Copyright is owned by the Author of the thesis.  Permission is given for 
a copy to be downloaded by an individual for the purpose of research and 
private study only.  The thesis may not be reproduced elsewhere without 
the permission of the Author. 
 



NEW ZEALAND IRONSANDS CATALYSIS

OF THERMO-CATALYTIC METHANE

DECOMPOSITION

A thesis presented for the degree of

Master of Science

in

Nanoscience

at Massey University, Palmerston North,

New Zealand

Samuel Thomas Powick

School of Fundamental Sciences

2020



Abstract

The use of hydrogen gas as an energy carrier is a proposed pathway to elim-

inating greenhouse gas emissions from fossil fuels. However, emissions-free

production of hydrogen is more costly than traditional high-emissions hy-

drogen production processes such as Steam Methane Reformation (SMR).

To address this, a process called Thermo-catalytic Methane Decomposition

(TCMD) is being commercially developed. This process uses methane (natu-

ral gas) to produce hydrogen gas and high quality solid carbon which can be

sold to offset the price of the hydrogen produced. TCMD has the potential

to be cost-competitive with SMR. A key feedstock in the TCMD process is

a low cost catalyst, because the carbon produced deposits on the catalyst

and deactivates it. The most commercially viable choice of catalyst has been

identified as iron ore, or hematite, due to its high activity and lifetime, and

its low cost [13, 23]. The TCMD process could have applications in New

Zealand to supply the heavy transport market, but for this to happen, a

domestic source of iron ore is required for use as a catalyst. New Zealand’s

primary source of iron ore is in the form of titanomagnetite found in iron-

sands, which has different properties to hematite. As a result, this research

was commissioned to evaluate the effectiveness of New Zealand ironsands as

a catalyst for TCMD. The effects of temperature, flow rate, catalyst com-

position, and aggregate particle size on ironsand catalytic activity, lifetime,

and carbon by-product quality are evaluated relative to a hematite control.

ii



Contents

Abstract ii

Acknowledgements vi

Nomenclature vii

List of Figures x

List of Tables xii

1 Introduction 1

1.1 Project Aims . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2 Background 7

2.1 Hydrogen Production from Methane . . . . . . . . . . . . . . 7

2.1.1 Steam Reformation . . . . . . . . . . . . . . . . . . . . 7

2.1.2 Partial Oxidation and Autothermal Reformation . . . . 9

2.2 Thermo-Catalytic Decomposition of Methane . . . . . . . . . 10

2.2.1 The Catalyst . . . . . . . . . . . . . . . . . . . . . . . 13

2.2.2 Metal Catalyst Reaction Mechanism . . . . . . . . . . 16

2.2.3 Reactor Design . . . . . . . . . . . . . . . . . . . . . . 21

2.3 Iron Ore Catalysis of Thermal Methane Decomposition . . . . 22

2.4 Ironsands Catalysis of Thermal Methane Decomposition . . . 27

3 Materials And Methods 30

3.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . 30

3.2 Analytical Model . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.2.1 Methane Conversion . . . . . . . . . . . . . . . . . . . 33

3.2.2 Thermodynamic Equilibrium Limit . . . . . . . . . . . 34

iii



3.2.3 Characterisation of Carbon/Catalyst Particles . . . . . 35

3.3 Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . 38

3.4 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

4 Results and Discussion 42

4.1 Catalyst Activity and Decay . . . . . . . . . . . . . . . . . . . 50

4.1.1 Temperature Effects . . . . . . . . . . . . . . . . . . . 56

4.1.1.1 Results . . . . . . . . . . . . . . . . . . . . . 56

4.1.1.2 Discussion . . . . . . . . . . . . . . . . . . . . 66

4.1.2 Flow Rate Effects . . . . . . . . . . . . . . . . . . . . . 68

4.1.2.1 Results . . . . . . . . . . . . . . . . . . . . . 68

4.1.2.2 Discussion . . . . . . . . . . . . . . . . . . . . 73

4.1.3 Catalyst Composition and Particle Size Effects . . . . . 73

4.2 Carbon Quality . . . . . . . . . . . . . . . . . . . . . . . . . . 76

4.2.1 Temperature Effects . . . . . . . . . . . . . . . . . . . 77

4.2.1.1 Results . . . . . . . . . . . . . . . . . . . . . 77

4.2.1.2 Discussion . . . . . . . . . . . . . . . . . . . . 83

4.2.2 Flow Rate Effects . . . . . . . . . . . . . . . . . . . . . 85

4.2.2.1 Results . . . . . . . . . . . . . . . . . . . . . 85

4.2.2.2 Discussion . . . . . . . . . . . . . . . . . . . . 87

4.2.3 Catalyst Composition and Particle Size Effects . . . . . 89

4.3 Combined Effects . . . . . . . . . . . . . . . . . . . . . . . . . 91

4.3.1 Absolute Conditions . . . . . . . . . . . . . . . . . . . 91

4.3.2 Control-Relative Conditions . . . . . . . . . . . . . . . 93

4.3.3 Discussion of Optimal Conditions . . . . . . . . . . . . 94

4.4 Investigation Into Errors . . . . . . . . . . . . . . . . . . . . . 97

4.4.1 Gas Chromatography Error . . . . . . . . . . . . . . . 97

4.4.2 Replication Error . . . . . . . . . . . . . . . . . . . . . 100

5 Conclusion 105

iv



References 108

A Standards and Baselines 117

A.1 Standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

A.2 Baselines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

B Raw Data 121

C Methane Conversion Equation Validity 127

v



Acknowledgements

Thank you to my supervisors: Prof. Richard Haverkamp, and Assoc. Prof.

Mark Waterland for reviewing this late at night while Mrs Waterland watches

Grey’s Anatomy without him in the next room. Also, thanks to the funding

partners that made this research possible: The Energy Education Trust of

NZ, and Hiringa Energy. Andrew and Cathy, thanks for continually seeing

potential in me and giving me opportunities to work on things that I’m

excited about.

Steve, Olaf, and Stan, thanks for a five star workshop experience and for

the complimentary life advice. John Sykes and John Edwards, thanks for

the consistent and patient help with instruments and fittings. Thank you

to all the staff and other post-grads in the School of Fundamental Sciences,

particularly Prof. Shane Telfer and his group, who trained me and let me

loose on their instruments. Sam Brooke and Robert McEwen, cheers for

being the motivation I needed to finish this, despite your best efforts.

Thanks to all my friends and family, for your unending encouragement

and confidence in me. And lastly, thank you Hannah, for being incredibly

supportive while I wrote this thing instead of helping with the wedding. I

couldn’t have made it this far without you. I’m done now.

vi



Nomenclature

β 1
2

Width of the PXRD diffraction peak at FWHM

λ Wavelength of X-ray radiation used in PXRD

θ Bragg angle

gC/gcatalyst Weight of carbon in grams per weight of catalyst used in

the experiment

gC/gFe Weight of carbon in grams per weight of iron

ATR Auto-Thermal Reforming

Avg PD Average particle diameter, also dc

BEV Battery Electric Vehicle

CCM Crackling Core Model

Cementite Fe3C, also iron carbide

CSM Cracking Shrinking Model

d Interplane distance of graphite (002) plane

dc Average particle diameter, also Avg PD

FCEV Fuel Cell Electric Vehicle

Ferrite α-Fe, the form of iron that is catalytically active for TCMD

FWHM Full Width Half Maximum

G Glenbrook ironsand

vii



g Fractional degree of graphitisation, also GD

GC Gas Chromatography

GD Fractional degree of graphitisation, also g

H Hematite control

HFCEV Hydrogen Fuel Cell Electric Vehicle

I Patea ironsand, unpurified

ID/IG Ratio of the intensity of the D peak and the G peak in the

graphite Raman spectrum. Used as a measure of carbon

disorder.

IP Magnetically purified Patea Ironsand

Iron Carbide Fe3C, also cementite

k Scherrer coefficient

KT Equilibrium constant of the reaction at temperature T

LOI Loss On Ignition

N Sample size

POX Partial Oxidation

PSA Pressure Swing Adsorption

PSD Particle Size Distribution

PXRD Powder X-Ray Diffraction

Q Quartz Tube

viii



Rt Retention Time

S Stainless steel tube

SCAt Sustained Conversion Average over t hour(s) from the point

of maximum conversion

SCD Surface Carbide Deposition

SCM Shrinking Core Model

SG Specific Gravity

SMR Steam Methane Reformation

t 1
2
max−tmax Time taken for the conversion ratio to decay from maxi-

mum to half maximum

tmax Time at which the maximum conversion ratio occurs during

an experiment

TCMD Thermo-catalytic Methane Decomposition

TEL Thermodynamic Equilibrium Limit

TGA Thermo-Gravimetric Analysis

TGA Half-temp The temperature at which a TGA sample loses half its mass

vol% Volume percentage

XCH4 Conversion of methane (%)

XRF X-Ray Fluorescence

ix



List of Figures

1 A simplified schematic of the steam reformation process . . . . 8

2 Effect of oxygen stoichiometry on enthalpy of methane decom-

position . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3 CO2 emissions of hydrogen production processes . . . . . . . . 12

4 The four stage reaction mechanism of TCMD on a metal catalyst 17

5 Bulk metal dusting process schematic . . . . . . . . . . . . . . 19

6 Carbon morphologies produced by TCMD . . . . . . . . . . . 20

7 Summary of carbon morphologies produced, by catalyst and

temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

8 Three stage process of the first cycle of iron particle metal

dusting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

9 The second cycle of iron particle metal dusting . . . . . . . . . 26

10 Schematic of the spinel cubic structure of titanomagnetite . . 28

11 Experimental setup schematic and photo . . . . . . . . . . . . 31

12 Methane conversion ratios over time at 850 ◦C, by flow rate . 51

13 Maximum conversion ratio and tmax trends with temperature . 55

14 Methane conversion ratios over time of all catalyst types at

850 ◦C and 0.67 L/min . . . . . . . . . . . . . . . . . . . . . . 56

15 Methane conversion ratios of hematite at 0.67 L/min, by tem-

perature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

16 Maximum conversion ratio and tmax trends with temperature . 58

17 Sustained conversion average and t 1
2
max−tmax trends with tem-

perature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

18 Carbon yield trends with temperature . . . . . . . . . . . . . 61

19 Methane conversion ratios of H and IP catalysts at 850 ◦C and

0.067 L/min . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

x



20 Methane conversion ratios of H and IP catalysts at 900 ◦C and

0.067 L/min . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

21 Methane conversion ratios of H and IP catalysts at 850 ◦C and

0.067 L/min . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

22 Maximum H2 production rate of IP and H catalysts at 850 ◦C,

by flow rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

23 Total H2 yields of IP and H catalysts at 850 ◦C, by flow rate . 69

24 Methane conversion ratios over time of all catalyst types at

850 ◦C and 0.013 L/min . . . . . . . . . . . . . . . . . . . . . 74

25 TGA half-temperature trends with temperature . . . . . . . . 78

26 PXRD graphitic degree trends with temperature . . . . . . . . 78

27 PXRD average particle diameter trends with temperature . . . 79

28 Raman ID/IG ratio trends with temperature . . . . . . . . . . 79

29 Trends in 0.067 L/min H and IP carbon characterisation values

by temperature . . . . . . . . . . . . . . . . . . . . . . . . . . 80

30 Carbon characterisation data by temperature, relative to the

control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

31 Mean discrepancies in GC volume percentage totals, by flow

rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

32 H2 GC standard curve . . . . . . . . . . . . . . . . . . . . . . 119

33 CH4 GC standard curve . . . . . . . . . . . . . . . . . . . . . 120

34 Gas chromatogram example . . . . . . . . . . . . . . . . . . . 123

35 Raman spectrum example . . . . . . . . . . . . . . . . . . . . 124

36 PXRD spectrum example . . . . . . . . . . . . . . . . . . . . 125

37 TGA spectrum example . . . . . . . . . . . . . . . . . . . . . 126

xi



List of Tables

1 Equilibrium constants and thermodynamic equilibrium limits . 35

2 XRF major oxides analysis of ironsands . . . . . . . . . . . . . 41

3 Iron content of catalysts used in this research. . . . . . . . . . 41

4 Particle size distribution data for catalysts used in this research 41

5 Complete summary of results . . . . . . . . . . . . . . . . . . 46

6 Summary of results, relative to the control . . . . . . . . . . . 49

7 Conversion ratios after 18.5 hrs of deactivated 0.067 L/min

runs, relative to the control . . . . . . . . . . . . . . . . . . . 62

8 Differences between replicate runs . . . . . . . . . . . . . . . . 101

9 Blank tube methane conversion baselines . . . . . . . . . . . . 122

xii



1 Introduction

In 2017, as a consequence of greenhouse gas emissions from human activities,

the average global temperature was between 0.8 and 1.2◦C higher than pre-

industrial levels. Between 2030 and 2052, this is likely to increase to 1.5◦C,

and then, if no action is taken, to 2◦C. The risks to humanity and the envi-

ronment are thought to be significantly higher at 2◦C of global warming than

at 1.5◦C. Because of this, an urgent and concerted international response is

required to reduce our emissions [26].

The 2016 Paris agreement aims to motivate this global response through

national policy commitments. To date, 184 countries have ratified the agree-

ment, including New Zealand, however the national commitments of each

country vary widely and are currently overwhelmingly insufficient to restrain

projected temperature increases. New Zealand’s current policies are rated as

’insufficient’ to meet its fair share of emissions reductions [11]. As part of

the Paris agreement, New Zealand has committed to reducing net greenhouse

gas emissions by 2030 to 11% below what they were in 1990, and in 2019,

the government introduced a target of zero net emissions by 2050 (with a

caveat for agricultural methane). Meeting these targets will require particu-

lar action in the agriculture, energy, and transportation sectors. Other than

agriculture, energy and transportation are the largest sources of emissions in

New Zealand, together contributing 39.8% of our total emissions in 2016 [34].

Almost half of emissions within the energy & transportation combined sector

are from road transportation emissions, which have increased by 82.1% since

1990.

A significant reduction in the use of fossil fuels for transportation is re-

quired if New Zealand is to keep pace with the rest of the world. Eleven

countries have announced bans on new internal combustion vehicles to re-

duce transport emissions, including Norway by 2025, Denmark and Sweden

1



by 2030, and France and Britain by 2040 [4]. As a result, vehicle manu-

facturers are moving towards the production of electric drive vehicles, in the

form of either Battery Electric Vehicles (BEVs) or Fuel Cell Electric Vehicles

(FCEVs). BEVs are powered by batteries charged from the electricity grid,

and have the potential to be carbon-neutral if the electricity used to power

them is produced sustainably from hydro-, wind, or solar power. Hydro-

gen FCEVs (HFCEVs) are powered by fuel cells that convert hydrogen from

an on-board tank into electricity that is fed to the electric drive. HFCEVs

also have the potential to be carbon-neutral if the hydrogen fuel is produced

by renewable means. Manufacturers such as Nissan, Volvo, and Ford are

developing BEVs, while hydrogen FCEVs are being developed by Toyota,

Hyundai, and Honda.

Globally, there has been higher uptake of BEVs than HFCEVs, for two

reasons. Firstly, less capital expenditure is required to run small numbers

of BEVs, because they can be supported by existing electrical infrastructure

that does not rely on economies of scale. HFCEVs require hydrogen produc-

tion infrastructure, as well as a network of hydrogen fueling stations, which

carry with them a high capital cost that has forced a high price of hydrogen

at the pump in cities where this infrastructure has been built. Secondly, the

price of BEVs has been lower than HFCEVs, in part due to the fact that the

development of FCEVs is several years behind that of BEVs. However, par-

ticularly in a New Zealand context, HFCEVs have important advantages over

BEVs. For instance, in New Zealand, most freight is transported by high-

emissions diesel-fuelled trucking [34]. Heavy vehicles made up 3.7% of all

vehicles in New Zealand in 2016, but contributed 25.5% of vehicle emissions

that year [36], so this is an important area to target for emissions reductions.

However, for heavy transport and long range applications, BEVs are limited

by the cost and weight of additional battery capacity, while hydrogen tanks

can simply be made larger to increase the range of the vehicle, without un-

2



due weight penalty. This is because hydrogen has a specific energy of 33.3

kWh/kg, compared to around 0.55 kWh/kg for Li-ion batteries [17].

In addition to fuelling trucks, hydrogen also has possible applications in

marine transport, such as fuel-cell powered ferries, stationary power genera-

tion, such as the use of fuel cells as emergency power sources in remote areas,

and trains. In the long term, a hydrogen-based economy could facilitate inter-

national trade of renewable energy through the export of renewably-produced

hydrogen to countries such as Japan.

While in its infancy, the development of hydrogen production and fuelling

infrastructure in New Zealand is gaining traction. In 2019, the New Zealand

government released its vision statement for hydrogen development, stating

that developing hydrogen will be a part of its upcoming renewables strat-

egy [35]. In addition, projects by Ports of Auckland, Hiringa Energy, and

Tuaropaki Trust are currently in progress to build hydrogen production and

fuelling infrastructure, with varying levels of support from the government.

A significant challenge that these projects face is that emissions-free hy-

drogen is more expensive to produce than hydrogen produced with traditional

methods. The most cost effective method for producing hydrogen is Steam

Methane Reformation (SMR), which produces 95% of hydrogen in the United

States and around 50% of hydrogen globally [44]. SMR produces hydrogen

from methane in the presence of a nickel catalyst according to the following

process:

CH4 (g) + H2O (g) 3 H2 (g) + CO (g)

CO (g) + H2O (g) H2 (g) + CO2 (g)

Across the entire process, SMR produces 0.43 mol CO2 per mol H2, and

the overall Global Warming Potential of SMR has been estimated at 13.7

kg CO2e per kg H2 [39]. In comparison, a standard light vehicle emits 2.31

kg of CO2 per kg of hydrocarbon fuel [46], and although this does not take

3



into account the other greenhouse gases associated with the production and

use of hydrocarbon fuels, it is clear that using SMR to produce hydrogen for

use as an alternative fuel is not a viable option from an emissions reduction

point-of-view.

Instead, there exists an inherently cleaner process called Thermo-Catalytic

Methane Decomposition (TCMD). At high temperatures and in the presence

of a catalyst, methane gas decomposes into hydrogen gas and solid elemental

carbon via the following scheme:

CH4 (g) 2 H2 (g) + C (s)

This process still uses a fossil fuel as a feedstock, but it produces no green-

house gases. Hazer Group, an Australian company, are in the pilot stage

of commercialising a process based on TCMD, using iron ore as a catalyst.

Along with being almost emissions-free, they claim to have two competitve

advantages over SMR: Firstly, iron ore is significantly cheaper than the nickel

catalyst used in SMR, and secondly, the carbon produced is high quality

graphite that can be sold to offset the cost of hydrogen production [15, 57,

58]. Because of this, they project a 75% reduction in commodity costs over

SMR, and 6 times the production rate and half the running costs of electrol-

ysis [24].

In order to adopt TCMD in New Zealand, there are two things that must

be considered. Firstly, there is uncertainty about the future of natural gas

(methane) in Taranaki. In April 2018, the government banned new offshore

oil and gas exploration permits, and will restrict on-shore exploration by

2021 [55]. The current natural gas reserves of Taranaki have a limited life-

time, with only about 10 years of production left, and a projected shortfall

beginning this year without further development [12]. The lifetime of gas

reserves doubles to 20 years with the development of reserves that are cur-

4



rently considered non-commercial - that is, the current gas demand means

it would be un-economical to invest in expanding the gas supply. As the

supply of gas decreases and the shortfall becomes apparent, this investment

will happen organically. In the mid to long term, methanol production, a

key source of demand, will gradually exit the market, extending the lifetime

of existing reserves and leaving them available for higher value use cases.

Unfortunately, this means that a new source of demand for natural gas that

requires capital expenditure, such as a commercial TCMD operation, would

be unable to compete with the existing petrochemical gas users that have

sunk costs in existing infrastructure. Even with drastic policy changes and

a high carbon price, new demand development of TCMD is unlikely to be

economical [12]. The two exceptions to this are if there is a major new gas

discovery in an existing field in the near future, or if the TCMD process is

tied to a major biogas development. Ecogas in Rotorua has been awarded

funds from the government to build an anaerobic digestion plant that would

produce methane of a quantity useful for TCMD [54].

Secondly, for TCMD to be cost effective in New Zealand, it requires a

source of iron oxide to use as a catalyst. The country’s only domestic source

of iron oxide is from ironsands, found on the west coast of the North Island.

However, titanium is present in the sand in the form of titanomagnetite,

TiFe3O4. Early steel manufacturers attempted to smelt the ironsands to

produce iron, but the titanium in the sand prevented the use of a normal blast

furnace. Instead, a new electric smelting method was successfully developed

to reduce the titanomagnetite, and today over 2 million tonnes of ironsand

is mined each year [45]. While the presence of titanium caused difficulties

in the steel smelting process, the effect it may have on catalysis is unknown.

Ironsands may perform differently to typical iron oxides when used as a

catalyst for TCMD.

Hiringa Energy have commissioned this research to determine if New

5



Zealand ironsands could be commercially viable as a thermo-catalytic methane

decomposition catalyst. It aims to understand the conditions necessary to

promote the best catalytic performance of the ironsands.

1.1 Project Aims

The project aims were as follows:

1. Investigate the use of titanomagnetite (iron sands) as a catalyst for

methane cracking to hydrogen and graphite and its effect on hydrogen

and graphite production efficiency and quality.

2. Compare the results of the research with the control process utilising

iron ore hematite catalyst Fe2O3.

3. Gain process knowledge in the application of iron ore catalysis in the

production of hydrogen and graphite.

Its desired outcome was to successfully develop an environment in which

the methane cracking catalytic performance of titanomagnetite (iron sands)

is equal to that of iron ore. The environment encompassed reactor design,

reaction conditions, and catalyst pre-treatment. Catalytic performance en-

compassed hydrogen yield, hydrogen purity, carbon yield, carbon quality,

and catalyst lifetime.

These project aims were developed in conjunction with Hiringa Energy

and the New Zealand Energy Education Trust, and could not easily be al-

tered. In particular, the requirement to compare ironsand with the hematite

control seemed sensible at the start of the project, because raw hematite

was identified in the literature as the most commercially viable catalyst, and

was actively being commercialised in Australia. However, this comparison

was time consuming, and in hindsight ended up constraining other poten-

tially beneficial avenues of investigation, such as characterising the structural

changes in ironsands during TCMD.

6



2 Background

A review of the literature related to thermo-catalytic methane decomposition

using ironsands is presented here, beginning with an overview of traditional

methods for hydrogen production from methane, then covering specific detail

relevant to the project.

2.1 Hydrogen Production from Methane

Methane is the most common feedstock for production of molecular hydro-

gen. It is favoured over other hydrocarbons because it has the highest ratio

of hydrogen to carbon in its chemical composition, and because it is the

most abundant fossil fuel. There are three established methods for produc-

ing hydrogen from methane: Steam Reformation, Partial Oxidation, and

Autothermal Reformation.

2.1.1 Steam Reformation

Steam methane reformation (SMR) occurs via the following scheme:

CH4 (g) + H2O (g) 3 H2 (g) + CO (g) (1)

CO (g) + H2O (g) H2 (g) + CO2 (g) (2)

The cracking step is highly endothermic, with an enthalpy of ∆h0 = 206

kJ/mol, while the Water Gas Shift step is mildly exothermic with an enthalpy

of ∆h0 = −41.1 kJ/mol. The total reaction,

CH4 (g) + 2 H2O (g) 4 H2 (g) + CO2 (g) (3)

7



Figure 1: A simplified schematic of the steam methane reformation pro-
cess. Desulfurisation prevents sulfur compounds in the natural gas stream
from poisoning the nickel catalyst inside the reformer. The reformer reacts
the methane with steam to produce H2 and CO. Heat recovery refers to
the recirculation of unconverted methane back into the burner to produce
steam. Next, the water gas shift reaction converts the CO from the reformer
into additional H2 using more steam. The pressure swing adsorption (PSA)
system is used to separate the H2 product from CO2 and any other gases
present in the exit stream. [44]

has an enthalpy of ∆h0 = 165 kJ/mol [13]. 4 moles of hydrogen and 1 mole

of carbon dioxide are produced per mole of methane. The reaction occurs

over a metal catalyst, normally nickel based. Figure 1 shows the standard

SMR process. First, there is typically a desulfurisation step before the feed

gas reaches the reforming reactor, because nickel is easily poisoned by sul-

furic compounds. The methane then flows through the reforming chamber,

where it is mixed with steam, heated to 850 ◦C and undergoes cracking ac-

cording to Equation 1. The resulting methane/hydrogen/carbon monoxide

mixture is cooled before reacting with more steam according to Equation 2.

Finally, the exit gas stream, a mixture of hydrogen, methane, and carbon

dioxide, is separated using a pressure swing adsorption system (PSA). The

hydrogen produced is between 97% and 99.99% pure, and the overall conver-

sion efficiency is between 74% and 85%. The cost to produce hydrogen from

8



large-scale SMR was around $2.1 USD/kg in the USA in 2019. [44, 27].

Stoichiometrically, SMR produces 5.5kg of CO2 per kg of H2, but the

overall Global Warming Potential of the process is estimated at 13.7 kg CO2e

per kg H2 [39].

2.1.2 Partial Oxidation and Autothermal Reformation

Partial oxidation (POX) reacts methane with oxygen according to the fol-

lowing scheme:

CH4 (g) + 0.5 O2 (g) 2 H2 + CO (4)

This has an enthalpy of ∆h0 = −36 kJ/mol [9], although the enthalpy of

reaction changes depending on the ratio of oxygen to methane, as shown in

Figure 2. The reaction becomes more exothermic as higher ratios of oxygen

are added [43].

Autothermal reforming (ATR) adds a water gas shift reaction to the

partial oxidation process as for SMR, giving an overall reaction of:

CH4 (g) + H2O + 0.5 O2 (g) 3 H2 + CO2 (5)

This has a total enthalpy of ∆h0 = −77 kJ/mol, and is self-sustaining.

Because of this, no methane recirculation is required for heat recovery in

the process, and the cost of hydrogen produced by large-scale autothermal

reformation is cheaper than SMR at around $1.5 USD/kg [44]. However,

there are very large capital costs associated with oxygen injection, and so

this method is less favoured than SMR [13]. In addition, stoichiometrically,

ATR produces 7.7 kg CO2 per kg of H2, more than SMR.

9



Figure 2: Methane decomposition reactions have a near-linear decrease in
enthalpy with increasing ratio of oxygen to methane [43]. This means that
large-scale process that use oxygen, such as partial oxidation or autothermal
reformation, have lower energy requirements than processes without, such
as SMR.

2.2 Thermo-Catalytic Decomposition of Methane

The cracking of methane into hydrogen and solid carbon has been present

in the literature for a century. It was first patented in 1918, using a simple

airless chamber heated to 1100 ◦C, and included designs for recirculated

hydrogen to heat the furnace [8].

The thermal decomposition of methane into hydrogen and carbon pro-

ceeds according to the following scheme:

CH4 (g) 2 H2 (g) + C (s) (6)

This reaction is mildly endothermic, with an enthalpy of ∆h0 =74.85

kJ/mol [1]. This equates to 37.8 kJ/mol of H2, compared with 41.3 kJ/mol

10



H2 for Steam Methane Reformation. Equation 7 shows the Gibbs energy

relationship dependence on temperature for TCMD:

∆G0(J/mol) = 89658.88 − 102.27 · T − 0.00428 · T 2 − 2499358.99

T
(7)

The Gibbs energy is negative when T > 545 ◦C, meaning that theoretically

methane decomposition should occur above this temperature [1]. However,

to achieve an efficient hydrogen conversion rate, the temperature required is

over 1200 ◦C [2]. To increase the conversion efficiency at lower temperatures,

a metal or carbon-based catalyst is usually used.

There are no CO2 emissions inherent in the TCMD process if the heat-

ing method is decarbonised by using hydrogen as fuel for combustion. As

a result, TCMD offers complete emissions reductions over SMR and ATR

for hydrogen production, which will increase its commercial competitiveness

as those established processes are forced to reduce their emissions through

methods such as carbon capture and storage (CCS) - something that is al-

ready happening. For example, H21 in the UK are developing a plan to

decarbonise their reticulated natural gas network using an ATR process with

CO2 sequestration, beginning conversion in 2022 [20]. Figure 3 shows the H2

and CO2 production efficiencies per unit of CH4 feedstock, with and without

CO2 capture and storage. It shows that TCMD is the only process than can

produce hydrogen with zero emissions.

Additionally, the carbon product from TCMD has significant value and a

large worldwide market, depending on the quality produced. TCMD can pro-

duce carbon nanotubes, carbon nanofibres, and other forms of high-quality

graphite which are used in lithium-ion batteries and have a wide variety of

potential uses, including in fuel cells [29]. A recent sharemarket evaluation

of Hazer Group TCMD-produced high-purity graphite gave a projected price

of US$10,000/tonne in 2021 [15].

11



Figure 3: H2:CH4 and CO2:CH4 production volume ratios of various pro-
cesses, with and without CO2 emissions-mitigation measures [41]. While
CO2 sequestration has the potential to reduce emissions from SMR and POX
processes substantially, TCMD is the only process that can reach zero emis-
sions. Plasma Decomposition produces no emissions, but uses large amounts
of electricity which must be accounted for.

Overall, the economical viability of TCMD as a commercial process de-

pends on the following factors:

1. Energy efficiency of the process

2. Heating method used

3. Cost effectiveness of the catalyst

4. Value of carbon product

5. Cost of regulatory requirements placed on existing SMR processes, such

as CO2 sequestration costs.

Factor (1) is favourable for TCMD due to its lower enthalpy, (5) will

become increasingly favourable for TCMD as governments experience pres-

sure to reduce emissions, and the remaining factors are all currently being

researched. This research focuses on (3) and (4).

12



2.2.1 The Catalyst

Use of catalysts to lower the reaction temperature of TCMD and increase

the rate of decomposition has been the subject of much research, as they are

considered necessary for the process to be economically viable. By extension,

the cost-effectiveness of the catalyst itself has a large influence on this process

viability. The critical factors that determine whether a catalyst is cost-

effective are:

1. Activity

2. Lifetime

3. Quality of carbon produced

Of these, the lifetime of the catalyst is the most challenging to address,

because the carbon produced in the reaction deposits on the surface of the

catalyst, deactivating it.

Characteristics which influence the above critical factors include:

1. Chemical composition

2. Stability and mechanical properties

3. Surface area

4. Pre-treatment

5. Particle size

For a given catalyst, the last three characteristics are variables that may be

altered to influence its performance.

Catalysts for TCMD can be divided into two groups: metallic and car-

bonaceous. The relevance of carbon-based catalysts to this research is lim-

ited, and comes mostly from the possibility of TCMD-deposited carbon itself

contributing to the reaction by catalysing further methane decomposition.

As such, they will be discussed here only briefly.

Carbon-based catalysts have the advantages of being low-cost, stable,

and potentially self-sustaining within a TCMD environment. Many types of

13



carbon have been investigated, including amorphous carbon, carbon black,

graphite, and highly ordered morphologies such as carbon nanotubes and

fullerenes. It was found that nanostructured carbon had low catalytic ac-

tivity, but fullerene soot had high activity [3]. In general, more disordered

carbon catalysts had higher activities than ordered forms [40]. Smaller par-

ticle sizes correlated with higher methane conversion rates, but over longer

reaction times, partial catalyst deactivation negated this effect. Tempera-

tures over 850 ◦C caused the carbon particle size to decrease, resulting in

slower deactivation of the catalyst relative to lower temperatures [3]. Higher

temperatures also increased the rate of methane decomposition by increasing

diffusion into the smaller pores of the carbon bulk, allowing the methane to

access more catalyst surface area than at lower temperatures. Catalyst life-

time was limited by deposited carbon fibre growth blocking pores, and this

process occurred much more rapidly in the initial stages of a reaction. In

one case, carbon deposits weighing less than a quarter of the total catalyst

body caused a 97% reduction in catalyst surface area [37]. This occurs early

in the reaction period (within 1-4 hrs), with the surface area vs time curve

following the methane conversion curve in a 1
x
-like relationship [3].

Metal catalysts typically have higher activities than carbonaceous cata-

lysts, depending on the metal used. One proposed hierarchy of metal catalyst

activities for TCMD is Co, Ru, Ni, Rh > Pt, Re, Ir > Pd, Cu, W, Fe, Mo,

although other reports claim the highest activities for Ni and Fe/Al com-

binations, with activities as high as 491 gC/gCatalyst [3]. Other metals are

sometimes used as promoters to improve the properties of a catalyst and

balance the deposition and diffusion rates. A multi-metallic Ni+Pd catalyst

on an Al2O3 support produced a record yield of 1498 gC/gNi+Pd [47].

Metal catalyst deactivation occurs by carbon encapsulation and poison-

ing. Encapsulation with solid carbon deactivates the catalyst simply by pre-

venting additional methane from accessing the catalyst surface. This is also

14



affected indirectly by sintering - structural changes in the catalyst particles

themselves during the reaction, which produce more encapsulating types of

carbon and increase the rate of carbon encapsulation. For example, Ni has a

high activity and produces carbon fibres at temperatures below 700 ◦C, but

above this temperature the catalyst sinters, loses surface area, and rapidly

deactivates [3]. This can be prevented by supporting the catalyst particles

on an inert material, so they cannot coalesce.

Poisoning deactivates the catalyst via foreign compounds binding to ac-

tive sites, lowering its activity. Sulfur in the methane feedstock is the most

common cause of poisoning, and is the reason that the SMR process typically

includes an initial desulfurisation step.

To extend the catalyst lifetime, there are two approaches: remove the

carbon from the catalyst, and slow the rate of deactivation. Separating the

carbon from the catalyst is challenging because the method used must be

compatible with a continuous operation of the process, as would be used in

a large-scale commercial facility. Carbon can be separated using off-stream

methods such as acid washing or sonication, however this can damage the

catalyst and makes the continuous operation of the process difficult, espe-

cially if the catalyst has a high cost and must be recycled. The deactivation

rate can be slowed by adding small amounts of water vapour, which helps

by removing carbon nanoonions and other encapsulating carbon types, but

carries a risk of lower carbon yields. Injecting oxidising gases like O2, CO2,

or steam into the process can be used to remove the carbon from the cat-

alyst particles, however, doing so produces CO2 emissions similar to those

produced by SMR, thus rendering the excercise pointless from an environ-

mental perspective. It also degrades the carbon product, and to a varying

extent, the catalyst [13], while contaminating the hydrogen exit gas stream

with carbon oxides, requiring more purification [3].

15



2.2.2 Metal Catalyst Reaction Mechanism

TCMD follows a four step mechanism, shown in Figure 4:

1. Surface reactions

2. Dissolution

3. Diffusion

4. Precipitation

Firstly, methane adsorbs to the surface of the catalyst particle. An indicator

of the rate of adsorption is the sticking coefficient (the ratio of adsorbing

atoms to atoms that hit the surface in a given time) of methane for a given

catalyst type. The metal used affects the orbital interactions with the de-

positing graphite and thus which face of the catalyst the carbon structures

will deposit or grow on, and different faces of a catalyst may have different

sticking coefficients [29]. For example, the Ni(1 1 0) face is the most reactive

towards methane and forms the gas/metal interface, while the Ni(1 1 1) face

is the least reactive towards methane and forms the metal/graphite inter-

face [61]. Adsorption is also governed by the activation energy of methane

adsorption to the catalyst [29].

Secondly, the methane is dehydrogenated repeatedly until the carbon

dissolves into the catalyst crystal structure. This is thought to be the rate

limiting step at temperatures between 740 and 770 ◦C.

Thirdly, the concentration gradient of carbon inside the catalyst particle

drives the diffusion of the carbon to the other sides of the particle. This

continues until the catalyst particle has become supersaturated with carbon.

When Fe is the catalyst, an iron carbide (Fe3C) phase is present, but this is

not observed with Ni. However, the activities of metal catalysts, including

Ni, have been shown to correlate with its carbon solubility [29]. For Fe

nanoparticles formed from ferrocene, the diffusion step was the rate limiting

step at temperatures between 762 and 830 ◦C [21].

16



Lastly, the carbon precipitates from the surface in different morphologies

[13].

Figure 4: The four stage reaction mechanism of TCMD on a metal catalyst.
Reproduced with permission from A. Cornejo [13].

Parameters of the reaction that may affect the activity, lifetime, and

carbon production quality of a metal catalyst include:

1. Temperature

2. Pressure

3. Flow rate

4. Feed gas composition

These parameters must be balanced to provide conditions that allow the

sustained growth of carbon fibres. This is because the rate at which carbon

deposits on the surface of the catalyst and the rate at which carbon diffuses

through the catalyst particles to form fibres must be balanced to prevent

carbon encapsulation by structures such as carbon nanoonions. If the carbon

17



deposition rate is higher than the diffusion rate, the catalyst particle will be

coated in carbon before it can diffuse away, and deactivation will occur. This

will also cause termination of fibre growth, resulting in amorphous carbon

morphologies that further encapsulate the catalyst. If, however, there is not

enough carbon supplied to the catalyst, nucleation of carbon fibres will not

occur [13].

Parameters can be chosen to delay deactivation by promoting metal dust-

ing of the catalyst particles. Metal dusting is when carbon layers precipitat-

ing from the catalyst particles cause the metal to be separated from the bulk,

as shown in Figure 5. This process allows the feed gas to access new surfaces

of catalyst, extending the reaction lifetime. Increasing the pressure increases

metal dusting. If metal dusting is high, it reduces the effect of having large

catalyst particle aggregate sizes at the start of the reaction, as eventually the

aggregates disperse [13].

Another factor shown to be important in the growth of carbon fibres is

the curvature of the catalyst particle surface. This means that carbon fibre

nucleation and growth favours smaller particle sizes. In one example, carbon

fibres grew preferentially on Co catalyst particles of 10-30 nm diameter, while

ignoring any larger particles [56]. Having said that, it was also observed

that the metal catalyst particle morphology changed along with the type of

carbon produced, moving from dense, rigid polyhedrons at low temperatures

to an elongated, liquid-like state at higher temperatures [29]. This partial

liquefaction increased the surface curvature of the catalyst particles, allowing

carbon fibre growth on larger particles. This occurs below the normal melting

point of the metal because the carbon fibre formation is exothermic [13].

A wide range of carbon morphologies have been observed as TCMD prod-

ucts, including single- and multi-walled carbon nanotubes, carbon nanofibres,

and carbon nanoonions as shown in Figure 6. These morphologies are influ-

enced by the reaction parameters above, along with the changes in shape of

18



Figure 5: Schematic of metal dusting observed in TCMD. An unsupported
metal catalyst, as might be used in a fixed bed reactor, initially only con-
tacts the feed gas on one surface. As the carbon saturates the metal and
begins to precipitate, the metal particle is forced away from the bulk by the
graphite layer, opening up more catalytic surfaces for the methane to con-
tact, and prolonging the reaction. EmendationsChange1.5Reproduced with
permission from A. Cornejo [13].

the quasi-liquid catalyst particle. For example, the catalyst particle shape

changes from templating the nucleation cap to templating the wall growth of

a carbon nanotube [25]. This can occur differently depending on the metal

used, and so different catalysts produce different carbon morphologies, as

shown in Figure 7.

Carbon morphology is also known to be temperature dependent. With-

out a catalyst, the hardness of the carbon produced was found to increase

with temperature, and the particle size to decrease [2]. Using a Ni cata-

lyst, low temperatures around 500 ◦C caused the formation of full carbon

nanofibres, while temperatures around 750 ◦C produced hollow fibres. This

was attributed to a higher rate of carbon atom diffusion through the metal

particles [53] and a higher nucleation rate disfavouring sites with a long dif-

fusion path at higher temperatures [29]. Using a Co catalyst, dense carbon

19



Figure 6: Various carbon morphologies produced by the TCMD process [29].
Highly ordered forms of carbon are desirable products of this process and
thus the catalysts and reaction conditions that produce them are favoured.

Figure 7: Representation of literature data on catalysts, preferred tempera-
ture range and carbon products related to TCMD. Catalysts: 1—Ni-based,
2—Fe-based, 3—carbon-based, 4—summary of data related to Co, Ni, Fe,
Pd, Pt, Cr, Ru, Mo, W catalysts, 5—non-catalytic decomposition. Carbon
products: CF—carbon filaments, TC—turbostratic carbon, GC—graphitic
carbon, AmC—amorphous carbon [39].

20



nanofibres were again produced at around 500 ◦C, hollow nanofibres between

600 ◦C and 700 ◦C, and various forms of single-walled nanotubes at higher

temperatures [56].

The feed gas composition also affected the morphology of the depositing

carbon. Adding 0.6 vol% of hydrogen to the methane feed over a nickel-

alumina catalyst at 500 ◦C prevented carbon nucleation entirely [28], while at

lower levels, increasing hydrogen percentage shifted the carbon morphology

from filament-like to bamboo-like. Hydrogen is thought to reduce the carbon

deposition rate and increase the carbon removal rate, and as such is a useful

tool for finding the balance required for carbon fibre growth as discussed

earlier [29].

2.2.3 Reactor Design

The design of the reactor for TCMD is critical to minimising catalyst deac-

tivation. Reactors have been trialled with fixed and fluidised catalyst beds,

single chambers and multiple chambers in series at different pressures.

The simplest reactor is a stainless steel tube contained within a furnace,

with a fixed quantity of catalyst loaded before heating, and gas intake and

exit streams. Solid carbon must be periodically removed from this reactor

to prevent the gas flow from being blocked, and this makes fixed bed reac-

tors impractical for continuous operation [3]. It also means that accurately

determining the maximum carbon yield of a given mass of catalyst may not

be possible due to reactor space limitations.

Fluidised bed reactors use a continuous flow of fine catalyst particles

through the reaction zone transported by a carrier gas, allowing deactivated

particles to be removed and fresh catalyst to be added without halting the

process. Fluidised beds are more effective at transferring heat to catalyst

but require a carbon scrubber to remove particles from the exit gas feed [3].

Also, the use of a fluidised bed creates a high catalyst turnover, meaning

21



that the catalyst must be low cost for such a process to be economical.

Multi-chamber fluidised bed reactors have been developed that attempt

to extend the lifetime of the catalyst by altering the parameters of each

chamber. It has been shown that multiple reactors in series are superior

to having them in parallel, depending on the design [13]. In one case, a

3-5 chamber series counter-flow reactor was designed, whereby the methane

feed gas flow was opposite to the fludised catalyst bed flow. The reactor

chambers were at increasing pressures, and the methane flowed from high

to low pressure, while the catalyst flowed from low to high. This meant

that as the catalyst deactivated, the increasing pressure would allow the

methane to access deeper pores, kinetically favouring hydrogen conversion

and extending the lifetime of the reaction. In the low pressure chambers,

hydrogen production was more thermodynamically favoured [13].

Separation of the hydrogen from the exit gas stream can be done at lab

scale using a gas chromatograph, and on a commercial scale using a pressure

swing adsorption system as is used for SMR.

2.3 Iron Ore Catalysis of Thermal Methane Decompo-

sition

The catalysis of TCMD with plain powdered iron ores such as hematite

(Fe2O3) and magnetite (Fe3O4) has been studied specifically, and identified

as suitable catalyst choice for a commercial TCMD process. These pow-

dered ores are referred to as unsupported polycrystalline particles, because

they are macroscopic aggregates of clusters of smaller iron crystals. Iron has

a lower activity than Ni in this form, but is more stable at the higher tem-

peratures needed for high hydrogen conversion (700-1000 ◦C) [13]. It is also

low cost (around $150/t), meaning much higher catalyst turnover rates can

be sustained in fluidised bed reactor continuous operation than for a high

22



cost catalyst such as Ni (around $20,000/t). The low cost also means that

chemical methods of carbon purification can be used to remove the carbon,

even if the catalyst would be damaged as a result.

Cornejo [13] describes a conceptual model for TCMD using unsupported

polycrystalline iron ore as a catalyst. They suggest that the mechanism

of iron catalysis is similar to that displayed in Figure 4, but contains an

additional intermediate step called the carbide cycle. Rather than simply

diffusing through the catalyst crystal in step 3 of Figure 4 as is observed

for Ni catalysts, carbon atoms saturate the Fe crystals, forming a cementite

(Fe3C) crystal [19]. This cementite crystal is stable until decomposition is

triggered by a low-surface energy carbon particle depositing on the surface.

The lower surface energy allows the carbon to leave its cementite form and

aggregate into carbon structures, while the crystal returns to ferrite. This

means that multiple cycles of carbon dissolution, cementite formation, and

graphite precipitation can occur. The catalyst crystal eventually deactivates

when the carbon structures on the particle prevent methane diffusing into

contact with the Fe.

The method of catalyst reduction affected the performance of iron as a

catalyst [13]. Reduction with methane during the reaction caused higher

activity than pre-reduction with hydrogen. It is thought that there is a syn-

ergistic effect on catalyst performance when catalyst reduction and methane

decomposition take place simultaneously. Hydrogen from the decomposition

of methane causes reduction localised to the decomposition site. Reduction

also produces water vapour, which in small amounts has been observed to

reduce the production of deactivating carbon morphologies and increase the

catalyst lifetime [60]. As a result, letting the catalyst be reduced within

the reaction itself is beneficial. In addition, cementite formation occurs more

easily in iron oxides, and pre-reduction would eliminate this possibility. How-

ever, because of the removal of oxygen from the iron oxides during the re-

23



Figure 8: First cycle of polycrystalline iron particle metal dusting. A: SCM1
- Shrinking Core Model, first cycle. Methane dissociates and carbon diffuses
into the boundary region of the reduced iron particle. B: SCD1 - Surface
Carbide Decomposition, first cycle. Formation of graphite skin around the
particle allowing saturation of crystal clusters. C: CCM1 - Crackling Core
Model, first cycle. Precipitation of graphite along the boundary of saturated
crystal clusters pushes the particle apart. Reproduced with permission from
A. Cornejo [13].

24



action, small amounts of water vapour, CO2, and CO were produced during

reduction, which had to be purified out of the exit gas stream.

In experiments done comparing different types of iron ore, it was shown

that the oxidation state of iron did not affect its hydrogen conversion rate,

and any impurities in low-grade ores only served to increase its catalytic

activity [13]. However, oxide catalysts must be highly uniform to produce

uniform carbon morphologies [29].

The lifetime of the iron ore catalyst is determined by metal dusting [13, 19,

50]. The metal dusting process follows an iterative multi-stage process that

incorporates three established models of solid-fluid reactions: The shrink-

ing core model (SCM) [59], crackling core model (CCM) [48], and cracking

shrinking model (CSM) [33]. First, the iron ore must be completely reduced

to ferrite (α-Fe). Next, methane adsorbs to the surface of the particle, and

as the methane dissociates, carbon atoms dissolve into the surface region of

the polycrystalline ferrite, forming iron carbide (or cementite, Fe3C), and

approximately following the shrinking core model. This is shown in Figure

8A. At a certain point, the boundary region of the polycrystalline particle be-

comes saturated with carbon, which then precipitates as layers of graphite on

the surface [19]. This is called Surface Carbide Decomposition (SCD) and is

shown in Figure 8B. As this graphite layer builds up, it inhibits the diffusion

of methane into contact with the iron surface. This slows the reaction rate,

and thus the rate of precipitated carbon build-up on the surface. Because of

this, freshly dissociated carbon is able to diffuse deeper into the particle and

saturate entire crystal clusters, not just the sub-surface boundary regions.

The precipitation of graphite from these saturated crystal clusters occurs

along all boundaries of the crystal cluster, including those internal to the

polycrystalline particle. Such precipitation forces the crystal clusters within

the particle to separate, according to the Crackling Core Model, and creates

graphite flakes. This step is shown in Figure 8C, and is the final stage of the

25



Figure 9: The second cycle of polycrystalline iron particle metal dusting.
SCM2 - Shrinking Core Model, second cycle. SCD2 - Surface Carbide De-
composition, second cycle. CCM2 - Crackling Core Model, second cycle.
This cycle involves the metal dusting of crystal clusters separated from the
bulk in the first cycle. If dusting continues until the iron particle is split into
individual crystals, then carbon fibers or carbon nanoonions will form. If de-
activation occurs before dusting is complete, cluster carbon nanoonions will
form. Graphite flakes are formed from the carbon shell separating during
dusting. Reproduced with permission from A. Cornejo [13].

26



first cycle of metal dusting.

The separated crystal clusters have surfaces that are newly exposed to

the feed of methane, which are again saturated with carbon according to

the SCM, beginning the cycle again [13]. The second cycle of metal dusting

is shown in Figure 9. The metal dusting cycle repeats until either the iron

particles have dusted into single crystals, or premature deactivation occurs.

Single iron crystals follow the carbide cycle as described earlier. Premature

deactivation occurs when methane can no longer contact the iron surface,

because the reaction kinetics provide insufficient energy for the methane to

diffuse through the carbon layer. In this case, the final carbon morphology

will be a cluster carbon nanoonion as shown in Figure 9. Reaction conditions

can be chosen to promote metal dusting and prevent premature deactivation.

Overall, the number of metal dusting cycles depends on the size of the poly-

crystalline iron particles, the size of the individual crystals within the particle,

and the reaction conditions.

At temperatures over 900 ◦C, the rate of methane pyrolysis (non-catalytic

methane decomposition) increases, which produces amorphous carbon that

accelerates the deactivation of the iron particle [13].

2.4 Ironsands Catalysis of Thermal Methane Decom-

position

This research is focused on the use of titanomagnetite from New Zealand iron-

sands as a TCMD catalyst. Titanomagnetite has the structure Fe3–xTixO4,

where 0 ≤ x ≤ 1. It is a solid solution of magnetite (Fe3O4, x = 0) and

ulvöspinel (Fe2TiO4,x = 1), with an inverse spinel cubic structure as shown

in Figure 10. Titanomagnetite’s structure can be restated as xFe2TiO4(1-x)Fe3O4

to better indicate its composition [30]. For New Zealand ironsand, x is com-

monly between 0.25 and 0.28 [51].

27



Figure 10: A schematic of the spinel cubic crystal structure unit cell. In
magnetite, the grey atoms are O, the green atoms are Fe2+, and the blue
atoms are Fe3+. In titanomagnetite, titanium is incorporated into the lattice
by substituting two green Fe3+ atoms with a Ti4+ atom and an Fe2+ [38]. In
naturally occurring titanomagnetite such as in ironsands, small amounts of
other metals such as Al3+ and Mg2+ are also substituted. Image by David
Schrupp, generated from data by W.H. Bragg and F.R.S. Cavendish [18].
Reproduced under Creative Commons Licence CC BY-SA 2.0 DE.

The literature does not contain information on the catalytic activity of

TiFe3O4 for TCMD, but it does have information on how the titanium in

the iron ore affects the formation and stability of cementite (Fe3C), the key

intermediate in the carbide cycle.

During the formation of cementite, titanomagnetite had a slower reduc-

tion rate than hematite, and was more effectively reduced when the feed gas

composition included up to 30% hydrogen, while hematite performed better

under methane autoreduction. This is relevant to TCMD because the re-

duction of the catalyst affects the methane decomposition step, as discussed

earlier, and so slower reduction may delay the reaction. The slow reduction of

28



titanomagnetite is due to its stability and crystal structure. By pre-oxidising

the titanomagnetite to titanohematite, its reduction rate was increased to

almost that of hematite [30].

The overall rate of cementite formation was also slower for titanomag-

netite than for hematite, even when the reduction rate had increased. This

was thought to be because of unreduced titanium, magnesium, and calcium

oxides within the crystal structure. CaO is a known suppressor of cementite

formation, although this may be simply due to the lower availability of metal

active sites due to the presence of oxygen [30].

Titanium in the iron ore increases the stability of cementite relative to

pure hematite, and this stability is at a maximum between 750 and 770 ◦C

[31]. Because cementite decomposition is an intermediate step in the TCMD

process, higher cementite stability is likely to slow the reaction and cause it

to require more energy to proceed.

29



3 Materials And Methods

3.1 Experimental Setup

Two different tubes were used as reaction vessels in experiments for this

project. The first was a 500 mm long, 38 mm OD stainless steel tube,

with a K-type thermocouple (Nickel-chromium, -270 ◦C to 1260 ◦C ±2.2◦C)

inset into one end cap, that allowed accurate temperature readings in the

reaction zone at the centre of the tube. The ends of the tube were sealed

with clamped Teflon gaskets for experiments at temperatures below 900◦C,

and copper gaskets for experiments at temperatures of 900◦C or higher. The

second was a 40 mm OD quartz tube, 1000 mm in length, with clamped

Teflon O-ring seals for experiments at all temperatures. At low flow rates

and high temperatures, the stainless steel tube seals proved less reliable, and

so the quartz tube was used.

The vessel was heated using a Carbolite 1200◦C Split Tube Furnace (HST

Model 12/400). The catalyst was deposited in the centre of the tube using a

long-handled scoop. All runs used 1.85 g of catalyst.

Cylinders of high purity methane and argon gas (99.995% and 99.99%

purity respectively, BOC Gases) were linked to the setup using a Swagelock

90◦ ball valve to allow on-stream gas selection. The gas line then passed

through a pressure gauge and a rotameter for flow control. A schematic and

photo of the experimental setup are shown in Figure 11.

The rotameter was calibrated for air at 100 kPa, so the nominal flow rate

was converted to the true flow rate of methane using

Corrected Flow Rate = Nominal Flow Rate ×
√

SGair

SGmethane

(8)

where SGair and SGmethane are the specific gravities of the respective gases at

30



(a) Experimental setup schematic. The gaskets on either end of the tube were
clamped, and could be opened to load and unload the tube.

(b) Photo of experimental setup, with stainless steel reactor tube.

Figure 11

31



100 kPa and 18 ◦C (SGair = 1 and SGmethane = 0.54785). Nominal flow rates

of 0.5, 0.25, 0.05, and 0.01 L/min were used in this research, corresponding

to real flow rates of 0.67, 0.34, 0.067, and 0.013 L/min respectively.

After the catalyst was loaded and the vessel sealed, the reaction chamber

was purged with argon for 45 minutes at approximately 0.5 L/min, while the

reaction vessel was heated to the desired temperature at a rate of 15 ◦C/min.

When the tube was at temperature, the argon was replaced with methane at

the desired flow rate. Samples were taken via a septum periodically from the

exit gas stream using plastic gas-tight syringes. The interval between samples

varied between 5 and 30 minutes, depending on the flow rate and purpose

of the run. At the beginning of an experimental run, during the activation

of the catalyst, sampling was typically done every 10 minutes for the first

hour, and then every 30 minutes for the remainder of the run. The samples

were analysed using gas chromatography to determine the conversion rate of

methane to hydrogen. The remaining gas was then vented.

Upon completion of the experiment, the methane supply was again re-

placed with argon and the vessel was allowed to cool to room tempera-

ture over a period of 3-4 hours. Any carbon produced by the experiment

was removed and weighed. Carbon samples were analysed using Thermo-

Gravimetric Analysis (TGA), Powder X-Ray Diffraction (PXRD), and Ra-

man Spectroscopy.

3.2 Analytical Model

This section describes the methods used to calculate the conversion of methane

to hydrogen, the maximum conversion efficiency for a given temperature, and

the level of graphitisation of carbon samples.

32



3.2.1 Methane Conversion

A gas chromatograph (GC) was used to measure the volume percentages of

CH4 and H2 in the syringe samples from the exit gas stream of the reaction

tube. To do this, the integrals of the CH4 peak (Retention time ≈ 7.6 mins)

and the H2 peak (Rt ≈ 2.1 mins) were input into their respective standard

curve equations (See Appendix A.1) to give the volume percentage (vol%).

To determine the percentage conversion efficiency of methane to hydrogen

(XCH4), two methods were used. The first uses only the volume percentage

of hydrogen [6]:

XCH4 =

(
%H2

200 − %H2

)
· 100 (9)

The second uses the volume percentages of both methane and hydrogen [16]:

XCH4 =

(
%H2

2

%CH4 + %H2

2

)
· 100 (10)

Assuming that methane and hydrogen are the only gases present in the exit

stream of the reactor, these methods produce the same result mathematically,

but sometimes differences between them arise due to inaccuracies in the GC

measurement. The average of both methods was used in this research in order

to minimise the effects of such inaccuracies. The mathematical equivalence

of these two equations is proven in Appendix C, the errors associated with

GC measurements are discussed in Section 4.4.1, and an example of a gas

chromatograph is shown in Appendix B.

The assumption that only H2 and CH4 were present in the exit stream

was expected to be valid based on similar experiments in literature [6, 13,

16], and was retained in all data analysis for this project. However, as dis-

cussed in Section 4.4.1, there was evidence of water vapour in the GC exit gas

samples later in the project. No correction for this was made, because the

33



gas chromatograph used to measure the exit gas stream was not equipped

to measure moisture content. Had this problem been anticipated, a sepa-

rate analytical technique or a gas chromatograph with a different detector

could have been found that would have allowed quantification of these water

volumes and improved the quality of a subset of the data.

Two tube reactors were used in this research: a stainless steel tube, and

a quartz tube. Blank baseline runs at each temperature and flow rate were

run for each tube, and these values have been subtracted from all methane

conversion ratio values presented in this research, unless otherwise stated.

Baseline values can be found in Appendix A.2. In the names of experimental

runs, a Q indicates that the quartz tube was used, otherwise, the steel tube

was used. Because of the baselining process, the tube type is a controlled

variable and was ignored when calculating trends and drawing conclusions.

As a result, on the same graph, some series names may contain a Q, and

others may not.

3.2.2 Thermodynamic Equilibrium Limit

The Thermodynamic Equilibrium Limit (TEL) is the maximum value of

XCH4 thermodynamically possible under a given set of reaction conditions.

To calculate the TEL of the conversion ratio of methane to hydrogen for a

given temperature, the equilibrium constant of the reaction at that temper-

ature (KT ) must be calculated. The TEL can then be expressed in terms of

KT , temperature, and pressure [14].

The equilibrium constants and thermodynamic equilibrium limits used

in this research were calculated using the method outlined in the Support-

ing Information of Cornejo et al. [14]. The KT and TEL values for each

temperature considered in this research are displayed in Table 1.

34



Table 1: Equilibrium constants and thermodynamic equilibrium limits for
the conversion of methane to hydrogen in TCMD. The pressure was constant
at 1 bar.

Temperature (◦C) KT TEL
800 21.849 91.9%
850 34.331 94.6%
900 51.988 96.4%
950 76.194 97.5%

3.2.3 Characterisation of Carbon/Catalyst Particles

To determine the level of graphitisation of the carbon produced in this re-

search, four indicators were used: mass-loss temperature, graphitic degree,

crystal size, and disorder.

The 50% mass-loss temperature of carbon in TGA is an indication of

the graphitisation of the sample. TGA measures the thermal stability of

a carbon sample by heating it under air and observing the 50% mass-loss

temperature. An example of a TGA curve can be found in Appendix B.

Amorphous carbon is more easily oxidised than ordered graphite, resulting

in mass loss at lower temperatures during TGA. Amorphous carbon typically

has a 50% mass-loss temperature of below 450 ◦C. Above 550 ◦C indicates

well graphitised carbon [5, 22]. However, the reliability of TGA 50% mass-

loss temperatures as a precise predictor of graphitisation is uncertain, and

may not be useful for distinguishing between samples with small mass-loss

temperature differences. For the remainder of this research, the temperature

at which a TGA carbon sample has lost half its mass is referred to as the

TGA half-temp.

The graphitic degree of a sample is calculated using Powder X-Ray Diffrac-

tion (PXRD) data. PXRD measures the x-ray diffraction spectrum of the

carbon/catalyst particle and uses it to calculate structural information about

the crystals in the sample. First, the position of the graphite 002 peak in the

35



PXRD spectrum is used to calculate the interplanar spacing d according to

Braggs Law:

d =
λ

2sinθ002
(11)

where λ is the radiation wavelength, and θ002 is the (002) reflection angle.

The layer spacing is then compared to that of ideal graphite to determine

the fractional degree of graphitisation g :

g =
(3.440 − d)

(3.440 − 3.354)
(12)

where 3.354 is the ideal layer spacing of a single crystal of graphite without

defects, and 3.440 is the layer spacing of turbostratic carbon (carbon with

crumpled or randomly folded layers) [7]. The higher the fractional degree of

graphitisation, the more graphitic the carbon sample is.

The average crystal size of the carbon/catalyst particles (called PXRD

Average Particle Diameter or Avg PD in this work) can also be calculated

from PXRD data, using diffraction peak broadening and the Scherrer for-

mula:

dc =
kλ

β 1
2

cos θ
(13)

where dc is the average diameter of a carbon/catalyst sample crystal, k is

the Scherrer coefficient (approximately equal to 1), λ is the is x-ray radiation

wavelength (1.5405 Å), β 1
2

is the Full Width at Half Maximum (FWHM) of

the diffraction peak in radians, and θ is the Bragg angle of the peak maxi-

mum [49]. The Scherrer formula may produce different values when different

diffraction peaks are used, due to anisotropy of the sample particles. How-

ever, it is useful as a comparative measure of graphitisation, as larger crystal

sizes tend to indicate higher graphitisation of the sample. In this research,

the (002) peak was used to calculate all crystal size values. An example

PXRD spectrum collected as part of this research is shown in Appendix B.

36



Because the FWHM is used to calculate crystal size, it is affected by peak

broadening caused by the presence of amorphous carbon or other materials

with low crystallinity. Graphitic degree, on the other hand, uses the Bragg

angle, and is not affected by the presence of these materials. As a result, these

two indicators of carbon quality sometimes diverge for the same sample. In

these cases, the graphitic degree is taken as an indicator of the core graphite

within the shell of amorphous carbon, while the crystal size measurement

encompasses the disorder of the amorphous carbon and gives a better average

indication of the graphitisation of the entire sample [13].

Disorder in the carbon sample implies that it is less graphitic, and Raman

spectroscopy can be used to estimate this. The Raman spectrum of graphite

contains two key peaks: The D band, at 1341 cm−1, and the G band, at

1568 cm−1. The intensity of the D band is an indication of disorder, as it

is produced by the vibrations of atoms with dangling bonds, such as those

in defects in the lattice. No D peak is present in defect-free, highly ordered

pyrolitic graphite. The intensity of the G band is produced by vibrations

of sp2 bonded atoms in the graphite lattice, and is high in highly ordered

graphite. The ratio of the intensities of the D and G peaks (ID/IG) is often

used to indicate the level of disorder in a carbon sample, where a low ratio

corresponds to a low level of disorder [32, 52]. An example of a Raman

spectrum collected in this research can be found in Section B.

Raman spectroscopy has limitations that affect the reliability of its pre-

dictions of carbon disorder. Carbon strongly absorbs visible light, meaning

most sub-surface scattering is re-absorbed by the sample. As a result, the

scattering received by the detector is predominantly from the surface of the

sample. While the sample particles were mixed, they were not ground or

crushed before characterisation, so the measures of disorder produced by

Raman analysis are only representative of the surface carbon. In addition,

Raman spectroscopy has a small focal point of less than 50 microns, which

37



gives a small collection volume and is therefore subject to sub-sampling in-

accuracies. Several spectra were typically averaged to account for this, but

it may still affect the results.

3.3 Instrumentation

Gas samples were analysed using a Shimadzu GC-2014 Gas Chromatograph

with an Alltech CTR I Concentric Packed Column (Category 8700) and

a Thermal Conductivity Detector. Nitrogen was used as the carrier gas.

Syringe samples were manually injected into the chromatograph.

All TGA measurements were carried out using an Alphatech Systems

TGAQ50 instrument, platinum sample holders, and alumina sample pans.

The carbon-coated catalyst particles were heated from room temperature to

850◦C at 10◦C/min in a linear temperature ramp, with an air flow rate of

0.4 L/min.

All Powder X-ray Diffraction measurements were carried out on a Rigaku

Spider X-ray diffractometer with Cu Kα radiation (Rigaku MM007 micro-

focus rotating-anode generator), monochromated and focused with high-flux

Osmic multilayer mirror optics, and a curved image plate detector.

All Raman Spectroscopy measurements were made using a home-built

Raman microscope based on an Olympus IX70 inverted fluorescence micro-

scope with 532 nm diode laser excitation. Excitation was spectrally filtered

using an OptiGrate Volume Bragg bandpass filter, then directed to the sam-

ple by an Iridian Spectral Technologies Raman edge filter. Excitation was

focused onto the sample with a NA = 0.65 (40x magnification) objective.

Typical excitation power at the sample was < 2 mW (to avoid sample dam-

age). Raman scattering was collimated with the same objective and residual

Rayleigh scattering was removed with an additional Iridian Spectral Tech-

nologies Raman edge filter before being focused in free space to the 50 micron

38



slit of a Princeton Instruments FERGIE spectrometer.

The major oxide compositions of ironsands used in this research were

determined using X-Ray Fluorescence. All XRF measurements were carried

out by SpectraChem Analytical, CRL Energy Ltd. Samples were oven dried

at 110 ◦C under air, and then heated to 1000 ◦C in air for 1 hour to oxidise

them. Weight changes from this process are displayed in relevant tables as

a percentage Loss On Ignition (LOI). Samples were prepared using borate

fusion before XRF analysis.

Particle size distribution measurements were made using a Malvern Mas-

tersizer 3000.

3.4 Materials

The raw materials used in this research are as follows:

• Patea ironsand (I): Raw ironsand taken from the beach at Patea, in

the Taranaki region.

• Magnetically purified Patea Ironsand (IP): Patea ironsand, hand puri-

fied using a magnet.

• Glenbrook ironsand (G): Ironsand from the Glenbrook sand mine at

the Waikato Heads. This sand was pre-purified using the industrial

magnetic concentrators at the mine.

• Hematite control (H): 99.4%, <5 µm, Milton Adams. Every effort was

made by the university to obtain hematite with a size distribution closer

to that of the ironsand, but at the time, this was the only commercially

available hematite at high purity.

The composition of each type of ironsand is shown in Table 2. After

magnetic purification, the iron oxide content of the Patea ironsand increased

from 52.0% to 80.5%. The iron oxide and corresponding elemental iron con-

tent of each material are listed in Table 3. In ironsand, the iron oxide present

39



is magnetite, or Fe3O4, which has 2.4% more Fe by weight than the hematite

control, Fe2O3.

The XRF analysis in Table 2 shows that purified Patea and Glenbrook

ironsands contain about 7.5% titanium. However, XRF does not show the

way in which the titanium is incorporated into the magnetite lattice as

ulvöspinel, as discussed in Section 2.4. Normally XRD analysis would to ob-

tain this kind of structural information. However, magnetite and ulvöspinel

have the same spinel cubic crystal structure, making it impossible to resolve

the difference with the available laboratory XRD instrument [51].

The XRF analysis also shows that calcium, silicon, and magnesium oxides

were removed during magnetic purification. For example, unpurified Patea

Ironsand contained 22.54% SiO2, which decreased to 4.37% after purification.

This shows that these elements were not incorporated into the magnetite

lattice, but were instead present as separate oxide crystals. The fact that the

percentage of titanium increased with purification shows that it was in fact

incorporated into the iron oxide lattice as titanomagnetite.

Particle size distribution data for Patea ironsand and hematite is shown

in Table 4.

40



Oxide Patea (I) Pure Patea (IP) Glenbrook (G)
Fe2O3 51.97 80.46 83.75
MnO 0.54 0.65 0.63
TiO2 4.81 7.44 7.77
CaO 8.62 1.51 0.73
K2O 0.14 0.08 0.03
SO3 < 0.01 < 0.01 < 0.01
P2O5 0.29 0.37 0.12
SiO2 22.54 4.37 2.63
Al2O3 4.31 3.94 3.69
MgO 7.57 3.48 2.99
Na2O 0.41 0.21 0.07
V2O5 0.34 0.54 0.58
LOI -1.57 -2.94 -3.07
SUM 99.97 100.11 99.93

Table 2: XRF major oxides analysis showing the composition of ironsands
used in this research. Values are expressed as weight percentages. LOI is
Loss On Ignition, a measure of the mass change when the sample was heated
before analysis.

Catalyst Type % Iron Oxide % Fe by Weight
Patea Ironsand (I) 51.97% 36.35%
Pure Patea (IP) 80.46% 56.28%
Glenbrook (G) 83.75% 58.58%
Hematite (H) 100.00% 69.94%

Table 3: Iron content of catalysts used in this research.

Catalyst D10 D50 D90 D4,3 D3,2

I 157 265 446 286 245
IP 35.4 180 296 182 44.0
G 15.7 128 202 124 25.8
H 0.426 0.984 2.78 60.9 2.01

Table 4: All values are in units of µm. Particle size distribution data of
catalysts used in this research. Dx values indicate that x% of particles had
a diameter lower than the given value. D4,3 refers to the volume moment
mean diameter. D3,2 refers to the surface area moment mean diameter.

41



4 Results and Discussion

The goal of this research was to investigate the effect of as many variables

as possible on the performance of ironsands for TCMD, and compare them

to the hematite control. This meant constructing an experimental matrix of

the highest priority variable combinations that would be manageable in the

time allotted. With the exception of reaction pressure, which was rejected for

technical reasons, catalyst type, temperature, flow rate, and catalyst particle

size were identified as being the most useful parameters of a commercial pro-

cess. Experiments covering all combinations of these variables were planned,

but as the work went on, and there were repeated setbacks related to exper-

imental equipment, it became clear that the experimental matrix would not

be completed. The final variables considered are listed here:

• Catalyst type:

– Hematite (H), Unpurified Patea Ironsand (I), Purified Patea Iron-

sand (IP), and Glenbrook ironsand (G)

• Temperature:

– 750, 800, 850, 900, and 950 ◦C

• CH4 flow rate:

– 0.67, 0.035, 0.067, and 0.013 L/min

• Particle Size:

– Natural size distribution, and <100 µm

It would have been beneficial to test additional particle sizes. In addition,

the characterisation of catalysts in this research would have benefited from

optical microscopy and SEM images. In hindsight, this should have been pri-

oritised over other types of characterisation from the beginning, but it was

not. In part, this was due to the fact that this research was commissioned

by an industry partner, and the comparison of ironsand with hematite was

a key deliverable included by them. Had there been less focus on compar-

42



ison with hematite, the project might have focused more strongly on the

characterisation of the ironsand itself.

To provide insight into the decision making process and the flow of ex-

perimental work, a timeline is given here:

• A short test run with hematite at 850 ◦C and 0.67 L/min CH4 flow

rate was carried out to prove that the experimental setup worked.

• 4-5 hour, 0.67 L/min runs of unpurified Patea ironsand (I) and hematite

(H) were carried out at 850 ◦C and 950 ◦C as a preliminary test of

temperature effects, and then additional runs at 750, 800, and 900 ◦C

for both H and I were done to finish off this section of the experimental

matrix.

– 800 ◦C was confirmed as the activation temperature for the reac-

tion, so no further experiments were carried out at temperatures

lower than 850 circC.

• Glenbrook sand (G) was obtained and tested at 0.67 L/min and 850 ◦C.

This sand performed much better than I and H under these conditions,

which incentivised the magnetic purification of the Patea ironsand.

• Purified Patea Ironsand (IP) was tested under the same conditions.

• IP was then sieved to <100 µm to investigate particle size effects at

0.67 L/min and 850 ◦C. There was no significant difference in activity

observed, so the priority of this variable was lowered.

• Following this, IP was used to begin exploring flow rate effects at 850
◦C, by stepping down the flow rate gradually during a single run and

measuring the changes in methane conversion ratio.

– During these experiments, problems with the seals at low flow

rates for the stainless steel tube caused delays while a quartz tube

was obtained.

– Additional flow meters also had to be obtained during this time.

– 0.013 L/min was found to be the lowest flow rate that was unaf-

43



fected by mass transfer issues in the new quartz tube.

• At 0.013 L/min, IP was used to attempt a full deactivation run overnight

in the quartz tube. The maximum allocated time for a single run on the

shared furnace was 24 hours, including heating and cooling, allowing

approximately 20 hours for deactivation. At this flow rate, the cata-

lyst had not deactivated in time. This was attempted twice before the

decision was made to raise the flow rate.

• Further flow rate studies were carried out to try and find a flow rate

that would allow deactivation in the 20 hr period while not sacrificing

methane conversion too much.

• A flow rate of 0.067 L/min was decided upon for complete deactivation

runs, and deactivation of both IP and H was observed within 18 hours

at this flow rate and 850 ◦C.

• A time consuming complete deactivation temperature study was then

carried out, with 18 hr H and IP runs at 0.067 L/min and 850, 900,

and 950 ◦C.

• A 0.013 L/min, <100 µm IP run was done.

• Throughout this time, duplicate runs were performed where time al-

lowed.

The results of each experiment are described in terms of dependent vari-

ables, which are loosely split into two categories: “Catalyst Activity and

Decay”, and “Carbon Quality”. These two categories are used to structure

the results section.

The Catalyst Activity and Decay section encompasses variables that pro-

vide information on how effectively the reaction produced hydrogen. This

includes conversion ratio data, along with time data documenting the times

of various points of interest, such as the time of maximum conversion ratio,

or time until loss of activity of the catalyst. It also includes carbon and

hydrogen yield values.

44



The Carbon Quality section encompasses variables that describe the level

of graphitisation of carbon produced in the reaction. These variables are

obtained from carbon characterisation data.

Both sections draw upon raw data included in two key summary tables.

The first, Table 5, contains raw data from each experimental run, with repli-

cate runs displayed individually. The second, Table 6, contains data from all

experimental runs, relative to each run’s respective control. Replicate runs

were averaged before the control value was subtracted to produce the values

in this table.

45



Table 5: Summary of all results obtained in this research. Missing values are explained below, after their
respective column headings. A: Max Conversion Ratio(%); B: tmax (min); C: SCA1 (%), value missing due
to insufficient samples after late maximum; D: t 1

2
max− tmax (min), low flow rate runs had no t 1

2
max value due

to slow decay; E: Carbon yield (g C / g Fe); F: Mass carbon (g); G: TGA Half-temp; H: PXRD Graphitic
Degree; I: PXRD Avg Particle Size (nm); J: Raman D/G Ratio, could not acquire a value from HQ850 or I850
runs, as no G peak was present. The I950 duplicate run was mishandled and its carbon was not characterised
by any method.

Table 5: Summary of Results

Catalyst

Type

Tube

Type

Temp.

(◦C)

ttotal

(hrs)

Flow

(L/min)

Size

(nm)
A B C D E F G H I J

H S 800 4 0.67 Normal 8.5 120 120 1.67 1.60 570 0.865 142.6 0.215

I S 800 4 0.67 Normal 4.7 55 91.2 95 3.70 1.31 615 0.484 142.8 0.453

G Q 850 5 0.013 Normal 73.2 210 90.7 1.69 1.06 620 0.836 193.8 0.325

G Q 850 5 0.013 Normal 79.0 180 94.8 1.68 1.05 614 0.836 188.5 0.158

H Q 850 4.5 0.013 Normal 65.6 180 91.0 1.00 0.73 584 0.996 222.8

H S 850 7 0.013 Normal 69.0 60 85.5 360 1.99 2.02 571 0.836 156.4 0.239

IP Q 850 5 0.013 <100 µm 70.4 180 95.8 1.59 0.85 620 0.880 187.1 0.477

IP Q 850 5 0.013 <100 µm 58.2 210 80.6 1.50 0.75 612 0.923 187.4 0.258

IP Q 850 4.5 0.013 Normal 73.1 180 94.8 1.79 1.05 623 0.793 191.7 0.445

IP Q 850 20 0.013 Normal 66.1 180 76.7 6.97 6.45 610 0.851 190.8 0.274

IP S 850 8 0.013 Normal 80.1 240 91.0 4.15 3.52 607 0.880 205.4 0.186

Continued on next page

46



Catalyst

Type

Tube

Type

Temp.

(◦C)

ttotal

(hrs)

Flow

(L/min)

Size

(nm)
A B C D E F G H I J

H Q 850 18.5 0.067 Normal 41.5 90 3.07 3.42 612 0.865 148.0 0.372

H Q 850 18.5 0.067 Normal 37.6 45 83.3 3.21 3.60 610 0.909 153.0 0.242

IP Q 850 18.5 0.067 Normal 70.9 90 67.6 4.96 4.36 622 0.880 204.0 0.280

IP Q 850 18.5 0.067 Normal 63.5 90 81.7 5.41 4.83 626 0.793 165.2 0.282

IP Q 850 15.5 0.067 Normal 4.92 4.31 623 0.778 159.6 0.276

IP S 850 4 0.34 Normal 35.7 35 45.3 20 2.86 2.17 616 0.602 163.8 0.322

G S 850 5 0.67 Normal 18.8 35 36.0 20 3.09 2.59 618 0.573 145.8 0.318

H S 850 5.5 0.67 Normal 11.9 45 47.7 45 2.01 2.05 570 1.024 186.7 0.228

I S 850 5 0.67 Normal 11.0 30 42.1 30 4.37 1.76 585 0.631 147.5

IP S 850 4 0.67 <100 µm 19.7 45 43.6 15 2.70 2.00 610 0.602 155.2 0.378

IP S 850 4.5 0.67 Normal 18.4 45 45.7 15 2.95 2.26 622 0.602 153.8 0.353

H Q 900 19 0.067 Normal 30.3 60 72.3 3.44 3.89 643 0.365 98.0 0.932

H Q 900 19.25 0.067 Normal 47.1 45 58.1 60 3.38 3.82 669 0.880 146.2 1.199

IP Q 900 18.5 0.067 Normal 59.7 45 31.7 30 9.42 9.00 617 0.793 156.4 0.268

IP Q 900 18.5 0.067 Normal 29.1 45 71.1 7.44 6.93 625 0.865 187.0 0.312

H S 900 4 0.67 Normal 13.6 15 46.4 20 1.81 1.79 610 0.836 181.9 0.264

I S 900 4 0.67 Normal 18.4 20 24.2 10 6.03 2.88 614 0.513 141.5 0.348

IP Q 950 20.5 0.013 Normal 56.2 80 93.8 5.57 4.99 632 0.938 229.3 0.224

H Q 950 18.5 0.067 Normal 51.1 30 33.9 30 7.36 8.97 695 0.793 106.1 1.386

Continued on next page

47



Catalyst

Type

Tube

Type

Temp.

(◦C)

ttotal

(hrs)

Flow

(L/min)

Size

(nm)
A B C D E F G H I J

H Q 950 18.5 0.067 Normal 56.8 30 28.9 30 5.37 6.39 684 0.793 110.4 1.300

IP Q 950 18 0.067 Normal 32.2 30 37.1 90 12.36 12.06 681 0.807 171.1 0.859

IP Q 950 18.5 0.067 Normal 26.1 30 19.6 45 11.44 11.11 680 0.851 199.0 0.896

H S 950 4 0.67 Normal 17.2 15 42.5 30 1.81 1.79 637 0.907 153.5 0.962

I S 950 4 0.67 Normal 16.0 15 67.2 75 7.36 3.77

I S 950 4 0.67 Normal 22.4 10 41.9 45 7.60 3.93 662 0.719 181.7 0.765

IP S 950 4 0.67 Normal 26.3 10 52.5 60 6.60 6.06 659 0.880 217.3 0.535

IP S 950 4 0.67 Normal 26.3 10 52.5 60 6.60 6.06 659 0.880 217.3 0.535

48



T
ab

le
6
:

C
on

tr
ol

-r
el

at
iv

e
D

ep
en

d
en

t
V

ar
ia

b
le

R
es

u
lt

s
S

u
m

m
a
ry

C
at

al
y
st

T
y
p

e
T

u
b

e
T

y
p

e
T

em
p

.
(◦

C
)

t t
o
ta
l

(h
rs

)
F

lo
w

(L
/m

in
)

S
iz

e
(n

m
)

A
B

C
D

E
F

G
H

I
J

I
S

8
00

4
0.

67
N

or
m

al
-3

.8
-6

5
-2

5
2.

03
-0

.2
9

45
.2

-0
.3

81
0.

2
0
.2

3
7

G
Q

8
50

5
0.

01
3

N
or

m
al

10
.4

15
1.

8
0.

69
0.

32
33

.3
-0

.1
59

-3
1
.7

0
.0

0
3

IP
Q

8
50

5
0.

01
3

<
10

0
µ

m
-1

.4
15

-2
.8

0.
55

0.
07

32
.5

-0
.0

94
-3

5
.5

0
.1

2
9

IP
Q

8
50

2
0

0.
01

3
N

or
m

al
0.

4
0

-1
4.

3
26

.8
-0

.1
45

-3
2
.0

0
.0

3
5

IP
Q

8
50

4
.5

0.
01

3
N

or
m

al
7.

5
0

3.
8

0.
79

0.
32

39
.7

-0
.2

03
-3

1
.0

0
.2

0
6

IP
S

8
50

8
0.

01
3

N
or

m
al

11
.2

18
0

5.
5

2.
16

1.
50

35
.7

0.
0
44

4
9.

0
-0

.0
5
3

IP
Q

8
50

18
.5

0.
06

7
N

or
m

al
27

.7
23

2.
05

0.
99

12
.6

-0
.0

70
2
5.

7
-0

.0
2
8

G
S

8
50

5
0.

67
N

or
m

al
6.

9
-1

0
-1

1.
7

-2
5

1.
08

0.
54

48
.6

-0
.4

52
-4

0
.9

0
.0

9
0

I
S

8
50

5
0.

67
N

or
m

al
-0

.9
-1

5
-5

.6
-1

5
2.

35
-0

.2
9

15
.4

-0
.3

93
-3

9
.2

IP
S

8
50

4
0.

67
<

10
0
µ

m
7.

8
0

-4
.1

-3
0

0.
68

-0
.0

5
40

.1
-0

.4
22

-3
1
.5

0
.1

5
0

IP
S

8
50

4
.5

0.
67

N
or

m
al

6.
5

0
-2

.1
-3

0
0.

93
0.

21
52

.2
-0

.4
22

-3
2
.9

0
.1

2
5

IP
Q

9
00

18
.5

0.
06

7
N

or
m

al
5.

7
-8

-1
3.

8
-3

0
5.

02
4.

11
-3

4
.8

0.
2
07

4
9.

6
-0

.7
7
5

I
S

9
00

4
0.

67
N

or
m

al
4.

8
5

-2
2.

2
-1

0
4.

22
1.

09
4.

2
-0

.3
23

-4
0
.4

0
.0

8
3

IP
Q

9
50

18
.5

0.
06

7
N

or
m

al
-2

4.
8

0
-3

.0
5.

54
3.

90
-9

.6
0.

0
36

7
6.

8
-0

.4
6
5

I
S

9
50

4
0.

67
N

or
m

al
2.

0
-3

12
.0

30
5.

66
2.

06
25

.0
-0

.1
88

2
8.

3
-0

.1
9
7

IP
S

9
50

4
0.

67
N

or
m

al
9.

0
-5

10
.0

30
4.

78
4.

27
22

.0
-0

.0
27

6
3.

9
-0

.4
2
7

T
a
b

le
6
:

V
al

u
es

sh
ow

n
a
re

th
e

d
iff

er
en

ce
s

of
th

e
ir

on
sa

n
d

va
lu

e
an

d
th

e
h

em
at

it
e

co
n
tr

o
l

va
lu

e
fo

r
a

gi
ve

n
ru

n
.

R
ed

co
lo

u
r

co
d

in
g

in
d

ic
at

es
th

e
ir

on
sa

n
d

p
er

fo
rm

ed
w

or
se

th
an

th
e

co
n
tr

ol
,

w
h

il
e

gr
ee

n
in

d
ic

at
es

b
et

te
r

p
er

fo
rm

an
ce

.
A

:
M

a
x

C
o
n
v
er

si
on

R
at

io
(%

);
B

:
t m

a
x

(m
in

);
C

:
S
C
A

1
(%

);
D

:
t 1 2

m
a
x
−
t m

a
x

(m
in

),
so

m
e

ru
n

s

h
ad

n
o
t 1 2

m
a
x

va
lu

e;
E

:
C

ar
b

o
n

y
ie

ld
(g

C
/

g
F

e)
;

F
:

M
as

s
ca

rb
on

(g
),

on
e

ru
n

h
a
d

a
lo

n
g
er

ru
n
ti

m
e

th
an

th
e

co
n
tr

o
l;

G
:

T
G

A
H

al
f-

te
m

p
;

H
:

P
X

R
D

G
ra

p
h

it
ic

D
eg

re
e;

I:
P

X
R

D
A

v
g

P
ar

ti
cl

e
S

iz
e

(n
m

);
J
:

R
am

a
n

I D
/
I G

R
at

io
.

49



4.1 Catalyst Activity and Decay

To determine if New Zealand ironsand in TCMD is more effective at pro-

ducing hydrogen than the hematite control, gas chromatography data was

collected from the exit gas stream of the reaction. Each sample produced a

conversion ratio value at the time the sample was taken. Figure 12 shows

an example of what this looked like at different flow rates, over a period of 5

hours.

The TCMD process has three distinct phases, as can be seen in Figure

12:

1. Activation - The time period before the conversion ratio reaches max-

imum

2. Stable Conversion - The period where the conversion ratio remains at

or near the maximum value

3. Deactivation - The period where the conversion ratio decreases to zero.

The activation phase is controlled by the reduction of the iron oxide to

elemental iron (ferrite) by methane. As the reduction proceeds, more active

sites become available for the methane to adsorb to, and its decomposition

accelerates. The stable conversion phase and the deactivation phase are

controlled by the metal dusting cycle described in Section 2.3. Metal dusting

requires that the methane be able to diffuse through the growing graphite

layer, decompose, and saturate the crystals within the iron particle with

carbon. If this occurs, then the carbon will precipitate within the particle,

initiating the dusting process. If the graphite layer becomes too thick for

the methane to diffuse to the surface of the iron particle, the particle will be

deactivated before it has dusted completely. As a result, the stable conversion

period will only persist as long as conditions allow the diffusion of methane

to the surface of the iron particle.

Different methods of representing these phases numerically will be used

50



0 6 0 1 2 0 1 8 0 2 4 0 3 0 0
0

2 0

4 0

6 0

8 0

1 0 0

 I P 8 5 0 _ 0 . 6 7
 I P 8 5 0 _ 0 . 3 4
 I P Q 8 5 0 _ 0 . 0 6 7
 I P Q 8 5 0 _ 0 . 0 1 3
 T E L 8 5 0

Me
tha

ne
 Co

nv
ers

ion
 Ra

tio
 (v

ol 
%)

T i m e  ( m i n )
Figure 12: Vol% of CH4 converted to H2 at 850 ◦C over time, by flow rate. IP
refers to purified Patea Ironsand, and the CH4 flow rate in L/min is delimited
by an underscore. The experiments were performed in the steel tube reactor,
unless the quartz tube is specified by a Q in the series name. TEL850 refers
to the Thermodynamic Equilibrium Limit (the maximum conversion %) at
850 ◦C.

later in this section, so it is important to have an understanding of what

the raw data typically looked like. Depending on the flow rate and other

factors, the three reaction phases had different durations, and often did not

have clear delineations. For example, for the runs shown in Figure 12, at

a high flow rate (0.67 L/min), the activation and deactivation phases were

pronounced, while the Stable Conversion phase was non-existent. However,

at low flow rates (0.013 L/min), the deactivation phase was prolonged, and

although not shown on this graph, complete deactivation could take over 20

hours. An ideal catalyst has a long period of stable conversion, and a slow

deactivation.

This section uses two metrics for evaluating the activity of the catalyst,

51



and two metrics for evaluating the decay of that activity. The first metric of

activity is the Maximum Conversion Ratio, which is the highest percentage

of methane converted to hydrogen at a single moment during a run. For

example, in Figure 12, the max conversion ratio for each run is the conversion

ratio at the highest point of the curve. The second metric of activity is

simply the yield of carbon extracted from the reaction tube at the end of the

run. The carbon yield is dependent on the cumulative volume of methane

converted across the entire run, so it is a measure of total or cumulative

activity.

The first metric of decay is t 1
2
max−tmax, which is the time it takes for the

conversion ratio to drop from max to half-max. This quantity is equivalent

to the Half Width Half Maximum (HWHM) for conversion ratio curves that

approach Gaussian, such as those in Figure 12 at 0.67 L/min. The second

metric of decay is the Sustained Conversion Average (SCAt), which is a

measure of how the conversion ratio is sustained relative to its maximum

value, in a given time period after that maximum value was reached. It is

calculated using the following formula:

SCAt =

(
XCH4,t

XCH4,max

)
· 100 (14)

where t is the time elapsed in hours from the time of maximum conversion

ratio, XCH4 is the mean conversion ratio of CH4 to H2 over time period

t, and XCH4,max is the maximum conversion ratio. SCAt is expressed as a

percentage of the maximum value. The higher the value, the lower the decay

in catalyst activity, because the average conversion ratio value approaches

the maximum. The value of t used in this research was always 1 hour.

It is important to note that the SCAt metric has limitations, and there

was some debate about how best to define it. Because it is constructed by

normalising the mean conversion of a run to the maximum for the same

52



run, the SCAt will be higher for a run that has consistently low activity,

and lower for a run that has a high initial peak which then decays rapidly.

This makes SCAt values meaningless when it comes to comparing catalytic

activity. For example, in Figure 12, the SCA1 value for the red IP850 0.34 run

was lower than the SCA1 value for the black IP850 0.67 run. If the intention

had been to define another metric of catalytic activity, the mean conversion

ratio (XCH4,t) would have been used instead, without any normalisation to

the maximum value. However, SCAt was intended to be a useful measure

of decay, not activity. SCAt immediately provides information on the decay

profile, but must be interpreted in conjunction with the max conversion value

to allow the absolute activity of the catalyst to be inferred. XCH4,t gives an

indication of activity over a given time from max, but must be interpreted

in conjunction with the max conversion value to allow the decay profile to

be inferred. It was for this reason that SCA1 was used as a metric of decay.

However, the usefulness of SCA1 for comparing decay trends was also lim-

ited by the data itself. Low sampling rates, particularly for 950 ◦C runs and

0.067 L/min runs, meant that there were often few or no data points for the

hour following the conversion maximum. Any metric that sought to compare

these runs with other runs that had richer data in the hour following the max-

imum was therefore very unreliable. Had there been a larger number of data

points following the maximum conversion ratio for all runs, the SCA1 met-

ric would certainly have proven more useful for comparing the decay of the

catalyst under different conditions. Ideally, an auto-sampler and on-stream

GC instrument would have been used to create a high resolution conversion

ratio plot. In that case, the SCA1 metric would have provided more reliable

information on catalyst decay, and could have been supplemented by SCA3

and SCA10 values for a richer perspective.

To quickly test whether XCH4,t would add additional value as an activity

metric, XCH4,1 was plotted against the max conversion ratio for each run

53



(Figure 13a). XCH4,1 was quite strongly correlated to the max conversion

ratio, with an R2 value of 0.84. This suggested that there would be limited

marginal utility in including XCH4,1 as an additional metric of catalyst ac-

tivity. The correlation between SCA1 and max conversion ratio is shown in

Figure 13b for comparison, with an R2 value of 0.34.

To evaluate the overall lifetime of the catalyst, a complete deactivation

run of Patea ironsand was initially attempted at 0.013 L/min. However, after

20 hrs, the conversion ratio was still 35.6%. As a result, the flow rate was

increased by a factor of 5 to 0.067 L/min. Experiments at this flow rate were

run for approximately 18.5 hrs. Figure 19 on Page 63 shows a set of complete

deactivation runs at 850 ◦C. Because the gas sampling was done manually,

there was a long period overnight where no samples were taken. This caused

some metrics of catalyst evaluation to be unreliable or subject to increased

error at this flow rate. For example, the sampling period at the beginning

of runs at 0.067 L/min was often shorter than for runs at other flow rates,

meaning normal metrics of decay like t 1
2
max−tmax weren’t always calculable.

However, the main purpose of these runs was to weigh the carbon produced

and calculate the total yields of C and H2 from the deactivated catalyst.

Figure 14 compares the hematite control run, H850, with four different

types of ironsand, at a CH4 flow rate of 0.67 L/min. As these runs were all

at the same flow rate, the shape and timing of the conversion ratio curves

was similar, but it was apparent that the Purified Patea Ironsand (IP850,

IP850<100 µm) and the Glenbrook ironsand (G850) had a higher catalytic

activity than the hematite control in these conditions. The difference be-

tween the maximum conversion ratio of each ironsand run and the maximum

conversion of the hematite control is referred to as the control-relative max

conversion value. Control-relative values for max conversion ratio and several

other key dependent variables are displayed in Table 6. Errors and uncer-

tainty pertaining to all measurements are discussed in Section 4.4.

54



(a) Correlation between the mean conversion ratio for 1 hour after maximum, and
the maximum conversion ratio across all runs. R2 = 0.84.

0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0

0

1 0

2 0

3 0

4 0

5 0

6 0

7 0

8 0

1-h
r M

ea
n X

CH
4 (%

)

M a x  X C H 4  ( % )
(b) Correlation between the 1 hour sustained conversion average and the maximum
conversion ratio across all runs. R2 = 0.34.

0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0

2 0 %

4 0 %

6 0 %

8 0 %

1 0 0 %

SC
A 1

M a x  X C H 4  ( % )
Figure 13

55



0 6 0 1 2 0 1 8 0
0

5

1 0

1 5

2 0

2 5

Me
tha

ne
 Co

nv
ers

ion
 Ra

tio
 (v

ol 
%)

T i m e  ( m i n )

 H 8 5 0
 I P 8 5 0
 I P 8 5 0 < 1 0 0 u m
 G 8 5 0
 I 8 5 0

Figure 14: Vol% of CH4 converted to H2 at 850 ◦C over time, by catalyst
type. H is the hematite control, IP is Purified Ironsand, G is Glenbrook
ironsand, <100 µm refers to the particle size distribution of the sand, and
the CH4 flow rate was 0.67 L/min. The experiments were performed in the
steel tube reactor.

4.1.1 Temperatur