Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Nano-sized carbon-supported molybdenum disulphide particles for hydrodesulphurization Solnickova, Lucie 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2017_february_solnickova_lucie.pdf [ 6.24MB ]
JSON: 24-1.0340640.json
JSON-LD: 24-1.0340640-ld.json
RDF/XML (Pretty): 24-1.0340640-rdf.xml
RDF/JSON: 24-1.0340640-rdf.json
Turtle: 24-1.0340640-turtle.txt
N-Triples: 24-1.0340640-rdf-ntriples.txt
Original Record: 24-1.0340640-source.json
Full Text

Full Text

LucieNano-Sized Carbon-Supported Molybdenum DisulphideParticles for HydrodesulphurizationbyLucie SolnickovaB.Sc., The University of British Columbia, 2014A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFMASTER OF APPLIED SCIENCEinTHE FACULTY OF GRADUATEAND POSTDOCTORAL STUDIES(Chemical and Biological Engineering)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)December 2016© Lucie Solnickova, 2016AbstractCanadian bitumen is a plentiful source of hydrocarbons. However, to obtain oils whichmay be sold to consumers, bitumen must be upgraded. Among other processes, bitumenupgrading includes lowering sulphur content and correcting the carbon to hydrogen ratiomostly by carbon rejection, which results in the formation of petroleum coke (PC), a by-product which must be stored or disposed of.This study’s focus was the preparation of a molybdenum disulphide (MoS2) catalyst fora more facile removal of intercalated sulphur from bitumen, by synthesising nano-sized MoS2particles. Simultaneously, this study attempted to use PC as a catalyst support.Carbon-supported MoS2 catalysts were successfully prepared by two methods using am-monium tetrathiomolybdate: reverse micelles using the water/IGEPAL CO-520/cyclohexanesystem, and incipient wetness impregnation from ultra pure water. MoS2 prepared by im-pregnation was supported on PC, and MoS2 prepared by reverse micelles was supported onboth PC and activated carbon (AC). Catalysts prepared by reverse micelles contained nano-sized MoS2 with low stacking order, and the catalyst prepared by impregnation consisted oflong sheets of MoS2 with a higher stacking order. The catalysts were screened for hydrodesul-phurization activity in a novel slurry-phase batch microreactor using dibenzothiophene as amodel compound.The overall rate constant for DBT conversion per gram of molybdenum for the MoS2/PCprepared by impregnation was greater than that for the catalysts prepared by reverse micellesiiin the temperature range of 350− 375 ◦C. MoS2 supported on AC and PC showed a similaractivity toward catalysing the HDS of DBT when the MoS2 was prepared by reverse micelles;therefore, PC is a good alternative support to AC for MoS2/C catalysts prepared by thismethod.The rate constant associated with hydrogenation was an order of magnitude greater forthe catalyst prepared by impregnation than that for the catalysts prepared by reverse mi-celles. It was concluded that the larger stacking order in MoS2/PC prepared by impregnationprovided more sites for hydrogenation, which resulted in an overall larger rate constant thanthat for the catalysts prepared by reverse micelles, whose MoS2 stacking orders were minimaldue to the small particle size.iiiPrefaceThe work presented in this document was conducted in the department of Chemical andBiological Engineering in Vancouver at the University of British Columbia.The M.A.Sc. work was performed by Lucie Solnickova under the direct supervision of Dr.Kevin J. Smith in the same department. Lucie Solnickova was responsible for the design ofexperiment and its execution, including catalyst synthesis, characterization, activity testing,data collection, data analysis, kinetic modelling, and preparation of this thesis.Dr. Ross Kukard commissioned the micro batch reactor used for activity testing. HaiyanWang prepared the activated petcoke used in this study.ivTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xivNomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxDedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxii1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1.1 Introduction to bitumen upgrading . . . . . . . . . . . . . . . . . . . 11.1.2 Feed separation and bitumen extraction . . . . . . . . . . . . . . . . 21.1.3 Bitumen upgrading processes . . . . . . . . . . . . . . . . . . . . . . 31.1.4 Regulations for the sulphur content of fuels . . . . . . . . . . . . . . . 51.2 Current processes for more complete sulphur removal from bitumen . . . . . 6v1.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.2.2 HDS catalysts and mechanism . . . . . . . . . . . . . . . . . . . . . . 61.2.3 Particle morphology as an indicator for activity . . . . . . . . . . . . Rim-edge model . . . . . . . . . . . . . . . . . . . . . . . . Brim sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111.3 Role of support in catalysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 121.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121.3.2 Types of support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Activation of petcoke . . . . . . . . . . . . . . . . . . . . . . 131.3.3 Metal-support interaction . . . . . . . . . . . . . . . . . . . . . . . . 141.3.4 Metal loading considerations . . . . . . . . . . . . . . . . . . . . . . . 151.4 Preparation of nano-sized MoS2 particles . . . . . . . . . . . . . . . . . . . . 161.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161.4.2 Water-in-oil reverse micelles . . . . . . . . . . . . . . . . . . . . . . . 161.4.3 Particle nucleation and growth . . . . . . . . . . . . . . . . . . . . . . 171.5 Effects of water-in-oil emulsion preparation on particle size . . . . . . . . . . 181.5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181.5.2 Water to surfactant ratio . . . . . . . . . . . . . . . . . . . . . . . . . 191.5.3 Metal concentration . . . . . . . . . . . . . . . . . . . . . . . . . . . 201.5.4 Microemulsion temperature . . . . . . . . . . . . . . . . . . . . . . . 211.5.5 Precursor reduction and sulphidation . . . . . . . . . . . . . . . . . . 211.5.6 Supporting the metal particles . . . . . . . . . . . . . . . . . . . . . . 221.6 Catalytic activity of MoS2 nanoparticles . . . . . . . . . . . . . . . . . . . . 231.6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231.6.2 Effect of surfactant on the catalysis . . . . . . . . . . . . . . . . . . . 231.6.3 Model compounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24vi1.6.4 Reactors for research laboratory-scale activity testing . . . . . . . . . 251.7 Literature review summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251.8 Study approach and objectives . . . . . . . . . . . . . . . . . . . . . . . . . . 262 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272.1 Catalyst preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272.1.1 Carbon supports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272.1.2 MoS2 preparation using reverse micelles . . . . . . . . . . . . . . . . 272.1.3 Typical MoS2 preparation via reverse micelles . . . . . . . . . . . . . 292.1.4 Making a stable micelle . . . . . . . . . . . . . . . . . . . . . . . . . . 312.1.5 Mo salt selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332.1.6 MoS2 preparation by incipient wetness impregnation . . . . . . . . . 332.2 Catalyst characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342.2.1 X-Ray diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352.2.2 X-Ray photoelectron spectroscopy . . . . . . . . . . . . . . . . . . . . 352.2.3 Surface area measurements . . . . . . . . . . . . . . . . . . . . . . . . 362.2.4 Metal loading analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 372.2.5 Transmission electron microscopy . . . . . . . . . . . . . . . . . . . . 372.2.6 Thermogravimetric analysis . . . . . . . . . . . . . . . . . . . . . . . 382.3 Activity measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392.3.1 Gas chromatography-mass spectrometry . . . . . . . . . . . . . . . . 423 Catalyst Properties and Characterization Results . . . . . . . . . . . . . . 443.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443.2 TGA results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443.3 BET results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473.4 Metal loading results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48vii3.5 Surface composition and Mo coverage . . . . . . . . . . . . . . . . . . . . . . 513.6 Deconvoluted narrow XPS scans . . . . . . . . . . . . . . . . . . . . . . . . . 533.7 Chemical states of Mo in the catalysts . . . . . . . . . . . . . . . . . . . . . 573.8 Chemical states of S in the catalysts . . . . . . . . . . . . . . . . . . . . . . 593.9 XRD results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 603.10 TEM results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 683.11 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 724 HDS of DBT Reaction Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . 754.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 754.2 Reaction pathways of HDS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 764.3 Kinetic model development . . . . . . . . . . . . . . . . . . . . . . . . . . . . 774.4 Modelling results of reactions without catalyst . . . . . . . . . . . . . . . . . 794.5 Modelling results of the prepared MoS2 catalysts . . . . . . . . . . . . . . . 824.6 Error associated with mol%C . . . . . . . . . . . . . . . . . . . . . . . . . . 934.7 Discussion of k’ results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 944.8 Comparison of accessed selectivities . . . . . . . . . . . . . . . . . . . . . . . 964.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1045 Conclusions and Recommendations . . . . . . . . . . . . . . . . . . . . . . 1065.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1065.2 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110A Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124A.1 Feed calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124A.2 Calculations for catalysts prepared by the micelle method . . . . . . . . . . . 126viiiA.3 Calculations for catalysts prepared by the incipient wetness impregnationmethod . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128A.4 Calculation to convert Tye’s k data . . . . . . . . . . . . . . . . . . . . . . . 128B MATLAB Codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130B.1 Main body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130B.2 Modelmulti code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133B.3 ODEfunm code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134B.4 Calculation of Jacobian matrix . . . . . . . . . . . . . . . . . . . . . . . . . 134B.5 Least square code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135C Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143C.1 Petcoke activation procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . 143C.1.1 Obtaining petcoke with a certain particle size . . . . . . . . . . . . . 143C.1.2 Chemical activation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144C.1.3 Washing the petcoke . . . . . . . . . . . . . . . . . . . . . . . . . . . 144C.2 ICP-OES sample preparation . . . . . . . . . . . . . . . . . . . . . . . . . . 145C.2.1 Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145C.2.2 General cautions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146C.2.3 Required equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . 146C.2.4 Making the matrix solution . . . . . . . . . . . . . . . . . . . . . . . 147C.2.5 Digesting the sample . . . . . . . . . . . . . . . . . . . . . . . . . . . 147C.2.6 Preparing the standards . . . . . . . . . . . . . . . . . . . . . . . . . 149D Detailed Sample Characterization Data . . . . . . . . . . . . . . . . . . . . 151D.1 TEM micrographs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151D.2 ATTM certificates of analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 154ixD.3 Values inputted into the MATLAB model . . . . . . . . . . . . . . . . . . . 156D.4 Arrhenius parameters for the thermal and thermocouple-induced reactions . 161xList of Tables1.1 Atomic composition of Canadian bitumen . . . . . . . . . . . . . . . . . . . 22.1 Calculated mass fractions of stable micellar water/AOT/n-heptane systems . 322.2 The inorganic salts used as Mo sources for catalyst preparation . . . . . . . . 332.3 Preparation methods and naming of the catalysts screened for kinetic parameters 342.4 Binding energies for Mo 3d3/2 spectral lines for selected compounds . . . . . 363.1 BET surface area, pore volume, and pore size of the prepared catalysts usedfor activity testing, and of their supports . . . . . . . . . . . . . . . . . . . . 473.2 Summary of Mo sources, preparation methods, and metal loading of some ofthe prepared catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493.3 Mo loading and preparation methods of the catalysts screened for kineticparameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513.4 Surface composition of the prepared catalysts screened for kinetic parametersas measured by XPS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 533.5 Chemical states and their relative amounts of Mo in MoS2/PC-RM . . . . . 583.6 Chemical states and their relative amounts of Mo in MoS2/AC-RM . . . . . 583.7 Chemical states and their relative amounts of Mo in MoS2/PC-WI . . . . . . 583.8 Chemical states and their relative amounts of S in MoS2/PC-RM from twoidentically made batches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59xi3.9 Chemical states and their relative amounts of S in MoS2/AC-RM . . . . . . 593.10 Chemical states and their relative amounts of S in MoS2/PC-WI . . . . . . . 603.11 Particle lengths of MoS2 in MoS2/AC-RM and MoS2/PC-RM . . . . . . . . 694.1 Summary of preparation methods for the catalysts whose kinetic activitieswere studied . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 754.2 Estimated reaction rate constants for HDS of DBT for the thermal and thermocouple-induced reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 814.3 Estimated reaction rate constants for HDS of DBT catalysed by MoS2/AC-RM 864.4 Pre-exponential factors and activation barrier energies for the rate constantsassociated with HDS of DBT catalysed by MoS2/AC-RM . . . . . . . . . . . 874.5 Estimated reaction rate constants for HDS of DBT catalysed by MoS2/PC-RM 894.6 Pre-exponential factors and activation barrier energies for the rate constantsassociated with HDS of DBT catalysed by MoS2/PC-RM . . . . . . . . . . . 904.7 Estimated reaction rate constants for HDS of DBT catalysed by MoS2/PC-WI 924.8 Pre-exponential factors and activation barrier energies for the rate constantsassociated with HDS of DBT catalysed by MoS2/PC-WI . . . . . . . . . . . 934.9 Errors associated with presented mol%C . . . . . . . . . . . . . . . . . . . . 934.10 Comparison of k′ values at 350 ◦C . . . . . . . . . . . . . . . . . . . . . . . . 944.11 Comparison of k′ values at 340 ◦C . . . . . . . . . . . . . . . . . . . . . . . . 954.12 Summary of kinetic parameters of the three catalysts at 350 ◦C . . . . . . . 102A.1 Composition of feed used for activity testing . . . . . . . . . . . . . . . . . . 126D.1 Measured concentrations of products . . . . . . . . . . . . . . . . . . . . . . 156D.2 Concentrations of products with forced carbon balance, inputted into theMATLAB code for kinetic modelling . . . . . . . . . . . . . . . . . . . . . . 159xiiD.3 Pre-exponential factors and activation barrier energies for the rate constantsassociated with the thermal and thermocouple-induced reactions . . . . . . . 161xiiiList of Figures1.1 World energy statistics, total primary energy supply from 1971 to 2014 . . . 21.2 A proposed heteroatom-containing molecule in bitumen . . . . . . . . . . . . 31.3 Schematic of a typical HDS unit in a petroleum refinery plant . . . . . . . . 71.4 Proposed mechanism for the hydrodesulphurization of thiophene . . . . . . . 81.5 3D structure of MoS2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91.6 Rim/edge model of an MoS2 particle . . . . . . . . . . . . . . . . . . . . . . 101.7 Model for nanoparticle formation and growth in a reverse micelle . . . . . . 181.8 Radius of a micelle formed in a solution of water/AOT/n-heptane as a func-tion of water to surfactant ratio . . . . . . . . . . . . . . . . . . . . . . . . . 201.9 Frequently used model compounds to study HDS . . . . . . . . . . . . . . . 242.1 Molecular structure of surfactant AOT . . . . . . . . . . . . . . . . . . . . . 282.2 Molecular structure of surfactant IGEPAL CO-520 . . . . . . . . . . . . . . 282.3 Phase diagram of the water/AOT/n-heptane system . . . . . . . . . . . . . 322.4 Process flow diagram of the batch microreactor used for activity testing . . . 403.1 Thermolytic decomposition of surfactant AOT under N2 by TGA . . . . . . 453.2 Thermolytic decomposition of surfactant Igepal under N2 by TGA . . . . . . 463.3 Mo content of various catalysts as measured by XPS and ICP-OES . . . . . 523.4 XPS spectra of Mo 3d energy level of the catalysts . . . . . . . . . . . . . . 55xiv3.5 XPS spectra of S 2p energy level of the catalysts . . . . . . . . . . . . . . . . 563.6 XPS spectrum of petcoke before metal addition . . . . . . . . . . . . . . . . 573.7 XRD spectrum of the APTM salt . . . . . . . . . . . . . . . . . . . . . . . . 613.8 XRD spectrum of the ATTM salt . . . . . . . . . . . . . . . . . . . . . . . . 623.9 XRD spectra of the supports prior to Mo addition . . . . . . . . . . . . . . . 633.10 XRD spectrum of MoS2/AC prepared from APTM by the micelle method . . 643.11 XRD spectrum of MoS2/AC prepared from ATTM by the micelle method . . 653.12 XRD spectra of the MoS2/C catalysts screened for activity . . . . . . . . . . 663.13 XRD spectrum of MoS2/PC-RM, K2SO4 present . . . . . . . . . . . . . . . . 673.14 Particle length distribution of MoS2/AC-RM . . . . . . . . . . . . . . . . . . 693.15 TEM image of MoS2/AC-RM . . . . . . . . . . . . . . . . . . . . . . . . . . 703.16 Particle length distribution of MoS2/PC-RM . . . . . . . . . . . . . . . . . . 703.17 TEM image of MoS2/PC-RM . . . . . . . . . . . . . . . . . . . . . . . . . . 713.18 TEM image of MoS2/PC-WI . . . . . . . . . . . . . . . . . . . . . . . . . . . 714.1 Proposed pathways of the HDS reaction of DBT . . . . . . . . . . . . . . . . 764.2 Measured data and modelled fit for HDS of DBT product concentrations as afunction of time for the thermal and thermocouple-induced reactions . . . . 804.3 HDS of DBT rate as a function of temperature, for the thermal and thermocouple-induced reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 814.4 Measured data and modelled fit for product concentrations of the HDS ofDBT reaction catalysed by MoS2/AC-RM as a function of time . . . . . . . 854.5 HDS of DBT rate as a function of temperature, catalysed by MoS2/AC-RM 864.6 Measured data and modelled fit for product concentrations of the HDS ofDBT reaction catalysed by MoS2/PC-RM as a function of time . . . . . . . 884.7 HDS of DBT rate as a function of temperature, catalysed by MoS2/PC-RM . 89xv4.8 Measured data and modelled fit for product concentrations of the HDS ofDBT reaction catalysed by MoS2/PC-WI as a function of time . . . . . . . . 914.9 HDS of DBT rate as a function of temperature, catalysed by MoS2/PC-WI . 924.10 Selectivity of the HDS of DBT reaction catalysed by MoS2/AC-RM . . . . . 984.11 Selectivity of the HDS of DBT reaction catalysed by MoS2/PC-RM . . . . . 994.12 Selectivity of the HDS of DBT reaction catalysed by MoS2/PC-WI . . . . . 1004.13 Comparison between the three catalysts for selectivity toward the DDS route 101D.1 Whole TEM image of MoS2/AC-RM . . . . . . . . . . . . . . . . . . . . . . 152D.2 Whole TEM image of MoS2/PC-RM . . . . . . . . . . . . . . . . . . . . . . 153D.3 Whole TEM image of MoS2/PC-WI . . . . . . . . . . . . . . . . . . . . . . . 154D.4 Certificate of analysis for ATTM batch 1 . . . . . . . . . . . . . . . . . . . . 155D.5 Certificate of analysis for ATTM batch 2 . . . . . . . . . . . . . . . . . . . . 155xviNomenclatureRoman symbolsA Pre-exponential factor, used in the Arrhenius equation, in cm3 g−1Mo s−1Ea Activation barrier energy, in kJ mol−1k′x Rate constant parameter for a catalyst, in cm3 g−1Mo s−1kx Rate constant parameter for thermal and thermocouple-induced reactions, in s−1Mn Number average molecular weight, in g mol−1mol% Percentage by moles, a measure of concentrationmol%C of x Cx/sum, a molar percentage of x in the reaction product mixtureMoS2 Molybdenum disulphideP Pressure, in Pappm Parts per million, a measure of concentrationsum Σ (CDBT + CBP + CCHB + CTHDBT ), in µmol mg−1Thiolate C-S-H functional groupt Reaction time, in hours or minutesT Reaction temperature, in ◦Cxviivol.% Percentage by volume, a measure of concentrationwt.% Percentage by weight, a measure of concentrationX ConversionGreek symbolsβ Peak width at half maximum intensity, in radiansθ Angle of reflection, in ◦ or radiansω Water to surfactant ratioAbbreviationsAC Activated Carbon, a supportAOT Aerosol OT, a surfactantAPTM Ammonium Polythiomolybdate, a Mo saltATTM Ammonium Tetrathiomolybdate, a Mo saltBE Binding Energy, in eVBP BiphenylCHB CyclohexylbenzeneDBT DibenzothiopheneDDS Direct DesulphurizationDecalin DecahydronaphthaleneDFT Density Functional TheoryDPE 1,1-diphenylethyleneGC-MS Gas Chromatography-Mass SpectrometryHDS HydrodesulphurizationHG Hydrogenationmol%C of xxviiiICP-OES Inductively Coupled Plasma-Optical Emission SpectroscopyIgepal IGEPAL CO-520, a surfactantKOH Potassium hydroxide, a strong baseODE Ordinary Differential EquationPC Petroleum CokePDF Powder Diffraction FilePetcoke Petroleum CokeRM Reverse MicelleSTM Scanning Tunnelling MicroscopyTEM Transmission Electron MicroscopyTGA Thermogravimetric AnalysisTHDBT 1,2,3,4-tetrahydrodibenzothiopheneTOF Tunover FrequencyWI Wetness ImpregnationXPS X-Ray Photoelectron SpectroscopyXRD X-Ray Diffractionmol%C of xxixAcknowledgementsThe largest thanks are owed to the supervisor of this work, Dr. Kevin Smith, without whomnone of this would have been possible. Dr. Smith’s insightful guidance made this work moremeaningful, especially for the author.Thanks are also due to the examiners, Drs. Elo˝d Gyenge and Dusko Posarac.A big thank you to Dr. Ross Kukard for building such a great microreactor, and toMajed Alamoudi for training me in its use.Thank you, Scott Ryken, for the insightful discussions throughout the project, and crucialedits of this document.Dr. Mina Alyani, thank you for your friendship, advice, and help. Sharing an office withyou made my time in CHBE very enjoyable.Thank you to fellow Catalysis Group members for your friendship and general assistance:Haiyan Wang, Shida Liu, Chujie Zhu, Pooneh Ghasvareh, Dr. Ali Alzaid, Dr. RahmanGholami, Xu Zhao, and Hamad Almohamadi.Thank you to the technicians who assisted with data collection:Maureen Soon in Earth and Ocean Sciences for performing the ICP-OES mea-surements,Marshall Lapawa and Dr. Yun Ling in the Department of Chemistry for advisingwith the GC-MS measurements,xxLan Kato and Jenny Lai in Earth and Ocean Sciences for training me in the useof the powder XRD instrument,Dr. Ken Wong in the Advanced Materials & Process Engineering Laboratory forperforming the XPS measurements,Bradford Ross in the Bio Imaging Facility for training me in the use of theTunnelling Electron Microscope and providing assistance as necessary.Thank you to all the support staff in Chemical and Biological Engineering, whose con-tribution to research is not insignificant.Additionally, I would like to acknowledge the funding received from NSERC and ShellCanada which financially enabled this research.Finally, my wonderful family, thank you for your continued support and love throughoutall these years, and for the opportunities you made possible.xxiThis thesis is dedicated to all family, friends,and mentors who helped me to reach this point.I could not have done this without your support.xxiiChapter 1Introduction1.1 Overview1.1.1 Introduction to bitumen upgradingHydrocarbons and coal are currently the world’s leading source of energy, and extrapolatingfrom the trend in Figure 1.1, the world’s energy supply will remain fossil fuel-based for theforeseeable future. After Venezuela and Saudi Arabia, Canada has the largest oil reserves inthe world, with proven reserves totalling 166 billion barrels as of 2014. It is expected thatthe oil-based fossil fuels will continue to be produced in regions with high reserves. Canadianoil reserves are in the form of bitumen in oil sands [1].Bitumen is a very complex mixture of hydrocarbons and heteroatoms. Compared withconventional offshore crude oil, bitumen has many more impurities, in the form of heavymetals and intercalated heteroatoms (S, O, N), as well as a higher C:H ratio. Table 1.1summarises the average atomic content of Canadian bitumen, and Figure 1.2 is an exampleof a hypothesised typical heteroatom-containing bitumen molecule. The exact bitumen com-position highly depends on the bitumen’s source and its processing method. The processof refining bitumen into intermediate crude oil includes the feed separation, upgrading, and1Figure 1.1: World energy statistics, total primary energy supply from 1971 to 2014, En-ergy demand (in megatonne of oil equivalent) on the ordinate, time in years on theabscissa. ©OECD/IEA 2016 Key World Energy Statistics, IEA Publishing. of impurities [2–5]. These steps are outlined in more detail below.Table 1.1: Atomic composition of Canadian bitumen [6]AtomAverage contentin bitumen (wt.%)C 83.0H 10.2S 5.1O 1.5N Feed separation and bitumen extractionCanadian bitumen as mined is a mixture of sand, water, and bitumen. The mixture isfirst slurried with hot water, causing the bitumen to separate as a froth which includes some2Figure 1.2: A proposed heteroatom-containing molecule in bitumen. Reprinted with permis-sion from [7]. Copyright (2004) American Chemical Societysolids and water. The froth is diluted with naphtha which allows the bitumen to be extractedby centrifugation and subsequently be sent for upgrading [4]. Liquid with an intermediatedensity is processed in smaller centrifuge separation vessels in order to extract bitumen fromthese middlings, and the bottom dense material is sent to a settling basin [4].1.1.3 Bitumen upgrading processesBitumen containing naphtha diluent is first distilled to remove the naturally-occurring lightgas oil components (boiling point of <535 ◦C) as well as the naphtha, which is recycled. Lightgas oil is sent directly for hydrotreating, a process which converts the distillate into a feedwith the correct specifications for jet fuel, diesel, and kerosene [5]. The heavier componentswhich remain even after vacuum distillation, termed residue or heavy gas oil, need to beupgraded [3, 4].Upgrading the residue is crucial, since about half the bitumen by volume is not light3gas oil. To convert the heavy gas oil into a mixture with specifications similar to those ofoffshore crude oil, processes to break the large molecules into components with a low boilingpoint that allows distillation, are employed [3–5].The residue stream contains a larger concentration of heteroatoms and metals than freshlyextracted bitumen, since only hydrocarbons have been removed thus far. Additionally, theresidue contains many compounds with unsaturated hydrocarbons and aromatics which needto have their carbon to hydrogen ratio corrected to reduce the viscosity and boiling pointof the feed into the specification range acceptable for diesel and gasoline. The C:H ratio isimproved by hydroprocessing, either by C rejection or H addition, also called hydroconversion[5].Hydroprocessing is performed at high temperatures, where the thermal cracking of largemolecules occurs radically − a beneficial side-effect which increases the concentration of lightoil products. If enough H2 is present in the system during hydroprocessing, this process isbasically hydroconversion. Light components produced by cracking are combined with lightcomponents from the initial distillation for hydrotreating [5].However, if the system is starved of H2 during hydroprocessing, the temperature is highenough (above 400 ◦C), or the residence time is long, a coking reaction significantly competeswith hydrogenation and the hydrocarbon radicals condense to petroleum coke. This solidcoke by-product is often derived from the very reactive asphaltene components of bitumen,defined as compounds insoluble in n-pentane or n-heptane. Many asphaltene moleculescontain at least one heteroatom, so removing asphaltene as coke also reduces the S, O, Ncontent of the residue. The exact characteristics of petcoke depend on the way it was made:delayed coke (made in a coking drum) contains more fixed C than fluid coke (made in afluidized bed), while fluid coke has a higher weight content of S, O and N [8]. Coke makesup to 20 wt.% of the final products derived from bitumen and is often seen as a low-valueby-product which must be stored or disposed of [9].4The bitumen upgrading process outlined above reduces the heteroatom content signifi-cantly, with the S content after upgrading at approximately 0.5 wt.% [3]. However, this Sconcentration is still too large for reasons discussed below, and the bitumen needs furtherupgrading to remove more S.1.1.4 Regulations for the sulphur content of fuelsS-containing molecules are corrosive to metals, and high-S containing feeds can cause damageto processing equipment [10, 11]. Additionally, costly noble metal catalysts which are useddown-stream in the refinery for hydrotreating have very low resistance to S poisoning [12].Furthermore, specifications for the maximum allowable S content in fuels sold to con-sumers have decreased by several orders of magnitude: from 500 ppm to 15 ppm allowableS concentration in the last decade [13]. Recently, the Government of Canada announcedthat new regulations will allow less than 12 ppm (0.0012 wt.%) S concentration starting in2020 [14], meaning that further innovation needs to occur in order to implement technologycapable of reducing the S content to such low levels.It should be noted that Canada is not alone in regulating S content of oil− the regulationsare similar in many countries. For example, the European Union’s S content regulationallows less than 50 ppm, and fuel with a S content of less than 10 ppm must be available forconsumers to purchase [15].This study focused on preparing a highly active catalyst for a more facile removal ofsulphur from bitumen.51.2 Current processes for more complete sulphurremoval from bitumen1.2.1 IntroductionThe S content of bitumen at 0.5 wt.% after initial upgrading is still unacceptably high forthe equipment and noble catalysts downstream in the refinery, and for the permissible Scontent in oil regulations set out by governments. Presently, the technology most-oftenused to obtain fuels at or below the maximum allowable S-content limit is hydroprocessing,specifically hydrodesulphurization (HDS).The initial MoS2/Al2O3 catalysts developed for HDS deactivate quickly. However, con-tinued development of the catalyst drove its improvement, and new catalysts have longerlife-spans and are able to achieve deeper (more complete) desulphurization [16]. Current re-search is aimed at deeper HDS at lower temperatures, using more active and more selectivecatalysts. An improved catalyst would enable refineries to keep their current equipment andsimply replace the catalyst for an improved one in the hydrotreater, avoiding the high costsassociated with a complete shut-down and refitting to accommodate a different desulphur-ization technology.1.2.2 HDS catalysts and mechanismIndustrially, HDS occurs in a dedicated unit, essentially a hydrotreater, by the catalyticaddition of hydrogen. During the process, the actual reaction occurs when the feed is passedthrough a fixed-bed reactor housing a metal sulphide catalyst under a H2 atmosphere.A process diagram of an HDS unit in a petroleum refinery is pictured in Figure 1.3. Tostart, a pump delivers the feed which is then combined with a H2 stream. The mixtureis preheated, and then further heated in a fired heater to vaporize the components of the6feed. This process is very energy intensive, since many of the molecules in the feed havehigh molecular weights and therefore have high boiling points. The now-vaporized streampasses through the fixed-bed reactor which itself is heated to 300− 400 ◦C and pressurized to3− 13 MPa. Much energy is required for this step also, since the reactor must be heated tokeep all components vaporized. The hot product stream passes through a heat exchanger topreheat the incoming feed while itself being cooled. The product stream must pass throughan additional cooler so that it may be separated in the gas separator into two phases: a H2rich gas, and an oil-rich liquid. Most of the unreacted H2 is separated from the gas streamand recycled into the early part of the cycle. The remaining sour gas, termed so for its highH2S content, is sent for further treatment. The liquid stream is sent to a distillation column(stripper) where the final bottom products are the desired desulphurized oil products. Thetop gas stream from the stripper contains light C1 to C4 hydrocarbons as well as some S-containing molecules. The gas stream is condensed into sour water, and the still-gaseousstream joins the sour gas for processing [17, 18].Figure 1.3: Schematic of a typical HDS unit in a petroleum refinery plant. Used under acreative commons license from [19]7Currently the most commonly used HDS catalyst is a NiMoS/Al2O3 catalyst, where Niis a promoter; Co can be used as a promoter in place of Ni. Although unsupported MoS2catalysts may be used, unsupported catalysts agglomerate faster than supported catalysts;agglomeration decreases the area on which the reaction occurs and consequently the rate ofreaction is reduced. The NiMoS catalyst precursor is prepared by impregnating mesoporousAl2O3 with aqueous Mo and Ni solutions, either sequentially or in tandem, followed bycalcining. The procedure yields an oxide which must be sulphided in situ prior to the HDSreaction. This preparation method is fast and easy, an important trait due to frequency withwhich catalyst is prepared. However, the method provides little control over the size of thecatalyst particles [20].Figure 1.4: Proposed mechanism for the hydrodesulphurization of thiophene which is hydro-genated to 2,5-dihydrothiophene before the release of the final product, butadiene. Figureadapted by author from [21]In the actual HDS reaction, organosulphur molecules are removed by reaction with H2over an MoS2 catalyst to form H2S and hydrocarbons. A proposed mechanism for HDS isoutlined in Figure 1.4. Desulphurization begins when a S atom intercalated in the oil moleculebinds to the Mo in the vacancy site where a S would have been, through Mo-S σ and pi bonds.There is electron back-donation into the thiophenic pi* antibonding orbitals, which weakens8the C-S bond [22]. Then two H atoms coordinated to S atoms in the catalyst migrate tothe substrate, weakening the S-C bonds which are then cleaved. The desulphurized productis released, while the S remains coordinated to the metal. The next step is the dissociativeaddition of H2, and subsequent release of S, in the form of H2S, from the Mo to regeneratethe vacant active site on the metal [23]. To complete the catalytic cycle, the S sites arerehydrogenated by the catalytic dissociation of H2.1.2.3 Particle morphology as an indicator for activityMoS2 exists as sheets, pictured in Figure 1.5. However, HDS does not occur at basal sites,since basal S ions are bonded too strongly to be removed and create a vacancy near the Moatom for C-S bond activation to occur [24, 25]. For unpromoted MoS2, the activity at edgesites is greater than at the corner sites [26].Figure 1.5: 3D structure of MoS2, Mo atoms in purple, S atoms in yellow. Adapted bypermission from Macmillan Publishers Ltd: Nature Communications ([27]), copyright (2014) Rim-edge modelHDS occurs only on rim and edge sites of the MoS2 particle since that is where the vacanciesare located (Figure 1.6). According to the Daage-Chianelli theory, it is the metal ions in atetrahedral site which are more active for HDS, rather than the octahedral sites. By thisreasoning, HDS is a structure-sensitive reaction, and hence turnover frequency is dependent9on particle size, with smaller particles offering more edge and rim sites per mass than largerparticles [22, 28]. Therefore, creating nano-sized MoS2 catalyst particles, rather than sheets,introduces more of these edge and rim sites per catalyst mass to catalyse the reactions,allowing the use of a smaller amount of catalyst to maintain the same conversion. WhenMoS2 crystallites are in the 1− 10 nm size range, the proportion of surface sites in variouscoordination numbers and molecular shapes changes noticeably with size and geometry [22].Furthermore, depending on the shape of the MoS2 nano-crystallite, different selectivitiesmay be accessed. This has also been explained by the rim-edge model which proposes thathydrogenation (HG) occurs exclusively on top and bottom (rim) perimeter sites, while directdesulphurization (DDS) by hydrogenolysis is accessed at the edge sites of the MoS2 crystallite[23, 29]. Therefore, it is theoretically possible to change the shape of the crystallite to favourmore edge or rim sites, depending on the treatment necessary: HG for highly aromatic feeds,DDS for very sour feeds.Figure 1.6: Rim/edge model of an MoS2 catalyst particle, some sites omitted for clarity.Figure adapted by author from [30] Brim sitesTo add on to the rim-edge model and the mechanism outlined above, Topsøe and co-workerssuggested that HDS activity can occur on an equilibrium MoS2 edge without the creation ofa vacancy at a Mo atom due to a “metallic brim site.” These metallic brim edge sites wereobserved by Scanning Tunnelling Microscopy (STM) and calculated by Density FunctionalTheory (DFT) as metallic in contrast to the basal plane of the fully sulphided Mo edge[31, 32]. The brim site is only exposed past the top layer of a multi-stacked particle, andnot observed in single-slab structures [33]. The authors claimed that the brim sites couldbind thiophene and be involved in HG reactions without the creation of a coordinativelyunsaturated Mo site.However, an earlier publication from the same research group stated that thiopheneadsorption on these brim sites was only observed below −73 ◦C by DFT, above which themolecules dissociate before finally concluding that the brim sites were likely inconsequentialto HDS catalysis [34]. Later, the group claimed that when H is present on sites adjacent to theweakly bound thiophene as S-H, both HG and ring-opening may occur [33]. This publicationfurther claimed that brim sites are free of steric-hindrance effects and allow the adsorptionof bulky atoms easier than non-brim sites. Additionally, the researcher claimed that brimsites are poisoned by basic N-containing substrates but unaffected by H2S, explaining whythe HG route is not suppressed by high H2S concentrations [33].Clearly this theory is still in the progress of being developed and tested, and should notbe overlooked.111.3 Role of support in catalysis1.3.1 IntroductionIn general, unsupported catalysts agglomerate faster than supported catalysts. Agglomera-tion lowers the surface area of the catalyst in the reaction mixture and results in decreasedactivity [35]. Therefore, supported catalysts offer one less deactivation mechanism to over-come.Furthermore, the support anchors and stabilises the nano-sized catalyst particles on itssurface, and limits sintering at reaction temperatures to ensure that the surface area of thecatalyst particles remains high. However, the support is not involved in creating active siteson the particle [26].1.3.2 Types of supportThe selection of a support material for the catalyst is crucial, since the metal-support in-teraction can impact the activity of the catalyst in several ways, such as the polarity ofthe support-metal bond, support-precursor interaction, and the sulphidability of the oxidicprecursor. These factors are discussed in detail below.A variety of oxidic supports, such as Al2O3, SiO2, TiO2 and zeolites are available; however,a particularly interesting support is carbon. Of special interest in this project is petroleumcoke (petcoke), a C source produced in great quantity as a by-product of hydroprocessing ofoil sands bitumen − an estimated 77 million tonnes of petcoke were stockpiled in Albertaby 2013, and approximately 4 million tonnes are created each year [8].Successfully replacing the support for the MoS2 HDS catalyst with petcoke at the siteof its creation would reduce the costs associated both with storing the petcoke, and thepurchasing and transportation costs of the material it would replace. However, petcoke hasvery low surface area and porosity, so it would first need to be activated to mesoporous12petcoke before being used as a catalyst support. Mesopores (pores with a diameter between2 − 50 nm) are essential for an HDS catalyst support in order to accommodate the MoS2crystallite, and to reduce the diffusion effects related to the large molecules which need todiffuse in to reach the MoS2 in order to be desulphurized [36, 37]. Activation of petcokeThe preparation of activated carbon consists of two steps: carbonization (heating in inertgas to ∼ 800 ◦C), followed by either physical or chemical activation. Petcoke can be similarlyactivated. In physical activation, steam or CO2 increases the porosity of the C material bygasification at ∼ 1000 ◦C [38]. In chemical activation, the coke is combined with a chemicalspecies (usually KOH, though NaOH, HNO3, H3PO4, or ZnCl2 can be used), and heatedto 450− 900 ◦C followed by subsequent washes to remove and/or neutralize the activatingspecies [8]. Practically chemical activation is faster, but has more costs associated with itdue to the waste stream created during washing [8]. On a research laboratory scale, chemicalactivation is more practical, due to the equipment necessary to inject steam or CO2 into agasifier needed for physical activation.Liang and co-workers reported preparing chemically activated petcoke with a surfacearea of 3234 m2/g and a pore volume of 1.78 cm3/g by mixing coke with an excess of KOHand heating at 900 ◦C followed by subsequent washing. The average pore size was 2.2 nm;however, the authors do not specify how much of the pore volume was due to mesopores[39].Chen and co-workers reported activating petcoke also by chemical activation but usedmicrowaves rather than conventional heating. They affected the activation of both delayedand fluid coke, and found that delayed coke activated by this method had a larger surfacearea (1131 m2/g vs. 440 m2/g), and more total pore volume (0.4795 cm3/g vs. 0.2247 cm3/g)than activated fluid coke. However, the fluid coke yielded a larger proportion of mesopores13(0.0842 cm3/g accounting for 37% of pores), than delayed coke (0.0571 cm3/g, accountingfor 12% of total pores) [40]. The authors related the differences in surface area and porevolume to the physical properties of the raw cokes: delayed coke is sponge-like enabling theKOH to spread through the internal pores of the raw coke, whereas fluid coke is more hardand smooth, making it difficult for KOH to penetrate to the core of each coke particle [40].Rambabu and co-workers reported a successful physical activation of fluid petcoke, af-fected by carbonizing the coke at 900 ◦C and subsequently injecting steam. Optimised foractivation time of 9 hours and a water flow rate of 15.0 g/h, the method’s yield was 35%.The obtained activated petcoke had a surface area of 482 m2/g and a total pore volume of0.231 cm3/g of which 45% were mesopores [41].A combined physical/chemical activation was related by Virala and co-workers whomactivated delayed petcoke first by NaOH or KOH at 800 ◦C and later by humidified N2 againat 800 ◦C without an intermediate washing of the coke. The activated petcoke with thelargest mesopore volume was prepared by heating a coke/KOH mixture for 2 hours in dryN2, followed by heating for 2 hours in humidified N2. The activated petcoke had a surfacearea of 2449 m2/g, and pore volume of 1.54 cm3/g of which 40% were mesopores; however,the method’s yield was only 2%. The activated petcoke with the best compromise betweenyield (27%) and mesopore volume (0.61 cm3/g of which 51% were mesopores) was preparedby heating petcoke with NaOH for 2 hours in dry N2, followed by heating for 30 minutes inhumidified N2 [42].1.3.3 Metal-support interactionIt has been demonstrated that one of the parameters which determine the activity of MoS2toward HDS is the covalent character of the catalyst’s Mo-S bond [43]. When MoS2 issupported on an oxidic material, Mo can bind to O on the support which decreases theelectron density at the metal centre. Reduced electron density increases the polarity of the14Mo-S bond, decreasing its covalency and consequently the HDS activity is lowered [43, 44].The C-Mo bond is less polarized than the corresponding Mo-O-Al interaction on Al2O3, orthe Mo-O-Si link on SiO2 [44].The magnitude of the activity decrease is proportional to the number of OH sites presenton the surface of the support: Al2O3 has the most, SiO2 has fewer, and C has very few OHsites [44]. Furthermore, Kibsgaard and co-workers found that the MoS2 particles supportedon oxidic supports are attached to the O through the particle’s edge sites, hence blockingthe very sites active for HDS as discussed above [45]. Therefore, supporting MoS2 on C asopposed to Al2O3 or SiO2 should result in an increase in HDS activity.The polarity of the support-metal bond determines the oxidation state of the metal.Additionally, the oxidation state of the metal in the oxidic precursor MoO3 dictates the easewith which the oxide can be sulphided, and since MoS2 not MoO3 is active for HDS [30],the ease of sulphidation directly impacts the catalyst’s activity; MoO3/C is more readilysulphided than MoO3/Al2O3 and MoO3/SiO2 [44, 46].A final point regarding the use of Al2O3 for HDS catalyst support to consider is thesolubility of Ni and Co (common promoters) in the support. For instance, Arteaga andco-workers reported a decrease of HDS activity during a CoMoS/Al2O3 catalysed reaction.They attributed the decrease of activity to the decrease of Co atoms at the surface dueto their diffusion into the support and formation of bulk CoAl2O4, a catalytically inactivespecies which is not sulphidable [47]. The diffusion effect is absent when C is used as asupport, since C does not solubilize the promoter [48].1.3.4 Metal loading considerationsMany factors are responsible for effecting catalytic activity, including the metal’s oxidationstate, MoS2 particle size and shape, atom packing, inter-particle metal-metal bonding, andthe aforementioned particle-support interaction [35, 49]. Some of these factors (atom pack-15ing, inter-particle metal-metal bonding) are due to metal loading. For example, as the Moloading increases, the MoS2 particles are inherently more closely spaced, there are moremetal-metal bonding interactions, and activity is decreased [50]. For Mo/C catalysts, amaximum conversion exists at approximately 25 wt.% Mo loading [51].Since HDS is a structure-sensitive reaction, the MoS2 particle size also has interestingimplications. As the particle size decreases the shape of the crystallite changes, and thenumber of corners, edges, and rim sites that its surface offers increases [35]. The particle’ssurface appearance impacts the activity directly by the nature of active site offered, whichaffects the oxidation state of the metal at that site.1.4 Preparation of nano-sized MoS2 particles1.4.1 IntroductionFor the past three decades, there has been a wide interest in the synthesis of nano-sizedmetal particles for application to catalysis. These nanoparticles have often been made usingthe reverse micelle method. The research has tended to focus on making the particles anddeducing the variables responsible for the particle’s size. Many research groups have used thesynthesised particles to catalyse reactions, after initial reports of nano-sized heterogeneouscatalysts being more selective than traditionally prepared catalysts with larger particle sizes[52].1.4.2 Water-in-oil reverse micellesAs reported extensively since the 1980’s, the size of an MoS2 particle may be controlledby synthesising it in a reverse micelle. Each micelle consists of a small droplet of watersurrounded by a layer of surfactant and suspended in an oil phase. The hydrophilic “heads”16of the surfactant molecules point inward to the water pool, while the lipophilic “tails” ofthe surfactant are oriented into the oil phase. A water-soluble metal salt may be dissolvedin the internal pool of water, and the micelle can act as a nano-reactor to facilitate thereduction and/or sulphidation of the salt. A reverse micelle system is an optically-clear,thermodynamically-stable solution which forms spontaneously.The micelle size can be controlled by changing the water to surfactant ratio, selection ofthe surfactant and oil phase, concentration of the metal salt inside the water pool, reductionand/or sulphidation conditions, and many other factors [53–59].1.4.3 Particle nucleation and growthTo isolate the metal particles from the microemulsion, the Mo must be reduced and precip-itated. The two steps have been demonstrated separately where the metal is first reducedfollowed by the addition of a precipitating agent, or can occur simultaneously [57]. Thechange of oxidation state either causes the particle to grow beyond its micelle’s size, orcauses its solubility to change such that it is no longer soluble in water. In either case, theparticle precipitates out of solution and may thusly be isolated. Among other factors, theaverage particle size is determined by the size of the micelle from which it was precipitated[53–59].A mechanism for particle formation from micelles is based on a statistical approach,rooted in the following assumptions [54, 55, 60, 61]:1. Metal ions are distributed throughout the micelles of the microemulsion according toa Poisson distribution,2. A micelle must contain a set minimum number of ions in order for a stable nucleus toform,3. Nucleation is much slower than growth,174. Material exchange between micelles occurs when two water cores coalesce to form atransient dimer which then decoalesces, leaving each core with a randomised solutecontent.Upon the introduction of a reducing agent, nucleation occurs only in the micelles whichmeet the criteria for a stable nucleus. As the reduction proceeds, micelles exchange materialamongst themselves by continuously coalescing and decoalescing. The unreduced Mo ionsintroduced into a micelle containing a reduced nucleus are reduced and incorporated intothe nucleus, making it larger (see Figure 1.7). This process continues until either a criticalnucleus size with a critical surface tension is reached which prevents further material exchangebetween micelles, or all Mo ions are reduced. In essence, the number of nuclei formed at theearly nucleation stage determines the total number of particles which will form [55].Figure 1.7: Model for nanoparticle formation and growth in a reverse micelle. Adapted bypermission from Macmillan Publishers Ltd: Nature Chemistry ([62]), copyright (2010)1.5 Effects of water-in-oil emulsion preparationon particle size1.5.1 IntroductionMicroemulsion systems contain many variables, each of which may change the particle’s finalsize or the particle size distribution. Extensive studies of these variables and their effects onthe particle size have been carried out by many groups [53–59]. The average particle size is18a function of the size of the micelles from which the metal was precipitated; however, theexact relationship is a complicated one.1.5.2 Water to surfactant ratioThe size of the micelle can be controlled by changing the ratio of water to surfactant, ω.The larger the ω, the larger the micelles formed, as shown in Figure 1.8 for a reverse micellesystem of water, surfactant AOT, and oil phase n-heptane. The linear relationship can beexplained by the fact that only a certain number of micelles can be formed from a set amountof surfactant. Adding more water molecules to the solution means that each existing micellemust swell in order to accommodate the water since no new micelles can form. The swellingwill only continue to a point, the critical micelle concentration, after which if more water isadded the microemulsion breaks and two phases form [61, 63].As discussed above, the critical nucleus size depends on the original size of the micelle,which in turn is set by ω. Therefore, the larger the ω, the larger will be the precipitatedparticles [56]. However, there are many more factors which influence the final particle sizethan ω.19Figure 1.8: Radius of a micelle formed in a solution of water/AOT/n-heptane as a functionof water to surfactant ratio, ω. Line added for clarity and is not a modelled fit. Plottedfrom data reported in [64]1.5.3 Metal concentrationThe metal salt concentration inside the water pool is limited by the salt’s solubility in water.Calculating the average salt concentration in the micelle is made more complicated by thefact that one is unable to truly account for all the water inside the water pool, as some watergoes toward hydrating the surfactant chains, which is discussed further below [60, 65].The salt concentration inside the water pool was reported to be a factor which impacts thesize of the isolated particles. Fischer and co-workers reported being able to vary the diameter20of Co3O4/Al2O3 particles which maintain their size even after reduction with ammonia; themicelle sizes were changed by varying ω as well as the salt concentration in the water pool[66]. A lower concentration of metal salt inside the micelle resulted in a smaller averageCo3O4 diameter [66].1.5.4 Microemulsion temperatureThe temperature at which the microemulsion is prepared influences the micelle size inversely.At constant water and surfactant concentration, the micellar diameter decreases as temper-ature increases, with the minimum diameter occurring at ∼ 2 ◦C below the temperature ofphase separation. This effect was argued to be due to hydration of the lipophilic chains ofthe surfactant, the magnitude of which increases with temperature. At higher temperatures,more water molecules hydrate the surfactant chains, leaving fewer inside the micelle andthus lowering the micellar diameter [65, 67].However, during the reduction of the salt inside the water pool, the particle size de-pendence on temperature is reversed. Reductions performed at lower temperatures tend toyield smaller crystallites, if all conditions other than temperature are constant. This was ex-plained by the decreasing oil phase viscosity with increasing temperature − oil phase whichis more viscous will result in more nuclei being formed. Thus, fewer nuclei and thereforelarger particles are observed when the reduction is affected at higher temperatures, whilereductions performed at lower temperatures give smaller crystallites [56, 67].1.5.5 Precursor reduction and sulphidationAs discussed above, the metal inside the micelle must be reduced and precipitated in orderfor the particle to be isolated. Furthermore, current industrial HDS catalyst preparationmethods use MoO3 precursors which must be sulphided since it is the MoS2 phase which is21active for HDS catalysis [30]. The reduction, precipitation, and sulphidation processes havea large impact on the final MoS2 particle size; some researchers have attempted to combinetwo or three of these processes to reduce the impact on particle size.Reduction of the Mo salt in the micelle has been reported by bubbling H2 or N2 throughthe microemulsion; however, due to particle agglomeration, large and irregularly-sized par-ticles precipitated as a result of this method [53, 68, 69]. Using 5% H2S in H2 to affect thesimultaneous in situ reduction and sulphidation at room temperature was also reported toproduce large agglomerated bulk particles [70].Some researchers affected the reduction/sulphidation with liquid agents such as N2H4,LiBH4, NaBH4, Li2S, (NH4)2CO3, or NH4OH instead of the gaseous H2S or H2. It wasreported that the amount of liquid reducing agent added to the microemulsion impacted theaverage particle size [54, 58]. Boutonnet and co-workers reported that the catalyst whosemetal was reduced by N2H4 was significantly more active, by an order of magnitude, thanthe catalyst prepared by reduction using H2 [69].Other researchers avoid the oxide to sulphide transformation, which can lead to incom-plete sulphidation or agglomeration of the particles, by utilising a Mo salt which is alreadya sulphide. Most often, this salt is ammonium tetrathiomolybdate (ATTM) which con-tains [MoS4]2−. Researchers have reported making a catalyst more active for HDS than aconventional catalyst by using the ATTM salt in place of an oxidic salt [71–73].1.5.6 Supporting the metal particlesFischer and co-workers reported several different methods of supporting Co3O4 particles, andthe large impact this procedure has on the particle size and its distribution. They found thatthe smallest particles resulted when the Co salt was first reduced for thirty minutes insidethe micelle with an aqueous reducing agent but not isolated from the emulsion, followed bythe addition of Al2O3 support and the destabilization of the emulsion [74].22Boutonnet and co-workers reported obtaining nano-sized Pt/pumice particles with a nar-row size distribution by initially reducing the Pt salt in the microemulsion with aqueousN2H4, and subsequently adding the support into the vessel and heating the solution to 60 or85 ◦C to remove the oil phase [52].1.6 Catalytic activity of MoS2 nanoparticles1.6.1 IntroductionWhile early research focused on the effect of the variables discussed above on the size ofthe particles isolated from the micelles, much research lately has focused on using thesenanoparticles to catalyse reactions.1.6.2 Effect of surfactant on the catalysisSome researchers have reported using the catalysts directly after separating them out ofthe rest of the solution with no adverse effects [52, 57]. However, others related that thesurfactant has a poisoning effect on the reaction being catalysed [70, 74–76].It has been suggested that not only does the surfactant poison the MoS2 catalyst outright,but that its thermal decomposition products scavenge H atoms from the solvent, reducingthe solvent’s efficacy as a hydrogen shuttle [76]. For instance, AOT begins decomposingat 200 ◦C [77], and the typical HDS reaction temperature is 350− 400 ◦C. Furthermore,the proposed mechanism for HDS suggests that the reaction starts by a C-S bond scission,followed by H transfer to a thiolate intermediate; thus, the concentration of H2 in the reactionmixture effects reactivity [78]. Therefore, it seems that whether or not the surfactant willpoison the reaction depends on the reaction mechanism; HDS is poisoned by surfactantremaining in the reaction mixture. To prevent the HDS reaction from being poisoned, it is23best to remove the surfactant and its constituents prior to reaction.Additionally, the boiling point of the surfactant is such that it begins to sublime undervacuum, precluding characterization methods performed at high vacuum, such as BET,TEM, and XPS.The drawback to heating metal catalysts is that elevated temperatures enable speciesto migrate, resulting in particle agglomeration or sintering. It was reported that sulphidingMoS2/Al2O3 at temperatures below 397◦C preserves the catalyst’s dispersion [44].1.6.3 Model compoundsIn lieu of using a real oil feed to study the activity of a novel MoS2 catalyst toward HDS,researchers often use thiophene to greatly simplify the reaction mixture. Many groups preferto use dibenzothiophene (DBT) or its substituted version 4,6-dimethyldibenzothiophene (4,6-DMDBT) as sterically-hindered S-containing molecules resistant to hydroprocessing (seeFigure 1.9 for the structures of these compounds). Often these model compounds are usedas the feed stock in search for improved HDS catalysts [79–81].Figure 1.9: Frequently used model compounds to study HDS. a) Thiophene; b) DBT; c)4,6-DMDBTThough a model compound does not accurately display the full variety of complexmolecules in an oil feedstock, it is postulated that the catalyst most successful at desulphur-izing a model compound will translate to the best candidate for treating a more complex24mixture. Using a model compound eases the deconvolution of the results of catalytic tests,especially in terms of selectivity.1.6.4 Reactors for research laboratory-scale activity testingIndustrially, most hydrotreating processes are either trickle flow or gas-phase processes.These reactor systems are generally too large to be viable in laboratories, and researchersoften use the equipment already available in their laboratories for catalyst testing. Some havereported catalytic testing of nanoparticles in batch reactors [82, 83], and some in continuousflow reactors [84, 85].No matter the regime, microreactors are often used in research to test the activity ofnovel catalysts, or to deduce mechanisms of reactions. The use of microreactors is advan-tageous since these small-scale reactors require the use of less catalyst and substrate, andconsequently less waste is generated. Furthermore, it is easier to fully control the reactionconditions in a microreactor than in a full scale one. The disadvantages include a large degreeof error inherent to any experiment dealing with small volumes, and frequent down-timesfor cleaning and maintenance [84, 86, 87].1.7 Literature review summaryFossil fuel-based oils will continue to be prominent in the near future. As the concentrationof S in feedstocks continuously increases, improved catalysts with higher HDS activities willneed to be implemented to meet S content regulations. Using C as a support increases theactivity of supported MoS2 catalysts toward HDS. Utilising petcoke, a low-value by-productof bitumen upgrading, as a catalyst support for bitumen upgrading would help the oilsandsbe more sustainable. Using reverse micelles to synthesise MoS2 catalysts offers control overparticle size by carefully controlling preparation conditions. Control over geometry, and25therefore selectivity, is also possible. Catalysts need to be fully cleansed of surfactant priorto being screened for HDS activity. Microreactors utilising DBT as a model compound areeffective to test catalysts for HDS activity.1.8 Study approach and objectivesThis work sought to prepare a highly active, petcoke-supported MoS2 catalyst for the HDSof the model compound DBT. A series of MoS2/C catalysts were prepared and comparedquantitatively in terms of their activity toward affecting the HDS of DBT reaction.Two MoS2/C catalysts were prepared using the reverse micelle method in the hopes ofsynthesising nano-sized particles − one on activated carbon and one on activated petcoke inorder to draw a comparison between the C supports.A conventional MoS2/C catalyst using incipient wetness impregnation was prepared inorder to draw a comparison between the reverse micelle approach and a conventional prepa-ration.26Chapter 2ExperimentalThe procedures used to prepare and analyse the catalysts are outlined in this chapter.2.1 Catalyst preparation2.1.1 Carbon supportsTwo types of carbon were used to support the MoS2: activated carbon (Sigma Aldrich, DarcoG-60 high purity powder, referred to as AC), and petroleum coke which was chemicallyactivated by a fellow group member (referred to as PC). The properties of the supports aregiven in Table 3.1 on page 47, and the petroleum coke activation procedure is detailed inAppendix C.In all preparations, the Mo loading target was 10% Mo by weight.2.1.2 MoS2 preparation using reverse micellesIn an attempt to prepare a more active MoS2 catalyst, it was desirable to synthesise par-ticles of small diameters which would provide many edge and rim sites to facilitate thehydrodesulphurization reaction [30]. For this reason, reverse micelles were utilized to syn-27thesise nano-sized MoS2 particles. There are many micelle systems cited in the literature([88] and references within), and this study utilised two water-in-oil microemulsion systems:water/Igepal/cyclohexane and water/AOT/n-heptane. AOT (Sigma-Aldrich, ≥96.0% byTLC) shown in Figure 2.1 is an anionic surfactant, and IGEPAL CO-520 (Sigma-Aldrich,Average Mn 441, referred to as Igepal in this document) shown in Figure 2.2 is a non-ionicsurfactant; both surfactants were used as received. Oil phase solvents cyclohexane (Ana-chemia, 99%), and n-heptane (Fisher, 99.8%) were also used as received. The reasoning forselecting these specific surfactant/oil systems is discussed in section 3.4.The Mo salt ammonium tetrathiomolybdate (ATTM, discussed in more detail below)was obtained from Strem chemicals and used as received.Figure 2.1: Molecular structure of surfactant AOTFigure 2.2: Molecular structure of surfactant IGEPAL CO-520282.1.3 Typical MoS2 preparation via reverse micellesA preparation for MoS2/PC by the reverse micelle method is presented here for one batchof MoS2/PC catalyst. The calculations to arrive at the masses and volumes of substancesused can be found in Appendix A.In a typical preparation, enough surfactant was added to a certain volume of oil to givea 0.1 or 0.3 M solution of surfactant in oil – for example, 48.6 g of Igepal was added to 367mL of cyclohexane (a 0.3 M surfactant in oil solution), and the solution was stirred for twohours to achieve thermodynamic equilibrium. During the mixing period, a 5 mM solution ofsolubilized ATTM was prepared by weighing out enough salt and dissolving it in ultra-purewater (e.g. 0.1912 g of ATTM in 147.04 g of water).After the surfactant and oil phase had been sufficiently mixed, enough acid (0.5 M H2SO4in ultra-pure water) was added to the surfactant/oil mixture such that there would be 10 H+protons for each ATTM molecule (added later), and this solution was mixed for one hour.The acid served as the reducing and precipitating agent [59, 75]. In this preparation, 1.9 mLof 0.5 M H2SO4 was added.During the mixing period of acid into the oil/surfactant mixture, nano-sized reversemicelles are hypothesised to have formed, indicated by the solution remaining opticallytransparent. To this solution of nano-sized water pools encapsulated by surfactant andsuspended in oil phase, the solubilized Mo salt was added dropwise. Continuing with thepreparation outlined thus far, 37.95 mL of the aqueous 5 mM ATTM solution was addeddropwise over approximately nine minutes in nine portions of 4.22 mL using a calibratedpipette. This mixture was allowed to stir for thirty minutes, during which some of theATTM salt would have been reduced to MoS3 (as reported by Wang and co-workers [89]).It should be mentioned that the amount of total water added from the acid and aqueousATTM solution was carefully calculated so that the molecular water to surfactant ratio, ω,was 20.29After the mixing period, enough C support was added such that the metal loading was10 wt.% Mo in the final MoS2 catalyst, 0.1516 g of PC in this case, and this mixture wasallowed to stir for one more hour. It is hypothesised that during this mixing period, moreof the ATTM salt was reduced, and the particles anchored on the support.Finally, to break the emulsion and separate the solid precursor, 200 mL of acetone wasadded dropwise over the course of two hours while the solution was being stirred. To completethe separation of the phases, the mixture was allowed to settle overnight. The next day, ablack precipitate and a clear and colourless supernatant were completely separate such thatmuch of the supernatant could be decanted. The remaining solution was gravity filteredthrough Whatman quantitative ashless filter paper, and the collected solid was washed with250 mL of methanol in several aliquots in an attempt to remove as much supernatant fromthe catalyst precursor as possible. The solid was isolated and allowed to air-dry. To preventthe oxidation of the metal, at no point during the preparation or subsequent activity testingwas the precursor allowed to heat above 65 ◦C in air.The final step in the preparation of the catalysts obtained by the reverse micelle methodwas the annealing step which transformed the catalyst precursor into catalyst. The heatingregime removed much of the remaining surfactant by thermolysis, completed the reductionof the Mo salt, and allowed the MoS2 to organize into sheets. The transition of MoS4 toMoS2 is complete at 320◦C [90].The annealing procedure was performed by transferring the solid into a quartz glass U-tube into which a plug of quartz wool was first inserted. The tube containing the precursorwas then positioned into a Barnstead Thermolyne Type 47900 muffle furnace equipped witha temperature controller. The tube was connected to ultra-high purity Ar or N2 gas viaa mass flow controller on one end, and directly to building ventilation on the other end.Finally, the gas flow rate was slowly increased until the surface of the powder was visiblymoving, indicating that the bed of solid would not rise into the ventilation. The precursor30was flushed with inert gas for ten minutes. After all air was presumably displaced from thepowder, the furnace was set to ramp at 10 ◦C per minute to 350 ◦C, and to hold at 350 ◦Cfor sixty minutes during which inert gas continued to be flushed through the solid. After thehold period, the furnace was shut off, and the furnace door was opened to allow for rapidcooling to ensure that the annealing step was as close to 60 minutes as possible.Table 2.3 summarises the preparation methods and naming conventions of the catalystsprepared and screened for kinetic parameters in this study.2.1.4 Making a stable micelleThe preparation outlined above was a typical preparation performed during the course of thiswork, and yielded approximately 100 mg of solid catalyst per preparation. The product massmay seem disproportionate to the volume of solvent used during the preparation; however,the volumes used were strictly calculated to maintain micellar integrity or to keep certainratios constant between preparations.The phase diagram for the water/AOT/n-heptane system is presented in Figure 2.3.Stable micelles will form only from solutions whose component concentrations are such thatthey fall in the L2 region. A molar concentration of surfactant in oil needs to be high enoughto accommodate enough water and subsequently metal in the micelle to enable isolation ofsolid after preparation (some solid is lost during washing and annealing). Throughout muchof this project and in the synthesis outlined above, a concentration of 0.3 M of surfactant inoil was used.The next constraint was the water to surfactant ratio, which was set to 20 to maintainpreparations as close to identical as possible. Finally, the concentration of the metal solutionand the ratio of acid to Mo salt was kept constant at 5 mM and 10:1, respectively. With theseconstraints, the mass fraction of each component could be calculated, and is presented inTable 2.1 along with similarly calculated values for a 0.1 M system; the detailed calculations31can be found in Appendix A.Figure 2.3: Phase diagram of the water/AOT/n-heptane system. Figure adapted from [91],used under a creative commons licenseTable 2.1: Calculated mass fractions of stable micellar water/AOT/n-heptane systemsConcentrationof surfactantin oil (M)n-heptanemassfractionAOTmassfractionWatermassfraction0.1 0.89 0.06 0.050.3 0.74 0.14 0.12322.1.5 Mo salt selectionAside from selecting a surfactant/oil system, there was the matter of the Mo-containingsalt. As mentioned in the literature review, many researchers begin with MoO3 beforesulphiding the oxide to obtain MoS2 – the species active for HDS; however, the sulphidationprocess leads to the restructuring of the particle, manifesting in large and/or irregularlysized particles [53, 68–70]. To avoid needing to sulphide the oxidic precursor, the catalystprecursor in this study was made from a Mo salt which already contained S; the Mo salts usedin this study are summarised in Table 2.2. Ammonium tetrathiomolybdate (ATTM) waschosen as the sulphidic salt containing [MoS4]2−. Two batches of ATTM salt were obtainedfrom Strem and used as received (certificates of analysis for the two batches can be found inAppendix D). However, one batch of Mo salt obtained was later determined to be ammoniumpolythiomolybdate (APTM) with different properties than ATTM. For instance, APTM isonly very sparingly soluble in water, whereas the solubility of ATTM is moderate. Needlessto say, the solubility impacted the preparation of the micelle-derived catalysts drastically.Table 2.2: The inorganic salts used as Mo sources for catalyst preparationAcronym Full name FormulaATTM Ammonium tetrathiomolybdate (NH4)2MoS4APTM Ammonium polythiomolybdate (NH4)2Mo3S13(H2O)2.1.6 MoS2 preparation by incipient wetness impregnationFinally, to be able to compare the micelle-derived MoS2 catalysts to a conventionally-prepared catalyst, an MoS2/PC catalyst was prepared using incipient wetness impregnation.In this preparation method, only enough solvent is added to the solid support such that theliquid is completely taken up by the pores of the support; the salt to be deposited onto thesupport is dissolved in the liquid. To prepare 0.3 g of the conventional MoS2 catalyst (of33which 0.05 g would be MoS2, calculation in Appendix A) on PC, 0.0814 g of ATTM wouldneed to be added to 0.25 g of the support.Solubility trials of ATTM in ultra-pure water were performed, and it was found by visualobservation that 0.0836 g of ATTM was dissolved in 0.7113 g of ultra-pure water. However,the pore volume of PC is 0.97 cm3/g, so only 0.2425 cm3 of liquid could be added to 0.25 g ofthe support at once. Therefore, the ATTM salt was added to the support in three additionsover three days. On each day, 0.0279 g of ATTM was dissolved in 0.2415 g of ultra-pure waterand added dropwise onto the PC support using a Pasteur pipette. The solid was allowed torest at room temperature for 24 hours after each addition. It was assumed that the porevolume was not changed by the step-wise impregnation, and that all the liquid evaporatedbetween additions.As in the catalyst preparation by reverse micelles, the solid was never allowed to heatabove 65 ◦C in air. The conventional catalyst’s preparation was completed with the sameannealing procedure employed for the reverse micelle-derived catalysts.Table 2.3 summarises the preparation methods and naming conventions of the catalystsprepared and screened for kinetic parameters in this study.Table 2.3: Preparation methods and naming of the catalysts screened for kinetic parametersCatalyst Mo source Support Preparation methodMoS2/AC-RM ATTM AC Reverse micelles: water/Igepal/cyclohexaneMoS2/PC-RM ATTM PC Reverse micelles: water/Igepal/cyclohexaneMoS2/PC-WI ATTM PC Incipient wetness impregnation2.2 Catalyst characterizationA series of analytical methods were used to characterize the prepared catalysts. It shouldbe noted that all the characterizations listed were performed ex situ due to instrumental34limitations. For more accurate data, in situ characterizations would have been preferable.2.2.1 X-Ray diffractionThe presence of the desired metal, its oxidation state, and potentially crystallite size may bestudied by Powder X-Ray Diffraction (XRD). The spectrum of each catalyst was collectedusing a Bruker D8 Focus X-Ray diffractometer (equipped with a LynxEye detector) withCo Kα radiation (λ=1.79 A˚, 35 kV), and scan range 2θ = 10 to 90◦ at 0.08◦/second. Eachsample was spun at 50 RPM during measurement to ensure complete powder averaging.By analysing the sample supported on a quartz plate, which does not produce an amor-phous “bump” in the 15− 30◦ 2θ region like glass, the 002 reflection of MoS2 centred at 16◦2θ can be fully visualized, and crystallite size calculated using the Scherrer equation (2.1)using the peak’s line-broadening (provided the crystallites are large enough).dc =0.89 λβ cos(θ)(2.1)β is the full width at half the maximum intensity and θ is the angle of reflection.Since XRD is not performed under vacuum, the method is not sensitive to the presenceof surfactant and can be used to compare the structure and chemical identity of species priorto, and after annealing.2.2.2 X-Ray photoelectron spectroscopyThe presence and concentration of the desired metal on the catalyst surface as well as itsoxidation state can be studied by X-Ray Photoelectron Spectroscopy (XPS). XPS spectrawere collected using a Leybold MAX200 X-Ray photoelectron spectrometer with an Al Kαsource. Survey scans were collected at 192 eV, and narrow scans at 48 eV.XPS provides the binding energies (BE) of atoms at the solid’s surface. Correcting for35the charging effects by referencing the BE scale to the C 1s signal at 284.8 eV enables thecomparison of signal positions between samples. BE can be used to determine a chemi-cal species’ identity by comparing the experimental BE value to those reported by otherresearchers (see Table 2.4 for binding energies in Mo complexes relevant to this work).Table 2.4: Binding energies for Mo 3d3/2 spectral lines for selected compoundsSpecies Binding energy (eV)MoS2 232.3 ± 0.1 [92]MoO3 235.6 ± 0.2 [93]The collected XPS spectra were deconvoluted into separate contributions by the speciespresent on the sample surface using the open-source program XPS Peak 4.1.Furthermore, XPS is only able to scan the top ∼1 nm of sample [94], so the bulk Moconcentration cannot be measured; however, XPS can give an estimate of surface metaldispersion by comparing the ratios of the signal intensities due to the metal and the supportelement. In the case of MoS2/C, particles with low dispersion will have a low Mo:C intensityratio; however, if particles with high dispersion are successfully prepared, the Mo:C intensityratio will be higher [95].XPS is performed under high vacuum, which precludes the analysis of unannealed samplessince surfactant sublimes under vacuum and can destroy the equipment.2.2.3 Surface area measurementsA Micromeritics ASAP 2020 analyser was used to measure the surface area, pore volume,and average pore size using N2 adsorption-desorption isotherms measured at −196 ◦C. Eachsample was degassed in vacuum at 200 ◦C for 4 hours prior to analysis.Surface area of each sample was calculated from the BET N2 isotherm. Pore volume wascalculated based on the total amount of N2 adsorbed at 0.995 of the relative pressure (P/P◦),36where P◦ is the saturated vapour pressure of N2 at −196 ◦C (101 kPa), and P is the pressurethat is varied for the measurement. Each sample’s desorption isotherm was analyzed by theBarrett-Joyner-Halenda (BJH) method to determine the average pore size.2.2.4 Metal loading analysisThe exact Mo loading of the catalysts was measured using Inductively Coupled Plasma-Optical Emission Spectroscopy (ICP-OES). Sample preparation involved completely dis-solving the pre-weighed sample by boiling it in aqua regia. A set of standards from which acalibration curve may be constructed was used for quantitative analysis. Sample digestionand standard solution preparation procedures can be found in Appendix C.The samples were analysed using a Varian 725-ES Optical Emission Spectrometer witha 10 ppm solution of Eu in 2 vol.% HNO3 analysed simultaneously to correct for matrixeffects. The computer program ICP Expert II was used to calculate the metal concentrationin each prepared sample using a calibration curve.During analysis, the metal ions in solution are excited by argon plasma at a very hightemperature (∼ 9000 ◦C). As the ions relax to ground state they emit a unique ultravioletand visible light spectrum, the intensity of which is proportional to the concentration of thespecies in solution. The Mo content is measured quantitatively by the use of a calibrationcurve. In this method, the total amount of Mo in the bulk sample is accounted for.2.2.5 Transmission electron microscopyThe average size of the MoS2 particles and their size distribution were measured usingTransmission Electron Microscopy (TEM). Prior to analysis, a drop of MoS2/C catalystdispersed in acetone was introduced onto a formvar-coated, carbon-stabilized copper grid,and allowed to air dry. Images were captured using an FEI Tecnai G2 electron microscope37operated at 200 kV, for which source electrons were generated from a LaB6 filament.The particle size measured by TEM is a more direct measurement than the volume-averaged size measured by XRD; however, the drawback to analysing particle size withTEM is that a large number of images must be captured to obtain a true average particlesize, and can introduce operator bias during particle measurements. In this work, severalimages of each sample were captured, and the open-source program ImageJ was utilized tomeasure the diameters of many particles per sample from all over the TEM grid, allowingfor a distribution of sizes to be calculated.As in XPS analysis, unannealed samples could not be studied by TEM due to the pres-ence of surfactant which sublimes at high vacuum and obscures the image, in addition torisking damage to the instrument. Therefore, no samples can be analysed by TEM prior toannealing, and the effect of the annealing treatment on MoS2 dispersion and/or particle sizecannot be studied by this method.2.2.6 Thermogravimetric analysisThermogravimetric Analysis (TGA) was used to evaluate the progress of surfactant decompo-sition by thermolysis. By heating a sample under annealing conditions in a TGA instrument,the mass loss from the sample may be monitored. Surfactant thermolysis experiments ina TGA revealed the temperature at which the surfactants decomposed, and the degree ofdecomposition.∼5 mg of surfactant was placed into an aluminum pan and loaded into a Shimadzu TGA-50 thermogravimetric analyzer. A constant flow of N2 at 40 mL/min purged the system ofliberated themolysis products while the sample was ramped at 10 ◦C/min to 350 ◦C, and heldat 350 ◦C for sixty minutes to simulate the annealing process during catalyst preparation.382.3 Activity measurementsThe activity and selectivity of the prepared catalysts toward the hydrodesulphurization andhydrogenation reactions were investigated by using dibenzothiophene (DBT) as a modelreactant. The catalyst activity was tested using a novel batch microreactor, pictured inFigure 2.4. The reactor operates isothermally [86], and the temperature was controlled by a800 W Lindberg 55031 tubular furnace and continuously monitored by an OMEGA type Kthermocouple immersed in the reaction mixture. The reactor’s small dimensions reduce theheat and mass transfer effects [86].It was previously found by Kukard that the stainless steel thermocouple and reactorwalls of the reactor used in this study are active for cracking reactions [86]. To avoidintroducing activity from the walls and also for easier clean-up, a glass insert in which thereaction would proceed was utilised. These inserts were commissioned from CANSCI GlassProducts Ltd., and measured 260 mm in length, 6 mm in outer diameter and 4 mm in innerdiameter. Between reactions the inserts were cleaned with aqua regia to remove all catalystand products from the walls, and the thermocouple was wiped with acetone-moistened papertowel until no more black solid was removed.Prior to testing, the catalyst was ground and sieved to a maximum particle diameterof 63 µm to eliminate internal diffusion effects. To perform a reaction, a known amountof catalyst was suspended in acetone and transferred to a clean and dry glass insert byPasteur pipette. To remove the acetone without oxidizing the catalyst, the reactor tubewas immersed in a water bath maintained at 65 ◦C with a hotplate for 24 hours. Since theboiling point of acetone is 56 ◦C, this process removed all the acetone without allowing thecatalyst to become oxidized. Additionally, this slow evaporation ensured that all the solidremained at the bottom of the tube instead of being carried upward as might occur with afast evaporation.39Figure 2.4: Process flow diagram of the batch microreactor used for activity testing.Reprinted with permission from [86]. Copyright (2015) American Chemical Society40Once the acetone was removed from it, the insert was allowed to cool to room temperaturebefore feed containing the model compound was added to it. The volume of liquid feed addedto the insert was calculated such that the Mo concentration would be 900 ppm during thereaction. The feed consisted of 2 wt.% DBT and 0.4 wt.% CS2, with the balance composedof decahydronaphthalene (Sigma-Aldrich, ≥99%, referred to as decalin). The calculationsinvolved with the feed preparation are given in Appendix A. The purpose of the CS2 was tosulphide any MoO3 present in the catalyst, but was also kept low to prevent the poisoningeffect of H2S [80]. The liquid volume was sufficient to allow the vortex mixer to agitatethe contents at 2000 RPM, approximately 150 µL [86]. In addition to keeping the contentswell-mixed, the use of a vortex mixer at the base of the reactor also minimized externaldiffusion effects.The glass insert containing all the components of the reaction was placed inside the steelreactor, the thermocouple was inserted into the tube, and the reactor was hermetically sealedusing a stainless steel gasket. The reactor was purged of air by three cycles of adding ∼ 0.7MPa of H2 and its subsequent venting. After purging, the system was pressurized to 4.14MPa and tested for leaks with a PerkinElmer leak detector (N9306089) to ensure it waswell-sealed.With the reactor loaded, well-sealed and pressurized, the reaction could proceed. Re-actions were performed at reaction temperatures, T , of 350 ◦C, 365 ◦C, or 375 ◦C and atreactions times, t, of 60 min, 120 min, or 180 min. The pressure at reaction temperaturereached on average 5.08 MPa.In order for the reaction mixture to achieve reaction temperature, the tubular furnacesurrounding the reactor tube was activated and controlled by an OMEGA CN8201 temper-ature controller. The heating program was composed of a fast temperature ramp to 5 ◦Cbelow target T followed by a 120 second hold, and a slow ramp to final T . This programensured that the target temperature was not overshot by more than 2 ◦C. The reaction time41was considered to be t = 0 when the temperature first crossed temperature T . Regardlessof T , the time to reach it was ∼21 minutes, and all liquid products include the productsproduced during this heat-up period.After t minutes, the furnace control was set to 0, the furnace door opened, and a fanwas turned on to help cool the system faster. Once below 30 ◦C, the system was slowlydepressurized by opening the needle valve and directing the gas through a bubbler. Uponreaching atmospheric pressure, the system was opened and the liquid product extracted by along stainless steel needle attached to a glass syringe to prevent the adsorption of substratesto the walls, which might occur if a plastic syringe was used. The liquid products were storedat −20 ◦C unless being actively diluted for GC-MS analysis.2.3.1 Gas chromatography-mass spectrometryGas Chromatography-Mass Spectrometry (GC-MS) was used to quantify the products of thereaction. An Agilent 7890B GC separated the components in each sample, and they weredetected by an Agilent 5977A MS.To quantify the products, a calibration curve was constructed for each analyte, and aknown amount of internal standard was added to each calibration sample. Calibration curvesof biphenyl (BP), cyclohexylbenzene (CHB), and dibenzothiophene (DBT) were constructed.The internal standard used in this study was 1,1-diphenylethylene (Sigma-Aldrich, 97%,referred to as DPE), a compound with similar chemical characteristics as the analytes. Theproduct 1,2,3,4-tetrahydrodibenzothiophene (THDBT) was quantified by the DBT curve,since pure THDBT is not commercially available. No bicyclohexyl was detected as a productin this study.Each calibration curve plotted the known concentration ratio of analyte:DPE againstthe area ratio of analyte:DPE for each analyte diluted from pure reference samples, wherethe areas were measured by integrating the peak areas of chromatograms obtained by the42GC-MS.Prior to analysis, each sample was diluted with decalin to place the concentrations ofthe analytes in the linear range of the detector, and a known quantity of internal standardwas added. This method enabled the calculation of each analyte’s concentration in thereaction product mixture from the known concentration of internal standard and calculatedchromatogram areas of the internal standard and analyte.43Chapter 3Catalyst Properties andCharacterization Results3.1 IntroductionThe results of the catalyst characterization by the methods outlined in the Chapter 2 arepresented, summarised, and discussed in this chapter; a comprehensive summary is given atthe end of this chapter. The kinetic activity testing results are presented in Chapter 4.3.2 TGA resultsThe water/AOT/n-heptane system for nanoparticle preparation using reverse micelles hasbeen extensively studied [49, 59–62, 75, 88, 91, 96–98], and was therefore determined to bea suitable system to use since the effect of many variables (e.g. water:surfactant ratio, metalconcentration in the water pool) have been reported. Additionally, using the same systemas other researchers would enable comparisons to be drawn between particles prepared byother groups with the particles prepared during this project.44However, the preliminary activity results with catalysts prepared with APTM as a Mosalt in the water/AOT/n-heptane system were poor, and an explanation for these resultswas sought. It was hypothesised that some of the surfactant from the preparation remainedon the catalyst even after annealing, covering the metal and blocking the active sites, whichwould cause low activity. Studied by Luwang and co-workers, the decomposition of AOT wasreported to occur in two stages, the first at 340 ◦C, and the final decomposition occurringat 550− 700 ◦C which corresponds to the evaporation of SO2 due to the decomposition ofthe SO3 group [98].Figure 3.1: Thermolytic decomposition of surfactant AOT under N2 by TGAThe annealing procedure during catalyst preparation was reproduced in a TGA instru-ment, with the results of AOT thermolysis presented in Figure 3.1. The results indicate thatindeed the AOT was not fully burned off after heating at 350 ◦C for one hour, and that14.5 wt.% of the surfactant remained on the catalyst surface. However, the decompositionprofile was not identical to the one reported by Luwang [98], with the initial decomposition45occurring at a lower temperature.Since the goal of this project was to make an MoS2 catalyst with high activity towardHDS of DBT, a different system would need to be used to prepare the catalyst if leftoversurfactant was causing suppressed activity. AOT is an anionic surfactant, and adding saltsinto the water pool of a micelle bounded by an ionic surfactant was found to lead to anincrease in the micellar size [99], so the search for a different surfactant system was lim-ited to non-ionic surfactants. A non-ionic surfactant often mentioned in the literature isIGEPAL CO-520 (referred to as Igepal in this document), reportedly used as part of the wa-ter/Igepal/cyclohexane system. Only 0.7 wt.% of Igepal’s mass remained after thermolysisat annealing conditions, as shown in Figure 3.2, a good improvement over AOT.Figure 3.2: Thermolytic decomposition of surfactant Igepal under N2 by TGA463.3 BET resultsTable 3.1 presents the catalyst surface properties measured by BET. AC is used for manypurposes because of its high surface area, including as a catalyst support; the surface areaof the AC used in this study was measured to be 1136 m2/g, while the surface area of PCwas measured to be 2030 m2/g.Table 3.1: BET surface area, pore volume, and pore size of the prepared catalysts used foractivity testing, and of their supportsSupportor catalystBETsurface area(m2/g)Specific porevolume(cm3/g)Pore size(nm)Mo content(wt.%)Activated Carbon (AC) 1136 0.98 3.4 0MoS2/AC-RM 464 0.51 4.4 9.3Petroleum Coke (PC) 2030 0.97 1.9 0MoS2/PC-RM 927 0.49 2.1 7.8MoS2/PC-WI 1677 0.85 2.0 8.7The PC was prepared by chemical activation at 800 ◦C, so it was assumed that theannealing step at 350 ◦C during catalysis preparation did not impact the structure of PCmuch. However, it is possible that some pores collapsed during the annealing step of theMoS2/AC-RM catalyst. Furthermore, since all three catalysts presented in Table 3.1 under-went the same annealing process, it is possible to discuss the effect the catalyst preparationshad on the surface of the supports.Comparing PC to MoS2/PC-WI, it can be seen that the surface area decreased by 17%,pore volume decreased by 12%, and pore size remained statistically unchanged after wetimpregnation of the support and subsequent annealing. The small loss of surface area andpore volume was attributed to MoS2 occupying space on the surface of the catalyst.47Comparing PC to MoS2/PC-RM, the changes to the C surface are much more drastic:the surface area decreased by 54%, and the pore volume decreased by 49%. The pore sizeremained statistically unchanged.From the three PC data sets, it was concluded that the reverse micelle preparation methodwas responsible for surface area and pore volume loss on the surface of the PC support. Sinceall other factors except the use of surfactant were the same, it is hypothesised that the 0.17wt.% of Igepal that remained after annealing somehow decreased the surface area of the PCsupported catalyst. The exact mechanism is unknown.A significant loss in surface area and pore volume was also observed in the MoS2/AC-RM catalyst when compared to AC: 60% of the surface area was lost, and the pore volumedecreased by 48%. The catalyst had a larger pore size than AC before catalyst prepara-tion, suggesting that the MoS2 filled the smaller pores preferentially over the larger pores.However, AC was not treated at elevated temperatures prior to catalyst preparation, so it ispossible that some of the pores collapsed during annealing, resulting in the decreased surfacearea and pore volume. Conversely, it is possible that a similar mechanism responsible forthe loss of surface area on PC during the preparation of MoS2/PC-RM may be responsiblefor the surface area loss in the MoS2/AC-RM catalyst as well.3.4 Metal loading resultsThe preliminary catalyst preparations were plentiful, and it was not practical to fully char-acterize the kinetic activity of each catalyst. However, the success of the preparations canbe partially ascertained by the amount of Mo which was possible to be supported with eachpreparation method. The results are summarised in Table 3.2. Due to the combined errorsassociated with the ICP-OES instrument, incomplete metal extraction during digestion priorto analysis and the fact that the sample is not perfectly homogeneous, the error with the48ICP-EOS analysis was calculated to be 4% from performing some digestions and analysesmultiple times. The errors for values presented in Table 3.2 include this calculated error.Table 3.2: Summary of Mo sources, preparation methods, and metal loading of some of theprepared catalysts. Target loading was set to 10 wt.% Mo in all cases. The error associatedwith this measurement was calculated to be 4%#MosourceSupportMo content(wt.%)Reverse micellesystem1 APTM PC 1.2 ± 0.1 water/AOT/n-heptane2 APTM AC 4.0 ± 0.2 water/AOT/n-heptane3 APTM AC 6.8 ± 0.3 water/AOT/n-heptane4 APTM AC 5.9 ± 0.2 water/AOT/n-heptane5 APTM PC 0.4 ± 0.02 water/Igepal/cyclohexane6 ATTM AC 0.5 ± 0.02 water/AOT/n-heptane7 ATTM AC 9.1 ± 0.4 water/Igepal/cyclohexane8 ATTM AC 9.6 ± 0.4 water/Igepal/cyclohexane9 ATTM PC 8.5 ± 0.3 water/Igepal/cyclohexane10 ATTM PC 7.1 ± 0.3 water/Igepal/cyclohexaneComparing catalysts # 2, 3, and 4 in Table 3.2 allows for the calculation of error asso-ciated with the preparation of MoS2/AC using the water/AOT/n-heptane reverse micellesystem, calculated to be 26%. This high error is likely due to the fact that APTM is notvery soluble in water, as described in Section 2.1.5, so it was difficult to support the Moconsistently between preparations. However, even the high error does not account for thefact that when APTM was attempted to be supported on PC, only 12% of the intended Moended up on the support (# 1); using a different system of water/Igepal/cyclohexane didnot yield an improved catalyst (# 5). It was concluded that APTM is a poor choice of Mo49salt for preparing MoS2/C catalysts by either of the two reverse micelle systems used in thisstudy.When the ATTM salt was used, the preparation using the water/AOT/n-heptane reversemicelle system was also not successful in supporting a significant amount of Mo on AC (#6). A vast improvement in the amount of Mo supported on C was obtained when thewater/Igepal/cyclohexane system was used, with 93% of the Mo added attaching to the AC(# 7 and 8). Using the same system to prepare MoS2/PC was also successful (# 9 and10). It is hypothesised that it is the anionic character of AOT which interferes negativelywith the anionic [MoS4]2− salt, though the exact interaction is unknown, and the non-ionic character of Igepal which mediates a successful MoS2/C preparation. Furthermore,the repeated preparations presented in catalysts # 7 and 8 and # 9 and 10 allowed forthe calculation of error associated with the preparation of MoS2/C from ATTM using thewater/Igepal/cyclohexane reverse micelle system, at 9%.These findings helped to shape the final selection of catalysts which were fully screenedfor catalytic activity as discussed in Chapter 4. The final selection of salt was ATTM, andthe water/Igepal/cyclohexane system. Of the two reverse micelle systems tested, this oneproved more effective and consistent in preparing MoS2/C. The preparation methods of thecatalysts synthesised for activity testing are summarised in Table 3.3.The errors for Mo content values presented in Table 3.3 represent the calculated errorassociated with both the preparation and the ICP-OES analysis for the RM catalysts, andonly the analysis for the WI catalyst because its preparation error cannot be calculated sinceits preparation was not repeated.From Table 3.3, it can be observed that all three catalysts contain the statistically sameamount of Mo in the bulk: 8.6 ± 1.5 wt.% Mo.50Table 3.3: Mo loading and preparation methods of the catalysts screened for kinetic param-etersCatalystMosourceSupportMo content(wt.%)MethodMoS2/AC-RM ATTM AC 9.3 ± 0.8 water/Igepal/cyclohexaneMoS2/PC-RM ATTM PC 7.8 ± 0.7 water/Igepal/cyclohexaneMoS2/PC-WI ATTM PC 8.7 ± 0.3 Impregnation3.5 Surface composition and Mo coverageMeasuring Mo content by ICP-OES quantifies the total metal content in the catalyst bulk;however, some of the Mo measured by this method may be inaccessible to the DBT molecules.XPS analysis is a better method to measure how much Mo is on the surface of the catalystavailable to catalyse the HDS reaction.By comparing the values of Mo content measured by ICP-OES with those measured byXPS, it was possible to obtain a sense of the metal dispersion on the catalyst surface; theseresults for some of the catalysts prepared and discussed above are summarised in Figure 3.3.All data presented in the figure are for annealed catalysts, prepared from either ATTM orAPTM Mo salt (as indicated), supported on AC or PC (also indicated), and prepared eitherby reverse micelles from the water/Igepal/cyclohexane or water/AOT/n-heptane system, orby incipient wetness impregnation.From Figure 3.3 it can be seen that the catalyst with the highest surface concentrationof Mo was the one prepared by reverse micelles using the water/Igepal/cyclohexane systemand ATTM salt, supported on PC. Additionally, the amount of Mo on the surface of thisMoS2/PC catalyst showed a linear correlation between the amount of metal detected on thesurface and in the bulk. This indicates that when more Mo was supported on the PC by51Figure 3.3: Mo content of various catalysts as measured by XPS and ICP-OES. A connectingline was added for clarity and does not represent a modelled fitthis method, it remained as many small particles rather than agglomerating into fewer andlarger particles.In addition to Mo, the quantities of S, C, and O on the catalyst surface can be measuredby XPS. The results of these surface concentration measurements for the catalysts screenedfor activity are given in Table 3.4.Due to the error associated with the low surface S and Mo concentrations, the error inthe S:Mo ratio presented in Table 3.4 is such that no comparisons of this ratio between thecatalysts may be made; in all three catalysts the ratio was close to the expected value of 2.52Table 3.4: Surface composition of the prepared catalysts screened for kinetic parameters asmeasured by XPSAtomic % Atomic ratioSample O S C Mo Si S : MoMoS2/AC-RM 4.4 0.9 93.7 0.4 0.7 2.5MoS2/PC-RM 6.0 2.5 90.7 0.8 0.0 3.1MoS2/PC-WI 6.9 1.3 90.5 0.5 0.8 2.7It was observed by XRD that AC contained highly crystalline SiO2 and PC did not (XRDdata are presented below). However, the surface Si concentration results of the catalystspresented in Table 3.4 do not correlate with the Si observations from XRD. It is likely thatthe Si observed on the surfaces of MoS2/AC-RM and MoS2/PC-WI were the remnants ofquartz wool used during the annealing step of catalysis preparation. The effect of Si on theHDS of DBT was not studied, and the values are included for completeness.3.6 Deconvoluted narrow XPS scansAll collected XPS spectra were deconvoluted into separate peaks using the open-source pro-gram XPS Peak, each of which results from a different chemical species on the surface of thesample.Two separate batches of MoS2/PC-RM were prepared and analysed by XPS to furtherquantify the repeatability in the catalyst preparation. Figures 3.4a and 3.4b are the de-convoluted spectra of the Mo scans of these two batches, Figures 3.5a and 3.5b are thedeconvoluted S scans. Comparing these figures, it appears that the Mo species compositionwas very similar, while the S composition differed in the amount of SOx species present(Table 3.8).Figures 3.4c and 3.5c show the deconvoluted spectra of Mo and S of the species on53MoS2/AC-RM prepared by the micelle method, respectively. The spectra look similar totheir MoS2/PC-RM counterparts. Figures 3.4d and 3.5d show the deconvoluted spectra ofMo and S of the species on the surface of the MoS2/PC-RM catalyst prepared by incipientwetness impregnation, respectively.54Figure 3.4: XPS spectra of Mo 3d energy level for: a) MoS2/PC-RM (batch 1); b) MoS2/PC-RM (batch 2); c) MoS2/AC-RM; d) MoS2/PC-WI55Figure 3.5: XPS spectra of S 2p energy level for: a) MoS2/PC-RM (batch 1); b) MoS2/PC-RM (batch 2); c) MoS2/AC-RM; d) MoS2/PC-WI56Figure 3.6 is the deconvoluted XPS scan in the Mo region of the activated PC support.This scan shows that there was no Mo present on the surface of the support before catalysispreparation, as expected. Additionally, this scan shows that the K-β peak incorporated inthe deconvolutions of the Mo spectra above is justified, since this signal is significant abovenoise. The K-β signal is a satellite of the C 1s peak, and is therefore proportional to theconcentration of C in the sample. Since C was the support and accounts for ∼90% of thesample by weight, the K-β peak was large in the Mo region of the scans and needed to beaccounted for during peak deconvolution.Figure 3.6: XPS spectrum of petcoke before metal addition, presence of a K-β peak at 236.09eV evident3.7 Chemical states of Mo in the catalystsThe deconvolution of XPS spectra into individual peaks enabled the relative area of each peakto be calculated. The ratio of peak areas corresponds directly to the relative concentrationsof those species on the surface of the catalyst.57For Mo, the two species present on the catalyst surface were MoS2 and MoO3 (Tables 3.5− 3.7). The catalyst prepared by incipient wetness impregnation had a higher proportionof oxide present on the surface than the catalysts prepared by the reverse micelle method,whose proportion of MoS2 to MoO3 on the surface was statistically the same.Table 3.5: Chemical states and their relative amounts of Mo in MoS2/PC-RM, from the Monarrow XPS scanStateMo 3d3/2 B.E.(eV)Mo 3d5/2 B.E.(eV)Area(%)std error(%)MoS2 232.3 ± 0.2 229.1 ± 0.2 77 3MoO3 235.4 ± 0.2 232.2 ± 0.2 23 3Table 3.6: Chemical states and their relative amounts of Mo in MoS2/AC-RM, from the Monarrow XPS scanStateMo 3d3/2 B.E.(eV)Mo 3d5/2 B.E.(eV)Area(%)MoS2 232.0 ± 0.2 228.8 ± 0.2 79MoO3 235.4 ± 0.2 232.2 ± 0.2 21Table 3.7: Chemical states and their relative amounts of Mo in MoS2/PC-WI, from the Monarrow XPS scanStateMo 3d3/2 B.E.(eV)Mo 3d5/2 B.E.(eV)Area(%)MoS2 232.3 ± 0.2 229.1 ± 0.2 68MoO3 235.7 ± 0.2 232.5 ± 0.2 32583.8 Chemical states of S in the catalystsA similar deconvolution was performed on the collected narrow S XPS scans, and the resultsare summarised in Tables 3.8 − 3.10. Again, the catalysts prepared from the reverse micelleshad a similar surface composition with MoS2 accounting for a total of 78 to 88% and an oxideaccounting for the balance, while for the catalyst made by incipient wetness impregnationthe MoS2 contributed 65% of the S species and the oxide the balance. It should be notedthat the MoS2 in the prepared catalysts had two separate binding energies, and both energieshave been attributed to MoS2 [100].Table 3.8: Chemical states and their relative amounts of S in MoS2/PC-RM from two iden-tically made batches, from the S narrow XPS scan.StateS 2p1/2 B.E.(eV)S 2p3/2 B.E.(eV)Area(%)StateS 2p1/2 B.E.(eV)S 2p3/2 B.E.(eV)Area(%)MoS2 163.1 ± 0.2 161.9 ± 0.2 64 MoS2 163.0 ± 0.2 161.8 ± 0.2 29SOx 169.6 ± 0.2 168.4 ± 0.2 14 SOx 170.8 ± 0.2 169.6 ± 0.2 22MoS2 164.8 ± 0.2 163.6 ± 0.2 22 MoS2 163.8 ± 0.2 162.7 ± 0.2 49Table 3.9: Chemical states and their relative amounts of S in MoS2/AC-RM, from the Snarrow XPS scanStateS 2p1/2 B.E.(eV)S 2p3/2 B.E.(eV)Area(%)MoS2 162.7 ± 0.2 161.6 ± 0.2 39SOx 169.8 ± 0.2 168.6 ± 0.2 21MoS2 163.9 ± 0.2 162.8 ± 0.2 4059Table 3.10: Chemical states and their relative amounts of S in MoS2/PC-WI, from the Snarrow XPS scanStateS 2p1/2 B.E.(eV)S 2p3/2 B.E.(eV)Area(%)MoS2 163.0 ± 0.2 161.9 ± 0.2 26SOx 169.5 ± 0.2 168.4 ± 0.2 35MoS2 164.0 ± 0.2 162.8 ± 0.2 393.9 XRD resultsAll XRD data were collected by using a quartz plate to support the samples in order toavoid the large amorphous peak around 15◦ 2θ, which would result from using a glass plateand obscure the signals resulting from the sample.The intensity of a peak on an XRD spectrum is directly related to the degree of crys-tallinity of the sample. Essentially, more crystalline samples give rise to sharp and intensepeaks when studied by XRD. In terms of MoS2, several reflections are possible depending onthe shape and dimensions of the MoS2 particles which make up the sample, and the intensityof those peaks is indicated by the crystallinity, or degree of structural order, of the MoS2particles. If there is little long-range order in the crystallites, the peak will be very broad; ifthere is practically no long-range order, there will be no peak.60Figure 3.7: a) Collected XRD spectrum of the APTM salt, compared with b) the corre-sponding PDF for (NH4)2Mo3S13(H2O) (PDF # 04-010-0953)Figures 3.7 and 3.8 present the collected XRD spectra of the two batches of Mo salt,aligned with their corresponding Powder Diffraction Files (PDF). Comparing to the PDF,the first batch of ATTM was confirmed to actually be APTM, and the second batch ofATTM was indeed ATTM.61Figure 3.8: a) Collected XRD spectrum of the ATTM salt, compared with b) the corre-sponding PDF for (NH4)2MoS4 (PDF # 04-015-4295)62Figure 3.9 presents the collected spectra of the two C supports used in this study. Thespectrum for AC shows a very crystalline impurity of SiO2 at 32.5◦ 2θ, and very crystallinegraphitic C signal at 24.3 and 25.6◦ 2θ. The XRD spectrum of PC shows C which was veryamorphous, indicated by the broadness of the peak centred at ∼ 25◦ 2θ.Figure 3.9: XRD spectra of the supports prior to Mo addition: a) AC; b) PC63Figure 3.10: XRD spectra of MoS2/AC prepared from APTM by the reverse micelle methoda) before annealing; b) after annealingFigure 3.10 presents the XRD spectra of an MoS2/AC sample prepared from APTM usingthe reverse micelle method and the catalyst’s precursor prior to annealing. The distinctivedouble peaks of the Mo salt (compared to Figure 3.7a) are visible in the precursor’s spectrum,indicating that the APTM salt was not impacted by the acid-mediated reduction inside themicelle, likely because the salt was not inside the micelle during the reduction due to its poorsolubility in water. After annealing, the catalyst presented a broad peak centred at ∼ 16◦2θ indicative of poorly crystalline MoS2 composed of crystallites with moderate stacking.The average crystallite stack height of this MoS2/AC was estimated to be 2.3 nm from theSherrer equation (Equation 2.1).64Figure 3.11 presents the XRD spectra of catalyst MoS2/AC-RM and its precursor priorto annealing. In this catalyst, the unannealed sample’s spectrum does not resemble thatof its Mo salt (compared with Figure 3.8a), and looks rather like slightly-crystalline MoS2,indicated by a slight “bump” around 16◦ 2θ. The annealed sample does not show this16◦ 2θ peak, indicating that the annealing process did not cause the particles to agglomerateto larger stack heights, but that the MoS2 was well-dispersed and nano-crystalline. Thepresence of MoS2 on this sample was confirmed by ICP-OES and XPS (vide supra). Whilethe 16◦ 2θ (reflection 002) peak disappeared, a 39◦ 2θ (reflection 103) peak appeared. The103 reflection indicates some interlayer rotation in the crystallites [101, 102], though thedegree of this rotation is unknown.Figure 3.11: XRD spectra of MoS2/AC prepared from ATTM by the reverse micelle methoda) before annealing; b) after annealing65Figure 3.12 compares the XRD spectra of the three prepared catalysts which were usedfor activity screening. In all three samples, the peak around 16◦ 2θ which is most indicativeof stacked crystalline MoS2 is absent, indicating that the MoS2 on the catalyst was well-dispersed and nano-crystalline. The presence of MoS2 in these samples was confirmed byICP-OES and XPS, vide supra. All three catalysts display a peak related to the 103 reflectionof MoS2 to different intensities, indicating some interlayer rotation in all three MoS2 catalysts[101, 102]; however, the degree of this twisting is unknown. Additionally, MoS2/PC-WIdisplays a peak related to the 105 reflection also indicative of interlayer rotation, which isnot present for the catalysts prepared by reverse micelles.Figure 3.12: XRD spectra of the catalysts screened for activity a) MoS2/PC-WI; b)MoS2/PC-RM; c) MoS2/AC-RM66Additionally, the peaks present at 24.9, 34.8, and 36.0◦ 2θ observed in the spectrum ofMoS2/PC-RM in Figure 3.12b match the PDF of K2SO4, as presented in Figure 3.13. ThisK2SO4 is likely a product of the leftover K+ ions from the petcoke’s activation combiningwith SO2−4 ions from the H2SO4 acid used to reduce the Mo salt. However, no K was detectedon the surface of the catalyst by XPS, so it was concluded that the K2SO4 was buried in thepores of the support. The effect of K2SO4, if any, on the HDS reaction was not studied.Figure 3.13: a) XRD spectrum of MoS2/PC-RM, compared with b) the PDF for K2SO4(PDF # 04-006-8317)673.10 TEM resultsThe three catalysts screened for activity were studied by TEM to compare their MoS2 sheetstructures. Each sample was mounted on a TEM grid as outlined in section 2.2.5, and severalimages (∼15) were captured of each sample. Later, these images were visually searched fordiscrete particles which were measured using the open-source program ImageJ. Histograms ofthe particle sizes were constructed from the measurements, and are displayed in Figures 3.14and 3.16 for MoS2/AC-RM and MoS2/PC-RM, respectively. To obtain a robust measurementof particle size distribution, the aim was to measure at least 100 particles per catalyst fromall over the grid. Only several particles were measured per image in an attempt to get a fairdistribution by measuring particles from many images, representative of a larger area of thegrid. Small sections of TEM images after processing are presented in Figures 3.15, 3.17, and3.18. Complete images from which the sections were taken can be found in Appendix D.Catalysts MoS2/AC-RM and MoS2/PC-RM had their MoS2 well-dispersed on the surface,as shown by the narrow particle size distribution which fits a lognormal distribution well.The average diameter for the two catalysts prepared by the micelle method was statisticallythe same at the 95% confidence interval: 3.8 ± 0.8 nm for MoS2/AC-RM and 4.7 ± 1.0 nmfor MoS2/PC-RM, summarised in Table 3.11.Particles on MoS2/PC-WI were difficult to measure. As can be seen from Figure 3.18, theMoS2 was not arranged in discrete particles, but rather in long sheets which twist and bend,and continue for great lengths. Therefore, it was not possible to measure the long sheetsand instead only visible small pieces of MoS2 were measured. Obviously this skews theparticle size distribution significantly, and the particle length distribution is not reportedhere due to its large error. Additionally, from the alternating dark and light lines in theTEM micrographs it can be qualitatively observed that there are more layers of MoS2 onMoS2/PC-WI than on the MoS2/C catalysts prepared by reverse micelles.68Table 3.11: Particle lengths of MoS2 in the MoS2/C catalysts prepared by reverse micelles,determined using TEMCatalyst Average length (nm)MoS2/AC-RM 3.8 ± 0.8MoS2/PC-RM 4.7 ± 1.0Figure 3.14: Particle length distribution of MoS2/AC-RM, from measuring 101 particlesimaged using TEM69Figure 3.15: TEM image of MoS2/AC-RM, ellipses drawn around particles counted for thedistribution displayed in Figure 3.14Figure 3.16: Particle length distribution of MoS2/PC-RM, from measuring 100 particlesimaged using TEM70Figure 3.17: TEM image of MoS2/PC-RM, ellipses drawn around particles counted for thedistribution displayed in Figure 3.16Figure 3.18: TEM image of MoS2/PC-WI, ellipses drawn around several particles whichwere able to be distinguished from the otherwise long sheets713.11 SummaryA summary of the findings and conclusions which can be drawn from the characterizationresults follows.From TGA:ˆ Thermolysis of surfactant AOT was incomplete at 350 ◦C, 14.5 wt.% remained;ˆ Thermolysis of surfactant Igepal was nearly complete at 350 ◦C, 0.7 wt.% remained.From BET:ˆ Both MoS2/AC-RM and MoS2/PC-RM lost 57 ± 3% of surface area and ∼ 48% ofpore volume compared to the respective AC and PC supports;ˆ MoS2/PC-WI lost 17% of surface area and 12% of pore volume compared to PC.From ICP-OES:ˆ APTM was a poor choice of Mo salt for both reverse micelle systems studied;ˆ ATTM was a good choice of Mo salt when used with the water/Igepal/cyclohexanereverse micelle system, or incipient wetness impregnation;ˆ The water/AOT/n-heptane reverse micelle system was a poor choice when ATTM orAPTM were used, possibly due to the ionic character of AOT;ˆ The error associated with the Mo loading for MoS2/C catalysts prepared from thewater/Igepal/cyclohexane reverse micelle system was 9%;ˆ All three catalysts screened for activity contained 8.6 ± 1.5 wt.% Mo.72From XPS:ˆ MoS2/PC-RM had the largest surface dispersion of MoS2 of the three catalysts screenedfor activity;ˆ All three catalysts had a surface S:Mo ratio of ∼ 2;ˆ MoS2/PC-WI had a larger surface MoO3 to MoS2 ratio than the catalysts prepared byreverse micelles.From XRD:ˆ The AC used had an SiO2 impurity, PC did not;ˆ The C in AC was graphitic, C in PC was amorphous;ˆ MoS2/C prepared by the reverse micelles using APTM did not have its Mo reduced byH+ in the micelle; its crystallite stack height was ∼ 2.3 nm by XRD after annealing;ˆ Annealing completed the transformation of Mo salt to MoS2;ˆ MoS2/C prepared by the reverse micelle using ATTM did have its Mo reduced by H+in the micelle; its crystallite stack height was too small to be estimated by XRD;ˆ The MoS2 of all three catalysts screened for activity was well-dispersed and nano-crystalline;ˆ MoS2/PC-WI had a K2SO4 impurity, likely buried inside the particle.From TEM:ˆ The MoS2 was well-dispersed in MoS2/AC-RM and MoS2/PC-RM, and the particleswere small: 3.8 ± 0.8 nm and 4.7 ± 1.0 nm, respectively;73ˆ The MoS2 in MoS2/PC-WI existed as long sheets which twist and bend, rather thansmall particles;ˆ The MoS2 stack heights in MoS2/PC-WI were greater than those in the particles ofMoS2/AC-RM and MoS2/PC-RM.74Chapter 4HDS of DBT Reaction Kinetics4.1 IntroductionIn order to quantitatively compare the three prepared catalysts, their activity toward catalysingthe HDS of DBT reaction was tested in a novel batch microreactor. Following the reaction,the liquid products were quantified by GC-MS, and the product concentrations were usedfor kinetic modelling of the reaction. This allowed for the determination of the rate constantand Arrhenius parameters for each catalyst. The reaction activity data are presented in thischapter; a comprehensive summary is given at the end of this chapter.Table 4.1 summarises the catalyst preparation methods and naming conventions used inthis document.Table 4.1: Preparation methods for the catalysts whose kinetic activities were studiedName Support Preparation methodMoS2/AC-RM AC Reverse micelles: (water/Igepal/cyclohexane)MoS2/PC-RM PC Reverse micelles: (water/Igepal/cyclohexane)MoS2/PC-WI PC Incipient wetness impregnation754.2 Reaction pathways of HDSBased on the observed products of the HDS reaction in the microreactor as well as previouslypublished work [30, 32, 80, 83, 85, 103, 104], the HDS of DBT was assumed to follow tworoutes in parallel, as shown in Figure 4.1.Figure 4.1: Proposed pathways of the hydrodesulphurization reaction of dibenzothiophene,the dotted arrows indicate reactions assumed to be insignificant. Adapted from [83], withpermission from Elsevier Ltd, copyright (2005)In the Direct Desulphurization route (DDS), S is removed from DBT and proceeds tobiphenyl (BP). The rate of hydrogenation of BP to cyclohexylbenzene (CHB) is compara-tively slower than the other rates especially in the presence of H2S [80, 83, 103], and wasignored in this study.In the Hydrogenation route (HG), DBT is first hydrogenated to 1,2,3,4-tetrahydro-dibenzothiophene (THDBT) which may later be further hydrodesulphurized by the DS route76to CHB. The reverse of the HG, DS, and DDS reactions are insignificant. CHB can be fur-ther hydrogenated to bicyclohexyl; however, that reaction was ignored in this study due toits slow rate similar to the hydrogenation of BP to CHB mentioned above. No bicyclohexylwas observed in the reaction mixture.4.3 Kinetic model developmentOnce the concentrations of DBT, BP, CHB, and THDBT were quantified by GC-MS as afunction of reaction temperature and time, a MATLAB program was used to estimate therate constant parameter of each of the reaction components. The codes can be found inAppendix B.In the present study as in previous studies, [68, 80, 83–85, 105], the HDS of DBT reactionwas assumed to be pseudo first-order in the concentration of DBT, and zero-order in theconcentration of H2 since H2 is in large excess in the reactor and is in good contact with theliquid due to the vigorous mixing by the vortex mixer. Furthermore, it was assumed thatthe catalyst was not deactivating as the reaction progressed.The kinetic parameters were estimated by minimizing the sum of squares of the measuredreactant concentrations and the concentrations predicted by the model using Levenberg-Marquardt non-linear regression. The errors in the kinetic parameters were estimated fromthe diagonal of their covariance matrices. The series of MATLAB programs calculated thereactant concentration at each iteration by numerical integration of the 4 ODEs describingthe system, presented in Equations 4.1 − 4.4.77dCDBTdt= −(k′DDSCcat + kDDS)CDBT − (k′HGCcat + kHG)CDBT (4.1)dCBPdt= (k′DDSCcat + kDDS)CDBT (4.2)dCTHDBTdt= (k′HGCcat + kHG)CDBT − (k′DSCcat + kDS)CTHDBT (4.3)dCCHBdt= (k′DSCcat + kDS)CTHDBT (4.4)In Equations 4.1 − 4.4, parameters k′x are rate constants for the catalyst for the rel-evant reaction pathway as defined in Figure 4.1, and parameters kx are the rate constantcontributions of the thermal and thermocouple-induced reactions.Due to the small volume and number of moles which are associated with the use of amicroreactor, the carbon balance rarely sums to 100%. That is to say that summing theconcentrations of products and unreacted DBT in the product mixture rarely equals to thenumber of moles of DBT added to the reaction in the feed. Therefore, to force the carbonbalance, the concentration of each component (DBT, BP, CHB, and THDBT) in the productmixture (in units of µmol/mg) were summed, and this sum was set to be the total number ofmoles in the reaction. The sum was subsequently used to calculate the mole fraction of eachcomponent, designated as mol%C , with mol%C of all the components always summing to100. The raw data in µmol/mg and converted data in mol%C can be found in Appendix D.The difference between calculated and measured sum values was on average 6%. This er-ror could be due to thermal cracking resulting in products other than BP, CHB, and THDBTthe concentration of which were not measured. However, due to the low temperatures atwhich the HDS reaction was studied, this error is more likely due to the small volumes andnumber of moles used in this microreactor study.With the mole balance constraint in place, the mol%C concentration data for each com-ponent were entered into the MATLAB program outlined above.784.4 Modelling results of reactions without catalystIn order to calculate the rate constants for each component of the HDS reaction catalysedby MoS2/C as presented in Figure 4.1, a set of reactions was performed in the same manneras a catalytic reaction set, but without the presence of MoS2/C catalyst, to evaluate thecontribution to reactivity by the thermal and thermocouple-induced reactions. As with thecatalysed reactions, the product mixture was analysed by GC-MS for the concentration ofthe three main products.Figure 4.2 presents the measured product concentrations along with the modelled fitobtained from the MATLAB program at various temperatures. As can be seen, the productconcentrations were very low. The Arrhenius plot in Figure 4.3 was fit using a line whichwas within error of the points; therefore, the calculated kDDS and kHG values are robust andwere subtracted from the corresponding values obtained for the catalysts, vide infra. kDScould not be calculated due to a very low concentration of THDBT and the subsequent higherror associated with kDS.The low concentration of THDBT suggests that the thermal and thermocouple-inducedreactions proceeded preferentially through the DDS route, and not the HG route.The MATLAB program output the rate constant parameters for each component of theHDS reaction based on the concentrations of products, discussed in more detail below, andthe kx parameters due to the thermal and thermocouple-induced reactions are presented inTable 4.2.From Figure 4.2 it can be seen that the contributions of the thermal and thermocouple-induced reactions were small at all temperatures; the conversion reached a maximum of 7.4%after 3 hours at 375 ◦C, calculated by summing the mol%C for BP, CHB and THDBT, orsubtracting the mol%C of DBT from 1.79Figure 4.2: Measured data and modelled fit (line) for product concentrations of the HDS ofDBT reaction as a function of time, for the thermal and thermocouple-induced reactions ata) 350 ◦C; b) 365 ◦C; c) 375 ◦C80Table 4.2: Estimated reaction rate constants for HDS of DBT for the thermal andthermocouple-induced reactionsParameter 350 ◦C 365 ◦C 375 ◦CkDDS(s−1) (2.44 ± 0.23) E-06 (4.12 ± 0.16) E-06 (6.07 ± 0.33) E-06kHG(s−1) (1.01 ± 0.26) E-06 (1.18 ± 0.19) E-06 (1.62 ± 0.38) E-06Figure 4.3: HDS of DBT rate as a function of temperature, for the thermal and thermocouple-induced reactions814.5 Modelling results of the prepared MoS2 catalystsTo calculate the rate constants for each component of the HDS reaction as presented in Fig-ure 4.1, each prepared catalyst’s activity toward the HDS reaction was investigated by usingDBT as a model reactant in a batch microreactor. The reactor specifications and loadingprocedure can be found in section 2.3. Reactions were performed at reaction temperatures,T , of 350 ◦C, 365 ◦C, or 375 ◦C and at reactions times, t, of 60 min, 120 min, or 180 min. Theproduct mixture of each reaction was analysed by GC-MS for the concentration of the threemain products (BP, CHB, and THDBT) as well as the starting material DBT, as outlinedin section 2.3.1.Figures 4.4 − 4.5 and Tables 4.3 −4.4 present the data for the MoS2/AC-RM catalystprepared by the reverse micelle method.Figures 4.6 − 4.7 and Tables 4.5 − 4.6 present the data for the MoS2/PC-RM catalystprepared by the reverse micelle method.Figures 4.8 − 4.9 and Tables 4.7 − 4.8 present the data for the MoS2/PC-RM catalystprepared by incipient wetness impregnation.It should be mentioned that Figures 4.4, 4.6, and 4.8 represent the product concen-trations in the reaction mixture, including the products generated from the thermal andthermocouple-induced reactions discussed above. The modelled fits presented in these fig-ures were generated by the MATLAB program detailed in Appendix B, and the data intabular format can be found in Appendix D.Furthermore, it is worth reiterating that the figures present the mol%C concentrations ofproducts and starting material DBT in the reaction mixture as mole factions of the sum, thetotal number of moles of DBT, BP, CHB, and THDBT measured in the reaction mixture.This in effect forced the mole balance, and at every point along the abscissa in Figures 4.4,4.6, and 4.8, the mol%C concentrations of DBT, BP, CHB, and THDBT always sum to 1.82Furthermore, the MATLAB program outputs the rate constant parameters for each com-ponent of the HDS reaction based on the concentrations of products. The rate constantparameters associated with the thermal and thermocouple-induced reactions, kx, presentedin Table 4.2 were subtracted from the program’s output for the catalyst’s rate constant pa-rameters. Therefore, the rate constant parameters k′x presented in Tables 4.3, 4.5, and 4.7are the parameters associated only with the specified catalyst and reaction pathway; theratios of k′DDS to k′HG for each catalyst at each T are also presented in those tables, and arediscussed below.Even before using the MATLAB program to deconvolute the rate constants, it was ob-vious that the desulphurization of the partially hydrogenated intermediate, THDBT, by theDS route was very fast for all catalysts. This conclusion was made based on the low con-centration of THDBT, and the fact that all CHB present in the product proceeded throughTHDBT (the HG of BP to form CHB was assumed to be insignificant, vide supra). There-fore, once the rate constants were separated by the kinetic modelling outlined above, it wasnot surprising that k′DS, compared with k′DDS and k′HG, had the largest magnitude for allthree catalysts at all temperatures tested.However, it can be seen from the figures presenting product concentrations as a function oftime that CHB concentration was never very high for the catalysts prepared by the reversemicelle method (Figures 4.4 and 4.6), and moderate when prepared by incipient wetnessimpregnation (Figure 4.8). This is because the HG step needs to precede the DS step, andas presented in Tables 4.3, 4.5, and 4.7, k′HG had the smallest magnitude of the rate constantparameters for all three catalysts at all temperatures tested. Thus, it can be surmised thatthe direct desulphurization route was the most significant contributor to the desulphurizationof DBT for all three catalysts. The degree of the catalyst’s preference for DDS over HG isquantified by the k′DDS to k′HG ratio.Having the k′x parameters due only to catalyst, it was then possible to calculate both83the pre-exponential factor, A, and the activation barrier energy, Ea, from the interceptand slope of ln(k′x) plotted against T−1, respectively, as factors for the Arrhenius equation(Equation 4.5). These plots are presented in Figures 4.5, 4.7, and 4.9. Calculating theArrhenius parameters for the HDS reactions catalysed by the three catalysts outlined aboveallowed for their quantitative comparison, and therefore the comparison of the preparationmethods and C supports. The calculation of Arrhenius parameters for the thermal andthermocouple-induced reactions can be calculated in a similar way. However, the A and Eavalues for the thermal and thermocouple-induced reactions are not significant outside of thisstudy and are not presented here; they can be found in Appendix D.k = A e (−Ea/RT ) (4.5)The catalysts can also be compared qualitatively by examining Figures 4.4, 4.6, and 4.8,since the plotted concentration DBT is directly related to conversion. The mole fraction ofDBT (its mol%C), at t = 0 is 1, and decreases as the reaction proceeds; the conversion, X,can easily be calculated by:X = 1−mol%C of DBTwhere mol%C of DBT =CDBTΣ(CDBT + CBP + CCHB + CTHDBT )For instance, the conversion of DBT catalysed by MoS2/PC-RM (and including thethermal and thermocouple-induced reactions) after 3 hours was 11% at 350 ◦C, 19% at365 ◦C, and 26% at 375 ◦C. Of the three prepared catalysts, MoS2/PC-WI had the greatestconversion at the times and temperatures tested.84Figure 4.4: Measured data and modelled fit (line) concentrations for the products of theHDS of DBT reaction catalysed by MoS2/AC-RM as a function of time at a) 350◦C; b)365 ◦C; c) 375 ◦C85Table 4.3: Estimated reaction rate constants for HDS of DBT catalysed by MoS2/AC-RMParameter 350 ◦C 365 ◦C 375 ◦Ck′DDS(cm3 g−1Mo s−1) (1.05 ± 0.06) E-02 (1.52 ± 0.04) E-02 (2.1 ± 0.05) E-02k′HG(cm3 g−1Mo s−1) (4.33 ± 0.72) E-03 (5.77 ± 0.50) E-03 (6.60 ± 0.62) E-03k′DS(cm3 g−1Mo s−1) (2.07 ± 0.58) E-01 (2.90 ± 0.43) E-01 (3.64 ± 0.43) E-01k′DDS/k′HG 2.4 ± 0.4 2.6 ± 0.2 3.2 ± 0.3Figure 4.5: HDS of DBT rate as a function of temperature, catalysed by MoS2/AC-RM86Table 4.4: Pre-exponential factors, A, and activation barrier energies, Ea, for the rate con-stants associated with HDS of DBT catalysed by MoS2/AC-RM. The error associated withA is at least 2 orders of magnitude smaller than A and is omitted hereParameter A (cm3 g−1Mo s−1) Ea (kJ mol−1)k′DDS 5.0 E+05 91.7 ± 8.1k′HG 2.6 E+02 57.0 ± 4.6k′DS 4.5 E+05 75.6 ± 1.187Figure 4.6: Measured data and modelled fit (line) concentrations for the products of theHDS of DBT reaction catalysed by MoS2/PC-RM as a function of time at a) 350◦C; b)365 ◦C; c) 375 ◦C88Table 4.5: Estimated reaction rate constants for HDS of DBT catalysed by MoS2/PC-RMParameter 350 ◦C 365 ◦C 375 ◦Ck′DDS(cm3 g−1Mo s−1) (6.84 ± 0.42) E-03 (1.41 ± 0.06) E-02 (1.98 ± 0.10) E-02k′HG(cm3 g−1Mo s−1) (1.86 ± 0.48) E-03 (3.85 ± 0.62) E-03 (4.03 ± 1.12) E-03k′DS(cm3 g−1Mo s−1) (1.12 ± 0.37) E-01 (1.75 ± 0.47) E-01 (2.20 ± 0.88) E-01k′DDS/k′HG 3.7 ± 1.0 3.7 ± 0.6 4.9 ± 1.4Figure 4.7: HDS of DBT rate as a function of temperature, catalysed by MoS2/PC-RM89Table 4.6: Pre-exponential factors, A, and activation barrier energies, Ea, for the rate con-stants associated with HDS of DBT catalysed by MoS2/PC-RM. The error associated withA is at least 3 orders of magnitude smaller than A and is omitted hereParameter A (cm3 g−1Mo s−1) Ea (kJ mol−1)k′DDS 7.4 E+09 143.4 ± 11.4k′HG 1.6 E+06 106.3 ± 39.1k′DS 4.4 E+06 90.5 ± 5.890Figure 4.8: Measured data and modelled fit (line) concentrations for the products of theHDS of DBT reaction catalysed by MoS2/PC-WI as a function of time at a) 350◦C; b)365 ◦C; c) 375 ◦C91Table 4.7: Estimated reaction rate constants for HDS of DBT catalysed by MoS2/PC-WIParameter 350 ◦C 365 ◦C 375 ◦Ck′DDS(cm3 g−1Mo s−1) (3.96 ± 0.03) E-02 (7.01 ± 0.17) E-02 (9.66 ± 0.35) E-02k′HG(cm3 g−1Mo s−1) (3.31 ± 0.17) E-02 (5.24 ± 0.18) E-02 (6.66 ± 0.36) E-02k′DS(cm3 g−1Mo s−1) (5.62 ± 0.91) E-01 (7.48 ± 0.99) E-01 (8.45 ± 1.9) E-01k′DDS/k′HG 1.2 ± 0.1 1.3 ± 0.1 1.5 ± 0.1Figure 4.9: HDS of DBT rate as a function of temperature, catalysed by MoS2/PC-WI92Table 4.8: Pre-exponential factors, A, and activation barrier energies, Ea, for the rate con-stants associated with HDS of DBT catalysed by MoS2/PC-WI. The error associated withA is at least 4 orders of magnitude smaller than A and is omitted hereParameter A (cm3 g−1Mo s−1) Ea (kJ mol−1)k′DDS 4.2 E+08 119.6 ± 4.6k′HG 2.6 E+06 94.0 ± 5.0k′DS 2.0 E+04 54.3 ± 6.34.6 Error associated with mol%CFive reactions were performed twice, which enabled the calculation of the error associatedwith the graphically-presented mol%C for each of DBT, BT, THDBT, and CHB. The errorsare presented in Table 4.9, and were calculated using the following formula:s2e =∑ki=112(Yi1 − Yi2)2∑ki=1 ni − kTable 4.9: Errors associated with presented mol%C , concentration ratios of observed productsor starting material compared to total initial DBT concentrationSpecies Errormol%C CHB ± 0.3mol%C BP ± 0.5mol%C THDBT ± 0.2mol%C DBT ± 0.8934.7 Discussion of k’ resultsThe literature was searched for analogous work (HDS of DBT by unpromoted MoS2) in thehopes of comparing k′ values for catalysts prepared in this study to those prepared by otherresearchers. It was found that there is little convention for reporting k′ values, in terms ofthe units of the values and the data completeness − some researchers report k′ values at onlyone temperature without also reporting Arrhenius parameters to enable conversion to othertemperatures. Consequently, the data from this study were calculated for two temperaturesto enable comparison with other work. All compared data were presented or converted tothe units of molDBT s−1 g−1catalyst and are presented in Tables 4.10 and 4.11.Additionally, some researchers report k′ in units of g−1 for supported catalysts withoutindicating whether the unit is g−1catalyst or g−1metal. It was assumed that the researcher alwaysmeant g−1catalyst. However, if the reported value was in units of g−1metal, the discrepancy wouldbe an order of magnitude for a catalyst with a 10% metal loading. Due to these factors, onlyloose comparisons between the k′ values observed in this study and those reported by otherresearchers were drawn.Table 4.10: Comparison of k′ values at 350 ◦C. If multiple values were reported by the sameresearcher, the largest k′ is citedResearcher Catalyst k′DBT (molDBT s−1 g−1catalyst)Camacho-Bragado [106] Unsupported MoS2 3.7 E-07Romero [107] Unsupported MoS2 1.9 E-06Tye [105] Unsupported MoS2 3.32 E-06This workMoS2/AC-RM (2.0 ± 0.5) E-06MoS2/PC-RM (9.2 ± 2.8) E-07MoS2/PC-WI (5.4 ± 0.8) E-06Comparing k′ values for the catalysis of the HDS reaction toward DBT at 350 ◦C pre-94sented in Table 4.10, it was concluded that the catalysts prepared by reverse micelles inthis study catalysed the HDS of DBT at a comparatively similar rate to the unsupportedMoS2 catalysts prepared by Camacho-Bragado, Romero, and Tye. The rate constant of theMoS2/AC catalyst was ∼2x that of the MoS2/PC catalyst prepared by reverse micelles. Therate constant of the catalyst prepared by incipient wetness impregnation had an activitywhich was ∼3 to 6x greater than that for the catalysts prepared by reverse micelles, and wasalso larger than the k′ for unsupported MoS2 reported by the compared groups.It could be argued that the difference in catalytic activity between the RM and WIcatalysts prepared in this study is due to the lower surface area and pore volume of theRM catalysts compared to the WI catalyst. While it is true that MoS2/PC-RM had halfthe surface area (927 compared with 1677 m2/g) and pore volume (0.49 compared with 0.85cm3/g) of MoS2/PC-WI, the k′ value for MoS2/PC-WI is ∼6x that of the MoS2/PC-RMcatalyst. Therefore, it was concluded that the surface appearance of the support was notwholly responsible for the differences in the magnitudes of k′.Table 4.11: Comparison of k′ values at 340 ◦C. If multiple values were reported by the sameresearcher, the largest k′ is citedResearcher Catalyst k′DBT (molDBT s−1 g−1catalyst)Reinhoudt [82] MoS2/Al2O3 5.56 E-07Egorova [80] MoS2/Al2O3 1.37 E-03This workMoS2/AC-RM (1.6 ± 0.4) E-06MoS2/PC-RM (6.9 ± 2.1) E-07MoS2/PC-WI (4.5 ± 0.6) E-06The k′ values presented in Table 4.11 for comparison are for the HDS of DBT at 340 ◦Ccatalysed by unpromoted and supported MoS2 catalysts. As with the comparison at 350◦C,the k′ for MoS2/AC-RM was ∼2x greater than that for MoS2/PC-RM; the k′ for MoS2/PC-95WI was again ∼3x greater than that for MoS2/AC-RM. The k′ values for the catalystsprepared in this study were greater than the k′ for the MoS2/Al2O3 catalyst prepared byReinhoudt.The k′ for the MoS2/Al2O3 catalyst presented by Egorova is significantly greater than thatfor the catalysts prepared in this study, even accounting for the possible order of magnitudeinflation due to the fact that the article did not state whether the units for the k′ wereg−1catalyst or g−1metal. Even if the k′ value presented in [80] wass stated in the units of g−1metal, afterconversion to g−1catalyst the k′ value was still ∼2 orders of magnitude greater than the mostactive catalyst prepared in this study, MoS2/PC-WI. However, it should be noted that theregime in Egorova’s study was a fixed bed reactor in a continuous mode [80].4.8 Comparison of accessed selectivitiesWhile k′DBT is a measure of the overall rate of the conversion of DBT, selectivity measures howmuch of each product component – BP, CHB, THDBT – was produced from the convertedDBT. Together with TEM and XRD data, insight may be gained into the relationshipbetween the shape of the MoS2 crystallites and product selectivity, and to make a comparisonwith the expectations from the rim-edge model. The selectivity is governed by the magnitudeof the individual k′x components contributing to the overall k′DBT rate constant.Figures 4.10 − 4.12 present the proportion of each compound in the product mixtureafter reaction at a specific T , by catalyst. These figures allow for the comparison of theselectivities accessed at different T by a screened MoS2/C catalyst. Figure 4.13 graphicallycompares each catalyst’s selectivity toward the DDS route, where only the proportion of BP(the sole product of the DDS route) is plotted, and is effectively a graphical representationof the calculated k′DDS/k′HG ratios presented in Tables 4.3, 4.5, and 4.7.Observing the changes in the k′DDS/k′HG ratios presented in Tables 4.3, 4.5, and 4.7, it96was noted that the ratio for the each of the RM catalysts remained approximately constantat 350 ◦C and 365 ◦C, and significantly increased at 375 ◦C. The k′DDS/k′HG ratio remainedapproximately constant for all T when the reaction was catalysed by MoS2/PC-WI. Thecause for this effect is unknown, but a sintering event of the nano-crystalline MoS2 at thehighest T which would not effect the MoS2 in longer sheets as in MoS2/PC-WI is put forthas a hypothesis.97Figure 4.10: Selectivity of the HDS of DBT reaction catalysed by MoS2/AC-RM after subtracting the thermal andthermocouple-induced reaction contributions at a) 350 ◦C, b) 365 ◦C, c) 375 ◦C98Figure 4.11: Selectivity of the HDS of DBT reaction catalysed by MoS2/PC-RM after subtracting the thermal andthermocouple-induced reaction contributions at a) 350 ◦C, b) 365 ◦C, c) 375 ◦C99Figure 4.12: Selectivity of the HDS of DBT reaction catalysed by MoS2/PC-WI after subtracting the thermal andthermocouple-induced reaction contributions at a) 350 ◦C, b:) 365 ◦C, c) 375 ◦C100Figure 4.13: Comparison between the catalysts MoS2/AC-RM, MoS2/PC-RM and MoS2/PC-WI for selectivity towardthe DDS route at a) 350 ◦C, b) 365 ◦C, c) 375 ◦C101From the graphical comparisons, it can be observed that the catalysts prepared by thereverse micelle method showed similar selectivity, one that was more favourable toward DDS,than the catalyst prepared by incipient wetness impregnation which still favoured DDS butnot by such a wide margin. Table 4.12 summarises some of the data which were presentedabove to enable a discussion of the selectivity differences between the catalysts.Table 4.12: Summary of kinetic parameters of the three catalysts at 350 ◦CCatalystk′DDS(cm3 g−1Mo s−1)k′HG(cm3 g−1Mo s−1)k′DS(cm3 g−1Mo s−1)k′DDS/k′HGMoS2/AC-RM (1.05 ± 0.06) E-02 (4.33 ± 0.72) E-03 (2.07 ± 0.58) E-01 2.4MoS2/PC-RM (6.84 ± 0.42) E-03 (1.86 ± 0.48) E-03 (1.12 ± 0.37) E-01 3.7MoS2/PC-WI (3.96 ± 0.03) E-02 (3.31 ± 0.17) E-02 (5.62 ± 0.91) E-01 1.2From the TEM data presented in section 3.10, it was concluded that the MoS2 preparedby the reverse micelle method was significantly smaller than MoS2 made by impregnationwhose particle size could not be accurately measured due to the long sheets formed. Therim-edge model predicts that the smaller crystallites would be more active fore HDS, sincesmall particles offer more rim and edge sites per catalyst mass to facilitate the reaction.However, from the data presented in this chapter and summarised in Table 4.12, it wasconcluded that the MoS2/PC-WI catalyst is ∼5 to 8x more active for HDS of DBT than theMoS2/C catalysts prepared by the reverse micelle method, by comparing the k′DBT values(sum of the k′DDS + k′HG components).Furthermore, the rim-edge model states that HG occurs on the rim sites of MoS2, whileDDS is affected at edge sites [30]. By geometry, the small MoS2 crystallites prepared in thisstudy by the reverse micelle method had a larger ratio of rim to edge sites than the longMoS2 sheets made by the impregnation method; however, the k′DDS/k′HG ratio was largerfor the smaller MoS2 particles − the opposite of what the rim-edge model predicts. This102point is visualised in Figure 4.13 where the molar proportion of BP in the product mixtureproduced by MoS2/PC-WI is consistently lower than that by MoS2/PC-RM and MoS2/AC-RM. Clearly, the rim-edge model does not accurately predict the catalytic activity andselectivity of the MoS2/C prepared in this study. The rim-edge model has in fact alreadybeen questioned by some researchers who concluded that the activity of very small particlescould not be predicted by the rim-edge model [26, 104].Topsøe and co-workers found that Mo/Al2O3 catalysts prepared with a low metal loadingformed small crystallites which were less active (per mass of metal) for HDS than thoseprepared with a higher loading. They concluded that in small crystallites, more atoms arein a corner position rather than being at a true edge position, and that the corner atoms arenot as catalytically active as the edge atoms [26]. This is due to the fact that any S bindingto metal atoms at a corner are more weakly bound than if they were bound at an edge site.The weaker bond with a substrate’s S molecule means less electron back-donation which isnecessary for the C-S bond scission [26, 104].Hensen and co-workers related the low stacking degree in small MoS2 crystallites to a lowoverall catalytic activity of HDS of DBT. The authors concluded that since the sterically-hindered DBT molecule binds to MoS2 in an η6 planar geometry through the benzene ring[108], rather than η1 through the S atom like thiophene, stacking degree has a large impacton the magnitude of k′ for DBT, with k′ increasing with stacking order [104]. Furthermore,the article related that k′DDS/k′HG decreases as stacking order increases due to k′HG increasingwith stacking degree, even though the larger stacking order means that fewer edge sites areavailable [104].The activity trends found in this study mirrored those found by Topsøe and Hensen, andit was concluded that the high proportion of corner sites combined with a low MoS2 stackingorder in MoS2/AC-RM and MoS2/PC-RM resulted in a lower activity toward the catalysisof HDS of DBT. The MoS2 stacking order in MoS2/PC-WI was larger, which corresponded103to a greater HDS activity.Furthermore, the selectivity trends can also be explained by the hypotheses set out byTopsøe and Hensen. Referring to Table 4.12, it can be seen that k′HG of MoS2/PC-WI was∼1 order of magnitude greater the RM catalysts. While the k′DDS of MoS2/PC-WI was alsogreater than those of the RM catalysts, it was not by an order of magnitude. Therefore, itwas concluded that the larger magnitude of the overall rate constant for HDS of DBT of theMoS2/PC-WI was due to more hydrogenation occurring on the WI catalyst than on the RMcatalysts, due to the higher stacking order of the MoS2 in MoS2/PC-WI.4.9 SummaryA summary of the findings and conclusions which can be drawn from the kinetic modellingresults follows.ˆ The thermal and thermocouple-induced reactions were slow at all T and t; their con-tributions were accounted for;ˆ k′DBT values of catalysts prepared in this study are comparable to those of unsupportedMoS2 prepared by other researchers;ˆ The proportion of sites available for either DDS or HG depends on the catalyst prepa-ration;ˆ A similar proportion of DDS and HG sites was elucidated to be present on the catalystsprepared by the RM method, for both the AC and PC supports;ˆ MoS2 supported on AC and PC showed a similar activity toward catalysing the HDSof DBT when the MoS2 was prepared by the reverse micelles; therefore, PC is a goodalternative support to AC in these catalysts;104ˆ Long and stacked sheets of MoS2 were more active toward the HDS of DBT than nano-crystallites, as seen by the higher rate constant for MoS2/PC-WI compared with thatfor MoS2/PC-RM and MoS2/AC-RM;ˆ The rim-edge model did not correlate to the activity and selectivity observed for HDSof DBT catalysed by MoS2/C prepared by the reverse micelle method;ˆ The larger magnitude of k′DBT of MoS2/PC-WI was concluded to result from more HGoccurring on the WI catalyst than on the RM catalysts, due to the higher stackingorder of the MoS2 in MoS2/PC-WI.105Chapter 5Conclusions and Recommendations5.1 ConclusionsCarbon-supported molybdenum disulphide catalysts were successfully prepared by two meth-ods using ammonium tetrathiomolybdate as the molybdenum source: reverse micelles usingthe water/IGEPAL CO-520/cyclohexane system, and incipient wetness impregnation fromultra pure water. MoS2 prepared by impregnation was supported on activated petroleumcoke, and MoS2 prepared by the reverse micelle method was supported on both activatedpetroleum coke and activated carbon. In this way, comparisons between the two preparationmethods and two supports were possible.All three catalysts were extensively characterized, and it was found that the catalystsprepared by the reverse micelle method contained nano-sized MoS2. When prepared by thereverse micelle method, MoS2/AC was composed of supported particles measuring 3.8 ± 0.8nm and the MoS2/PC catalyst’s particles measured 4.7 ± 1.0 nm. The MoS2 in the MoS2/PCcatalyst prepared by incipient wetness impregnation consisted of long sheets, rather thansmall particles.The three catalysts were compared quantitatively by the use of a batch microreactor in106which the catalysed hydrodesulphurization of model compound dibenzothiophene was af-fected. The products biphenyl, cyclohexylbenzene, and 1,2,3,4-tetrahydrodibenzothiophenewere observed and quantified by Gas Chromatography-Mass Spectrometry. A pseudo first-order model was used to decouple the rate constant parameters of each of the three catalystsinto separate components corresponding to two parallel routes (hydrogenation and directdesulphurization) by which the HDS of DBT occurs.It was found that the rate constant per gram of Mo for the MoS2/PC prepared byimpregnation was ∼3 to 5x greater than that for the catalysts prepared by reverse micellesin the temperature range of 350 to 375 ◦C. MoS2 supported on AC and PC showed a similaractivity toward catalysing the HDS of DBT when the MoS2 was prepared by reverse micelles;therefore, PC is a good alternative support to AC for the MoS2 prepared by this method.Additionally, the rate constant component associated with the hydrogenation route wasan order of magnitude greater for the catalyst prepared by impregnation than that for theMoS2/C catalysts prepared by reverse micelles. It was concluded that the larger stackingorder in MoS2/PC prepared by impregnation provided more sites for hydrogenation to oc-cur, which resulted in an overall larger rate constant than that for the catalysts prepared byreverse micelles whose MoS2 stacking orders were minimal due to the small particle dimen-sions. These findings contradicted the rim-edge model set out by Daage and Chianelli, andcorroborated the hypotheses set out by Topsøe and Hensen.5.2 RecommendationsThis section sets out some recommendations for future work which would either improve on,or extend, the research performed in this study.The three catalysts were extensively characterized and it was concluded that the MoS2/Ccatalysts prepared by reverse micelles had a higher dispersion than MoS2/PC-WI. However,107O2 chemisorption could be performed to obtain a quantitative measure of MoS2 dispersionand the number of catalytic sites present on each catalyst. This would allow the rate con-stants to be calculated per active site which in turn would allow the turnover frequency(TOF) on each catalyst to be quantified. TOF is a measure of the speed of conversion peractive site; therefore, this measurement would help to further elucidate the effect of MoS2particle size on catalyst activity toward the HDS of DBT.The water to surfactant ratio in the preparation of MoS2/PC-RM should be increased inorder to make larger micelles and therefore larger MoS2 crystallites. The size and productselectivity of a catalyst with larger, but still nano-sized, crystallites would help to furtherclarify the effect of stacking order on the rate constant. Similarly, different reducing agentsshould be tested to aid in the synthesis of a more active MoS2/C catalyst.It was found that MoS2/PC-RM contained K2SO4, hypothesised to have formed due toK+ ions remaining from the activation of the petcoke. The effect of this impurity was notstudied. In order to eliminate any effects of the K2SO4 in MoS2/PC-RM, the activatedpetcoke should be washed more to remove all K+ ions, or the effect that K2SO4 has on theHDS of DBT reaction should be studied in a controlled manner, for example by adding someK2SO4 to a catalytic screen test.Some HDS occurred during the ∼21 minute reactor heat-up period before the systemreached reaction temperature. In this study, the products associated with this pre-reactionare counted as part of the products from the catalysed reaction. To obtain more accuratekinetic data, the reactor should be ramped to reaction temperature and then cooled quickly.The conversion from this test represents the reaction during the heat-up phase, and could besubtracted from the reaction conversion to give catalyst activity for the reaction time only.To elucidate whether cracking of DBT is occurring, the gaseous products of the reaction,which in this study were not studied, should be separated by gas chromatography andidentified. If products other than BP, CHB or THDBT are detected, the reaction kinetics108of the cracking should be calculated to improve the kinetic model.The HG reaction was found to be slow for the catalysts prepared by reverse micelles, soignoring the hydrogenation of BP to CHB or CHB to bicyclohexane in the kinetic model wasacceptable. However, for the MoS2/PC-WI catalyst where HG is very fast, these HG routesshould be included to improve the results of the modelling. One way to further discriminatebetween the HDS and HG activities would be to perform the catalytic tests with toluene ordiphenylethylene as the reactant instead of, or in addition to, DBT, and monitor the levelof hydrogenation. Another option would be to use BP or CHB as the molecule in the feedinstead of DBT to obtain the precise rate constant parameters associated with hydrogenationof these compounds.Finally, an MoS2/AC-WI catalyst should be prepared by incipient wetness impregnationof activated carbon in a preparation similar to that for MoS2/PC-WI. This would enableanother direct comparison of the AC and PC supports. Furthermore, a catalyst preparedby incipient wetness impregnation using the ATTM salt should be prepared on alumina toascertain the effect of the metal-support interaction.109Bibliography[1] Alberta Energy. Oil Sand: Facts and Statistics. 2014. url: http : / / www . (visited on 2015-12-10).[2] R. N. Hunter and J. Read. The Shell Bitumen Handbook. ICE Publishing, 2014.[3] M. R. Gray. Tutorial on Upgrading of Oilsands Bitumen. 2001. url:∼gray/Library/Tutorials/Upgrading/Index.htm (visited on 2015-12-04).[4] Syncrude Canada Ltd. Our Process. 2016. url: http : / / www . syncrude . ca / our -process/ (visited on 2016-11-15).[5] D. K. Banerjee. Oil Sands, Heavy Oil and Bitumen - From Recovery to Refinery.PennWell, 2012.[6] O. P. Strausz. “The Chemistry of the Alberta Oil Sand Bitumen”. In: Symposium onOil Sand and Oil Shale. 1977, pp. 171–176.[7] J. M. Sheremata, M. R. Gray, H. D. Dettman, and W. C. McCaffrey. “QuantitativeMolecular Representation and Sequential Optimization of Athabasca Asphaltenes”.In: Energy & Fuels 18 (2004), pp. 1377–1384.[8] J. M. Hill, A. Karimi, and M. Malekshahian. “Characterization, Gasification, Acti-vation, and Potential Uses for the Millions of Tonnes of Petroleum Coke Produced inCanada Each Year”. In: The Canadian Journal of Chemical Engineering 92 (2014),pp. 1618–1626.110[9] J. G. Speight. “New approaches to hydroprocessing”. In: Catalysis Today 98 (2004),pp. 55–60.[10] S. Papavinasam, P. Rahimi, and S. Williamson. “Corrosion Conditions in the Pathof Bitumen from Well to Wheel”. In: NACE 2012 Northern Area Eastern ConferenceProceedings. 2012.[11] Shell Global Interview with Jose Bravo. “Making sense of the sulphur challenge”. In:Impact 1 (2012), pp. 5–7.[12] Y. H. Shokrlu and T. Babadagli. “In-Situ Upgrading of Heavy Oil/Bitumen DuringSteam Injection by Use of Metal Nanoparticles: A Study on In-Situ Catalysis andCatalyst Transportation”. In: SPE Reservoir Evaluation & Engineering 16 (2013),pp. 333–344.[13] Environment Canada. Sulphur in Diesel Fuel Regulations. 2012. url: (visited on 2015-12-04).[14] Environment Canada. Sulphur in Gasoline Regulations. 2016. url: (visited on 2016-11-15).[15] The European Parliament And The Council Of The European Union. Directive 2003/17/EC Of The European Parliament And Of The Council. 2003. url: on 2016-11-20).[16] J. M. Oelderik, S. T. Sie, and D. Bode. “Progress in the catalysis of the upgrading ofpetroleum residue: A review of 25 years of R&D on Shell’s residue hydroconversiontechnology”. In: Applied Catalysis 47 (1989), pp. 1–24.[17] O. Ocic. In: Oil Refineries in the 21st Century: Energy Efficient, Cost Effective, En-vironmentally Benign. John Wiley & Sons, 2008. Chap. Instruments for DeterminingEnergy and Processing Efficiency of an Oil Refinery.111[18] Siemens. Refining Industry Hydrotreater/HDS Unit. 2013. url: (visited on2016-10-13).[19] M. Beychok. Hydrodesulfurization. Creative commons license: 2013. url: hydrodesulfurization (visited on 2016-11-15).[20] J. Haber, J. Block, and B. Delmon. Manual Of Methods And Procedures For CatalystCharacterization. Tech. rep. 1995.[21] M. Neurock. “Reaction coordinate analyses of transition metal catalyzed C-H andC-S activation steps”. In: Applied Catalysis A: General 160 (1997), pp. 169–184.[22] S. Harris and R. R. Chianelli. “Catalysis by Transition Metal Sulfides: A Theoreticaland Experimental Study of the Relation between the Synergic Systems and the BinaryTransition Metal Sulfides”. In: Journal of Catalysis 98 (1986), pp. 17–31.[23] R. Prins, V. de Beer, and G. A. Somorjai. “Structure and function of the catalystand the promoter in Co-Mo hydrodesulfurization catalysts”. In: Catalysis Reviews:Science and Engineering 31 (1989), pp. 1–41.[24] S. Kasztelan. “A Descriptive Model of Surface Sites on MoS2 (WS2) Particles”. In:Langmuir 6 (1990), pp. 590–595.[25] M. Salmeron, G. Somorjai, A. Wold, R. Chianelli, and K. Liang. “The adsorption andbinding of thiophene, butene and H2S on the basal plane of MoS2 single crystals”. In:Chemical Physics Letters 90 (1982), pp. 105–107.[26] H. Topsøe, R. Candia, N.-y. Topsøe, and B. S. Clausen. “On the State of the Co-Mo-SModel”. In: Bulletin des Socie´e´te´s Chimiques Belges 93 (1984), pp. 783–806.112[27] A. P. Nayak, S. Bhattacharyya, J. Zhu, J. Liu, X. Wu, T. Pandey, C. Jin, A. K.Singh, D. Akinwande, and J.-F. Lin. “Pressure-induced semiconducting to metallictransition in multilayered molybdenum disulphide”. In: Nature Communications 5(2014), pp. 1–9.[28] S. Harris and R. R. Chianelli. “Periodic Effects in Catalysis: The Relation BetweenTrends in Catalytic Activity and Calculated Electronic Structure of Transition-metalSulfides”. In: Chemical Physics Letters 101 (1983), pp. 603–605.[29] R. R. Chianelli. “Fundamental Studies of Transition Metal Sulfide Hydrodesulfuriza-tion Catalysts”. In: Catalysis Reviews: Science and Engineering 26 (1984), pp. 361–393.[30] M. Daage and R. R. Chianelli. “Structure-Function Relations in Molybdenum SulfideCatalysts: The “Rim-Edge” Model”. In: Journal of Catalysis 149 (1994), pp. 414–427.[31] J. V. Lauritsen, M. V. Bollinger, E. Lægsgaard, K. W. Jacobsen, J. K. Nørskov, B. S.Clausen, B. S. Clausen, H. Topsøe, and F. Besenbacher. “Atomic-scale insight intostructure and morphology changes of MoS2 nanoclusters in hydrotreating catalysts”.In: Journal of Catalysis 221 (2004), pp. 510–522.[32] P. G. Moses, B. Hinnemann, H. Topsøe, and J. K. Nørskov. “The hydrogenation anddirect desulfurization reaction pathway in thiophene hydrodesulfurization over MoS2catalysts at realistic conditions: A density functional study”. In: Journal of Catalysis248 (2007), pp. 188–203.[33] H. Topsøe. “The role of Co–Mo–S type structures in hydrotreating catalysts”. In:Applied Catalysis A: General 322 (2007), pp. 3–8.[34] J. V. Lauritsen, M. Nyberg, R. T. Vang, M. V. Bollinger, B. S. Clausen, H. Topsøe,K. W. Jacobsen, E. Lægsgaard, J. K. Nørskov, and F. Besenbacher. “Chemistry of113one-dimensional metallic edge states in MoS2 nanoclusters”. In: Nanotechnology 14(2003), pp. 385–389.[35] J. M. Campelo, D. Luna, R. Luque, J. M. Marinas, and A. A. Romero. “SustainablePreparation of Supported Metal Nanoparticles and Their Applications in Catalysis”.In: ChemSusChem 2 (2009), pp. 18–45.[36] N. Pal and A. Bhaumik. “Mesoporous materials: versatile supports in heterogeneouscatalysis for liquid phase catalytic transformations”. In: RSC Advances 5 (2015),pp. 24363–24391.[37] M. D. Argyle and C. H. Bartholomew. “Heterogeneous Catalyst Deactivation andRegeneration: A Review”. In: Catalysts 5 (2015), pp. 145–269.[38] H. Marsh and F. Rodriguez-Reinoso. Activated Carbon. Elsevier, 2006.[39] C. Liang, P. Ying, and C. Li. “Nanostructured β−Mo2C Prepared by CarbothermalHydrogen Reduction on Ultrahigh Surface Area Carbon Material”. In: Chem. Mater.14 (2002), pp. 3148–3151.[40] H. Chen and Z. Hashisho. “Fast preparation of activated carbon from oil sands cokeusing microwave-assisted activations”. In: Fuel 95 (2012), pp. 178–182.[41] N. Rambabu, R. Azargohar, A. Dalai, and J. Adjaye. “Evaluation and comparisonof enrichment efficiency of physical/chemical activations and functionalized activatedcarbons derived from fluid petroleum coke for environmental applications”. In: FuelProcessing Technology 106 (2013), pp. 501–510.[42] L. D. Virla, V. Montes, J. Wu, S. F. Ketep, and J. M. Hill. “Synthesis of porous carbonfrom petroleum coke using steam, potassium and sodium: Combining treatments tocreate mesoporosity”. In: Microporous and Mesoporous Materials 234 (2016), pp. 239–247.114[43] S. Harris and R. R. Chianelli. “Catalysis by Transition Metal Sulfides: The Relationbetween Calculated Electronic Trends and HDS Activity”. In: Journal of Catalysis86 (1984), pp. 400–412.[44] B. Scheffer, P. Arnoldy, and J. A. Moulijn. “Sulfidability and HydrodesulfurizationActivity of Mo Catalysts Supported on Alumina, Silica, and Carbon”. In: Journal ofCatalysis 112 (1988), pp. 516–527.[45] J. Kibsgaard, B. S. Clausen, H. Topsøe, E. Lægsgaard, J. V. Lauritsen, and F. Be-senbacher. “Scanning tunneling microscopy studies of TiO2-supported hydrotreatingcatalysts: Anisotropic particle shapes by edge-specific MoS2–support bonding”. In:Journal of Catalysis 295 (2012), pp. 146–154.[46] M. Tre´panier, A. K. Dalai, and N. Abatzoglou. “Synthesis of CNT-supported cobaltnanoparticle catalysts using a microemulsion technique: Role of nanoparticle sizeon reducibility, activity and selectivity in Fischer–Tropsch reactions”. In: AppliedCatalysis A: General 374 (2010), pp. 79–86.[47] A. Arteaga, J. L. G. Fierro, P. Grange, and B. Delmon. In: Catalyst Deactivation1987. Elsevier, 1987. Chap. CoMo HDS Catalysts: Simulated Deactivation and Re-generation. Role of Various Regeneration Parameters.[48] S. Louwers and R. R. Prins. “Ni EXAFS Studies of the Ni-Mo-S Structure in Carbon-Supported and Alumina-Supported Ni-Mo Catalysts”. In: Journal of Catalysis 133(1992), pp. 94–111.[49] D. Astruc, F. Lu, and J. R. Aranzaes. “Nanoparticles as Recyclable Catalysts TheFrontier between Homogeneous and Heterogeneous Catalysis”. In: Angewandte Chemie44 (2005), pp. 7852–7872.115[50] B. Yan. “The Properties of Resin Supports and Their Effects on Solid-Phase OrganicSynthesis”. In: Combinatorial Chemistry & High Throughput Screening 1 (1998),pp. 215–229.[51] J. C. Duchet, E. M. van Oers, V. de Beer, and R. R. Prins. “Carbon-Supported SulfideCatalysts”. In: Journal of Catalysis 80 (1983), pp. 386–402.[52] M. Boutonnet, J. Kizling, V. Mintsa-Eya, A. Choplin, R. Touroude, G. Maire, and P.Stenius. “Monodisperse Colloidal Metal Particles from Nonaqueous Solutions: Cat-alytic Behavior In Hydrogenation of But-1-ene of Platinum, Palladium And RhodiumParticles Supported on Pumice”. In: Journal of Catalysis 103 (1987), pp. 95–104.[53] M. Boutonnet, J. Kizling, P. Stenius, and G. Maire. “The Preparation of MonodisperseColloidal Metal Particles from Microemulsions”. In: Colloids and Surfaces 5 (1982),pp. 209–225.[54] J. B. Nagy, A. Gourgue, and E. G. Derouane. Preparation of Monodispersed NickelBoride Catalysts Using Reversed Micellar Systems. Ed. by G. Poncelet, P. A. Jacobs,and P. Grange. Elsevier Science Publishers, 1983, pp. 192–202.[55] J. Nagy, I. Bodart-Ravet, and E. G. Derouane. “Preparation, Characterization andCatalytic Activity of Monodisperse Colloidal Metal Borides”. In: Faraday Discussionsof the Chemical Society 87 (1989), pp. 189–198.[56] M. L. Steigerwald and L. E. Brus. “Synthesis, Stabilization, and Electronic Structureof Quantum Semiconductor Nanoclusters”. In: Annual Review of Materials Science19 (1989), pp. 475–495.[57] M. Boutonnet, J. Kizling, R. Touroude, G. Maire, and P. Stenius. “MonodispersedColloidal Metal Particles from Nonaqueous Solutions: Catalytic Behaviour in Hy-drogenolysis and Isomerization of Hydrocarbons of Supported Platinum Particles”.In: Catalysis Letters 9 (1991), pp. 347–354.116[58] M. P. Pileni, L. Motte, and C. Petit. “Synthesis of Cadmium Sulfide in Situ in ReverseMicelles: Influence of the Preparation Modes on Size, Polydispersity, and Photochem-ical Reactions”. In: Chemistry of Materials 4 (1992), pp. 338–345.[59] K. Marchand, M. Tarret, J. Lechaire, L. Normand, S. Kasztelan, and T. Cseri. “Inves-tigation of AOT-based microemulsions for the controlled synthesis of MoSx nanoparti-cles: an electron microscopy study”. In: Colloids and Surfaces A 214 (2003), pp. 239–248.[60] P. D. Fletcher, A. M. Howe, and B. H. Robinson. “The Kinetics of Solubilisate Ex-change between Water Droplets of a Water-in-oil Microemulsion”. In: Journal of theChemical Society, Faraday Transactions 83 (1987), pp. 985–1006.[61] I. Ravet, A. Gougue, and J. B. Nagy. “Hydrogenation Activity of Colloidal CobaltBoride Particles Synthesized in the CTAB-1-Hexanol-Water Micellar Systems”. In:Surfactants in Solution. Ed. by K. Mittal and P. Botherel. Springer Science + BusinessMedia, 1984, pp. 697–712.[62] D. Tanaka, A. Henke, K. Albrecht, M. Moeller, K. Nakagawa, S. Kitagawa, and J.Groll. “Rapid preparation of flexible porous coordination polymer nanocrystals withaccelerated guest adsorption kinetics”. In: Nature Chemistry 2 (2010), pp. 410–416.[63] S. Modes and P. Lianos. “Luminescence Probe Study of the Conditions Affecting Col-loidal Semiconductor Growth in Reverse Micelles and Water-in-Oil Microemulsions”.In: Journal of Physical Chemistry 93 (1989), pp. 5854–5859.[64] M. Wong, J. K. Thomas, and T. Nowak. “Structure and State of H20 in ReversedMicelles. 3.” In: Journal of the American Chemical Society 99 (1977), pp. 4730–4736.[65] J. Kizling and P. Stenius. “Microemulsions Formed by Water, Aliphatic Hydrocar-bons, and Pentaethylene Glycol Dodecyl Ether: The Temperature Dependence of117Aggregate Size”. In: Journal of Colloid and Interface Science 118 (1987), pp. 482–492.[66] N. Fischer, E. van Steen, and M. Claeys. “Preparation of supported nano-sized cobaltoxide and fcc cobalt crystallites”. In: Catalysis Today 171 (2011), pp. 174–179.[67] H.-F. Eicke, R. Kubik, and H. Hammerich. “On the Temperature-Dependent Viscosityof the L2 Phase in Water/Aerosol OT/Isooctane (W/O) Thermodynamically StableEmulsions”. In: Journal of Colloid and Interface Science 90 (1981), pp. 27–33.[68] B. Pawelec, R. Navarro, J. L. G. Fierro, and P. T. Vasudevan. “Studies of molybde-num sulfide catalyst ex ammonium tetrathiomolybdate: effect of pretreatment on hy-drodesulfurization of dibenzothiophene”. In: Applied Catalysis A: General 168 (1998),pp. 205–217.[69] M. Boutonnet, J. Kizling, R. Touroude, G. Maire, and P. Stenius. “MonodispersedColloidal Metal Particles from Non-Aqueous Solutions: Catalytic Behaviour for theHydrogenation of But-1-ene of Platinum Particles in Solution”. In: Applied Catalysis20 (1986), pp. 163–177.[70] A. Hall, A. Duangchan, and K. J. Smith. “Characterization of Dispersed Hydropro-cessing Catalysts Prepared in Reversed Micelles”. In: Canadian Journal of ChemicalEngineering 76 (1998), pp. 744–752.[71] E. Boakye, L. R. Radovic, and K. Osseo-Asare. “Microemulsion-Mediated Synthesis ofNanosize Molybdenum Sulfide Particles”. In: Journal of Colloid and Interface Science163 (1994), pp. 120–129.[72] F. Zhang and P. T. Vasudevan. “TPD and HYD Studies of Unpromoted and Co-Promoted Molybdenum Sulfide Catalyst ex Ammonium Tetrathiomolybdate”. In:Journal of Catalysis 157 (1995), pp. 536–544.118[73] R. Huirache-Acun˜a, G. Alonso-Nu´n˜ez, F. Paraguay-Delgado, J. Lara-Romero, G.Berhault, and E. M. Rivera-Mun˜oz. “Unsupported trimetallic CoMoW sulfide HDScatalysts prepared by in situ decomposition of sulfur-containing precursors”. In: Catal-ysis Today 250 (2015), pp. 28–37.[74] N. Fischer, M. Minnermann, M. Baeumer, E. van Steen, and M. Claeys. “MetalSupport Interactions in Co3O4/Al2O3 Catalysts Prepared from w/o Microemulsions”.In: Catalysis Letters 142 (2012), pp. 830–837.[75] D. Ganguli and M. Ganguli. Inorganic Particle Synthesis via Macro and Microemul-sions. Springer, 2003.[76] A. Martino, J. P. Wilcoxon, and J. S. Kawola. “Synthesis and Characterization of CoalLiquefaction Catalysts in Inverse Micelles”. In: Energy & Fuels 8 (1994), pp. 1289–1295.[77] J. Marszalek. “The Relationship between Dynamic and Chemical Factors On theMorphology in AOT-Water-Polymer Systems”. PhD thesis. University of SouthernMississippi, 2008.[78] “Ab-initio energy profiles for thiophene HDS on the MoS2 (1010) edge-surface”. In:Hydrotreatment and hydrocracking of oil fractionsProceedings ofthe 2nd InternationalSymposium/7th European Workshop. Ed. by B. Delmon, G. Froment, and P. Grange.Vol. 127. Studies in Surface Science and Catalysis. Elsevier, 1999, pp. 309–317.[79] C. Pophal, F. Kameda, K. Hoshino, S. Yoshinaka, and K. Segawa. “Hydrodesulfuriza-tion of dibenzothiophene derivatives over TiO2-Al2O3 supported sulfided molybdenumcatalyst”. In: Catalysis Today 39 (1997), pp. 21–32.[80] M. Egorova and R. Prins. “Hydrodesulfurization of dibenzothiophene and 4,6-dimethyl-dibenzothiophene over sulfided NiMo/γ-Al2O3, CoMo/γ-Al2O3, and Mo/γ-Al2O3 cat-alysts”. In: Journal of Catalysis 225 (2004), pp. 417–427.119[81] S. T. Oyama and Y.-K. Lee. “The active site of nickel phosphide catalysts for thehydrodesulfurization of 4,6-DMDBT”. In: Journal of Catalysis 258 (2008), pp. 393–400.[82] H. R. Reinhoudt, C. H. M. Boons, A. D. van Langeveld, J. A. R. van Veen, S. T. Sie,and J. A. Moulijn. “On the difference between gas- and liquid-phase hydrotreatingtest reactions”. In: Applied Catalysis A: General 207 (2001), pp. 25–36.[83] N. Kagami, B. M. Vogelaar, A. D. van Langeveld, and J. A. Moulijn. “Reactionpathways on NiMo/Al2O3 catalysts for hydrodesulfurization of diesel fuel”. In: AppliedCatalysis A: General 293 (2005), pp. 11–23.[84] K. Inamura, K. Uchikawa, S. Matsuda, and Y. Akai. “Preparation of active HDScatalysts by controlling the dispersion of active species”. In: Applied Surface Science121/122 (1997), pp. 468–475.[85] A. Niquille-Ro¨thlisberger and R. Prins. “Hydrodesulfurization of 4,6-dimethyldibenzo-thiophene and dibenzothiophene over alumina-supported Pt, Pd, and Pt-Pd cata-lysts”. In: Journal of Catalysis 242 (2006), pp. 207–216.[86] R. S. Kukard and K. J. Smith. “Slurry-Phase Batch Microreactor for HydroconversionStudies”. In: Energy & Fuels 29 (2015), pp. 5274–5281.[87] N. Joshi and A. Lawal. “Hydrodeoxygenation of pyrolysis oil in a microreactor”. In:Chemical Engineering Science 74 (2012), pp. 1–8.[88] N. M. Correa, J. J. Silber, R. E. Riter, and N. E. Levinger. “Nonaqueous PolarSolvents in Reverse Micelle Systems”. In: Chemical Reviews 112 (2012), pp. 4569–4602.[89] H. W. Wang, P. Skeldon, and G. E. Thompson. “XPS studies of MoS2 formationfrom ammonium tetrathiomolybdate solutions”. In: Surface and Coatings Technology91 (1997), pp. 200–207.120[90] J. L. Brito, M. Ilija, and P. Hernandez. “Thermal and reductive decomposition ofammonium thiomolybdates”. In: Thermochimica Acta 256 (1995), pp. 325–338.[91] K. Kanan, H. Yousef, and I. Kayali. “Nanostructured Microemulsion Phase Behav-ior Using AOT or Extended Surfactant Combined with a Cationic Hydrotrope”. In:Journal of Surface Engineered Materials and Advanced Technology 2 (2012), pp. 53–60.[92] Z. Shuxian, W. K. Hall, G. Ertl, and H. Kno¨zinger. “X-Ray Photoemission Study ofOxygen and Nitric Oxide Adsorption on MoS2”. In: Journal of Catalysis 100 (1986),pp. 167–175.[93] J.-G. Choi and L. T. Thompson. “XPS study of as-prepared and reduced molybdenumoxides”. In: Applied Surface Science 93 (1996), pp. 143–149.[94] J. Zemek, S. Hucek, J. A., and I. S. Tilinin. “Photoelectron escape depth”. In: Journalof Electron Spectroscopy and Related Phenomena 76 (1995), pp. 443–447.[95] J. W. Niemantsverdriet. Spectroscopy in Catalysis: An Introduction. John Wiley &Sons, 2008.[96] D. Astruc. In: Nanoparticles and Catalysis. Wiley, 2008. Chap. Transition-metalNanoparticles in Catalysis: From Historical Background to the State-of-the Art.[97] I. Capek. “Preparation of metal nanoparticles in water-in-oil (w/o) microemulsions”.In: Advances in Colloid and Interface Science 110 (2004), pp. 49–74.[98] M. N. Luwang, R. S. Ningthoujam, N. S. Singh, R. Tewari, S. K. Srivastava, and R. K.Vatsa. “Surface chemistry of surfactant AOT-stabilized SnO2 nanoparticles and effectof temperature”. In: Journal of Colloid and Interface Science 349 (2010), pp. 27–33.121[99] X.-G. Lei, Z. Guo-Hua, Y.-C. Liu, and N. J. Turro. “Influence of Binding Strengthof Added Electrolytes on the Properties of Micelles and of Micellized Radical Pairs”.In: Langmuir 8 (1992), pp. 475–480.[100] G. C. Stevens and T. Edmonds. “Catalytic Activity of the Basal And Edge Planes ofMolybdenum Disulphide”. In: Journal of the Less-Common Metals 54 (1977), pp. 321–330.[101] J. Whelan, I. Banu, G. E. Luckachan, N. D. Banu, S. Stephen, A. Tharalekshmy, S. AlHashimi, R. V. Vladea, M. S. Katsiotis, and S. M. Alhassan. “Influence of decomposi-tion time and H2 pressure on properties of unsupported ammonium tetrathiomolybdate-derived MoS2 catalysts”. In: Journal of Analytical Science and Technology 6 (2015),pp. 1–8.[102] S. K. Jana, P. Mukhopadhyay, S. Ghosh, S. Kabi, A. Bag, R. Kumar, and D. Biswas.“High-resolution X-ray diffraction analysis of AlxGa1−xN/ InxGa1−xN/GaN on sap-phire multilayer structures: Theoretical, simulations, and experimental observations”.In: Journal of Applied Physics 115 (2014), pp. 174507-1–174507-13.[103] M. Houalla, N. K. Nag, A. V. Sapre, D. H. Broderick, and B. C. Gates. “Hydrodesulfu-rization of dibenzothiophene catalyzed by sulfided CoO-MoO3γ-Al2O3: The reactionnetwork”. In: AIChE Journal 24.6 (1978), pp. 1015–1021.[104] E. J. M. Hensen, P. J. Kooymna, Y. van der Meer, A. M. van der Kraan, V. H. J. deBeer, J. A. R. van Veen, and R. A. van Santen. “The Relation between Morphologyand Hydrotreating Activity for Supported MoS2 Particles”. In: Journal of Catalysis199 (2001), pp. 224–235.[105] C. T. Tye and K. J. Smith. “Hydrodesulfurization of dibenzothiophene over exfoliatedMoS2 catalyst”. In: Catalysis Today 116 (2006), pp. 461–468.122[106] G. A. Camacho-Bragado, J. L. Elechiguerra, O. A., S. Fuentes, D. Galvan, and M. J.Yacaman. “Structure and catalytic properties of nanostructured molybdenum sul-fides”. In: Journal of Catalysis 234 (2005), pp. 182–190.[107] R. Romero-Rivera, A. G. Camacho, M. Del Valle, G. Alonso, S. Fuentes, and J. Cruz-Reyes. “HDS of DBT with Molybdenum Disulfide Catalysts Prepared by In SituDecomposition of Alkyltrimethylammonium Thiomolybdates”. In: Top Catalysis 54(2011), pp. 561–567.[108] H. Topsøe, B. S. Clausen, and F. E. Massoth. “Hydrotreating Catalysis”. In: Catalysis:Science and Technology. Ed. by J. R. Anderson and M. Boudart. Springer BerlinHeidelberg, 1996, pp. 1–269.123Appendix ACalculationsA.1 Feed calculationIt was desired to prepared a feed which consisted of 2 wt.% DBT, enough CS2 to keep themolybdenum sulphided but not too much since it is a poison to HDS. The feed volume duringeach run should be 150µL with a Mo concentration of 900 ppm.Calculating the amount of catalyst is straight-forward. Assuming a 10 wt.% Mo content,and that the bulk of the feed is composed of decalin which has a density of 0.9 g/mL:mcatalyst = (900 gMo106 gfeed)(0.9 gfeedmL)(mL1000 µL)(150µL)(1 gcatalyst0.1 gMo)= 1.215 mgTo calculate how much sulphiding agent CS2 should be added, it was decided to use a0.5% mole fraction of H2S in the system. First, however, the amount of H2 in the systemneeded to be calculated. The number of moles of hydrogen in the system at average reaction124temperature (636 K) and pressure (5.08 MPa) was estimated using the ideal gas law:nH2 =PVRT=(5.08x106 Pa)[pi(2x10−3m)2(0.25m)](8.314m3PaK mol)(636.3K)= 3.0x10−3 molesIt was decided to use a 0.5% mole fraction of H2S in the system:nH2S = 0.005 nH2= 1.5x10−5 molesThe amount of liquid CS2 this corresponds to:VCS2 = (1.5x10−5 molH2S)(1molCS22molH2S)(76.13 gCS2molCS2)(mL1.266gCS2)= 4.5 x 10−4 mL= 0.45 µLSeeing how small the volume of CS2 is to the volume of decalin, the 2 wt.% amount ofDBT was calculated purely on the weight of decalin:mDBT = (0.02 gDBT1 gfeed)(0.9 gfeedmL)(mL1000 µL)(150µL)= 2.7 mgTo reduce the error associated with the small volumes, a larger batch of feed was prepared.Table A.1 summarises the feed which was prepared and used during the activity testingdescribed in Chapter 4.125Table A.1: Composition of feed used for activity testingDecalin CS2 DBTMass (mg) 13261.7 56 271.9wt.% 97.59 0.41 2.00A.2 Calculations for catalysts prepared by the micellemethodThe calculations to prepare a 10 wt.% Mo catalyst on PC using a 0.3M surfactant in oilsolution, water:surfactant ratio of 20, and a 5 mM ATTM solution are outlined here.Fist, to prepare a 0.3 M Igepal in cyclohexane solution, the number of moles of Igepal tobe added to the already measured 367 mL of cyclohexane:mIgepal = (0.367 Loil)(0.3molIgepalLoil)(441 gIgepalmolIgepal)= 48.55 gIgepalNext, the calculation of the total amount of water to be added, which includes thecontributions from the 0.5 M H2SO4 and 5 mM ATTM solutions:mwater = (48.55 gIgepal)(molIgepal441 gIgepal)(20molH2O1molIgepal)(15.015 gH2OmolH2O)= 39.67 gH2OThe constraints of a 5 x 10−3 molar solution of ATTM, a 0.5 M H2SO4 solution, aratio of 10 H+ ions per molecule of ATTM, and total volume of 39.67 g of water weresolved simultaneously to obtain the solutions for the amount of each solution needed. In the126preparation outlined in section 2.1.3, the obtained volumes were 37.95 mL of 5 mM ATTMsolution and 1.9 mL of 0.5 M H2SO4 solution, with approximately 0.5% error.From 37.95 mL of 5 mM ATTM solution, the mass of formed Mo was calculated:mMo = (5 x 10−3 molATTML)(37.95 x 10−3 L)(1molMo1molATTM)(95.94 gMomolMo)= 0.0182 gMoThen the total mass of catalyst with a 10% Mo loading will be:mcatalyst = (0.0182 gMo)(gcatalyst0.1 gMo)= 0.182 gcatalystOf that 0.182 g of catalyst, some will be MoS2 with the balance of C support. To calculatethe mass of MoS2:mMoS2 = (0.0182 gMo)(molMo95.94 gMo)(molMoS2molMo)(160.07 gMoS2molMoS2)= 0.0304 gMoS2Then to calculate the amount of PC support that needs to be added:mPC = mcatalyst −mMoS2= 0.182 gcatalyst − 0.0304 gMoS2= 0.1516 gPC127A.3 Calculations for catalysts prepared by the incipi-ent wetness impregnation methodThe calculations to prepare a 10 wt.% Mo catalyst on PC by incipient wetness impregnationare outlined here. They are similar to the calculations for preparing MoS2/PC by reversemicelles, only simpler.A 0.3 g amount of catalyst with a 10 wt.% Mo loading contains 0.03 g of Mo. To calculatethe mass of MoS2 this corresponds to:mMoS2 = (0.03 gMo)(molMo95.94 gMo)(molMoS2molMo)(160.07 gMoS2molMoS2)= 0.050 gMoS2To calculate the amount of ATTM to be added to the 0.250 g of PC to make the 10 wt.%MoS2/PC by wetness impregnation:mATTM = (0.05 gMoS2)(molMoS2160.07 gMoS2)(molATTMmolMoS2)(260.28 gMoS2molATTM)= 0.0814 gATTMA.4 Calculation to convert Tye’s k dataTye reported kDBT in units of [mL mol Mo−1e s−1] [105]. To convert to units of [molDBT s−1 g−1catalyst],the following calculation was performed, shown for the AHM derived MoS2 (Note: the re-128searcher used a 900 ppm feed of DBT in n-hexadecane, whose density is 0.77 g/mL):kDBT =791 mLmol Moe s= (791 mLmol Moe s)(0.107 mol Moemol Mo)(900 g DBT106 g feed)(0.77 g feedmL)(mol Mo95.94 g Mo)(mol DBT184.26 g DBT)= 3.32 x 10−6 molDBT s−1 g−1catalyst129Appendix BMATLAB CodesThe MATLAB code for a Levenberg-Marquardt Non-Linear Regression was created by R.Shrager and modified by R. Muzic and A. Jutan. The code below has been used by thisresearch group for many years, and was combined with a series of ODEs prepared describingthe HDS of DBT reaction by the author. To calculate the rate constants presented inChapter 4, the model was solved simultaneously using a 4th order Runge–Kutta algorithm.B.1 Main bodydiary MicellePC375Clog.txtclear allglobal nvar nx x0 y0;global verbose;global n1 n2 n3 H2;verbose(1:2) = 1;% x is the indep variable vector e.g. time measurements% y is matrix of responses% Columns of y are responses y1, y2 (e.g. mol frac of component 1 and 2)% Rows of y are y values at the value of the indep variable (time) in x% First row of y is initial value of response% The program uses the Levenberg-Marquardt method to estimate parameters% and calculate statistics - done in leasqr.m and dfdp.m% These two matlab m-files are designed for single response130% The input data is re-arranged to yield a single response vector y% The L-M requires the model to be calculated - this is done in modelmulti.m% and assumes the model is a series of ODEs, with the number of ODEs equal% to the number of responses. The ODEs are calculated in ODEfunm.m.% Note that this function must use the correct model for each y % Also note that the rateconstants are in units of per minute% INPUT% input number of responses% Order of reactions:n1=1; % order of CAn2=1; % order of CBn3=1; % order of CCH2=1; % Concentration of Hydrogen in liquid.T=[0 3600 7200 10800]’; %time in secondsnt=length (T);x(1:nt-1)=T(2:nt);nx=length(x);CAX=[1 0.91254 0.83929 0.74013]’; % Conc of DBT (mole frac.)CBX=[0 0.07113 0.13033 0.21439]’; % Conc of BP (mole frac.)CCX=[0 0 0.01279 0.02719]’; % Conc of CHB (mole frac.)CDX=[0 0.01632 0.01758 0.01828]’; % Conc of THDBT (mole frac.)for j=1:nt-1y1(j)=CAX(j+1);y2(j)=CBX(j+1);y3(j)=CCX(j+1);y4(j)=CDX(j+1);endnvar=4;x0=0;oldx=x;nx = length(x);y = [y1’ y2’ y3’ y4’];newy=y(:);oldy=reshape(newy,nx,nvar);x=x’;newx=[x;x;x;x];y01(1:nx)=CAX(1);y02(1:nx)=CBX(1);y03(1:nx)=CCX(1);y04(1:nx)=CDX(1);newy0=[y01’;y02’;y03’;y04’];%INPUT DATA NOW IN CORRECT COLUMN FORMAT131y0=newy0;x=newx;y=newy;%Provide initial parameter guessestheta=[0.002 0.002 0.03];np=length(theta);pin=theta;%Begin calculation by calling L-M least squares routine[f,p,kvg,iter,corp,covp,covr,stdresid,Z,r2]=leasqr(x,y,pin‘modelmulti’)disp(‘RESPONSE:’)if kvg ==1disp (‘PROBELM CONVERGED’)elseif kvg == 0disp(‘PROBLEM DID NOT CONVERGE’)endformat shortEngoldf=reshape(f,nx,nvar);oldr=reshape(y-f, nx, nvar);tspan=0:0.1:180;C0 = [y0(1);CBX(1);CCX(1);CDX(1)];[t,Y]=ode45(@ODEfunm,tspan,C0,[],p);CA=Y(:,1);CB=Y(:,2);CC=Y(:,3);CD=Y(:,4);disp (‘X-values:’)disp (oldx’)disp (‘Y-values:’)disp (oldy)disp (‘f-values - i.e. model calculated responses:’)disp (oldf)disp (‘Residuals:’)disp (oldr)disp (‘Standardized residuals:’)disp (stdresid)disp (‘Estimated parameter values are:’)disp (p)disp (‘Covariance of estimated parameters - sqrt of diagonal gives CL’)disp (covp)disp (‘R2 values are:’)disp (r2)figure132ax1 = gca;hold onp1=plot(oldx(:),oldy(:,1),‘bx’);title(‘T=375C, Micelle/PC’)xlabel(‘Time (min)’)ylabel(‘Concentration of DBT (mol frac)’)plot(tspan,CA,‘b–’)ax2 = axes(‘Position’,get(ax1,‘Position’),...‘XAxisLocation’,‘top’,...‘YAxisLocation’,‘right’,...‘Color’,‘none’,...‘XColor’,‘none’,‘YColor’,‘k’);ylabel(ax2,‘Concentration of BP, CHB, THDBT (mol frac)’)linkaxes([ax1 ax2],’x’);hold onp2=plot(oldx(:),oldy(:,2),‘k*’,‘Parent’,ax2);p3=plot(oldx(:),oldy(:,3),‘r+’,‘Parent’,ax2);p4=plot(oldx(:),oldy(:,4),‘gd’,‘Parent’,ax2);plot(tspan,CB,‘k–’,‘Parent’,ax2);plot(tspan,CC,‘r–’,‘Parent’,ax2);plot(tspan,CD,‘g–’,‘Parent’,ax2);legend([p1,p2,p3,p4],‘DBT’,‘Biphenyl’,‘CyclohexylBenzene’,‘1,2,3,4-THDBT’)diary offB.2 Modelmulti codefunction f = modelmulti (x,pin)% Solve a simple system of 2 ODE’s - 2 response variables% find the solution at sepcified x values - corresponding to measured data% first data point in x corresponds to initial conditionglobal nvar nx x0 y0global verboseglobal n1 n2 n3 n4 H2nxx=length(x);yzero=reshape(y0,nx,nvar);for i = 1:nxxf = x(i);xoo=x0;yzed=yzero(i,:);[xmodel,ymodel] = ode45 (@ODEfunm,[xoo,xf], yzed,[],pin);133yfinal(i,:)=ymodel(end,:);endf = yfinal(:);B.3 ODEfunm codefunction yprime=ODEfunm(xatx,yatx,p)global nvar nx x0 y0 xstepglobal verboseglobal n1 n2 n3 n4 H2k1=p(1);k2=p(2);k3=p(3);yp(1)=-k1*yatx(1)-k2*yatx(1);yp(2)=k1*yatx(1);yp(3)=k3*yatx(4);yp(4)=-k3*yatx(4)+k2*yatx(1);yprime =[yp(1);yp(2)’;yp(3)’;yp(4)’];B.4 Calculation of Jacobian matrixfunction prt=dfdp(x,f,p,dp,func)% numerical partial derivatives (Jacobian) df/dp for use with leasqr% ——–INPUT VARIABLES———% x=vec or matrix of indep var(used as arg to func) x=[x0 x1 ....]% f=func(x,p) vector initialsed by user before each call to dfdp% p= vec of current parameter values% dp= fractional increment of p for numerical derivatives% dp(j)>0 central differences calculated% dp(j)<0 one sided differences calculated% dp(j)=0 sets corresponding partials to zero; i.e. holds p(j) fixed% func=string naming the function (.m) file% e.g. to calc Jacobian for function expsum prt=dfdp(x,f,p,dp,‘expsum’)%——–OUTPUT VARIABLES———% prt= Jacobian Matrix prt(i,j)=df(i)/dp(j)%================================m=length(x);n=length(p); %dimensionsps=p; prt=zeros(m,n);del=zeros(n,1); % initialise Jacobian to Zero134for j=1:ndel(j)=dp(j) .*p(j); %cal delx=fract(dp)*param value(p)if p(j)==0del(j)=dp(j); %if param=0 delx=fractionendp(j)=ps(j) + del(j);if del(j)∼=0, f1=feval(func,x,p);if dp(j) <0, prt(:,j)=(f1-f)./del(j);elsep(j)=ps(j)- del(j);prt(:,j)=(f1-feval(func,x,p))./(2 .*del(j));endendp(j)=ps(j); %restore p(j)endreturnB.5 Least square codefunction [f,p,kvg,iter,corp,covp,covr,stdresid,Z,r2]= ...leasqr(x,y,pin,F,stol,niter,wt,dp,dFdp,options)%function[f,p,kvg,iter,corp,covp,covr,stdresid,Z,r2]=% leasqr(x,y,pin,F,stol,niter,wt,dp,dFdp,options)% Version 3.beta% {}= optional parameters% Levenberg-Marquardt nonlinear regression of f(x,p) to y(x), where:% x=vec or mat of indep variables, 1 row/observation: x=[x0 x1....xm]% y=vec of obs values, same no. of rows as x.% wt=vec(dim=length(x)) of statistical weights. These should be set% to be proportional to (sqrt of var(y))ˆ-1; (That is, the covariance% matrix of the data is assumed to be proportional to diagonal with diagonal% equal to (wt.ˆ2)ˆ-1. The constant of proportionality will be estimated.),% default=ones(length(y),1).% pin=vector of initial parameters to be adjusted by leasqr.% dp=fractional incr of p for numerical partials,default= .001*ones(size(pin))% dp(j)>0 means central differences.% dp(j)<0 means one-sided differences.% Note: dp(j)=0 holds p(j) fixed i.e. leasqr wont change initial guess: pin(j)% F=name of function in quotes, of the form y=f(x,p)% dFdp=name of partials M-file in quotes default is prt=dfdp(x,f,p,dp,F)135% stol=scalar tolerances on fractional improvement in ss,default stol=.0001% niter=scalar max no. of iterations, default = 20% options=matrix of n rows (same number of rows as pin) containing% column 1: desired fractional precision in parameter estimates.% Iterations are terminated if change in parameter vector (chg) on two% consecutive iterations is less than their corresponding elements% in options(:,1). [ie. all(abs(chg*current parm est) <options(:,1))% on two consecutive iterations.], default = zeros().% column 2: maximum fractional step change in parameter vector.% Fractional change in elements of parameter vector is constrained to be% at most options(:,2) between sucessive iterations.% [ie. abs(chg(i))=abs(min([chg(i) options(i,2)*current param estimate])).],% default = Inf*ones().% OUTPUT VARIABLES% f=vec function values computed in function func.% p=vec trial or final parameters. i.e, the solution.% kvg=scalar: =1 if convergence, =0 otherwise.% iter=scalar no. of interations used.% corp= correlation matrix for parameters% covp= covariance matrix of the parameters% covr = diag(covariance matrix of the residuals)% stdresid= standardized residuals% Z= matrix that defines confidence region% r2= coefficient of multiple determination% All Zero guesses not acceptable% Richard I. Shrager (301)-496-1122% Modified by A.Jutan (519)-679-2111% Modified by Ray Muzic 14-Jul-1992% 1) add maxstep feature for limiting changes in parameter estimates% at each step.% 2) remove forced columnization of x (x=x(:)) at beginning. x could be% a matrix with the ith row of containing values of the% independent variables at the ith observation.% 3) add verbose option% 4) add optional return arguments covp, stdresid, chi2% 5) revise estimates of corp, stdev% Modified by Ray Muzic 11-Oct-1992% 1) revise estimate of Vy. remove chi2, add Z as return values% Modified by Ray Muzic 7-Jan-1994% 1) Replace ones(x) with a construct that is compatible with versions% newer and older than v 4.1.% 2) Added global declaration of verbose (needed for newer than v4.x)136% 3) Replace return value var, the variance of the residuals with covr,% the covariance matrix of the residuals.% 4) Introduce options as 10th input argument. Include% convergence criteria and maxstep in it.% 5) Correct calculation of xtx which affects coveraince estimate.% 6) Eliminate stdev (estimate of standard deviation of parameter% estimates) from the return values. The covp is a much more% meaningful expression of precision because it specifies a confidence% region in contrast to a confidence interval.. If needed, however,% stdev may be calculated as stdev=sqrt(diag(covp)).% 7) Change the order of the return values to a more logical order.% 8) Change to more efficent algorithm of Bard for selecting epsL.% 9) Tighten up memory usage by making use of sparse matrices (if% MATLAB version >= 4.0) in computation of covp, corp, stdresid.% Modified by Sean Brennan 17-May-1994% verbose is now a vector:% verbose(1) controls output of results% verbose(2) controls plotting intermediate results% References:% Bard, Nonlinear Parameter Estimation, Academic Press, 1974.% Draper and Smith, Applied Regression Analysis, John Wiley and Sons, 1981.%set default args% argument processingplotcmd=‘plot(x(:,1),y,‘’+‘’,x(:,1),f); shg’;%if (sscanf(version,‘%f’) >= 4),vernum= sscanf(version,‘%f’);if vernum(1) >= 4,global verboseplotcmd=‘plot(x(:,1),y,‘’+‘’,x(:,1),f); figure(gcf)’;end;if (exist(‘OCTAVE VERSION’))global verboseend;if(exist(‘verbose’)∼=1), %If verbose undefined, print nothingverbose(1)=0 %This will not tell them the resultsverbose(2)=0 %This will not replot each loopend;if (nargin <= 8), dFdp=‘dfdp’; end;if (nargin <= 7), dp=.001*(pin*0+1); end; %DTif (nargin <= 6), wt=ones(length(y),1); end; % SMB modificationif (nargin <= 5), niter=20; end;if (nargin == 4), stol=.0001; end;137y=y(:); wt=wt(:); pin=pin(:); dp=dp(:); %change all vectors to columns% check data vectors- same length?m=length(y); n=length(pin); p=pin;[m1,m2]=size(x);if m1∼=m ,error(‘input(x)/output(y) data must have same number of rows ’) ,end;if (nargin <= 9),options=[zeros(n,1) Inf*ones(n,1)];nor = n; noc = 2;else[nor noc]=size(options);if (nor ∼= n),error(‘options and parameter matrices must have same number of rows’),end;if (noc ∼= 2),options=[options(noc,1) Inf*ones(noc,1)];end;end;pprec=options(:,1);maxstep=options(:,2);% set up for iterationsf=feval(F,x,p); fbest=f; pbest=p;r=wt.*(y-f);sbest=r’*r;nrm=zeros(n,1);chgprev=Inf*ones(n,1);kvg=0;epsLlast=1;epstab=[.1 1 1e2 1e4 1e6];% do iterationsfor iter=1:niter,pprev=pbest;prt=feval(dFdp,x,fbest,pprev,dp,F);r=wt.*(y-fbest);sprev=sbest;sgoal=(1-stol)*sprev;for j=1:n,if dp(j)==0,nrm(j)=0;elseprt(:,j)=wt.*prt(:,j);nrm(j)=prt(:,j)’*prt(:,j);if nrm(j)>0,nrm(j)=1/sqrt(nrm(j));138end;endprt(:,j)=nrm(j)*prt(:,j);end;% above loop could ? be replaced by:% prt=prt.*wt(:,ones(1,n));% nrm=dp./sqrt(diag(prt’*prt));% prt=prt.*nrm(:,ones(1,m))’;[prt,s,v]=svd(prt,0);s=diag(s);g=prt’*r;for jjj=1:length(epstab),epsL = max(epsLlast*epstab(jjj),1e-7);se=sqrt((s.*s)+epsL);gse=g./se;chg=((v*gse).*nrm);% check the change constraints and apply as necessaryochg=chg;for iii=1:n,if (maxstep(iii)==Inf), break; end;chg(iii)=max(chg(iii),-abs(maxstep(iii)*pprev(iii)));chg(iii)=min(chg(iii),abs(maxstep(iii)*pprev(iii)));end;if (verbose(1) & any(ochg ∼= chg)),disp([‘Change in parameter(s): ’ ...sprintf(‘%d ’,find(ochg ∼= chg)) ‘were constrained’]);end;aprec=abs(pprec.*pbest);% ss=scalar sum of squares=sum((wt.*(y-f))ˆ2).if (any(abs(chg) >0.1*aprec)), % only worth evaluating function ifp=chg+pprev; % there is some non-miniscule changef=feval(F,x,p);r=wt.*(y-f);ss=r’*r;if ss<sbest,pbest=p;fbest=f;sbest=ss;end;if ss<=sgoal,break;end;139end;end;epsLlast = epsL;% if (verbose(2)),% eval(plotcmd);% end;if ss<eps,break;endaprec=abs(pprec.*pbest);% [aprec chg chgprev]if (all(abs(chg) <aprec) & all(abs(chgprev) <aprec)),kvg=1;if (verbose(1)),fprintf(‘Parameter changes converged to specified precision\n’);end;break;elsechgprev=chg;end;if ss>sgoal,break;end;end;% set return valuesp=pbest;f=fbest;ss=sbest;kvg=((sbest>sgoal)—(sbest<=eps)—kvg);if kvg ∼= 1 , disp(‘ CONVERGENCE NOT ACHIEVED! ’), end;% CALC VARIANCE COV MATRIX AND CORRELATION MATRIX OF PARAMETERS% re-evaluate the Jacobian at optimal valuesjac=feval(dFdp,x,f,p,dp,F);msk = dp ∼= 0;n = sum(msk); % reduce n to equal number of estimated parametersjac = jac(:, msk); % use only fitted parameters%% following section is Ray Muzic’s estimate for covariance and correlation%% assuming covariance of data is a diagonal matrix proportional to%% diag(1/wt.ˆ2).%% cov matrix of data est. from Bard Eq. 7-5-13, and Row 1 Table 5.1if vernum(1) >= 4,Q=sparse(1:m,1:m,(0*wt+1)./(wt.ˆ2)); % save memory140Qinv=inv(Q);elseQinv=diag(wt.*wt);Q=diag((0*wt+1)./(wt.ˆ2));end;resid=y-f; %un-weighted residualscovr=resid’*Qinv*resid*Q/(m-n); %covariance of residualsVy=1/(1-n/m)*covr; % Eq. 7-13-22, Bard %covariance of the datajtgjinv=inv(jac’*Qinv*jac); %argument of inv may be singularcovp=jtgjinv*jac’*Qinv*Vy*Qinv*jac*jtgjinv; % Eq. 7-5-13, Bard %cov of parm estd=sqrt(abs(diag(covp)));corp=covp./(d*d’);covr=diag(covr); % convert returned values to compact storagestdresid=resid./sqrt(diag(Vy)); % compute then convert for compact storageZ=((m-n)*jac’*Qinv*jac)/(n*resid’*Qinv*resid);%% alt. est. of cov. mat. of parm.:(Delforge, Circulation, 82:1494-1504, 1990%%disp(‘Alternate estimate of cov. of param. est.’)%%acovp=resid’*Qinv*resid/(m-n)*jtgjinv%Calculate Rˆ2 (Ref Draper & Smith p.46)r=corrcoef(y,f);if (exist(‘OCTAVE VERSION’))r2=rˆ2;elser2=r(1,2).ˆ2;end% if someone has asked for it, let them have itif (verbose(2)), eval(plotcmd); end,if (verbose(1)),disp(‘ Least Squares Estimates of Parameters’)disp(p’)disp(‘ Correlation matrix of parameters estimated’)disp(corp)disp(‘ Covariance matrix of Residuals’ )disp(covr)disp(‘ Correlation Coefficient Rˆ2’)disp(r2)sprintf(‘ 95%% conf region: F(0.05)(%.0f,%.0f)>= delta pvec”*Z*delta pvec’,n,m-n)Z% runs test according to Bard. p 201.n1 = sum((f-y) <0);n2 = sum((f-y) >0);nrun=sum(abs(diff((f-y)<0)))+1;141if ((n1>10)&(n2>10)), % sufficent data for test?zed=(nrun-(2*n1*n2/(n1+n2)+1)+0.5)/(2*n1*n2*(2*n1*n2-n1-n2).../((n1+n2)ˆ2*(n1+n2-1)));if (zed <0),prob = erfc(-zed/sqrt(2))/2*100;disp([num2str(prob) ‘% chance of fewer than ’ num2str(nrun) ‘ runs.’]);else,prob = erfc(zed/sqrt(2))/2*100;disp([num2str(prob) ‘% chance of greater than ’ num2str(nrun) ‘ runs.’]);end;end;end% A modified version of Levenberg-Marquardt% Non-Linear Regression program previously submitted by R.Shrager.% This version corrects an error in that version and also provides% an easier to use version with automatic numerical calculation of% the Jacobian Matrix. In addition, this version calculates statistics% such as correlation, etc....% Version 3 Notes% Errors in the original version submitted by Shrager (now called version 1)% and the improved version of Jutan (now called version 2) have been corrected.% Additional features, statistical tests, and documentation have also been% included along with an example of usage. BEWARE: Some the the input and% output arguments were changed from the previous version.% Ray Muzic <>% Arthur Jutan <>142Appendix CSample PreparationC.1 Petcoke activation procedureThe following procedure was developed by Haiyan Wang and used to activate the PC usedas a support in this study. The activation procedure has 3 steps and results in activated PCwith the physical properties presented in Table 3.1.C.1.1 Obtaining petcoke with a certain particle sizeThe first step is to crush the petcoke and sieve it to obtain petcoke within a certain particlediameter range:1. Add raw petcoke into a mortar, crush until pieces are smaller2. Use two sieve trays to sieve the petcoke into the desired particle diameter of 106−180µm3. The larger pieces can be crushed more, the smaller pieces are not used for this activation143C.1.2 Chemical activationThe second step is the actual activation of petcoke with potassium hydroxide. Both thepetcoke and KOH should be stored in a desiccator to enable accurate mass measurements:1. Weigh out x g of crushed petcoke, set aside2. Weigh out y g of KOH pellets so that z = x :y ∼ 3 (i.e. 3 parts KOH to 1 part petcoke)3. Place KOH into a mortar and crush with a pestle to turn KOH into a fine powder (Note:use a powder mask or respirator to avoid inhaling KOH, a very caustic substance)4. Add the petcoke to the mortar and physically mix the two powders5. Place the mortar in an oven set to 100◦C and leave it for two hours to evaporate themoisture from the samples introduced during the mixing process6. Weight a clean ceramic boat, add the dried sample to it, and obtain the mass of thedried sample7. From the ratio z, calculate the amount of petcoke (m1) present in the ceramic boat8. Place the ceramic boat into the exact centre of a horizontal furnace and seal the furnace9. Purge the furnace with N2 for 10 minutes10. Begin the heating program: ramp at 5◦C/minute to 800◦C, hold for 2 hoursC.1.3 Washing the petcokeThe last step is the removal of excess KOH:1. Line a Buchner funnel with filter paper and wet it with deionized (DI) water2. Quantitatively transfer the PC into the funnel, using DI water as needed1443. Add DI water to the funnel and mix the slurry to solubilize KOH4. Connect a pump to the filter flask and turn it on5. Disengage the pump and add more water to wash the sample, turn the pump on again6. Repeat step 5 3 times7. Rinse the sample alternating 1M HCl with DI water until the runoff water has a neutralpH8. Transfer the washed sample to a glass dish and dry the sample in an oven set to 100◦Covernight9. Weigh the amount of activated petcoke (m2) obtained10. The carbon burn-off can be calculated from m1−m2m1C.2 ICP-OES sample preparationThe following procedure was developed by Lucie Solnickova and is a guideline for preparingsamples for ICP-OES analysis in the EOS lab. The samples discussed below are molybdenumsulphide on carbon; this method can be adapted to suit other samples. Contact for ICP-OEStechnician: Maureen Soon msoon@eos.ubc.caC.2.1 PurposeHave a ∼ 5 wt.% Mo in MoS2/C catalyst. Want a solution containing ∼ 25 ppm Mo in 2vol.% HNO3 and bulk DI water. Need 6 standard solutions: 0, 5, 12.5, 25, 37.5, 50 ppmmade using AHM in 2 vol.% HNO3 and bulk DI water.145C.2.2 General cautions1. When combining acid and water, always add acid to water2. Always add acid dropwise, no matter what you are adding it to3. Glassware should either be clean and dry (for solids), or clean and rinsed with theliquid which will make up the bulk of the final volumeC.2.3 Required equipment1. Fumehood2. 1 x 100mL beaker per sample3. 1 x hotplate per 3 samples4. 1 x 25.00 mL volumetric flask per 1 or 2 samples5. 1 x 15 mL Falcon tube per samples6. 1 x 50.00 mL volumetric flask7. Pasteur pipettes calibrated to 1mL8. Concentrated nitric acid9. Concentrated hydrochloric acid10. Ultrapure water11. Ammonium heptamolybdate146C.2.4 Making the matrix solutionThe ICP OES internal standard is Europeum in in 2 vol% HNO3. The instrument is unfitto handle stronger acid. The procedure below outlines the preparation of a 2 vol% HNO3solution, which must be the matrix for all samples analysed. This preparation will sufficefor 10 samples and 6 standards, and should be freshly prepared for every sample analysis.1. To a 500.00 mL volumetric flask add 250 mL ultra-pure water2. Add 14.2 mL of 69% HNO3 to the 500.00 mL volumetric flask with swirling, checkingthat it does not get too hot3. Make it up to the mark with ultra-pure water to obtain 2 vol.% HNO3 in water solutionC.2.5 Digesting the sampleCAUTION! Boiling acid is very dangerous, exercise caution and use a fume hood!1. Place ∼ 10 mg of sample into a 100 mL beaker, weighing to at least 3 significant figures2. Add 10mL conc HNO3 to the solidˆ Bring to a boil and reflux for 10 mins, cool3. Add 2 mL conc HNO3 and 6 mL conc HClˆ Bring to a boil and reflux for 10 mins, coolˆ There will be much bubbling here as CO2 is produced4. Add 2 mL conc HNO3 and 6 mL conc HClˆ Bring to a boil and reflux for 10 mins, coolˆ Again a lot of bubbling1475. Add 1 mL conc HNO3 and 3 mL conc HClˆ Bring to a boil and reflux for 10 mins, cool6. Add 1 mL conc HNO3 and 3 mL conc HClˆ Bring to a boil and reflux for 10 mins, cool7. Add 1 mL conc HNO3 and 3 mL conc HClˆ Bring to a boil and reflux for 10 mins, cool8. If there are solids, filter the contents through an ashless filter paper into another beaker9. Boil the solution until it is almost dry10. Add 2 mL of 2 vol.% HNO3ˆ Boil this to dryness to remove all the acid, cool the beakerˆ The metal will remain in the beaker, but all acid must be removed to be sure thatthe final solution contains only 2 vol.% HNO3 solution11. Using 5mL portions of 2 vol.% HNO3 solution, dissolve all the solids on the beaker andtransfer all to a 25.00 mL volumetric flask12. Make the solution up to the line with 2 vol.% HNO3 solution13. Transfer the solution to a Falcon tube the day of the measurementˆ This solution can be stored for approximately a week in a well-covered glasscontainer to prevent evaporation148C.2.6 Preparing the standardsMeasuring mass is more accurate than measuring volume – after preparing the stock, thestandards were prepared using a pipette and weighed in order to calculate the true concen-tration.Knowing the precise concentrations of the standard solutions before doing the analy-sis is helpful, because the software can do the regression instantly. Otherwise the metalconcentration will need to be calculated manually later.Fresh standards were always prepared on the day of analysis for the most accurate results.1. Prepare 50.00 mL of a 50. ppm stock solution:ˆ Place 4.333 mg of dry AHM into a small vialˆ In several portions, quantitatively transfer the AHM to a 50.00 mL volumetricflaskˆ Dilute up to the mark with 2 vol.% HNO3 solutionˆ Make sure all of the solid is dissolved2. 0 ppm standard:ˆ Pour 15mL of the 2 vol.% HNO3 solution into a Falcon tube3. 5 ppm standard:ˆ Transfer 2 mL of the stock to a vialˆ Add 18 mL of 2 vol.% HNO3 solutionˆ Transfer 15 mL to a Falcon tube4. 12.5 ppm standard:ˆ Transfer 4 mL of the stock to a vial149ˆ Add 12 mL of 2 vol.% HNO3 solutionˆ Transfer 15 mL to a Falcon tube5. 25.0 ppm standard:ˆ Transfer 9 mL of the stock to a vialˆ Add 9 mL of 2 vol.% HNO3 solutionˆ Transfer 15 mL to a Falcon tube6. 37.5 ppm standard:ˆ Transfer 12.9 mL of the stock to a vialˆ Add 4.3 mL of 2 vol.% HNO3 solutionˆ Transfer 15 mL to a Falcon tube7. 50 ppm standard:ˆ Pour the rest of the 50 ppm stock into a Falcon tube150Appendix DDetailed Sample CharacterizationDataD.1 TEM micrographsThe TEM images shown in Section 3.10 were taken from the larger images to make the MoS2particles easier to see. Below are the whole images from which the snips were taken.151Figure D.1: Whole TEM image of the MoS2/AC-RM prepared by the micelle method, ellipsesdrawn around particles counted for the distribution displayed in Figure 3.14152Figure D.2: Whole TEM image of theMoS2/PC-RM prepared by the micelle method, ellipsesdrawn around particles counted for the distribution displayed in Figure 3.16153Figure D.3: Whole TEM image of the MoS2/PC-RM prepared by incipient wetness impreg-nation, ellipses drawn around several particles which were able to be distinguished from theotherwise long sheets which skew the particle size distributionD.2 ATTM certificates of analysisAs mentioned in the Experimental section, two batches of ATTM were purchased from Strem.The first batch turned out to be APTM, and the second batch was ATTM. Below are thecertificates of analysis, claiming that both batches are ATTM.154Figure D.4: Certificate of analysis for ATTM batch 1 (actually APTM)Figure D.5: Certificate of analysis for ATTM batch 2 (actually ATTM)155D.3 Values inputted into the MATLAB modelTable D.1: Measured concentrations of productsCatalystT(◦C)t(h)DBT µmolat X=0µmol/mg in original sample mol%CCHB BP THDBT DBT CHB BP THDBT DBTThermal 350 1 0.130 0.00078775 0.002 0.000 0.127 0.607 1.216 0.000 98.177Thermal 350 2 0.128 0.00083780 0.002 0.000 0.125 0.656 1.732 0.000 97.611Thermal 350 3 0.123 0.00086500 0.003 0.001 0.119 0.702 2.396 0.479 96.423Thermal 365 1 0.122 0.00074374 0.002 0.000 0.119 0.610 1.448 0.000 97.942Thermal 365 2 0.092 0.00038381 0.003 0.000 0.089 0.417 2.860 0.502 96.222Thermal 365 3 0.111 0.00061227 0.005 0.001 0.105 0.552 4.331 0.648 94.469Thermal 375 1 0.119 0.00069892 0.002 0.001 0.115 0.588 1.967 0.432 97.013Thermal 375 2 0.118 0.00073676 0.006 0.001 0.111 0.623 4.714 0.828 93.835Thermal 375 3 0.122 0.00080411 0.007 0.001 0.113 0.659 6.022 0.691 92.628MoS2/PC-RM 350 1 0.098 0.00000000 0.003 0.001 0.094 0.000 2.704 0.909 96.387MoS2/PC-RM 350 2 0.102 0.00000000 0.005 0.001 0.095 0.000 5.378 1.214 93.408MoS2/PC-RM 350 2 0.093 0.00000000 0.006 0.001 0.085 0.000 6.304 1.520 92.176MoS2/PC-RM 350 3 0.093 0.00110443 0.008 0.002 0.083 1.185 8.118 1.619 89.078156CatalystT(◦C)t(h)DBT µmolat X=0µmol/mg in original sample mol%CCHB BP THDBT DBT CHB BP THDBT DBTMoS2/PC-RM 365 1 0.105 0.00062970 0.005 0.001 0.098 0.600 4.443 1.208 93.750MoS2/PC-RM 365 2 0.094 0.00124431 0.010 0.002 0.081 1.329 10.504 2.021 86.146MoS2/PC-RM 365 3 0.100 0.00177120 0.015 0.002 0.081 1.777 14.976 1.884 81.364MoS2/PC-RM 375 1 0.095 0.00078660 0.006 0.002 0.087 0.826 6.627 1.624 90.923MoS2/PC-RM 375 1 0.093 0.00000000 0.007 0.002 0.085 0.000 7.113 1.632 91.254MoS2/PC-RM 375 2 0.101 0.00129753 0.013 0.002 0.085 1.279 13.033 1.758 83.929MoS2/PC-RM 375 3 0.101 0.00275033 0.022 0.002 0.075 2.719 21.439 1.828 74.013MoS2/AC-RM 350 1 0.124 0.00074293 0.004 0.002 0.117 0.601 3.133 1.229 95.037MoS2/AC-RM 350 2 0.109 0.00148799 0.007 0.002 0.099 1.364 6.570 1.746 90.321MoS2/AC-RM 350 3 0.104 0.00199079 0.011 0.002 0.090 1.906 10.283 1.583 86.229MoS2/AC-RM 350 3 0.099 0.00220968 0.011 0.002 0.083 2.230 11.519 2.051 84.199MoS2/AC-RM 350 3 0.067 0.00154105 0.008 0.001 0.057 2.295 11.493 2.026 84.186MoS2/AC-RM 365 1 0.110 0.00094328 0.006 0.002 0.101 0.858 5.688 1.589 91.866MoS2/AC-RM 365 2 0.108 0.00254568 0.012 0.002 0.092 2.363 10.700 1.989 84.948MoS2/AC-RM 365 2 0.099 0.00197460 0.010 0.002 0.084 2.004 10.164 2.161 85.670157CatalystT(◦C)t(h)DBT µmolat X=0µmol/mg in original sample mol%CCHB BP THDBT DBT CHB BP THDBT DBTMoS2/AC-RM 365 3 0.107 0.00349097 0.017 0.002 0.085 3.248 15.527 1.766 79.459MoS2/AC-RM 375 1 0.109 0.00113457 0.008 0.002 0.099 1.040 7.056 1.631 90.273MoS2/AC-RM 375 2 0.105 0.00290096 0.015 0.002 0.085 2.772 14.606 1.813 80.809MoS2/AC-RM 375 3 0.106 0.00462883 0.023 0.002 0.077 4.385 21.352 1.748 72.516MoS2/PC-WI 350 1 0.088 0.00317919 0.009 0.004 0.073 3.609 9.815 4.131 82.444MoS2/PC-WI 350 2 0.094 0.01089399 0.019 0.004 0.060 11.595 20.555 4.022 63.828MoS2/PC-WI 350 3 0.087 0.01670870 0.024 0.003 0.043 19.136 27.450 3.621 49.792MoS2/PC-WI 365 1 0.091 0.00669792 0.016 0.004 0.064 7.366 17.694 4.456 70.484MoS2/PC-WI 365 1 0.092 0.00697962 0.017 0.004 0.065 7.548 17.946 4.537 69.969MoS2/PC-WI 365 2 0.095 0.01707782 0.029 0.004 0.045 17.924 30.486 4.025 47.565MoS2/PC-WI 365 3 0.086 0.02260616 0.035 0.003 0.026 26.188 40.192 2.998 30.622MoS2/PC-WI 375 1 0.090 0.00900900 0.022 0.004 0.055 9.997 23.890 4.975 61.138MoS2/PC-WI 375 2 0.088 0.01839302 0.032 0.003 0.034 20.883 36.409 3.795 38.913MoS2/PC-WI 375 3 0.087 0.02612525 0.043 0.002 0.016 30.011 49.042 2.101 18.845158Table D.2: Concentrations of products with forced carbon balance, inputted into the MAT-LAB code for kinetic modellingCatalystT(◦C)t(h)Concentration (mol%C) Conversion(%)CHB BP THDBT DBTThermal 350 1 0.607 1.216 0.000 98.177 1.823Thermal 350 2 0.656 1.732 0.000 97.611 2.389Thermal 350 3 0.702 2.396 0.479 96.423 3.577Thermal 365 1 0.610 1.448 0.000 97.942 2.058Thermal 365 2 0.417 2.860 0.502 96.222 3.778Thermal 365 3 0.552 4.331 0.648 94.469 5.531Thermal 375 1 0.588 1.967 0.432 97.013 2.987Thermal 375 2 0.623 4.714 0.828 93.835 6.165Thermal 375 3 0.659 6.022 0.691 92.628 7.372MoS2/PC-RM 350 1 0.000 2.704 0.909 96.387 3.613MoS2/PC-RM 350 2 0.000 5.378 1.214 93.408 6.592MoS2/PC-RM 350 2 0.000 6.304 1.520 92.176 7.824MoS2/PC-RM 350 3 1.185 8.118 1.619 89.078 10.922MoS2/PC-RM 365 1 0.600 4.443 1.208 93.750 6.250MoS2/PC-RM 365 2 1.329 10.504 2.021 86.146 13.854MoS2/PC-RM 365 3 1.777 14.976 1.884 81.364 18.636MoS2/PC-RM 375 1 0.826 6.627 1.624 90.923 9.077MoS2/PC-RM 375 1 0.000 7.113 1.632 91.254 8.746MoS2/PC-RM 375 2 1.279 13.033 1.758 83.929 16.071MoS2/PC-RM 375 3 2.719 21.439 1.828 74.013 25.987159CatalystT(◦C)t(h)Concentration (mol%C) Conversion(%)CHB BP THDBT DBTMoS2/AC-RM 350 1 0.601 3.133 1.229 95.037 4.963MoS2/AC-RM 350 2 1.364 6.570 1.746 90.321 9.679MoS2/AC-RM 350 3 1.906 10.283 1.583 86.229 13.771MoS2/AC-RM 350 3 2.230 11.519 2.051 84.199 15.801MoS2/AC-RM 350 3 2.295 11.493 2.026 84.186 15.814MoS2/AC-RM 365 1 0.858 5.688 1.589 91.866 8.134MoS2/AC-RM 365 2 2.363 10.700 1.989 84.948 15.052MoS2/AC-RM 365 2 2.004 10.164 2.161 85.670 14.330MoS2/AC-RM 365 3 3.248 15.527 1.766 79.459 20.541MoS2/AC-RM 375 1 1.040 7.056 1.631 90.273 9.727MoS2/AC-RM 375 2 2.772 14.606 1.813 80.809 19.191MoS2/AC-RM 375 3 4.385 21.352 1.748 72.516 27.484MoS2/PC-WI 350 1 3.609 9.815 4.131 82.444 17.556MoS2/PC-WI 350 2 11.595 20.555 4.022 63.828 36.172MoS2/PC-WI 350 3 19.136 27.450 3.621 49.792 50.208MoS2/PC-WI 365 1 7.366 17.694 4.456 70.484 29.516MoS2/PC-WI 365 1 7.548 17.946 4.537 69.969 30.031MoS2/PC-WI 365 2 17.924 30.486 4.025 47.565 52.435MoS2/PC-WI 365 3 26.188 40.192 2.998 30.622 69.378MoS2/PC-WI 375 1 9.997 23.890 4.975 61.138 38.862MoS2/PC-WI 375 2 20.883 36.409 3.795 38.913 61.087MoS2/PC-WI 375 3 30.011 49.042 2.101 18.845 81.155160D.4 Arrhenius parameters for the thermaland thermocouple-induced reactionsTable D.3: Pre-exponential factors, A, and activation barrier energies, Ea, for the rate con-stants associated with the thermal and thermocouple-induced reactions. The error associatedwith A is at least 2 orders of magnitude smaller than A and is omitted hereParameter A (s−1) Ea (kJ mol−1)kDDS 4.1 E+04 121.9 ± 4.6kHG 1.3 E-01 61.2 ± 19.5161


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items