INVESTIGATION OF MATERIALS FOR
THE ELECTROCHEMICAL DECOMPOSITION OF AMMONIA
By
Daniel J. Little

A DISSERTATION
Submitted to
Michigan State University

in partial fulfillment of the requirements
for the degree of
Chemistry–Doctor of Philosophy

ABSTRACT
INVESTIGATION OF MATERIALS FOR
THE ELECTROCHEMICAL DECOMPOSITION OF AMMONIA
By
Daniel J. Little
The ability to store renewable energy in the form of chemical bonds in fuels is
essential for the realization of a clean energy future. There have been decades of work
on the subject of using renewable energy for the electrochemical splitting of water
molecules to form hydrogen gas (H2) as a fuel. However, development of infrastructure
for the efficient transportation and storage of large quantities of the low-density H2
remains as a major challenge preventing a commercial disruption. Instead of reducing
protons (H+) from water to produce H2, therefore, an alternative solution would be to
couple the oxidation of water to the reduction of nitrogen gas (N2) from the air and produce
ammonia (NH3). NH3 is a higher-density fuel for which the transportation and storage
infrastructure is already in place at a national industrial scale.
There is plenty of ongoing research into the synthesis of NH3 from renewable N2
and H2, but there have been relatively few studies into efficiently splitting it back apart to
recover the H2 fuel. Furthermore, what work has been done in NH3 splitting has largely
been in aqueous conditions, which are corrosive to the transportation and storage
infrastructure and involve an inherent loss of energy density compared to liquid ammonia
(NH3(l)). In this work, we investigate the half reactions involved the electrochemical
splitting of anhydrous NH3(l), and propose new Earth-abundant anode and cathode
materials to replace the traditionally used noble metals.

For all who have supported me along
the pathway of academic achievement.

iii

ACKNOWLEDGEMENTS

First, I would like to thank my wife, Kayli, my parents, Jeannine and John, my
sister, Sarah, and the rest of my family for their love and support through my academic
career. Without them I would have never gotten to where I am today, quite literally in the
case of my parents. Their willingness to put up with my eccentricities throughout my
entire life has been the catalyst for the development of my scientific career.
I would like to acknowledge all of the educators–both in science and otherwise–
who have inspired me, bared with me, and otherwise facilitated my career in chemistry.
The lessons I learned from them will never be forgotten, though I am going to continue
my pursuit to be a science educator whether they like it or not.
I owe a lot of gratitude to Ben Levine, and the members of his research group
during the summer and fall of 2012: Garrett Meek, Yinan Shu, and Scott Fales for their
willingness to introduce me to theoretical chemistry and continue to support me even
after I discovered it was not my area of interest. They have all been excellent friends
and colleagues, willing to answer the many questions I posed to them during my tenure
as a graduate student. I owe my excitement and willingness to use computer
programming to assist my chemical experimentation to them, which likely saved me
months of time over these five years.
Credit is due to all of the “Ammoniaks”–members of the ammonia fuel
collaboration–especially Jason Thornton, Arianna Savini, Dillon Edwards, Faezeh Habib
Zadeh, and of course Tom Hamann and Mitch Smith. I could make an extensive list of

iv

things that I learned from working with each one of them–though I will not do so here for
the convenience of the reader. Still, they, and the other collaborators, helped build me
from a first-year graduate student who did not understand which electrode was which,
into the scientist I am now.
Of course I must thank all of the members of the Hamann lab, past and present,
for accepting me into their group, putting up with my slightly obsessive tendencies, and
always being there to talk about both chemistry, and things completely unrelated to
chemistry. Obviously I learned much from those ahead of me in the Ph.D. program, but I
actually learned just as much from teaching things to the newer members and
answering their questions. I am a better scientist today having worked in such a
supportive research group, and I plan to take group dynamic and mechanic ideas into
my future career.
I need to acknowledge the constructive, collaborative mentality of all of the other
students, postdocs, and faculty of the fourth floor. The ability to freely discuss research–
in addition to non-research–with others in the department is critical to the success of the
entire department. On that note, I think it is important to acknowledge all of the women
and men who make the Michigan State Department of Chemistry work–from
administrators, to support staff, to those who keep the building clean, and those who fix
the elevators all too frequently. They do not receive enough credit for what they do, and
academic research would be nearly impossible without their hard work.
For acknowledgements specific to the experiments within, I need to first thank Dr.
Eli Fahrenkrug and the Electron Microbeam Analysis Laboratory at the University of

v

Michigan for assistance with the Auger Electron Spectroscopy measurements. Also, I
would like to thank Dr. Kai Sun and the Michigan Center for Materials Characterization
for assistance with and use of the Kratos XPS, which is supported by the University of
Michigan College of Engineering and NSF grant #DMR-0420785. I would also like to
specifically thank Hamed Hajibabaei for assistance with the SEM at the Composite
Materials and Structures Center, and Dr. Kathryn Severin for use of the ICP-OES in the
AP Labs of the Michigan State University Department of Chemistry. All of the work
presented in this dissertation was supported financially by Michigan State University.

vi

PREFACE

I jumped at the ammonia project back in the fall of 2012 because it was clear that
in the worst-case scenario, I would be engaged for five years and learn all of the skills
needed to be a successful, independent researcher, and in the best-case scenario I
would change the world.
That sounds dramatic, but it is definitely true. I believe that in order to make my
research as a chemist truly meaningful, I should work on projects that are both
extremely innovative, and also attempt to solve the most important problems in the
world we live in. As I see it, the problem-of-choice is giving access to clean energy for
everyone on Earth, consequently helping to solve problems of politics, economics, and
the environment all at once. Thus, when starting my graduate work in chemistry, I was
drawn to Professor Tom Hamann’s research group to study photoelectrochemistry: the
conversion of sunlight into useful electricity or fuels.
When interviewing with Professor Hamann, however, I was introduced to a new
project, on which I would be the first graduate student, that offered me the potential to
do ground-breaking research in a virtually unexplored field: liquid ammonia electrolysis.
The idea was to use ammonia as a fuel in place of–or at least as a transportation and
storage medium for–renewable hydrogen gas. There are likely hundreds if not
thousands of research groups worldwide–both academic and industrial–studying the
renewable formation of hydrogen from water splitting, but this project actually sought to
answer the question, “Then what?” The public already has apprehension about moving

vii

away from traditional hydrocarbon fuels, and design and implementation of new
hydrogen infrastructure will be expensive and complicated.
So I jumped on board into a new research project where it was completely
unclear what I would learn, what I would do, and whether or not what I was attempting
to do would actually work. I owe a lot of credit to Professor Hamann, Professor Mitch
Smith, and coworkers like Dr. Jason Thornton, Dr. Arianna Savini, and Faezeh Habib
Zadeh and other collaborators on “the ammonia project” who helped me learn all of the
information in the chapters that follow. While I know my research by itself is not exactly
world-changing, it opens the door for future students on the ammonia project
collaboration, and our combined work can come together as something spectacular.

viii

TABLE OF CONTENTS

LIST OF TABLES ................................................................................................. xi
LIST OF FIGURES ............................................................................................... xii
KEY TO ABBREVIATIONS ............................................................................... xviii
CHAPTER 1: THE POTENTIAL ROLE OF AMMONIA
IN A RENEWABLE ENERGY FUTURE ........................................................... 1
Energy and the Global Climate ........................................................ 2
Ammonia as an Energy Carrier ....................................................... 4
The State of Ammonia Electrolysis Research ................................. 8
APPENDICES ................................................................................ 15
Appendix 1.1: Calculation of Ecell for the Electrolysis
of Liquid Ammonia .............................................................. 16
Appendix 1.2: Background Theory for 2-Electrode
and 3-Electrode Measurements .......................................... 18
Appendix 1.3: Discussion of the Term “Onset Potential” .......... 21
REFERENCES .............................................................................. 23
CHAPTER 2: EXPERIMENTAL DESIGN FOR THE ELECTROCHEMICAL
INVESTIGATION OF ANHYDROUS AMMONIA ........................................... 27
Electrochemical Experimentation in Stock Liquid Ammonia .......... 28
Distillation of Liquid Ammonia for Electrochemistry in
Rigorously Dry Conditions ........................................................ 31
Evaluation of Electrochemical Headspace
Via Gas Chromatography ......................................................... 34
Elemental Analysis of the Electrochemical Cell Contents
Via ICP-OES............................................................................. 36
Voltammetry in N,N-Dimethylformamide........................................ 37
Sampled Time Voltammetry........................................................... 38
Fabrication of Working Electrodes from
Transition Metal Wires .............................................................. 39
Determination of the Electrochemically Active Electrode
Surface Area ............................................................................ 40
Instrumentation .............................................................................. 42
REFERENCES .............................................................................. 43

ix

CHAPTER 3: IDENTIFYING THE PROBLEMS WITH LIQUID AMMONIA
ELECTROLYSIS ON PLATINUM ELECTRODES ......................................... 45
Introduction .................................................................................... 46
Experimental .................................................................................. 47
Results and Discussion ................................................................. 47
Voltammetry in Liquid Ammonia ......................................... 47
Voltammetry in N,N-Dimethylformamide ............................. 54
Conclusions ................................................................................... 59
REFERENCES .............................................................................. 62
CHAPTER 4: INVESTIGATION OF IRON NITRIDE AS AN
EARTH-ABUNDANT ELECTROCATALYST FOR
THE OXIDATION OF LIQUID AMMONIA ...................................................... 64
Introduction .................................................................................... 65
Experimental .................................................................................. 67
Results and Discussion ................................................................. 68
Conclusions ................................................................................... 83
REFERENCES .............................................................................. 85
CHAPTER 5: EVALUATION OF AMORPHOUS COBALT PHOSPHIDE
AS AN EARTH-ABUNDANT AMMONIUM
REDUCTION ELECTROCATALYST ............................................................. 88
Introduction .................................................................................... 89
Experimental .................................................................................. 91
Results and Discussion ................................................................. 92
Conclusions ................................................................................. 103
REFERENCES ............................................................................ 105
CHAPTER 6: CONCLUSIONS AND FUTURE WORK...................................... 107
Introduction .................................................................................. 108
Ammonia Oxidation Catalysts ...................................................... 109
Ammonium Reduction Catalysts .................................................. 111
REFERENCES ............................................................................ 114

x

LIST OF TABLES

Table 1.1: Energy density comparison of various fuels and
commercial lithium ion batteries.12–14 ................................................. 5
Table 5.1: Average electrochemical performance of fifteen similar CoPx
electrodes compared to a Pt disk in 0.1 M NH4NO3 in NH3(l)......... 100

xi

LIST OF FIGURES

Figure 1.1: An ammonia fuel cycle involving the reduction of N2 coupled
with H2O oxidation (lower left), the splitting of NH3 to N2 and
H2 (top), and the oxidation of H2 as a fuel (lower right)
(Reproduced from Little et al.)23 ........................................................ 7
Figure 1.2: Electrode configuration diagrams for a 2-electrode (a) and
3-electrode (b) experiment. In the 3-electrode case, the working,
reference, and counter electrodes are labeled W.E., R.E., and
C.E. respectively. In both cases, the red “VA” represents where
a potential is applied, the green “VM” represents where a
potential is measured, and the blue “IM” represents where the
current is measured ........................................................................ 18
Figure 1.3: Illustration of how the onset potential would be calculated from
a sample curve ................................................................................ 22
Figure 2.1: Diagram of the electrochemical cell used for voltammetry
in NH3(l) as well as in DMF with dissolved NH3(g) .......................... 29
Figure 2.2: Diagram of the custom Ag/AgNO3 reference electrode
for use in NH3(l) solutions................................................................ 30
Figure 2.3: Powder XRD spectrum of KNH2 synthesized from KH and liquid
NH3. A reference spectrum for KNH2 is shown in the red, dotted
line. The large amorphous feature at 20° is due to the sample
holder, which included a dome to maintain a nitrogen
atmosphere for the powder. (Reproduced from Little et al.)4 .......... 31
Figure 2.4: Diagram of the NH3(l) distillation flask including two glass valves
for the gas inlet and outlet, as well as a black, plastic screw-cap
containing a white Teflon septum. ................................................... 32

xii

Figure 3.1: The electrolysis of NH3(l) using platinum disk electrodes. Panes
(a) and (b) show successive cyclic voltammograms with 1.0 M
NH4PF6 and 1.0 M KPF6 as electrolytes, respectively, while pane
(c) shows STVs for both conditions. The STV for 1.0 M KNH2 is
also depicted in pane (c) ................................................................. 48
Figure 3.2: AES spectra of a Pt electrode cleaned electrochemically in
sulfuric acid (a), and of a similar electrode cleaned
electrochemically and then anodically poisoned in 1.0 M
NH4PF6 in liquid ammonia (b). Sputtering with Ar+ allowed for
depth-profiling ................................................................................. 49
Figure 3.3: (a) STVs of NH3(l) comparing current densities at varying potential
using three different concentrations of either NH4PF6 (triangles)
or KPF6 (circles) as the supporting electrolyte, as well as 1.0 M
KNH2 (black squares). For all STVs, the markers represent actual
points of data, and there are straight lines connecting them for
trend clarity. (b) CVs in NH3(l) with 1 M NH4PF6 electrolyte (red)
and 1 M KPF6 electrolyte (green) with a Pt disk working electrode
at 100 mV/s ..................................................................................... 50
Figure 3.4: Cathodic CVs in NH3(l) of 1.0 M KPF6 (green) and 1.0 M NH4PF6
(black). Also included are STVs measured at 1 and 100 ms in
1.0 M NH4PF6 (dark red and light red triangles respectively), and
at the same times in 1.0 M KPF6 (dark blue and light blue circles
respectively) .................................................................................... 52
Figure 3.5: (a) STVs showing electrolysis of NH3 dissolved in DMF using
platinum disks as both the anode and the cathode. The red curve
has 0.1 M NH4PF6, while the green curve has 0.1 M TBAPF6.
(b) CV (red) and STV (blue) of the anodic current with NH3 dissolved
in 0.1 M NH4PF6 in DMF. In black is the CV without dissolved NH3.
In green is the CV with NH3 dissolved in 0.1 M TBAPF6. (c) CVs
(black and green with NH4PF6 and TBAPF6 respectively) and
STVs of the cathodic current with NH3 dissolved in DMF. The black
and light and dark red curves contain 0.1 M NH4PF6 as the
supporting electrolyte, while the light and dark blue curves
contain 0.1 M TBAPF6 ..................................................................... 55

xiii

Figure 3.6: (a) Quasi-reversible redox wave for NH4+ in DMF. (b) Cathodic
waves measured in 0.1 M NH4PF6 with increasing concentrations
of HClO4 using a Pt disk electrode in DMF ..................................... 57
Figure 3.7: Tafel plot of the CV shown in black in Figure 3.5(c). The value of
! close to 0.5 indicates a likely 1-electron reaction, quasireversible reaction ........................................................................... 58
Figure 3.8: Diagram showing where improvements can be made by either
avoiding electrode poisoning (indicated by arrow with “poison”
icon), or via catalysis (indicated by arrows with “catalysis” ions).
The formal potentials for NH4+ reduction, H+ reduction, and NH3
oxidation are indicated on the top black scale. Both the STV (red)
and CV (green) were recorded using a Pt disk electrode in 0.1 M
NH4PF6 in DMF ............................................................................... 60
Figure 4.1: (a) CV of a Pt disk electrode in distilled NH3(l) (black), and five
successive CVs of a Pt disk electrode in distilled NH3(l) after the
addition of degassed Millipore water to make the solution 10 %
water (colors). (b) CVs of an Fe disk electrode in the same
conditions as (a), except that only 1 % water was added. The
vertical black arrows indicate the order of successive anodic
scans. Dashed boxes indicate the overpotential required to reach
10 mA cm-2 current. Small, black, diagonal arrows indicate the
scan direction. All curves were measured at 100 mV s-1 with
0.1M KPF6 as the supporting electrolyte ......................................... 69
Figure 4.2: Successive CVs of distilled NH3(l) on an Fe disk electrode with
the order of the scans indicated by the vertical black arrow. All
scans were measured at 100 mV s-1 with 0.1 M KPF6 as
supporting electrolyte ...................................................................... 71
Figure 4.3: Chronoamperograms conducted on an Fe disk at seven different
potentials near the onset of the anodic wave shown in Figure 4.2.
Potentials here are reported versus NHE, and verified using
electron solvation between experiments. All scans used a
0.1 M KPF6 electrolyte. Periodic sudden jumps in the current seen
at the highest four potentials are due to the effervescence of

xiv

evolved gas bubbles, thus suddenly changing the exposed
surface area. ................................................................................... 72

Figure 4.4: Current efficiencies for the dissolution of an Fe disk electrode in
distilled NH3(l) as a function of the net charge passed at the
working electrode during the experiment. Potentials of 1.2, 1.4,
or 1.6 V versus NHE were applied for varying amounts of time
and the current was integrated to obtain the total charge passed.
For all points, 0.1M NH4NO3 was used as the supporting
electrolyte to avoid the presence of fluorine from the PF6- anion .... 73
Figure 4.5: N2 and H2 gas measured by GC from the headspace of the
electrochemical cell. Measurements were taken every 30 minutes
with a potential of 1.35 V versus NHE applied during the intervals.
Molar amounts approximately correspond to the gas present in
the headspace of the cell. The supporting electrolyte was 0.1 M
NH4NO3. (Inset) The H2:N2 evolution ratio over the course of
the experiment ................................................................................ 76
Figure 4.6: Anodic current of the Pt disk electrode in distilled NH3(l) before
and after dissolution of Fe electrode. Five essentially overlapping
scans are depicted .......................................................................... 77
Figure 4.7: SEM images of Fe foil after use as an anode in 0.1 M NH4NO3
in dry NH3(l) ..................................................................................... 78
Figure 4.8: The surface area of an Fe disk electrode as a function of the
number of wide cathodic scans measured. The surface area was
measured from CV capacitance between each of the wide scans. . 78
Figure 4.9: Photos of custom Fe disk electrodes taken before (a) and after
(b) electrolysis showing the formation of a black film covering the
electrode surface ............................................................................. 79
Figure 4.10: XPS of the Nitrogen 1s region (a), the Fe 2p region (b), and the
Oxygen 1s region (c) of an in situ modified Fe electrode surface ... 80

xv

Figure 4.11: Raman spectrum of in situ modified iron anode after exposure to
the air on the lab bench overnight ................................................... 81

Figure 4.12: Steady-state sampled time voltammograms (STVs) of anodic
behavior of Pt and in situ modified Fe in distilled NH3(l). All points
are sampled from 180 seconds into a chronoamperogram at the
respective potential and are connected with straight lines .............. 82
Figure 5.1: SEM images of the Cu disk substrate, (a) and (b), and CoPx disk
electrode corresponding to the curves shown in Figures 5.3, 5.8,
and 5.9, (c) and (d) .......................................................................... 93
Figure 5.2: SEM images of a CoPx film deposited on a Cu disk substrate by
holding -1.1 V versus Ag/AgCl rather than application of the
pulsed deposition method. .............................................................. 94
Figure 5.3: Cathodic J-V curves in 0.1 M NH4NO3 in NH3(l) with a Pt disk
(black) or CoPx disk (blue). Linear sweeps were measured at
100 mV s-1, and current densities were calculated using the
electrochemically active surface area ............................................. 95
Figure 5.4: Composite of J-V curves in 0.1 M NH4NO3 in NH3(l) from fifteen
CoPx electrodes fabricated by the pulsed deposition method ......... 96
Figure 5.5: Comparison of the µmoles of H2 measured in the headspace of
the electrochemical cell (red diamonds) with the theoretical
maximum amount of H2 if the Faradaic efficiency for H2 evolution
were 100 % (blue circles) ................................................................ 97
Figure 5.6: First and second cathodic cyclic voltammograms of a CoPx
electrode in 0.1 M NH4NO3 in NH3(l) showing the large cathodic
wave only evident on the first scan of each CoPx electrode,
indicated by the black arrow ............................................................ 98
Figure 5.7: Cyclic voltammogram of a CoPx electrode in 0.1 M NH4NO3 in
NH3(l) showing hysteresis due to H2 bubble formation. The scan
direction is indicated by the black arrows. ....................................... 99

xvi

Figure 5.8: Chronoamperogram of the same CoPx electrode shown in
Figure 5.3 measured at -1.67 V versus NHE in 0.1 M NH4NO3
in NH3(l) ......................................................................................... 100
Figure 5.9: Cathodic Tafel plots corresponding to the J-V curves depicted in
Figure 5.3 in 0.1 M NH4NO3 in NH3(l) with a Pt disk (black) or
CoPx disk (blue). Tafel fits are shown in dashed red lines ............ 101
Figure 5.10: Cathodic J-V curves in 0.1 M NH4PF6 in DMF with a Pt disk (black)
or CoPx disk (blue). Linear sweeps were measured at 100 mV s-1,
and current densities were calculated using the electrochemically
active surface area ........................................................................ 102
Figure 5.11: Cathodic Tafel plots corresponding to the J-V curves depicted in
Figure 5.10 in 0.1 M NH4PF6 in DMF with a Pt disk (black) or
CoPx disk (blue). Tafel fits are shown in dashed red lines ............ 103
Figure 6.1: The electrolysis of NH3(l) with a 0.1 M NH4NO3 electrolyte using
FeNx and CoPx as the anode and cathode respectively
(Spartan green), compared to two Pt electrodes (red). ................. 108

xvii

KEY TO ABBREVIATIONS

AES

Auger Electron Spectroscopy

Ag

Elemental silver

Ag+

Silver(I) cation

AgCl

Silver chloride

AgNO3

Silver nitrate

Ar

Elemental argon

Ar+

Argon ion

C3 H8

Propane

C8H18

Octane

CH4

Methane

cm

Centimeters (1 x 10-2 meters)

CO2

Carbon dioxide

CoP

Cobalt phosphide

CoPx

Amorphous cobalt phosphide

Cu

Elemental copper

CV

Cyclic Voltammogram, or Cyclic Voltammetry

DMF

N,N-Dimethylformamide

e–

Electron

xviii

Erxn

The thermodynamic minimum cell potential needed to drive a reaction

EDS

Energy Dispersive X-ray Spectroscopy

F

Faraday constant

Fe

Elemental iron

[Fe(NH3)6][Cl2]

Iron(II) hexamine chloride complex

Fe4N

Iron nitride

FeNx

Amorphous iron nitride

GDP

Gross Domestic Product

GT

Gigatons (1 x 109 metric tons)

H

Elemental hydrogen

H+

Free proton

H•

Hydrogen atom

H2

Molecular hydrogen

H2(g)

Molecular hydrogen in the gas state

H2 O

Water

HClO4

Perchloric acid

He

Elemental Helium

Ir

Elemental iridium

kg

Kilogram (1 x 103 grams)

xix

KH

Potassium hydride

KPF6

Potassium hexafluorophosphate

kW

Kiloohms (1 x 103 ohms)

L

Liters

Li

Elemental lithium

mA

Milliamps (1 x 10-3 amps)

MJ

Megajoules (1 x 106 joules)

n

Integer number of electrons

N

Elemental nitrogen

N(ad)

Nitrido “poison” species adsorbed on an electrode surface

N2

Molecular nitrogen

N2(g)

Molecular nitrogen in the gas state

N2Hx+y,(ad)

Hydrazido species adsorbed on an electrode surface (x,y = 1 or 2)

NxHy,(ad)

Unspecified ammonia oxidation intermediate (x,y = 1 or 2)

NxHy-1,(ad)

Unspecified ammonia oxidation intermediate (x,y = 1 or 2)

NxH–y-1,(ad)

Unspecified ammonia oxidation intermediate (x,y = 1 or 2)

NH(ad)

Imido species adsorbed on an electrode surface

NH2,(ad)

Amido species adsorbed on an electrode surface

NH2–

Amide anion

NH3

Ammonia
xx

NH3,(ad)

Ammonia molecule adsorbed on an electrode surface

NH3(g)

Ammonia in the gas state

NH3(l)

Ammonia in the liquid state

NH3(aq)

Saturated solution of ammonia in water

NH4+

Ammonium ion

NH40•

Neutral, reduced ammonium Rydberg radical species

NH4Br

Ammonium bromide

NH4Cl

Ammonium chloride

NH4I

Ammonium iodide

NH4NO3

Ammonium nitrate

NH4PF6

Ammonium hexafluorophosphate

NHE

Normal Hydrogen Electrode

Ni

Elemental nickel

NO3–

Nitrate anion

NOx

Unspecified nitrogen oxide molecule

M

Molar (moles solute divided by liters solvent)

mM

Millimolar (millimoles of solute divided by liters of solvent)

mV

Millivolts (1 x 10-3 volts)

O

Elemental oxygen

xxi

O2

Molecular oxygen

OH–(ad)

Hydroxide species adsorbed on an electrode surface

PF6–

Hexafluorophosphate anion

pH

Negative base ten logarithm of the molar concentration of protons

PH3

Phosphine

Pt

Elemental platinum

PtCl4•xNH3

Ammoniated platinum(IV) chloride complex

PV

Photovoltaic

Rh

Elemental rhodium

s

Seconds

S°

Standard entropy of a chemical species

SCE

Standard Calomel Electrode

STP

Standard Temperature and Pressure (25 degrees Celsius
and 1 atmosphere)

STV

Sampled Time Voltammogram, or Sampled Time Voltammetry

T

Temperature

TBAPF6

Tetrabutylammonium hexafluorophosphate

TCD

Thermal Conductivity Detector

TW

Terra-watts (1 x 1012 watts)

V

Volts

xxii

XPS

X-ray Photoelectron Spectroscopy

XRD

(Thin film) X-ray Diffraction

a

Butler-Volmer reaction barrier symmetry coefficient

°C

Degrees Celsius

DG°f

Gibbs free energy for the formation of a chemical species under
standard conditions
Change in Gibbs free energy for a chemical reaction

DGrxn
nK

DGrxn

Change in Gibbs free energy for a chemical reaction at a specified
temperature “n” Kelvin

DH°f

Change in enthalpy for the formation of a chemical species under
standard conditions

DS°f

Change in entropy for the formation of a chemical species under
standard conditions

µF

Microfarads (1 x 10-6 farads)

xxiii

CHAPTER 1:
THE POTENTIAL ROLE OF AMMONIA IN A RENEWABLE ENERGY FUTURE

1

Energy and the Global Climate
Carbon dioxide (CO2) acts as a prominent greenhouse gas in the Earth’s
atmosphere due to its high transparency to the incident solar visible radiation, and a
high absorptivity for infrared radiation emitted from the surface of the Earth.1 According
to data from the Vostok ice core, the CO2 concentration in the atmosphere has been in
equilibrium between 200 and 300 ppm for the past 420,000 years.2 However, due to
what are almost certainly anthropogenic causes, the CO2 concentration has been
increasing exponentially since the industrial revolution to over 400 ppm today, and the
rate of increase is not slowing.3 Climate scientists expect this increased atmospheric
concentration of CO2 to increase the global mean surface temperature by 2-5 °C,4 and
this predicted increase in the average temperature is already occurring. All of the five
hottest years on record, (global average since 1880), have occurred since 2010, and all
of the twelve hottest years on record have occurred since 1998.5 The risks associated
with such climate change could include mass extinctions and destabilization of global
and regional food and water supplies.4
The combustion of fossil fuels for energy–either for production of electricity, or in
the transportation, industrial, or residential and commercial sectors–is the largest source
of greenhouse gas emission.6 It has been demonstrated that the global average energy
consumption can be closely modeled as proportional to human population, global
average gross domestic product (GDP) per capita, and the global energy intensity, or
energy consumption per unit GDP.7,8 Due to increased conservation and improvements
in efficiency due to technology, the total global energy intensity has actually been

2

steadily decreasing for the past century, and this trend is projected to continue.8
However, the global population is projected to increase to between 9 and 10 billion by
the year 2050 (from about 7.5 billion today), and it is not in the best interest of any
country to intentionally decrease its own per capita GDP.9 Since the rate of increase of
the global population and per capita GDP is faster than the decrease in energy intensity,
the global energy consumption is projected to increase to 26 TW (from 18 TW today) by
the year 2040.10,11 86 % of the energy consumed globally today is derived from the
combustion of fossil fuels, with approximately equal portions of oil, natural gas, and coal
consumed.8,10,11 By most estimates, there are no foreseeable supply constraints that will
limit continued consumption of fossil fuels.4,11 However, relying on fossil fuel combustion
to meet the world’s increasing energy demand will result in an associated increase in
carbon dioxide (CO2) emissions from 33 GT today to a projected 46 GT in 2040.4 Thus,
in order to mitigate the worst possible outcomes of climate change, large-scale, carbonneutral alternative energy resources must be implemented in the coming decades.
Solar energy is by far the largest energy resource available, and efficient
photovoltaic (PV) technologies comprised of Earth-abundant elements already exist.
While their relatively high cost compared to fossil fuels has hampered their commercial
implementation, conventional PV’s are on the verge of reaching grid-parity, and next
generation PV’s may prove disruptive in overcoming the cost-efficiency paradigm.
Intrinsic limitations of solar energy, however, are its diffusivity and geographic,
seasonal, and daily variability. Thus, in order to effectively utilize the power of the sun,

3

efficient and scalable methods to concentrate, store and transport solar energy are most
likely required.

Ammonia as an Energy Carrier
The above has motivated research in artificial photosynthetic systems that use
solar photons to drive the water splitting reactions, thereby storing solar energy in the
bonds of molecular hydrogen (H2). H2 gas formed from photoelectrochemical water
splitting could be used with either conventional combustion engines or in fuel cells, with
either process featuring only water as the byproduct–thus eliminating greenhouse
gasses from the fuel cycle. However, while the specific energy, (energy per unit mass),
of H2 is unrivaled, H2 gas has a relatively low energy density (energy per unit volume).
Table 1.1 shows a comparison of various potential fuel sources, as well as a
commercial lithium ion battery technology. Using state-of-the-art carbon fiber storage
tanks a volumetric energy density of only 5.3 MJ L-1 can be achieved for H2 gas at 700
bar pressure.12 This is over a six-fold reduction from the 34.4 MJ L-1 for gasoline.13
Extreme cryogenic temperatures are needed to liquefy H2, and large volumes of
hydrogen cannot be stored at high pressures. Consequently, the storage and
transportation of hydrogen represents a significant technical challenge that needs to be
overcome before it can become a widely-used fuel.
One possible avenue to store and transport hydrogen is to convert it to ammonia
(NH3). Synthesis of NH3 (largely via the Haber-Bosch process) is already one of the

4

Table 1.1: Energy density comparison of various fuels and commercial lithium ion
batteries.12–14
Energy Source
Gasoline (C8H18)
Propane (C3H8, 14 bar)
Natural Gas
(CH4, 250 bar)
Hydrogen (H2)
(700 bar, Carbon fiber tank)
Ammonia (NH3, 10 bar)
Tesla Powerwall 2
(Commercial Li Ion Battery)

Specific Energy
/ MJ kg-1
46.7
48.9

Energy Density
/ MJ L-1
34.4
19.0

55.5

10.4

142

5.3

22.5

13.6

0.405

0.361

largest industrial processes worldwide, and as a result, the infrastructure for the storage
and transport of NH3 already exists. It is held in tanks at the 50,000 ton scale for up to
years at a time, and is shipped via conventional ground transportation or via pipeline.15
Pipeline networks dedicated to NH3 are already in place in the United States, and
existing natural gas pipelines can be used to transport NH3 as well with minimal
modifications.15 NH3 is also already distributed and sold commercially primarily for use
in agriculture as fertilizer. Safety protocols are already in place for these processes
involving NH3, and independent safety evaluations have been performed for NH3 fuel
systems, concluding the level of risk to be similar-to or less-than that of natural gas or
propane.16,17
Unlike hydrogen, NH3 can be liquefied easily under 10 bar of pressure, which
leads to an energy density of 13.6 MJ L–1–almost three-fold larger than H2. In addition,
NH3 can be decomposed back to nitrogen and hydrogen thermodynamically with less
than 10% of the energy needed for splitting H2O.18,19 NH3 “splitting” would generate
5

hydrogen that could be utilized in hydrogen fuel cells. The inter-conversion of hydrogen
and ammonia therefore potentially offers a more efficient route to store and transport
hydrogen than compression. Alternatively, NH3 itself could potentially be used in direct
ammonia fuel cells.20–22
Figure 1.1 shows a graphical depiction of a possible ammonia fuel cycle.23 This
cycle is completely closed and contains only oxygen (O2) and nitrogen (N2) from the air
and water (H2O) in addition to NH3. If the N2 reduction process in the lower-left corner,
and the NH3 splitting reaction at the top are driven using renewable energy, then there
are no greenhouse gas emissions or any other pollutants involved with this cycle.
Hydrogen fuel cell cars like the one suggested in the lower-right corner are already
available for consumers today. Additionally, there is extensive ongoing research into the
catalytic synthesis of NH3 using renewable energy, as suggested in the lower-left
corner, that may match, or even exceed the efficiency of the widely-adopted HaberBosch process.24–26
The large point of uncertainty for the realization of the fuel cycle in Figure 1.1 is
the NH3 splitting process at the top. If that reaction cannot be done efficiently with
minimal greenhouse gas emissions, then the entire cycle would be unreasonable for
implementation. There is a large body of promising work on the thermal decomposition,
or “cracking”, of NH3 using solid-state catalysts.27 However, these catalysts generally
incorporate expensive noble metals and require operation temperatures around 300 ºC
or greater to achieve sufficient conversion. More importantly, thermal cracking in this

6

Figure 1.1: An ammonia fuel cycle involving the reduction of N2 coupled with H2O
oxidation (lower left), the splitting of NH3 to N2 and H2 (top), and the oxidation of H2 as a
fuel (lower right). (Reproduced from Little et al.)23

way is a continuous-flow process, which is not conducive to the inherently intermittent
nature of solar (or wind) energy, or the periodic increases and decreases in fuel
demand over the course of the day and year. What is preferred, therefore, is an
electrochemical reactor that can switch between active and idle states quickly rather
than needing to run continuously
Such a reactor would drive the splitting of NH3 through electrolysis–where a
potential is applied across two electrodes in contact with NH3 in solution–evolving N2 at
the positive electrode or “anode”, and evolving H2 at the negative electrode or
“cathode”. The magnitude of the potential needed to drive the electrolysis is equal to

7

0.077 V–the thermodynamic minimum for the reaction–plus some additional potential
called an “overpotential”. Overpotential is needed to push the reaction in the forward
direction as there is no net reaction when the overpotential is zero. For a particular
catalyst/electrode, a larger overpotential corresponds to a larger driving-force for the
reaction to proceed, and thus generally a faster rate of reaction, measured as current.
Therefore, a small amount of overpotential is fundamentally required in electrolysis, but
more efficient catalysts can increase the current without increasing the potential.

The State of Ammonia Electrolysis Research
The electrolysis of NH3 to H2 and N2 has been studied primarily only in aqueous
solution. Vitse et al. used an alloy of platinum and iridium to catalyze the electrolysis of
a 1 M aqueous ammonia (NH3(aq)) solution and obtained current densities between 400
and 500 mA cm–2 with only 500 mV of applied overpotential.19 Under these aqueous
conditions, the reduction of the H2O solvent is the source of the H2 product, and
hydroxide ions (OH–) have been reported to participate in the oxidation of NH3(aq) as
proton acceptors as suggested by the mechanism below in Equations (1.1) through
(1.5)–originally postulated by Gerischer and Mauerer.28
−
−
!!!
⇀
NH3,(ad ) + OH(ad
! NH2,(ad ) + H2O + e
) ↽!!

−
−
!!!
⇀
NH2,(ad ) + OH(ad
! NH(ad ) + H2O + e
) ↽!!

!!!
⇀
NHx,(ad ) + NHy ,(ad ) ↽
!!
! N2Hx+y ,(ad ) (x,y = 1 or 2)
−
−
!!!
⇀
N2Hx+y ,(ad ) + (x + y)OH(ad
! N2 + (x + y)H2O + (x + y)e
) ↽!!

8

(1.1)
(1.2)
(1.3)
(1.4)

−
NH(ad ) + OH(ad
⎯⎯
→ N(ad ) + H2O + e − (N(ad ) = poison)
)

(1.5)

After investigating the NH3(aq) oxidation voltammograms carefully as a function
of pH, Katsounaros et al. were able to conclude that the rate determining process
involves decoupled electron and proton transfers–as opposed to the proton-coupled
electron transfer processes suggested by Gerischer and Mauerer.29 Thus, in the revised
mechanism, either equation (1.1) or (1.2) should be replaced by equations (1.6) and
(1.7). Even in the revised mechanism it is widely accepted that OH– plays a critical role
in the NH3(aq) oxidation mechanism by acting as a base and accepting the protons as
they are removed from the nitrogen species.
−
!!!
⇀
NxHy ,(ad ) + OH− ↽
!!
!(NxHy −1)(ad ) + H2O

−
−
!!!
⇀
(NxHy −1)(ad
! NxHy −1 ,(ad ) + e
) ↽!!

(1.6)
(1.7)

An important point to emphasize in the oxidation of NH3(aq) is the formation of
N(ad) as shown in (1.5). The reactions of (1.3) and (1.5) are competing pathways, and
thus at sufficiently high positive overpotential, N(ad) formation occurs faster than N–N
bond formation leading to a poisoned anode surface. This is observed in voltammetry
as a peak followed by a sharp decline in anodic current as potential is swept in the
positive direction.30 The “poisoned” electrode is then ineffective for further oxidation of
NH3(aq). However, poisoned anodes can then be restored, generally, by sweeping to,
or the application of, sufficiently negative potentials.30 This is because the N(ad) poison
species exists exclusively on the surface of the anode, and thus does not cause a
significant change in electrode morphology or bulk composition.

9

The cathodic half reaction during NH3(aq) electrolysis is the reduction of H2O to
H2 as shown in equation (1.8), making NH3(aq) electrolysis further analogous to H2O
electrolysis under basic conditions. Thus noble metal proton reduction catalysts
effective for H2O electrolysis are also generally effective as cathodes for NH3(aq)
electrolysis.31 Since this reaction occurs at the same potential as simple proton
reduction–after accounting for changes in pH–the thermodynamic potential needed to
drive the entire cell (both half reactions), Erxn, is minimized.

1
−
!!!
⇀
H2O + e − ↽
!!
! H2 + OH
2

(1.8)

Since both anodic and cathodic half reactions are affected by the presence of
water, (or OH–), the electrolysis of NH3(aq) is not directly analogous to the electrolysis of
liquid ammonia (NH3(l)). Still, there is information that can be gleaned from the larger
body of NH3(aq) electrolysis research. For instance, the problem of anodic poisoning as
shown in (1.5), where at sufficiently positive overpotentials the rate of N(ad) formation
exceeds that of N–N bond formation, can potentially be a problem regardless of solvent.
The formation of the N(ad) poison species is a significant barrier for NH3(aq) oxidation as
it sets a maximum anodic overpotential limit that practical devices cannot exceed.29,32 In
general, however, the electrolyses of NH3(aq) and NH3(l) should be treated as different
processes, and it is dangerous to make assumptions about NH3(l) based upon what is
known in aqueous conditions.
For practical considerations, if the electrolysis efficiencies were identical, NH3(l)
would be desired over NH3(aq) for implementation into the energy framework. A
saturated NH3(aq) solution is approximately 14 M at STP, whereas NH3(l) is 40 M—
10

nearly triple the energy density. Further, the alkaline NH3(aq) solutions are corrosive to
metallic storage and transport infrastructure, whereas there is already a substantial
existing pipeline system for the distribution of NH3(l). In addition, electrolysis of NH3(l)
eliminates the possibility of forming NOx compounds as unwanted side products.30,33 If
the liquid state is sustained using decreased temperatures–such as -65 ºC in the case
of our NH3(l) experiments–there is a thermodynamic penalty, raising Erxn from 0.077 V
to 0.10 V (calculation vide infra). However, practical devices could operate at ambient
temperature simply by using elevated pressure to sustain the NH3(l).18,34
In contrast to the electrolysis of NH3(aq), the electrolysis of NH3(l) has received
almost no attention. The primary example is a paper by Hanada et al, who reported 7.2
mA cm–2 from the electrolysis of NH3(l) with 2.0 V applied between two platinum
electrodes at 9.63 bar and 25°C.18 This was only achieved with a relatively high
concentration (1 M) potassium amide (KNH2) electrolyte, which in solid form is known to
detonate. They demonstrated using gas chromatography that H2 and N2 are the
products generated at the cathode and anode respectively, and attributed these
products to the reduction of NH3 and oxidation of amide (NH2–). However, this
mechanism was not substantiated. While they showed that increasing the concentration
of NH2– increased the rate of NH3(l) electrolysis, they did not report controls where the
NH2– concentration was changed but the ionic strength of the solution was held
constant. In addition, since only two electrodes were used, the overpotentials for each
of the half reactions were not identified.

11

As will be discussed in detail in Chapter 3, we were able to demonstrate that the
half reactions of the electrolysis of NH3(l) on Pt electrodes are the reduction of
ammonium (NH4+) and the oxidation of NH3.23 In fact, the reduction of NH3 in NH3(l) is
impossible with a Pt cathode as electron solvation–a process where free electrons are
dissolved directly into solution–occurs at more positive potentials.35 We also
demonstrated that NH2– electrolytes actually decrease the rate of NH3 oxidation by
showing much larger anodic currents using either NH4+ or “neutral” electrolytes.
These findings were later confirmed by Dong et al, who studied the electrolysis of
NH3(l) using ammonium halide electrolytes, as well as ammonium nitrate (NH4NO3), and
potassium amide (KNH2).36 They found that anodic current increased by switching from
ammonium chloride (NH4Cl), to bromide (NH4Br), to iodide (NH4I), and that NH4NO3 had
a similar performance to NH4I. KNH2 drastically underperformed all three of the above,
however, in agreement to what we observed, and what Hanada et al. reported.18,23
Dong et al. were able to further improve anodic performance by switching to a
nanostructured rhodium-platinum-iridium (Rh-Pt-Ir) alloy which reduced the anodic
overpotential by approximately 0.7 V.36 However, they report that while the geometric
areas of their Pt electrode and Rh-Pt-Ir electrode are the same, the nanostructured RhPt-Ir electrode has an electrochemical active surface area over 360 times larger.
In 2016, Goshome et al. also reported the electrolysis of NH3(l) with an NH4Cl
electrolyte, investigating the activity of both Pt and stainless steel anodes.34 They
achieved current densities similar to our report and that of Dong et al, though they also
saw significant anodic dissolution after sustaining a cell potential of 2 V for 8 hours in

12

the case of stainless steel, or 50 hours in the case of Pt. They determined the
dissolution products to be [Fe(NH3)6][Cl]2 and PtCl4•xNH3 respectively via powder x-ray
diffraction (XRD). However, only one variety of stainless steel was tested as an anode,
and the type used was not specified in the paper. Also, chemical processes occurring at
the anode could not be studied specifically due to the lack of a reference electrode.
There has also been some work studying the NH3 electrolysis reaction pathways
using alternative solvents. Buzzeo et al. studied both the oxidation and reduction half
reactions for the case of NH3 dissolved in N,N-dimethylformamide (DMF) or the ionic
liquid 1-Ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide on Pt electrodes.37
They concluded that the oxidation of NH3 and reduction of H+ were the important half
reactions.

However,

their

electrochemical

cell

incorporated

a

silver

wire

pseudoreference electrode, making it difficult to quantitatively compare reaction
potentials. Still, the technique demonstrates the possibility of learning mechanistic
information about NH3 electrolysis through the use of an organic solvent.
The objective of our research is to more closely explore both the anodic and
cathodic half reactions of the electrolysis of NH3(l), identify the chemical problems
hindering the efficiency of the overall process, and suggest solutions to those problems
involving sustainable, Earth-abundant materials. First, the electrolysis of NH3(l) using Pt
as both the anode and cathode is investigated as a function of the relative
concentrations of NH4+ and NH2– in the electrolyte. By coupling the NH3(l) results with
data from the electrolysis of NH3(g) dissolved in DMF, the half reactions occurring at
each electrode are clarified and the requirements of large overpotentials for each are

13

discussed. Second, a new NH3(l) electrolyte preparation method is introduced and
discussed, and the characteristics of an Earth-abundant Fe/FeNx anode material will be
highlighted. Then, electrodeposited CoPx is presented as an Earth-abundant alternative
cathode material. Finally, possible future research directions are suggested that should
clarify the route toward better NH3 splitting catalysis.

14

APPENDICES

15

Appendix 1.1: Calculation of Ecell for the Electrolysis of Liquid Ammonia
The thermodynamic potential for electrolysis, Ecell, is related to the free energy
change of the reaction, DGrxn, by (1.9) where n is the number of electrons involved in
the electrochemical reaction, and F is Faraday’s constant.

Ecell =

−ΔGrxn
nF

(1.9)

The reaction of interest is the decomposition of ammonia (NH3), shown by (1.10).

1
3
NH3 (l) ⎯⎯
→ N2 (g) + H2 (g)
2
2

(1.10)

This decomposition reaction is simply the inverse of the formation of NH3 from its
elements at standard state. The free energy change of formation (DG°f) is related to the
changes in entropy (DS°f) and enthalpy (DH°f) by (1.11):
ΔGo f = ΔHo f − TΔSo f

(1.11)

NH3 is not liquid under standard conditions (T = 298 K and 1 atm pressure), however,
where values for DH°f andDS°f are available. Thus, Ecell at the temperature of this study,
203 K, was estimated by using DH°f and DS°f at 298 K, (which are easily obtained from
tabulated values)38 and a temperature of 203 K in equation (1.11). DS°f is determined
from S° for the reactant and both products as shown in (1.12) through (1.14).38
⎡1
3
ΔSo f = ⎢ So N2 (g) + So H2 (g)
2
⎣2

(

)

(

)⎤⎥ − S (NH (l))
o

3

⎦

⎡1
⎤
3
ΔSo f = ⎢ 191.6 J mol−1 K −1 + 130.7 J mol−1 K −1 ⎥ − 103.3 J mol−1 K −1
2
⎣2
⎦

(

) (

) (

16

(1.12)

)

(1.13)

ΔSo f = 188.6 J mol−1 K −1

(1.14)

Plugging into (1.11):
203 K

(

) (

)(

)

ΔGrxn ≈ 67200 J mol−1 − 203 K 188.6 J mol−1 K −1 = 28900 J mol−1

(1.15)

Plugging the result from (1.15) into (1.9):
Ecell

(
( )(

)

− 28900 J mol−1
− 203 K ΔGrxn
=
=
= −0.100 V = −100 mV
nF
3 96485 C mol−1

17

)

(1.16)

Appendix 1.2: Background Theory for 2-Electrode and 3-Electrode Measurements
In different parts of this project, electrochemical measurements were made that
employed either two electrodes, an anode and cathode, as depicted in Figure 1.2(a), or
three electrodes with the addition of a reference electrode as depicted in Figure 1.2(b).
Cells employing only two electrodes have the advantage of more closely resembling
practical devices, but it is far simpler to investigate specific electrochemical reactions
with the addition of the reference electrode. For the sake of clarity, we will discuss some
important differences between the two experiments.

Figure 1.2: Electrode configuration diagrams for a 2-electrode (a) and 3-electrode (b)
experiment. In the 3-electrode case, the working, reference, and counter electrodes are
labeled W.E., R.E., and C.E. respectively. In both cases, the red “VA” represents where
a potential is applied, the green “VM” represents where a potential is measured, and the
blue “IM” represents where the current is measured.
In the 2-electrode configuration there is a positively charged “anode”, which is
where an oxidation reaction occurs, and a negatively charged “cathode” where a
reduction reaction occurs. Experimentally, a potentiostat is used to apply a known
potential between the anode and cathode–as indicated by the red “VA” in Figure 1.2.

18

Then, the electrons flowing into the anode, through the circuit, and then out of the
cathode are measured as a current–as indicated by the blue “IM”. In order for Faradaic
current to flow in this configuration, the potential applied between the anode and
cathode must be greater than the thermodynamically limiting Ecell, which for the case of
NH3 electrolysis is specified in equation (1.16). After that point, further increasing the
potential will result in an increase in the measured current. This additional potential is
called “overpotential”.
Unfortunately, in the 2-electrode configuration it is impossible to determine
whether the anodic or cathodic reaction is rate limiting, i.e. requires more overpotential
to increase the current. Thus, for most electrochemical experiments a 3-electrode
configuration is used. Here a “working” electrode–“W.E.” in Figure 1.2(b)–can serve as
either the anode or the cathode as its potential is set with respect to a “reference”
electrode, “R.E.”. A reference electrode is actually an electrochemical half-cell that will
hold a fixed potential so long as only a small magnitude current flows through the cell.
For that reason, a third “counter” electrode–“C.E.” in Figure 1.2(b)–is used to complete
the circuit for current measurements.
The potential difference between the working and reference electrode is known,
and since the potential of the reaction in the reference electrode is constant, the
potential at the surface of the working electrode specifically is known, providing the
advantage over the 2-electrode configuration. The potential at the counter electrode is
unknown in this configuration, as the potentiostat simply allows it to become any
potential necessary to allow current to flow through the circuit and prevent excess

19

charge from building up at the working electrode. In the 3-electrode configuration, either
oxidation or reduction reactions can occur at the working electrode, depending upon the
potential with respect to the reference, and thus the current/overpotential relationship for
each half-reaction can be studied independently.

20

Appendix 1.3: Discussion of the Term “Onset Potential”
The term “onset potential” is frequently used in discussions comparing catalytic
efficiencies of different materials. However, the term is not rigorously defined. In fact, as
the scale of a plot is continually magnified, an electrolysis J-V curve will continue to hold
the same shape, making it arbitrary to define where the “onset” of the wave is on the
potential axis.
In Chapter 3, due to the current density frequently reaching a linear, ohmic
regime where the slope is dictated by the solution resistance, the “onset potential” is
defined as the point where a line drawn through the linear portion of a redox wave of the
J-V curve would intersect with the baseline. A sample of this method is depicted in
Figure 1.3. With this definition, it is possible for two CV curves to have similar onset
potentials, but significantly different current densities at their onsets. Therefore,
materials will also be compared by the overpotentials required to reach a certain current
density, particularly in Chapters 4 and 5. These current densities are chosen arbitrarily
as well, however this is a convenient method for compensating for different shapes of
catalytic curves.
A common method to quantitatively compare the shapes of two CV or J-V curves
is through the use of a Tafel slope.39 In this method, the base-ten logarithm of the
absolute value of the current density is plotted as a function of potential, making it
simpler to compare the exponentially-curved regions of two CV or J-V plots. The slope
of the linear region of a Tafel plot indicates the additional overpotential required to

21

create a ten-fold increase in the magnitude of current, and can also serve as a useful
metric for quantitatively comparing the performances of two different materials.

Figure 1.3: Illustration of how the onset potential would be calculated from a sample
curve.

22

REFERENCES

23

REFERENCES

(1)

Harries, J. E.; Brindley, H. E.; Sagoo, P. J.; Bantges, R. J. Nature 2001, 410,
355–357.

(2)

Petit, J. R.; Jouzel, J.; Raynaud, D.; Barkov, N. I.; Barnola, J.-M.; Basile, I.;
Bender, M.; Chappellaz, J.; Davis, M.; Delaygue, G.; Delmotte, M.; Kotlyakov, V.
M.; Legrand, M.; Lipenkov, V. Y.; Lorius, C.; PÉpin, L.; Ritz, C.; Saltzman, E.;
Stievenard, M. Nature 1999, 399, 429–436.

(3)

Tans, P.; Keeling, R. Atmospheric CO2 at Mauna Loa Observatory
https://scripps.ucsd.edu.

(4)

Field, C. B.; Barros, V. R.; Dokken, D. J.; Mach, K. J.; Mastrandrea, M. D.; Bilir, T.
E.; Chatterjee, M.; Ebi, K. L.; Estrada, Y. O.; Genova, R. C.; Girma, B.; Kissel, E.
S.; Levy, A. N.; MacCracken, S.; Mastrandrea, P. R.; White, L. L. Climate Change
2014: Impacts, Adaptation and Vulnerability. Contributions of the Working Group
II to the Fifth Assessment Report of the Intergovernmental Panel on Climate
Change; Cambridge University Press: Cambridge, United Kingdom and New
York, NY, USA, 2014.

(5)

NOAA National Centers for Environmental Information. State of the Climate:
Global Analysis for Annual 2016; 2017.

(6)

United States Environmental Protection Agency. Draft Inventory of U.S.
Greenhouse Gas Emissions and Sinks: 1990-2015; 2017.

(7)

Kates, R. W. Environment 2000, 42, 10–19.

(8)

Lewis, N. S.; Nocera, D. G. Proc. Natl. Acad. Sci. U. S. A. 2006, 103, 15729–
15735.

(9)

United Nations Department of Economic and Social Affairs, P. D. World
Population Prospects: The 2015 Revision; 2015.

(10) BP Statistical Review of World Energy, June 2016; 2016.

24

(11) U.S. Energy Information Administration. International Energy Outlook 2016; 2016.
(12) Hua, T.; Ahluwalia, R.; Peng, J.-K.; Kromer, M.; Lasher, S.; McKenney, K.; Law,
K.; Sinha, J. Technical Assessment of Compressed Hydrogen Storage Tank
Systems for Automotive Applications; 2010.
(13) Zamfirescu, C.; Dincer, I. J. Power Sources 2008, 185, 459–465.
(14) Powerwall 2 https://www.tesla.com/powerwall.
(15) Bartels, J. R. A feasibility study of implementing an Ammonia Economy, Iowa
State University, 2008.
(16) Duijm, N. J.; Markert, F.; Paulsen, J. L. Safety Assessment of Ammonia as a
Transport Fuel; Roskilde, Denmark, 2005.
(17) Quest Consultants Inc. Comparative Quantitative Risk Analysis of Motor
Gasoline, LPG, and Anhdryous Ammonia as an Automotive Fuel; Norman, OK,
2009.
(18) Hanada, N.; Hino, S.; Ichikawa, T.; Suzuki, H.; Takai, K.; Kojima, Y. Chem.
Commun. (Camb). 2010, 46, 7775–7777.
(19) Vitse, F.; Cooper, M.; Botte, G. G. J. Power Sources 2005, 142, 18–26.
(20) Afif, A.; Radenahmad, N.; Cheok, Q.; Shams, S.; Kim, J. H.; Azad, A. K. Renew.
Sustain. Energy Rev. 2016, 60, 822–835.
(21) Boggs, B. K.; Botte, G. G. J. Power Sources 2009, 192, 573–581.
(22) Lan, R.; Tao, S. Electrochem. Solid-State Lett. 2010, 13, B83.
(23) Little, D. J.; Smith, III, M. R.; Hamann, T. W. Energy Environ. Sci. 2015, 8, 2775–
2781.
(24) Tanabe, Y.; Nishibayashi, Y. Chem. Rec. 2016, 1–29.
(25) Shipman, M. A.; Symes, M. D. Catal. Today 2016, 1–12.

25

(26) Bandy, J. A.; Zhu, D.; Hamers, R. J. Diam. Relat. Mater. 2016, 64, 34–41.
(27) Yin, S. F.; Xu, B. Q.; Zhou, X. P.; Au, C. T. Appl. Catal. A Gen. 2004, 277, 1–9.
(28) Gerischer, H.; Mauerer, A. Electroanal. Chem. Interfacial Electrochem. 1970, 25,
421–433.
(29) Katsounaros, I.; Chen, T.; Gewirth, A. A.; Markovic, N. M.; Koper, M. T. M. J.
Phys. Chem. Lett. 2016, 7, 387–392.
(30) Le Vot, S.; Reyter, D.; Roué, L.; Bélanger, D. J. Electrochem. Soc. 2012, 159,
F91.
(31) Palaniappan, R.; Ingram, D. C.; Botte, G. G. J. Electrochem. Soc. 2014, 161,
E12–E22.
(32) Li, Z.-F.; Wang, Y.; Botte, G. G. Electrochim. Acta 2017, 228, 351–360.
(33) Bunce, N. J.; Bejan, D. Electrochim. Acta 2011, 56, 8085–8093.
(34) Goshome, K.; Yamada, T.; Miyaoka, H.; Ichikawa, T.; Kojima, Y. Int. J. Hydrogen
Energy 2016, 2–7.
(35) Itaya, K.; Malpas, R. E.; Bard, A. J. Chem. Phys. Lett. 1979, 63, 411–415.
(36) Dong, B.-X.; Tian, H.; Wu, Y.-C.; Bu, F.-Y.; Liu, W.-L.; Teng, Y.-L.; Diao, G.-W.
Int. J. Hydrogen Energy 2016, 41, 14507–14518.
(37) Buzzeo, M. C.; Giovanelli, D.; Lawrence, N. S.; Hardacre, C.; Seddon, K. R.;
Compton, R. G. Electroanalysis 2004, 16, 888–896.
(38) Jolly, W. L. Chem. Rev. 1952, 50, 351–361.
(39) Bard, A. J.; Faulkner, L. R. Electrochemical Methods: Fundamentals and
Applications; 2nd ed.; Wiley and Sons Inc., 2000.

26

CHAPTER 2:
EXPERIMENTAL DESIGN FOR THE ELECTROCHEMICAL INVESTIGATION OF
ANHYDROUS AMMONIA

27

Electrochemical Experimentation in Stock Liquid Ammonia
Anhydrous ammonia (Airgas) was liquefied at ambient pressure by passing the
gas directly into a pre-assembled electrochemical cell like the one shown in Figure 2.1
submerged in a bath of dry ice in methanol. All electrochemical experiments were
conducted in this bath, which was kept between -70 and -65 °C. After a sufficient
amount of NH3(l) was collected–as designated by a horizontal mark on the side of the
cell–the flow of NH3(g) at the inlet was replaced by a gentle flow of dry N2. The outlet
was connected to a silicon oil bubbler at all times. An amount of ammonium
hexafluorophosphate

(NH4PF6)

(Fluka,

98%,

recrystallized),

potassium

hexafluorophosphate (KPF6) (Acros, 99%, recrystallized), or ammonium nitrate
(NH4NO3) (Jade Scientific, reagent, dried and stored under vacuum) required to
produce an NH3(l) solution of the desired electrolyte concentration was added to the
electrochemical cell, along with all electrodes as shown in Figure 2.1, and the cell was
sealed using virgin, rubber septa.
The two-electrode electrolysis of NH3(l) was performed using two identical
platinum metal (Pt) disks as the anode and cathode. Three-electrode voltammetry in
NH3(l) was performed using one of the Pt disk working electrodes, a high surface area
Pt mesh as a counter electrode, and a custom-made silver/silver nitrate (Ag/AgNO3)

28

Figure 2.1: Diagram of the electrochemical cell used for voltammetry in NH3(l) as well
as in DMF with dissolved NH3(g).

reference electrode. (The cell in Figure 2.1 depicts the three-electrode setup.) The
Ag/AgNO3 reference–as shown in the diagram in Figure 2.2–was fabricated by melting
a glass Pasteur pipette around a small piece of Pt wire at one end, and filling the inside
of the pipette with a methanol solution of 0.1 M tetrabutylammonium perchlorate (TBAP)
(Fluka, 99%), and 0.01 M AgNO3 (Jade Scientific, Reagent ACS). A silver wire was then
inserted into the open end of the pipette, sealed in place using a rubber septum, and
connected to the potentiostat.1 The resulting reference held a sufficiently stable
potential in any solvent it was tested in, including NH3(l). The potential of the reference
electrode was measured versus a commercial saturated calomel electrode (SCE),
allowing all electrochemical measurements to be reported versus NHE after

29

conversion.2 Unless otherwise noted, all cyclic voltammograms (CVs) were conducted
at a scan rate of 100 mV s–1.

Figure 2.2: Diagram of the custom Ag/AgNO3 reference electrode for use in NH3(l)
solutions.

Special care was taken when conducting voltammetry in NH3(l) with potassium
amide (KNH2). KNH2 was synthesized as reported by Hanada et al. from anhydrous
ammonia and potassium hydride (KH) (Alfa Aesar, 30% w/w in mineral oil), and was
kept under a nitrogen atmosphere at all times.3 Characterization of the resulting powder
is shown in Figure 2.3. NH3(l) was formed in situ inside a sealed flask containing KNH2,

30

and

was

quenched

with

isopropanol

immediately

following

electrochemical

measurements.

Figure 2.3: Powder XRD spectrum of KNH2 synthesized from KH and liquid NH3. A
reference spectrum for KNH2 is shown in the red, dotted line. The large amorphous
feature at 20° is due to the sample holder, which included a dome to maintain a nitrogen
atmosphere for the powder. (Reproduced from Little et al.)4

Distillation of Liquid Ammonia for Electrochemistry in Rigorously Dry Conditions
From Chapter 4 and on, NH3(l) was first thoroughly dried before use. To
accomplish this, anhydrous ammonia (Airgas) was liquefied at ambient pressure by
condensing the gas in a custom flask submerged in a methanol/dry ice bath as depicted
in Figure 2.4. When a sufficient amount of NH3(l) was collected (approximately 200 mL),
about 1 cm3 of metallic sodium was added to the flask through the center neck, resulting

31

in the characteristic deep-blue color of solvated electrons that indicates the removal of
trace water in the liquid. The flask was then lifted out of the bath, one end of a cannula
was inserted in the sealed electrochemical cell, and then other was inserted through the
virgin septum, keeping the tip above the NH3 solution. As the flask warmed, dry NH3(g)
was transferred to sealed electrochemical cell.

Figure 2.4: Diagram of the NH3(l) distillation flask including two glass valves for the gas
inlet and outlet, as well as a black, plastic screw-cap containing a white Teflon septum.

The methodology for electrochemical experiments conducted with distilled NH3(l)
was similar to that of the stock NH3(l) experiments, except that after assembly and
before solvent condensation, the cell was then purged of air by carefully pulling a
vacuum and refilling the flask with dry nitrogen using a Schlenk line, repeated three

32

times. This way, all electrochemical experiments could be reasonably considered free of
air and water contamination. When conducting gas chromatography experiments, argon
gas (Airgas, 99.99%) or helium gas (Airgas, 99.99%) was used to fill the headspace of
the cell in place of nitrogen so that the nitrogen gas evolved during the reaction could be
measured.
From Chapter 4 and on, a custom Ag/AgCl reference electrode was used in
place of the Ag/AgNO3 reference previously described. The Ag/AgCl reference
electrode was designed similarly to the Ag/AgNO3 reference, however the electrolyte
was saturated in both ammonium chloride (NH4Cl) (Fischer, >99.6%) and NH3(g) in
methanol. The silver wire was first mechanically polished, cleaned in 3 M nitric acid, and
then pre-coated with AgCl by applying 0.4 mA cm–2 in a 0.1 M solution of hydrochloric
acid.5 The wire was then sealed in the solution and allowed to reach equilibrium over
three days. The completed reference was demonstrated to be stable in NH3(l) by
measuring the potential for electron solvation to be -2.67 ± 0.01 V over a period of
hours.6 Potentials measured with this reference were converted to NHE by measuring
the reversible redox wave for ferrocene in acetonitrile with 0.1 M tetrabutylammonium
hexafluorophosphate (TBAPF6) (Aldrich, 98%, recrystallized), which has been reported
at 0.31 V versus SCE.2 This was confirmed via measurements with a commercial SCE
electrode. SCE values were then converted to NHE by adding 241 mV.2

33

Evaluation of Electrochemical Headspace Via Gas Chromatography
Gas chromatography (GC) was conducted using an Agilent 7820A GC System
incorporating a 30 m HP-PLOT/U column for separating NH3 from N2 and H2, and a 50
m long 5Å Molsieve for separating N2 and H2 from each other, and a thermal
conductivity detector (TCD). The thermal conductivity detector was calibrated for N2 and
H2 by injecting known quantities of pure N2 or pure H2 using either a 100 µL or 25 µL
syringe. The TCD senses changes in the thermal conductivity of the gas in the outlet
due the presence of analyte mixed with the carrier gas, and thus the sensitivity for
different gasses depends on the carrier gas chosen. Since He and H2 have very similar
thermal conductivities, the TCD sensitivity for H2 is very low when He is used as a
carrier gas–though it is very high when Ar is used as the carrier. The opposite is true for
N2 detection. The choice of carrier gas thus impacts the quantity of product required for
detection and should be carefully considered before carrying out an experiment.
The GC system also featured a pneumatic bypass valve which allowed for flow
through both the Plot/U and Molesieve columns to the detector when “off”, or straight
from the Plot/U column to the detector, bypassing the Molesieve, when “on”. This
feature is important as larger molecules such as NH3, H2O, or CO2 take an extremely
long time–up to hours–to pass through the 5Å Molsieve and then experience extreme
peak broadening. Thus, the separation procedure involves injection with the bypass
valve “off”, then switching the valve to “on” after the non-dipole-containing gases (H2,
He, N2, O2, and Ar) have passed the valve onto the Molsieve. (Of course, Ar is
significantly larger and heavier than the others, but still quickly passes onto the

34

Molsieve because in is non-polar.) These non-dipole-containing gases then sit stagnant
on the Molsieve while the remaining gases elute from the Plot/U directly to the detector.
After this elution step is complete, the bypass valve is turned back to “off” and the
remaining gases elute in increasing order of their size.
A sample port incorporating a low-bleed rubber septum and septum injector nut
(Vici Valco) was connected to one of the narrow, quarter-inch necks of the
electrochemical flask via a Swagelok Ultra-torr Vacuum Fitting. Gaseous samples were
extracted from the headspace of the cell using a Hamilton 500 µL Model 1750 SampleLock syringe. In order to not overfill the GC inlet, only 100 µL samples were extracted
and injected. Even with the Sample-Lock feature of the syringe (a Teflon valve that
allowed for manual sealing of the syringe), great care was taken to inject samples into
the GC as quickly as possible after extraction from the cell headspace. This was done
to minimize any sample leakage and N2 contamination from the atmosphere into the
sample.
The complete separation method took approximately 30 minutes, and as a result
samples were extracted from the cell headspace every 30 minutes. During bulk
electrolysis experiments, the potential of interest was applied for 30-minute intervals
between injections, instead of continuously, in order to easily identify the charge passed
between each separation and quantification of the headspace.
In order to subtract any N2 originating from air contamination, several samples of
laboratory air were injected through the GC in order to precisely identify the N2:O2 ratio
from the air. Since the headspace of the cell was thoroughly purged with vacuum and

35

Ar, the initial partial pressure of O2 in the cell was assumed to be 0. Thus, the area of
the O2 peak in the gas chromatograph could be attributed exclusively to air
contamination. Therefore, when analyzing the GC data, the fraction of the N2 peak
proportional to the O2 peak was subtracted, and the remainder was assumed to result
from electrochemical N2 production.

Elemental Analysis of the Electrochemical Cell Contents Via ICP-OES
Elemental analysis of the liquid-phase of the NH3(l) electrochemical cell was
conducted through Inductively-Coupled Plasma Optical Emission Spectroscopy (ICPOES). After all electrochemical experiments were complete, the NH3(l) solution was
carefully poured into a vial and left in the fume hood to slowly evaporate to dryness. The
remainder of the cell was rinsed with three 5 mL aliquots of Millipore water, which were
poured into a separate vial for collection. (Three 5 mL aliquots of Millipore water was
sufficient for collection of trace dissolved nickel or copper which are reported to dissolve
at 100% Faradaic efficiency with positive applied bias.7–12) These two vials were
quantitatively diluted by a factor of 250 in 2% nitric acid (HNO3) in order to be certain of
the total volume and sufficiently dilute the supporting electrolyte to avoid salt
depositions in the ICP instrument. A standard “Fe stock” solution was made by
dissolution of a precisely known amount of iron(III) nitrate (Spectrum, 99%) into 2%
HNO3 using a 100.0 mL volumetric flask. A standard dilution series for iron (Fe)
quantification was then generated using aliquots of the “Fe stock” measured with a 100–
1000 µL Eppendorf pipettor and additional volumetric flasks.

36

In order to conduct the ICP-OES measurements, we were required to switch to
NH4NO3 as the supporting electrolyte instead of the NH4PF6 or KPF6 used previously.
This decision was made to prevent any source of fluorine–PF6– in this case–from being
injected into the instrument. Nitrate (NO3–) was chosen as the replacement anion to
avoid halide ions that can be oxidized at the very positive potentials we employed.
NH4NO3 is also very soluble in NH3(l), the nitrogen is already in its highest oxidation
state so it cannot be oxidized further, and since all samples for ICP-OES are dissolved
in 2% nitric acid (HNO3), using NO3– as the supporting anion keeps the analyte as
simple as possible. In order to keep the findings consistent, all electrochemical
experiments involving the investigation of Fe electrode corrosion in Chapter 4–
specifically those preceding ICP-OES, XPS, and GC–were conducted with 0.1 M
NH4NO3 as the supporting electrolyte.

Voltammetry in N,N-Dimethylformamide
Electrochemical experiments conducted in N,N-Dimethylformamide (DMF) used
the same electrodes as in NH3(l) with respect to both the two-electrode and threeelectrode configurations. The electrolytes in DMF consisted of varying concentrations of
either NH4PF6 or TBAPF6, depending on the desired concentration of NH4+. When
dissolved NH3 was required, anhydrous NH3 gas (Airgas) was bubbled for 2-3 minutes
until the solution was saturated.

37

Sampled Time Voltammetry
In voltammetry experiments, where the potential is changed as a function of time,
both potential and time are convoluted together in the resulting CV or J-V curve. Thus,
deviations in the standard Butler-Volmer-esque shape of a voltammogram indicate
chemical changes occurring over time. However, when the working electrode surface is
chemically changing as a function of both time and potential–such as in the case of
anodic poisoning (Chapter 3), dissolution (Chapter 4), or the formation of a catalytic film
(Chapter 4)–it is necessary to be able to experimentally deconvolute potential and time.
In order to accomplish this, Sampled Time Voltammetry (STV) was sometimes used in
place of simple CV or linear sweep voltammetry. Instead of simply sweeping the
potential at a linear scan rate, a sequence of chronoamperograms at regular potential
intervals was used. The current from each of the chronoamperograms was extracted at
a constant point and plotted as a function of potential, thus generating a J-V curve
where all points were collected after the same length of time. The key advantage of this
method is that the working electrode can be freshly regenerated between each
chronoamperogram so that at each point the only variables across the entire experiment
are the potential applied and the resulting current.
Unless otherwise noted, the current from the point 1 ms into each
chronoamperogram was plotted as a function of potential. A time of 1 ms is large
enough to ensure a negligible contribution from charging current, but also a relatively
small degree of electrode fouling (assuming solution resistance of 1 kΩ and an

38

electrode capacitance of 20 μF cm–2, the contribution from charging current should be
below 1 mA after only 10–5 s).

Fabrication of Working Electrodes From Transition Metal Wires
Unless otherwise mentioned, all working electrodes used were manufactured by
attaching a small piece of wire of the element of interest, (iron, copper, etc.), to a
relatively longer piece of Cu wire insulated with polyvinyl chloride. A polypropylene
pipette tip was then melted with a heat gun around the end of interest, enveloping all
conductive wire tightly in plastic. The very end was then removed carefully with an
electric belt sander resulting in a small disk electrode of the desired element. Disk
electrodes were then polished using 1000 grit sandpaper followed by a typical
commercial electrode polishing pad utilizing 50 nm alumina particle paste (BASi). The
Pt working electrode was made similarly to the Fe electrodes, except that the Pt wire
(0.5 mm, 99.99%, Aldrich) was fixed permanently in borosilicate glass. The disk and
surrounding glass were finely ground, and before each electrochemical experiment the
electrode was polished by the same typical commercial electrode polishing pad. The
disk-end of the Pt electrode was curved to a 90-degree angle–as depicted on the left of
Figure 2.1–to encourage electrochemically formed bubbles to become detached.
The Fe plate electrodes used in Chapter 5 were manufactured by mechanically
bending a long piece of the aforementioned Fe wire to attach it to a 0.5 x 0.5 cm piece
of Fe foil (0.25 mm thick, 99.99%, Alfa Aesar). No epoxy, adhesive, nor anything not

39

composed of Fe was used so that only Fe was exposed to the solution. The Fe plate
could easily be detached from the wire for instrumental characterization.
The Pt mesh counter electrode used in nearly all electrochemical experiments
was manufactured similarly to the Fe plate electrode, except that the Pt wire
mechanically attached to the Pt mesh was only approximately 1 cm long. The 1 cm Pt
wire was mechanically attached to a piece of insulated Cu wire, and the junction
between the wires was covered with melted polypropylene plastic to insulate all Cu from
being exposed to the solution.

Determination of the Electrochemically Active Electrode Surface Area
For most macroscopic electrodes, like those used in this study, there can be a
significant

difference

between

the

geometric

active

surface

area

and

the

electrochemically active surface area. This is due to micro- or nanostructures of the
electrode surface that provide additional active sites for electrochemical reactions
without an increase in the macroscopic size of the electrode. In this study, current
densities as a function of both geometric and electrochemically active surfaces are
employed, as appropriate, and identified in the text.
When required, the electrochemically active surface areas of electrodes were
determined by measuring the double-layer charging current as a function of CV scan
rate. When the Faradaic current at the working electrode is essentially zero, the
electrode can be modeled as a parallel-plate capacitor where the magnitude of the
current at a constant scan rate is proportional to the capacitance per unit active area of

40

the electrode. For the well-studied system of a platinum electrode in an aqueous
electrolyte, the double-layer capacitance by this approximation is about 20 µF cm–2. We
took advantage of this information to approximate the electrochemically active surface
areas of FeNx electrodes and CoPx electrodes in Chapters 4 and 5 respectively.
When determining the growth of the electrochemical surface area of the Fe/FeNx
electrode in NH3(l) in Chapter 4, we made the very coarse assumption that the doublelayer capacitance would be approximately the same as a platinum electrode in water.
This is undoubtedly untrue, but since we were only interested in comparing the change
of one Fe/FeNx electrode, rather than compare one electrode to another, both the trend
of electrochemically active surface area increasing with successive anodic scans, and
the magnitude of that increase, should still be correct.
We employed a more quantitative approach to compare the electrochemically
active surface areas of Pt and CoPx electrodes in DMF solutions in Chapter 5. We could
not isolate a region without significant Faradaic current for a CoPx in aqueous solution
due to either H2 evolution processes, or film dissolution, depending on the potential
region. First, the electrochemically active surface area of our Pt electrode was
determined in an aqueous solution using the 20 µF cm–2 assumption. Then, that
electrode was transferred to the DMF solution, and the dependence of the charging
current on the scan rate was re-evaluated. Since the surface area of that Pt electrode
was known from the aqueous measurement, the double-layer capacitance of the Pt
electrode in DMF could be calculated. Then, this value was used to model the

41

capacitance of a CoPx electrode in DMF so that the electrochemically active surface
area of all CoPx electrodes could be approximated in the DMF solution.

Instrumentation
All electrochemical experiments were conducted using either a Gamry Reference
600 potentiostat, or a Metrohm µAutolabIII. Elemental analysis of the electrochemical
solution was measured using a Varian 710-ES Axial ICP-OES. X-ray Photoelectron
Spectroscopy (XPS) was measured with a Kratos Axis Ultra XPS. Raman spectroscopy
was conducted using a Renishaw inVia Raman Microscope employing a RL532C100
laser source. Scanning Electron Microscopy (SEM) and Energy Dispersive X-Ray
Spectroscopy (EDS) measurements were recorded using a Carl Zeiss Variable
Pressure SEM EVO LS25 with an Ametek EDAX Apollo X Energy Dispersive X-ray
Detector.

42

REFERENCES

43

REFERENCES

(1)

East, G. A.; del Valle, M. A. J. Chem. Educ. 2000, 77, 97.

(2)

Bard, A. J.; Faulkner, L. R. Electrochemical Methods: Fundamentals and
Applications; 2nd ed.; Wiley and Sons Inc., 2000.

(3)

Hanada, N.; Hino, S.; Ichikawa, T.; Suzuki, H.; Takai, K.; Kojima, Y. Chem.
Commun. (Camb). 2010, 46, 7775–7777.

(4)

Little, D. J.; Smith, III, M. R.; Hamann, T. W. Energy Environ. Sci. 2015, 8, 2775–
2781.

(5)

Sawyer, D. T.; Sobkowiak, A.; Roberts, J. L. Electrochemistry for Chemists;
Second.; John Wiley & Sons, Inc.: New York, 1995.

(6)

Itaya, K.; Malpas, R. E.; Bard, A. J. Chem. Phys. Lett. 1979, 63, 411–415.

(7)

Bergstrom, F. W. J. Am. Chem. Soc. 1924, 46, 2631–2636.

(8)

Jolly, W. L. J. Am. Chem. Soc. 1956, 78, 4849–4849.

(9)

Vijh, A. K. J. Electrochem. Soc. 1972, 119, 861.

(10) Brown, O. R.; Thornton, S. A. J. Chem. Soc. Faraday Trans. 1 Phys. Chem.
Condens. Phases 1974, 70, 14.
(11) Brown, O. R.; Thornton, S. A. J. Chem. Soc. Faraday Trans. 1 Phys. Chem.
Condens. Phases 1974, 70, 1269.
(12) Heusler, K. E.; Kutzmutz, S. J. Electroanal. Chem. Interfacial Electrochem. 1990,
285, 93–101.

44

CHAPTER 3:
IDENTIFYING THE PROBLEMS WITH LIQUID AMMONIA ELECTROLYSIS
ON PLATINUM ELECTRODES

Electrolysis of Liquid Ammonia for Hydrogen Generation, Little, D. J.; Smith, III, M. R.;
Hamann, T. W. Energy Environ. Sci. 2015, 8, 2775–2781. - Adapted by permission of
The Royal Society of Chemistry
45

Introduction
As of June, 2015, the only two reports in the literature investigating the
electrolysis of liquid ammonia (NH3(l)) were from the Ichikawa group who reported 7.2
mA cm-2 from the electrolysis of NH3(l) with 2.0 V applied between two platinum
electrodes at 9.63 bar and 25°C.1,2 This was only achieved with a relatively high
concentration (1 M) potassium amide (KNH2) electrolyte, which can detonate in solid
form. They demonstrated using gas chromatography that hydrogen gas (H2) and
nitrogen gas (N2) were the products generated at the cathode and anode respectively,
and attributed these products to the reduction of NH3 and oxidation of NH2–. However,
this mechanism was not substantiated. In addition, the two-electrode cell in those
studies could not determine separate overpotentials for the anodic and cathodic
reactions.
With these reports as a starting point, we began investigating the electrolysis of
ammonia (NH3)–both as NH3(l) and as NH3 dissolved in N,N-dimethylformamide
(DMF)–using platinum electrodes, with and without a third reference electrode. This
allowed us to identify the anodic and cathodic reactions occurring, as well as the
standard reduction potentials and overpotentials associated with these reactions, and
the current controlling processes. As a result, comparable electrolysis efficiencies were
realized with simpler and safer electrolytes. This allowed us to identify the specific
reaction barriers that needed to be surmounted in order to improve the electrolysis
efficiency of ammonia splitting.

46

Experimental
Electrochemical experiments were conducted in both NH3(l) and DMF as
described in Chapter 2. In this chapter, the NH3(l) was prepared without distillation over
sodium, and the Ag/AgNO3 reference electrode was used. All electrochemical
experiments were conducted using a Metrohm μAutolab(III) using the Nova 1.7 software
package.

Results and Discussion
Voltammetry in Liquid Ammonia
Figure 3.1 shows the cyclic voltammetry (CV) curves corresponding to the
electrolysis of NH3(l). The potential between the two disks was cycled between 0 and
2.0 V at 100 mV s-1 several successive times. From the curves depicted in Figure 3.1(a)
for NH4PF6 and 3.1(b) for KPF6, it is clear that the overall current decreases over the
course of the experiment, particularly during the first scan, for both NH4PF6 and KPF6
electrolytes. This is characterized both by the 2-fold decrease in the maximum current
at 2.0 V from scan 1 to scan 2, and also by the very large hysteresis in scan 1. This
behavior suggests that the surface of at least one of the Pt electrodes is changed over
the course of the experiment making it less active for the respective electrolysis half

47

Figure 3.1: The electrolysis of NH3(l) using platinum disk electrodes. Panes (a) and (b)
show successive cyclic voltammograms with 1.0 M NH4PF6 and 1.0 M KPF6 as
electrolytes, respectively, while pane (c) shows STVs for both conditions. The STV for
1.0 M KNH2 is also depicted in pane (c).
reaction. We attribute this fouling to a poisoned state which builds up on the platinum
surface during the ammonia oxidation half reaction, vide infra, analogous to species
originally suggested by Gerischer and Mauerer, and then confirmed by others for the
oxidation of NH3 in aqueous solution.5–7 Indeed, after intentionally poisoning a planar Pt
electrode, a strong surface and near-surface nitride signal was detected by Auger
electron spectroscopy (AES) as shown in Figure 3.2. The CVs of the poisoned electrode

48

(scans 2-5) are similar to that reported for the 1.0 M KNH2 electrolyte by Hanada et
al.,1,2 however with a slightly smaller current density.

Figure 3.2: AES spectra of a Pt electrode cleaned electrochemically in sulfuric acid (a),
and of a similar electrode cleaned electrochemically and then anodically poisoned in 1.0
M NH4PF6 in liquid ammonia (b). Sputtering with Ar+ allowed for depth-profiling.

The anodic poisoning of the Pt electrodes represents a major hurdle preventing
the accurate determination of the rate of ammonia electrolysis on Pt. Thus, in order to
determine the CV response for ammonia electrolysis in the absence of electrode
fouling, sampled time voltammetry (STV) measurements were employed. The current
onset potential is similar, but with an order of magnitude larger current density. The STV
with 1.0 M KNH2 was also measured, which showed similar performance to the other
two electrolytes.
During the electrolysis on two electrodes, the potential drop between the anode
and cathode must be the sum of the thermodynamic potential for the reaction, Erxn, and
the overpotentials for both the oxidation and reduction half reactions in order for the
49

reaction to be driven at a desired rate. But it is impossible to determine both the
oxidation and reduction overpotentials without the addition of a third, reference
electrode. In order to avoid fouling of the working electrode as seen in Figures 3.1(a)
and (b), and also to prevent large-amplitude noise resulting from the formation of
bubbles at the surface of the working electrode at high overpotentials, a similar STV
procedure to that described above was conducted with a three-electrode cell. The
anodic behavior from the STV measurements for three different concentrations of
NH4PF6 and KPF6 electrolytes, as well as 1.0 M KNH2 are plotted in Figure 3.3(a). After
measuring the curve of 1.0 M KPF6, sodium amide (NaNH2) was added to the solution
until saturated (approximately 6 mM), and the STV was re-measured. Figure 3.3(a)
shows the resulting data.

Figure 3.3: (a) STVs of NH3(l) comparing current densities at varying potential using
three different concentrations of either NH4PF6 (triangles) or KPF6 (circles) as the
supporting electrolyte, as well as 1.0 M KNH2 (black squares). For all STVs, the markers
represent actual points of data, and there are straight lines connecting them for trend
clarity. (b) CVs in NH3(l) with 1 M NH4PF6 electrolyte (red) and 1 M KPF6 electrolyte
(green) with a Pt disk working electrode at 100 mV/s.
50

The magnitude of current density at a given potential increases roughly
proportionally with the concentration of both KPF6 and NH4PF6 electrolytes. The anodic
waves for both electrolytes are linear between 0.7 and 2.2 V versus NHE. There is no
significant difference between the behavior of the KPF6 and NH4PF6 electrolytes.
Further, the addition of NH2– to the KPF6 electrolyte had no noticeable effect on the
anodic current, as the sparingly soluble salt changed the ionic strength of the solution
very little. These results are consistent with the rate of oxidation being limited by the
solution resistance. This is in general agreement with the previous report by Hanada et
al. of liquid ammonia electrolysis where the current density increased with increasing
concentration of amide electrolyte.1 In that paper, however, it was suggested that amide
ions were oxidized at the anode and ammonia reduced at the cathode.1 In Figure 3.3(a),
there is ten-fold higher current with 1.0 M NH4PF6 or KPF6 as the electrolyte compared
to when 1.0 M KNH2 is used. The auto-ionization constant for NH3 is approximately
10-33 M2, thus the concentration of NH2– in “neutral” NH3(l) is ~3 × 10-17 M. In a NH3(l)
solution with a 1.0 M NH4+ electrolyte, the concentration of NH2– is ~10-33 M. Upon the
addition of 6 mM NaNH2 to “neutral” NH3(l), the concentration of NH2– is ~10-3 M, and
the concentration of NH2– with 1.0 M KNH2 is 100 M. If NH2– were the species being
oxidized, the solution of 1.0 M KNH2 should show anodic current 1033 larger than the
solution of 1.0 M NH4PF6, where instead it is a factor of 10 smaller. This demonstrates
that the overpotential for oxidation of NH3 is the same as that for oxidation of NH2–, both
of which would produce a NH2(ads) species adsorbed on the electrode surface. The

51

overall reaction of NH3 (and NH2–) to form N2 is a three electron-transfer process. This
indicates that the rate-limiting step for N2 generation is not the first electron transfer.

Figure 3.4: Cathodic CVs in NH3(l) of 1.0 M KPF6 (green) and 1.0 M NH4PF6 (black).
Also included are STVs measured at 1 and 100 ms in 1.0 M NH4PF6 (dark red and light
red triangles respectively), and at the same times in 1.0 M KPF6 (dark blue and light
blue circles respectively).

The presence of NH4+ has a much more profound effect on the cathodic current
density at negative applied potentials in NH3(l). Figure 3.3(b) shows a cathodic wave
with an onset of approximately -2.5 V versus NHE for the 1.0 M KPF6 electrolyte which
is consistent with the current attributed to producing solvated electrons by Itaya et al.8
The formal potential for electron solvation was reported as -2.69 V versus Ag/AgNO3,
and the onset of the wave in Figure 3.3(b) was measured as -2.69 V versus the custom

52

Ag/AgNO3 reference. Clearly, NH3 is not the species being reduced at the cathode. The
cathodic current for the 1 M NH4PF6 exhibits a wave beginning at approximately -1.1 V
versus NHE; thus a current onset at ~1.4 V lower applied potential. The cathodic wave
in the presence of NH4+ must be due to either the reduction of protons dissociated from
NH4+, or the reduction of the NH4+ ion directly to NH40•. With NH3(l) solvent, there is a
negligible concentration of protons in the solvent, thus it is difficult to distinguish
between these two mechanisms. Some insight can be gleaned from the measured STV
curves, however, as the electrodes may have initial hydrogen species bound to the
electrode surface from the acid treatment as suggested by Figure 3.4. The STV curves
in both 1.0 M KPF6 and 1.0 M NH4PF6 exhibit a wave with an onset of approximately 0
V versus NHE, which is consistent with H(surface) reduction. This initial current decays
to the steady-state CV of the respective electrolyte once the surface hydrogen species
are consumed. Thus, we suggest reduction of the NH4+ ion directly to NH40• is the first
electron transfer step, which controls the cathodic overpotential. From these results, we
propose the electrolysis of NH3(l) to occur via the following scheme (equations 3.1–3.5):

1
NH3 ⎯anode
⎯⎯→ N2 + 3H+ + 3e −
2

Overall:

(3.1)

+
!!!
⇀
3H+ + 3NH3 ↽
!!
! 3NH4

(3.2)

3NH+4 + 3e − ⎯cathode
⎯⎯⎯
→ 3NH0•
4

(3.3)

3
3NH0•
⎯⎯
→ 3NH3 + H2
4
2

(3.4)

1
3
NH3 ⎯⎯
→ N2 + H2
2
2

(3.5)

53

Voltammetry in N,N-Dimethylformamide
In order to attain greater control over all redox active species in solution, the
electrolysis of NH3 was also conducted in DMF at room temperature. In DMF, the
concentration of H+ can be manipulated in addition to NH4+, as well as the concentration
of NH3 (which is 4 M at saturation).9 The electrolysis of a saturated NH3 solution on two
identical Pt disk electrodes–directly analogous to the electrolysis of Figure 3.1–is
depicted in Figure 3.5(a). The CVs have a similar shape and current magnitude to those
shown in Figure 3.1(c), although the difference between the currents in the TBAPF6 and
NH4PF6 electrolytes in DMF is more pronounced than the difference between the KPF6
and NH4PF6 electrolytes in NH3(l). Specifically, the CVs for the electrolysis in the
NH4PF6 electrolyte in both DMF and NH3(l) are very similar, the KPF6 current density is
slightly (but consistently) lower, and the current density with the TBAPF6 electrolyte is
significantly lower at high potentials. These discrepancies are likely due to differences in
ion mobility, especially between the small NH4+ cation and the bulky TBA+ cation.10,11 A
decreased mobility results in decreased solution conductivity, and thus lower current
densities at high potentials.
Analogous to the experiments in NH3(l), the oxidation and reduction half
reactions were isolated using a three-electrode cell with a conventional Ag/AgNO3
reference electrode. Figure 3.5(b) shows the CV and STV curves at positive potentials
versus NHE with NH3 dissolved in DMF. The STV has an onset of approximately 0.5 V
versus NHE, while the CV waves do not begin until about 1 V. Both of these waves are
attributed to the oxidation of dissolved NH3 as there is no anodic activity under the same

54

Figure 3.5: (a) STVs showing electrolysis of NH3 dissolved in DMF using platinum disks
as both the anode and the cathode. The red curve has 0.1 M NH4PF6, while the green
curve has 0.1 M TBAPF6. (b) CV (red) and STV (blue) of the anodic current with NH3
dissolved in 0.1 M NH4PF6 in DMF. In black is the CV without dissolved NH3. In green is
the CV with NH3 dissolved in 0.1 M TBAPF6. (c) CVs (black and green with NH4PF6 and
TBAPF6 respectively) and STVs of the cathodic current with NH3 dissolved in DMF. The
black and light and dark red curves contain 0.1 M NH4PF6 as the supporting electrolyte,
while the light and dark blue curves contain 0.1 M TBAPF6.

conditions in the absence of dissolved NH3. This large shift from the STV onset to the
CV onset is attributed to the 1 ms pre-poisoned surface similar to what was seen in
NH3(l).

55

Figure 3.5(c) shows the CV and STV curves at negative potentials versus NHE
with NH3 dissolved in DMF. Two STV curves are shown which were obtained from the
same set of measurements, sampled at different times: 1 ms and 100 ms. The STV
curves sampled at 1 ms both have a current onset of approximately 0 V versus NHE,
which is consistent with the assignment of initial reduction of surface adsorbed
hydrogen species. The STV curve of the NH4PF6 electrolyte at longer times is in good
agreement with the CV, which both have current onsets of approximately -0.6 V versus
NHE. The STV curve of the TBAPF6 electrolyte at longer times shows essentially zero
current at any potential, which is consistent with the CV curve in the same electrolyte.
These results agree with the assignment of the cathodic current arising from the direct
reduction of the NH4+ ion.
Voltammetry in DMF, instead of NH3(l), allows for further elucidation of the
reduction of NH4+. Figure 3.6(a) shows a quasi-reversible CV of a solution containing 20
mM NH4PF6 and 100 mM TBAPF6 in DMF, corresponding to the reduction of NH4+ and
oxidation of the reduced product. A formal potential of -0.60 ± 0.04 V versus NHE was
determined for NH4+ in DMF, which is consistent with the position of the wave for NH4+
reduction seen in aqueous conditions by Berkh et al.12 However, to the best of our
knowledge, this is the first quasi-reversible CV of NH4+ reduction in non-aqueous
conditions suitable for reporting a formal potential. The reduction potential of NH4+ also
coincides with the cathodic current onset potentials with NH4+ shown in Figure 3.5(c).
Tafel analysis of the black CV in Figure 3.5(c) suggests a single-electron transfer
process, further indicating that the cathodic wave results from the direct reduction of

56

NH4+ to NH40• (Figure 3.7). This is contrary to the mechanism reported by Buzzeo et al,
who suggested that the reduction wave for NH4PF6 in DMF is the result of reducing H+
that had dissociated from the NH4+.13 Their cathodic CV actually has a similar shape to
that of Figure 3.6(a), but since it is reported using a silver wire pseudo-reference, it is
impossible to quantitatively compare the two plots.

Figure 3.6: (a) Quasi-reversible redox wave for NH4+ in DMF. (b) Cathodic waves
measured in 0.1 M NH4PF6 with increasing concentrations of HClO4 using a Pt disk
electrode in DMF.

In order to confirm that the cathodic wave is not due to a coincidentally large
reduction overpotential of protons that have dissociated from the weak acid, small
aliquots of perchloric acid (HClO4), which can be thought of simply as a source of H+,
were added to a solution with 0.1 M NH4PF6 in DMF. The CVs of these solutions are
depicted in Figure 3.6(b). A cathodic wave with an onset of about -0.1 V versus NHE
becomes larger as the H+ concentration is increased, and thus is attributed to the
reduction of H+ in solution. There is also a second cathodic wave with an onset of about
57

Figure 3.7: Tafel plot of the CV shown in black in Figure 3.5(c). The value of ! close to
0.5 indicates a likely 1-electron reaction, quasi-reversible reaction.

-0.7 V versus NHE whose magnitude with respect to baseline does not change with
increasing H+ concentration. If the mechanism for the reduction of NH4+ first involved
the dissociation of H+, followed by the reduction of H+, (like that previously reported),13
increasing the concentration of H+ would cause the cathodic wave to shift to more
positive potentials, not just introduce a second wave. This wave is therefore attributed to
the direct reduction of NH4+ as shown in the equations 3.6 through 3.8:
3NH+4 + 3e − ⎯⎯
→ 3NH0•
4

(3.6)

3NH0•
⎯⎯
→ 3NH3 + 3H•
4

(3.7)

3
3H• ⎯⎯
→ H2
2

(3.8)

58

This scheme is also consistent with that of Berkh et al., where they emphasized
that the NH40• species could be stabilized by remaining adsorbed on the Pt electrode
surface.12 The net reaction shown in (3.6) through (3.8) corresponds to reactions (3.3)
and (3.4) proposed for electrolysis of NH3(l). It is reasonable to extend this proposed
mechanism of direct NH4+ reduction as the overpotential determining step for the
cathodic reaction in NH3(l) as well, where even with the same 0.1 M NH4PF6 the
concentration of H+ is lower than in DMF by one order of magnitude. This reaction also
largely accounts for the differences in overpotential between liquid ammonia electrolysis
and aqueous ammonia electrolysis. In aqueous systems, the reduction of water to H2
determines the overpotential, and good catalysts for this reaction are well-developed. In
NH3(l) it is the reduction of NH4+ to H2 which determines the cathodic overpotential, and
to the best of our knowledge, no effort has been invested to catalyze this reaction.

Conclusions
In order to make the electrolysis of NH3 a feasible source of fuel for a hydrogen
economy, several kinetic hurdles must be overcome. The first of these is the
phenomenon of poisoning. An anode material must be found where the overpotential for
ammonia oxidation is the same 1 s into the electrolysis as it is after 1 µs. This may
involve either a pretreatment of an existing electrode material to keep the surface
identity constant, or development of a new anode material that is a superior catalyst for
the oxidation of NH3. The overpotential gains from such a material are shown with the
arrow labeled “Poison” in Figure 3.8.

59

Figure 3.8: Diagram showing where improvements can be made by either avoiding
electrode poisoning (indicated by arrow with “poison” icon), or via catalysis (indicated by
arrows with “catalysis” ions). The formal potentials for NH4+ reduction, H+ reduction, and
NH3 oxidation are indicated on the top black scale. Both the STV (red) and CV (green)
were recorded using a Pt disk electrode in 0.1 M NH4PF6 in DMF.
As seen in Figure 3.1(c), the electrolysis of NH3 is already proceeding to a small
degree after only 100 mV of applied potential (the first point collected in the STV). This
is consistent with the formal potential of ammonia oxidation, which can be deduced from
the difference of the formal potential for the reduction of protons (0.0 V versus NHE)
and the free energy change of ammonia electrolysis. For this reason, the formal
potential for NH3 oxidation is 0.10 V versus NHE as shown on Figure 3.8.
Once electrode poisoning is no longer a concern, a catalyst is still needed to
lower the overpotential for both NH3 oxidation, as well as NH4+ reduction as much as
60

possible. The overpotential gains from this catalysis are shown with arrows labeled
“Catalysis” in Figure 3.8. In addition, a catalyst needs to be developed that will allow for
the reduction of H+ instead of the initial reduction NH4+ to NH40•. That would allow for a
sufficient rate of proton reduction closer to the thermodynamic limit of 0 V versus NHE.
Catalytic materials for both of these reactions–based on Earth-abundant metals–are the
subjects of subsequent chapters.

61

REFERENCES

62

REFERENCES

(1)

Hanada, N.; Hino, S.; Ichikawa, T.; Suzuki, H.; Takai, K.; Kojima, Y. Chem.
Commun. (Camb). 2010, 46, 7775–7777.

(2)

Dong, B.-X.; Ichikawa, T.; Hanada, N.; Hino, S.; Kojima, Y. J. Alloys Compd.
2011, 509, S891–S894.

(3)

East, G. A.; del Valle, M. A. J. Chem. Educ. 2000, 77, 97.

(4)

Bard, A. J.; Faulkner, L. R. Electrochemical Methods: Fundamentals and
Applications; 2nd ed.; Wiley and Sons Inc., 2000.

(5)

Gerischer, H.; Mauerer, A. Electroanal. Chem. Interfacial Electrochem. 1970, 25,
421–433.

(6)

De Vooys, a. C. a; Koper, M. T. M.; Van Santen, R. a.; Van Veen, J. a R. J.
Electroanal. Chem. 2001, 506, 127–137.

(7)

Le Vot, S.; Reyter, D.; Roué, L.; Bélanger, D. J. Electrochem. Soc. 2012, 159,
F91.

(8)

Itaya, K.; Malpas, R. E.; Bard, A. J. Chem. Phys. Lett. 1979, 63, 411–415.

(9)

Gerrard, W. Solubility of Gases and Liquids: A Graphic Approach; 1976.

(10)

Izutsu, K. Electrochemistry in Nonaqueous Solutions; First.; Wiley-VCH:
Weinheim, 2002.

(11)

Krumgalz, B. S. J. Chem. Soc. Faraday Trans. 1 1983, 79, 571.

(12)

Berkh, O.; Shacham-Diamand, Y.; Gileadi, E. J. Electrochem. Soc. 2008, 155,
F223.

(13)

Buzzeo, M. C.; Giovanelli, D.; Lawrence, N. S.; Hardacre, C.; Seddon, K. R.;
Compton, R. G. Electroanalysis 2004, 16, 888–896.

63

CHAPTER 4:
INVESTIGATION OF IRON NITRIDE AS AN EARTH-ABUNDANT
ELECTROCATALYST FOR THE OXIDATION OF LIQUID AMMONIA

Adapted with permission from:
As Precious as Platinum: Iron Nitride for Electrocatalytic Oxidation of Liquid Ammonia,
Little, D. J.; Edwards, D. O.; Smith, M. R.; Hamann, T. W. ACS Appl. Mater. Interfaces
2017, 9, 16228–16235. Copyright 2017 American Chemical Society.
64

Introduction
We showed previously that the electrolysis of liquid ammonia (NH3(l)) occurs on
platinum (Pt) electrodes via the half reactions depicted in equations (4.1) through (4.3).1
E0 for (4.1) and (4.2) were reported to be 0.1 V and -0.6 V versus NHE respectively,
resulting in an E0cell for the overall NH3(l) splitting process in (4.3) of 0.7 V. There is a
large overpotential of over 1 V needed to drive the anodic half reaction (4.1) at the rates
that would be required for on-site NH3(l) electrolysis. This barrier arises from both
anodic poisoning and a lack of efficient catalysts. Thus, the problem that we aim to
address is the development of a highly-catalytic material that resists anodic poisoning.
Further, if the electrolysis of NH3 is going to be economically competitive in the energy
market, the process must be as inexpensive as possible. Pt is the only anode material
that has really been studied in depth for the electrolysis of NH3(l), however.1,2 The most
efficient NH3 electrolysis device reported operates in aqueous conditions–which results
in a loss of energy density compared to NH3(l) and is corrosive to storage and
distribution media–and uses precious metals such as platinum, ruthenium, and iridium
for the anode.3,4 Remarkably, the electrolysis of NH3(l) using Earth-abundant materials,
such as first-row transition metals, remains essentially unstudied.

1
4NH3 ⎯anode
⎯⎯→ N2 + 3NH+4 + 3e −
2

(4.1)

3
3NH+4 + 3e − ⎯cathode
⎯⎯⎯
→ H2 + 3NH3
2

(4.2)

1
3
NH3 ⎯overall
⎯⎯→ N2 + H2
2
2

(4.3)

65

Nickel (Ni)–the first-row transition metal in Group 10 above Pt–is a natural first
choice for investigating Earth-abundant materials. A Ni/Ni(OH)2 electrode was shown to
be active for NH3 oxidation in aqueous conditions.5 Unfortunately, Ni corrodes easily
with positive applied bias in NH3(l) forming the stable hexamine complex, [Ni(NH3)6]2+.6,7
The neighbors of Ni in the first-row of the periodic table, cobalt (Co) and copper (Cu),
also rapidly corrode in NH3(l) and form stable ammine complexes.8–11 Interestingly,
unlike the ammine complexes of Ni, Co, and Cu, iron hexamine is not very stable,
converting to iron oxide upon exposure to air.12,13 The high reactivity of iron hexamine
makes iron (Fe) a potentially advantageous element for NH3 catalysis, since the
formation of stable amines could cripple catalytic rates. Also, there are known catalysts
for the Haber-Bosch process–the formation of NH3 from nitrogen and hydrogen–
employing Fe.14 However, to the best of our knowledge, there are no reports in the
literature where elemental Fe is employed as an anode for the electrolysis of NH3.
By comparison, the corrosion behavior of various transition metals in NH3(l) have
been better studied.6,15–18 This work was largely motivated by the commercial interest in
storing NH3(l) in stainless steel containers. Thus, corrosion studies have focused
primarily on stainless steel components such as nickel, zinc, and iron. The Fe corrosion
study by Ahrens et al. demonstrated two separate potential regions for corrosion
(evolving Fe2+ and Fe3+ respectively) with a large, 0.5 V wide passive region between
them.16 They also discussed the possibility of nitridation of the Fe anode surface, which
could possibly serve as a catalyst for NH3 oxidation.

66

In 2016, Goshome et al. reported the electrolysis of NH3(l) with a 5 M ammonium
chloride electrolyte using a stainless steel anode.19 They observed similar electrolysis
current density with stainless steel and Pt anodes, but reported that over time the steel
anode corroded with positive applied bias to form [Fe(NH3)6]Cl2 in solution. However,
there are numerous varieties of stainless steel containing different additives in varied
amounts. Therefore, mechanistic conclusions on the chemistry of electrolysis on
stainless steel cannot be made without studying electrolysis using pure Fe anodes.
Additionally, since the desired electrolysis was reported in addition to corrosion, it is
important to identify all of the anodic chemical pathways in order to design the ideal
electrode.
In this chapter, we compare the electrochemical behavior of elemental Pt and Fe
anodes in stock NH3(l) versus distilled NH3(l). These results illustrate how the presence
of even a small amount of water in NH3(l) can critically change the poisoning behavior of
electrodes. The extent of different anodic processes on Fe–both corrosion and NH3(l)
oxidation–are determined for multiple potential regimes. Finally, an in situ modified Fe
electrode with an anodic performance similar to Pt in NH3(l) is characterized.

Experimental
Electrochemical experiments were conducted in NH3(l) as described in Chapter
2. Before each experiment, the anhydrous ammonia was distilled over sodium before
condensing in a pre-assembled electrochemical cell under inert atmosphere. The
custom Ag/AgCl reference electrode with an electrolyte saturated with both NH4Cl and

67

dissolved NH3(g) was used in all electrochemical experiments in this chapter. The Fe
working electrodes used were manufactured using Fe wire (1.0 mm, 99.9%, Aldrich). All
electrochemical experiments were conducted using a Gamry Reference 600
potentiostat.

Results and Discussion
Figure 4.1(a) shows the anodic current density versus applied potential curves
derived from cyclic voltammetry (CV) measurements which correspond to the oxidation
of NH3(l) on a Pt disk electrode. Only one scan is shown in Figure 4.1(a) for Pt in pure
distilled NH3(l) because successive scans all overlapped perfectly. When an aliquot of
degassed Millipore water was added to the distilled NH3(l) to make the solution 10 %
water, however, the previously reported poisoning behavior was reproduced, indicated
by decreasing anodic current in successive scans.1 This behavior was not evident in the
pristine distilled NH3(l) electrolyte. We showed in Chapter 3 via Auger electron
spectroscopy that the formation of a nitrogen-containing species accompanied the
anodic poisoning of Pt in NH3(l). However, all experiments in that study were performed
using stock NH3(l), rather than distilled, and the headspace of the flask was exposed to
the air. It is now clear that water is crucial to the poisoning process. An additional anodic
process involving both NH3 and H2O at the Pt surface would be consistent with
“nitrogen-containing”. Indeed, adsorbed OH– ions from H2O have been proposed to
have a role in the mechanism of anodic poisoning of Pt surfaces in aqueous NH3

68

oxidation by Gerischer and Mauerer.20 It appears this mechanism extends to the related
process in NH3(l) as well.

Figure 4.1: (a) CV of a Pt disk electrode in distilled NH3(l) (black), and five successive
CVs of a Pt disk electrode in distilled NH3(l) after the addition of degassed Millipore
water to make the solution 10 % water (colors). (b) CVs of an Fe disk electrode in the
same conditions as (a), except that only 1 % water was added. The vertical black
arrows indicate the order of successive anodic scans. Dashed boxes indicate the
overpotential required to reach 10 mA cm-2 current. Small, black, diagonal arrows
indicate the scan direction. All curves were measured at 100 mV s-1 with 0.1M KPF6 as
the supporting electrolyte.

From this result, we hypothesize that water is a required ingredient for electrode
poisoning, and use of distilled NH3(l) is essential for studying NH3 oxidation at
transitional metal anodes. As can be seen in Figure 4.1(b), a polished Fe disk anode
behaves profoundly different in distilled NH3(l) than in the presence of trace amounts of
water, even with an order of magnitude less water than was needed to illustrate the
effect with Pt. In rigorously dry solutions, the current onset appears to be approximately
the same as the Pt anode. In the presence of water, however, the Fe anodes are

69

poisoned until the measured current is extremely small. All further experiments were
conducted in distilled (dry) NH3(l).
Interestingly, the CV response of Fe electrodes shown in Figure 4.1(b) does not
correspond to the initial CV of a pristine electrode. Figure 4.2 shows the initial and
subsequent CVs measured in distilled NH3(l) leading-up to that pictured in Figure 4.1(b).
It is clear that in the absence of water, the anodic current is initially higher than in the
presence of trace water, but then increases with successive scans and eventually
reaches a constant, reproducible response. Note that even for this CV, there is a large
hysteresis present at the foot of the wave. This indicates a kinetically slow process
occurring at the electrode where the current takes several seconds to increase even
when sufficient overpotential is applied.
The slow kinetics of this process are evident in the chronoamperograms in Figure
4.3, where several different specific potentials were held for 180 seconds each on a
fresh Fe disk. Even at the highest overpotentials, the current takes several seconds to
reach steady-state, but at the lowest overpotentials steady-state is barely reached by
the full 180 seconds. Simple electron-transfer processes do not generally involve
kinetics on a seconds time-scale, implying that the chemistry at the Fe surface is more
complicated.
The most obvious process that can compete with NH3(l) oxidation is the corrosion
of the Fe electrode. A detailed study of the corrosion of Fe in NH3(l) was published by

70

Figure 4.2: Successive CVs of distilled NH3(l) on an Fe disk electrode with the order of
the scans indicated by the vertical black arrow. All scans were measured at 100 mV s-1
with 0.1 M KPF6 as supporting electrolyte.

Ahrens et al.16 Utilizing a distillation procedure similar to ours and Fe wire anodes, they
identified two different corrosion processes for Fe in NH3(l): One process occurs at
relatively low potentials (reported around 0 V versus Tl/TlCl, or approximately -0.34 V
versus NHE)21 where the anode dissolves into Fe2+ ions in solution. A second process
is seen at relatively high potentials (reported around 1 V versus Tl/TlCl, which is
approximately 0.66 V versus NHE)21 where the anode dissolves as yellow/brown Fe3+
ions in solution. In the potential range between these two processes–roughly 0.5 to 0.9
V versus Tl/TlCl–there is a passive region where corrosion is reported to not occur.
While it was mentioned that nitrogen (N2) evolution occurs simultaneously with

71

Figure 4.3: Chronoamperograms conducted on an Fe disk at seven different potentials
near the onset of the anodic wave shown in Figure 4.2. Potentials here are reported
versus NHE, and verified using electron solvation between experiments. All scans used
a 0.1 M KPF6 electrolyte. Periodic sudden jumps in the current seen at the highest four
potentials are due to the effervescence of evolved gas bubbles, thus suddenly changing
the exposed surface area.

the corrosion process that leads to Fe3+, the authors did not comment on the relative
efficiencies of the Fe3+ and N2 forming processes. We therefore determined the
efficiency for Fe3+ dissolution by applying a given potential on an Fe electrode,
measuring the molar quantity of Fe ions in solution, and comparing that to the number
of electrons passed during the experiment. Figure 4.4 shows a plot of the current
efficiency of anodic dissolution as a function of the net charge passed at the anode. A
potential of 1.2, 1.4, or 1.6 V versus NHE was held at an Fe disk electrode in a solution
of 0.1 M NH4NO3 in distilled NH3(l) for varying times, where each individual experiment
72

is represented by one point in Figure 4.4. Current variation with each experiment was
compensated by simply potting the time-integrated current versus charged passed in
Coulombs (C). The background Fe–i.e. any Fe dissolved in solution from passive
processes due to Fe in the glass, supporting electrolyte, or from the electrode during
preparation–was constant with time. This constant background was therefore subtracted
from every point when calculating the dissolution efficiency.

Figure 4.4: Current efficiencies for the dissolution of an Fe disk electrode in distilled
NH3(l) as a function of the net charge passed at the working electrode during the
experiment. Potentials of 1.2, 1.4, or 1.6 V versus NHE were applied for varying
amounts of time and the current was integrated to obtain the total charge passed. For
all points, 0.1M NH4NO3 was used as the supporting electrolyte to avoid the presence of
fluorine from the PF6– anion.

73

At the lowest potential of 1.2 V–the potential corresponding to the wave onset
from Figure 4.2–the initial current efficiency for Fe dissolution is approximately 100 %.
The calculated current efficiency was actually larger than 100 % for the experiment with
the smallest amount of charge passed. This could be due to the relatively large error in
collecting and measuring such a small quantity of dissolved Fe (2.5 x 10–7 moles), an
underestimate in the amount of expected Fe as a result of Fe2+ instead of Fe3+
dissolution, and/or the dissolution of stray Fe0 or surface FeOx particles as a result of
the rapid, aggressive changes in electrode surface morphology. The current efficiency
for Fe dissolution seems to level-off at about 20 percent on approximately the same
time-scale as the steady-state current formation shown in Figure 4.3. At the higher
potential of 1.4 V, there is no initial spike in dissolution efficiency. Instead it quickly
levels off to values less than 5 percent. Essentially the same behavior was seen at 1.6
V. These results are in agreement with the suggestion by Ahrens et al. that there are
other chemical processes competing with Fe3+ dissolution.16 At lower overpotentials,
e.g. 1.2 V versus NHE, nearly all of the anodic current initially goes toward Fe3+
dissolution. However, as time passes, other chemical processes compete with
dissolution and lower its efficiency. As the potential increases further–past 1.2 V versus
NHE–the rate of the other chemical processes dominates the rate of dissolution,
resulting in only a very small fraction of current going toward Fe3+ dissolution.
If at steady-state only this small fraction of the anodic current is going toward
dissolution of Fe3+, the remaining current must be involved in electrode surface
modification and/or oxidation of the NH3(l) solvent to N2. Bubbles rapidly evolved from a

74

large Fe anode at 1.4 V versus NHE, suggesting N2 evolution. GC analysis of the
headspace in the electrochemical cell during a similar reaction indicates an N2:H2 ratio
of 1:7 at the beginning of the experiment, which approached 1:4 at the end (Figure 4.5).
This ratio suggests that while H2 evolution is always the primary reaction occurring at
the Pt cathode, another process competes with N2 evolution at the anode–otherwise the
evolution ratio would be closer to the theoretical value of 1:3 for pure NH3(l) electrolysis.
These results agree with the findings of the ICP experiments in Figure 4.4, and since a
potential slightly lower than 1.4 V versus NHE was used, the imperfect ratio of evolved
gasses is expected. Unfortunately, it is difficult by this method to precisely relate the
measured N2 to an evolution efficiency at the anode as the solubility of N2 in the NH3(l)
solvent depends upon the exact temperature of the reaction vessel, and because the
exact pressure in the headspace of the cell was not measured and will fluctuate with the
temperature of the solvent as well. However, if we assume that all of the evolved N2 is
present in the headspace of the flask, and compare the moles of N2 measured to the
coulombs of charge passed at the electrode, we can calculate that there is about a 20%
anodic faradaic efficiency initially, which increases to over 71% by the end of the
experiment.
Since a large concentration of Fe3+ (presumably in the form of [Fe(NH3)6]3+) is
dissolved in solution after steady-state has been reached–as was observed by the
appearance of a yellow color–it is conceivable that the dissolved Fe acts as a catalyst
for oxidation to N2. This would explain the increase in current as the Fe dissolves, as

75

Figure 4.5: N2 and H2 gas measured by GC from the headspace of the electrochemical
cell. Measurements were taken every 30 minutes with a potential of 1.35 V versus NHE
applied during the intervals. Molar amounts approximately correspond to the gas
present in the headspace of the cell. The supporting electrolyte was 0.1 M NH4NO3.
(Inset) The H2:N2 evolution ratio over the course of the experiment.

well as the decreased overall efficiency for dissolution as the catalyzed N2 evolution
process becomes dominant. In order to test this possibility, NH3(l) oxidation was
measured with a Pt disk electrode while an increasing concentration of Fe3+ introduced.
There was no change in the CVs in response to the continually increasing Fe3+
concentration, displayed in Figure 4.6. Thus, dissolved Fe3+ species such as
[Fe(NH3)6]3+ do not appear to act as a catalyst.

76

Figure 4.6: Anodic current of the Pt disk electrode in distilled NH3(l) before and after
dissolution of Fe electrode. Five essentially overlapping scans are depicted.

An increase in current density calculated during the initial CVs may simply arise
from corrosion increasing the active surface area. The anodic corrosion of Fe in distilled
NH3(l) does result in pitting, as clearly shown by SEM images in Figure 4.7, rather than
uniform surface corrosion. As a result, the active surface area of the anode increases.
Since only the geometric surface area is accounted for when calculating current density,
the unaccounted increase in actual surface area causes the apparent current density to
be inflated. We note, however, that for practical applications a higher surface area
electrode such as the one that develops is desired. From CV capacitance
measurements taken between anodic sweeps similar to those shown in Figure 4.2, the
active surface area of the Fe electrode was found to increase with successive scans

77

Figure 4.7: SEM images of Fe foil after use as an anode in 0.1 M NH4NO3 in dry NH3(l).

Figure 4.8: The surface area of an Fe disk electrode as a function of the number of
wide cathodic scans measured. The surface area was measured from CV capacitance
between each of the wide scans.

78

(Figure 4.8). However, the increase in surface area approached a limit with an area
approximately 25% larger than the initial area. This is consistent with the concomitant
decrease in dissolution efficiency as well as the H2:N2 evolution ratio approaching 3:1.
Another possibility is that modification of the electrode surface consumes some
of the current. Ahrens et al. proposed the formation of Fe4N, which they suggested
would inhibit further corrosion, however did not provide direct evidence.16 Indeed, after
extended periods of positive applied potential a black film forms on the Fe electrode
surface, which could indicate a change in its composition (Figure 4.9). Since the
resulting film was too amorphous for characterization with x-ray diffraction (XRD), and
too thin for detection with energy dispersive x-ray spectroscopy (EDS), x-ray
photoelectron spectroscopy (XPS) measurements of the modified film were used to
determine the composition of the modified electrode. Three peaks are resolved in the
Nitrogen 1s region at 406.9 eV, 399.4 eV, and 396.4 eV, shown in Figure 4.10a. The
peak at 406.9 (violet) can be assigned to the small amount of residual NO3– from the

Figure 4.9: Photos of custom Fe disk electrodes taken before (a) and after (b)
electrolysis showing the formation of a black film covering the electrode surface.

79

Figure 4.10: XPS of the Nitrogen 1s region (a), the Fe 2p region (b), and the Oxygen 1s
region (c) of an in situ modified Fe electrode surface.

electrolyte.22–24 Since most of the electrolyte was rinsed away with fresh, distilled NH3(l)
this peak is barely distinguishable from the background. The peak at 399.4 eV (blue) is
assigned

to

an

NHx

species

bonded

to

surface

Fe

atoms.23,25,26

Unfortunately, it is difficult to determine from these XPS measurements whether this
feature is due to bound NH3 or N2H4, or covalently bound –NH2. The peak at 396.4 eV
(green) is assigned to FeNx–possibly Fe4N–due to its close proximity to previous reports
of FeNx.27–30 We note that while the peak at 396.4 eV has nearly a full eV lower binding
energy than most reports for FeNx, it is still in the same region as other metal nitrides.31
80

In the Fe 2p region, there are two peaks comprising the 2p3/2 spin state–at 710.6
and 712.2 eV–and two corresponding peaks comprising the 2p1/2 spin state–at 724.0
and 725.9 eV. The satellite peaks at 718.9 and 731.6 eV (violet and pink respectively)
each were only fit to one peak for simplicity due to the low signal. Since the
corresponding 2p3/2 and 2p1/2 peaks were fit with the positions and heights constrained
to fit the appropriate ratio, the use of only one peak to fit each of the satellites should
not have affected our interpretation. The lowest-energy peak of 707.7 eV (orange) is
assigned to the small amount of exposed atomic Fe0. The lower-energy peak pair of
710.6 and 724.0 (dark and light blue) is assigned to Fe2O3 which likely formed upon
exposure of the electrode to air. The corresponding Fe2O3 peak in the Oxygen 1s region

Figure 4.11: Raman spectrum of in situ modified iron anode after exposure to the air on
the lab bench overnight.
81

is found in Figure 4.10(c) at 530.0 eV (blue). This assignment is confirmed by Raman
spectroscopy measurements, shown in Figure 4.11.32 The second peak comprising the
Oxygen 1s feature at 530.9 eV (green) is due to adventitious CO which is present in
nearly every XPS spectrum. Since there are only two Oxygen states detected, the
second pair of peaks in the Fe 2p region–at 712.2 and 725.9 eV (dark and light green)–
are assigned to FeNx. The Fe-N bond is less polar than the Fe-O bond, consistent with
the slightly higher binding energy of these FeNx photoelectrons compared to the Fe2O3
photoelectrons assigned above. Thus, the combined XPS results of N and Fe regions

Figure 4.12: Steady-state sampled time voltammograms (STVs) of anodic behavior of
Pt and in situ modified Fe in distilled NH3(l). All points are sampled from 180 seconds
into a chronoamperogram at the respective potential and are connected with straight
lines.
82

confirm that the Fe electrode forms FeNx in response to applied bias in NH3(l). The
formation of this nitride film is likely what inhibits further electrode dissolution or pitting.
The FeNx electrode actually shows strikingly similar anodic performance to a Pt disk
electrode at steady-state in distilled NH3(l) as depicted in Figure 4.12. Sampled time
voltammetry (STV) was used to compensate for the slight decrease of the Pt electrode
current with time and the drastic increase of the Fe electrode current in the same time.1
Within the error of the STV method used, the onsets of NH3(l) oxidation are exactly the
same for both Pt and FeNx.

Conclusions
Removing the small amount of water that is present in stock NH3(l) strongly
affects the electrolysis to N2 and H2 on non-noble, Earth-abundant metal anodes like Fe.
At very short times in distilled NH3(l), corrosion of the Fe competes with electrolysis,
thus lowering current efficiency. However, the surface converts to FeNx during the
electrolysis of distilled NH3(l), which is far more resistant to further corrosion. While the
FeNx anode continues to dissolve with applied positive bias, the rate of N2 evolution
increases much faster than the rate of dissolution. Thus, at the high overpotentials used
to drive rapid NH3(l) electrolysis, the efficiency of corrosion is rather low. Even though
the anodic overpotential required for the in situ generated, nanostructured FeNx surface
to reach a 10 mA cm-2 current density exceeds 1 V, its performance is similar to Pt
electrodes.1 This demonstrates that noble metals are not a requirement for efficient NH3
electrolysis. Our FeNx anode results open the door for research to develop Fe and other

83

Earth-abundant electrode materials for NH3 oxidation, which is the subject of ongoing
investigation in our labs.

84

REFERENCES

85

REFERENCES

(1)

Little, D. J.; Smith, III, M. R.; Hamann, T. W. Energy Environ. Sci. 2015, 8, 2775–
2781.

(2)

Hanada, N.; Hino, S.; Ichikawa, T.; Suzuki, H.; Takai, K.; Kojima, Y. Chem.
Commun. (Camb). 2010, 46, 7775–7777.

(3)

Wang, Y.; Guo, X.; Li, J.; Yang, Y.; Lei, Z.; Zhang, Z. Open J. Appl. Sci. 2012, 2,
241–247.

(4)

Vitse, F.; Cooper, M.; Botte, G. G. J. Power Sources 2005, 142, 18–26.

(5)

Kapałka, A.; Cally, A.; Neodo, S.; Comninellis, C.; Wächter, M.; Udert, K. M.
Electrochem. commun. 2010, 12, 18–21.

(6)

Heusler, K. E.; Kutzmutz, S. J. Electroanal. Chem. Interfacial Electrochem. 1990,
285, 93–101.

(7)

Bergstrom, F. W. J. Phys. Chem. 1924, 29, 160–165.

(8)

Vijh, A. K. J. Electrochem. Soc. 1972, 119, 861.

(9)

Brown, O. R.; Thornton, S. A. J. Chem. Soc. Faraday Trans. 1 Phys. Chem.
Condens. Phases 1974, 70, 14.

(10) Jolly, W. L. J. Am. Chem. Soc. 1956, 78, 4849–4849.
(11) Brown, O. R.; Thornton, S. A. J. Chem. Soc. Faraday Trans. 1 Phys. Chem.
Condens. Phases 1974, 70, 1269.
(12) Klocke, D. J.; Hixson, A. N. Ind. Eng. Chem. Process Des. Dev. 1972, 11, 141–
146.
(13) Asch, L.; Shenoy, G. K.; Friedt, J. M.; Adloff, J. P.; Kleinberger, R. J. Chem. Phys.
1975, 62, 2335.
(14) Smil, V. Enriching the Earth: Fritz Haber, Carl Bosch, and the Transformation of
World Food Production; MIT Press, 2001.
(15) Jones, D. A.; Wilde, B. E. Corrosion 1977, 33, 46–50.

86

(16) Ahrens, M.; Schmitt, W.; Heusler, K. E. Mater. Corros. 1981, 32, 167–173.
(17) Miles, M. H.; Kellett, P. M. J. Electrochem. Soc. 1968, 115, 1225–1227.
(18) Ahrens, M.; Heusler, K. E. Electrochim. Acta 1982, 27, 239–243.
(19) Goshome, K.; Yamada, T.; Miyaoka, H.; Ichikawa, T.; Kojima, Y. Int. J. Hydrogen
Energy 2016, 2–7.
(20) Gerischer, H.; Mauerer, A. Electroanal. Chem. Interfacial Electrochem. 1970, 25,
421–433.
(21) Bard, A. J.; Faulkner, L. R. Electrochemical Methods: Fundamentals and
Applications; 2nd ed.; Wiley and Sons Inc., 2000.
(22) Folkesson, B. Acta Chem. Scand. 1973, 27, 287–302.
(23) Hendrickson, D. N.; Hollander, J. M.; Jolly, W. L. Inorg. Chem. 1969, 8, 2642–
2647.
(24) Aduru, S.; Contarini, S.; Rabalais, J. W. J. Phys. Chem. 1986, 90, 1683–1688.
(25) Campbell, J. H.; Bater, C.; Durrer, W. G.; Craig, J. H. Thin Solid Films 1997, 295,
8–10.
(26) Binder, H.; Sellmann, D. Z. Naturforsch 1977, 33b, 173–179.
(27) Torres, J.; Perry, C. C.; Bransfield, S. J.; Fairbrother, D. H. J. Phys. Chem. B
2003, 107, 5558–5567.
(28) Biwer, B. M.; Bernasek, S. L. J. Electron Spectros. Relat. Phenomena 1986, 40,
339–351.
(29) Cocke, D. L. J. Electrochem. Soc. 1989, 136, 3655–3662.
(30) Wang, X.; Zheng, W. T.; Tian, H. W.; Yu, S. S.; Xu, W.; Meng, S. H.; He, X. D.;
Han, J. C.; Sun, C. Q.; Tay, B. K. Appl. Surf. Sci. 2003, 220, 30–39.
(31) Naumkin, A.; Kraut-Vass, A.; Gaarenstroom, S.; Powell, C. NIST X-ray
Photoelectron Spectroscopy Database, 2012, Version 4.
(32) de Faria, D. L. A.; Venâncio Silva, S.; de Oliveira, M. T. J. Raman Spectrosc.
1997, 28, 873–878.

87

CHAPTER 5:
EVALUATION OF AMORPHOUS COBALT PHOSPHIDE AS AN EARTHABUNDANT AMMONIUM REDUCTION ELECTROCATALYST

88

Introduction
As we previously showed, the reduction of ammonium (NH4+) on a platinum (Pt)
cathode in NH3(l) occurs by the reactions shown in equations (5.1) through (5.3).1 Berkh
et al. proposed that the NH40• formed in (5.1) is likely stabilized by adsorption to the Pt
cathode surface, though they were unable to demonstrate this experimentally.2 The sum
of equations (5.1) through (5.3) is equivalent to (5.4), but the reduction of NH4+ directly
to the radical species results in a required potential of -0.6 V versus NHE for the
reduction process. Thus, in order to allow for the economic viability of on-site NH3(l)
electrolysis, a new cathode material is needed that will catalyze the hydrogen (H2)
evolution reaction in NH3(l), and ideally be comprised of Earth-abundant elements,
rather than precious metals like Pt.

Overall:

3NH+4 + 3e − ⎯⎯
→ 3NH0•
4

(5.1)

3NH0•
⎯⎯
→ 3NH3 + 3H•
4

(5.2)

3
3H• ⎯⎯
→ H2
2

(5.3)

1
NH+4 + e − ⎯⎯
→ H2 + NH3
2

(5.4)

There is a large body of research into alternative H+ reduction materials in
aqueous media employing Earth-abundant elements as a replacement for long-used
and highly efficient Pt.3 A very promising candidate from among these is cobalt
phosphide (CoP).4–7 In acidic conditions, CoP has been shown to operate at

89

overpotentials only slightly larger than Pt.5 Additionally, it can be easily nanostructured
to maximize the active surface area and thus boost performance without modifying the
electrode geometry.6,7
Palaniappan and coworkers have studied an array of cathode materials for
NH3(aq) electrolysis, including both noble and transition metals, nanostructured on
carbon supports.8,9 However, in aqueous conditions, water is the species reduced at the
cathode, rather than NH4+, and so it is unclear whether these materials will be suitable
cathodes in NH3(l) as well.10 In NH3(l) conditions, it would be ideal for the cathode
surface to have a slightly larger proton affinity than NH3 so as to extract and reduce H+
to H2 directly without forming NH40•. Free phosphine (PH3) has a similar proton affinity to
NH3,11 so it is reasonable to hypothesize that exposed phosphide sites in a material like
amorphous CoP could accept H+ from NH4+ and serve as an efficient reduction platform.
Saadi et al. were able to electrodeposit amorphous CoP–with a significant
fraction of amorphous cobalt oxide contaminant–onto copper (Cu) disk substrates as
catalysts for H+ reduction in acidic conditions.5 The cobalt oxide was removed in
operando, leaving approximately 20 µm diameter islands of CoP exposed on the Cu
surface as confirmed by scanning electron microscopy (SEM), energy dispersive x-ray
spectroscopy (EDS), and x-ray photoelectron spectroscopy (XPS). The film performed
at the same current density for H+ reduction as a Pt cathode for over 24 hours with a
mere 50 mV additional overpotential.
In this chapter, we apply a similar electrodeposited amorphous cobalt phosphide
electrode to the reduction of NH4+ in NH3(l), and compare its performance to our Pt disk

90

results.1 The reductions of NH4+ and H+ dissolved in DMF are also compared to NH4+
reduction in NH3(l) for both electrodes. Cobalt phosphide is shown to be an inexpensive,
simple

to

fabricate,

Earth-abundant

alternative

cathode

material

featuring

electrochemical performance similar to Pt and with the potential for morphological
engineering for commercial devices.

Experimental
Copper (Cu) disk electrodes were prepared using Cu wire (Arcor, 24 gauge).
After polishing, the electrodes were sonicated in a detergent solution (Sparkleen,
Fisherbrand), followed by deionized water, followed by isopropanol for 15 minutes each.
The electrodes were then kept in the isopropanol until they were rinsed with deionized
water immediately before immersion into the electrodeposition solution.
The CoPx film deposition was based on the method from Saadi et al.5 The
deposition bath consisted of a 10 mL solution with 0.15 M boric acid (Strem, 99.99%),
0.1 M sodium chloride (CCI, 99%), 1.2 M hypophosphorous acid (Spectrum, reagent),
and 0.2 M cobalt chloride hexahydrate (Alfa Aesar, 99.9%). A concentrated solution of
sodium hydroxide (Macron, reagent ACS) was then added until the pH of the solution
was 5.0. Nitrogen gas was bubbled through the solution for at least 30 minutes prior to
deposition to remove any dissolved oxygen. Immediately before electrodeposition, the
potential at the prepared Cu disk was cycled at 100 mV s-1 between -0.36 and -1.16 V
versus NHE, (-0.5 and -1.3 V versus a homemade Ag/AgCl reference electrode),12 in
the deposition bath to reduce the copper oxide layer in preparation for electrodeposition.

91

A pulsed deposition method was used to electrodeposit the CoPx film, involving a 1
second -0.96 V versus NHE pulse (-1.1 V versus the Ag/AgCl reference), followed by 3
seconds at open circuit. This was repeated 20 times for a total of 20 seconds deposition
time. After all 20 cycles were complete, the electrode was gently rinsed with deionized
water, rinsed again with acetone, and then dried under a gentle flow of nitrogen gas.
The electrochemically active surface areas of electrodeposited CoPx disks were
determined by fitting the charging current from cyclic voltammograms at scan rates
between 10 and 1000 mV s-1 in a solution of 0.1 M NH4PF6 (Fluka, 98%, recrystallized)
in N,N-dimethylformamide (Macron, reagent ACS) as described in Chapter 2. All
electrochemical experiments were conducted using a Metrohm µAutolabIII potentiostat
and employed the custom Ag/AgCl reference electrode.

Results and Discussion
A total of fifteen CoPx electrodes were fabricated by the same pulsed deposition
method since inconsistency in the film electrodeposition can result from small
differences in the Cu disk substrates. By comparison of the SEM images of the bare Cu
disk substrate in Figures 5.1(a) and (b) to those of the CoPx film in Figures 5.1(c) and
(d), it can be seen that the pulsed deposition method results in CoPx films conformal
with the Cu substrate morphology. The Co:P ratio of the CoPx film was measured by
EDS to be approximately 5 ± 1. Due to the amorphous nature of the film and variation in
stoichiometry, we will refer to the material as CoPx. The excess Co in the

92

Figure 5.1: SEM images of the Cu disk substrate, (a) and (b), and CoPx disk electrode
corresponding to the curves shown in Figures 5.3, 5.8, and 5.9, (c) and (d).

electrodeposited CoPx film parallels a report by Saadi et al. where there was an excess
of amorphous cobalt oxide deposited in addition to the desired CoP electrocatalyst.5
While the cobalt oxide was removed in operando by sulfuric acid (H2SO4) in the
literature report, our CoPx was unchanged in the cathodic operating conditions in NH3(l).
The films generated by pulsed deposition are different from those produced on similar
Cu disk substrates by simply holding the potential at -1.44 V versus NHE for 20
seconds. As depicted in Figure 5.2, “held depositions” produced CoPx microspheres,
about 1 µm in diameter, with a similar 5:1 Co:P stoichiometry. These microspheres
93

increased the electrode roughness factor by nearly ten-fold, and the produced a high
level of variation in electrode composition. The best of these performed similar to an
average electrode formed by pulsed deposition, after normalizing the electrochemically
active surface area. Thus, in the interest of experimental consistency, the pulsed
deposition

method

was

selected

for

all

further

electrochemical

studies.

Figure 5.2: SEM images of a CoPx film deposited on a Cu disk substrate by holding 1.1 V versus Ag/AgCl rather than application of the pulsed deposition method.

94

The cathodic performance of CoPx films deposited on Cu disks was measured
using linear sweep voltammetry in 0.1 M NH4NO3 in NH3(l). These were compared to
similar sweeps with a Pt disk cathode, and representative J-V curves are depicted in
Figure 5.3. Figure 5.3 shows a CoPx cathode with performance close to the average,
but Figure 5.4 shows a composite of the J-V curves of all fifteen CoPx. The onsets of
these waves are similar to those previously reported for the reduction of NH4+ in NH3(l),
where the large overpotential is likely due to the reduction pathway involving NH40• as
an intermediate.1

Figure 5.3: Cathodic J-V curves in 0.1 M NH4NO3 in NH3(l) with a Pt disk (black) or
CoPx disk (blue). Linear sweeps were measured at 100 mV s-1, and current densities
were calculated using the electrochemically active surface area.

95

Figure 5.4: Composite of J-V curves in 0.1 M NH4NO3 in NH3(l) from fifteen CoPx
electrodes fabricated by the pulsed deposition method.
To verify the cathodic process occurring, -1.85 V versus NHE was applied at a
CoPx electrode surface in 0.1 M NH4NO3 in NH3(l) for 150 minutes and the gasses in
the headspace were measured using gas chromatography. Figure 5.5 shows the
production of H2 gas from a CoPx electrode with approximately 100 % Faradaic
efficiency. This shows that H2 evolution is the only reaction of consequence at a CoPx
cathode in NH3(l).
As can be seen in Figure 5.6, the first cathodic scan with a CoPx cathode in
NH3(l) features a large cathodic pre-wave which is likely due to the reduction of surface
oxides formed from the storage of the electrode in air. This wave is not present in all
subsequent scans confirming the irreversibility of the process. To clearly differentiate

96

Figure 5.5: Comparison of the µmoles of H2 measured in the headspace of the
electrochemical cell (red diamonds) with the theoretical maximum amount of H2 if the
Faradaic efficiency for H2 evolution were 100 % (blue circles).

between this irreversible reduction behavior and the NH4+ reduction process, the third
successive linear sweep is depicted in Figures 5.3, 5.4, and 5.9.
Linear potential sweeps are shown in place of full cyclic voltammograms (CVs)
due to lack of clarity from the large hysteretic behavior with CoPx electrodes, which can
be seen in Figure 5.7. The cathodic current will steadily decrease, approaching a
constant plateau with time at a constant potential, as shown in Figure 5.8, periodically
interrupted by sudden jumps in the current density. The same behavior can be
produced by externally manually agitating the electrode during or between cathodic

97

Figure 5.6: First and second cathodic cyclic voltammograms of a CoPx electrode in 0.1
M NH4NO3 in NH3(l) showing the large cathodic wave only evident on the first scan of
each CoPx electrode, indicated by the black arrow.

scans. This phenomenon is attributed to the formation and later displacement of H2 gas
bubbles at the surface of the cathode, similar to the anodic behavior reported for Pt
anodes in NH3(l) at sufficiently positive overpotentials.1 Thus, a current density of about
-70 mA cm-2 is representative of the steady-state current density in Figure 5.8, rather
than -65 mA cm-2, as this is approximately what the current density jumps to as bubbles
are displaced.
While the overpotential required for 10 mA cm-2 is approximately the same for the
Pt and CoPx electrodes, their Tafel behaviors are very different, as shown in Figure 5.9
and Table 5.1. While the onset for NH4+ reduction on Pt is 200-300 mV lower than that

98

Figure 5.7: Cyclic voltammogram of a CoPx electrode in 0.1 M NH4NO3 in NH3(l)
showing hysteresis due to H2 bubble formation. The scan direction is indicated by the
black arrows.

of CoPx, the CoPx Tafel slope is more than two-fold steeper. This leads to intersection of
J-V and Tafel plots at about -1.3 V versus NHE. Consequently, the current density for
NH4+ reduction on CoPx exceeds that of Pt at potentials lower than 1.3 V versus NHE. It
is difficult, however, to extract any quantitative kinetic information from the Tafel plots in
Figure 5.9 as there are several different slopes in different potential regions, and it is
unclear which, if any, are suitable to be fit to a Butler-Volmer-type electron transfer.
The Pt Tafel behavior is unusually poor with respect to typical Pt aqueous H+
reduction characteristics, where slopes of 50 to 120 mV dec-1 are common.6 Again, this
is likely due to the alternative NH4+ reduction pathway involving the radical

99

Figure 5.8: Chronoamperogram of the same CoPx electrode shown in Figure 5.3
measured at -1.67 V versus NHE in 0.1 M NH4NO3 in NH3(l).

Table 5.1: Average electrochemical performance of fifteen similar CoPx electrodes
compared to a Pt disk in 0.1 M NH4NO3 in NH3(l).
Electrode

V for
10 mA cm-2

mV dec-1

Pt

-1.46

298

CoPx

-1.48 ± 0.04

130 ± 18

intermediate.1,2 In our previous work we probed the NH4+ reduction more closely using
NH4PF6 dissolved in DMF. A comparison of NH4+ reduction with the same Pt and CoPx
electrodes as those depicted in Figures 5.3, 5.8, and 5.9 is shown in Figure 5.10. The Pt
cathodic onset is approximately the same as previously reported,1 while the CoPx
reduction onset is about 300 mV more negative, as observed in NH3(l).

100

Figure 5.9: Cathodic Tafel plots corresponding to the J-V curves depicted in Figure 5.3
in 0.1 M NH4NO3 in NH3(l) with a Pt disk (black) or CoPx disk (blue). Tafel fits are shown
in dashed red lines.

In DMF, however, the Tafel slopes for Pt and CoPx are similar to each other: about 160
and 225 ± 10 mV dec-1 respectively (Figure 5.11). The Tafel behavior of CoPx indicates
slightly poorer relative NH4+ reduction performance in DMF, which could easily be
attributed to a solvent effect. For instance, if NH4+ reduction intermediates are better
stabilized by the NH3(l) environment than the DMF environment, the kinetic barrier for
NH4+ reduction would be lower in NH3(l) resulting in more favorable Tafel behavior. The
Pt Tafel behavior in DMF, unlike in NH3(l), is approximately what is expected for the
reduction of NH4+, and is similar to what we previously reported.1 The onset of -0.6 V

101

Figure 5.10: Cathodic J-V curves in 0.1 M NH4PF6 in DMF with a Pt disk (black) or
CoPx disk (blue). Linear sweeps were measured at 100 mV s-1, and current densities
were calculated using the electrochemically active surface area.

versus NHE corresponds with the potential of the direct reduction of NH4+. This further
emphasizes how unexpectedly large the Tafel slope for Pt in Table 5.1 is. It is unclear
what the cause of the relatively poor performance of a Pt cathode in NH3(l) is with
respect to DMF when the species reduced is presumably the same: NH4+. It may be the
case that a more “innocent” organic solvent would provide a model system that more
accurately reflects the cathodic behavior of both Pt and CoPx in NH3(l) and would allow
for in situ spectroscopy experiments to probe key intermediates in the NH4+ reduction
reaction. This problem, in addition to further investigation of Earth-abundant cathode
materials is the subject of continued research in our laboratory.

102

Figure 5.11: Cathodic Tafel plots corresponding to the J-V curves depicted in Figure
5.10 in 0.1 M NH4PF6 in DMF with a Pt disk (black) or CoPx disk (blue). Tafel fits are
shown in dashed red lines.

Conclusions
By a slight modification to a previously reported electrodeposition solution and
application of a pulsed deposition method, amorphous CoPx films were deposited
conformally onto Cu disk substrates. These films demonstrated electrochemical
performances similar to a Pt disk for the reduction of NH4+ in NH3(l), even after
accounting for the electrochemically active surface area of the electrodes. Due to the
simplicity of the deposition procedure, similar CoPx should be simple to deposit onto
micro- or nanostructured Cu substrates, such as Cu foam, in order to increase cathode
performance without increasing geometric area. While it appears that the direct NH4+
reduction pathway involving the unstable Rydberg radical intermediate is not bypassed

103

through use of CoPx, these cathodes are comparable to expensive, noble metal
cathodes previously reported.

104

REFERENCES

105

REFERENCES

(1)

Little, D. J.; Smith, III, M. R.; Hamann, T. W. Energy Environ. Sci. 2015, 8, 2775–
2781.

(2)

Berkh, O.; Shacham-Diamand, Y.; Gileadi, E. J. Electrochem. Soc. 2008, 155,
F223–F229.

(3)

McCrory, C. C. L.; Jung, S.; Ferrer, I. M.; Chatman, S. M.; Peters, J. C.; Jaramillo,
T. F. J. Am. Chem. Soc. 2015, 137, 4347–4357.

(4)

Shi, Y.; Zhang, B. Chem. Soc. Rev. 2016, 45, 1529–1541.

(5)

Saadi, F. H.; Carim, A. I.; Verlage, E.; Hemminger, J. C.; Lewis, N. S.; Soriaga, M.
P. J. Phys. Chem. C 2014, 118, 29294–29300.

(6)

Popczun, E. J.; Read, C. G.; Roske, C. W.; Lewis, N. S.; Schaak, R. E. Angew.
Chemie Int. Ed. 2014, 53, 5427–5430.

(7)

Callejas, J. F.; Read, C. G.; Popczun, E. J.; McEnaney, J. M.; Schaak, R. E.
Chem. Mater. 2015, 27, 3769–3774.

(8)

Palaniappan, R.; Botte, G. G. J. Phys. Chem. C 2013, 117, 17429–17441.

(9)

Palaniappan, R.; Ingram, D. C.; Botte, G. G. J. Electrochem. Soc. 2014, 161,
E12–E22.

(10) Gerischer, H.; Mauerer, A. Electroanal. Chem. Interfacial Electrochem. 1970, 25,
421–433.
(11) Hunter, E. P. L.; Lias, S. G. J. Phys. Chem. Ref. Data 1998, 27, 413–656.
(12) Sawyer, D. T.; Sobkowiak, A.; Roberts, J. L. Electrochemistry for Chemists;
Second.; John Wiley & Sons, Inc.: New York, 1995.

106

CHAPTER 6:
CONCLUSIONS AND FUTURE WORK

107

Introduction
As demonstrated in previous chapters, by administration of proper electrolyte
conditions, noble metals can be completely replaced by Earth-abundant, first row
transition metal-based materials like iron nitride (FeNx) and cobalt phosphide (CoPx) for
the electrolysis of liquid ammonia (NH3(l)). Figure 6.1 shows that the 2-electrode
electrolysis of NH3(l) can occur with nearly the same efficiency using FeNx and CoPx as
the anode and cathode respectively compared to two platinum (Pt) disk electrodes. This
represents a significant shift toward sustainability, scalability, and economic feasibility of

Current Density / mA cm-2

electrochemical NH3 splitting in a NH3 fuel cycle like that described in Chapter 1.

60.0
50.0
FeN
CoPx
CoPx
x /FeNx
/ Pt
Pt Pt

40.0
30.0
20.0
10.0
0.0
1.0

1.2

1.4

1.6

1.8

2.0

2.2

2.4

2.6

Cell Potential / V
Figure 6.1: The electrolysis of NH3(l) with a 0.1 M NH4NO3 electrolyte using FeNx and
CoPx as the anode and cathode respectively (Spartan green), compared to two Pt
electrodes (red).

108

However, there is still a large need for catalysis of both half-reactions if there is to
be any energetic benefit to using NH3 as a hydrogen fuel carrier. Here, future research
directions are proposed that could reduce the electrolysis overpotential substantially
and allow for market implementation.

Ammonia Oxidation Catalysis
A clear next step in the investigation of NH3(l) oxidation catalysts is to synthesize
pure,

crystalline

iron

nitride

materials–presumably

in

a

variety

of

different

stoichiometries–and compare their anodic performance to the in situ generated FeNx
presented in Chapter 4. This will allow us to learn more about the NH3 oxidation
pathway on iron nitride and potentially provide insight into better anode design.
Additionally, it is worth revisiting the poison species on a Pt electrode and comparing it
to the nitride species of FeNx to investigate how the presence of a surface nitride
decreases the performance of a Pt electrode, but increases performance of an Fe
electrode.
However, unless a breakthrough occurs while investigating crystalline nitride
materials, it is likely that simply searching through pure, solid state materials may not be
the best solution for an efficient NH3(l) electrochemical oxidation reaction. The best
catalyst we have found in NH3(l)–FeNx as described in Chapter 4–still converges to
approximately the same onset as a Pt surface. Significant gains can be made by simply
increasing the active surface area of an anode by nanostructuring, as shown by Dong et
al.,1 however when normalizing by the resulting roughness factor, the current is still

109

approximately the same as on a Pt surface. These observations indicate that an outersphere electron transfer is likely the rate-limiting step of the NH3(l) oxidation process.
This is supported by the observation by Katsounaros et al. that in aqueous media NH3
oxidation proceeds by a rate-limiting electron transfer, instead of the previously
proposed proton-coupled electron transfer.2 This is in agreement with our observation
that the rate of the NH3 oxidation reaction is independent of NH4+ concentration.3 If it is
true that an outer-sphere process is rate-limiting under the current conditions, then it is
unlikely that a bulk material will be able to efficiently drive the NH3(l) oxidation reaction
unless it contains specific catalytically active sites.
In order to gather more specific knowledge about what type of site would be
ideal, there is a large effort from the Smith and Hamann groups at Michigan State
University to design homogeneous molecular NH3 oxidation catalysts. Inspired by
molecular catalysts for electrochemical water oxidation employing ligands that can
support high oxidation states, they are working toward designing new molecules that
can support oxidation of NH3 to N2 at relatively modest overpotentials. Other research
groups have already been able to demonstrate N-H bond cleavage using molecular
molybdenum complexes,4,5 and work toward a complete catalytic cycle is well
underway. Synthesizing variations of functional molecular catalysts may delineate
factors that are rate-limiting, which could inform design of next-generation catalysts.
Ideally, relevant structures can then be built into heterogeneous electrodes in a
commercially viable NH3(l) electrolyzer.

110

There are a variety of feasible ways to incorporate catalytic sites from
homogeneous molecular catalysts into a heterogeneous electrode platform. The
simplest of these involves simply immobilizing the molecular catalyst on an electrode
surface through one of its ligands. This has the advantage of eliminating current
limitations due to catalyst diffusion, thus minimizing the quantity of catalyst needed in a
device. This strategy also only minimally changes the electronics of the catalyst.
However, synthesis of molecular catalysts is often time-consuming and expensive which
can be a major inhibition to commercial implementation.
Ideally, a bulk material can be fabricated that closely mimics the active site of the
molecular catalyst, eliminating the need for molecular synthesis altogether. For
example, if the key moiety in a molecular catalyst is a pendent amine ligand coordinated
to an Fe metal center, this could be potentially mimicked simply by attaching pendent
amines to a bulk iron nitride material in the form of a self-assembled monolayer. The
monolayer can then serve a similar function to the critical ligand in the molecule and
facilitate the oxidation reaction.

Ammonium Reduction Catalysis
While it is useful that electrodeposited CoPx may replace Pt as a NH4+ reduction
catalyst in NH3(l) similar to its function in aqueous H+ reduction, the problem of lowering
the cathodic overpotential by avoiding the NH40• still remains to be solved. It is a useful
step to move away from expensive, noble metals like Pt, but work still needs to be done

111

to design NH4+ reduction catalysts that will extract H+ from NH4+, and reduce it directly
to H2, thus avoiding energetically unfavorable radical species.
Analogous to the next step in NH3 oxidation, electrodes of pure CoP
nanoparticles, or bulk crystalline CoP should be fabricated and compared to the
electrodeposited CoPx. Additionally, there are a multitude of reported transition metal
alloys and phosphide materials that perform similarly in aqueous conditions, but may
diverge in NH3(l).6–8 Since Pt and CoPx are essentially the only two cathode materials
investigated in NH3(l), it is impossible to make generalizations about the H2 evolution
reaction in NH3(l) at large.
Should all cathode materials investigated converge to similar required
overpotentials, it will be necessary to more closely probe the reduction intermediates
using in situ spectroscopy. Incorporating either Pt–or possibly gold–nanoparticles, it
should be possible to conduct Surface Enhanced Raman Spectroscopy (SERS)
measurements to identify vibrational modes of intermediate species.9 The difficult part
of this experiment is the design of an in situ Raman apparatus that can incorporate an
electrochemical cell containing NH3(l) that would need to be kept cold or pressurized
during the experiment without the possibility of leaking into the instrument.
As a parallel approach to studying different cathode materials, NH4+ reduction
should be studied in an array of alternative solvents in addition to N,Ndimethylformamide (DMF). While NH3 should be stored and transported in pure liquid
form to maximize its benefits over other hydrogen storage media, a commercial
electrolyzer does not necessarily need to rely on a NH3(l) electrolyte. We showed in

112

Chapters 3 and 5 that the H2 evolution reaction occurs at more favorable overpotentials
in DMF compared to NH3(l). (In water, H2 evolution occurs at the optimum 0 V versus
RHE, though this is due to the reduction of H2O.)10 A possible route to diminishing the
NH4+ reduction overpotential, therefore, is to conduct the reaction in a solvent that can
destabilize the NH4+ ion enough to promote H+ reduction without the production of
Rydberg radical intermediates, while being sufficiently anhydrous to eliminate the
anodic poisoning process described in Chapters 3 and 4.
A different pathway toward NH4+ reduction catalysis involves a similar approach
to the NH3 oxidation techniques previously discussed. A proton reduction catalyst–
inspired by existing proton reduction catalysts that are active in aqueous conditions–
could be discovered to efficiently reduce NH4+ in NH3(l). Such a catalyst, again, would
need to feature a moiety with a greater proton affinity than NH3 in order to extract H+
from NH4+ and avoid the unwanted radical pathway. This molecular catalyst could then
either be immobilized onto an electrode surface, or else an electrode surface could be
fabricated to mimic the critical moieties of the catalyst.

113

REFERENCES

114

REFERENCES

(1)

Dong, B.-X.; Tian, H.; Wu, Y.-C.; Bu, F.-Y.; Liu, W.-L.; Teng, Y.-L.; Diao, G.-W.
Int. J. Hydrogen Energy 2016, 41, 14507–14518.

(2)

Katsounaros, I.; Chen, T.; Gewirth, A. A.; Markovic, N. M.; Koper, M. T. M. J.
Phys. Chem. Lett. 2016, 7, 387–392.

(3)

Little, D. J.; Smith, III, M. R.; Hamann, T. W. Energy Environ. Sci. 2015, 8, 2775–
2781.

(4)

Bezdek, M. J.; Guo, S.; Chirik, P. J. Science (80-. ). 2016, 354, 730–733.

(5)

Bhattacharya, P.; Heiden, Z. M.; Wiedner, E. S.; Raugei, S.; Piro, N. A.; Kassel,
W. S.; Bullock, R. M.; Mock, M. T. J. Am. Chem. Soc. 2017, 139, 2916–2919.

(6)

Callejas, J. F.; Read, C. G.; Roske, C. W.; Lewis, N. S.; Schaak, R. E. Chem.
Mater. 2016, 28, 6017–6044.

(7)

Cox, C. R.; Lee, J. Z.; Nocera, D. G.; Buonassisi, T. Proc. Natl. Acad. Sci. 2014,
111, 14057–14061.

(8)

McCrory, C. C. L.; Jung, S.; Ferrer, I. M.; Chatman, S. M.; Peters, J. C.; Jaramillo,
T. F. J. Am. Chem. Soc. 2015, 137, 4347–4357.

(9)

Le Ru, E. C.; Etchegoin, P. G. Principles of Surface-Enhanced Raman
Spectroscopy; Elsevier, 2009.

(10) Palaniappan, R.; Ingram, D. C.; Botte, G. G. J. Electrochem. Soc. 2014, 161,
E12–E22.

115