View to Its Use in An

Florida State University Libraries

Electronic Theses, Treatises and Dissertations The Graduate School

2017 Electro-Oxidation of Glycerol with Electroless CuNiMoP: Production of Fine Chemicals and Prospects for Co-Generation of Energy Oyidia Elendu

Follow this and additional works at the DigiNole: FSU's Digital Repository. For more information, please contact [email protected] FLORIDA STATE UNIVERSITY

COLLEGE OF ENGINEERING

ELECTRO-OXIDATION OF GLYCEROL WITH ELECTROLESS CUNIMOP:

PRODUCTION OF FINE CHEMICALS AND PROSPECTS FOR CO-GENERATION OF

ENERGY

OYIDIA ELENDU

A Dissertation submitted to the Department of Biomedical and Chemical Engineering in partial fulfillment of the requirements for the degree of Doctor of Philosophy

2017 Oyidia Elendu defended this dissertation on April 11, 2017. The members of the supervisory committee were:

Yaw Yeboah Professor Co-Directing Dissertation

Egwu Kalu Professor Co-Directing Dissertation

Peter Kalu University Representative

John Telotte Committee Member

The Graduate School has verified and approved the above-named committee members, and certifies that the dissertation has been approved in accordance with university requirements.

ACKNOWLEDGMENTS

I would like to first thank my supervisors, Dr. Egwu E. Kalu and Dr. Yaw Yeboah, for the opportunity to do this work. They believed that I could do it and thus, guided, encouraged and often pushed me beyond what I thought was adequate. I see the individual strengths they each brought to bear on this project, and I am grateful.

I would also like to thank my committee members: Dr. Telotte always provided a fresh perspective to the challenges I encountered, and often his suggestions helped to find my way out of a maze. Dr. Onyeozili provided invaluable help in the identification of organic species encountered in electro-synthesis aspect of the project. She often went far and beyond the call of duty, coming down to work with our equipment, if she thought it necessary. I also thank Dr. Peter

Kalu for his help in understanding metal and composite behavior in the electro-catalysts.

I thank my colleagues in the Electrochemical and Renewable Energy Research Group in the FAMU-FSU College of Engineering – past and present. Shannon Anderson, Ever Velasquez,

James Akrasi, Ruben Nelson, Venroy Watson, Wasu Chaitree, Joel Sankar and Joyce Kosivi: thank you for being sounding boards for ideas, willing guinea pigs for presentations, and all the lunch dates. I could not have asked for a better group of colleagues. You guys are truly awesome.

I am grateful to the Eziyis, Kosivis, Ufodikes and Oforis. It would be next to impossible to fail with these guys rooting for you. If any of you are ever in any city where I am, know that you have at least one friend there already.

I think of my family and each person’s contribution to today: my husband, Chidi, it was providence that made our paths cross. Thank you for your unquestioning support, kindness and love; my son Dirim, who brought laughter to many days; my father, Chief Godwin Eze who instilled in me the thirst for knowledge; my mother, Mrs. Jane Eze and mother-in-law, Mrs. iii

Comfort Elendu who looked after my family when I left Nigeria for the USA; my sisters, Dr.

Onyinye Anyaso and Dr. Uzochi Anemene; my brothers – Eze, Ojukwu and Chidi; my aunts- Mrs.

Ngozi Kalu and Mrs Victoria Chuku – voices of reason in a quest in which it was easy to lose focus. I celebrate every one of them today, and honor the memory of those who have passed on.

I acknowledge that this work would not have been possible without funding from the

National Science Foundation and the Department of Biomedical and Chemical Engineering,

College of Engineering, FAMU-FSU College of Engineering.

Above all, my most humble appreciation to God for His infinite grace and mercy throughout my academic pursuit.

TABLE OF CONTENTS LIST OF TABLES ...... viii

LIST OF FIGURES ...... ix

ABSTRACT ...... xiv

1. INTRODUCTION ...... 1

1.1 Background ...... 1

1.2 Sustainable Uses of Glycerol ...... 2

1.3 Catalysts for Glycerol Oxidation ...... 10

1.4 The Technique of Electroless Deposition ...... 13

1.5 Economic Considerations for Glycerol Electro-Oxidation ...... 16

1.6 Research Goals and Organization of Text ...... 17

2. METHODS AND MATERIALS ...... 20

2.1 Introduction ...... 20

2.2 Electroless Bath Formulation ...... 21

2.3 Substrate Preparation ...... 23

2.4 Electrode Preparation ...... 23

2.5 Deposit Characterizations ...... 25

2.6 Fabrication of Electrochemical Reactor...... 26

2.7 Cyclic Voltamograms ...... 27

2.8 Constant Potential Oxidation ...... 28

2.9 Oxidation Product Analysis ...... 28

3. CATALYST SYNTHESIS AND CHARACTERIZATION ...... 30

3.1 Introduction ...... 30

3.2 Choice of Main Active Materials ...... 30

3.3 Electroless Deposition of Copper and Nickel: Reducing Agent Study ...... 32 v

3.4 Effect of Time of Deposition on Deposit Characteristics ...... 36

3.5 Effect of Cu2+ and Dual Reducing Agents in the Electroless Bath ...... 47

4. ELECTROCHEMICAL PERFORMANCE OF CUNIMOP...... 50

4.1 Introduction ...... 50

4.2 Combined Effect of Copper, Nickel, Molybdenum and Phosphorus ...... 50

4.3 Reactions at CuNiMoP Anode ...... 52

4.4 Performance of CuNiMoP/C under Unstirred Cell Conditions ...... 54

4.5 Performance of CuNiMoP/C under Stirred Conditions ...... 59

4.6 Determination of Kinetic Parameters for CuNiMoP/C ...... 61

4.7 Chronoamperometry of Glycerol Oxidation on CuNiMoP ...... 64

5. GLYCEROL CONVERSION AND OXIDATION PRODUCT YIELD DURING POTENTIOSTATIC OXIDATION OF GLYCEROL ON CUNIMOP ...... 67

5.1 Introduction ...... 67

5.2 Thermodynamic Considerations for 3-Carbon Oxidation Products ...... 67

5.3 Controlled Potential Electro-Oxidation of Glycerol ...... 68

5.4 Reaction Mechanism and Pathways...... 79

6. STUDIES IN CATALYST STABILITY ...... 85

6.1 Introduction ...... 85

6.2 Catalyst Loss in Use...... 85

6.3 Copper Behavior under Alkaline Conditions ...... 87

7. CONCLUSIONS AND FURTHER WORK ...... 89

7.1 General Conclusions ...... 89

7.2 Directions for Future Work ...... 90

APPENDICES ...... 91

A. HPLC METHOD DEVELOPMENT ...... 91

B. SPECIATION IN THE MIXED REDUCING AGENT ELECTROLESS BATH ...... 103

REFERENCES ...... 112

BIOGRAPHICAL SKETCH ...... 118

vii

LIST OF TABLES

Table 1: Specific energy of common fuels. Source: DOE, Stanford University, College of the Desert, and Green Econometrics Research ...... 7

Table 2: Cost of common metals used as catalysts. A troy ounce is about 31.1g...... 11

Table 3: Chemicals used in the formulation of Cu-Ni-Mo-P electroless bath ...... 22

Table 4: Reversible potentials for Cu(II), Ni(II), formaldehyde and hypophosphite ions ...... 33

Table 5: Corrosion potentials of Cu and Ni in mixtures of reducing agents ...... 34

Table 6: Loading data for CuNiMoP/C ...... 41

Table 7: Cost of fabricating different catalyst samples. Electricity and labor cost based on graduate student residence and remuneration...... 46

Table 8: Performance of 15, 30, 45 minute samples in glycerol oxidation ...... 46

Table 9: Bath constituents (50 ml plating solution) ...... 47

Table 10: Reversible potentials of species at the anode ...... 54

Table 11: Electrochemical parameters for glycerol oxidation with different catalysts. Oxidations are in alkaline environment ...... 63

Table 12: Predictions of thermodynamic properties of 3-carbon oxidation products ...... 68

Table 13: Used/unused ratio of elements in catalysts ...... 87

Table 15: Concentration versus area data for glycerol ...... 95

Table 16: Retention times for various standards...... 102

viii

LIST OF FIGURES

Figure 1: Formation of biodiesel through trans-esterification of vegetable oil ...... 2

Figure 2: Examples of products from glycerol valorization...... 3

Figure 3: Biomass value pyramid [12]...... 4

Figure 4: Electrolysis cell (A) and fuel cell (B) ...... 6

Figure 5: Schematic showing direction of ionic flow of protons using a cation exchange membrane for alcohol fuel cell under acidic conditions. A-anode. M – Membrane. C - Cathode . 8

Figure 6: Schematic showing direction of flow of hydroxyl ions using an anion exchange membrane for alcohol fuel cell under alkaline conditions. MEA – membrane electrode assembly. A-anode. M – Membrane. C – Cathode ...... 9

Figure 7: Summary of experiments and methods ...... 20

Figure 8: Reactor used for electrochemical studies. Actual reactor (left) and schematic (right) . 26

Figure 9: CVs at different scan rates for 1M glycerol in 1M NaOH solution ...... 31

Figure 10: Oxidation of formaldehyde and sodium hypophosphite by different metals [5] ...... 34

Figure 11: Open circuit behavior of Cu and Ni with respect to changes in pH ...... 36

Figure 12: Effect of deposition time on deposit mass obtained by weighing substrates pre- and post-deposition ...... 37

Figure 13: CuNiMoP/C at 1000x magnification. A:0 min, B:15 min, C:30 min and D:45 min .. 39

Figure 14: Deposit thickness as a function of time ...... 40

Figure 15: Morphology of plain substrate, 15-, 30- and 45 minutes electroless deposits ...... 42

Figure 16: SEM of 45-minute sample at 12000x and 50000x ...... 43

Figure 17: XRD of different CuNiMoP samples ...... 43

Figure 18: CVs showing glycerol electro-oxidation with CuNiMoP/C (a) 15 minutes, (b) 30 minutes, (c) 45 minutes (d) plain and Pd catalysed carbon cloth. Conditions: 25 oC, 1 M glycerol in 1M NaOH, scan rate 10 mV/s ...... 45

Figure 19: Amounts of material deposited on support for formaldehyde, hypophosphite and mixed reducing agent baths as function of Cu/Ni ratio ...... 48

Figure 20: Relative amounts of copper, nickel, molybdenum and phosphorus in deposits from various baths. Mixed reducing agent (left), pure hypophosphite (middle) and pure formaldehyde (right)...... 49

Figure 21: Behavior of electroless copper (Cu/C), electroless nickel (NiMoP/C) and electroless CuNiMoP/C. Scan rate 10 mV/s in 1 M glycerol in 4 M NaOH. Deposition time for each catalyst was 15 minutes...... 51

Figure 22: Applied potential at which glycerol oxidation should be done ...... 53

Figure 23: CVs at different scan rates in 1 M glycerol + 4 M NaOH...... 55

1/2 Figure 24: ip vs. ν for the forward reaction ...... 56

Figure 25: Ep vs ln (ν)1/2 for the forward scan ...... 57

1/2 Figure 26: ip vs. ν for the backward reaction ...... 58

Figure 27: Ep vs ln (ν)1/2 for the backward scan ...... 58

Figure 28: Rotating disc experiment at different speeds. 1 M glycerol in 4 M NaOH ...... 59

Figure 29: Koutecky-evich plots from oxidation peaks on forward scan (-0.2 V and 0.16 V) and backward scan (-0.24V) ...... 60

Figure 30: Current profile during constant potential oxidations ...... 65

Figure 31: Energy density at different potentials ...... 66

Figure 32: Constant potential oxidation at 0.9 V. Conditions: 25 0C, 1 M glycerol + 4 M NaOH. Atmospheric pressure...... 69

Figure 33: Glycerol conversion as a function of applied potential ...... 71

Figure 34: Product buildup on catalyst surface...... 72

Figure 35: Glycerol concentration as a function of time ...... 73

Figure 36: Glycerol conversion (A) and products formed (B) at 0.5 V ...... 74

Figure 37: Glycerol conversion (A) and products formed (B) at 0.7 V ...... 76

Figure 38: Glycerol conversion (A) and products formed (B) at 0.9 V ...... 77

Figure 39: Glycerol conversion (A) and products formed (B) at 1.1 V ...... 77 xi

Figure 40: Glycerol conversion (A) and products formed (B) at 1.1 V ...... 79

Figure 41: Mechanism of formic acid formation ...... 80

Figure 42: Mechanism of formation of glyceraldehyde/DHA ...... 81

Figure 43: Mechanism of glyceric acid formation ...... 82

Figure 44: Mechanism of tartronic acid formation ...... 82

Figure 45: Mechanism of mesoxalic acid formation ...... 83

Figure 46: Mechanism of glycerol oxidation on CuNiMoP/C...... 84

Figure 47: Preferred product(s) at given potentials ...... 84

Figure 48: Atomic percentages of CuNiMoP/C from EDS. Balance is carbon ...... 86

Figure 49: Constant potential oxidations at 0.7 V ...... 87

Figure 50: Speciation of copper at 25 0C and 80 0C ...... 88

Figure 51: UV-Vis absorbance of tartronic acid and glycerol ...... 93

Figure 52: Chromatograms for standards (Glyceraldehyde, DHA, glycerol, mesoxalic acid and tartronic acid)...... 94

Figure 53: Calibration curves: Glycerol, glyceraldehyde, DHA, mesoxalic and tartronic acids .. 96

Figure 54: Glycerol chromatograms. Decreasing concentrations of glycerol and sulfuric acid ... 97

xii

Figure 55: Glyceraldehyde chromatograms at different concentrations (mg glyceraldehyde per ml of solution) ...... 98

Figure 56: DHA chromatograms showing different concentrations ...... 100

Figure 57: DHA presents multiple peaks, five of which show linear behavior...... 100

Figure 58: Mesoxalic acid chromatograms ...... 101

Figure 59: Tartronic acid chromatograms...... 101

Figure 60: Copper speciation in the mixed reducing agent bath ...... 105

Figure 61: Nickel speciation in mixed reducing agent bath ...... 107

Figure 62: Molybdenum speciation at 80 oC ...... 108

Figure 63: Saturation index for molybdenum containing compounds in mixed reducing agent bath ...... 109

Figure 64: Phosphorus speciation in mixed reducing agent bath ...... 110

Figure 65: Saturation index data for phosphorus species ...... 110

xiii

ABSTRACT

Precious metals are the state of the art electro-catalysts for the oxidation of organic compounds, and so are a logical choice for the electro-oxidation of glycerol. Two factors that hinder their use in this regard for commercial applications include their cost and susceptibility to poisoning by the carbonyl (CO) species formed during the electro-oxidation process. The use of inexpensive transition metals as the principal metals in a catalyst composite is thus appealing. In this work, an electro-catalyst composite consisting of copper (Cu), nickel (Ni), molybdenum (Mo) phosphorus (P) was synthesized and used in studying the electro-oxidation of glycerol. The synthesis technique used was electroless deposition, in which autocatalytic reactions in an aqueous bath containing ions of the composites caused deposition of Cu, Ni, Mo and P on an activated substrate. The electrocatalyst was characterized using scanning electron microscopy (SEM),

Energy Dispersive X-ray Spectroscopy (EDX) and X-ray Diffraction (XRD). Various electrochemical techniques, including Cyclic voltammetry (CV), Linear Sweep Voltammetry

(LSV), Chronoamperometry and Electrochemical Impedance Spectroscopy (EIS) were used to characterize the electrocatalyst. Product formation and glycerol conversions were determined with a combination of voltammetry and high pressure liquid chromatography. Suitable conditions for the co-deposition of all four components were found to be dependent on the use of a mixture of reducing agents, ratios of metal ions in solution, pH, oxygen content of the electroless bath and temperature. It was found that amount of deposited material varied non-linearly with time. CV results showed that the composite catalyst as prepared was active for glycerol oxidation in alkaline media and the activity was comparable to that of Pt under similar conditions and better than mono- or bi-metallic combinations of the components. Tafel analysis of the LSV results indicated

xiv exchange current densities ranging from 0.18 – 0.6 mA cm-2 for different deposition times and

Tafel slopes of 120 – 130 mV/decade. Constant potential oxidation of glycerol between 0.5 - 1.3V

(vs. Ag/AgCl) on CuNiMoP/C catalyst showed that the prepared catalyst selectively favored the production of formic acid at the lower potentials. At intermediate potential of 0.9 V, glyceraldehyde/dihydroxyacetone (DHA) are favored while at 1.1 V tartronic acid and mesoxalic acid are the major products. At potentials higher than 1.1 V, competing parasitic reactions reduced the glycerol conversion and product yield. Glycerol conversion of 62% was achieved at 1.1 V compared to 3 – 38% with Pt catalyst under similar conditions. Rates of reaction, measured in terms of current density, were found to be low for potentials lower than 0.7 V. Constant potential oxidations at 0.7 V for three different 24-hour cycles showed catalyst deactivation from leaching of Cu. These results have potential importance in direct alkaline glycerol fuel cell applications.

CHAPTER 1

INTRODUCTION

1.1 Background

Since antiquity (c. 2800 BC), glycerol has been made when fats are heated in ash to make soap [1]. It became an important military resource (as nitroglycerine) during the world wars in first half of the twentieth century, when demand outstripped supply from the soap industry. The chemical industry then stepped in and started the production of glycerol from petrochemical feedstock to meet the deficit. This situation persisted until early in the twenty-first century when glycerol derived from biomass sources drove global production figures to an estimated 200 million tons by 2012 [1]. This is largely due to government legislations and interventions [2] with respect to biofuels and renewable energy, which rendered petroleum derived glycerol non-sustainable.

Currently, synthetic or petroleum-derived glycerol constitutes only about 0.0025% of the total global glycerol output [1, 3].

The fastest growing segment of the renewable energy market is liquid biofuels, one of which is biodiesel. In the US alone, an average of 99,000 barrels/day (b/d) of biodiesel was produced in 2016, and there are projections of 104,000 b/d and 111,000 b/d production in 2017 and 2018 respectively [4]. By stoichiometry, 1 mole of glycerol is made for every 3 moles of biodiesel produced (see Figure 1). Hence, large quantities of glycerol will continue to be made if the biodiesel industry continues to grow.

The main problem with biodiesel derived glycerol is its contamination with methanol, soap and catalysts that originate from the trans-esterification process. It has a variable glycerol content

of (65-85) % and is often yellow or brown [5]. Attempts have been made to modify the trans- esterification process to increase the crude glycerol purity to 90-95% [6]. Estimated purification cost to obtain up to 98 weight % glycerol is about $0.15/kg [7]. This means that the cost of purification can be as much as five times the cost of crude glycerol and limits commercial adoption of this resource.

Figure 1: Formation of biodiesel through trans-esterification of vegetable oil

1.2 Sustainable Uses of Glycerol

Sustainability, as a concept, is increasingly being designed into processes and technologies.

On the energy terrain, these efforts are geared towards replacing fossil fuels with alternative energy. Many oil companies are re-positioning themselves, not just as oil giants, but as energy companies. Existing energy infrastructure, particularly for liquid fuels, was designed to handle distribution of fossil fuels. To utilize such, in the short term, glycerol derived from biodiesel can be used in the manufacture of glycerol tert-butyl ether, an additive that can increase the octane rating [8] of gasoline. In the longer term, glycerol can be reformed into hydrogen or used directly in a fuel cell to generate energy. It can also be combusted directly for low grade heat [9, 10].

Another sustainable avenue to glycerol use is the manufacture of high value chemicals from glycerol feedstock. [11]. Glycerol is a highly-functionalized molecule, and this property makes possible its use as a building block to produce many commodity chemicals (Figure 2). For

2 instance, biodegradable plastics can be made from polylactic acid and absorbent materials can be made from acrolein/acrylic acid.

To profitably utilize low cost glycerol, materials made from it have to be many times more valuable. One way of playing at the top of the biomass value pyramid is through the manufacture of high value chemicals where fluctuations in the price of crude glycerol will not unduly affect the profit margins on products made from it (see Figure 3).

Figure 2: Examples of products from glycerol valorization.

Figure 3: Biomass value pyramid [12]

1.2.1 Glycerol electro-oxidation as a source of chemicals

All oxygenated derivatives of glycerol oxidation are all high value chemicals. Some examples are:

1. Dihydroxyacetone (DHA) is an oxidation product of glycerol that can be used as

a building block in the synthesis of other organic chemicals[13]. DHA, along with

its isomer, glyceraldehyde, can also be produced by chemical oxidation of

glycerol. The amount of product formed is influenced by the acidity of the medium.

Glyceraldehyde is an important sugar in carbohydrate metabolism, and its presence

can indicate a favored pathway in tracing the mechanism of reaction. DHA is used

in manufacture of artificial suntan lotions.

2. Tartronic acid has pharmaceutical applications in the treatment of bone conditions

like osteoporosis and obesity. It is also an anticorrosion agent used in boilers or

other high temperature applications where steam has to be handled, because of its

ability to scavenge oxygen. This property makes it useful in food preservation and

packaging[14] where its high cost ($1564/kg) discourages use in this regard.

3. Mesoxalic acid is a highly functionalized molecule whose calcium salt is a potent

commercial hypoglycemic agent (Mesoxam) and the chlorophenylhydrazone is an

active anti-HIV species[15]. The Ohashi group in Japan has investigated the effect

of mesoxalate in the treatment of diabetes [16]. There have been notable

improvements in the synthesis of mesoxalic acid from tartronic acid, via

consecutive oxidation. However, the supported catalysts used in the electro-

synthesis have low stability in the oxidative environment in addition to existing

problems about selectivity[15].

Various approaches to glycerol oxidation to make these chemicals have been explored.

First, photo catalysis with metal oxide (e.g. TiO2 [17])-aqueous glycerol slurries under ultra violet light irradiation; or photo-electrolysis with metal oxides e.g. Bi2WO6 under ambient conditions[18]. A second oxidation route is by micro-organisms, as in the production of dihydroxyacetone[19]. Thirdly, thermo-catalytic oxidations have been explored in batch operations[14]. These methods all have attendant disadvantages: high cost if ultraviolet irradiation is used (photo catalysis); high temperature and pressure regimes (thermo-catalytic); large reactors, pH/dissolved oxygen content sensitivity and long reaction times (oxidation by microbes). These disadvantages can be circumvented through the use of electrochemical reactors.

Electrochemical reactors may be electrolysis cells or fuel cells (see Figure 4), the difference being that in the electrolysis cell, there is no intention to produce energy. In water electrolysis for instance, electricity supplied to the cell is enough to split water and produce protons. These then 5

combine with electrons flowing from the anode side to form hydrogen gas. Fuel cells on the other

hand, for example the hydrogen fuel cell, produce electricity through oxidation of hydrogen at

the anodes.

Figure 4: Electrolysis cell (A) and fuel cell (B)

These reactors offer several advantages over more traditional oxidation techniques. First, the processes can utilize fewer reagents and lead to an essentially cleaner process that can be easily automated. Second, the oxidation can be carried out at room temperature and atmospheric pressure with a versatility that can promote different reactions based on reaction conditions. Reactors can be sized for different throughputs (from microliters to thousands of cubic meters)[20], though liquid handling issues may complicate the process.

1.2.2 Glycerol on the energy landscape: direct glycerol fuel cell

Traditionally, fuel cells generate energy by oxidizing hydrogen directly to water with the production of energy without any intermediate combustion step. Because of hydrogen safety and handling issues, more innocuous fuels (methanol, ethanol, and glycerol) are being promoted for

6 use in fuel cells. Since these are miscible with water, this has given rise to aqueous direct alcohol fuel cells (DAFC). In comparison to hydrogen, some advantages offered by alcohols include[21]:

1. Lower cost;

2. Relatively high energy density (see

3. Table 1); and

4. Fewer handling challenges

Table 1: Specific energy of common fuels. Source: DOE, Stanford University, College of the Desert, and Green Econometrics Research

Fuel Specific Energy (kJ/g) Energy Density (kWh/gal) lbs CO2/gal

Propane 50.4 26.8 13 Ethanol 29.7 24.7 13 Gasoline 46.5 36.6 20 Diesel 45.8 40.6 22 Biodiesel 39.6 35 19 Methane 55.8 27.0 3 Oil 47.9 40.5 20 Wood 14.9 11.3 9 Coal 30.2 22.9 19 Hydrogen 141.9 10.1 0 Glycerol 18.1 [9] 23.8 15

Early direct alcohol fuel cells contained the alcohol (for instance, methanol) on the anode side, while oxygen/air mixtures were reduced on the cathode side. The anode and cathode compartments were separated by a membrane that allowed the passage of specific ionic species.

These membranes were cation exchange membranes since the species that were transported across

7 them were protons: thus, the electrolytes had to be acidic. For such a membrane to be considered successful, it should conduct these protons while simultaneously providing an adequate barrier to methanol transport. In addition, it must be thermally stable while in use [22, 23]. This is because methanol conducted across the membrane will react at the cathode without producing electricity, leading to a waste of fuel. Such delinquent methanol may also corrode the cathode, which may not have been fabricated to work in environments that contain methanol. Attempts have been made to modify these cation exchange membranes by incorporating pervaporation composites used to break up ethanol- water azeotropes[22]. However, these alcohol cross-over problems persist.

Figure 5: Schematic showing direction of ionic flow of protons using a cation exchange membrane for alcohol fuel cell under acidic conditions. A-anode. M – Membrane. C - Cathode

The development of anion exchange membranes is in part, a response to these alcohol cross-over problems. With a membrane that transports hydroxyl ions, the direction of ionic flow is reversed since the hydroxyl ions flow from the cathode side to the anode side (see Figure 5 and Figure 6). This eliminates possible precipitation that may occur when cations are transported. It also reduces fuel losses that may occur with the drag of the alcohol across the cation exchange membrane.

Under alkaline conditions, water is produced at the anode and consumed at the cathode ensuring better water management.

The emergence of these membranes revived interest in alkaline fuel cells [24]. They had become less popular as advances were made in the use of polymer (solid) electrolyte fuel cells since the latter were not prone to leakages. Advances in anion exchange membranes have largely removed this limitation, and have enabled the use of alkaline electrolyte through which anions are transported.

The ratio of sodium hydroxide to glycerol in the initial reaction mixture is an important parameter in glycerol electro-oxidation. Zhang et al (2012) found that higher glycerol conversions were obtained at a base to glycerol ratio of 4 [25]. They also suggested that these alkaline conditions allowed for faster kinetics as hydroxyl ions catalyzed the initial dehydration step in glycerol electro-oxidation. Not only that, the oxygen reduction reaction occurring on the cathode

side in an alkaline medium occurs at a lower potential compared to that in acidic medium (0.4 V

(SHE) versus 1.23 V (SHE), thus reducing the overall energy cost.

The applied potential at the anode where oxidations occur is an important parameter in

glycerol electro-oxidation as it affects products formation [25]. This is intuitive as potential is a

measure of the energy input and different energy inputs should cause different kinds/levels of

chemical change. Theoretical voltage requirements for a specific reaction can be predicted if

corresponding Gibbs free energies are known. These equilibrium energies give the maximum

voltages that can be obtained from the fuel cell, and thus predictions of possible voltage

requirements of a fuel cell utilizing such a compound can be computed. Where such

thermodynamic quantities are not readily available, they can be estimated by group contribution

methods[26]. In these schemes, an atom or group of atoms is added to (an) original atom(s) that

must have the ability of forming at least two bonds, and this gives a specific increase in the

thermodynamic quantities (e.g. enthalpies of formation and entropies) associated with the original

structure[27]. R2 (coefficient of determination) values greater than 0.99 have been obtained between empirical data and predictions built on these models[28].

1.3 Catalysts for Glycerol Oxidation

Glycerol oxidation is not a spontaneous reaction and so requires both an energy input and catalysis. Transition group metals, notably, the platinum group metals are known to have excellent catalytic properties with respect to hydrocarbons. They have been used in catalytic converters for automobiles (where they catalyze the complete combustion of incompletely oxidized hydrocarbon exhaust fumes), hydrogenation catalysts for hydrogen production and for hydro treating purposes in the petroleum industry. Many of these properties arise from their physical structures at the 10

atomic level; they possess incompletely filled d-sub shells which give them different oxidation

states. As widespread as their use is in these areas, they suffer from economic and ongoing

technical challenges. For example, platinum based catalysts are intolerant to CO concentrations

greater than 50 ppm, and these conditions are difficult to eliminate during hydrocarbon reactions

[29, 30]. Despite these technical drawbacks, platinum group metals continue to be state of the art

catalysts for most hydrocarbon applications. Yet, they are the most expensive metals in common

use (see Table 2).

Table 2: Cost of common metals used as catalysts. A troy ounce is about 31.1g.

Metal Symbol Unit of measure Cost (USD) Platinum Pt Troy ounce 836.00 Palladium Pd Troy ounce 502.00 Rhodium Rh Troy ounce 660.00 Iridium Ir Troy ounce 525.00 Ruthenium Ru Troy ounce 42.00 Osmium Os Troy ounce 400.00 Rhenium Re Rhenium 1150.00 Gold Au Troy ounce 1090.00 Silver Ag Troy ounce 14.13 Copper Cu Pound 2.11 Nickel Ni Pound 4.66

In a study of various carbon and metal oxide supports for gold catalysts, Sobzack et al

(2010) found that even though carbon supported gold catalysts give the best glycerol

conversions[38], Nb2O5 supported gold catalysts gave better performance than V2O5 or Ta2O5.

They went on to add small amounts of copper to their niobium supported gold catalysts for even 11

better results. Liang et al(2011) found that alloying copper into platinum catalysts for glycerol

oxidation increased the conversion from 61.6% to 81.6%, even though they claim that copper itself

showed no activity for glycerol oxidation[39]. The presence of copper also prevented C-C bond

cleavage and led to improved yields of tartronic acid. Copper-chromium oxide spinel and CuO

have also been investigated as support for gold catalysts[40], and glycerol conversions of 20-46%

were obtained between 333 and 353 K in autoclaves using these catalysts. As good as these

catalysts are, stability and leaching remained a problem.

The catalyst support material is an important consideration in the electro-oxidation process.

Since only the catalytic material accessible to the reactants participate in the reaction, the degree

of dispersion of the metals on the support material is important. Conductive supports made from

polymers like polyaniline allow for better electron transport and thus better kinetics[31],[32].

Other attempts to increase catalyst efficiency address the inclusion of organo-catalysts like 2,2,6,6-

tetramethylpiperidin-1-oxyl (TEMPO) in conjunction with N-hydroxyphthalimide (NHPI); these are proton and electron transfer mediators in themselves[33]. Other supports include Vulcan [34], multi-walled carbon nanotubes[35], metal oxides like CeO2, SiO2, Mn3O4 and NiO [36], [37].

In a study of Ni promoted Pt and Pd catalysts for glycerol oxidation, Li et al (2014) found

that conversions of 40-99% could be obtained even after five consecutive uses, and concluded that

improved stability was due to the presence of Ni[41]. Copper and nickel are cheap, readily

available transition metals, and belong to the same group as the platinum group metals. Tungsten

and chromium are known to increase the strength and corrosion resistance of iron alloys.

Molybdenum being in the same group may give desirable mechanical properties – strength and

thermal stability, for example, - to deposits[42, 43]. Phosphorus is known to improve the activity

of catalysts by increasing their acidity[44], in trace amounts even though high phosphorus content

12 is an indication of non-crystallinity. Hence, the inclusion of Mo and P in Cu/Ni based catalysts can improve the range of conditions under which they can be used.

1.3.1 Catalyst fabrication methods

In general, catalyst fabrication methods have tended to lower precious metal catalyst loading through more effectively distributed material on the surface of a support. Fabrication methods can generally be grouped under[29]:

1. Thin film methods: The catalyst is either directly coated onto a membrane, or printed on a

poly- tetra-fluoro-ethylene (PTFE) blank and then transferred onto the membrane.

Unfortunately, it is difficult to achieve uniformity in the distribution of the catalyst using

this fabrication method.

2. Sputtering methods: A thin catalyst film is deposited on the surface of the support. This

fabrication technique has been perfected by the glass industry

3. Electro-deposition: Thin films can be achieved by the immersion of a substrate in a

commercial plating bath through which electricity is pulsed.

4. Impregnation methods

5. Co-precipitation methods

6. Sol gel methods

7. Electroless deposition: This will be discussed in detail in the section that follows.

1.4 The Technique of Electroless Deposition

Electroless deposition is an example of the formation of a metallic coating on the surface of a substrate by the chemical reduction of the metal ions from an aqueous solution in which the

13 substrate is immersed. It is a distinct phenomenon from conventional electro-deposition in that there is no input of electricity in the entire reduction process. Wurtz (as quoted by Muller et al[45]) first described this phenomena in 1844; he reacted copper sulfate with hypo-phosphorus acid and obtained deposits which he proposed were hydrides of copper (a result confirmed by Muller et al through x-ray diffraction studies). Extreme caution was taken to avoid contact of the coatings with air in these systems as the hydrides react very readily with air, and thus compromise the analysis of the original electroless deposits.

Not much was done with this discovery until about a century later when Brenner and

Riddell[46], who were working with nickel plating baths discovered that autocatalytic reductions on certain surfaces explained several observations. They coined the term, electroless plating, for this phenomenon. In the earliest baths, ammonium chloride was used to maintain the pH in the region (8-10) with nickel chloride as the metal ion precursor. Much pioneering work done in the

1940s and 50s in electroless deposition laid the groundwork for the science. Working first with electroless Ni, it was found that [46]:

1. The use of ammonium chloride in electroless baths posed a problem since ammonia

was easily liberated at the plating temp (90oC), giving off a disagreeable smell.

2. Complexing agents affect electroless nickel plating. Typical complexing agents were

acetate and citrate salts of sodium. Acetate baths were generally considered better than

citrate baths as plating was faster with the former. Unfortunately, acetate baths became

turbid with use and gave electroless nickel deposits that were dull and rough. Tartatric

acid was also found not to be good for nickel plating. This led to citrate baths being

better preferred over baths containing other complexants.

3. Very pronounced pH effects were evident. In general, Ni plated better in acid solution,

and the deposition of multi-metals was better in such solutions. In fact, co-depositions

of P and W (possibly Mo) with Ni and Co were possible, and the amount of the various

elements added in the electroless nickel was driven by market forces.

Over the next few decades, electroless deposition was researched extensively, and applications can be classified under three broad groups [47]: alloys, composites and metallic coatings. The deposition of metal alloys on the substrates for catalyst purposes falls under thin layer coatings.

1.4.1 Electroless deposition of Cu and Ni

Ternary and quaternary alloys containing Cu, Ni, P and other elements have been produced as alloys and composites, some dating from antiquity. Copper-zinc alloys (brasses) and copper-tin alloys (bronzes) are common examples. Copper and its alloys are ductile, malleable and easy to solder/weld using many different techniques. They possess electrical and antimicrobial properties.

It is one of the commonest metals in use, surpassed only by iron and aluminum[48]. In nickel- phosphorus alloys, tungsten, chromium or copper can be added to improve the electrical conductivity or lower the temperature coefficient of resistance[48]. These qualities are exploited in the fabrication of heat exchange equipment, electrical wiring, soldering material etc. The alloys described above are typically produced by mechanical alloying of the constituent metals, involving a sequence of ball milling and heat treatment steps[47].

Metallic coatings of Cu and Ni have been formed using electroless deposition (e.g.

NiCuSnP) with chloride baths [49]. Tin, molybdenum and tungsten salts are used in electroless

15 baths as stabilizers, and the mechanism of their co-deposition in electroless deposits is little understood, however, in larger concentrations these ions act as catalyst poisons.

A typical electroless plating bath will contain metal ion precursors, stabilizers, reducing agents, pH adjusters and chelating agents. The deposition of copper-rich Cu-Ni-Mo-P for catalyst application requires a reducing agent that can reduce both Ni and Cu. Typically, to plate Cu, a formaldehyde bath is used under alkaline conditions, while Ni is plated in acidic hypophosphite baths. To electrolessly co-deposit nickel and copper onto one substrate in catalytically relevant amounts presents challenges that are absent when a single metal is being plated: First, the bath must contain constituents that can adequately chelate both Cu and Ni, and thus, ensure the stability of the electroless bath. Second, as already mentioned the reducing agent must be capable of reducing both metals. Third, pH and temperatures should be optimized for a co-deposition process.

Successful electroless deposition of Cu-Ni-P has been carried out in acid hypophosphite baths [12,

18 – 20]. However, very low amounts of Cu are deposited with Ni in such acidic conditions (2-

4%)[50]. For copper-rich Cu/Ni deposits, a higher pH at which both metals can be co-deposited is necessary.

1.5 Economic Considerations for Glycerol Electro-Oxidation

The concept of a bio refinery, in which biomass sources are used to co-generate power, fuels and chemicals, has been examined by investigators interested in microbial fermentation[13].

Microbial fermentation plants tend to be expensive, a number of which are already in existence[13], as for example the process based on the use of a designer microbe to produce 1,3-

PDO (I,3 propane-diol) from corn has been running in a Bio-PDO™ plant in Tennessee.

One additional challenge that the industry may suffer from is the fluctuation in glycerol production which is contingent on seasonal variations of the biomass itself. Diversification of feedstock may have to be built into the original design of the biodiesel production process to ensure year-round supply of raw materials. Currently, process routes to the oxidation of glycerol are still at the research stage and none has been commercialized. This makes it difficult to compare the economics of the electro-oxidation routes, with more established technologies that have been implemented industrially. However, Katryniok et al[14], based on annual production rate of 10 tons of glycerol oxidation products per annum made an analysis and found that 55-86% of the entire raw material cost is based on catalyst. Cheaper catalyst alternatives would make these types of processes more attractive.

1.6 Research Goals and Organization of Text

From the foregoing, the electro-oxidation of glycerol to valuable chemicals in a fuel cell offers several advantages over other oxidation routes. We propose to use an alkaline fuel cell to achieve the dual objectives of high value chemical production and energy generation. Part of the identified economic and technical challenges associated with using biodiesel-derived glycerol in this regard is the high cost and high susceptibility to poisoning of state of art precious metal catalysts used for these processes. We hypothesize that:

Composite non-precious metal electro-catalysts can be synthesized with attendant

economic consequences for the co-generation of energy and fine chemicals in a direct

alkaline glycerol fuel cell.

The non-precious metals/materials chosen were copper, nickel, molybdenum and phosphorus, and these were fabricated using a novel mixture of formaldehyde and sodium hypophosphite reducing agents to overcome synthesis barriers arising from deposition behaviors of all catalyst constituents. Post fabrication, the behavior of these non-precious metal electro catalysts will be characterized to determine the effect of potential on product selectivity/ glycerol conversion, and energy generation capacity.

A continuing problem with electro-catalysts is the mechanical and chemical stability of the catalyst under process conditions, which correlate directly with the catalytic ability and overall cost [20]. These issues are yet to be satisfactorily addressed by research. These problems are compounded by the cost of the precious metal state of the art catalysts. The stability of the fabricated catalyst will also be studied.

Based on the foregoing, the major goals and objectives of this project are:

1. Synthesize non-precious metal catalyst composites through electroless deposition

a. Determine favorable conditions for the co-deposition of copper, nickel,

phosphorus and molybdenum

b. Study the effect of different deposition times of physical and electrochemical

characteristics of catalysts

c. Study speciation in mixed reducing agent electroless bath

2. Analyze the electrochemical performance of CuNiMoP/C, with a view to its use in an

alkaline glycerol fuel cell

3. Use synthesized catalysts in glycerol electro-oxidation, determine products formed,

and establish mechanism of reaction through monitoring time dependent product

evolution.

In Chapter 2, we describe the development and fabrication of copper, nickel, molybdenum and phosphorus catalyst composites using formaldehyde and sodium hypophosphite mixed reducing agents. The techniques used for the physical and electro-chemical characterizations of the catalysts are described. The analytical tools used to study chemicals generated will also be described.

In Chapter 3, we discuss the results of catalyst fabrication studies, and their implications towards glycerol electro-oxidation. Various physical characterizations using scanning electron microcopy (SEM), energy dispersive spectroscopy (EDS) and x-ray diffraction (XRD) done on

CuNiMoP/C are discussed. Activity towards glycerol oxidation will also be shown.

In Chapter 4, we discuss electrochemical characterization of the catalysts. These results show the energy generation capacity of the prepared catalyst.

In Chapter 5, we discuss potential-product relationships, glycerol conversions and product yields. These results show the favored mechanism of reaction on CuNiMoP/C, and glycerol conversions at given applied potentials.

In Chapter 6, we discuss catalyst de-activation mechanisms, and the implications of these in the stability of the fabricated catalyst composites.

Finally, in Chapter 7, conclusions based on all studies done are drawn, and directions for future work are delineated.

CHAPTER 2

METHODS AND MATERIALS

2.1 Introduction

In this chapter, materials and methods used to achieve the objectives stated in section 1.6 are described. The global objective of this project is to develop a catalyst system for oxidation of glycerol in an alkaline glycerol fuel cell, capable of producing both high value chemicals and energy. CuNiMoP/C electrocatalysts were synthesized using a mixed reducing agent protocol and characterized using scanning electron microscopy (SEM), Energy Dispersive X-ray Spectroscopy

(EDS) and X-ray Diffraction (XRD), Cyclic voltammetry (CV), Linear Sweep Voltammetry

(LSV) and Chronoamperometry. Product formation and glycerol conversions were determined with a combination of voltammetry and high pressure liquid chromatography. A summary of the methodology is provided in Figure 7.

Figure 7: Summary of experiments and methods

2.2 Electroless Bath Formulation

An electroless plating bath contains a reducing agent (which is oxidized) and metal ions of the metal to be reduced. Different reducing agents are active for different metals. It is known that copper is reduced in formaldehyde baths at pH values greater than 12 while nickel is reduced in hypophosphite baths between pH values between 9 and 10. It was postulated that a mixture of reducing agents at the right pH could lead to the co-deposition of copper and nickel at fractions greater than is possible with the use of a single reducing agent. That pH was determined through a reducing agent study.

2.2.1 Reducing agent study

Formaldehyde solution (Sigma Aldritch ACS reagent 37 wt. % in H2O containing 10-15% methanol as stabilizer to prevent polymerization) was mixed with differing amounts of 1 M sodium hypophosphite monohydrate (Aldritch Chemicals) to make 10 ml of different electrolyte solutions.

Copper foil was cleaned in dilute HNO3 to eliminate any oxides on the surface of the metal.

It was used as the working electrode in a conventional three electrode cell (with the reference electrode as Ag/AgCl in saturated KCl, and the counter electrode as a platinum wire). Corrosion potentials of the copper electrode were measured in different mixtures of the reducing agents using a potentiostat (Solartron SI Electrochemical Interface). The process was repeated with nickel foil, and the corrosion potentials also noted. The two sets of results were compared to see if there was any reducing agent mixture in which both nickel and copper were sufficiently cathodic. This was ascertained to be the mixture of 6 ml 37 wt. % formaldehyde solution and 4 ml of 1 M hypophosphite solution. These results will be shown in Chapter 3.

2.2.2 Optimal deposition pH

With this reducing agent mixture, the optimal deposition pH was investigated. The pH was changed using either dilute sulfuric acid or sodium hydroxide within a pH range of 3 to 12. The corrosion potentials of strips of copper as before were monitored in these pH moderated environments. The same procedure was carried out for nickel. Graphs of the corrosion potentials as a function of pH were generated for these two metals. The intersection between these two graphs was the optimal pH for the co-deposition of copper and nickel.

2.2.3 Bath composition

An electroless bath was formulated based on a mixture of reducing agents [1].

Table 3: Chemicals used in the formulation of Cu-Ni-Mo-P electroless bath

Material (250 ml plating bath) Amount (g)

NiSO4.6H2O 3.36

CuSO4.5H2O 1

Na2MoO4 1

NaH2PO2.H2O 1.35 K-Na-tartrate 2.5 Gluconic acid potassium salt 4.75 Sodium citrate 2 Formaldehyde (ml) 10

Gluconic acid potassium salt, sodium citrate and sodium potassium tartrate were used as stabilizers and complexing agents. The amount of materials used are given in

Table 3. A second bath was made by keeping quantities of every other component constant, and

changing amounts of copper and nickel to 0.5 g and 1 g respectively.

2.3 Substrate Preparation

Electroless deposition is usually initiated by small amounts of a material that is used to activate the substrate surface. These act as primary nucleation points, and autocatalyzes further deposition of the required metal ions from an aqueous bath. In this case, palladium was the seed material chosen.

Palladium (Pd) ink was made by dissolving palladium acetate trimer (Aesar) in ammonium hydroxide. This solution was stirred into a mixture of ethanol and polyvinyl butyral (Solutia Inc.

Butvar B-98) for 5 hours to stabilize the ink formed. The Pd ink was then coated on a carbon cloth. The coated substrate was put in the oven set at 375oC, and the polymeric material burnt off over 24 hours. This process activates the carbon cloth by ensuring that Pd2+ is reduced to Pd0. For

preparation of catalyst powders, the ink was mixed with the substrate (Al2O3 or carbon powder).

All procedural steps, as done for carbon cloth were also carried out for powders.

2.4 Electrode Preparation

2.4.1 Direct electroless deposition on carbon cloth

Because carbon cloth could be clamped directly to the potentiostat, direct deposition was

done on the prepared substrate. These electrodes were used for the unstirred bath experiments.

Argon gas was bubbled through the electroless bath for 30 minutes to eliminate any

dissolved oxygen that may act as catalyst poisons during the deposition. The bath was heated to a

temperature of 80oC, and the pH raised to 10.5 ± 0.5 with the drop wise addition of 20 wt.% NaOH.

These conditions were maintained for the whole experiment. Once these conditions were achieved, the Pd catalyzed carbon cloth was submerged in the electroless bath. Evidence of electroless deposition was shown by effervescence as the deposition occurred. The bath was continually stirred at 100 revolutions per minute (rpm) during the deposition process. At the expiration of the stipulated electroless deposition time, the plated carbon cloth was removed from the bath and rinsed in de-ionized (DI) water at 5oC twice and then once with isopropyl alcohol. The carbon

cloth supported CuNiMoP was then dried in the oven at 60oC for 12 hours.

2.4.2 Electrode fabrication from carbon powder

Electrodes for rotating disc experiments were made from electroless deposits on powder.

These experiments were necessary to determine transport limitations associated with glycerol

oxidation on the synthesized catalysts. All processing steps as already described for the carbon

cloth were followed: pH and temperature maintained at 10-11 and 800C respectively, continuous

stirring during the entire deposition period; rinsing in DI water and isopropyl alcohol, and drying

in the oven at 600C. Thereafter, 5 mg of CuNiMoP supported on carbon was mixed with 0.5 ml

nafion and 2.5 ml isopropyl alcohol. The mixture was sonicated for 10 minutes to ensure uniform

mixing. One drop of the ink so formed was dropped on a polished glassy carbon rotating disk

(diameter 0.5 cm). The electrode was left standing for 24 hours to evaporate the solvents at room

temperature.

2.5 Deposit Characterizations

It is known that the structure and composition of a material affects properties and behavior. The microstructure of the catalyst was viewed at different magnifications, and the elemental composition determined by energy dispersive spectroscopy (EDS).

2.5.1 Morphology and deposit composition

The morphology of the CuNiMoP/C deposits was investigated with a Jeol-740 IF Field

Emission Scanning Microscope. The carbon cloths were attached to sample holders using double sided adhesive tapes. As the metal deposits were conductive, no further sample preparation was necessary. They were then moved into the sample chamber. A vacuum was created within the chamber, and operation of the SEM deferred till the attached vacuum gauge pressure was about

10-5 pounds per square inch gauge (psig). An adequate working distance was chosen, and electron

gun was primed for the strength of the electron beam to be 20 kV. A beam of electrons was shot

at the sample, and the camera focused. When the brightness had been manipulated satisfactorily,

images were then captured. The spent catalysts were also similarly treated.

Another scanning electron microscope, FEI Nova, was used to capture energy dispersive

spectroscopy (EDS) data. This gives information on the metals on the surface of the carbon cloth,

as the elements on the sample respond in specific ways to incident radiation. At rest, the atoms of

an element have electrons which occupy different energy levels. If energy, in the form of x-rays

or electrons beams, is incident on one of such electrons, it is forced to go to a higher energy level

creating an electron hole, which another electron at a higher level drops down to fill. The energy

dispersed by this second electron as it drops down is unique for each element, and is thus an

25 identifier for the element. Through the intensity of the ensuing emissions, the amount of the element in the sample can be determined.

2.6 Fabrication of Electrochemical Reactor

To simulate fuel cell conditions, an electrochemical reactor was fabricated out of Teflon.

Separation between the anode and cathode side compartments was achieved by an anion exchange membrane (AMI-7001S) that allows the passage of hydroxide ions. These components were held in place by a brace (see Figure 8). The reactor was also equipped with covers that had perforations for the electrodes on both sides, and two others for the bubbling of gases on the cathode side, and sample collection on the anode side.

Figure 8: Reactor used for electrochemical studies. Actual reactor (left) and schematic (right)

2.7 Cyclic Voltamograms

Cyclic voltammograms (CVs) were used to check for the presence of deposits on the surface substrate surface, and subsequently, their activity for glycerol electro-oxidation. These were done in quiescent electrolytes (unstirred experiments) or stirred conditions (rotating disk electrodes).

All CVs pertaining to catalyst activity were done in 1 M glycerol in 4 M NaOH. Prior to each experiment, argon gas was bubbled through this solution for 10 minutes.

2.7.1 Unstirred experiments

Carbon cloth supported CuNiMoP was used as the working electrode, while a Pt wire was used as the cathode. Prior to the electrochemical measurements, CuNiMoP/C electrodes were scanned 50 times at 150 mV/s in 1M NaOH to clean the surfaces. The reference electrode was

Ag/AgCl in saturated KCl. The electrodes were connected to a Solartron SI 1287 Electrochemical

Interface. Perturbations were sent to the surface of the anode at different scan rates (voltage per second) and the current response measured. The results showed the types of chemical activities that were taking place on the surface of the anode.

2.7.2 Stirred experiments

The same experimental conditions were used to run CVs in rotating disk experiments, the only difference being the imposition of stirring on the electrolyte. This was done to study transport effects on glycerol electro-oxidation using CuNiMoP. The electrode was used to run CVs at different rotation speeds ranging from 100 to 1600 revolutions per minute (rpm).

2.8 Constant Potential Oxidation

It was expected that product yield and glycerol conversion would vary depending on applied

potential. Potentiostatic oxidations were done between 0.3 V and 1.3 V. Argon gas was bubbled

through 25 ml of 1 M glycerol in 4 M glycerol for 10 minutes. This solution was put in the anode

side of the reactor fabricated in section 2.6. 30 ml of 4 M NaOH was put in the cathode side

compartment and oxygen flow into it maintained for the duration of the oxidation reaction.

CuNiMoP/C (working electrode) and Ag/AgCl in saturated KCl (reference electrode) were secured

in the anode side compartment. A platinum wire was used as the counter electrode, and was placed

in the cathode side compartment. Results were obtained as current-time plots over 24-hour

oxidation periods. Samples were taken periodically during the oxidation, and were analyzed via

high pressure liquid chromatography (HPLC).

2.9 Oxidation Product Analysis

2.9.1 Determination of optimal detection wavelength in HPLC

Samples of all aqueous standards (glycerol, mesoxalic acid, tartronic acid, DHA, lactic acid, formic acid and glyceraldehyde) were run through the UV-Vis detector (Shimadzu SPD-20A

Prominence UV-Vis Detector) at different wavelength from 160 to 800 nm. 190 nm was determined as the wavelength at which all chemicals were adequately detected (see Appendix A).

2.9.2 Product determination using high performance liquid chromatography

As the reactions were carried out in aqueous alkaline medium, high pressure liquid chromatography equipped with UV-vis detector was chosen as the way to determine glycerol

conversion and the evolution of any new products. This is because the analysis times are relatively

short, and appropriate calibration can give both quantification and qualitative information on

product yield and glycerol conversion.

To analyze the liquid products of oxidation, a Hi-plex column (7.7mm x 300mm, Agilent)

was used under ambient conditions. The mobile phase was 5mM H2SO4 and the flow rate was 0.7 ml/min (isocratic mode). The injection volume was set at 10 µL. The uv-vis detection was at 190 nm, as glycerol and various oxidation products had an adequate UV reflection at this wavelength

(see Appendix A).

CHAPTER 3

CATALYST SYNTHESIS AND CHARACTERIZATION

3.1 Introduction

Results obtained on the fabrication of electroless CuNiMoP, and the use of these catalysts in an electrochemical reactor to show activity towards glycerol electro-oxidation, are discussed in this chapter. In addition, challenges encountered during the catalyst fabrication and the ways in which they were addressed are reported. Investigations on the electrochemistry of glycerol oxidation on CuNiMoP conducted are shown.

3.2 Choice of Main Active Materials

Copper had shown activity for glycerol oxidation when it was added to precious metal

catalysts [1]. It was hypothesized that by itself, it could also oxidize glycerol. To explore this

hypothesis, strips of copper, gold and platinum foils were used to oxidize the same concentration

of glycerol in sodium hydroxide solution (see cyclic voltammograms (CVs) in Figure 9), and their

catalytic activity towards glycerol oxidation was compared.

As expected, the magnitudes of the currents for the gold electrode are an order of magnitude

higher than those obtained for the platinum electrode. The gold foil performed better in alkaline

medium compared to platinum, as higher current densities were associated with noted oxidation

peaks. There are also two distinct oxidation peaks for gold between -0.2V and 0.2V (one of the

forward scan, and the other on the reverse scan), compared to one for platinum within the same

range. Gold also has a fourth oxidation peak that is almost superimposed on the first peak at about 30

-0.5 V. Both precious metal catalysts however show an oxidation peak on the reverse scan at about

-0.5 V. This behavior was consistent for all scan rates used (10, 25, 50 and 100 mV/s).

10 mV/s 10 mV/s 10 mV/s -2 Copper foil Gold foil -3 Platinum foil 2.5x10 25 mV/s -2 25 mV/s 8.0x10 25 mV/s 2.0x10 50 mV/s 50 mV/s 50 mV/s -2 100 mV/s 100 mV/s 100 mV/s

2 2.0x10 6.0x10-3 1.5x10-2 -2

Amp/cm 1.5x10 4.0x10-3 -2 1.0x10-2 1.0x10

2.0x10-3 5.0x10-3 5.0x10-3 0.0x100 0.0x100

0 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.0x10 -0.6 -0.4 -0.2 0.0 0.2 0.4 -0.6 -0.4 -0.2 0.0 0.2 0.4 V vs V (Ag/AgCl)

Figure 9: CVs at different scan rates for 1M glycerol in 1M NaOH solution

The copper foil however exhibited a different response. In place of the multiple oxidation peaks seen with the precious metals, there is a single oxidation hump. The magnitudes of currents for the copper electrode were about 50% higher than those for gold, the better performer of the precious metals in alkaline medium. The copper oxidation peaks straddle the potential region in which the two primary gold peaks are seen (between -0.2V and 0.2V).

In addition, the first oxidation peak encountered with the gold electrode has an onset potential of -0.6 V; this is lacking in the copper electrode. The onset potential for the second gold peak at about -0.4 V coincides with the onset potential for the single peak for the copper electrode.

The double oxidation peaks encountered for gold and platinum are considered characteristic of alcohol oxidation where some incompletely oxidized material on the forward scan are further oxidized on the reverse scan. This is also lacking on the plain copper electrode. However, the

31 magnitude of currents on the single copper oxidation peak suggests that the rate of reaction is so fast on the copper electrode that the reaction proceeds to completion on the forward scan alone.

It is possible that in addition to alcohol oxidation, there is also copper oxidation to various oxides taking place on the surface of the copper electrode [2]. The range within which copper is oxidized is narrower, between -0.25 V to 0.1 V (i.e. about 0.3 V) compared to a range of 0.6 V for the copper electrode in mixture of sodium hydroxide and glycerol in this work. Thus, other oxidation processes must be occurring within the wider potential range. Unfortunately, copper oxidation also occurring within this range highlights a challenge in using copper catalysts: they are easily leached out. Thus, electro-catalysts fabricated from copper must address this leaching problem at the lower potentials. This can be done by the inclusion of adatoms that impart greater stability to the copper deposits.

The inclusion of other metals during alloy synthesis can enhance mechanical properties.

Chromium in steel makes stainless steel corrosion resistant and strong. A copper composite, instead of plain copper, may prove less resistant to leaching. Ni as a co-catalyst has already been investigated in PdNi composites [3]. This led to the investigation of Ni as an inclusion for an enhanced copper electro-catalyst. The synthesis method for nickel inclusions lead directly to the co-deposition of molybdenum and phosphorus.

3.3 Electroless Deposition of Copper and Nickel: Reducing Agent Study

Electroless deposition of metal ions takes place when the right potentials are reached. These potentials, normally written as standard reduction potentials are presented in Table 4 for the copper

– nickel-mixed reducing agent system in both acid and alkaline environments

Table 4: Reversible potentials for Cu (II), Ni (II), formaldehyde and hypophosphite ions

Reaction Potential (V vs SHE) Cu2+ + e- ↔ Cu+ + 0.158 Cu2+ + 2e- ↔ Cu + 0.340 Ni2+ + 2e- ↔ Ni + 0.230 + - HCOOH + 2H + 2e ↔ HCHO + H2O (pH=0) + 0.056 - - - HCOO + 2H2 + 2e ↔ HCHO + 3OH (pH=14) - 1.070 + - H3PO3 + 2H + 2e ↔ H3PO2 + H2O (pH=0) - 0.500 - - HP + 2 H2O+ 2e ↔ H2P + 3OH (pH=14) -1.650 O O For a reducing agent to be oxidized by a metal, it must be less cathodic than the metal. This condition ensures that the reducing agent is oxidized. Thus, both nickel and copper should be reduced from formaldehyde and sodium hypophosphite baths. In practice, formaldehyde does not reduce nickel. Ohno et al. (1985) investigated the ability of different metals to catalyze the oxidation of different reducing agents [5]. By fixing the current at 10-4 A/cm2, they measured potentials at which metals oxidized different reducing agents. The experimental conditions were:

1. NaH2PO2 conditions: (0.2 M NaH2PO2 + 0.2 M Na-citrate + 0.5 M H3BO3, pH 9 and

temperature 343K).

2. Formaldehyde conditions: (0.1 M HCHO + 0.175 M Na-EDTA, pH 12.5 and temperature

298 K).

Figure 10 is a summary of the results for formaldehyde and sodium hypophosphite oxidation by various metals. The smaller the magnitude of the difference between Er (the oxidation potential of the reducing agent) and the oxidation potential of the metal, the more effective the redox couple.

Thus, for formaldehyde (HCHO) in Figure 10, copper would oxidize formaldehyde much better

than either nickel or cobalt. They also found that in acidic hypophosphite baths, copper and silver

simply dissolved and they could not properly account for their oxidation capacities.

Figure 10: Oxidation of formaldehyde and sodium hypophosphite by different metals [5] Copper can be deposited from acidic hypophosphite baths [6].

This most likely arises from copper displacement of deposited nickel since copper is more cathodic than nickel, and would be preferentially discharged in the presence of copper and nickel ions.

Hence, there may be no distinct interaction between the copper and the hypophosphite reducing agent.

Table 5: Corrosion potentials of Cu and Ni in mixtures of reducing agents

Formaldehyde/ Corrosion potentials V vs V Ag/AgCl in saturated KCl

Hypophosphite (vol/vol) Ni Cu

2/8 -0.321 0.018

4/6 -0.289 0.017

6/4 -0.234 -0.022

8/2 -0.252 0.039

It has also been shown that copper is deposited in formaldehyde baths at pH greater than

12, while nickel is deposited from hypophosphite baths at pH 9-10 [7, 8]. To co-deposit both metals

from a single bath requires the right mixture of electrolytes, and a favorable pH environment. To

study this, mixtures of 37 wt% formaldehyde solution and 1 M NaH2PO2 were used as electrolytes for copper and nickel, and the corrosion potentials noted. Table 5 is a summary of the results of the study of the corrosion potentials and shows that an electrolyte containing 6 ml 37 wt% formaldehyde solution and 4 ml sodium hypophosphite solution gave corrosion potentials for both metals that were sufficiently cathodic.

With this mixture of reducing agents, the co-deposition behavior of copper and nickel with respect to pH was studied. Figure 11 shows plots of the corrosion potentials of Cu and Ni in a 6 ml formaldehyde solution and 4 ml hypophosphite solution mixture over a pH range of 4-11. When the pH is below 7, Ni is much more cathodic than Cu. Further, as the mixed reducing agent bath pH becomes more basic, the corrosion potential of copper changes more rapidly than that of nickel.

At pH greater than 10, there is a sharp drop in the corrosion potential of copper. The graphs coincided when pH was about 10.2 (see Figure 11).

This confirmed our initial hypothesis that both reducing agents are active within a potential window within which co-deposition of both metals is possible. Outside this window, only one reducing agent is active for one metal. This is observed for the hypophosphite bath which is active for the deposition of nickel when pH is less than 10 although the co-deposition of copper can only occur when the copper ion concentration is below a certain level in such a bath. This formed the basis for all subsequent experiments that were carried out at pH of 10.5 ± 0.5, using a mixture of reducing agents.

Figure 11: Open circuit behavior of Cu and Ni with respect to changes in pH

3.4 Effect of Time of Deposition on Deposit Characteristics

3.4.1 Deposit mass

Electroless deposit thickness is function of immersion time in an electroless solution. In electroless deposition of Ni-P on polished steel plates Assashi-Sorkhabi et al. (2004) found that coating thickness increased with deposition time and pH, over a 3.5 hour period [51]. However, the quantity of phosphorus in these deposits decreased with time. Similar results were obtained on electroless nickel deposits on polished magnesium alloy plates[52]. It was expected that electroless CuNiMoP deposition would show similar deposition characteristics.

Figure 12: Effect of deposition time on deposit mass obtained by weighing substrates pre- and post-deposition

Figure 12 shows a deviation from the behavior described above. Electroless deposition was linear

during the first 20 minutes of deposition. Beyond this time, deposition became non-linear with

respect to time, reaching a maximum between 30 and 40 minutes. Deposition amounts thereafter

fell off, and decreased slightly from 35 and 45 minutes. The results suggest a mechanism in which

the electroless deposition rate decreases as the residence time is extended in a mixed reducing

agent bath. The loss in activity could be attributed to surface poisoning from the by-products of

the electroless deposition reaction. Alternate explanation could be that since electroless deposition

is a surface phenomenon, all the ions coming to the surface of the carbon cloth are reduced until

there are no more ions in the immediate vicinity of the surface. From speciation analysis, dominant

copper and nickel species under deposition conditions are hydroxides (see Appendix B). Prolonged

exposure to these deposition conditions could lead to dissolution of already deposited material.

This results in a loss of active material from the substrate.

The results in Figure 12 highlight differences between deposition on polished metal

plates[51] and porous carbon cloth (this work). Noted differences arise because actual surface area

exposed to deposition is higher than geometric area in porous materials, compared to polished

metal surfaces. The results presented imply that the activity for glycerol electro-oxidation with the

prepared CuNiMoP should be highest for a sample with 30 minute deposition time compared

compared to 15- or 45-deposition times samples.

3.4.2 Morphology: SEM

Images of carbon cloth substrate were taken pre- and post electroless deposition to investigate

any changes in the morphology. This was expected to have a direct impact on the properties of the

supported catalyst. Figure 13 is a palette of SEM images showing the changes in morphology of

different electroless deposits from 0-, 15-, 30- to 45 minutes. Surface roughness/coverage

increased with increases in deposition time.

Deposit thickness was estimated by measuring changes in the cross-section of carbon

fibers. The annular thickness of the deposits was calculated from equation 3.1

(3.1) (� − �0) 2

Where and represent the diameter of a coated carbon fiber and bare carbon fiber respectively.

These two� quantities� were measured from SEM images. Figure 14 shows changes in the annular

thickness of carbon fibers with as deposition time increases. Dividing the annular distance by 2,

estimates of deposit thickness can be obtained. Thus, thicknesses corresponding to 15-, 30- and 45

minute samples are 0.133 ± 0.005 µm, 0.255 ± 0.009 µm and 0.202 ± 0.011 µm respectively.

Figure 13: CuNiMoP/C at 1000x magnification. A:0 min, B:15 min, C:30 min and D:45 min

For a given length of carbon fiber, h, mass of deposits on substrate can be obtained from

(3.2) 2 2 �(� − � ). ℎ. � ℎ. (2��)

Where is the composite density, R is the outer diameter of the annular ring formed by the deposit and r is� the inner radius of the annulus, which corresponds to the radius of the bare substrate.

Figure 14: Deposit thickness as a function of time

For unit length of fiber, equation 3.2 reduces to a relationship between composite geometry and density.

(3.3) 2 2 (� − � )� (2�)

Composite density was calculated by multiplying fractional contribution of each component by its density:

(3.4) � = �. � Where and refers to the mass fraction and density of jth component (Cu, Ni, Mo or P).

The porous� nature� of the carbon cloth substrate is seen at high magnification (x50, 000) in the different samples. From SEM images at high magnification, r was measured as 4.4 µm. Using equations 3.1 to 3.4, loading was then calculated (see Table 6)

Table 6: Loading data for CuNiMoP/C

Time Density Deposit mass per unit area Normalized with respect

(min) (kg/m3) g/cm2 to 15 min sample

SEM Weight/area

calculations

15 8574.6 0.000113 0.0008 1.00 1.00

30 8226.2 0.000204 0.0016 1.81 2.00

40 8271.3 0.000163 0.0011 1.45 1.38

Loading results from SEM are calculated for a single carbon fiber of unit length, while that from weight /area calculations were obtained by weighing a carbon cloth of known dimensions pre- and post-electroless deposition. Thus, the latter is bigger. However, when loadings are normalized with respect to the 15-minute sample, similar ratios are obtained for either set of experiments. This confirms non-linearity of deposition behavior.

Figure 15 shows increase in surface roughness as deposition time increases. At high magnification, the cross section of the carbon fiber is seen to be spongy. On the surface of each fiber, spherical CuNiMoP deposits are visible, and these were used to determine particle sizes.

Figure 15: Morphology of plain substrate, 15-, 30- and 45 minutes electroless deposits

3.4.2.1 Estimation of particle sizes

Estimates of the average particle size of the electroless deposits were obtained from high magnification SEM images (see Figure 16). Scales provided on SEM images were used to calculate diameters of the spherical electroless deposits. The average estimated particle diameter was thus 271.4 ± 43.2 nm. This analysis is extrapolated to give deposit thickness of 10-9 - 10-6 m.

Figure 16: SEM of 45-minute sample at 12000x and 50000x

Figure 17: XRD of different CuNiMoP samples 43

3.4.3 X-ray diffraction

The presence/crystallinity of copper and nickel on the carbon substrate were investigated using X-ray diffraction on CuNiMoP deposited on alumina. 2 theta values indicative of the presence of elemental copper and nickel were seen. In addition, oxides of copper and nickel were also present. The sharpness of peaks in the spectra indicate that crystalline materials were deposited (see Figure 17).

3.4.4 Activity towards Glycerol Electro-Oxidation

The activity of CuNiMoP as a function of deposition time was also studied (see Figure 18).

The plain carbon cloth showed no activity towards glycerol oxidation. However, the palladium catalyzed carbon cloth showed some activity for glycerol electro-oxidation, evidenced by a peak of 0.1 mA at -0.05 V. This is expected as palladium is a known electro-catalyst for glycerol oxidation [14]. However, in the present approach used for the fabrication of CuNiMoP catalyst, the purpose of palladium on the substrate is to act as a primary initiation (nucleation) point for the electroless deposits, and so any exposed palladium would be covered by CuNiMoP. Any oxidations seen on the 15, 30 and 45 minute samples are due to the deposited CuNiMoP, not palladium, as seen from the magnitude of current on the activated substrate. Although all three samples showed activity for glycerol oxidation, the maximum current density (4.3 mA/cm2) was obtained on the 30-minute sample.

Since the same geometric areas of woven carbon cloth and volumes of electroless bath were used for all samples, any differences in the cost of fabrication arise from any time dependent materials/utilities used. Relative to the cost of the 15-minute sample, processing costs doubled for the 30-minute sample, and tripled for the 45 minute sample. It is expected that similar trends would be seen in a scaled-up plant (see Table 7).

Table 7: Cost of fabricating different catalyst samples. Electricity and labor cost based on graduate student residence and remuneration.

Item 15 minute 30 minute 45 minute

Electricity ($ 0.12 /kWh) 0.03 0.06 0.09

Labor ($20/h) 5.00 10.00 15.00

pH adjusters ($0.001/minute) 0.02 0.03 0.05

Total ($) 5.05 10.09 15.14

The specific current is a measure of how effectively the quantity of catalytic material

deposited is used. A high specific current means that small amounts of catalytic material give high

rates of reactions. A comparison of the specific currents (Table 8) shows that the 15-minute sample

(4.84 A/g) utilizes the deposited material better than the 30- (2.65 A/g) and 45-minute (1.40 A/g)

samples. The 0.6 mA/cm2 advantage gained by using the 30-minute sample does not justify the extra labor and material cost.

Table 8: Performance of 15, 30, 45 minute samples in glycerol oxidation

Sample ID 15 min 30 min 45 min

Deposition per unit area (g/cm2) 0.0008 0.0016 0.0011

Peak current (A/cm2) 0.0037 0.0043 0.0015

Specific current (A/g) 4.84 2.65 1.40

3.5 Effect of Cu2+ and Dual Reducing Agents in the Electroless Bath

We also investigated whether similar amounts of copper, nickel, molybdenum and phosphorus could be obtained using a single reducing agent. If this could be done, it would reduce the complexity of electroless deposition and materials cost. In addition to the mixed reducing agent baths, single reducing agent baths were formulated, as shown in Table 9.

Table 9: Bath constituents (50 ml plating solution)

Formaldehyde Sodium Hypophosphite Hypophosphite and

Material alone alone Formaldehyde

NiSO4.6H2O(g) 0.672 0.672 0.672

CuSO4.5H2O (g) 0.025 – 0.541 0.025 – 0.101 0.025 - 0.671

Na2MoO4.H2O(g) 1 1 1

NaH2PO2.H2O(g) - 0.5 0.5

K-Na-tartrate (g) 0.5 0.5 0.5

Gluconic acid potassium 0.95 0.95 0.95 salt (g)

Formaldehyde (ml) 2 - 2

Amounts of material deposited on the supports were analyzed by energy dispersive X-rays

(EDX). The pure hypophosphite bath had every constituent except the formaldehyde, while the

pure formaldehyde bath did not have any hypophosphite. The amounts of copper sulfate in the

electroless baths were varied while the amounts of every other component were kept constant (see

Figure 19). For the mixed reducing agent system, as the amount of copper was increased, the amount of material deposited on the support increased. The bath remained stable and productive, even at relatively high copper contents (Cu/Ni ratio of 1). For the pure hypophosphite bath, the amount of deposited material increased until about Cu/Ni ratio of about 0.12, from where it gradually decreased before the bath abruptly stopped depositing material at a ratio of 0.17. It appeared that at this ratio, copper effectively became a poison to the deposition process. For the pure formaldehyde bath, the bath abruptly started precipitating copper at high copper to nickel ratios.

0.8

Formaldehyde + 0.7 Sodium hypophosphite

0.6

Sodium hypophosphite 0.5

Formaldehyde 0.4

0.3 CuNiMoP fraction on substrate surface 0.2

0.1

0.0 0.0 0.2 0.4 0.6 0.8

2+ 2+ Cu /Ni ratio in plating bath

Figure 19: Amounts of material deposited on support for formaldehyde, hypophosphite and mixed reducing agent baths as function of Cu/Ni ratio

The relative amounts of copper, nickel, molybdenum and phosphorus deposited for each of the reducing agent(s) varied (see Figure 20). While the mixed reducing agent bath deposits

48 copper, nickel molybdenum and phosphorus for Cu/Ni ratios from 0-1, the formaldehyde alone bath deposits copper almost exclusively. The hypophosphite alone bath was not biased towards copper as it yielded more Ni in the deposits until Cu/Ni of 0.1 when more copper was contained in the deposits. Thus, to achieve a Cu/Ni ratio of 8 in the electroless deposits, a ratio of Cu/Ni in the electroless bath should be about 0.2 - 0.5. The speciation calculations for all the species involved in the electroless deposition is attached in Appendix B.

Figure 20: Relative amounts of copper, nickel, molybdenum and phosphorus in deposits from various baths. Mixed reducing agent (left), pure hypophosphite (middle) and pure formaldehyde (right).

CHAPTER 4

ELECTROCHEMICAL PERFORMANCE OF CUNIMOP

4.1 Introduction

In the combustion of hydrocarbons for power generation purposes, part of the chemical energy contained in the fuel is lost as waste heat, and so the efficiency of these systems (30-40%) which is limited by Carnot considerations, is lower than can be obtained with hydrogen-oxygen fuel cells, for example (60-80%). Glycerol has a high specific energy content (18.1 kJ/g [9]), and has been investigated as a fuel for energy generation purposes. Fuel cells can be used to recover this chemical energy without an intermediate combustion step. These fuel cell systems are equipped with catalysts on the anode side to facilitate the oxidation reactions that release energy.

In this chapter, the performance of CuNiMoP/C as the anode side catalyst in a divided electrochemical cell is investigated using voltammetry.

4.2 Combined Effect of Copper, Nickel, Molybdenum and Phosphorus

Previously in Chapter 3 while characterizing CuNiMoP, it was shown that appreciable oxidation rates could be obtained. However, this composite as prepared, consists of four separate materials. Fewer components in the catalyst composite could significantly reduce both production time and cost, especially if comparable activity could be achieved. Cyclic voltammetry (CV) experiments were conducted with electroless Cu/C and NiMoP/C (see Figure 21). Electroless Cu/C showed some activity for glycerol electro-oxidation, but the maximum peak current obtained on it is eight times lower than that obtained on electroless CuNiMoP/C. NiMoP/C on the other hand

50 showed negligible activity towards glycerol oxidation. Unfortunately, both molybdenum and phosphorus are co-deposited with copper and nickel, and so could not be studied in isolation. From these results, neither electroless copper nor nickel can give similar current densities to that obtained on electroless CuNiMoP. Thus, the noted activity towards glycerol electro-oxidation is due to the presence of all four elements in the catalyst composite.

10 CuNiMoP/C

8 ) 2 6

Cu/C I (mA/cm 4 NiMoP/C

-2 -0.4 0.0 0.4 0.8 E (Volts)

It has been reported that nickel, and nickel modified with copper and cobalt, are active for glycerol electro-oxidation[53, 54]. Such activity was not seen with NiMoP/C fabricated electrolessly in this work. However, CuNiMoP/C was at least six times as active as such nickel

51 composites towards glycerol oxidation. Other reasons for the higher activity of these nickel catalyst may arise from catalyst fabrication methods. Thus, catalyst micro-structure has an important effect on its behavior as an electro catalyst.

4.3 Reactions at CuNiMoP Anode

The question then arises: What are the electrochemical events occurring at the CuNiMoP anode? These can be broken into the following steps:

1. Transport of glycerol and hydroxyl ions from the bulk electrolyte to the CuNiMoP surface

2. Adsorption of reactant molecules to active sites on the catalyst

3. Reactions and bond re-arrangements on the catalyst surface

4. Desorption of products

5. Transport of products back to the bulk of the electrolyte.

Steps 2, 3 and 4 are surface phenomena involving a series of reactant-electron interactions.

Cyclic voltammetry can be used to estimate the number of electrons involved in such reactions.

In terms of species, different electro-chemical reactions are possible at the CuNiMoP/C anode. First the individual components in the catalyst could be oxidized. However, oxidation potentials greater than 0.6 V are unlikely to lead to dissolution of copper and/or nickel components of the CuNiMoP catalyst composites. Second, oxygen could be evolved at the anode. Third, glycerol and its derivatives could be oxidized. We determine the most likely events occurring from the electrochemistry of the species at the anode (data from IUPAC at 298 K), assuming that the catalyst support is not oxidized.

For ease of comparison, all potentials have been written with species as being reduced (see

Table 10). Considering the three carbon oxidation products of glycerol and the major components in the catalyst (copper and nickel), the most electro-negative species are the organic species.

Oxygen evolution in alkaline media, which is possible, would still occur at higher potentials than the glycerol oxidation. Thus, even though activities of the various ions in solution deviate from conditions under which thermodynamic reversible potentials are calculated, these considerations show that glycerol electro-oxidation is the most likely electrochemical reactions occurring at applied anodic potentials greater than 0.7 V vs SHE (0.9 V vs V Ag/AgCl).

Figure 22: Applied potential at which glycerol oxidation should be done

Possible oxygen evolution could be occurring at the anode. This is the reverse of oxygen reduction reaction occurring at the cathode. If it occurs, gas bubbles would be formed at the electrode. A penalty for such a parasitic reaction would be a reduction in current efficiencies associated with desired glycerol oxidation. At the higher potentials, oxygen evolution could also arise because water decomposition potential has been approach. Thus, a good operating voltage should lead to desired product formation while not crossing the thresholds for other competing reactions. In addition, low rates of reaction will be obtained at low voltages. These conditions mean that applied potentials should fall between 0.4 V and 1.23 V vs SHE (Figure 22)

Table 10: Reversible potentials of species at the anode

Reaction Reduction potential, V vs SHE

Cu2+ + 2e-→Cu 0.34

Cu+ + e-→Cu 0.52

Cu2+ + e-→ Cu+ 0.16

Ni2+ + 2e-→Ni -0.23

- - 4OH → O2+4H2O +2e -0.40

Glyceraldehyde -0.419

C3H6O3 + 2H2O + 2e- → C3H8O3 + 2OH-

Glyceric acid -0.739

C3H6O4 + 3H2O + 4e- → C3H8O3 + 4OH-

Hydroxypyruvic acid -0.768

C3H4O5 + 5H2O+ 6e-→ C3H8O3 + 6OH-

Tartronic acid -0.719

C3H4O5 +6H2O+ 8e- → C3H8O3 + 8OH-

Mesoxalic acid -0.716

C3H2O5 + 8H2O+ + 10e- → C3H8O3 + 10OH-

4.4 Performance of CuNiMoP/C under Unstirred Cell Conditions

Having determined that the most likely event at the anode is the oxidation of the organic species, we then analyze the electrochemistry of these reactions using cyclic voltammetry. This is done by generating a series of CVs in the same electrolyte at different scan rates with the same

electrode. The oxidation peaks obtained at these different scan rates is a function of the number of

electrons involved in the anode side reaction and is described by the Randle-Sevcik equation. The

Randle Sevcik equation relates the peak currents from cyclic voltammograms to the square root of

scan rate of the voltage that an electrode is perturbed with. At a temperature of 25 oC, this can be

expressed as:

(4.1) i = 268,600 n AD Cv

ip = peak current from CVs, n = number of electrons transferred in the redox reaction; A =

electrode area in cm2; D = diffusion coefficient in cm2/s; C = concentration in mol/cm3; ν

= scan rate in V/s

For known diffusivities, electrode area and concentration of reactants, the number of electrons involved in a redox reaction was estimated. The estimates of number of electrons are dependent on the accuracy of diffusion coefficient, area and concentration of the diffusing species.

3.0

10 mV/s 2.5 25 mV/s 50 mV/s 100 mV/s 2.0 125 mV/s ) 2

1.5 i (mA/cm 1.0

0.5

0.0

-0.5 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 E vs V Ag/AgCl

Figure 23: CVs at different scan rates in 1 M glycerol + 4 M NaOH

Figure 23 shows the variation of peak currents with scan rates. For a reaction to be reversible, any oxidations encountered on the forward scan must be reversed on the backward scan.

However, for the system under consideration, the oxidation encountered on the forward scan is followed by a second oxidation hump on the reverse scan. Hence, irreversible reactions are taking place at the anode.

Peak currents for the forward scans were plotted against the square root of the scan rate as shown in Figure 24. It is expected that if the electrode reaction is due to the formation of a distinct species, a straight line will be obtained.

1/2 Figure 24: ip vs. ν for the forward reaction

Similarly, formation of distinct species at the anode will also give a linear relationship between peak potentials and natural logs of the square root of applied scan rates (see Figure 25).

Figure 25: Ep vs ln (ν)1/2 for the forward scan

From the slopes of the Randle-Sevcik equation, it was determined that a 2-electron process is occurring during the forward scan (most likely oxidation to glyceraldehyde) and a 14-electron process (complete oxidation to carbon dioxide) is occurring on the reverse scan. A single electrode

1/2 reaction is taking place on the electrode surface on the backward scan. From the ip vs. ν plot, a slope of 0.137 is obtained (Figure 26). Applying a diffusion coefficient of 9.3 x 10-6cm2/s

(diffusion coefficient of glycerol in water solution1) to the Randle-Sevcik equation gives the

1/2 number of electrons involved as 13.8. This is supported by the Ep vs ln (ν) plot (Figure 27),

where a constant potential of 0.348 ± 0.004 V was obtained on the backward scan. This peak

symmetry around a single potential indicates that the same reaction is occurring on the electrode

surface at that potential. This is close to the number of electrons involved in the complete oxidation

of glycerol to carbon dioxide (14). So, one of the reactions on the forward scan produces an

intermediate which is further oxidized to carbon dioxide, on the backward scan.

1 http://bionumbers.hms.harvard.edu 57

1/2 Figure 26: ip vs. ν for the backward reaction

Figure 27: Ep vs ln (ν)1/2 for the backward scan

4.5 Performance of CuNiMoP/C under Stirred Conditions

In a quiescent electrolyte, it is assumed that the rate of reaction at the catalyst surface is limited by transport of reactants to the reaction zone. To properly account for such transport effects, rotating disk experiments that impose known convection rates on the electrolyte are used.

Conclusions on the relationships between the stirring rates and the limiting currents achieved with the catalysts can be drawn from empirical relationships like the Koutecky-Levich equation:

(4.2) 1 1 = w i where is the limiting current0 on.620nFA the anode,D v F isC the Faraday’s constant, A is the electrode

area, D is thei diffusion coefficient of the reduced species, v is the kinematic viscosity of the solution, C is the concentration (1M) and w is the angular rotation rate.

Figure 28: Rotating disc experiment at different speeds. 1 M glycerol in 4 M NaOH

Results on rotating disk experiments done on CuNiMoP in 1 M glycerol + 4 M NaOH at different speeds are shown in Figure 28. Higher rates of reaction were obtained at higher scan 59 rates. Figure 29 shows Koutecky-Levich plots obtained from the CVs in Figure 28. The number of electrons are calculated at each oxidation peak (-0.20 V, 0.16 V and -0.24 V). Each of these plots has an intercept on the y-axis, indicating that there is a kinetic component to the current sensed at the electrode. In addition, a 2 electron process takes place at -0.2 V, followed by two separate single electron processes at 0.16 V and -0.24 V. These later single electron processes are likely to be the initiation and conclusion of a two electron process.

Figure 29: Koutecky-evich plots from oxidation peaks on forward scan (-0.2 V and 0.16 V) and backward scan (-0.24V)

A difference exists between the stirred and unstirred electrolyte conditions. While complete

oxidation to carbon dioxide is seen in the unstirred bath, fresh electrolyte is presented at the catalyst

surface in the stirred condition and this discourages carbon dioxide formation.

4.6 Determination of Kinetic Parameters for CuNiMoP/C

Modeling of glycerol oxidation kinetics is complicated by the large number of products formed, high viscosity of glycerol and number of electrons transferred in electro-oxidation. Han et al (2014) modelled glycerol electro-oxidation on gold catalysts by assuming that only tartronic acid, the 8-electron product, was formed under alkaline conditions [55]. Posing the problem as a

Fickian type diffusion in one dimension coupled to Butler-Volmer equations, they obtained numerical solutions from which they could predict kinetic parameters, like exchange current densities. These kinetic parameters are important because they provide an avenue to compare the performance of different electro catalysts. For CuNiMoP, these were estimated from analyses of results from voltammetry.

Current density at the CuNiMoP anode can be expressed by the Butler-Volmer equation as:

�� (4.3) �� = �0,��( ��) ��

where is the anodic current density, is the transfer coefficient, F is Faraday’s constant,

� R is molar gas constant,� T is temperature and� is the applied over potential at the anode and

, is the exchange current density. � �

The exchange current density from equation 4.3 is the current when there is zero over

potential on the anode. Thermodynamically, it is the current when the rate of forward reactions on

the anode is exactly balanced by the rate of the backward reaction. It is a measure of electron

transfer rates when there is no imposed over-voltage. It can be estimated from Tafel plots by

choosing a very narrow portion on a cyclic voltamogram where Butler Volmer kinetics are obeyed.

Within this region, current varies linearly with potential:

(4.4)

� = � + � log �

Where the quantities a and b are obtained from the slope and intercept of from equation

4.4. The exchange current density is estimated from extrapolating the graph to zero over potential.

Literature values for exchange current density for alkaline glycerol oxidation is scarce. The only

instance it was directly quoted was as 4 x 10-4 A/cm2, where it was fitted from numerical calculations[55]. For simpler alcohols like methanol and ethanol, more data is available. Despite this, exchange current densities calculated for even the same catalysts vary significantly based on synthesis method. Again, where the Tafel region exactly lies is left to the discretion of the investigator. Barring differences that arise from catalyst synthesis methods, different results can be obtained on the same metal catalysts. For instance, both working with Pd catalysts in 1 M ethanol and 1 M NaOH, Shen et al (2006) calculated an exchange current density of 2.7 x 10-5

A/cm2 [56], two orders of magnitude higher than the 1.24 x 10-7 A/cm2, obtained by Zhang et al

(2011) [57]. Despite these obvious discrepancies, the important thing to note is that very low exchange current densities imply poor catalytic ability.

The kinetic parameters for the CuNiMoP/C electro-catalysts were determined from Tafel

plots. The potential versus current data for this analysis was taken within a 100 mV region in which

Tafel equation was obeyed on the highest oxidation peaks obtained at a given scan rate. Log (i)

versus over-potential was plotted for the CuNiMoP/C catalysts. The exchange current density

obtained this analysis was 0.18 A/m2.

Table 11: Electrochemical parameters for glycerol oxidation with different catalysts. Oxidations are in alkaline environment

Catalyst Loading Scan rate I0 Tafel slope

(µg) mV/s (A/m2) i i i 55 wt% Au/C [55] - 25 4* 12.6 n/a n/a

CuNiMoP/C 750 10 0.18 12.8 14.27 130

Pt/CCE[58] 500 5 46.8 2.44 166

Pd/CCE[58] 500 5 51.8 19.90 136

Au/CCE [58] 500 5 58.0 0.678 177

Pt/C[37] 300 50 21 n/a n/a

Pd/C[37] 300 50 18 n/a n/a

Pt3Ni1/C [37] 100 50 0.18 0.24 1.75 166

Pt2Ni1/ C [37] 100 50 0.20 0.27 1.84 160

Pt1Ni1/C[37] 100 50 0.15 0.26 1.83 162

is peak current on the forward scan, is the peak current on the reverse scan, ή is the

overpotentiali (V vs V Ag/AgCl) and Vonset iis the onset potential for the anode reactions. Han et al

(2014) obtained exchange current density from fitted data[55].

Table 11 compares the electrochemical parameters of precious metal catalysts with those of CuNiMoP/C in alkaline environment. Comparable onset potentials and Tafel slopes are obtained on all the catalysts. This is because the scan rate used was low. Higher current densities are obtainable on pristine CuNiMoP, as for example in Figure 21 where 8 mA/cm2 was obtained.

The ratio if/ib is used to estimate the susceptibility of a catalyst towards poisoning[59]. The lower this ratio is, the less likely the catalyst is to be poisoned. At 0.678, Au/CCE is less likely to be poisoned than Pd/CCE (19.90) [58], and this is a major reason why it is the state of art catalyst for glycerol oxidation in alkaline environment. This ratio for CuNiMoP (2.40) is comparable to that of FeCo@Fe@Pd/C (2.33) [60] and Pt/CCE (2.44) [58]. Comparable Tafel slopes with literature was obtained on CuNiMoP/C.

4.7 Chronoamperometry of Glycerol Oxidation on CuNiMoP

Applied potential affect both rates of reaction and products obtained during glycerol oxidation [61]. These rates of reactions, measured as current density, would show which potentials are adequate to drive desired oxidations. Figure 30 shows potentiostatic oxidations in 1 M glycerol

+ 4 M NaOH at 0.7 V, 0.9 V, 1.1 V and 1.3 V (0.3 V and 0.5 V are inset). Rates of reaction at 0.3

V and 0.5 V decay to less than 1 mA/cm2 after 15 h, and imply that these conditions are not economically viable. The higher the deviation from equilibrium or decomposition potential, the greater the driving force sustaining the reaction. Much better sustained rates of reaction are obtained above 0.9 V, as obtained from thermodynamic considerations for three carbon oxidation products.

Figure 30: Current profile during constant potential oxidations

Current densities are important as they show what rates of reactions are obtainable on a catalyst. Values lower than 10 mA/cm2 will not support profitable commercialization. Thus, 0.9

V or higher is necessary to sustain commercial production. Based on these curves, energy input into the electrochemical system can be found by integrating the area under current versus time plots, and multiplying by the applied potential (Figure 30 and Figure 31).

3 8x10 0.3 V 0.5 V ) 2 0.7 V 3 0.9 V 6x10 1.1 V 1.3 V

3 4x10 Energy density (kJ/cm

3 2x10

0 0x10 0 4 4 4 4 0x10 2x10 4x10 6x10 8x10 Time (s)

Figure 31: Energy density at different potentials

CHAPTER 5

GLYCEROL CONVERSION AND OXIDATION PRODUCT YIELD DURING POTENTIOSTATIC OXIDATION OF GLYCEROL ON CUNIMOP

5.1 Introduction

It has been established that CuNiMoP is active for glycerol oxidation in alkaline medium.

In this chapter, we discuss results of 24 hour oxidations at constant potentials with emphasis on glycerol conversion, product selectivity and hence mechanism of reaction.

5.2 Thermodynamic Considerations for 3-Carbon Oxidation Products

Voltage thresholds, representing minimum energy requirements, should be reached before glycerol electro-oxidation can occur. These cell voltages, which are the thermodynamic equilibrium values, E0, were obtained from predictions based on group additivity values [26].

These compare well with literature values (Table 12) [12].

The importance of the predicted reversible potentials is that a species can, in theory, be obtained by fixing the applied electro-oxidation potential slightly more anodic (higher than) to the reversible thermodynamic value necessary for its formation. For instance, the reversible cell voltage to produce glyceraldehyde is 0.802 V, giving an anode half-cell potential of 0.402 V vs

SHE. An applied potential of a value that is at least slightly greater than 0.402 V (0.602 V vs

Ag/AgCl) will be required to produce glyceraldehyde. This also implies that to skew selectivity towards tartronic and mesoxalic acids, the potential applied at the anode should be in excess of 0.6

V vs SHE. At applied anode potential more than 0.6 V, other possible species with formation or

decomposition reversible potentials below 0.6 V will compete with tartronic and mesoxalic acids

production. Hence, product distribution will not only depend on the nature of the electrode or

electro catalyst (CuNiMoP) but will be a function of the applied anodic potential.

Table 12: Predictions of thermodynamic properties of 3-carbon oxidation products

Product n ΔH ΔS ΔG V vs (kJ/mol) (J/mol.K) (kJ/mol) EV vs SHE ESHE [62]

Glyceraldehyde 2 -437.79 -0.95 -154.839 0.802 n.r.

Glyceric acid 4 -730.38 -1.07 -410.164 1.063 1.140

Tartronic acid 8 -1162.43 -0.90 -893.306 1.157 1.170

Mesoxalic acid 10 -1283.48 -0.69 -1076.91 1.116 1.117

5.3 Controlled Potential Electro-Oxidation of Glycerol

To study the effect of applied potential on product formation and glycerol conversion, controlled potential oxidations were carried out at different voltages. Figure 32 shows typical of current / time behavior observed at a constant applied oxidation or anodic potential. As the reaction proceeds, the rate of reaction decays exponentially and arises from reduced amount of glycerol in the reaction mixture and/or catalyst fouling. The progress of glycerol conversion and product distribution was monitored via periodic sampling of the contents of the reactor.

Figure 32: Constant potential oxidation at 0.9 V. Conditions: 25 0C, 1 M glycerol + 4 M NaOH. Atmospheric pressure.

5.3.1 Effect of applied potential on glycerol conversion

The anodic over potential for glycerol oxidation process can be described by Tafel equation i.e.

(5.1) � = � + � log �

where  is the overpotential, a is intercept, b is the Tafel slope, and i the current. With

Faraday’s law relating the current in eqn. [1] to the rate of conversion of glycerol i.e:

� � (5.2) � �� = = � ��

Substituting for i in eqn. [5.2] using eqn. [5.1], the connection between the applied potential

and the rate of reaction or conversion of glycerol is apparent. At any given applied potential,

multiple competing reactions may take place forming different products and each reaction is

described by the Tafel equation with its own partial current density but same over potential as

other reactions on the electrode.

Results from the electro-oxidation of glycerol using electroless CuNiMoP/C show that

conversion of glycerol is a function of applied potential at the anode. This is in agreement with

eqns. [1] and [2] combined i.e.

(5.3) () 10 Ä� = Ä� ��

where c is the concentration change and t is the time of oxidation. At the thermodynamic ideal, the overpotential is zero. Thus the concentration and time are functions of that are dependent on calculated Tafel parameters for the system.

(5.4) () 10 Δ� = Δ� ��

Results of glycerol conversion after 24 hours at 0.5 V, 0.7 V, 0.9 V, 1.1 V and 1.3 V vs

Ag/AgCl are shown in Figure 33.

Figure 33: Glycerol conversion as a function of applied potential

From eqn. [5.3], it was expected that conversion versus potential would describe a sigmoidal behavior, with conversion gradually becoming independent of applied potential at high voltages i.e.

 (5.5) nFΔ� � � = � + � log = � + � log(nF) + � log Δ� Δ�

Such behavior was upheld until 1.1 V, after which there was a sharp decrease at 1.3 V.

Oxidations at these two potentials were repeated to ensure reproducibility. The catalysts used were

made in two different batches but following the same production protocol. Conversions of 63.7%

71 and 61.8% were obtained at 1.1 V. Similarly, deviations from a mean conversion of 42% was 3.9% conversion at 1.3 V.

This result can be explained by several phenomena occurring in the electrolytic cell.

Sodium mesoxalate had limited solubility in alkaline media during HPLC standard calibrations.

It crystallizes as a deliquescent solid; it is quickly formed at 1.3 V in quantities that enable

crystallization. This behavior thus occludes the catalyst surface, leading to a drop in total

glycerol converted at 1.3 V. This would happen if the mesoxalate is not easily desorbed after

formation, effectively becoming a catalyst poison.

Figure 34 is an SEM image of spent catalyst at 1.3 V. It is possible that this behavior can be reduced in an actual fuel cell in which the fluids are re-circulated. Another possible reason for the observed behavior could be competing oxygen evolution reaction (OER). As the applied potential approaches water decomposition potential, partial current used for the OER will limit available current for glycerol conversion, thus leading to a lowering of glycerol converted. In addition, gas bubbles formed at the electrode surface compromise the speed of electro-oxidation.

Figure 34: Product buildup on catalyst surface 72

In addition, the time dependent behavior of glycerol at each potential was monitored (see Figure

35). Glycerol conversion is constant after the first twelve hours of the reaction, except for the oxidation at 1.1 V. At this potential, CuNiMoP achieves high glycerol conversion after 24 hours of continuous usage.

Figure 35: Glycerol concentration as a function of time

5.3.2 Effect of potential on product selectivity

A specific amount of energy input into a reacting system should affect selectivity and yield of specific oxidation products. Constant potential oxidations were carried out at 0.5 V, 0.7 V, 0.9

V, 1.1 V and 1.3 V for 24 hour sessions. Additional plots and figures related to this analysis are collated in Appendix C. 73

Oxidation at 0.5 V

Once dissolved in base, glyceraldehyde and dihydroxyacetone turn the solution yellow/ amber. This is usually the first visible indication that an electro-oxidation has occurred in the electro-chemical reactor. At 0.5 V, less than 0.5 mg/ml glyceraldehyde is produced for the different times samples were taken (see Figure 36). However, the major oxidation product for the entire duration of the experiment is formic acid, even though the total glycerol conversion is low (5.3%).

Figure 36: Glycerol conversion (A) and products formed (B) at 0.5 V

Within the first two hours of reaction, 1 mg/ml of tartronic acid is produced. This is consumed to make formic acid because its gradual decay from the fourth hour corresponds to a smooth increase in formic acid production. Prior to that, multiple contributions to formic acid formation distort its behavior. Once the concentrations of tartronic and mesoxalic acids decrease, formic acid behavior becomes an exponential rise. Small quantities of tartronic acid and mesoxalic acid are formed, but these are consumed over the course of the synthesis to form formic acid through carbon bond

74 scission reactions. At the end of 24 hours, only formic acid and glyceraldehyde are produced in quantities 0.5 mg/ml or greater.

Oxidations at 0.7 V

At 0.7 V, a higher glycerol conversion of 14.7% is achieved (see Figure 37). Higher quantities of glyceraldehyde are formed, and its concentration versus time behavior is like that of tartronic acid and formic acid. Lactic acid evolution is seen in small amounts at this potential. This is because of homogeneous reactions that arise from glyceraldehyde interactions with base, and so its formation follows no clear pattern.

Glyceraldehyde is a particularly difficult product to isolate in alkaline or aqueous media.

Once in solution it breaks up into many fractions, and this complicates its identification. Even though glycerol oxidation does not occur in the absence of base [63], its presence allows for any glyceraldehyde formed to undergo other chemical reactions. Glyceraldehyde/DHA undergo base catalyzed benzylic acid re-arrangement reactions to yield lactic acid at room temperature [64, 65].

These chemical reactions may be reduced by lowering the base content in the reactor, but unfortunately, this will negatively impact glycerol conversion.

1 mg/ml tartronic acid is seen after 4 hours of reaction, reaching a maximum of 5.7 mg/ml after 24 hours. For the first three hours of reaction, tartronic acid generation is balanced by consumption. Beyond 4 hours, generation outstrips consumption, and higher quantities are then seen in the unstirred bath.

No mesoxalic acid was seen in the product mix at the end of the reaction. Rather, oxalic acid was formed, gradually reaching a maximum 1.1 mg/ml after 24 hours. The behavior of oxalic at this potential is very like that of formic acid and suggests:

1. First glycerol oxidation to mesoxalic acid

2. Oxidation of mesoxalic acid to oxalic acid and formic acid

Figure 37: Glycerol conversion (A) and products formed (B) at 0.7 V

Oxidations at 0.9 V

At 0.9 V, the total number of major projects formed has increased to 5 (from two at 0.5 V and three at 0.7 V) (Figure 38). 10 mg/ml of tartronic acid is formed exponentially after 7 hours, and it is consumed gradually until 24 hours when only 4 mg/ml is seen. Mesoxalic acid profile is similar: it is produced and consumed during reaction, with a peak of 2 mg/ml at 5 hours.

Glyceraldehyde is also formed at this potential, reaching a plateau of 2 mg/ml at 10 hours, which is maintained until the end of the reaction. More lactic acid is also seen at this potential compared to the lower potentials, and corresponds to increased quantities of glyceraldehyde seen. Formic acid produced peaks at 5.5 mg/ml after 5 hours and reaches a plateau of 5 mg/ml which is

76 maintained until the end of the reaction. The preferred product at this potential is dihydroxyacetone, which rises exponentially to reach 8 mg/ml after 24 hours.

Figure 38: Glycerol conversion (A) and products formed (B) at 0.9 V

Figure 39: Glycerol conversion (A) and products formed (B) at 1.1 V

Oxidations at 1.1 V

The attraction for this project was to use non-precious metals to synthesize the 3-carbon oxidation products. For this reason, more emphasis is given in the potential range in which thermodynamic calculations suggest high likelihood of their formation (0.9 V to 1.1 V vs

Ag/AgCl).

Glycerol conversion of 62 % was achieved at 1.1 V (see Figure 39). DHA is an important transient at 1.1 V as it rises to 22 mg/ml after 5 hours, before being consumed completely by the

24th hour. Similar quantities of tartronic acid are at the 7th and 24th hour, and imply that the amounts seen at 17th and 20th hour are distortions to a plateau reached after 7 hours. Alternatively, it could be formed quickly and completely consumed during the first 17 hours. A secondary production then starts again, reaching 16 mg/ml after 24 hours. While this is happening, the quantities of mesoxalic acid, glyceraldehyde and mesoxalic acid rise uniformly. Results at this potential show that at 1.1 V, selectivity to DHA is high, while at 24 hours, glyceraldehyde, tartronic and mesoxalic acids are preferred. However, unwanted formic acid formation is also increasing.

Oxidations at 1.3 V

A glycerol conversion of 42% was achieved after 24 hours at 1.3 V (see Figure 40). The reasons for this lower performance have already been given in the preceding sections on glycerol conversion. Over the potential range explored, the largest quantity of glyceric acid was seen in the product mix, rising linearly up to 0.5 mg/ml after 24 hours.

At 1.3 V, 11. 3 mg/ml tartronic acid is seen after 14 hours, rising to a value of 12.3 mg/ml after 24 hours. An average of 12 mg/ml mesoxalic acid was obtained at 1.3 V after the first 10 hours. Higher quantities of formic acid are evidence that unwanted carbon scission products are

78 seen at this higher potential. This expected, given that mexoxalic acid is the deepest oxidized three carbon product. Any further oxidation must necessarily lead to breakdown products. Mesoxalic production reaches a maximum at 18 hours before falling off. Within the same time frame that this is occurring, DHA production rises. This relationship suggests that DHA and mesoxalic acid lie on the same reaction pathway.

Figure 40: Glycerol conversion (A) and products formed (B) at 1.1 V

5.4 Reaction Mechanism and Pathways

Formic acid is a single carbon oxidation product, and is likely formed through glyceraldehyde to tartronic acid. After 24 hours at each potential, its yield is inversely proportional to glycerol conversion (Figure 41). In general, formation of formic acid is favored at lower potentials, with the highest selectivity obtained at 0.5 V (total glycerol conversion of 5.3%). At this voltage, the presence of mesoxalic acid, tartronic acid and glyceric acid in the product mix show that some 3-carbon oxidation products are formed. Since the cell is unstirred, these remain

79 near the catalytic surface, and thus allows for bond scission reactions to occur. Hence, precursors to the formation of formic acid are not easily desorbed after formation. A comparison with 0.7 V

(glycerol conversion 14.7%) shows higher yields of glyceraldehyde. Further analysis shows that there is an exponential decay in the quantity of formic acid formed until 1.1 V, when production again increases, mirroring overall glycerol conversion.

Figure 41: Mechanism of formic acid formation

The amount of glyceraldehyde/DHA formed increases with increase in glycerol conversion as the potential is increased. Yield then declines at 1.1 V before rising again at 1.3 V (Figure 42).

Glyceraldehyde is the first oxidation product of glycerol and occurs by the abstraction of a hydrogen molecule. 80

Figure 42: Mechanism of formation of glyceraldehyde/DHA

Glyceric acid quantities seen from analysis of liquid oxidation products using CuNiMoP/C were small. It constituted 10% of the all products at 0.5 V, decreasing to 3% or less at all other potentials. Considering its thermodynamics, its formation is a four-electron process. Since 8- and

10-electron products are seen in higher quantities, it is produced and consumed quickly. Figure 43 shows the likely mechanism for glyceric acid formation.

Figure 44 and Figure 45 show the mechanism for the formation of tartronic acid and mesoxalic acid respectively. Low quantities of both chemicals are produced at the lower potentials.

Productions of booth reach a maximum at 1.1 V before falling off at 1.3 V when glycerol conversion decreases.

Figure 43: Mechanism of glyceric acid formation

Figure 44: Mechanism of tartronic acid formation

Figure 45: Mechanism of mesoxalic acid formation

From the presence, and quantities of identified products, the oxidation mechanism on

CuNiMoP is proposed as shown in Figure 46. Glycerol can be oxidized through the glyceraldehyde pathway or the dihydroxyacetone pathway. Glyceraldehyde is oxidized to glyceric acid which is further oxidized to tartronic acid or hydroxypyruvic acid. Tartronic acid can be oxidized to mesoxalic acid or into simpler single or double carbon molecules. Glyecrol could also be oxidized through a second oxidative route: from dihydroxyacetone to hydroxypyruvic acid to mesoxalic acid. Further oxidation of mesoxalic acid leads to the formation of carbon dioxide.

Product-potential relationships described above also show that products are preferentially formed at specific potentials (Figure 47). This agrees with the thermodynamic considerations of three carbon oxidation products.

Figure 46: Mechanism of glycerol oxidation on CuNiMoP/C

Figure 47: Preferred product(s) at given potentials

CHAPTER 6

STUDIES IN CATALYST STABILITY

6.1 Introduction

The ability to re-use and recycle electro-oxidation catalysts, while lowering the overall cost of

running the plant would also be important in making the process more environmentally friendly.

In this chapter, we report the investigation of the stability of CuNiMoP/C after three consecutive

24-hour oxidations of 1 M glycerol in 4 M NaOH at 0.7 V.

6.2 Catalyst Loss in Use

Active material leaching while in use remains a serious concern in catalyst development[66]. This may arise from reaction conditions or the particle characteristics of the deposited material. Using EDS, we studied the stability of electroless CuNiMoP. Figure 48 shows the relative proportions of copper, nickel, molybdenum and phosphorus in different electroless samples before and after in 1 M glycerol + 4 M NaOH.

91.7% of total deposit mass at 15 minutes’ deposition time is copper while nickel constitutes 7.1 %. Molybdenum and phosphorus make up the remaining 2%. At 30 minutes, 87.1% of deposit mass is the copper fraction while 9.8% is nickel. Molybdenum constitutes 1.1% while phosphorus is 1.9%. For the 45-minute sample, the deposited material consists of 87.5 % copper,

9.7% nickel, 1.1 % molybdenum and 1.82 % phosphorus.

Comparing quantities of these elements in before and after oxidations (see

Table 13), 90% of copper is lost from the 15-minute sample, 45% from the 30-minute

sample and 80% from the 45 minute deposits. Copper loss from the 30-minute sample was the

source of significant material loss. For three consecutive 24 hour oxidations, copper loss

corresponds to 90% and 50% decreases in activity for the 15- and 30- minute samples respectively

after three days (Figure 49).

Figure 48: Atomic percentages of CuNiMoP/C from EDS. Balance is carbon

By this analysis, the most stable catalyst is the 30-minute sample, in which the only material leached out is copper. Hence, the instability of the CuNiMoP/C catalyst is a direct result of copper leaching. 86

Table 13: Used/unused ratio of elements in catalysts

Deposition time (min) Cu Ni Mo P

15 0.09 0.41 1.00 -

30 0.55 0.99 1.01 1.00

45 0.20 0.51 0.73 0.43

Figure 49: Constant potential oxidations at 0.7 V

6.3 Copper Behavior under Alkaline Conditions

All electro-oxidations of glycerol carried out during this project were done under alkaline

conditions. The advantages associated with this were a more facile oxygen reduction reaction and

thus faster kinetics. Speciation analysis were done at 25 0C and 80 0C for copper in the mixed reducing agent bath (see Figure 50). During these analysis, mixed reducing agent bath components were used to run simulations in Visual Minteq (see Appendix B). Copper deposition, as used to

87 prepare electroless CuNiMoP/C was done at 80 0C. Deposition most likely occurs from copper

-2 +2 hydroxides (Cu(OH)2 (aq), Cu(OH)4 and Cu3(OH)4 ) as these are the dominant species at deposition pH of 10.5.

Potentiostatic oxidations, on the other hand, were done at 25 0C in 4 M NaOH. At these

- -2 conditions, the major species existing are Cu(OH)3 and Cu(OH)4 . These ionic copper hydroxides are the preferred species at high pH and room temperature. This explains copper loss from the

CuNiMoP/C electro-catalyst while in use. Another loss route may be due to the electrode rinsing in between oxidations. However, all catalyst components should be washed off if this is the case.

Consideing the 30-minute sample, there was no nickel loss. Thus, loss through rinsing may have contributed an insignificant amount to the total catalyst loss.

Figure 50: Speciation of copper at 25 0C and 80 0C

CHAPTER 7

CONCLUSIONS AND FURTHER WORK

7.1 General Conclusions

This project has focused on the electro-oxidation oxidation of glycerol under alkaline conditions with electroless CuNiMoP/C. The anode side phenomena in a divided electrochemical cell were studied using a series of physical and electrochemical approaches.

The anode side catalyst was electrolessly fabricated from copper, nickel, molybdenum and phosphorus using a mixture of formaldehyde and sodium hypophosphite as reducing agents.

Copper was found to be the major active ingredient in this electro-catalyst, even though it is easily leached out. The inclusion of molybdenum, nickel and phosphorus improved the performance and mechanical properties of the catalyst.

The electrochemical performance of CuNiMoP showed that activity towards glycerol oxidation could be compared to those of precious metal catalysts in direct glycerol fuel cells. This implies lower catalyst costs in the utilization of biodiesel derived glycerol

The oxidation pathways for glycerol on CuNiMoP were determined. For 24 hour oxidations at given potentials, glyceraldehyde and DHA are the major products at 0.7 V and 0.9 V, while higher yields of tartronic acid and mesoxalic acid were obtained at 1.1 V and 1.3 V. This is in line with estimates from group additivity calculations. Product selectivity and yield vary depending on applied potential, with oxidation at 1.1 V leading to both high glycerol conversion and yields of

3-carbon oxidation products.

Catalyst de-activation was also investigated. It was determined that the major de-activation mechanism arose from copper leaching of the catalyst composite. In addition, fouling arising from carbonyl (CO) species encountered during the electro-oxidation process may also occlude the catalyst surface.

7.2 Directions for Future Work

The main contribution at the beginning of this project was that a mixture of reducing agents could be used for the synthesis of electro-catalysts. This catalyst containing about 80% - 90% copper based on electroless deposition time was de-activated by copper leaching. Optimization of this catalyst to ensure greater cohesion of catalyst alloy to support can improve catalyst life. Other mechanisms to catalyst de-activation should be explored.

Electrochemical characterization of CuNiMoP/C as prepared gave various electrochemical parameters that are comparable to those of precious metals. However, all investigations carried out for constant potential oxidations were done in batch, unstirred electrolytes in a divided electrochemical reactor. Even though this reactor had essential elements of a fuel cell – anode, cathode, anion exchange membrane – it differed from an actual fuel cell as the reactor was not operated under flow conditions. Secondly, the inter-electrode distance between the anode and the cathode in an actual fuel cell will be smaller. Using CuNiMoP/C under actual fuel cell conditions will create a better understanding of performance as a fuel cell anode.

APPENDIX A

HPLC METHOD DEVELOPMENT

High pressure liquid chromatography (HPLC) is a versatile tool for quick analysis and identification of the components of a liquid mixture. Its usefulness hinges on two cardinal points:

A. An eluent will transport various solutes across a stationary phase depending on

1. The flow rate of the mobile phase

2. The characteristics of the stationary phase (ion exchange (anion or cation), etc)

3. The composition of the mobile phase

4. The column temperature

5. Injection volume of the analyte

B. If all the conditions in (A) above are kept constant, any material that has the same retention

time as a standard which has been run with the conditions must be the same substance as

the standard.

To detect the presence of substances, the chromatograph is coupled to an equipment that can sense relative amounts of the analyte. Typically UV-Vis and infra-red spectrometers are used. For the

UV-Vis, the wavelength has to be fixed as the same material can have different absorbance values at different wavelengths. The addendum to (B) above is that different concentrations of a substance will send different intensities of signals to the detector. This enables the calibration of concentrations of a particular standard, so that its presence in an unknown liquid mixture can be

91 quantified. These conditions, which have to be set before the analysis of a sample is done, constitute the HPLC method.

Choice of column

This decision is one of the most important as it is actually in the column that separation occurs. As the oxidation of glycerol leads to the formation of carboxylic acids and aldehydes, ion exchange or ion exclusion columns are typically used. Beltran-Prieto et al give an excellent review of HPLC methods commonly used for analysis of glycerol and its oxidation products[67]. Typical columns include: Aminex HPX-87H [34, 68-71]; OA-1000 column (Alltech) [63, 72]; Zorbax

SAX column[73]; Metacarb 67H column[74]; IC Pak[75]; Hitachi GL-C610-S column[76]. These columns contain materials that ionize, and thus by coulombic interactions, retain oppositely charged ions that contact them. Hence, any substance that does not ionize will not be properly separated via this mechanism.

Choice of wavelength

The choice of wavelength is informed by the UV visibility of the target compound, in an eluent that will not distort the spectra. Given the ubiquity of methods, we decided to determine the appropriate wavelength in house using water or dilute sulphuric acid as eluent. Using Cary 60 UV-

Vis equipment, solutions of glycerol in water were scanned from 800 to 200 nm. The same experiment was done in 15% acetonitrile (balance water). The results (see Figure 1) showed that below 260, glycerol showed absorbance behavior. The same series of experiments were also done for tartronic acid, and the similar results were obtained, that is, better spectra were obtained at higher frequencies. Thus the HPLC analyses were run at 190 nm.

Figure 51: UV-Vis absorbance of tartronic acid and glycerol

Summary of HPLC Conditions

These are the conditions that were used to run the HPLC

Column: Hi-plex

Oven temperature: 40 oC

Mobile phase: 5mM H2SO4

Flow rate: 0.7 ml/min (isocratic mode)

Injection volume: 10 µL

Wavelength: 190 nm

Calibration

To get exact retention times, aqueous solutions of the pure substances were run through the HPLC for 30 minutes to ensure that all the material was eluted. This gave a rough estimate of where the elution times should be. Subsequently, shorter analysis times were chosen. In addition, to replicate conditions under which the electro-oxidation may occur, standards were dissolved in

1M NaOH. 1 ml of the resulting solution were then mixed with 6ml of 1M H2SO4 solution. This formed the initial stock solution which was then diluted and used to generate concentrations versus area relationships in the chromatogram.

Figure 52: Chromatograms for standards (Glyceraldehyde, DHA, glycerol, mesoxalic acid and tartronic acid). 94

Figure 52 shows chromatograms of standards calibrated. The peaks with the red dots are the characteristic peaks used to identify the compounds. The peaks with asterix appear at the same point in all the chromatograms and maybe due to the eluent interactions with the column

Calibration protocol: Example with glycerol

1. 3.4765g of glycerol was dissolved in 10 ml of 1M NaOH solution. The density of glycerol

is 1.26g/cm3. Correcting for the added volume of glycerol, this gives 272.47 g glycerol per

liter of solution.

Table 14: Concentration versus area data for glycerol

Conc (mg/ml) Peak area

0 0

38.92 6,629,663.00

19.46 3,642,038.00

9.73 1,873,785.00

4.87 931,082.00

4.87 932,004.00

2.43 467,282.00

2.43 467,318.00

1.22 229,481.00

1.22 229,417.00

2. 1 ml of this sample was added to 6 ml of 1M sulfuric, to give the quantity of glycerol after

this dilution as 38.92 mg/ml. This was used as a first calibration point.

3. The next point was obtained by taking 1 ml of the solution from (2) and diluting it with 1

ml of 1M sulfuric acid.

4. Step 3 was repeated to give successively more dilute solutions.

5. The different concentrations were then run separately through the HPLC. Table 1 shows

the concentration versus area data obtained. Some of the experiments were duplicated.

Figure 53: Calibration curves: Glycerol, glyceraldehyde, DHA, mesoxalic and tartronic acids

Glycerol chromatograms

Figure 54: Glycerol chromatograms. Decreasing concentrations of glycerol and sulfuric acid

Glyceraldehyde

Glyceraldehyde was one of the more expensive chemicals, and was handled very carefully.

Despite this, its chromatogram showed multiple peaks. This may be why it is so expensive, because it is difficult to isolate it in its pure form. However one peak had to be chosen to calibrate it. To do this, different chromatograms were plotted to isolate the ones that were common.

Observations

1. There is a characteristic peak that comes out at 6.2 min, which was seen in all the samples

(see Figure 2). If sample was mixed with distilled water (ie there was no dilution with

sulfuric acid), the 6.2 min peak was inverted (tartronic acid in Fig 2). If the stock solution

was made with 6ml of 1M sulfuric acid, but each subsequent dilution was with DI water,

the 6.2 min elution became smaller. This confirms that the peak is due to column-solvent

interactions.

2. The peak at 9.08 min was the most prominent peak in the glyceraldehyde columns. This

was used to calibrate glyceraldehyde

3. The method developed could be used to detect glyceraldehyde concentrations in

concentrations as low as 0.2 mg/ml.

Figure 55: Glyceraldehyde chromatograms at different concentrations (mg glyceraldehyde per ml of solution)

Dihydroxyacetone calibration

Dihydroxyacetone, like glyceraldehyde, presented multiple peaks with the HPLC method.

A total of 12 chromatograms were generated, and 16 peaks seen, though they did not occur in all.

The results are presented in the table 2below.

Analysis

1. The peaks with retention times of 6.473, 7.584, 8.304, 9.024, 10.096 and 11.699 minutes

show fairly linear behavior in the concentration versus area plots.

2. For chromatograms corresponding to area7 and area8, the peaks at 9.024 and 10.096

minutes show inconsistencies. These peaks should be duplicates. If however, all the areas

under the peaks from 9.024 mins to 14.9 minutes are added for these two chromatograms,

a constant value is obtained. For this reason, these retention times would not be used to

characterize DHA

3. 11.699 minutes also is too close to glycerol at 11.8.

4. This limits the characteristic peaks to those occurring at 6.473, 7.584, 8.304, 9.024 minutes.

5. The peak that comes out at 6.2 is still seen, and is attributed to eluent/column interactions

Figure 57 shows DHA fractions which increase linearly with increasing concentration. A

consolidated curve has average area under peaks for the five peaks analyzed on the y axis.

Figure 56: DHA chromatograms showing different concentrations

Figure 57: DHA presents multiple peaks, five of which show linear behavior.

100

Mesoxalic acid chromatograms

Figure 58: Mesoxalic acid chromatograms

Figure 59: Tartronic acid chromatograms

101

Retention time analysis

Each analyte was run through the column at least twice. Analysis of the peaks led to the following retention times associated with each chemical.

Table 15: Retention times for various standards

Substance Number of samples Retention times (min)

1 Glycerol 12 11.813 +/- 0.013

2 DHA 12 6.465+/- 0.0098

7.587 +/- 0.0051

8.313 +/- 0.0049

9.021 +/- 0.0018

10.064 +/- 0.025

3 Tartronic acid 7.796 +/- 0.0338

4 Glyceraldehyde 12 9.082 +/- 0.018

5 Mesoxalic acid 6.808 +/- 0.0441

102

APPENDIX B

SPECIATION IN THE MIXED REDUCING AGENT ELECTROLESS BATH

Metal ions in mixed reducing agent baths are expected to exist as either free metal ions or complexes with organic/inorganic ligands. The existence of these species are pH dependent, and a knowledge of which ones predominate at the deposition pH can be used to predict which species participate in electroless deposition.

Solution chemistry techniques can be used to study such speciation. The distribution of an element among various species in an aqueous environment can be determined using mass balances and mass action laws [9]:

(B.1) x = β c

(B.2) A x M = γ

where = total mass of component j; = stoichiometric coefficient of moles of j in species I;

= activityM coefficient of species I; A = number of components; = overall equilibrium

formationγ constant N β

103

These two equations were used in the formulation of the Visual Minteq2 program that was used to predict the speciation of the ions in the mixed reducing agent bath. The main inputs into the program were: temperature (which was either set at 25 oC (ambient conditions) or 80 oC

(deposition temperature); ionic strength of the solution (calculated by summing up the product of the molality of each solution component and the square of the charge of its ions in solution) and the concentration of each component. Iterations were run for the ion speciation at each pH, and the results recorded.

Copper speciation and primary deposition mechanism

The differences between the predominant species in the mixed reducing agent bath at room temperature and the deposition temperature are highlighted in Figure 9.

- +2 The predominant species at the deposition pH (10.25) are Cu(OH)2 (aq), Cu(OH)3 and Cu3(OH)4

o 2- at 25 C. Negligible amounts of Cu(OH)4 are seen at room temperature. At the higher temperature

2- however, Cu(OH)4 is seen to become very important, going on to become the only species existing at pH>11.5. This dominance shows that copper deposition in alkaline media involves copper ions chelated by hydroxide ions in a tetrahedral structure and coincides with the region of maximum rate of copper deposition [7]. Surprising the copper organic ligands do not appear to be important in the mechanism of deposition at the deposition temperature.

2 Downloaded from http://vminteq.lwr.kth.se/

104

1.0 A 1.0 B Cu+2 Cu(OH)2 (aq) Cu(OH)3- 0.8 0.8 2+ Cu(OH)4-2 2+ Cu-(Citrate)2-4 0.6 0.6 Cu-(Tartrate)2-2 Cu-Citrate-

Fraction of Cu Cu-Formate+ 0.4 0.4 Fraction of Cu Cu-Tartrate (aq) Cu2(OH)2+2 0.2 0.2 Cu2-(Citrate)2-2 Cu2OH+3 Cu3(OH)4+2 0.0 0.0 CuH-Citrate (aq) 4 6 8 10 12 9.0 9.5 10.0 10.5 11.0 11.5 12.0 CuH-Tartrate+ pH pH CuH2-Citrate+ 1.0 1.0 CuHPO4 (aq) C D CuOH+

0.8 0.8 0

2+ A = Copper speciation at 25 C 2+ between pH =3 and pH = 13 0.6 0.6 0 B= Copper speciation at 25 C

0.4 0.4 between pH =9 and pH =12 Fraction of Cu

Fraction of Cu 0 C = Copper speciation at 80 C 0.2 0.2 between pH =3 and pH = 13

0 D = Copper speciation at 25 C 0.0 0.0 between pH =9 and pH = 12 4 6 8 10 12 9.0 9.5 10.0 10.5 11.0 11.5 12.0 pH pH

Figure 60: Copper speciation in the mixed reducing agent bath

This analysis also highlights one important consideration in electroless copper deposition:

Electroless copper deposition can only happen at elevated temperatures because this is the temperature at which the species responsible for the deposition exist. A mechanism for copper deposition once a substrate is introduced can thus be postulated: the chelated copper (II) ions attack the formaldehyde, and oxidize it to formic acid, with the generation of hydrogen gas, and the release of two electrons in one half cell reaction:

(B.3) Cu(OH) + 2HCHO → 2HCOO + H + 2e

105

This process releases free Cu (II) ions in the vicinity of the 2 electrons released from the formaldehyde oxidation, and a further reaction takes place to complete the redox couple in the other half reaction:

(B.4) Cu + 2e → Cu

Nickel, phosphorus and molybdenum speciation

Nickel

The distribution of nickel ions in solution was also examined. Figure 11 shows the variation of nickel containing species across a pH range from 3 to 12. Organic ligands play a far more important role in nickel speciation as evidenced by their predominance at 10.25. Because much lower quantities of nickel are deposited compared to copper, the deposition of nickel does not occur from the reactions of the citrate containing species. The next species of interest is

NiOH+. This is the species that Ni deposition primarily proceeds from, as it is not visible in

Figure 35 (B) at pH 10.25 for 25 oC, but has a clear presentation at 80 oC.

In a 2013 paper, Cavallotti et al [10] used density functional theory calculations to investigate the deposition of NiP from hypophosphosphite baths. They found that the most likely mechanism for the electroless nickel involved the formation of NiOH+. They postulated a two- step mechanism: First free nickel ions are adsorbed on the substrate

(B.5) Ni + surf → Ni

106

A B 0.8 0.8 Ni+2 Ni(OH)2 (aq)

2+ Ni(OH)3- 0.6 0.6 Ni-(Citrate)2-4

2+ Ni-Citrate- Ni-Formate+ 0.4 0.4 Ni-Tartrate (aq) Fraction of Ni Fractionof NiH-(Citrate)2-3 0.2 NiH-Tartrate+ Fraction of Ni of Fraction 0.2 NiH2-Citrate+ NiH2PO4+ 0.0 0.0 NiHPO4 (aq) 4 6 8 10 12 9.0 9.5 10.0 10.5 11.0 11.5 12.0 NiOH+ pH pH

1.0 C 0 D A = Nickel speciation at 25 C 0.8 between pH =3 and pH = 13 0.8

0 0.6 B = Nickel speciation at 25 C 0.6 between pH =9 and pH = 12 2+ 2+ 0.4 0 0.4 C = Nickel speciation at 80 C between pH =3 and pH = 13

Fraction of Ni of Fraction 0.2 0.2 0 Fraction of Ni of Fraction D = Nickel speciation at 80 C between pH =9 and pH = 12 0.0 0.0 4 6 8 10 12 9.0 9.5 10.0 10.5 11.0 11.5 12.0

pH pH

Figure 61: Nickel speciation in mixed reducing agent bath

This adsorbed nickel then interacts with hypophosphite ions and water molecules to form a nickel-phosphorus complex

(B.6) Even though no evidenceNi + for2H freeO + nickel H P Oions→ atNiOH pH =( 10.25H PO was) found+ H inO these simulations, they pointed out one interesting fact: NiOH+ does not exist in the pH range (5 to 9) where nickel deposition normally occurs. This is a strong support for the evidence from the speciation analysis that nickel deposition occurs from NiOH+ once a substrate is introduced in the electroless bath 107

Molybdenum and phosphorus speciation

The total amount of metal distributed among the various species should always sum up to the initial amount dissolved. If predictions of aqueous species do not account for the total metal mass, the possibility that some species may have been precipitated exists.

Figure 62 shows the predominant molybdenum species in the bath. The total concentration of sodium molybdate dissolved in the electroless bath was 1.00 g, and this quantity could not be accounted for in the bath. Solubility data, using saturation index was thus examined.

0 0.008 Molybdenum speciation in (80 C)

H2Mo6O21-4 H3Mo8O28-5 0.006 HMo7O24-5 HMoO4- Mo7O24-6 0.004 Mo8O26-4 MoO3(H2O)3(aq) Mo7O24-6 Mo8O26-4 0.002 MoO3(H2O)3(aq)

Concentration of molybdenum containing species 0.000 4 6 8 10 12 pH

Figure 62: Molybdenum speciation at 80 oC

The saturation index is a measure of the tendency of a mineral to be in solution in an aqueous medium. Mathematically,

108

IAP SI = log( ) K

Where SI is the saturation index, IAP is the ionic activity product and KSP is the solubility product.

If SI is negative, the mineral is undersaturated, if positive, the solid has the potential to deposit, while if it is unity, then the mineral is in chemical equilibrium with water (saturated).

The solubility data for the various molybdenum species across the pH range of interest to see if they were likely to exist as solid at that pH. The results are shown using the saturation indices in

Figure 63. The results show that the formation of a nickel-molybdenum complex is likely.

4.5

-0.5 9 9.5 10 10.5 11 11.5 12

-5.5

-10.5 Saturation index

-15.5

-20.5 pH

CuMoO4(s) H2MoO4(s) MoO3(s) NiMoO4(s)

Figure 63: Saturation index for molybdenum containing compounds in mixed reducing agent bath

109

Similar behavior is obtained for the phosphorus component of the deposits. The amount of the

phosphorus containing compound in the electroless bath was 1.35 g. A mass balance showed that

all the phosphorus was not in solution. Figure 14 showed the speciation of phosphorus in the bath.

0.05 CuHPO4 (aq) H2PO4- 0.04 H3PO4 HPO4-2 NiH2PO4+ 0.03 NiHPO4 (aq)

0.02

0.01 Conc. phosphorusof containing species (M)

0.00 4 6 8 10 12 pH

Figure 64: Phosphorus speciation in mixed reducing agent bath

Saturation data also predict the formation of nickel-phosphorus complexes (Figure 65).

-2 9 9.5 10 10.5 11 11.5 12

-7

-12 Saturation index

-17 pH

Cu3(PO4)2(s) Cu3(PO4)2:3H2O(s) Ni3(PO4)2(s)

Figure 65: Saturation index data for phosphorus species

These speciation results explain some of the deposition data already obtained. Since nickel, phosphorus and molybdenum deposition occur today, any decreases in amount of nickel deposited

110 will similarly affect the amount of phosphorus and molybdenum deposited. It is also known that molybdenum cannot be deposited from electroless bath, except as components alloyed too another metal[10]. Its inclusion in the deposits arise from reactions between contiguous molybdenate and nickel ions:

(B.7) () Ni + MO → NiMnO

111

REFERENCES

[1] R. Ciriminna, C.D. Pina, M. Rossi, M. Pagliaro, Understanding the glycerol market, European Journal of Lipid Science and Technology, 116 (2014) 1432-1439.

[2] E.P.C.o.t.E. Union, Directive 2009/28/EC of the European Parliament and of the Council of 23 April 2009 on the promotion of the use of energy from renewable sources and amending and subsequently repealing Directives 2001/77/EC and 2003/30/EC, European Parliament - Legislative Observatory, Official Journal of the European Union, 2009.

[3] M. Nanda, Z. Yuan, W. Qin, M. Poirier, X. Chunbao, Purification of crude glycerol using acidification: effects of acid types and product characterization, Austin Journal of Chemical Engineering, 1 (2014) 1004.

[4] U.S.E.I. Administration, Short term energy outlook, in: I.S.a. Analysis (Ed.), 2017.

[5] J.C. Thompson, B.B. He, Characterization of crude glycerol from biodiesel production from multiple feedstocks, Applied Engineering in Agriculture, 22 (2006) 261-265.

[6] M. Bloch, Improved glycerin quality via solid catalyst trasnesterification technology: The Esterfip‐H process, Proc. Bio-Oil Int. Conf., Vigo, Spain, 2006.

[7] J.A. Posada, J.C. Higuita, C.A. Cardona, Optimization on the Use of Crude Glycerol from the Biodiesel Production to Obtain Poly-3-Hydroxybutyrate, World Renewable Energy Congress- Sweden; 8-13 May; 2011; Linköping; Sweden, Linköping University Electronic Press, 2011, pp. 327-334.

[8] M. Bowker, P.R. Davies, L.S. Al-Mazroai, Photocatalytic reforming of glycerol over gold and palladium as an alternative fuel source, Catalysis letters, 128 (2009) 253-255.

[9] M. Gupta, N. Kumar, Scope and opportunities of using glycerol as an energy source, Renewable and Sustainable Energy Reviews, 16 (2012) 4551-4556.

[10] M.E. Jamróz, M. Jarosz, J. Witowska-Jarosz, E. Bednarek, W. Tęcza, M.H. Jamróz, J.C. Dobrowolski, J. Kijeński, Mono-, di-, and tri-tert-butyl ethers of glycerol: A molecular spectroscopic study, Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 67 (2007) 980-988.

[11] M. Rossi, C. Della Pina, M. Pagliaro, R. Ciriminna, P. Forni, Greening the construction industry: Enhancing the performance of cements by adding bioglycerol, ChemSusChem, 1 (2008) 809-812.

[12] L. Research, The Bio-based Chemical Industry through 2030, 2014.

112

[13] Biodiesel biorefinery: opportunities and challenges for microbial production of fuels and chemicals Biotechnology for biofuels, 5 (2012).

[14] Selective catalytic oxidation of glycerol: perspectives for high value chemicals, Green Chemistry, 13 (2011) 1960-1979.

[15] Oxidation of tartronic acid and dihydroxyacetone to sodium mesoxalate mediated by TEMPO, Tetrahedron letters, (2004) 6381-6383.

[16] Hypoglycemic effect of mesoxalate produced by infusion into the pancreatic circulation in the dog, J. Pharmacolol. Exp Ther 115 (1955) 343-349.

[17] V. Augugliaro, H.H. El Nazer, V. Loddo, A. Mele, G. Palmisano, L. Palmisano, S. Yurdakal, Partial photocatalytic oxidation of glycerol in TiO 2 water suspensions, Catalysis Today, 151 (2010) 21-28.

[18] Y. Zhang, N. Zhang, Z.-R. Tang, Y.-J. Xu, Identification of Bi 2 WO 6 as a highly selective visible-light photocatalyst toward oxidation of glycerol to dihydroxyacetone in water, Chemical Science, 4 (2013) 1820-1824.

[19] Z.-C. Hu, Y.-G. Zheng, Y.-C. Shen, Dissolved-oxygen-stat fed-batch fermentation of 1,3- dihydroxyacetone from glycerol by Gluconobacter oxydans ZJB09112, Biotechnology and Bioprocess Engineering, 15 (2010) 651-656.

[20] Electrocoagulation/Electrodeposition, 4th European Summer School on Electrochemical EnergyPalic, Serbia and Montenegro, 2006.

[21] C. Lamy, A. Lima, V. LeRhun, F. Delime, C. Coutanceau, J.-M. Léger, Recent advances in the development of direct alcohol fuel cells (DAFC), Journal of Power Sources, 105 (2002) 283- 296.

[22] B.S. Pivovar, Y. Wang, E. Cussler, Pervaporation membranes in direct methanol fuel cells, Journal of Membrane Science, 154 (1999) 155-162.

[23] S. Wasmus, A. Küver, Methanol oxidation and direct methanol fuel cells: a selective review, Journal of Electroanalytical Chemistry, 461 (1999) 14-31.

[24] E. Antolini, E. Gonzalez, Alkaline direct alcohol fuel cells, Journal of Power Sources, 195 (2010) 3431-3450.

[25] Z. Zhang, L. Xin, W. Li, Electrocatalytic oxidation of glycerol on Pt/C in anion-exchange membrane fuel cell: cogeneration of electricity and valuable chemicals, Applied Catalysis B: Environmental, 119 (2012) 40-48.

[26] E.S. Domalski, E.D. Hearing, Heat capacities and entropies of organic compounds in the condensed phase. Volume III, Journal of physical and chemical reference data, 25 (1996) 1-525. 113

[27] N. Cohen, S. Benson, Estimation of heats of formation of organic compounds by additivity methods, Chemical Reviews, 93 (1993) 2419-2438.

[28] X. Ma, H.H. Schobert, Estimating heats of formation of hydrocarbon radicals by a combination of semiempirical calculation and family correlation with experimental values, The Journal of Physical Chemistry A, 104 (2000) 1064-1074.

[29] M.L. Barbelli, G.F. Santori, N.N. Nichio, Aqueous phase hydro genolysis of glycerol to bio-propylene glycol over Pt-Sn catalysts, Bioresour Technol, 111 (2012) 500-503.

[30] J.-H. Wee, K.-Y. Lee, S.H. Kim, Fabrication methods for low-Pt-loading electrocatalysts in proton exchange membrane fuel cell systems, Journal of Power Sources, 165 (2007) 667-677.

[31] A.N. Grace, K. Pandian, Pt, Pt–Pd and Pt–Pd/Ru nanoparticles entrapped polyaniline electrodes–A potent electrocatalyst towards the oxidation of glycerol, Electrochemistry communications, 8 (2006) 1340-1348.

[32] E. Venancio, W. Napporn, A. Motheo, Electro-oxidation of glycerol on platinum dispersed in polyaniline matrices, Electrochimica acta, 47 (2002) 1495-1501.

[33] M. Rafiee, B. Karimi, S. Alizadeh, Mechanistic Study of the Electrocatalytic Oxidation of Alcohols by TEMPO and NHPI, ChemElectroChem, 1 (2014) 455-462.

[34] M. Simões, S. Baranton, C. Coutanceau, Enhancement of catalytic properties for glycerol electrooxidation on Pt and Pd nanoparticles induced by Bi surface modification, Applied Catalysis B: Environmental, 110 (2011) 40-49.

[35] P.S. Fernández, M.E. Martins, C.A. Martins, G.A. Camara, The electro-oxidation of isotopically labeled glycerol on platinum: New information on C–C bond cleavage and CO 2 production, Electrochemistry Communications, 15 (2012) 14-17.

[36] C. Xu, R. Zeng, P.K. Shen, Z. Wei, Synergistic effect of CeO 2 modified Pt/C catalysts on the alcohols oxidation, Electrochimica acta, 51 (2005) 1031-1035.

[37] C. Xu, Z. Tian, P. Shen, S.P. Jiang, Oxide (CeO 2, NiO, Co 3 O 4 and Mn 3 O 4)-promoted Pd/C electrocatalysts for alcohol electrooxidation in alkaline media, Electrochimica Acta, 53 (2008) 2610-2618.

[38] I. Sobczak, K. Jagodzinska, M. Ziolek, Glycerol oxidation on gold catalysts supported on group five metal oxides—A comparative study with other metal oxides and carbon based catalysts, Catalysis Today, 158 (2010) 121-129.

[39] D. Liang, J. Gao, J. Wang, P. Chen, Y. Wei, Z. Hou, Bimetallic Pt―Cu catalysts for glycerol oxidation with oxygen in a base-free aqueous solution, Catalysis Communications, 12 (2011) 1059-1062. 114

[40] I. Sobczak, K. Szrama, R. Wojcieszak, E.M. Gaigneaux, M. Ziolek, Cu x Cr y O z mixed oxide as a promising support for gold–The effect of Au loading method on the effectiveness in oxidation reactions, Catalysis today, 187 (2012) 48-55.

[41] Y. Li, S. Chen, J. Xu, H. Zhang, Y. Zhao, Y. Wang, Z. Liu, Ni promoted Pt and Pd catalysts for glycerol oxidation to lactic acid, CLEAN–Soil, Air, Water, 42 (2014) 1140-1144.

[42] F.W. Schneble, J.F. McCormack, R.J. Zeblisky, 1976.

[43] A. Bienholz, H. Hofmann, P. Claus, Selective hydrogenolysis of glycerol over copper catalysts both in liquid and vapour phase: Correlation between the copper surface area and the catalyst's activity, Applied Catalysis A: General, 391 (2011) 153-157.

[44] H.D. Simpson, Gamma alumina catalyst support for molybdenum, nickel, phosphorus components, 1983.

[45] H. Müller, A.J. Bradley, CCXVII.—Copper hydride and its crystal structure, Journal of the Chemical Society (Resumed), 129 (1926) 1669-1673.

[46] A. Brenner, G.E. Riddell, Deposition of nickel and cobalt by chemical reduction, Journal of Research of the National Bureau of Standards, 39 (1947) 385.

[47] R. Agarwala, V. Agarwala, Electroless alloy/composite coatings: A review, Sadhana, 28 (2003) 475-493.

[48] J.R. Davis, Copper and copper alloys, ASM international2001.

[49] Z. Bangwei, X. Haowen, X. Xiewen, Autocatalytic deposition of nickel-tin-copper- phosphorus amorphous alloys, Metal finishing, 97 (1999) 35-41.

[50] Influence of pH solution on electroless copper plating using sodium hypophosphite as reducing agent, Int. J. of Electrochemistry, 7 (2012) 2009-2018.

[51] H. Ashassi-Sorkhabi, S.H. Rafizadeh, Effect of coating time and heat treatment on structures and corrosion characteristics of electroless Ni–P alloy deposits, Surface and coatings Technology, 176 (2004) 318-326.

[52] R. Ambat, W. Zhou, Electroless nickel-plating on AZ91D magnesium alloy: effect of substrate microstructure and plating parameters, Surface and Coatings Technology, 179 (2004) 124-134.

[53] B. Habibi, N. Delnavaz, Electrooxidation of glycerol on nickel and nickel alloy (Ni–Cu and Ni–Co) nanoparticles in alkaline media, RSC Advances, 6 (2016) 31797-31806.

115

[54] V. Oliveira, C. Morais, K. Servat, T. Napporn, G. Tremiliosi-Filho, K. Kokoh, Studies of the reaction products resulted from glycerol electrooxidation on Ni-based materials in alkaline medium, Electrochimica Acta, 117 (2014) 255-262.

[55] X. Han, D.J. Chadderdon, J. Qi, L. Xin, W. Li, W. Zhou, Numerical analysis of anion- exchange membrane direct glycerol fuel cells under steady state and dynamic operations, international journal of hydrogen energy, 39 (2014) 19767-19779.

[56] P.K. Shen, C. Xu, Alcohol oxidation on nanocrystalline oxide Pd/C promoted electrocatalysts, Electrochemistry Communications, 8 (2006) 184-188.

[57] Z. Zhang, L. Xin, K. Sun, W. Li, Pd–Ni electrocatalysts for efficient ethanol oxidation reaction in alkaline electrolyte, International Journal of Hydrogen Energy, 36 (2011) 12686-12697.

[58] E. Habibi, H. Razmi, Glycerol electrooxidation on Pd, Pt and Au nanoparticles supported on carbon ceramic electrode in alkaline media, international journal of hydrogen energy, 37 (2012) 16800-16809.

[59] M.O. Thotiyl, T.R. Kumar, S. Sampath, Pd supported on titanium nitride for efficient ethanol oxidation, The Journal of Physical Chemistry C, 114 (2010) 17934-17941.

[60] O.O. Fashedemi, K.I. Ozoemena, Comparative electrocatalytic oxidation of ethanol, ethylene glycol and glycerol in alkaline medium at Pd-decorated FeCo@ Fe/C core-shell nanocatalysts, Electrochimica Acta, 128 (2014) 279-286.

[61] L. Xin, Z.Y. Zhang, Z.C. Wang, W.Z. Li, Simultaneous Generation of Mesoxalic Acid and Electricity from Glycerol on a Gold Anode Catalyst in Anion-Exchange Membrane Fuel Cells, Chemcatchem, 4 (2012) 1105-1114.

[62] L. Xin, Z. Zhang, Z. Wang, W. Li, Simultaneous Generation of Mesoxalic Acid and Electricity from Glycerol on a Gold Anode Catalyst in Anion‐Exchange Membrane Fuel Cells, ChemCatChem, 4 (2012) 1105-1114.

[63] S. Carrettin, P. McMorn, P. Johnston, K. Griffin, C.J. Kiely, G.J. Hutchings, Oxidation of glycerol using supported Pt, Pd and Au catalysts, Physical Chemistry Chemical Physics, 5 (2003) 1329-1336.

[64] L. Chen, S. Ren, X.P. Ye, Lactic acid production from glycerol using CaO as solid base catalyst, Fuel Processing Technology, 120 (2014) 40-47.

[65] H. Kishida, F. Jin, Z. Zhou, T. Moriya, H. Enomoto, Conversion of glycerin into lactic acid by alkaline hydrothermal reaction, Chemistry Letters, 34 (2005) 1560-1561.

[66] E. Antolini, J.R. Salgado, E.R. Gonzalez, The stability of Pt–M (M= first row transition metal) alloy catalysts and its effect on the activity in low temperature fuel cells: a literature review and tests on a Pt–Co catalyst, Journal of Power Sources, 160 (2006) 957-968. 116

[67] J.C. Beltrán-Prieto, J. Pecha, V. Kašpárková, K. Kolomazník, Development of an HPLC method for the determination of glycerol oxidation products, Journal of Liquid Chromatography & Related Technologies, 36 (2013) 2758-2773.

[68] S. Demirel, K. Lehnert, M. Lucas, P. Claus, Use of renewables for the production of chemicals: Glycerol oxidation over carbon supported gold catalysts, Applied Catalysis B: Environmental, 70 (2007) 637-643.

[69] S. Carrettin, P. McMorn, P. Johnston, K. Griffin, G.J. Hutchings, Selective oxidation of glycerol to glyceric acid using a gold catalyst in aqueous sodium hydroxide, Chem. Commun., (2002) 696-697.

[70] W.C. Ketchie, Y.-L. Fang, M.S. Wong, M. Murayama, R.J. Davis, Influence of gold particle size on the aqueous-phase oxidation of carbon monoxide and glycerol, Journal of Catalysis, 250 (2007) 94-101.

[71] S. Liebminger, M. Siebenhofer, G. Guebitz, Oxidation of glycerol by 2, 2, 6, 6- tetramethylpiperidine-N-oxyl (TEMPO) in the presence of laccase, Bioresource technology, 100 (2009) 4541-4545.

[72] N. Dimitratos, C. Messi, F. Porta, L. Prati, A. Villa, Investigation on the behaviour of Pt (0)/carbon and Pt (0), Au (0)/carbon catalysts employed in the oxidation of glycerol with molecular oxygen in water, Journal of Molecular Catalysis A: Chemical, 256 (2006) 21-28.

[73] D. Liang, J. Gao, J. Wang, P. Chen, Z. Hou, X. Zheng, Selective oxidation of glycerol in a base-free aqueous solution over different sized Pt catalysts, Catalysis Communications, 10 (2009) 1586-1590.

[74] N. Dimitratos, J.A. Lopez-Sanchez, J.M. Anthonykutty, G. Brett, A.F. Carley, R.C. Tiruvalam, A.A. Herzing, C.J. Kiely, D.W. Knight, G.J. Hutchings, Oxidation of glycerol using gold–palladium alloy-supported nanocrystals, Physical Chemistry Chemical Physics, 11 (2009) 4952-4961.

[75] K. Musialska, E. Finocchio, I. Sobczak, G. Busca, R. Wojcieszak, E. Gaigneaux, M. Ziolek, Characterization of alumina-and niobia-supported gold catalysts used for oxidation of glycerol, Applied Catalysis A: General, 384 (2010) 70-77.

[76] W. Hu, B. Lowry, A. Varma, Kinetic study of glycerol oxidation network over Pt–Bi/C catalyst, Applied Catalysis B: Environmental, 106 (2011) 123-132.

117

BIOGRAPHICAL SKETCH EDUCATION Tallahassee, USA Florida State University Aug. 2012 – May 2017  PhD. Chemical Engineering, May 2017  Thesis: Co-generation of power and chemicals from direct glycerol fuel cells using CuNiMoP as anode.

London, UK Imperial College London Oct. 2002 – Sept. 2003  M.Sc. Advanced Chemical Engineering, November 2003  Thesis: Optimal volumetric gas cylinder filling.

Owerri, Nigeria Federal University of Technology Jun. 1994 – Aug. 1999  B. Eng. Chemical Engineering, January 2000  Thesis: Design of a pilot palm kernel oil refining plant.

CAREER HISTORY Research/Teaching Assistant Florida State University Aug. 2012 – May 2017  Assist the professors in journal reviews, tutorials, proctoring and administration of examinations.

Advisory Principal Nextzon Business Services Nov. 2006 – Dec. 2009  Performed business strategy development for several start-up companies in Nigeria.  Participated in mergers and acquisitions, contracts and compliance policy development for clients

Business Dev. Officer Interconnect Clearing House Jan. 2004 – Oct. 2006  Coordinated business plans development, engineering operations and billing reconciliations.

Youth Corper (Teacher) Obaji Comprehensive Sec. School Jul. 2000 – Aug. 2001  Taught senior students physics, mathematics and chemistry, and set-up/supervise lab experiments.

TECHNOLOGY EXPERIENCE Working experience of: Microsoft Suite, Matlab, Material Studio, Comsol Multiphysics, SEM, PSI- Plot, HPLC, RDE, Polymath and LAMMPS. ·

AWARDS AND RECOGNITIONS ·  Commonwealth Scholar, Imperial College London, 2002 – 2003  Best Academic Student, 1993 Graduating Class, Federal Government Girl’s College (FGGC) Owerri.  Federal Government of Nigeria University Scholarship – 1993  FGGC Owerri best student in English, Igbo, Chemistry, Technical Drawing and History, 1991 – 1992

118