, J. & Grady, W. M. K. Battery energy storage

for enabling integration of distributed solar power generation. IEEE Trans. Smart Grid 3, 850

- 857 (2012).

(5) Moslehi, K. & Kumar, R. A reliability perspective of the smart grid. IEEE Trans. Smart Grid

1, 57 - 64 (2010).

(6) Tarroja, B., Mueller, F., Eichman, J. D., Brouwer, J. & Samuelsen, S. Spatial and temporal

analysis of electric wind generation intermittency and dynamics. Renew. Energy 36, 3424 -

3432 (2011).

20

Page 22: A Benchmarking Approach for Routine Determination of Flow

(7) Bruno, R., Carbone, V., Veltri, P., Pietropaolo, E. & Bavassano, B. Identifying intermittency

events in the solar wind. Planet. Space Sci. 49, 1201 - 1210 (2001).

(8) Dell, R. . & Rand, D. A. . Energy storage a key technology for global energy sustainability. J.

Power Sources 100, 2 - 17 (2001).

(9) Dunn, B., Kamath, H. & Tarascon, J. M. Electrical Energy Storage for the Grid: A Battery of

Choices. Science (80-. ). 334, 928 - 935 (2011).

(10) Soloveichik, G. L. Battery Technologies for Large-Scale Stationary Energy Storage. Annu.

Rev. Chem. Biomol. Eng. 2, 503 - 527 (2011).

(11) Noack, J., Roznyatovskaya, N., Herr, T. & Fischer, P. The Chemistry of Redox-Flow Batter-

ies. Angew. Chemie - Int. Ed. 54, 9776 - 9809 (2015).

(12) Tokuda, N. et al. Development of a redox flow battery system. SEI Tech Rev, 88 - 94 (1998).

(13) Leung, P. et al. Progress in redox flow batteries, remaining challenges and their applications

in energy storage. RSC Adv. 2, 10125 (2012).

(14) Nguyen, T. & Savinell, R. F. Flow batteries. Electrochem. Soc. Interface 19, 54 - 56 (2010).

(15) Bartolozzi, M. Development of redox flow batteries. A historical bibliography. J. Power

Sources 27, 219 - 234 (1989).

(16) Pan, F. & Wang, Q. Redox species of redox flow batteries: A review. Molecules 20, 20499 -

20517 (2015).

(17) Wang, W. et al. Recent progress in redox flow battery research and development. Adv. Funct.

Mater. 23, 970 - 986 (2013).

(18) Yang, Z. et al. Electrochemical energy storage for green grid. Chem. Rev. 111, 3577 - 3613

(2011).

21

Page 23: A Benchmarking Approach for Routine Determination of Flow

(19) Weber, A. Z. et al. Redox flow batteries: A review. J. Appl. Electrochem. 41, 1137 - 1164

(2011).

(20) Skyllas-Kazacos, M., Chakrabarti, M. H., Hajimolana, S. a., Mjalli, F. S. & Saleem, M.

Progress in Flow Battery Research and Development. J. Electrochem. Soc. 158, R55 (2011).

(21) Ponce de Len, C., Fras-Ferrer, A., Gonzlez-Garca, J., Sznto, D. A. & Walsh, F. C. Redox flow

cells for energy conversion. J. Power Sources 160, 716 - 732 (2006).

(22) Lahiri, K., Raghunathan, a., Dey, S. & Panigrahi, D. Battery-driven system design: A new

frontier in low power design. Des. Autom. Conf. 2002. Proc. ASP-DAC 2002. 7th Asia South

Pacific 15th Int. Conf. VLSI Des. Proceedings. 261 - 267 (2002).

(23) Sauer, D. U. & Wenzl, H. Comparison of different approaches for lifetime prediction of

electrochemical systems-Using lead-acid batteries as example. J. Power Sources 176, 534 -

546 (2008).

(24) Soloveichik, G. L. Flow Batteries: Current Status and Trends. Chem. Rev. 115, 11533 - 11558

(2015).

(25) N H Hagedorn, N. and Thaller, L., Redox storage systems for solar applications, NASA TM

81464, National Aeronautics and Space Administration, U S Dept. of Energy, (1980).

(26) Randall,G. Hagedorn, N. and Ling, J., Single cell performance studied on the Fe/Cr redox

energy storage system using mixed reactant solutions at elevated temperature, NASA TM-

83385, National Aeronautics and Space Administration, U S Dept. of Energy, (1983).

(27) Thaller, L., Electrically rechargeable REDOX flow cell, U.S. Pat. 3,996 (1976).

(28) Kaneko, H. et al. Vanadium redox reactions and carbon electrodes for vanadium redox flow

battery. Electrochim. Acta 36, 1191 - 1196 (1991).

(29) Maria Skyllas-Kazacos , George Kazacos, G. P. & H. V. Recent Advances with UNSW

vanadium-based redox flow batteries. Int. J. Energy Res. 34, 182 - 189 (2010).

22

Page 24: A Benchmarking Approach for Routine Determination of Flow

(30) Skyllas-Kazacos, M. Novel vanadium chloride/polyhalide redox flow battery. J. Power

Sources 124, 299 - 302 (2003).

(31) Rychcik, M. & Skyllas-Kazacos, M. Characteristics of a new all-vanadium redox flow battery.

J. Power Sources 22, 59 - 67 (1988).

(32) Sun, B. & Skyllas-Kazacos, M. Modification of graphite electrode materials for vanadium

redox flow battery applicationI. Thermal treatment. Electrochim. Acta 37, 1253 - 1260 (1992).

(33) Sun, B. & Skyllas-Kazacos, M. Chemical modification of graphite electrode materials for

vanadium redox flow battery application-part II. Acid treatments. Electrochim. Acta 37, 2459

- 2465 (1992).

(34) Rychcik, M. & Skyllas-Kazacos, M. Evaluation of electrode materials for vanadium redox

cell. J. Power Sources 19, 45 - 54. (1987).

(35) Aaron, D. S. et al. Dramatic performance gains in vanadium redox flow batteries through

modified cell architecture. J. Power Sources 206, 450 - 453. (2012).

(36) Houser, J., Clement, J., Pezeshki, A. & Mench, M. M. Influence of architecture and material

properties on vanadium redox flow battery performance. J. Power Sources 302, 369 - 377.

(2016).

(37) Lin, K. et al. A redox-flow battery with an alloxazine-based organic electrolyte. Nat. Energy

1, 1 - 8 (2016).

(38) Huskinson, B. et al. A metal-free organic-inorganic aqueous flow battery. Nature 505, 195 -

198 (2014).

(39) Liu, T., Wei, X., Nie, Z., Sprenkle, V. & Wang, W. A Total Organic Aqueous Redox Flow

Battery Employing a Low Cost and Sustainable Methyl Viologen Anolyte and 4-HO-TEMPO

Catholyte. Adv. Energy Mater. 6, (2016).

23

Page 25: A Benchmarking Approach for Routine Determination of Flow

(40) Darling, R. M., Gallagher, K. G., Kowalski, J. A., Ha, S. & Brushett, F. R. Pathways to low-

cost electrochemical energy storage: a comparison of aqueous and nonaqueous flow batteries.

Energy Environ. Sci. 7, 3459 - 3477 (2014).

(41) Hu, B., DeBruler, C., Rhodes, Z. & Liu, T. L. Long-Cycling Aqueous Organic Redox Flow

Battery (AORFB) toward Sustainable and Safe Energy Storage. J. Am. Chem. Soc. 139, 1207

- 1214 (2017).

(42) Beh, E. S. et al. A Neutral pH Aqueous Organic-Organometallic Redox Flow Battery with

Extremely High Capacity Retention. ACS Energy Lett. 2, 639 - 644 (2017).

(43) Lin, K., Chen, Q., Gerhardt, M. R., Tong, L. & Bok, S. Alkaline Quinone Flow Battery.

Science. 349, 1529 - 1532 (2015)

(44) Milshtein, J. D. et al. Towards Low Resistance Nonaqueous Redox Flow Batteries. J. Elec-

trochem. Soc. 164, A2487 - A2499 (2017).

(45) Gong, K., Fang, Q., Gu, S., Li, S. F. Y. & Yan, Y. Nonaqueous redox-flow batteries: organic

solvents, supporting electrolytes, and redox pairs. Energy Environ. Sci. 8, 3515 - 3530 (2015).

(46) Suttil, J. A. et al. Metal acetylacetonate complexes for high energy density non-aqueous redox

flow batteries. J. Mater. Chem. A 3, 7929 - 7938 (2015).

(47) Sevov, C. S. et al. Evolutionary Design of Low Molecular Weight Organic Anolyte Materials

for Applications in Nonaqueous Redox Flow Batteries. J. Am. Chem. Soc. 137, 14465 - 14472

(2015).

(48) Wei, X. et al. Radical Compatibility with Nonaqueous Electrolytes and Its Impact on an All-

Organic Redox Flow Battery. Angew. Chemie - Int. Ed. 54, 8684 - 8687 (2015).

(49) Mcdermott, C. A., Kneten, K. R. Mccreery, R. L., Electron Transfer Kinetics of Aquated

Fe+3/+2,Eu+3/+2 and V+3/+2 at Carbon Electrodes, J. Electrochem. Soc., Vol. 140, No.

9,(1993).

24

Page 26: A Benchmarking Approach for Routine Determination of Flow

(50) Jordan, J. & Javick, R. Electrode Kinetics By Hydrodynamic Voltammetry-Study of Ferrous-

Ferric , and Iodide-Iodine Systems. Electrochim. Acta 6, 23 - 33 (1962).

(51) Jahn, D. & Vielstich, W. Rates of Electrode Processes by the Rotating Disk Method. J. Elec-

trochem. Soc. 109, 849 (1962).

(52) Angell, D. H. Dickinson, T. The kinetics of the ferrous/ferric and ferro/ferricyanide reactions

at platinum and gold electrodes. Part I. Kinetics at bare-metal surfaces. J. Electroanal. Chem.

35,55 - 72 (1972).

(53) Suzuki, J. Hydrodynamic Voltammetry with the Convection Electrode. Bull. Chem. Soc. Jpn.

43, 755 - 758 (1970).

(54) Breckenridge, J. H. & Harris, W. E. Effect of thorium on the reactions of iron(I1) and iron(II1)

at a platinum electrode. 2 - 4 (1970).

(55) Stulikova,M. and Vydra, F., Voltammetry with Disk Electrodes and Its Analytical Applica-

tions., J. Electroanal. Chem., 38, 349 - 357 (1972).

(56) Galus, Z. & Adams, R. N. The Investigation of the Kinetics of Moderately Rapid Electrode

Reactions Using Rotating Disk Electrodes. J. Phys. Chem. 67, 866 - 871 (1963).

(57) Weber, J., Samec, Z. Mareek, V. The effect of anion adsorption on the kinetics of the

Fe3+/Fe2+ reaction on Pt and Au electrodes in HClO4. J. Electroanal. Chem. 89, 271 - 288

(1978).

(58) Randles, J. E. B. and Somerton, K. W. Kinetics of rapid electrode reactions. Discuss. Faraday

Soc. 1, 11 (1947).

(59) Agarwal, H. and Qureshi, S. Faradaic rectification studies of the ferrous-ferric redox couple

in the acid media. 19, 607 - 610 (1974).

(60) Hung, N. G. and Nagy, Z. Kinetics of the Ferrous/Ferric Electrode Reaction in the Absence

of Chloride Catalysis. J. Electrochem. Soc. 134, 2215 (1987).

25

Page 27: A Benchmarking Approach for Routine Determination of Flow

(61) Anson, F. C. The Effect of Surface Oxidation on the Voltammetric Behavior of Platinum

Electrodes. J. Am. Chem. Soc. 81, 1554 - 1557 (1959).

(62) Zhong, S & Kazacos, M. S. Electrochemical behaviour of vanadium (V)/ vanadium (IV)

redox couple at graphite electrodes. J. Power Sources 39, 1 - 9 (1992).

(63) Sum, E., Rychcik, M. & Skyllas-Kazacos, M. Investigation of the V (V)/V (IV) system for

use in the positive half-cell of a redox battery. J. Power Sources 16, 85 - 95 (1985).

(64) Aaron, D. et al. In Situ Kinetics Studies in All-Vanadium Redox Flow Batteries. ECS Elec-

trochem. Lett. 2, A29 - A31 (2013).

(65) Oriji, G., Katayama, Y. & Miura, T. Investigations on V(IV)/V(V) and V(II)/V(III) redox

reactions by various electrochemical methods. J. Power Sources 139, 321 - 324 (2005).

(66) Agar, E., Dennison, C. R., Knehr, K. W. & Kumbur, E. C. Identification of performance

limiting electrode using asymmetric cell configuration in vanadium redox flow batteries. J.

Power Sources 225, 89 - 94 (2013).

(67) Fink, H., Friedl, J. & Stimming, U. Composition of the Electrode Determines Which Half-

Cells Rate Constant is Higher in a Vanadium Flow Battery. J. Phys. Chem. C 120, 15893 -

15901 (2016).

(68) Aaron, D., Tang, Z., Papandrew, A. B. & Zawodzinski, T. A. Polarization curve analysis of

all-vanadium redox flow batteries. J. Appl. Electrochem. 41, 1175 - 1182 (2011).

(69) Garsany, Y., Baturina, O. A., Swider-Lyons, K. E. & Kocha, S. S. Experimental Methods

for Quantifying the Activity of Platinum Electrocatalysts for the Oxygen Reduction Reaction.

Anal. Chem. 82, 6321 - 6328 (2010).

(70) Gasteiger, H. A., Kocha, S. S., Sompalli, B. & Wagner, F. T. Activity benchmarks and re-

quirements for Pt, Pt-alloy, and non-Pt oxygen reduction catalysts for PEMFCs. Appl. Catal.

B Environ. 56, 9 - 35 (2005).

26

Page 28: A Benchmarking Approach for Routine Determination of Flow

(71) McCrory, C. C. L., Jung, S., Peters, J. C. & Jaramillo, T. F. Benchmarking Heterogeneous

Electrocatalysts for the Oxygen Evolution Reaction. J. Am. Chem. Soc. 135, 16977 - 16987

(2013).

(72) McCrory, C. C. L. et al. Benchmarking Hydrogen Evolving Reaction and Oxygen Evolving

Reaction Electrocatalysts for Solar Water Splitting Devices. J. Am. Chem. Soc. 137, 4347 -

4357 (2015).

(73) Bard, A. J.; Faulkner, L. R. Electrochemical Methods:Fundamentals and Applications; Wiley:

New York, (1980).

(74) Scholz, F. Electroanalytical methods: Guide to experiments and applications, 2, rev. and

extended ed., Springer: Berlin, (2010).

(75) Janoschka, T. et al. An aqueous, polymer-based redox-flow battery using non-corrosive, safe,

and low-cost materials. Nature 527, 78 - 81 (2015).

(76) Ding, Y., Zhao, Y., Li, Y., Goodenough, J. B. & Yu, G. A high-performance all-metallocene-

based, non-aqueous redox flow battery. Energy Environ. Sci. 10, 491 - 497 (2017).

(77) Pournaghi-Azar, M. H. & Ojani, R. Electrode kinetic parameters of the ferrocene oxidation

at platinum, gold and glassy carbon electrodes in chloroform. Electrochim. Acta 39, 953 - 955

(1994).

(78) Rosso, K. M., Smith, D. M. A. & Dupuis, M. Aspects of aqueous iron and manganese (II/III)

self-exchange electron transfer reactions. J. Phys. Chem. A 108, 5242 - 5248 (2004).

(79) Brunschwig, B. S., Logan, J., Newton, M. D. & Sutin, N. A Semiclassical Treatment of

Electron-Exchange Reactions. Application to the Hexaaquoiron(II)-Hexaaquoiron(III) Sys-

tem. J. Am. Chem. Soc. 102, 5798 - 5809 (1980).

(80) Silverman, J. & Dodson, R. W. The exchange reaction between the two oxidation states of

iron in acid solution. J. Phys. Chem. 56, 846 - 852 (1952).

27

Page 29: A Benchmarking Approach for Routine Determination of Flow

(81) Jolley, W. H., Stranks, D. R. & Swaddle, T. W. Pressure Effect on the Kinetics of the Hex-

aaquairon(II/III) Self-Exchange Reaction in Aqueous Perchloric Acid. Inorg. Chem. 29, 1948

- 1951 (1990).

(82) Logan, J. & Newton, M. D. Ab initio study of electronic coupling in the aqueous Fe2+ - Fe3+

electron exchange process. J. Chem. Phys. 78, 4086 - 4091 (1983).

(83) Tembe, B. L., Friedman, H. L., Newton, M. D. The theory of the Fe2+ - Fe3+ electron ex-

change in water. J. Chem. Phys. 76, 1490 (1982).

28

Page 31: A Benchmarking Approach for Routine Determination of Flow

Supporting Information:

A Benchmarking Approach for Routine

Determination of Flow Battery Kinetics

Tejal V. Sawant and James R. McKone∗

Department of Chemical and Petroleum Engineering, Swanson School of Engineering, University

of Pittsburgh, Pittsburgh, PA 15261, USA

E-mail: [email protected]

Tabulated kinetics of Fe redox chemistry

Table S1 collects reported values of electron-transfer kinetics for aqueous Fe3+/2+ redox chemistry

using various techniques. The most commonly used measuring technique was RDE voltammetry,

and HClO4 was the electrolyte of choice in most cases. The reported rate constant (k0) varied over

the range from 10−3 to 10−1 cm/s for Pt and 10−5 to 10−2 cm/s for Au. The reported values for α

were generally close to 0.5 or slightly higher, although several studies did not explicitly report an

α value.

1

Page 32: A Benchmarking Approach for Routine Determination of Flow

Table S1: Kinetics of Fe3+/2+ as reported in literature

Electrode Rate constantk0 (cm/s) α

Measuringtechnique

Supportingelectrolyte [Fe] Reference

Pt 11 x 10−3 0.78 RDE 0.1 M HClO4 “few millimolar” Jordan1

Pt 2.4 x 10−3 0.63 RDE 1 M HClO4 10 mM Jahn2

Pt 5 x 10−3 AC impedance 0.1 M HClO4 1 mM Randles3

Pt 10.5 x 10−2 0.54faradaic

rectification1 M HCl 2 mM Agarwal4

Pt 9 x 10−3 0.5 RDE 1 M HClO4 10 mM Angell5

Pt 7.6 x 10−3 0.63 RDE 0.1 M HClO4 0.5 mM Suzuki6

Pt 1.5 x 10−3 RDE 0.5 M HClO4 5 mM Weber7

Pt 10 x 10−3 galvanostaticpulse

0.1 M HClO4 2 mM Anson8

Pt 1 x 10−3 RDE 0.1 M HClO4 10 mM Breckenridge9

Pt 4.3 x 10−3 0.46 RDE 1 M H2SO4 1 mM Galus10

Au 10 x 10−3 0.5 RDE 0.5 M H2SO4 10 mM Angell5

Au 7.9 x 10−3 RDE 0.5 M HClO4 5 mM Weber7

Au 1.2 x 10−5 0.55coulostatic

pulse0.5 M HClO4 10 mM Hung11

Simulations of RFB overpotential performance

RFB charge-discharge curves were modeled using commercial software (DigiElch) to simulate

hydrodynamic voltammetry data, from which overpotential values were extracted as a function of

state of charge (SOC). Fixed parameters for the simulation, which were chosen to approximate the

behavior of a flowing aqueous Fe3+/2+ RFB positive electrolyte, are collected into Table S2.

Electron-transfer rate constants were varied between 10−2 and 10−5 cm/s. Voltammo-

grams were simulated over a range of SOC values by varying the relative ratios of Fe3+ and Fe2+

while keeping the total Fe concentration fixed at 1 M. We then extracted a series of overpotential

values by recording the applied potential required from the simulation at current densities of ±10

mA/cm2, which was slightly less than ∼10 % of the mass-tranport limited current density at 0

% SOC. Figure S1 depicts a subset of the simulated RDE data. The points at which each of the

simulated polarization curves intersected the line at J = ±10 mA/cm2 were taken as the simulated

overpotential values at the corresponding SOC. The point at which each curve intersected J = 0

2

Page 33: A Benchmarking Approach for Routine Determination of Flow

Table S2: Parameters used for simulation on DigiElch software

Parameter Valuek0 1 × 10−2 and 1 × 10−5 cm/sα 0.5[Fe2+] + [Fe3+] 1ME0 0 V

Estart-0.5 V for k0=1 x 10−2 cm/s-1 V for k0=1 x 10−5 cm/s

Eend0.5 V for k0=1 x 10−2 cm/s1 V for k0=1 x 10−5 cm/s

scan rate 0.001 V/selectrode radius 0.25 cmdiffusivity 3 x 10−6 cm2/spre-equilibrium disabledkinematic viscosity 0.01 cm2/sseries resistance 4 Ω

Figure S1: Simulated rotating disc electrode data for k0 values of (a) 10−2 cm/s and (b) 10−5 cm/s.Horizontal demarcations correspond to the simulated RFB operating current density values duringcharge and discharge.Dotted lines represent the operating cathodic and anodic current densities

was the equilibrium potential of the system at that SOC.

Determination of purge time

We explicitly measured the time required to remove atmospheric oxygen from the RDE electro-

chemical cell by monitoring the voltammetric response of a Pt electrode for evidence of dissolved

3

Page 34: A Benchmarking Approach for Routine Determination of Flow

O2(aq). During the N2 purge, we continuously ran cyclic voltammograms at 200 mV/s in the po-

tential range from -0.25 V to 1.25 V versus Ag/AgCl. The data in Figure S2 shows the results of

these voltammetry experiments, wherein we observed characteristic Pt surface features as well as

a progressively diminishing negative shift in the voltammetric profile due to steady-state oxygen

reduction at negative applied potentials. This oxygen reduction response was found to decrease

asymptotically over the span of only ∼20 CV cycles. We took this as a clear indication that the

solution was fully purged of O2 after ∼5 minutes of purge time. Note that this timescale likely

depends on several factors, including total cell volume, purge gas flow rate, the specifics of solu-

tion agitation, and even the size of the purge tube opening (since it influences gas bubble size in

the cell). Nevertheless, this purge test can be easily replicated any time an electrochemical cell

configuration is created or modified.

Figure S2: Cyclic voltammogram of Platinum for the determination of purge time in 0.5M H2SO4at a scan rate of 200mV/s depicting the upward shift in CV after nitrogen purge for 20 cycles

Further evidence for the importance of O2 removal is depicted in Figure S3, which shows

representative cyclic voltammograms for Pt and Au electrodes in low-concentration Fe RFB elec-

trolyte before and after N2 purging. Interestingly, the voltammetric features vary only slightly for

the Pt electrode, whereas Au only shows clear peaks after the purge is complete.

4

Page 35: A Benchmarking Approach for Routine Determination of Flow

Figure S3: Cyclic voltammograms of (a) Pt and (b) Au electrodes to show influence of N2 purgein aqueous electrolyte containing 5mM FeCl2, 5mM FeCl3, and 0.5 M HCl.

Voltammetric scan rate dependence

We carried out cyclic voltammetry to observe the scan rate dependence of Fe3+/2+ redox chemistry

using Pt and Au electrodes, as shown in Figure S4. The scan rates were varied from 10–200 mV/s

and the observed voltammetric peak-to-peak separations (∆Ep) are collected into Table S3. The

∆Ep values were found to range from 60-150 mV, which indicates relatively fast quasi-reversible

electron transfer. Figure S5 further plots the observed peak current densities as a function of square

root of scan rate, which yields a clear linear relationship as expected from a diffusional process.

5

Page 36: A Benchmarking Approach for Routine Determination of Flow

Figure S4: Cyclic voltammograms of (a) Pt and (b) Au electrodes depicting scan rate dependencein low-concentration Fe RFB electrolyte containing 5mM FeCl2, 5mM FeCl3, and 0.5 M HCl.

Table S3: ∆Ep values for 5mM FeCl2 and 5mM FeCl3 in 0.5 M HCl at Pt and Au electrodes

Scan rate(mV/s) ∆Ep,Pt ∆Ep,Au10 60 11020 70 11550 75 135100 80 145200 90 150

Figure S5: Peak current density as a function of square root of scan rate for (a) Pt and (b) Auelectrodes in low-concentration Fe RFB electrolyte containing 5mM FeCl2, 5mM FeCl3, and 0.5M HCl.

6

Page 37: A Benchmarking Approach for Routine Determination of Flow

Evaluation of residual electrode surface roughness

Electrochemical kinetics depend in part on the morphology and surface roughness of the associated

electrodes. Our electrode preparation methods involve abrasive polishing, which yields microscop-

ically smooth electrode surfaces that nonetheless still exhibit polishing damage that manifests as

surface roughness on the nanoscale. For example, Figure S6(a) shows a scanning electron mi-

croscopy (SEM) results for a Pt electrode after polishing with 5, 1, and 0.05 µm alumina slurries;

where submicron scratches clearly remain on the surface.

Figure S6: Scanning electron micrograph of Pt electrode (a)after sequential polishing with 5, 1,and 0.05 µm alumina slurries and (b)after sonicating in water for 30 mins followed by annealingin a Bunsen flame

Thus, we made measurements of electrochemically active surface area, ECSA, following

the protocol reported by Garsany for Pt in fuel cell catalysts.12 Briefly, Pt surface voltammograms

were collected in 0.5 M H2SO4 using the electrode preparation and cell setup identical to that

described in the main text. ECSA was calculated from the hydrogen adsorption charge(QH) ac-

cording to the following equation:

ECSA[cm2] =QH [µC]

210[µCcm−2](1)

7

Page 38: A Benchmarking Approach for Routine Determination of Flow

whereQH is the hydrogen adsorption charge, extracted from the area under the voltammetric curve

from 0.012 to -0.279 V versus Ag/AgCl in the negative scan direction as shown in Figure S7.

Figure S7: Platinum cyclic voltammogram in 0.5M H2SO4 representing the area under the curvefor calculating QH

Background double-layer capacitance was removed by subtracting the observed cathodic

current at 0.012 V versus Ag/AgCl from the total current. Figure S7 shows a representative Pt CV,

with the area used to calculate QH highlighted.

Using a variety of polishing protocols, we consistently found the ratio of ECSA to geo-

metric area of the electrode to fall in the range of 3 to 4. We further sonicated the Pt electrode for

30 mins in water and then heated it for a few seconds in a Bunsen flame in an effort to anneal out

polishing damage. Using this technique, we obtained the ratio of ECSA to geometric surface area

to be approximately 1.5. Figure S6(b) shows the scanning electron micrograph of this electrode.

Interestingly, evidence for scratches remains along with the emergence of grain structure contrast

and nanoscale pits. Thus, we conclude that polishing damage manifests as roughness on at least

two size scales: relatively large submicron scratches and further nano or molecular scale roughness

that cannot be readily imaged by SEM but is nevertheless responsible for most of the increase in

8

Page 39: A Benchmarking Approach for Routine Determination of Flow

surface area. Annealing can be used to heal the smaller-scale polishing damage; indeed, this is

common in single crystal studies,13 but this approach may not yield the same results for all elec-

trode materials. Thus, we chose to use only polishing and electrochemical cleaning protocols as

the best compromise between residual roughness, ease of implementation, and replicability across

a variety of electrode materials.

Results of conventional Butler-Volmer fits

Figure S8 represents nonlinear least squares fits of the same kinetics data as in Figure 6 in the main

text while instead constraining the fit to a conventional Butler-Volmer model where anodic and

cathodic symmetry factors sum to 1. The result was a clear, systematic deviation in the fits from the

empirical data where the slopes of the fit lines are too steep in all cases (i.e., symmetry factors are

too large). This deviation, along with the improved fit if the symmetry factors are unconstrained,

is consistent with a reaction mechanism involving both electrochemical and chemical steps. Thus

it is apparent that interfacial catalysis is operative for Fe3+/2+ redox chemistry even though it only

involves the transfer of one electron.

Figure S8: Transport free polarization data for (a) Pt and (b) Au electrodes in low-concentration FeRFB electrolyte containing 5mM FeCl2, 5mM FeCl3, and 0.5 M HCl. The black lines correspond tofits using the Butler-Volmer equation where the symmetry factors were constrained to αox+αred =1

9

Page 40: A Benchmarking Approach for Routine Determination of Flow

References

(1) Jordan, J. & Javick, R. Electrode Kinetics By Hydrodynamic Voltammetry-Study of Ferrous-

Ferric , and Iodide-Iodine Systems. Electrochim. Acta 6, 23 - 33 (1962).

(2) Jahn, D. & Vielstich, W. Rates of Electrode Processes by the Rotating Disk Method. J. Elec-

trochem. Soc. 109, 849 (1962).

(3) Randles, J. E. B. and Somerton, K. W. Kinetics of rapid electrode reactions. Discuss. Faraday

Soc. 1, 11 (1947).

(4) Agarwal, H. and Qureshi, S. Faradaic rectification studies of the ferrous-ferric redox couple in

the acid media. 19, 607 - 610 (1974).

(5) Angell, D. H. Dickinson, T. The kinetics of the ferrous/ferric and ferro/ferricyanide reactions

at platinum and gold electrodes. Part I. Kinetics at bare-metal surfaces. J. Electroanal. Chem.

35,55 - 72 (1972).

(6) Suzuki, J. Hydrodynamic Voltammetry with the Convection Electrode. Bull. Chem. Soc. Jpn.

43, 755 - 758 (1970).

(7) Weber, J., Samec, Z. Mareek, V. The effect of anion adsorption on the kinetics of the Fe3+/Fe2+

reaction on Pt and Au electrodes in HClO4. J. Electroanal. Chem. 89, 271 - 288 (1978).

(8) Anson, F. C. The Effect of Surface Oxidation on the Voltammetric Behavior of Platinum Elec-

trodes. J. Am. Chem. Soc. 81, 1554 - 1557 (1959).

(9) Breckenridge, J. H. & Harris, W. E. Effect of thorium on the reactions of iron(I1) and iron(II1)

at a platinum electrode. 2 - 4 (1970).

(10) Galus, Z. & Adams, R. N. The Investigation of the Kinetics of Moderately Rapid Electrode

Reactions Using Rotating Disk Electrodes. J. Phys. Chem. 67, 866 - 871 (1963).

10

Page 41: A Benchmarking Approach for Routine Determination of Flow

(11) Hung, N. G. and Nagy, Z. Kinetics of the Ferrous/Ferric Electrode Reaction in the Absence

of Chloride Catalysis. J. Electrochem. Soc. 134, 2215 (1987).

(12) Garsany, Y., Baturina, O. A., Swider-Lyons, K. E. & Kocha, S. S. Experimental Methods

for Quantifying the Activity of Platinum Electrocatalysts for the Oxygen Reduction Reaction.

Anal. Chem. 82, 6321 - 6328 (2010).

(13) Clavilier, J., Faure, R., Guinet, G. & Durand, R. Preparation of monocrystalline Pt micro-

electrodes and electrochemical study of the plane surfaces cut in the direction of the 111 and

110 planes. J. Electroanal. Chem. 107, 205 - 209 (1979).

11