UNDERSTANDING THE SOLID/SOLID AND LIQUID/SOLID INTERFACE PHENOMENA FOR ALTERNATIVE ENERGY APPLICATIONS By Thanaphong Phongpreecha A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Chemical Engineering Doctor of Philosophy 2018 ABSTRACT UNDERSTANDING THE SOLID/SOLID AND LIQUID/SOLID INTERFACE PHENOMENA FOR ALTERNATIVE ENERGY APPLICATIONS By Thanaphong Phongpreecha In many alternativ e energy technologies, interfacial phenomena are critical in determining process efficiency and feasibility. Pre viously , a complete application of d ensity functional theory (DFT) as a tool to study interfaces has not only led to fundamental insights into the system, but also predictive models and recommendations for materials that were experimentally validated . In this thesis, i t is first demonstrated that such an a pproach can also be used to solve a well - characterized interface problem s related to sustainable energy technologies . This first problem is a solid/ solid interface between the solid oxide fuel cell (SOFC) and its sealant , specifically an - stabilized zirco nia (YSZ), and the commercial Ag - CuO sealing braze. The current material, CuO, which helps molten Ag to wet on YSZ surface s an d later adhere once the Ag is solidified, has been previously characterized to form a CuO - rich layer between Ag and YSZ. However, over time it is reduced by and results in formation of pores . Therefore, a lternative oxides are needed that are thermodynamically stable in high operating temperature s and also provide similar adhesion properties to CuO . To that end , this thesis has identified two mechanisms that enable only CuO to work so far. First, it was found that, un like most other Ag/oxides, the Ag/ CuO interface has a particularly that of Ag/YSZ. Second , the dissolved oxygen in molten Ag from air during the brazing process diffuse s to the interface and partially improve s adhesion. The high adhesion of A g/CuO was proposed to be unique atomic structure. Therefore, a descriptor based on the s structural and chemical features was developed to pre dict Ag/oxide adhesion. The descriptor expedites screening of new oxides as no exp ensive calculation is needed, lead ing to several recommendations such as CuAlO 2 and Cu 3 TiO 4 . In the second problem, a more complex and less well - characterized interface in sustainable energy technologies is investigated. It should be n oted that d ue to the comp lexity, the screening model can not be developed yet. This second interface is a liquid/solid int erface between solvated lignin and its catalyst. Lignin is a highly heterogeneous polymer in lignocellulosic biomass with many functional gr oups. Its valorization is essential to the economic feasibility of biorefinery , and its c atalytic hydrogenolysis in a liquid phase is well - studied technology route to add value to the lignin . Many experiments have shown that solvent cho ice can have large impacts on product types and yields. However, the solvent effect on the reaction is still unclear . An understanding of the sol vated lignin/ catalyst interface would be valuable. For this, we are interested in the adsorption. To achieve th is , spectroscopic and wet chemistry techniques are first used to help characterize the system by identify ing that the quantity of ether linkage s in lignin is the most critical functional group that determines the hydrogenolysis yie lds. Using a lignin dimer t hat contains a characteristic ether linkage, DFT show s that , in vacuum, the adsorption of the dimer is much stronger on Ni(111) than Cu(111). Upon solvation with ethanol, it was found that the dimer - metal interactions weaken so significantly that for Cu( 111), it no longer adsorbs onto the metal surface . T his implies that Cu may not provide high ca talytic activity, which agrees with hydrogenolysis experiments performed in this thesis, although there could be other contributing factors . Lastly, to circumven t large DFT calculations, a model based on a thermodynamic cycle was developed to predict adsorption energy of a solvated lignin dimer for a given pair of solvent and catalyst. Although the model is not suitable for screening purposes , it provides a valuab le , quantitative insight to the solvent effects . Copyright by THANAPHONG PHONGPREECHA 2018 v ACKNOWLEDGEMENT S Through a series of events, I have been fortunate to meet a number of smart, inspiring, and supportive people throughout my Ph.D . study , whose help I tremendously appreciate . First of all, I would like to thank my two principal investigators, Dr. David Hodge and Dr. Yue Qi, who have graciously taken me into their group s , are always o pen to discussions, and have provided support and guidance without reservation . I would also like to thank my committee members , Dr. Dennis Miller and Dr. James Jackson, who always welcome d me to their of fice, were happy to give me technical guidance , and also allowed me to use their precious laboratories. And last but not least, thank you Dr. Eric Hegg and Dr. Jason Nicholas for your support in my Ph.D. study . To my family, thank you Mom and D ad (Vilai P hongpreecha and Siripot Phongpreecha), for all of your sacrifi ces since the day I was born . You always love me, believe in me, support me , and encourage me to go one step fu rther. Of course, you too, Supreeya Phongpree cha and Suchawadee Phongpreecha, my li ttle sisters, for taking care of Mom and D for your moral support . To my friends, my Ph.D. life would have been so arduous and lonely without you guys. First of all, I would like to thank my former Hodge - lab mates , Jacob Crowe, Ryan Stoklosa, Lisaura Maldonado - Pereira , Muyang Li, and Glen Li for all the knowledge you passed to me and the fun time s in the lab. Thank you my 2014 Chemical Engineering student crew, Aritra Chakraborty, Eric Straley, Jake Claes, Preetum Giri, Sayli Bote, and Kirti Bhardwaj for sticking through the first couple years of graduate school together. My Chinese squad ( ), Jialin Liu, Pengchao Hao, Zhongyu Zhang, and Xinting Lin for not only helping me with my lab vi struggles, but also for all the good time s we had at Sichuan. My Hegg - lab mates, Aditya Bhalla, Namita Bhalla, Gracielou Klinger, Julius Campeciño , Elise Rivett, and Clarisse Mebane for teaching me biochemistry techniques and all the laughter you gave me. My Jackson - lab mates Benjamin Appiagyei and Tayeb Kakeshpour for always willing to help me with my poor chemistry . My Qi - lab mates Tridip Das, Hongkang Tian, Christine James, Kwangjin Kim, Yunsong Li, and Yuxiao Lin for the fun activities and teaching me theoretical works . My lab manager friends Cliff Foster, Dr. Dan Holmes, Dr. Li Xie, and Phillip Hill for all the good talks, helping out with using equipment, and not killing me when I broke it. My undergraduate friends , Nicholas Hool and Kendall Christy, I thank you for your hard work and sincerity, and I am happy to see your professional life taking off. For m y Thai friends, I would like to first thank my undergraduate advisors who inspired and supported me to pursue a Ph.D. study Dr. Siwarutt B oonyarattanakalin , Dr. Wanwipa Siriwatwetchakul, and Dr. Apichit Svangariyaskul . My Thai cook landlords, Ekkarach Thaweerak , Sorawit meejinda , and Pronthep promna , who fed me countless number of meals. Also, m y Thai friends who have not abandoned me and always kept me in company Sittipol Puarattana - aroonkorn , Charkrid Patamatamkul , Suwat Phinhongthong, Sathika Jariyatamkitti , Sudaporn Gunnaleka , Tonnam Balankura , Nattaya Siritrai , Will Wongsirikul, Woranit Muangmala , and many others who I may have failed to mention . Lastly, thank you Isaac Sarver for proof reading my English for this thesis. Thank you all for making this journey a relishing ride . I wish you all the best. vii TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ........................ x LIST OF FIGURES ................................ ................................ ................................ ..................... xi CHAPTER 1 Ba ckground and Motivations ................................ ................................ ............... 1 1.1 Selected Interfaces in Solid Oxide Fuel Cell and Biorefinery & Problem Identification ..... 1 1.1.1 Interfaces are Key Components in Energy Conversion and Storage Devices ................ 1 1.1.2 Sealing of the Electrolyte in SOFC ................................ ................................ ................ 5 1.1.3 Catalytic Hydrogenolysis of Lign in in Liquid Phase ................................ ..................... 7 1.2 Contributions of DFT Calculations to the Understanding of Interface Interactions ........... 11 1.2.1 Contributions to the Gas/Solid Interface Interactions ................................ .................. 13 1.2.2 Contributions to the Solid/Solid Interface Interactions ................................ ................ 15 1.2.3 Contributions to the Liquid/Solid Interface Interactions ................................ .............. 18 1.3 Current Studies on Adhesio n at Metal - Zirconia Interfaces & the Challenges for Ag/Oxides Interface for the Sealing of the SOFC ................................ ................................ ....................... 20 1.3.1 Adhesion at the YSZ Electrolyte Surface ................................ ................................ ..... 20 1.3.2 Trends and Correlation in Adhesion Energy ................................ ................................ 23 1.4 Current Computational Studies on Liquid/Metal Interface of Catalysts & the Challenges for Lignin Decomposition Processes ................................ ................................ .............................. 26 1.4.1 Characteristics of Solvent and Metal at the Interface ................................ ................... 27 1.4.2 The Effect of Solvent on the Adsorption of Aromatics on Catalyst Surfaces .............. 29 1.5 Motivation a nd Thesis Outline ................................ ................................ ............................ 30 CHAPTER 2 Designing Metal Oxides to Enhance Adhesion of Silver Brazes on YSZ ....... 33 2.1 Summary ................................ ................................ ................................ ............................. 33 2.2 Introduction ................................ ................................ ................................ ......................... 33 2.3 Methods ................................ ................................ ................................ ............................... 36 2.3.1 Geometry optimization ................................ ................................ ................................ . 36 2.3.2 Slab Models for Surfaces and Interfaces ................................ ................................ ...... 38 2.3.3 Ab Initio Molecular Dynamics ................................ ................................ ..................... 41 2.4 The Role of the Oxygen at the Ag( l )/YSZ interface during Brazing Process ..................... 42 2.5 Enhancement of Interface Adhesion by CuO ................................ ................................ ...... 45 2.6 Identification of Other Potential Oxide - Forming Alloying Elements ................................ . 47 2.6.1 Formulation of a Descriptor Using and in a Predictive Model ..................... 47 2.6.2 Descriptor - Based Prediction of Interface Energy and ................................ ........ 50 2.6.3 Identification of Complex Oxides and Ternary Braze Alloy Design ........................... 53 2.7 Conclusion ................................ ................................ ................................ ........................... 56 CHAPTER 3 Experimental Identification of the Yields Determining Functional Groups in Lignins ................................ ................................ ................................ ................................ .......... 58 3.1 Summary ................................ ................................ ................................ ............................. 58 3.2 Introduction ................................ ................................ ................................ ......................... 59 viii 3.3 Methods ................................ ................................ ................................ ............................... 63 3.3.1 Lignin Fractionation ................................ ................................ ................................ ..... 63 3.3.2 Lignin Extraction from Birch ................................ ................................ ....................... 63 3.3.3 Lignin Acetylation ................................ ................................ ................................ ........ 64 3.3.4 NMR Analysis of Lignin ................................ ................................ .............................. 64 3.3.5 Gel Permeation Chromatography ................................ ................................ ................. 66 3.3.6 Thioacidolysis Depolymerization of Lignin ................................ ................................ . 66 3.3.7 Catalytic Oxidative Cleavage of Lignin ................................ ................................ ....... 66 3.3.8 Catalytic Hydrogenolysis of Lignin ................................ ................................ ............. 67 3.3.9 Yield Prediction Model Development ................................ ................................ .......... 68 3.4. Lignin Fractionation Process and Recovery ................................ ................................ ....... 69 3.5 Trends in Properties of Fractionated Lignins ................................ ................................ ...... 70 3.6 Trends in Monomer Yields from Depolymerization of Fractionated Lignins .................... 73 3.7 Correlating Lignin Properties and Monomer Yields for Fractionated Alkali Lignins ........ 76 3.8 Predic O 4 Content ................................ ........................... 79 3.9 Conclusion ................................ ................................ ................................ ........................... 85 CHAPTER 4 Adsorption of the Yield Determining Lignin Dimer on Catalyst Surfaces .... 86 4.1 Summary ................................ ................................ ................................ ............................. 86 4.2 Introduction ................................ ................................ ................................ ......................... 87 4.3 Computat ional Methods ................................ ................................ ................................ ...... 91 4.3.1 Adsorption of Lignin Dimer and Monomers from DFT Calculations .......................... 91 4.3.2 Solvation Energy of Lignin Dimer by DFT Implicit and Explicit Model .................... 94 4.3.3 Additional Classical Molecular Dynamics ................................ ................................ ... 96 4.4. Development of Computational Approaches ................................ ................................ ..... 9 6 4.4.1 T he Significance of van der Waals Correction on Adsorption of Aromatics ............... 96 4.4.2 Determining Preferable Adsorption Site using Guaiacol ................................ ............. 98 4.4.3 Representing Polymeric Lignin Structure with a Dimer ................................ .............. 99 4.4.4 Identifying the most Stable Dimer Adsorption Conformation in Gas Phase ............. 101 4.5 Adsorption Energy of Lignin in Solvated Environments ................................ .................. 105 4.5.1 Comparison o f the Implicit and Explicit Model for Solvated Lignin Adsorption ...... 105 4.5.2 Calculation of Solvated Lignin Adsorption Energy from a Thermodynamic Cycle .. 108 4.6 Conclusion ................................ ................................ ................................ ......................... 111 CHAPTER 5 Catalytic Hydrogenolysis of Lignin to Aromatic Monomers ........................ 113 5.1 Summary ................................ ................................ ................................ ........................... 113 5.2 Introduction ................................ ................................ ................................ ....................... 114 5.3 Experimental Methods ................................ ................................ ................................ ...... 118 5.3.1 Biomass and Lignin Compositions ................................ ................................ ............. 118 5.3.2 Catalytic Hydrogenolysis of Lignin and Biomass ................................ ...................... 119 5.3.3 Enzymatic Hydrolysis ................................ ................................ ................................ 120 5.4. Monomer Yields from Native Lignin in Biomass Using Different Catalysts .................. 121 5.5 Lower Monomer Yields from Low Quality Lignin ................................ .......................... 125 5.6 Characteristics of the Pretreated Biomass and Enzymatic Hydrolysis ............................. 127 5.7 Conclusion ................................ ................................ ................................ ......................... 130 ix CHAPTER 6 Summary and Outlook ................................ ................................ ...................... 131 APPENDICES ................................ ................................ ................................ ........................... 135 APPENDIX A: Supp lementary Materials for Chapter 2 ................................ ......................... 136 APPENDIX B: Supplementary Materials for Chapter 3 ................................ ......................... 148 APPENDIX C: Supplementary Materials for Chapter 4 ................................ ......................... 151 BIBLIOGRAPHY ................................ ................................ ................................ ..................... 152 x LIST OF TABLES Table 1.1. Major products from selective hydrogenolysis reaction of lignin showing a wide range of possible products depending on reaction conditions and catalysts. ................................ ............ 9 Table 2.1 . The parametric detail of the interface cells with t heir corresponding k - points used in the present study. ................................ ................................ ................................ ................................ 38 Table 2.2. d ), work of adhesion ( ), calculated wetting angles ( ), and experimental ( ) for Ni/YSZ interface and systems involved in Ag CuO/YSZ braze. ................................ ................................ ................................ ................................ ....................... 44 Table 2.3. The surface and interface energy involved in Ni ................................ ................................ ................................ ................................ ....................... 45 Table 3.1. Percent recovery of each fraction through CO 2 acidification at room temperature and at 115 °C (SLRP TM ). The pH 2 fraction was obtained by acidification with H 2 SO 4 . ....................... 70 Table 4.1. Calculated adsorption energies (eV) of benzene on close - packed metal surfaces with different calculation methodology and comparison to the existing literature. .............................. 97 Table 4.2. d ) in Å for different adsorption conformations of the threo - dimer (shown in Figure 4.4) and conditions on Ni(111) and Cu(111) surfaces. ................................ ................................ ............. 104 Table 4.3. Adsorption energy ( ) in eV and average aryl carbon atoms distan ce to Ni(111) d ) in Å for the two different adsorption conformations of the erythro - dimer (shown in Figure 4.5) ................................ ................................ ................................ ................................ ... 105 Tabl e 4.4. Comparison of solvation energy ( ) of similar aromatic molecules in alcohol solvents using different implicit models, an explicit model, and from experiments. ................. 107 Table A1 . List of all the studied metals and oxides, including their crystal structures, lattice constants, surface energies from t he present work, along with comparisons to experiment values, and the calculated values for the descriptor ( ) as shown in Eqn. 2.4. All computational values are based on LDA functional unless otherwise stated 136 Table B1 . The Summary of quantified lignin properties and monomer yields 148 Table C1 . Comparison between density calculated from NPT dynamics using COMPASS II force field and experimental density of compounds similar to the dimer 151 xi LIST OF FIGURES Figure 1.1. Schematic diagram of SOFC in Nissan vehicle using bio - ethanol as a fuel (adapted from 6 ). ................................ ................................ ................................ ................................ ............ 2 Figure 1.2. Schematic diag ram of a single SOFC with a solid oxide - ion conducting electrolyte (A, adapted from 11 with permission), and of stacked SOFC showing sealing in each unit which is in 12 ). ................................ ............................. 3 Figure 1.3. The hypothetical biorefinery concept with an examples on how each component of the lignocellulosic biomass (cellulose, hemicellulose, and lignin) ca n be catalytically converted to useful fuels and chemicals (adapted from 14 with permission). ................................ ...................... 4 Figure 1.4. Hypothetical structure of softwood and hardwood lignin. Bonds abundance is obtained from Adler. 27 ................................ ................................ ................................ ................................ ... 7 Figure 1.5. Examples of changes in lignin functional ities at the O 4 linkage after going through different types of lignocellulosic biomass pretreatment (adapted from 14 with permission). ......... 8 Figure 1.6. Experimental, theoretical and computational tools plotted over their respective time and length scale domain of applicability with examples in the characterization and modeling of deformation and failure of biological protein material s (adapted from 46 with permission). ....... 12 Figure 1.7. DFT calculated scaling relations for adsorption energies of CH x species plotted against the adsorption energy of C for a number of different transition metals (adapted from 51 with permission). ................................ ................................ ................................ ................................ ... 13 Figure 1.8. DFT - calculated scaling relation between CO dissociative adsorption energy and an activation energy of C - O bond dissociation for methanation reaction (A). Experimentally measured rate as a function of the Fe/Ni content in catalysts (B) (adap ted from 55 with permission). ................................ ................................ ................................ ................................ ....................... 15 Figure 1.9. Universal scaling relation for adhesion energy of different metal interfaces (adapted from 69,70 with permission). ................................ ................................ ................................ ........... 17 Figure 1.10. The equation of state for different metals predicted by the scaling relation with results from experiments (adapted from 72 with permission). ................................ ................................ .. 18 Figure 1.11. The schematic diagram of Winterbottom analysis for calculation of interface energy and the high angle annular dark field micrographs of Ni partic le on YSZ surface captured by HRTEM (A). The Ni(111)/YSZ(111) obtained from HRTEM (B) (adapted from 64 with permission). ................................ ................................ ................................ ................................ ... 21 Figure 1.12. The surface and Ni/ ZrO 2 interface termination of ZrO 2 (111) at different oxygen chemical potentials (A, adapted from 95 with permission) . The oxygen chemical potential at each side and in electrolyte of an operating SOFC at 1000 °C (B, adapted from 96 with permission) . Together, these indicate that ZrO 2 (111) is terminated with O layer at the interface. ................... 22 xii Figure 1.13. The relationship between the work of adhesion ( ) of metals/ - Al 2 O 3 (0001) and 65 with permission) and the relationship between unit surface area (B, adapted from 101 with permission). ................................ .............................. 23 Figure 1.14. The relationship between the work of adhesion ( ) of metals on various oxides as 103 with permission). ................................ ................................ ................................ ................................ ....................... 25 Figure 1.15. Illustration of solvent simulation in an explicit (A) and an implicit model (B) (adapted from 118 with permission ). ................................ ................................ ................................ ............. 26 Figure 1.16. The average number of hydrogen bond along the z - coordinate (A, adapted from 124 with permission). The water str ucture at a water/Pt interface. The water molecule with blue ball represents watA, and the water molecules with red balls are of watB. The HOMO of the interface is partially located on watA, as represented with pink and cyan isosurfaces. The dashed lines i ndicate the hydrogen bonds formed between water molecules with the number showing the vibrational frequencies (B, adapted from 125 with permission). ................................ .................... 28 Figure 1.17. Difference in adsorption energies between in vacuum and solvated of molecules on Pt(111) surface using a PCM model to represent water medium. The changes are classified in color by nature of the adsorbates (adapt ed from 126 with permission). ................................ .................. 29 Figure 2.1. Hypothetical mechanisms of how Ag - CuO braze enables wetting starting from the solid phase (left) to melting and wetting (right) during the heating process in air. While the wetting angle of Ag on YSZ > 90 ° , Ag - O clusters in Ag( l ) solution may alter interfacial energy ( Mechanism I ), or addition of Cu may facilitate formation of CuO interla yer on YSZ surface ( Mechanism II ) leading to reduction of the wetting angle. The SEM images used for representation of stainless steel and YSZ microstructure was taken from literature. 147 ................................ ......................... 35 Figure 2.2. The relaxed interface structures for investigation of Ag CuO/YSZ brazes. These include Ag( l )/YSZ (a), Ag/YSZ (b), Ag/CuO (c), and CuO/YSZ (d). Only the top half of the sandwich model of the solid/solid interface is shown due to symmetry. ................................ ...... 39 Figure 2.3. Ag( l )/YSZ interface structures with dissolved oxygen atoms at the beginning and end O clusters were investigated, including in the bulk Ag( l ), at the free surface of the Ag( l ), and at the Ag( l )/YSZ interface. The Ag O at the interface is the most favorable structure. ................................ ................................ .. 42 Figure 2.4. Side view and top view of surface and interface structures to illustrate the formulation of th e descriptor, , in Eqn. 2.3. The and were demonstrated for a) an unrelaxed rock salt CaO {100} surface, where f - Al 2 O 3 (0001). The relaxed interfaces of Ag/corundum - oxide demonstrate that - Al 2 O 3 and other corundum - oxides f 3+ , weakens Ag O interfacial bonding, except c) in Fe 2 O 3 , whose cation, Fe 3+ , submerges beneath the oxyge n layer after relaxation so it does not interfere with the Ag O bonds ( f the same cation in all the corundum oxides. ................................ ................................ ................. 48 xiii Figure 2.5. The correlation a) between of the interface energy ( ) of Ag/oxides and the descriptor ( ), which followed a logarithmic function, and b) between the predicted work of adhesion ( ) at the Ag/oxides interface and the DFT calculated ones ( ). .................. 51 Figure 2.6. The relaxed structure of the Ag/CuAlO 2 by LDA functional showing an interface gap - shift in the oxygen's p - states. This indicates strong interactions between Ag and CuAlO 2 (b). ... 56 Figure 3.1. 2D 1 H - 13 C HSQC NMR spectra showing lignin fraction SLRP - 1 composition and linkages. The contours in aliphatic region (left) are used to quantify the fraction of inter - unit linkages, while those in the aromatic region (right) are used for S/G ratios. ............................... 72 Figure 3.2. Trends in monomer yields f rom different depolymerization techniques, including (A) Cu - catalyzed oxidation, (B) quantitative thioacidolysis, and (C) Ni/C - catalyzed hydrogenolysis. ................................ ................................ ................................ ................................ ....................... 74 Figure 3.3. Correlation map for fractionated alkali lignin properties and monomer yields demonstrating strong correlations between most lignin properties and its susceptibility to depolymerization via - O - 4 cleavage. ................................ ................................ .......................... 77 Figure 3.4. Selected linear correlation between fractionated lignin properties and monomer yields, - O - 4 content is d irectly correlated to molecular weight and inversely correlated to phenolic hydroxyl content, that (B) monomer yields from thioacidolysis (g/g) were - O - 4 contents, and that (C) as well as molecular weight and phenolic hydroxyl contents. ................................ ................................ ................................ .......... 78 Figure 3.5. Lignin chain model of chain length n used for formulating prediction rationale. The i th monomer unit, n th linkages of lignin are represented by m i and d n . A triad is composed of three monomer units. ................................ ................................ ................................ ............................. 80 Figure 3.6. Pr - O - 4 content for lignin polymer populations of chain length of 100 monomers (left) and 25 monomers (right) with results from quantitative thioacidolysis of fractionated poplar lignin and native birch biomass include d for validation. ................................ ................................ ................................ ................................ ...... 82 Figure 3.7. Model predictions of monomer yields from lignin polymer populations of different chain lengths v ersus monomer yields from thioacidolysis, oxidation, and hydrogenolysis of fractionated lignin. ................................ ................................ ................................ ........................ 83 Figure 3.8. Predicted yield distributions for monomers, dimers, and oligomers from the cleavage - O - 4 in a population model of lignin polymers of n = 100 o - O - 4 contents. . 84 Figure 4.1. Thermodynamic adsorption energy cycle of lignin dimer (lig) in solvated (A and B) wi th ethanol (EtOH) and vacuum (C and D) state. It should be noted that, by definition, the from state A to B is intended to encompass both partial loss in lignin solvation energy and ethanol adsorption energy. The is fraction of solvation energy retained when dimer adsorbs relative to full solvation. The is fraction of the surface area covered by adsorbed dimer. ......................... 93 Figure 4.2. Adsorption energy ( ) of guaiacol at four different sites, top, bridge, FCC, and HCP, with two orientations, 0° and 30° with respect to the hexagonal ring formed by of nickel atoms on the top layer of the slab. ................................ ................................ ................................ 98 xiv Figure 4.3. Top view and side view of relaxed hypothetical lignin of 10 G - - O - 4 linkages and possible adsorption conformations of the poly mer on metal surfaces (A). The approximated adsorption energy of the polymer ( ) for a given number of units adsorbed on Ni and Cu and conformation energy ( ) for a given number of unit fixed to lie horizontal (B). ................................ ................................ ................................ ............................... 99 Figure 4.4. Relaxed adsorption conformations of lignin threo - dimer where in A) and B) both rings adsorb to the surface with both aryl rings completely planar, C) only one ring is completely planar, D) and E) only one ring adsorbs. The change in angles between the aromatic ring to the atoms next to it in conformation A and B compared to relaxed free dimer is shown in the enlarged images. The di stances (Å) from the surface to all oxygen atoms in the molecule are also shown for conformation A and B. ................................ ................................ ................................ ................ 103 Figure 4.5. Relaxed adsorption conformations of lignin erythro - dimer where both rings adsorbed on the Ni(111) with one conformation yielding an ethyl oxygen atom away from the surface (I) and the other adsorbed onto the surface (II) with distances from the surfaces to all oxy gen atoms in the molecule and some bond distances within the molecule. ................................ ................. 105 Figure 4.6. The values for solvent accessible surface area (SASA) using a 1.25 Å probe radius (grey mesh), the number of neighboring atoms (enlarged solvent atoms), and the number of bonds within a 2.5 Å radius before and after adsorption for relaxed A and D conformation from Table 4.2 in ethanol. The average value is shown in parentheses. ................................ .................... 108 Figure 5.1. Monomer yields from hydrogenolysis reaction at 200 °C (patterned) and 220 °C using different metal/C catalysts with whole biomass. The results are arranged according to group and from left to right, 4 - propylguaiacol (4 - PG), 4 - propylsyrin gol (4 - PS), 4 - propenylguaiacol (4 - PnylG), 4 - propenylsyringol (4 - PnylS), 4 - ethylguaiacol (4 - EG), 4 - ethylsyringol (4 - ES). The very small green color in Pd corresponds to syringol. ................................ ................................ ........ 121 Figure 5.2. Monomer yields from hydrogenolysis reactions of lignin dimer (guaiacylglycerol - beta - guaiacyl ether) using Pd/C in ethanol with and without presence of H 2 gas. The 4 - vinylG and 4 - PnolG refer to 4 - vinylguaiacol and 4 - propenylguaiacol, respectively. ................................ ...... 124 Figure 5.3. Monomer yields from hydrogenolysis reaction using Ni/C and ethanol at 220 °C of alkali lignin that was sequentially precipitated through CO 2 acidification. The S refers to syringol. ................................ ................................ ................................ ................................ ..................... 126 Fi gure 5.4. Images of raw birch biomass (A), and after pretreatment at 220 °C in ethanol with no catalyst (B), and with Co/C catalyst (C). ................................ ................................ .................... 127 Figure 5.5. Percent recovery of glucan, xylan, and lignin after pretreatment at 200 (dotted line) and 220 °C in ethanol with different metal/C catalysts (A). The correlation between total monomeric yields and lignin removal (100 lignin recovery) (B). ................................ .............................. 128 Figure 5.6. Enzymatic hydrolysis yields of the biomass after the reaction at 200 (dotted line) and 220 °C in e thanol with different metal/C catalysts (A). The sugar yields are based on the untreated biomass compositions. The correlation between lignin removal and glucose yields (B). .......... 129 Figure 6.1. Lignin dimer adsorption on Pd (111) with 0 and 0.33 ML hydrogen coverage. ...... 132 xv Figure 6.2. Ab intio shows alcohol decomposition leads to H and ethoxy adsorbing species on Ni(111). ................................ ................................ ................................ ................................ ....... 133 Figure A1. The relaxed bulk structure of CuAlO 2 (a) and its (0001) O surface (b). The relaxed bulk structure of TiCu 3 O 4 (c) and its (010) O surface (d) 139 Figure A2 . Pair radial distribution function of Ag - Ag in liquid Ag with 2 O atoms dissolved showing the nearest Ag - Ag distance of 2.8 Å, which is the same to the experimental value of pure silver measured in reducing atmosphere. ................................ ................................ .................... 143 Figure A3. Projected density of states (PDOS) of interface atoms in Ni/YSZ, Ag /YSZ, and Ag/CuO in their no - contact positions (slabs were 10 Å apart) and in their interface position as calculated by LDA functional. ................................ ................................ ................................ .... 144 Figure A4. Projected density of states (PDOS) of CuO(111) surface with stretched lattice parameters to match that of the metals in the interface. It exhibited a bandgap of 0.25 eV. ...... 145 Figure A5. Surface structures showing the difference between the position of the terminated metallic atom of most corundum oxides (represented by Al 2 O 3 ) and the Fe 2 O 3 . These are (0001) surface of initial Al 2 O 3 slab (a), relaxed Al 2 O 3 slab (b), initial Fe 2 O 3 slab (c), relaxed Fe 2 O 3 slab (d), relaxed Ag(111)||Al 2 O 3 (0001) interface (e), and relaxed Ag(111)||Fe 2 O 3 (0001) interface (f). ................................ ................................ ................................ ................................ ..................... 147 Figure A6. The correlation between W adh and the oxide formation energy of (0001) surface corundum type oxide and {001} rock salt type oxide. Note that Fe 2 O 3 was not included for the reason that its geometry was unlike other corundum oxides after surface relaxation. ............... 147 Figure B1. Trends of phenolic hydroxyl contents (black) in fractionated lignin showing lower values in fractions precipitated at higher pH, while aliphatic hydroxy l content (red) do not show any specific trend 149 Figure B2. Trends of S/G ratios in fractionated lignin by HSQC NMR (red) and thioacidolysis (black) showing higher S/G values in fractions precipitated at higher pH and SLRP samples .. 149 Figure B3. - O - 4 content (black) as measured by 13 C - NMR and molecular weight among fractionated lignin showing similar increasing trend in fractions precipitated at higher pH and SLRP samples. ................................ ................................ ................................ ............................ 149 Figure B4 . Kinetics of Cu - catalyzed oxidation reactions of lignin sample SLRP - 2 showing time for optimal yield at the study conditio n is at 7 minutes. ................................ ............................. 150 Figure C1 . The change in fraction of surface accessible surface area (SASA) of the dimer in solvated adsorb in reference to fully solvated state ( ) as estimated by SASA as a function of probe radius 151 1 CHAPTER 1 Background and Motivations 1.1 Selected Interfaces in Solid Oxide Fuel Cell and Bio refinery & Problem I dentification 1.1.1 Interfaces are Key Components in Energy Conversion a nd Storage Devices To decrease our dependence on fossil fuels, there has been a surge in development and deployment of alternative energy in the past decade. From 2008 to 2017, the generation of non - hydro renewable energy in the U.S. has more than tripled, 1 most of which can be contributed to the wind and solar energy sector . However, these technologies alone would not suffice because of, first, the intermittent nature of the energy source and, second, limited land - based transmission infrastructure in certain regions. 2 In light of the first shortcoming, battery and fuel cells have emerged as the solution for energy storage /conversion device. Additionally, both of them can also be used for transportation applicatio ns supported by alternative fuels. The second shortcoming can be addressed by other renewable energy sources, particularly lignocellulosic biorefinery, which produce s liquid fuels that can leverage existing pipeline networks. Together, these technologies c ould expand the usage and applications of renewable resources, and hence, reduce our reliance on petroleum. Two of the renewable energy technologies that are of the interest in this thesis are the solid oxide fuel cell (SOFC) and the lignocellulosic bioref inery. A SOFC is an electrochemical conversion device that produces electr icity directly from oxidizing fuel s such as methane or syngas using ceramic or oxide electrolytes. Lignocellulosic biorefinery is a process of chemical conversion and separation of b iofuels and chemicals from non - food woody biomass, such as poplar or switchgrass , b oth of which have great potentials. For SOFC, it has been praised for fuel flexibility, low emissions, the highest efficiency (60 - 90%) of any hydrocarbon - based electricity g eneration technology, and most importantly, the highest gravimetric and volumetric power 2 densities of any energy conversion technology . 3 This makes SOFC suitable for applications such as offset ting intermittent renewable power sources . 4 On the other hand, lignocellulosic biorefinery offers not just energy from biofuels, such as ethanol, but also a platform for renewable chemicals. 5 As an example of how these technologies could expand the use of alternative fuels, bioethanol can also be integrated with SOFC as an energy storage in a vehicle as currently deployed by Nissan e - NV200 ( Figure 1.1). 6,7 Together, they provide an alternative t o battery - based electric vehicles, while also leverag ing the existing petroleum infrastructure. 8 Figure 1 . 1 . Schematic diagram of SOFC in Nissan vehicle using bio - ethanol as a fuel (ada pted from 6 ). A diagram of a SOFC shown in Figure 1.2A shows that it is composed of anode, c athode, and a solid electrolyte, which are usually made up of nickel, lanthanum strontium manganite (LSM), yttria - stabilized zirconia (YSZ) , respectively . An SOFC generates electricity through oxidation of fuels on the anode side. The oxygen is supplied fr om the cathode side, transporting through the solid electrolyte. This is possible through the high ionic conductivity of YSZ or other electrolyte materials at higher temperatures . 9,10 In practice, each SOFC is sealed and s tacked 3 together to provide more power, where sealing is applied in between each unit and is in contact with its electrolyte , and the cell frame, which is typically made out of stainless steel ( Figure 1.2B). Therefore, it is crucial that the seala nt has a s trong binding to both YSZ and stainless steel. Figure 1 . 2 . Schematic diagram of a single SOFC with a solid oxide - ion conducting electrolyte (A , adapted from 11 with permission), and of stacked SO FC showing sealing in each unit which is in (B, adapted from 12 ). The biofuel used in Fig 1 .1 , such as ethanol, can be generated from lignocellulosic biomass through b iorefinery. L ignocellulosic biomass is mainly composed of cellulose, hemicellulose, and lignin. Biorefinery is a sustainable processing of these biomass es into a spectrum of marketable products. Beyond ethanol, Figure 1.3 shows these products, catalytic conversion methods, and which part of the biomass it came from . In practice, lignin is the only part that is still underutilized because it is cheaply bur ned at - site for energy production. As lignin accounts up to up to ~30 % of plant's mass and is one of a few potential natural sources of aromatics, 13 its valorization represents a vital improvement in the economic viabil ity of cellulosic bioref ineries. In literature, numbers of studies have achieved this through either liquid - phase catalytic oxidative or reductive approaches ( Figure 1.3). Therefore, an understanding of how lignin valorization reactions proceed on the catalyst surface is importan t for catalyst and solvent design to improve yields, selectivity, and other desired catalyst properties . 4 Figure 1 . 3 . The hypothetical biorefinery concept with an examples on how each component of the lignocellulosic biomass (cellulose, hemicellulose, and lignin) can be catalytically converted to useful fuels and chemicals (adapted from 14 with permission). In both of these technologies , a number of interfaces are involved and are critical in determining the feasibility and efficiency of the technologies. Interface is a boundary between two spatial regions occupied by different matter, at which material properties differ from the bulk phase. Fundam ental to the operation of SOFC, the oxygen gas is reduced ( ) and diffuse into the cathode (LSM) surfa ce through its oxygen vacancies, then through the electrolyte, and finally to the anode surface (Ni cermet/YSZ) where it is used for catalytic oxidation of H 2 ( ) . 15 To prevent these gases from leaking , a sealant is applied Therefore, the adhesion at the SOFC/sealant/frame interface determines the strength of the seal. In lignocellulosic biorefinery, catalytic process is at the heart o f its depolymerization process for all major components. For 5 example, efficient conversion of carbohydrates in biomass to furfurals, which are valuable feedstock for chemicals production, was conducted in organic solvents and typically with zeolite catalys ts. 16 18 For lignin, its hydrogenolysis to aromatic monomers are usually conducted in liquid phase with various metal - supported catalysts such as Ni, Pd, and Pt . 5,19 These reactions occur at the liquid / solid interface of the solvated adsorbate with the catalyst. Therefore, their interactions at the interface largely determine the type of products, yields, and selectivity. Two specific interfaces among these are of particular interest in this thesis. The first interface is the seal between th e SOFC and the stainless steel frame , particularly at the interface of its early degradation. The other interface is the liquid / solid interface between solvated ligni n and metal surfaces. Th is interface is critical in understanding of the effect of solvent on adsorbate - surface interactions, particularly in a context of catalytic reaction pathways. 1.1. 2 Sealing of the Electrolyte in SOFC For SOFC, the interface problem of an interest is the adhesion between the device electrolyte and its sealant. Currently, due to the extreme operating conditions of SOFC, its sealant microstructure degrade s prematurely, inhibiting a long shelf life of SO FC. However, finding new replacement materials that would exhibit similarly strong adhesion to the electrolyte, and hence seal it, has proven to be a challenging task. This is because the sealing material has to be able to withstand extreme temperature (~8 00 - 1000 °C), and a highly oxidizing and reducing environments for prolonged period of time. Apart from these, the material needs to adsorb mechanical stress and the differences in thermal expansion. 11 Most importantly, it also needs to have st rong binding to YSZ so that the seal can be hermetic. 6 Current technology employs a Ag - CuO braze to seal the exposed YSZ surface and mechanically supported by ferritic stainless steel frame . 20 Ag is used be cause of its many desirable braze characteristics including stability in air above 160 °C 21 and sufficient braze strength and ductility to withstand stress from expansion mismatch with YSZ. 22 Because liquid Ag used in the initial stages of the air brazing process needs to completely wet , i.e. spread on, on YSZ surface in order to form a coherent surface upon solidification, a small amount of Cu or CuO is added to promote its wettability. The wettability can be measur ed by contact angle , with an angle of more than 90° considered non - wetting. The contact angle of Ag on YSZ in air is 110 °C, 23 while adding just 4 wt% of CuO, which forms a CuO - rich layer between Ag and YSZ (SEM micrograph) , 24 reduces it to 45° and keep s it adhered strongly upon solidification. 25 However, overtime the high hydrogen diffusivity in Ag (8 ×10 - 5 at 750 °C) 21 leads to reduction of CuO, and hence, formation of pores in the braze, limiting the lifespan of SOFC. Replacing CuO with many other common monovalent metal oxides did not lead to similar Ag wetting effect. 26 This emphasizes the need to find alternative oxides that could be more c hemically complicated. However, currently it is not well understood how CuO helps Ag wet on YSZ . Having such understanding would be useful in facilitating materials discovery. Therefore, the objectives of Chapter 2 of this thesis are to identify what are the underlying mechanisms of Ag - CuO/YSZ and to leverage the insight for designing new materials. 7 1.1.3 Catalytic Hydrogenolysis of Lignin in Liquid Phase Figure 1 . 4 . Hypothetical structure of softwood and hardwood lignin. Bonds abundance is obtained from Adler. 27 For lignocellulosic biorefinery , the interface of interest is the adsorption of solvated lignin on metal surfaces. This represent s an interface during lignin catalytic depolymerization to valuable chemicals (henceforth referred to as lignin upgrade/valorization) , which is crucial to the economic feasibility of lignocellulosic biorefinery. Despite most upgrading reactions being performed in a liquid phase, there is still limited understanding of the solvent effect on surface adsorption mechanisms. Understanding this more could aid in cata lyst and solvent design. It is very challenging to understand and model the extremely complicated lignin upgrade topic, both from the perspective of lignin structure and its upgrade. This is because, f irst, l ignin is a complex phenylpropanyl polymers synt hesized in plant from three monomeric precursors, namely p - coumaryl, coniferyl, and sinapyl alcohol, which lead to its three monolignol units of p - 8 hydroxyphenyl, guaiacyl, and syringyl, respectively ( Figure 1.4). D ifferent types of plants have different proportions of these units in lignin . These units are then randomly connec ted by several types of ether and carbon cross - linkages shown in red in Figure 1 .4 . Secondly, isolation processes of lignin from biomass further induce more changes to its natural s tructure, mostly either introducing undesirable functionalization ( Figure 1.5) or degrading it by forming more recalcitrant C C linkages. These resulted in even more complex functionalities in lignin, including but not limited to variable changes in amount s of methoxyl groups, phenolic and aliphatic hydroxyl groups, polymer chain length, and linkages depending on the process. Each of these functionalization has been reported to affect depolymerization products and yields, especially the abundance of C - O lin kages, which are the primary target as they are the most easily cleaved. 28 31 Figure 1 . 5 . Examples of changes in lignin functionalities at the O 4 linkage after going through different types of lignocellulosic biomass pretreatment (adapted from 14 with permission ). Despite a number of reports on the effect from each functional group , quantitative relationships that relate these groups collectively with product yields have not been fully demonstrated yet. F or m odeling of the liquid / solid interface in the present study, this knowledge is vital in confirming that the section of lignin selected for atomistic modeling, which will be 9 limited in number of atoms, contains all the major yield - deciding functionality . Additionally, for the general lignin community, the understanding of this relationship would be useful for benchmarking the quality of lignin from different sources. Therefore, prior to modeling lignin adsorption, Chapter 3 of this thesis will employ a n umber of experimental techniques (spectroscopies and wet chemistries) to characterize how functionalities relate to its monomer yields. Table 1 . 1 . Major products from selective hydrogenolysis reaction of lignin showing a wide range of possible products depending on reaction conditions and catalysts. Catalyst T(°C), H 2 (bars) Solvent P1 P2 P3 P4 P5 P7 Ni/C * 200 - 200 °C, 0 MPa C 2 H 5 OH Ni/C 32 220 °C, 2 MPa Dioxane Ni/Al 2 O 3 33 250 °C, 3 MPa CH 3 OH Raney Ni 34 200 °C, 0 MPa CH 3 OH Pd/C * 200 - 200 °C, 0 MPa C 2 H 5 OH Pd/C 35 225 °C, 3.4 MPa CH 3 OH Pd/C, Ru/C 36 200 - 200 °C, 2 MPa C 2 H 5 OH+ H 2 O PdZn/C 37 225 °C, 3.4 MPa CH 3 OH Ni 7 Au 3 38 130 - 160 °C, 1 MPa Alkali H 2 O CuPMO 39 180 - 220 °C, 4 MPa CH 3 OH * Reactions done in the present thesis (See Chapter 6). 10 Catalytic depolymerization of lignin yields a mixture of valuable aromatic chemicals and degraded compounds. Particularly, liquid - phase hydrogenolysis will be the focus of the present thesis as it has been studied extensively recently. Apart from the ligni n chemistry, catalytic hydrogenolysis of lignin would not only result in different yields, but could vary significant ly in types of products as shown in Table 1. 1. These difference s could stem from the choice of catalyst, H 2 pressure, and solvent. For the last factor, as can be seen in the case of Pd/C in Table 1.1 where reaction temperature and H 2 pressure are similar, with different solvents, products could differ. In the present thesis, the effect of solvent will be highli ghted. Currently, the understanding of the solvent effects primarily comes from experimental approaches. By varying solvent choice and using Ni catalyst , it was suggested that in low temperatures (<100 °C), aprotic non - polar solvents, such as decalin and n - heptane, result in the highest yields with complete conv ersion to hydrogenated products. Through a thorough discussion , it was proposed that other solvents, including methanol, ethanol, and 1,4 - dioxane , among others, result in low conversion s because of strong interactions with the Ni surface. However, these solvents are also very commonly used for lignin reactions due to their high solubility. 40 Therefore, the paper highlight ed a very interesting question quote : Is a compromise between catalytic activity and substrate solubility required for the efficient conversion of lignin? . Certainly, one of the solution s is to heat the reaction up significantly (>200 °C), w hich shows that alcohols can give similar yields. This is in agreement with the fact that most hydrogenolysis studies employ alco hols and Ni catalysts , which require s high temperatures . 5,19 Note that in many cases high reaction temperatures also lead to undesirable products and lower selectivity. 41,42 However, high yields from alcohol solvents at high temperatures are not necessarily true to all catalyst s as a separate stu dy comparing solvents at 300 °C using zeolite based catalysts suggests that other solvents, 11 including methyl isobutyl ketone and acetone, resulted in higher yields than alcohols. 43 There are several possible causes to the discrepancy , and one of which relates back to the effect of solvent, i.e. the change in the adsorption of lignin and its reaction intermediates could vary significantly depending on a specific combination of catalyst and solvent. T o answer the aforementioned question, hypothetically, t here may exist a combination of solvent and catalyst that can dissolve lignin, and not at the cost of catalytic activity because of its interaction with the surface . T o find (or design ) that solvent, how these solvents affect the lignin - metal interactions at the surface needs to be understood first . However , studies on how the - O - 4 lignin dimer adsorbs on the surface even in a gas state is scarce and still inconclusive. More importantly, it was reported that different isomers of the model yielded different adsorption conformations on Ni surface, which subsequently leads to different reaction paths. 44 This stresses the significance of identifying adsorption structure , and particularly, how it changes when solvated . Therefore, Chapter 4 of this thesis aims to clar ify how lignin adsorb s on metal surfaces, and demonstrate s how it changes with the effect of solv ent s , using a representative section of lignin. Additionally, the effect of these changes between different catalysts will be briefly compared to experiments through catalytic hydrogenolysis in Chapter 5. 1.2 Contributions of DFT C alculations to the U ndersta nding of Interface Interactions The use of mu lti - scale modeling and simulation methods has become a widely accepted strategy for furthering our understanding in essentially all fields of chemistry, biochemistry, condensed matter physics, and materials science . 45 Different modeling methods have different strengths, shortcomings, and hence, various suitable applications. In theory, material properties an d reaction mechanisms can be describable by quantum mechanics (QM). However, as shown in Figure 1. 6 , simulation of large system size or long dynamics time, such as modeling of materials 12 crack propagation or nanoparticles sintering, are proven to be impractical for QM due to computational expense. 46 This w as addressed by using approximations and simpler equations as shown in the higher scale at the expense of energy and force description accuracy. In this thesis, density functional theory ( DFT ) calculations, which is one of the flavors of QM, will be the pr imary computational tool for study of the interface s . DFT is a method of obtaining an approximate solution to the Schrödinger equation of a ground - state many - body system . In DFT , the interacting system of fermions is described as a function of density, instead of its many - body wavefunctions . This allows the system to depend on spatial coordinates, instead of a number of electrons in a system. It should be noted that the energetics term derived from DFT does not include a contribution from entropy from tr anslation and rotational modes. 47 I n this section, general concepts and applications that DFT calculations can contribute to different types of interfaces will be discussed. These will also include energetic terms that will frequently be brought up. Figure 1 . 6 . Experimental, theoretical and computational tools plotted over their respective time and length scale domain of applicability with examples in the characterization and modeling of deformation and failure of biological protein materials (adapted from 46 with permission). 13 1.2 .1 Contributions to the Gas/Solid Interface Interactions For gas/solid interfaces, DFT calculations are regularly used for the understanding of molecular adsorption and reaction on catalyst surface. DFT is selected in these applications to provide insights at the interface including geometry, energy, electronic structure, and more. Particularly, the adsorption energy of gas phase molecule ( ), which is a measurement of the interaction between adsorbat e and surface , wa s also measured , and is typically calculated in DFT by : Eqn. ( 1 . 1 ) where is the energy of the cell with adsorbate on metal surface and and are cells with individual adsorbate molecule and metal surface, respectively . Therefore, negative means it is an exothermic adsorption ( the system gains energy). Using this concept, DFT calculations have shown accurate description of for many systems, including aromatics on metal surfaces, compared to experimental methods. 48 50 Figure 1 . 7 . DFT calculated scaling relations for adsorption energies of CH x species plotted against the adsorption energy of C for a number of different transition metals (adapted from 51 with permission ). 14 With analyses of and faster access to the en ergetics at gas/solid interface, DFT has revealed several univ ersal adsorption tre nds with fundamental insight and can be used for materials search, particularly in catalytic context. For example, scaling relations were developed to show that of different surface intermediates that bind to the surface through the same atom(s) scale with each other. An example of the relation is shown in Figure 1.7 for simple CH x species on various metal surfaces. 51 With electronic structure analyses, it was demons trated that the trend originating from that adsorption strength can be approximated by meta d electrons, and is independent of their sp electrons. G iven that the variations in adsorption energies and transition state energies are governed by the same basic physics, it is not surprising that the relation can also be used to relate with transit ion state energy, which can be used for calculation of reaction rate, leading to significantly faster catalyst screening and design. The approach has a wide impact over catalyst search for numbers of reactions, such as ammonia synthesis , 52,53 alcohol synthesis , 54 methanation , 55 hydrogen evolution , 56 acetylene hydrogenatio n , 57 ethanol decomposition, 58 and many more. 59,60 An example of this is shown for methanation (CO + 3H 2 CH 4 + H 2 O) , where C - O dissoc iation was found to be the rate - limiting step. In Figure 1.8 A , it is shown that the activation ene rgy for C - O dissociation scaled with the dissociative adsorption energy of CO. The activation energy can then be used to quickly approximate act ivity using microkinetic models . 61 The result shown in Figure 1.8A indicated that Co and Ru should have the high activity; however, due to their high price, DFT calculations also suggest that alternatives such as Fe - Ni alloys with the composition between FeNi and FeNi 3 should yield similar activities . 55 The design was then confirmed by experiments where FeNi and FeNi 3 showed the highest activity (at different metal loading , Figure 1.8B ) when alloy composition was varied, and were reportedly close to Co. Th ese examples show that DFT has proven to be an efficient tool for exploration of gas/solid interface , 15 starting from accurate a dsorption energy, to energy barrier determination, and finally to develop ment of strategies for materials design. Therefore, DFT would be a suitable tool to provide insights to lignin/metals interactions in the present thesis. Figure 1 . 8 . DFT - calculated scaling relation between CO dissociative adsorption energy and an activation energy of C - O bond dissociation for methanation reaction (A). Experimentally measured rate as a function of the Fe /Ni cont ent in cata lysts (B) (adapted from 55 with permission). 1.2 .2 Contributions to the Solid/Solid Interface Interactions At the solid/solid interface, materials of different phase s adhere together, and the strength can be measured by several methods, such as scratch testing 62 and direct pull - off. 63 M easurement 16 from these methods, however, would also depend on macroscopic properties, such as cracks, grain boundaries, and other defects, apart from the chemistry of the m aterials. 63 To exclude interferences from these macroscopic features , analyses of images such as t he ones from HRTEM were used to quantify pure interface adhesion that only depends on surface chemistry and atomic geometry. 64 This microscopic adhesion insight is useful in many applications, such as determining deposited nanoparticle coarsening. However, the procedure is significantly more complicated as it involves sample preparation such as deposition, annealing, and equilibration of metals under elect ron microscope, 64 making the amount of data in the literature limited. Conveniently, the adhesion in this level is also describable using DFT calculations, with an appropriate setup and a well - defined scope. DFT has been used extensively to measure microscopic work of adhesion between two materials. Herein, the work of adhes ion ( ) is defined as the reversible thermodynamic work required to separate the interface from the equilibrium state of two phases to a separation distance of infinity. and could be calculated from DFT by : Eqn. ( 1 . 2 ) where is t he surface energy of material 1, is the interface energy at the boundary of material 1 and 2 , is the energy of the cell of slab 1 and is the energy of the interface cell . Note that by definition, surface energy is the amount of work per area required to create a surface . Similarly, the amount of work per area required to increase that interface area is called interface energy. From Eqn. 1.2, t he higher positive is, the stronger the interface adhesion. Using this setup, previous studies have demonstrated accurate , including that of metal/oxide, compared to experiments. 65 68 17 Figure 1 . 9 . Universal scaling relation for adhesion energy of different metal interfaces (adapted from 69,70 with permission). With the ability to focus on just the effect from surface chemistry and geometry, DFT has contributed many valuable insights and predictive capability to the field of solid/solid adhesion. For example, it was used to construct the universal binding - energy relation (UBER). 69,70 This relation describe s the relationships between atomic separation and binding energy, or adhesion in this the equilibrium adhesion energy and the equilibriu m to fit a universal scale curve as shown in Figure 1.9 . The equilibrium interstitial electron density is used for calculation of scaled distance (a * ), which one could think of this as an analogy to scaling a length unit to a dimensionless parameter. The origin of this scaling behavior was identified to be the exponential decay of electron density into a vacuum from surfaces or interstitial sites. 71 This relation has an impact in many fields since it is essentially based on interatomic interaction s . For example, by describing the interface of the same materials ( i.e. cohesive energy), this relation has been used to derive an equation of state (EOS) for metals. 72 The predicted EOS c ompared to experimental data are shown in Fig 1.1 0. This has a significant impact in predicting materials properties. T he universal relation itself has also been used as a foundation 18 in many other solid/solid relations , such as predicting crack growth, 73,74 predicting bulk modulus , 75 thermal expansion, 76 and melting temperatures . 76,77 Therefore, these examples show that D FT is an efficient tool for exploration of solid/solid interface that does not only give reliable interface adhesion, but also facilitate s in development of prediction models. Figure 1 . 10 . The equation of state for different metals predicted by the scaling relation with results from experiments (adapted from 72 with permission). 1.2 .3 Contributions to the Liquid/Solid Interface I nteractions I n liquid/solid interfaces, most of the works were contributed to pure liquid, i.e. no solute. Particularly, DFT has been useful in estimating wetting angle that result only from microscopic interactions . 26,78,79 In macro scale, wetting angle could be affected by several factors such as surface roughness. 80 However, in a microscopic leve l, the contact angle ( ) can be directly estimated fr om DFT - calculable , Eqn. ( 1 . 3 ) 19 where is the liquid surface energy, which will be a pproximated using DFT - calculated solid surface energy. The detail s of effects from this estimation are discussed in Section 2 .3.2 . Using this approximation, studies have been able to predict wetting angle of Pd and Pt on ZrO 2 for 125 ° and 120° compared to 100° and 95 ° measured in experiments. 79,81 This may not be absolutely accurate, but given that the curre nt thesis focuses on one metal, . Ag, the order in the trend of the wetting angles of Ag/oxides should not change, and hence DFT calculations with this estimation will be employed with additional comparison to experiments when available. For simulation of solute - solvent system at the interface, one of the applications of DFT is to measure the adsorption energy of the solvated adsorbate. Experimentally, the adsorption energy data is much more limited compared to gas phase adsorption, which could be due to that traditional instrument, such as TPD, is not compatible and s pecific experimental setup is required. 82 Even for DFT, it is still challenging to m odel solvent effects as there are many methods with different pros and cons, and it is still an actively developing field. The detail to these methods will be discussed in Section 1.4. In the past, studies mostly used DFT for calculation of adsorption ener gy of solvated small molecules on surfaces. For example, adsorption of CO or H 2 on Cu 2 O(111), which is of an interest for catalytic synthesis applications. 83,84 shows that H 2 does not adsorb on the surface when solvated with polar solvents such as water or methanol. This indicates st ronger interactions of H 2 with the solvents than the oxide surface. However, due to limited amounts of literature on adsorption energy in liquid phase, studies comparing calculated and experimental adsorption energy were not found. To this end, validation of the system has been mostly relied on comparing the solute - solvent interaction, i.e. solvation energy, and past studies have shown several successes in calculating accurate solvation energy of molecules compared to experiment, 85,86 for example methanol in water, which is - 0.21 computationally and - 0.22 experimentally. 85 Nonetheless, with 20 metal atoms included and many other pairs of solute - solvent unexplored, more procedures on solvation and methods to validate this for our thesis is required to ensure reliability of the calculation. 1.3 Current Stud ies on Adhesion at Metal - Zirconia Interfaces & the Challenges for Ag/Oxides Interface for the Sealing of the SOFC As discussed in Section 1.1.1, during the sealing process where Ag is brazed on yttrium - stabilized zi rconia (YSZ), which is an SOFC electrolyt e, two types of interfaces are involved liquid/solid and solid/solid. Liquid Ag (with small amount of CuO) is initially wet on YSZ, then it forms a seal upon solidification. B ecause eventually all components are solid and during actual SOFC operations, the strength of the braze/YSZ interface is determined by the adhesion of these interfaces, the following sections will introduce current literature on the solid/ solid adhesion of metals with the YSZ surface, and, to aid in formulating adhesion prediction , the trend s in adhesion among metal/oxide interfaces. 1.3.1 Adhesion at the YSZ Electrolyte Surface In an SOFC sealant, pure Ag adhesion on zirconia is so weak that it requires ad dition of CuO. 87 This is partly due to the surface natu re of zirconia, and learning why is as important as identifying other oxides that it would strongly adhere to. Simulation of oxide surface and interface is challenging and require s multifaceted information because adhesion with it can change depending on surface terminations, defects, and environments. Therefore, to learn about zirconia characteristics at the interface and potentially apply them to the current thesis, this section reviews the nature of zirconia surfaces and interfaces, along with what can affect its adhesion. It should be noted that due to limited literature on Ag/ZrO 2 (and YSZ), the focus will be on Ni/ZrO 2 which has been studied most extensively due to its broad technological applications, including steam 21 reforming anode catalysts, 88 coatings, 89 and electrolysis. 90 Since both Ag and Ni are face - centered cubic (FCC) metals, it is expected that most of the insight from previous stu dies are applicable to the work in this thesis. Figure 1 . 11 . The schematic diagram of Winterbottom analysis for calculation of interface energy and the high angle annular dark field micrographs of Ni particle on YSZ surface captured by HRTEM (A). The Ni(111)/YSZ(111) obtained from HRTEM (B) (adapted from 64 with permission). The type of surface that is exposed at the interface from the metal and oxide has a huge impact on the adhesion. Previous experiments using high - resolution transmission electron microscopy (HRTEM) have shown that both Ni and YSZ expose their (111) surface at the interface. 64,91 Furthermore, when the image was combined with the Winterbottom Analysis 92 of Ni parti cle on a YSZ substrate ( Figu re 1.11 ), a solid/solid interface energy was determined to be 1.8 J/m 2 . 64 With the knowledge of YSZ and Ni surface energies, 93,94 this yields the of 1.48 J/m 2 . This is 3 times higher than that of the Ag/YSZ interface computed in the p resent thesis. Additionally, this number will also be useful for benchmarking purposes in computational studies. Computationally, studies have shown that ZrO 2 (111) has an O - terminated surface (henceforth referred to as ZrO 2 (111) O ) in the SOFC operat ing conditions and maintained this even after forming an interface with Ni. Using DFT calculations, it was shown that ZrO 2 (111) O is preferred across a wide range of oxyge n chemical potentials ( Figure 1.12 A). 95 Neverthel ess, when 22 forming an interface with Ni(111), the region with oxygen chemical potential below - 8.25 eV can - favor Zr - terminated ZrO 2 (111) instead . However, this is significantly below the SOFC operating condition of 0 (cathode side) to - 2 eV (anode side, Fi gure 1.12 B), 96 With the O termination, the bonding characteristic of Ni and O at the interface was shown to be site - dependent according to the charge distribution analysis. Ni atoms on top of ZrO 2 strong covalent bond characteristic s , whereas Ni atoms at the hollow position and other positions form an ionic bond. 95 Figure 1 . 12 . The surface and Ni/ ZrO 2 interface termination of ZrO 2 (111) at different oxygen chemical potentials (A, adapted from 95 with permission) . The oxygen chemical potential at each side and in electrolyt e of an operating SOFC at 1000 °C (B, adapted from 96 with permission) . Together, these indicate that ZrO 2 (111) is terminated with O layer at the interface. The position of oxygen vacancy, a natural point defect in oxides, was previously shown to decrease the of Ni/ZrO 2 and Ni(YSZ). 97,98 The decre ase was reportedly significantly larger in case of Ni/ ZrO 2 (111) O as the oxygen vacancy was found through its formation energy to prefer to be at the outer most layer of the interface. 99 This was explained by the weaker Ni - O bond compar ed to Zr - O. A separate study also suggested similarly as first principles tensile tests indicate 23 fracture at the Ni - O interface. 95 In contrast, for YSZ(111), it was suggested that the effect from oxygen vacancy should be negligible as its first multilayer. 100 Additionally , it was also found that as the simulated surface depth increases, calculations indicated more stability if the positions of the oxygen vacancy locate as a mixture of at subsurface and deeper multilayers. 15 1.3.2 Trends and Correlation in Adhesion Energy Figure 1 . 13 . The relationship between the work of adhesion ( ) of metals/ - Al 2 O 3 (0001) and ide formation energy energy (A, adapted from 65 with permission) and the relationship between of metals on several ox nergy per unit surface area (B, adapted from 101 with permission). For the challenges in the SOFC sealant, identifying what properties of the metal oxides could correlate with the interfacial energy or adhesion has an important implication in identifying oxides, apart from CuO, that would allow Ag to wet on. However, this does not exist for the Ag/oxides system yet, at least not a reliable one. Therefore, a brief history of how trends in metal/oxides adhesion which was developed and applied over the years will be discussed in this section. Multiple studies have been able to relate of metal/oxide s to several properties of the adhering m etals themselves. By correlating the the 2 4 properties, one could circumvent the experimental measurement or expensive DFT interface calculations, and hence, expedite the material design process . ergy, the photon their on SiO 2 , in which metals with higher plasmon energy such as Fe and Ni yield more adhesion. 102 Similarly, a correlation between and electron density at the boundary of the Wigner - Seitz cells of the corresponding metals also suggests Fe and Ni, wh ich have higher electron density, to yield higher than other metals such as Pb or Sn. 103,104 Al though not for purely predictive purposes, DFT calculations also indicate that the of HfO 2 and ZrO 2 (001) with various metals are positively correlated with the bond overlap population (BOP) at th e interface. 105 Gen erally, it was found that metals with sp valence band s tend to have higher BOP due to the localized nature of the d band. Lastly, perhaps the most c ommon correlation studied is between the 65,101,106 108 For example, using DFT calculations, it was found that metals with higher corresponding oxide formation energy display higher on the - Al 2 O 3 (0001 ) surface ( Figure 1.13 A). 65 A recent experimental study further confirms the trend wit h several more oxides ( Figure 1.13 B), although the relationship is based on oxide formation energy per area . 101 This relies on the assumption that the strength of metal oxide surfaces are terminated wi th full or partial oxygen layers, which may not be true depending on oxygen chemical potential in a given condition. 109 Nonetheless, leveraging this adhesion relationship, studies have been able to understand and design metal oxi de - supported catalysts that have higher thermal stability. 110 This knowle dge has also become useful in many other fields such as for finding appropriate metallic interconnects in a semiconductor technology. 111 114 25 Up to this point , it has been shown that there is a significan t body of work on correlations with variation of metals on a single metal oxide surface. H owever, studies from a reversed point of view, i.e. correlations from a particular metal on various oxide substrates, is scarce. This could be due to the heterogeneity of different types of oxide surfaces. To the best of our knowledge, only two studies were dedicated to identify ing such a relationship. 102,103 The first study is based on a collection of numbers of previous experimental results, which showed a linear correlation between higher of Ni, Fe, and Co with wider band gaps of the oxide subst rates as well as higher formation energy of oxides ( Figure 1.1 4 ). 103 The relationship was proposed to originate from the electronic charge trans fer capability of the oxide. This prediction has been useful in many applications, such as designing alloys for composite steels by screening for metal oxide s that Fe and Ni could wet on, among other criteria. 115 However, a separate study employing dielectric continuum theory suggests that the correlations between of Ag/oxides and the band gap of oxides is weak (R 2 ~ 0.6). 102 This conclusion is similar to the one obtained from DFT calculations performed in the present thesis (results not shown). Therefore, this stresses the need to find a new descrip tor that can describe Ag/oxides adhesion more accurately . Figure 1 . 14 . The relationship between the work of adhesion ( ) of metals on various oxides as n energy or its band gap energy (adapted from 103 with permission). 26 1.4 Current C omputational S tudies on Liquid / Metal Interface of Catalysts & the C hallenges for L ignin D ecomposition P rocess es Despite the fact that almost all lignin hydrogenolysis stud ies were done in a liquid phase, there i s a very limited number of studies at the interface, with only a few of these studies being solvated . A number of existing computationa l studies suggests that solvents can affect the properties of metal at the interf ace, and activity, in many other reductive reactions such as phenol hydrogenation 116 and hydrodeoxygenation. 117 Therefore, in this section, an overview of methods to simulate liquid in DFT, and how those solvents change surface and interface properties with aromatic compounds , will be discussed. Figure 1 . 15 . Illustration of solvent simu lation in an explici t (A) and an implicit model (B) (adapted from 118 with permission). In ab - intio calculations, solvent eff ects can be simulated by 3 means: an explicit model , an implicit model , and a hybrid model. Explicit models consider molecular details of each solvent molecule, while implicit model s treat solvent s as a continuous medium with a fixed dielectric constant where solutes are placed inside the cavity in the medium, ultimately allowing the calculation to be significan tly faster. These two mod els are illustrated in Figure 1.15 . Hyb rid 27 models treat the first (and second) solvation sphere explicitly while the rest is treated implicitly. The free energy of solvation can be defined by: Eqn. ( 1 . 4 ) where is the free energy required to form the solute cavity, is the van der Waals interaction, is the electrostatic interaction be tween solute and solvent, and is the hydrogen bonding term. 119 By treating each of these terms differently, solvation energy from the explicit model and even from different flavors of implicit models can be significantly different. For example, in implicit models, a polarizable continuum model (PCM) calculated the cavity of the solute by overlapping van der Waals spheres, 120 whereas in a conductor - like screening model (COSMO) cavity is calculated based on the solv ent accessible area. 121 With this in mind, the difference in results from the following literature should be expected. 1.4 .1 Chara cteristics of S olvent and M etal at the I nterface Challenges in solvation of the lignin dimer model adsorbed on metal surfaces in DFT simulations depends on the method used for simulation, which in itself is also a challenge. Due to the relatively large siz e of the lignin dimer, explicit solvation would be very expensive, whereas implicit solvation may not accurately describe the interaction. To learn of the past performance s and effects from each method, the present section and Section 1.4.2 summarize relev ant works on these with and without the presence of an adsorbate. U sing an explicit model , a study found that upon solvation with water molecules, Pt and Ni electron density at the interface is redistributed toward the subsurface layer. 116 This was shown to lower the metal work function, and subsequently, make the liquid - covered metal a higher chemical reducti on strength. On the other hand, by testing various implicit solvation models, it 28 was found that most continuum models undersolvate metal surfaces, and underestimate the surface charge as a function of applied potential compared to experiments. 122 This is most likely because their parameterizations were fit to molecular solvation energies, with no ions or metallic surfaces included in the datasets. 123 Figure 1 . 16 . The average number of hydrogen bond along the z - coordinate (A, adapted from 124 with permission). The water structure at a water/Pt interface. The water molecule with blue ball represents watA, and the water molecules with red balls are of watB. The HOMO of the inter face is partially located on watA, as represented with pink and cyan isosurfaces. The dashed lines indicate the hydrogen bonds formed between water molecules with the number showing the vibrational frequencies (B, adapted from 125 with permission). On the solvent side, representation of water molecules on the Pt(111) surface explicitly has shown their different characteristics compared to the ir bulk . First, there is about one fewer number of average hydrogen bonds among water within a 1 Å thick layer above the interface compared t o the bulk as shown in Figure 1.16 A. This is a result of interaction with the metal surface. 124 However, it is shown that this interface characteristic is significantly dependent on the metal. As shown in Figure 1.16 A, the nu mber of hydrogen bonds at the interface layer with Au, which is a rather inert surface, recovers much quicker than that of Pt. Furthermore, it was also observed that some of the hydrogen bonds at the interface actually became stronger as evidenced by lower 29 vibrational frequencies shown in the interaction of watA (blue) compared to watB water molecules in Figure 1.16 B. 125 1. 4 .2 The E ffect of S olvent on the A dsorption of A romatics on C atalyst S urfaces Figure 1 . 17 . Difference in adsorption energies between in vacuum and solvated of molecules on Pt(111) surface using a PCM model to represent water medium. The changes are classified in color by nature of the adsorbates (adapted from 126 with permission). Using a PCM method , it was reported that among different typ es of molecules that were solvated in water while adsorbing on Pt(111) , the change in adsorption energy in aromatic compounds was the largest. A mong the five categories of molecules, aromatics are on average the most affected with adsorption energy in a solvated state in creased by ~0.3 eV compared to the value in a vacuum state ( Figure 1.1 7 ). 126 These results imply that the hydrophobic , aromatic rings drive these molecules to adhere more strongly to the surface. Noted that the effect from int eraction with water was amplified by the larger cavity sizes these aromatic adsorbates occupied. However, 30 a separate study modeling phenol adsorption on Ni(111) under explicit representation of water molecules suggested differently that the adsorption ener gy of phenol actually decreased by 0.11 eV when solvated. 116 On the other hand , a recent study has also att empted to incorporate the effect of solvation using ethylene glycol as a solvent to a lignin dimer model surface reaction. The study employed PCM method with thermodynamic cycle, and reported that the adsorption energy of the dimer decreased over 91%. Thes e existing discrepancies in results and met hods from previous studies pose a challenge in accurately modeling the solvated lignin dimer and stress the importance of care ful defining and benchmarking solvation for this thesis . 1.5 Motivation and Thesis Outline Similar to the e xamples of DFT contribution in Section 1.2, where a complete usage of DFT builds from desired energetics, to fundamental insights, to prediction model s , and finally to candidate suggestions, which could then be experimentally tested, the present thesis aims to use this approach to solve SOFC sealing interface problem. Therefore, it would encompass adhesion values, the mechanism of how Ag - CuO adhere strongly on YSZ sur face, and predicting alternative oxides that are ready to be experimented on. For solvated lignin adsorption problem, a similar approach will be used but only up to getting the insight of how solvation affects lignin adsorption due to the complexity of liq uid simulation and of the lignin itself . Because of the limit of DFT simulation size compared to the complexity of lignin, experimental identification of the mo st important functional group in lignin s on their mono mer yields will be conducted and serve as a basis for the simulated lignin section. Lastly, catalytic hydrogenolysis of lignin using the investigated metal surfaces will be conduct to observe the agreement between computational and experimental, although it should be noted that future works are required to investigate other contributing factors. In pursuing these goals, the chapters are outlined as follow s : 31 Chapter 2 : Designing Metal Oxides to Enhance Adhesion of Silver Brazes on YSZ This chapter explains the underlying mechanism s that allow the Ag - CuO system to adhere on YSZ surface s and presents a descriptor for prediction of Ag/metal oxide adhesion. The effect of diffused oxygen in Ag on the adhesion was first investigated. The microscopic adhe sion of Ag on all related metal ox ide surfaces were then extensively studied . B ased on the analyses of their structures and results, the descriptor was proposed and validated. Chapter 3: Identification of the Yield Determin ing Functional Groups in Lignin This chapter explicates the correla tions among the complex functionalitie s of technical lignin as well as their relationship to depolymerization yields from various r eactions. Almost all functionalities of lignin s , including polymer chain length, amount of aliphatic and phenolic hydroxyl co - O - 4 linkages, will be experimentally quantifi ed and correlated. The lignin was also subjected to multiple types of reactions, and a model relating the functionalit y to reaction yields was developed and validated. C h apter 4: Adsorption of the Yield Determining Lignin Dimer on Catalyst Surfaces This chapter examines the adsorption of the lignin dimer with the - O - 4 linkage , which was previously shown to be the pri mary yield determining factor , on Cu and Ni surfaces in vacuum and solvated states. First, t and its conformation were identified. Ethanol was then added to the simulation to examine its effect on the adsorption conformation and energy. From these results, a thermodynamic model was the n developed to connect the adsorption energy in vacuum and solvated state s. This model highlights the importance of solvent selection and can 32 Chapter 5: Catalytic Hydrogenolysis of Lignin to Aromatic Monomers This chapter explores the hydrogenolysis of lignin using ethanol as a solvent with a number of metal - supported catalysts. Using lignin s directly from whole biomass, th e product yields and selectivities from hydrogenolysis and hydrodeo xygenation were quantified. Lignin removal from the pretreated biomass was also investigated. By performing enzymatic hydrolysis on these biomasses, the correlation between lignin removal, aromatic monomer yields, and sugar yields were developed. Chapter 6 : Summary and Outlook This chapter exhibits a summary of the works performed in this thesis and discusses possible future research directions. 33 CHAPTER 2 Designing Metal Oxides to Enhance Adhesion of Silver Brazes on YSZ 2 .1 Summary Ag - CuO is a broadly used reactive air brazing (RAB) system for effectively bonding ceramics and metal interfaces, especially for sealing yttria stabilized zirconia (YSZ) to metals in solid - oxide fuel cells (SOFCs). To understand the superior performance of this braze, density func tional theory (DFT) calculations were employed to investigate two mechanisms that can potentially increase the work of adhesion ( ) and hence reduce the wetting angle of Ag on YSZ. It was found while the formation of Ag dissolved O clusters at the Ag - YSZ interface can promote wetting, a much greater wetting angle reduction comes from the formation of CuO interlayers between Ag and YS Z. Further, the of an Ag/CuO and CuO/YSZ interface was found to be significantly higher than that of an Ag/YSZ interface. Based on simulation obtained - insights into metal to oxide bond formation, a simple descriptor was developed to predict Ag/oxide interface energies, predict Ag/oxide , and search for potential oxide interlayers capable of promoting the wetting and adhesion of Ag on YSZ. Many simple metal oxides (single cation) were examined, however their with Ag was less than that of an Ag/CuO interface. Expanding the search to multi - cation oxides led to several promising candidates, such as CuAlO 2 , CuGaO 2 , and Cu 3 TiO 4 ; all of which are also stable in the reducing SOFC conditions. Depending upon their solubility in molten Ag, these newly - identified oxides could either be pre - applied as wetting promoting interlayers or directly incorporated into Ag to form new reactive air brazes. 2 .2 Introduction Adhesive metal/metal - oxide interface bonds are critical in a variety of applications including the formation of durable passivation coatings, oxide - supported catalysts, solders , and brazes . 127 For instance, silver - based brazes have been used to separate the fuel (i.e. anode) and 34 oxidant (i.e. cathode) chambers in solid oxide fuel cells due to their superior toughness and gas impermeability compared to glass or mica seals . 128 132 In this application, sealing is accomplished by brazing a yttria - stabilized zirconia (YSZ) electrolyte layer to a stainless steel (SS) support. Unfortunately, wetting and adhesion problems of conventional silver or copper based brazes often result in braz e joint manufacturing defects (pores) and as - produced joint failure at the braze - YSZ interface . 23,87,133 To help overcome the poor wetting and adhesion issues of Ag on YSZ, a technique called reactive air brazing (RAB) has recently been developed . 25,134,135 Reactive air b razing is performed in air, and the braze alloy is intentionally designed to contain minor elements (such as Cu) that oxidize during brazing to promote wetting and adhesion. In fact, the addition of as little as 4 wt% CuO reduces the wetting angle of Ag on YSZ from 110 ° to ~45 ° . 22,128,136,137 Furt her, the fact that brazing can be done in the air prevents mechanical and electrochemical degradation of the oxygen - partial - pressure - sensitive SOFC cathode materials during brazing . 138 140 Lastly, Ag - CuO reactive air brazes have also shown an ability to bond metals to a variety of oxides including alumina, lanthanum strontium cobalt iron oxide, etc. 24,141 144 As depicted in Figure 2. 1, two mechanisms have been suggested to facilitate the wetting and adhesion of Ag C uO reactive air brazes on oxide surfaces. In Mechanism I , the presence of oxygen dissolved in molten Ag is hypothesized to cause a 20 ° - 35 ° drop in Ag on oxide wetting angles when brazing is performed in the air instead of inert atmospheres . 145 (Note, unlike most molten metals at elevated t emperatures in air, Ag is thermodynamically more stable than AgO under brazing conditions , 146 suggesting that wetting improvements via the formation of crystalline AgO are unlikely). In Mechanism II , CuO is hypothesized to improve wetting and adhesion by forming a CuO - rich phase between the Ag and YSZ . 24 35 Figure 2 . 1 . Hypothetical mechanisms of how Ag - CuO braze enables wetting starting from the solid phase (left) to melting and wetting (right) during the heating process in air. While the wetting angle of Ag on YSZ > 90 ° , Ag - O clusters in Ag( l ) solution may alter interfacial energy ( Mechanism I ), or addition of Cu may fac ilitate formation of CuO interlayer on YSZ surface ( Mechanism II ) leading to reduction of the wetting angle. The SEM images used for representation of stainless steel and YSZ microstructure was taken from literature . 147 In the first half of the present study, the primary objective is to use atomistic modeling to determine, without interference from other fa ctors, the extent to which these two mechanisms occur. We consider this is the basic step toward materials design. To examine Mechanism I , Ab Initio Molecular Dynamics (AIMD) was used to track the evolution of dissolved O 2 in liquid Ag near the Ag( l )/YSZ i nterface in Section 3.1. In addition, O 2 was initially placed at the Ag( l ) surface, the Ag( l )/YSZ interface, or the Ag( l ) center to determine if the dissolved oxygen lowered the wetting angle b y altering either the Ag liquid/ vapor surface energy ( ), the solid / liquid interfacial energy ( ), or both. To examine Mechanism II and determine the effect of a CuO interlayer between Ag and YSZ, DFT simulations were performed in Section 3.2 to calculate the work of adhesion, , of the Ag/YSZ, Ag/CuO, and Ag/YSZ interfaces. Ag wetting angles were then calculated from the computed surface and interface energies. Previous DFT - based surface studies have laid essential groundwork on the surface structures of these materials. Combined DFT 36 and thermodynamic calc ulations have identified the lowest energy facets for CuO to be the stoichiometric (111) for the entire O chemical potential region where CuO is stable . 148,149 For YSZ, which has a cubic ZrO 2 (c - ZrO 2 ) structure, the O - terminated - (111) surface was shown to be the most stable in the entire oxygen chemical potent ial range encountered in a SOFC. 150,151 This is in accordance with the experimental HRTEM images of Ni/YSZ interfaces . 91 The surface structure of 9% - Y - doped YSZ (111) , 15,100,152 which is shown experimentally to be in cubic phase at room temperature, has also been modeled vi a DFT and will be used as a basis for the present study . 153 In the second half of the present work (Section 3.4.3), the objective was to use the conclusion gained from previous sections (Section 3.4.1 and 3.4.2 ) that Mechanism II is the dominant one to develop a simple descriptor to predict Ag/oxide interface energies, predict Ag/oxide , and search for new oxide interlayers capable of promoting the wetting and adhesion of Ag on YSZ, while maintaining stability in reducing env ironments. Although many theoretical and DFT studies have shown that for a given oxide surface, metals with lower d - band occupancy (particularly bcc metals) can induce higher metal - O charge transfer, and hence higher 105,154 156 , studies of the reverse relationship, i.e. identifying the oxides yielding a higher on a given metal (such as Ag), are scarce . 103,105 New brazing materials with other oxides are needed because the reduction of CuO on the anode side of Ag/CuO SO FC braze joints (CuO is thermodynamically unstable in the water - hydrogen mixtures found under SOFC operation ) 146 induces pores that mechanically weaken the joint and limit joint lifetimes. 157 160 2 .3 Methods 2 .3.1 Geometry optimization All calculations in this study were performed using the plane wave DFT implemented in Vienna Ab Initio Simulation Package (VASP). The exchange correlations were treated with both 37 the Local Dens ity Approximation (LDA) and the Generalized Gradient Approximation (GGA) by . 161,162 Projector augmented wave (PAW) potentials were used to represent the effective electron - core interaction . 163 The electronic convergence was achieved self - consistently with a plane - wave cutoff energ y of 500 eV. The two convergence criteria were 10 - 5 eV for the total energy, and 0.03 eVÅ - 1 for forces during structural relaxation via conjugate gradient minimization. The k - spacing and k - point with the details of the cell dimensions for the main interfaces are shown in Table 2. 1 (the rest is shown in Table A 1 in the appendix ). Essentially, a k - spacing of studied interfaces are set at around 0.04 A - 1 or below. Given that the interface dime nsions are different in each interface, this constraint on the k - spacing resulted in various k - points, i.e. the longer dimension results in lower k - point as k - space is a reciprocal space. This spacing allows us to achieve an energy convergence of 1 - 2 meV/a tom compared to the energy calculated by higher k - points. Gaussian smearing was employed with a sigma of 0.02 to achieve less than 1 meV/atom contribution from entropy. A spin - polarized calculation was performed with dipole moment correction along the perp endicular direction of the slabs. For strongly correlated transitional metal oxide, on - site coulomb interaction as introduced by Dudarev (+ U eff ) was employed . 164 The values of U eff were taken from previous studies that employed LDA functional. These were 3.20 , 165 4.30 , 165 2.80 , 166 5.00 , 167 6.00 , 168 7.50 , 169 and 7.50 169 eV for the transition metals in - Cr 2 O 3 - Fe 2 O 3 , Rh 2 O 3 , NiO, c - CeO 2 , CuO, and Cu 2 O, respectively. In the case for GGA - PBE functional calculation of both CuO and Cu 2 O, U eff of 8.50 eV was used . 170 The details of these oxides, including surface termination, lattice constants, and surface ene rgies are tabulated in the appendix ( Table A 2). 38 Table 2 . 1 . The parametric detail of the interface cells with their corresponding k - points used in the present study. Interface a, b, c (Å) k - points LDA PBE PBE&LDA Ag/YSZ 7.37, 7.37, 58.91 7.49, 7.49, 61.70 90 ° , 90 ° , 120 ° 4×4×1 Ag/CuO 12.31, 2.84, 53.98 13.96, 2.97, 54.29 90 ° , 90 ° , 60 ° 2×9×1 CuO/YSZ 11.56, 7.26, 50.57 12.20, 7.31, 51.14 90 ° , 90 ° , 95 ° 2×3×1 2 .3.2 Slab Models for Surfaces and Interfaces Slab models were constructed to compute surface energies. The surface orientation of each oxide used in this study was the lowest energy identified in the literature ( See Appendix A) . All the oxide surfaces were stoichiometric in this study except the delafossite oxides and Cu 3 TiO 4 , as detailed in Table A 2. Particularly for the O - terminated (111) slab of YSZ, which adopts a cubic ZrO 2 structure at SOFC operating and brazing temperatures , 171 the surface structure was taken from p revious computational studies using (Y 2 O 3 ) 0.09 (ZrO 2 ) 0.91 with the most preferable sites for Y and oxygen vacancies ( identified . 15,100 Specifically, the was found to be at the subsurface of the first YSZ multilayer with Y substituting for Zr at the second nearest neighbors positions (2NN), in t he first and second multilayers. 100 This is in agreement with experimental observation that Y tends to segregate to the surface . 172,173 The minimum number of layers in each slab was determined to give converged surface energy, , within ±0.02 J/m 2 ; six layers for both Ag and Ni slabs, whereas for certain oxides, such as ZrO 2 , NiO, and CuO, three layers were sufficient, and oxides of co rundum type usually required five layers, while Cu 2 O needed four. 39 Figure 2 . 2 . The relaxed interface structures for investigation of Ag CuO/YSZ brazes. These include Ag( l )/YSZ (a), Ag/YSZ (b), Ag/Cu O (c), and CuO/YSZ (d). Only the top half of the sandwich model of the solid / solid interface is shown due to symmetry. As shown in Figure 2 .2 , all interface simulations utilized, unless otherwise stated, a sandwich model where two metal slabs sandwiched an oxide slab and 16 Å of vacuum was used to preclude interactions of the top and bottom metal slabs due to periodicity. In the sandwich model, the number of oxide layers in the middle was a double of that from converged surface energy calculation. Using thi s setup, the interatomic distances in the interior of each slab retained bulk characteristics, eliminating the need for atomic position constraints. To match the metal/oxide lattices, the sandwich models were constructed using Virtual NanoLab to generate s everal interface configurations. All models were constructed with mismatch strains below 6% in all directions. The interfacial lattice mismatch strain was partitioned equally between adjoining slabs. The relative lateral positions of the metal slab on the oxide slab were varied to at least 3 positions in order to 40 identify the most preferable position. All sandwich models were constructed to have at least inversion symmetry to ensure that the two interfaces were identical, less affected by the dipole moment, and that the interface energy, , was uniquely defined. The average interface width ( ) was defined as the z - distance between the averaged z - positions of the atoms in the first layer of each side of the interface. For YSZ and CuO, only oxygen atoms of the outermost layer were taken into account. For liquid Ag, only atoms within ±0.5 Å range from the bottom most atom s were used to compute the average. The work of adhesion of the system was defined and calculated using the equation: Eqn. ( 2 . 1 ) where is the surface energy of the slab i , is the solid/ solid interface energy, E i is the calculated energy of the slab i , E 12 is the calcul ated total energy of the interface system with slab 1 and 2, and A is the model interface area. Therefore, was computed based on the energy difference between a simulation cell with slab 1 and slab 2 far apart and where slab 1 and 2 shared an interfa ce. Then, the interface energy was computed using the relationship . The wetting angle corresponding to the Eqn. ( 2 . 2 ) where is the solid oxide surface energy from DFT calculation report ed in Table A 2, is the DFT calculated solid/ solid interface energy from Eqn. 2. 1, is the solid metal surface energy calculated by DFT listed in Table A 2, and is the estimated contact angle. Except for one liquid Ag on YSZ interface calculation, all the other interfaces were solid/ solid interfaces. It 41 should be noted that although the absolute magnitude of the values obtained from the solid/ solid interface calculations, instead of liquid/ solid, may be overestimated due to the higher surface energy of a solid compared to its liquid , 174 but the calculated trends in wettability are expected to hold at brazing temperatures. 2 .3.3 Ab Initio Molecular Dynamics Ab - initio molecular dynamics (AIMD) was used to determine the adsorption and wetting behavior of Ag( l ), with and without dissolved oxygen, on YSZ. At a brazing temperature of 1000 ° C, the solubility of O 2 in Ag is 1.93 mL, which translates to approximately 2 a toms of oxygen per 200 Ag atoms. 175 Hence, in these AIMD simulations, the Ag( l )/YSZ system was modeled as a ~1 nm thick molten Ag layer containing two oxygen atoms (for a total of 102 atoms) supported on a YSZ slab three unit cells thick. The basal plane YSZ atoms were fixed in position, the YSZ had a surface area of 181.44 Å 2 , and a vacuum gap of 13 Å was placed above the liquid surface in the init ial structure. This initial structure corresponds to the experimentally measured density of 9.30 g/mL for Ag( l ) at 1000 ° C. 176 As shown in Figure 2. 3, three different initial structures were generated by inserting the two oxygen atoms at either the YSZ interface, within the bulk of the Ag( l ), or at the free surface of the molten Ag. A lthough the dissolved oxygen concentration used here was double the experimentally observed amount due to computational limitations, the calculated energy trends with oxygen position should still be applicable to real - world systems. Both the molten Ag and the Ag( l )/YSZ interfaces were initially relaxed using classical MD with a COMPASS force field (Ag = ag_m, O in Ag = o_ag, Y = y3o, Zr = zr4o, and O in YSZ = o2z) implemented in Material Studio. 177,178 The force field was previously shown to give a melting temperature of metal closer to first - principle calculations than emb edded atom model (EAM) method. 179 The evolution of these initial structu res was simulated by constant - temperature, 42 constant - volume ensemble (NVT) AIMD using spin - polarized DFT calculations with GGA - PBE - point k - point. The constant temperature was controlled at 1280K by t and the integration time step is 1 fs for a total of 3 ps. The average energy for the c alculation was then obtained by sampling and minimizing 3 frames from the last 0.5 ps of the AIMD run using (2 × 2 × 1) k - points. 2 .4 The Role of the Oxygen at the Ag( l )/YSZ interface during Brazing Process Figure 2 . 3 . Ag( l )/YSZ interface structures with dissolved oxygen atoms at the beginning and end initial positions of Ag O clusters were investigated, including in the bulk Ag( l ), at the free surface of the Ag( l ), and at the Ag( l )/YSZ interface. The Ag O at the interface is the most favorable structure. AIMD was employed to investigate the energetics in Ag( l )/YSZ interface in order to examine the mechanism of dissolved O 2 in enhancing wettability, i.e. Mechanism I in Figure 2. 1. AIMD was first performed to investigate dissolved O 2 in bulk liquid Ag. Dissociation of O 2 into Ag O clusters was promptly o bserved, where O atoms formed 2.2 - 2.3 Å bonds with the Ag atoms surrounding them. This bond length is 5~10% longer than the Ag - O bond length in crystalline Ag 2 O (2.08 Å in GGA - PBE relaxed structure), a reasonable increase for the liquid phase. Analysis 43 of the system by pair radial distribution (RDF) shown in Figure A 2 revealed that the nearest Ag - Ag distance peak was ~2.8 Å, in accordance with experimental measurements of pure Ag in a reducing atmosphere at 1050 ° C . 180 This indicated that the dilute concentration of oxygen in the Ag did not significantly alter the bulk behavior of the Ag. Figure 2. 3 shows three possible sites for Ag O clusters to be at , including in the bulk (a), at the surface of molten silver (b) , or at the interface of liquid Ag/YSZ (c). Although not the objective of the present work, oxygen vacancies in all these initial positions were observed to move around within the YSZ slab during the AIMD simulations . Figure 2. 3a and 2. 3b show that Ag with Ag - O clusters in the bulk of the Ag or placed at the Ag - free surface drifted away from the YSZ (yielding a GGA - calculated ~ 0 J/m 2 that may be a low considering the fact that GG A tends to underestimate binding energies). In contrast, Figure 2. 3c shows that Ag with Ag - O clusters at the Ag( l )/YSZ interface remained attached to the YSZ with a GGA - calculated = 0.43 ± 0.01 J/m 2 . The arrangement in Figure 2. 3c, denoted henceforth as {Ag( l )+2O}/YSZ, had a lower overall energy than those in Figure 2. 3a and 2. 3b, indicating that changes in the Ag( l )/YSZ interface, and not changes in the Ag free surface energy, are responsible for the Ag wetting enhancements expe rimentally observed when switching from inert Ag/YSZ brazing atmospheres to air . 25,87 Without the dissolved O 2 , the pure liquid Ag has an average (on YSZ from GGA - PBE functional) of only 0.11 ±0.01 J/m 2 , resulting in a contact angle of 149 ° (Table 2. 2). This is consistent with experimental observations showing that liquid Ag does not wet YSZ . 87 Although the results from GGA may be overestimated, the trend and the relative change of the wetting angles are expected to be reliable as will be shown in the next section. As shown in Table 2. 2, the {Ag( l )+2O}/YSZ wetting angle was 38 o less than that for pure Ag( l )/YSZ , in agreement with a 30 o drop observed experimentally 25,87 (Table 2. 2 also shows that the bin ding energy 44 underestimated by the GGA - PBE method resulted in wetting angle predictions that were ~25 o higher than those observed experimentally). Figure 2. 3c also shows that the two oxygen atoms originally dissolved in the Ag near the Ag/YSZ interface hopp ed into the YSZ substrate, filling two YSZ oxygen vacancies in the process. The resulting increase in the YSZ surface oxygen vacancy concentration likely contributed to the increased by providing additional sites for Ag - O bonding across the interface. Together, these results confirm Mechanism I (the hypothesis that oxygen dissolved in molten Ag is responsible for the ~20 ° - 35 ° drop in Ag on oxide wetting angles when brazing is performed in air instead of inert atmospheres). Table 2 . 2 . d ), work of adhesion ( ), calculated wetting angles ( ) , and experimental ( ) for Ni/YSZ interface and systems involved in Ag CuO/YSZ braze. d (Å) Ni[ ](111)||YSZ[ ](111) 2.03 0.49 87 Ag( l )/YSZ GGA 2.41±0.02 0.11±0.01 149 87 {Ag( l )+2O}/YSZ GGA 2.17±0.04 0.43±0.01 111 ° 181 87 25 45 2 .5 Enhancement of Interface Adhesion by CuO Table 2 . 3 . The surface and interface energy involved in Ni DFT computed GGA - PBE 1.94 0.70 2.26 0.49 LDA 2.68 1.42 2.68 1.42 Experiment 64 2.05 93 1.23 94 1.8 ±0.1 1.48 Before studying these solid interfaces, the accuracy and reliability of the results were evaluated. A Ni/YSZ interface was employed because of abundant existing computations and detailed experimental results, in contrast to limited studies of Ag adhesion on YSZ or ZrO 2 . Table 2. 3 shows the surface energy, interface energy, and from GGA and LDA functionals for Ni . The agreed well with the experimentally measured values. In fact, the LDA - predicted of 1.42 J/m 2 was within 5% of the 1.48 J/m 2 value obtained by converting Nahor et al. experimentally measured Ni[ ](111)||YSZ[ ](111) interface energy and surface energies. The interface energy was measured from Winterbottom analysis of the same Ni[ ](111)||YSZ[ ](111) interface, equilibr ated by annealing at 1350 ° C in high - purity inert gas and H 2 ( P O2 = 10 - 20 ) . 64,92 As is the primary focus of the subsequent sections, LDA functional was considered to give comparable results to the experiments. Further discussion on validation of the model and comparison to previous computational studies of related interfaces can be found in the appendix . To evaluate Mechanism II (the hypothesis that CuO interlayers atop YSZ promote Ag wetting and adhesion), values for Ag(s)/YSZ, Ag/CuO, and CuO/YSZ were calculated using the solid/ solid interfaces shown in Figure 2. 2a - d. Solid interfaces were used because components 46 are eventually solidified and during actual SOFC operations, the strength of braze/YSZ interfa ce was then determined by the adhesion of the two interfaces of Ag(s)/CuO/YSZ. Also, the work of adhesion with liquid Ag is likely to follow the same trend with solid Ag. As shown in Table 2. 2, the trend of the results from both LDA and GGA - PBE exchange c orrelations functionals agreed that the presence of CuO leads to a more adherent braze. The closer interface distances shown in Figure 2. d values of Table 2. 2 also correlated well with the increased . The higher resulted in reduction of wetting angle of ~48 ° when replacing Ag/YSZ with Ag/CuO. Particularly from the prediction by LDA functional, the increment in alone would theoretically be substantial enough to make Ag wettable on CuO (<90 ° ). On the other hand, the high for CuO/YSZ indicates that CuO can firmly adhere to YSZ. While the liquid/ solid interface mimics the braze processing condition and the solid / solid interface represents the solidified braze during SOFC operation, the adhesion and wetting angle between solid/solid and liquid/ solid are directly related. Hence the results from one can be translated to the other. Even though these simulations were performed on solid / solid interfaces, these conclusions should hold for a system with molten Ag for the following reasons: 1) both experimental measurements 174 and thermodynamic derivation 182 have shown that solid Ag surface energies are systematic ally higher by ~20 - 33% of those of liquid Ag, the computed on different interfaces should follow the same trend for liquid and solid Ag. 2) Wetting angles of many liquid metals are known to be a very weak function of temperature (for example, the of Ag( l ) on ZrO 2 only changes by 0.05 J/m 2 from 1000 to 1300 ° C ). 23,133,183 Thus, the absolute values of the and obtained from the solid/ solid interface will be overe stimated due to the higher surface energy of a solid compared to its liquid , 174 but the calculated trends in both values are expected to hold at brazing temperatures. 47 It should be noted that the experimental wetting angles of Ag/CuO listed in Table 2. 2 Ni[ ](111)||YSZ[ ](111) interface. 64 in practice, the experimental result has a combined effect from Mechanism I and II 25 2 .6 Identification of Other Potential Oxide - Forming Alloying Elements A descriptor that correlates a simple interface structure and bond strength to interfa ce energy at the Ag/oxide interfaces was identified to facilitate the sear ch for potential oxides. The descriptor was based on the results from the projected density of states (PDOS, Figure A 4), which is discussed in the Appendix A and indicates that interfacial bonding occurred primarily from the metal - oxygen bonds. The calculations pertaining to descriptors were performed using LDA functional only as results from previous sections indicate comparable to experiments. 2 .6.1 F ormulation of a Descriptor Using and in a Predictive Model The aforementioned descriptor was developed based on the definition of the interface energy, i.e. the excess energy introduced by the interface compared to the bulk structure. The first assumption used in this descriptor was that no new Ag - oxide phases formed during brazing. Since 48 the Ag - Ag bond strength is much lower than that of metal - oxide bonds, t he descriptor was simplified to only account for 1) the number of unsaturated cation and anion bonds per unit surface area, and 2) the energy changes per unit surface area caused when a portion of the oxygen dangling bonds on a surface was replaced with the newly formed Ag - O bonds across the oxide - Ag interface. This leads to the newly formulated the descriptor ( ): Figure 2 . 4 . Side view and top view of surface and interface structures to il lustrate the formulation of the descriptor, , in Eqn . 2. 3 . The and were demonstrated for a) an unrelaxed rock salt CaO {100} surface, where f - Al 2 O 3 (0001). The relaxed interfaces of Ag/corundum - oxide demonstrate that - Al 2 O 3 and o ther corundum - oxides f 3+ , weakens Ag O interfacial bonding, except c) in Fe 2 O 3 , whose cation, Fe 3+ , submerges beneath the oxygen layer after relaxation so it does not interfere with the Ag O bonds ( f arrows pinpoint the same cation in all the corundum oxides . Eqn. ( 2 . 3 ) where and were the number of dangling bonds on the surface cation and oxygen atoms, respectively, in an oxide surface unit cell, was the bond enthalp y of the oxide as defined by oxide formation enthalpy ( ) d ivided by the number of bonds in a unit cell, was the 49 bond enthalpy of silver oxide (Ag 2 O), and was the surface area of an oxide unit surface. was an empirical structure factor assigned to be 1 for all studied oxides, except most of the corundum oxides such as Al 2 O 3 , Cr 2 O 3 and In 2 O 3 which, as highlighted in Figure 2. 4b, contained protruding surface cations necessitating the use of f =2 (further explained in the next paragraph). Figure 2. 4 shows the surface and interface structures used to calculate the descriptors for three sample oxides: CaO - Al 2 O 3 , and Fe 2 O 3 . Figure 2. 4a shows that on the {100} surface of CaO, both Ca and O have one dangling bond (resulting in and ). Dividing an LDA - calculated - 6.78 eV (which is comparable to - 6.39 eV obtained by extrapolating NIST database data down to 0 K), by the 6 Ca - O bonds per unit cell yields - 1.13 eV/bond. In a similar fashion, LDA calculated the relaxed structure to be - 0.14 eV/bond. The relaxed CaO {100} surface has of 11.14 Å 2 . Together with f =1, this yielded a of 0.19 eV/Å 3 . Figure 2. 4b shows the surface structure of corundum - type oxides, represented by Al 2 O 3 (0001) and its relaxed interface with Ag. It has a 3, a , a LDA calculated relaxed structure Å 2 , and a LDA calculated relaxed structure - 17 .42 eV which resulted in a - 1.45 eV, with an f factor of 2, this yielded a of 0.85 eV/Å 3 . Among all the corundum oxides tested, namely Al 2 O 3, Cr 2 O 3, In 2 O 3, and Rh 2 O 3 all have protruding (0001) surface cation (pointed out by the yellow arrow in Figure 2. 4). The protruded cation prevents the silver slab to get closer to the oxygen layer ( Figure 2. d of Ag/Al 2 O 3 is 2.58 Å compared to 1.81 Å in Ag/Fe 2 O 3 (Fig 2. 4c). This behavior may account for the weak Ag/Al 2 O 3 bonding reported in the literature. 184 Consequently, to account for the increase in interfacial energy, f =2 is used for most of the corundum oxides with protruded cations. It is interesting to note that all the corundum type oxides studied here, except Fe 2 O 3 , had protruding 50 (0001) surface cations that necessitated the use of f = 2. In Fe 2 O 3 , whose cation, Fe 3+ , submerges beneath the oxygen layer after the relaxation so it does not interfere the Ag - O bonds ( Figure 2. 4c). This means that there is no need to increase the structure factor, so the calculation of for Fe 2 O 3 follows the similar manner as Al 2 O 3 but with f =1 for the smooth top O layer. Table A 2 in the appendix summarizes all the descriptor calculations performed in the present work. The unique behavior of Fe 2 O 3 may be due to its strong 3d electron interaction with the oxygen polyhedron. In M 2 O 3 corundum oxides bulk structure, the cation M 3+ occupies 2/3 of the o xygen octahedral sites. On the M 3+ terminated surface , the protruding M 3+ without unpaired electrons will relax toward the oxygen due to electrostatic interactions, as for Al 2 O 3 and In 2 O 3 . Comparing Fe 3+ (3d 5 ), Rh 3+ (4d 6 ), and Cr 3+ (3d 3 ) in Cr 2 O 3, Rh 2 O 3, and Fe 2 O 3, respectively, the interaction between the oxygen octahedral may attract Fe 3+ (3d 5 ), which has the most unpaired d - electrons to move underneath the top oxygen layer. Due to the complexity of computing strong correlated d - electrons in transition metal oxides, more research is warranted. It should be noted that the use of surface energetics, classical view of charge transfer, and metal - oxide formation energies, to formulate a descriptor to relate directly to interface energy and work of adhesion for Ag/oxide interface has been tried, but we did not find no strong correlation. 2 .6.2 Descriptor - Based Prediction of Interface Energy and Figure 2. 5a shows the DFT computed interface energy and its relationship with in various types of oxides. The interface energy monotonically increases with the descriptor. The simple descriptor defined in Eqn. 2. 3 enables assessment of oxide properties that will lead to high or low interface energy. Two factors can affect the interface energy; one is the oxide bond energy ( ), which has a chemical nature. The other is the density of the dangling bonds, which 51 depends on the geometry of the surface. The increase in either of these would increase the value of descriptor, and hence the interface energy . Figure 2 . 5 . The correlation a) between of the interface energy ( ) of Ag/oxides and the descriptor ( ), which followed a logarithmic function, and b) between the predicted work of adhesion ( ) at the Ag/oxides interface and the DFT calculated ones ( ). In Figure 2. 5a, the interplay between these two factors is evident. For example, ZrO 2 , which has the highest bond energy among oxides in this study ( - 2.39 eV/bond), did not show the highest interface energy. Al 2 O 3 has the highest interface energy, although its bond energy of - 1.45 eV/bond is not as large as ZrO 2 . This is due to a different surface geometry, as Al 2 O 3 contains 1.7 times larger surface dangling bond density than ZrO 2 . In fact, the oxides with higher interface energy with Ag (Fig 2. 5a) was dominated by corundum type oxides. On the other hand, in the lower region of interface energy, although Rb 2 O and CdO both had similarly low bond energy of - 0.46 and - 0.43 eV/bond, re spectively, Cu 2 O with the bond energy of - 0.51 eV/bond exhibited much lower interface energy due to its low dangling bond density of 0.07 bonds/Å 2 , compared to 0.11 and 0.18 bonds/Å 2 of Rb 2 O and CdO. This simple descriptor also means that if oxides of inte rest possessed the same surface geometry, and likely the same crystal structures, the descriptor could be 52 formation energy. This trend can be seen in Fig 2. 5, as d ifferent crystalline structures were labeled with different symbols. Therefore, for oxides within the same crystalline family, both interface energy and can be simplified to correlate with their formation energies as shown in Figure A 6 for corundum a nd rock salt oxides and suggested by Li et al . 103 The results plotted in Fig 2. 5a demonstrated that the descriptor is well - correlated with the DFT calculated interface energy, which increases with the descriptor. A fitting equation: Eqn. ( 2 . 4 ) was obtained with a coefficient of determination R 2 = 0.95. Hence, this equation can be used to estimate interface energy ( ) solely based on the descriptor. The predicted interface energy was then used to predict as defined by Eqn. 2. 1, with tabulated surface energies, which is easier compute with DFT calculations. In this study, the surfaces of these oxides were sel ected based upon a criterion of being the lowest surface energy. The individual surface energy of these oxides is tabulated in Table A 2 and discussed in Appendix A . The predicted was plotted against the DFT - calculated in Fig 2. 5b, which al so demonstrates satisfactory accuracy ( Figure 2. 5b). The conclusion obtained from the two ends of the plot indicated that ZrO 2 as well as YSZ were among the most difficult oxides for Ag to be wetted. On the other hand, while both Fe 2 O 3 and Rh 2 O 3 provided high adhesion, like CuO, both of them are not stable under SOFC reducing environment (Fe 2 O 3 would be reduced at temperature above 587 ° C). Within Figure 2. 5b, a candidate with the highest that would be stable in a reducing environment is In 2 O 3 , but its is only 1.03 J/m 2 , 0.4 J/m 2 lower than CuO. The descriptor was also employed to screen more single cation oxides but, among those, none gave a higher with Ag than CuO. For example, a rutile SnO 2 (110) was predicted based 53 on the de scriptor to have of 1.23 J/m 2 . For validation purposes, the DFT - calculation gave 1.10 J/m 2 , using a half - sandwich interface model. The translated to a borderline wetting contact angle of 90 ° , which when combined with the effect from disso lved oxygen in Ag( l ) could enable Ag wetting, although it would not be as good as CuO. In addition, SnO 2 is a bit more resistant to the reduction environment, as it is expected to be stable up to 647 ° C according to Ellingham Diagram. It have been suggested that V 2 O 5 could be another potential oxides; 24 however, due to the step - like structure of the (010) surface, a smooth surface of Ag used in the present descriptor mo del could not accurately determine of this interface . In conclusion, our screening of monocation oxides could not identify a candidate that would yield similar or higher than CuO, while not being cost - prohibitive or reduced in SOFC operating co nditions. 2 .6.3 Identification of Complex Oxides and Ternary Braze Alloy Design Since the simple oxide search did not predict many candidates that could potentially be 1) better than CuO in enhancing and 2) stable in a reducing SOFC environment, the descriptor was used to search for potential multi - cation oxides. First, the descriptor was used to explore multi - cation oxides with lower descriptor values that lead to lower interface energy (since a low interface energy is a necessary but not sufficient condition for a high work of adhesion). Then, the surface energies were computed by DFT. The stable surface for all delafossite oxides was chosen to be O - terminated (0001) surfaces, based on experime ntal a nd computational evidence. 185 187 Predicting the interface energy from the descriptor using Eqn. 2. 4, and the DFT computed surface energies, the was first predicted before the DFT interface model was constructed. From this, it was found that several oxides with a delafossite structure could exhibit strong adhesion due to their high surface energy relative to a low predicted interface energy ( Table 2. 4). For comparison, CuO is listed as well. While some delafossite oxides exhibited low adhesion ~0.7 J/m 2 with Ag, 54 such as CuCrO 2 and CuFeO 2 , several delafossite oxides, CuAlO 2 , CuGaO 2 , and CuInO 2, showed higher predicted (2.3~2.6 J/m 2 ) with Ag than CuO (1.43 J/m 2 ). Cu 3 TiO 4 (010) also yielded a predicted of 2.12 J/m 2 , which is higher than that of CuO. Cu 3 TiO 4 has a similar structure to delafossite, and its (010) surface is equivalent to the (0001) surface of delafossite oxides. The incre ase of on Cu 3 TiO 4 surface could also explain the observed enhanced braze wettability when TiO 2 is added to the Ag CuO system. 136 TiO 2 itself is not the likely cause of wettability enhancement, as predictions for both the rutile TiO 2 (110) and anatase TiO 2 (101) structures yielded of only 0.40 and 0.56 J/m 2 , respectively. However, TiO 2 could react with CuO or Cu 2 O to form Cu 3 TiO 4 . 188,189 Therefore, the addition of Cu and Ti to Ag, with modification of brazing gas environment and temperature conditions to purposely generate Cu 3 TiO 4 on the YSZ surface could als o further enhance the wetting of Ag. Also, as indicated by its formation energy ( - 13.12 eV) on the Ellingham diagram, Cu 3 TiO 4 will be stable in a reducing SOFC environment. Another interesting oxide is CuAlO 2 . The formation of CuAlO 2 would be favorable in certain conditions as the CuAlO 2 formation energy was calculated to be - 9.70 eV, which is comparable to - 9.74 eV for ½(Al 2 O 3 + Cu 2 O). Experimental phase diagrams indicate that CuAlO 2 is stable at high temperature, especially in the b razing and SOFC operating temperature range 190 . Putting the calculated formatio n energy of CuAlO 2 on the Ellingham diagram shows that it is stable in a reducing environment. To further investigate this promising oxide and also obtain the accurate , DFT calculations were performed ( Figure 2. 6a). The DFT - calcualted of Ag/CuA lO 2 . is 3.04 J/m 2 , almost twice of that at the Ag/CuO interface. The changes in PDOS through formation of this interface shown in Figure 2. 6b correspond well to its high , where oxygen p - states moved from high - energy antibonding ( - 2 to 0 eV) to bonding states at energies below the metal d - 55 states (~ - 8 eV). This high adhesion will lead to complete wetting (0 ° ). Moreover, DFT calculation using half - sandwich model also indicate strong adhesi on ( = 2.04 J/m 2 ) between CuAlO 2 and YSZ. Overall, this is in accordance with reports that the greater extent of CuAlO 2 formed from the unintended reaction of CuO with Al 2 O 3 , the lower contact angle of Ag . 145 The other delafossites, CuGaO 2 and CuInO 2 , and Cu 3 TiO 4 were predicted to yield similarly high are therefore are expected to yield similar results. However, the formation energy of CuInO 2 , which is the highest among investigated delafossites ( - 5.70 eV), was only - 0.52 eV below the hydrogen line in Elling ham diagram. Given that LDA tends to overestimate bonding energy, it is more prudent to assume that CuInO 2 would not be stable in a reducing SOFC environment. Therefore, several oxide interlayers, CuAl(or Ga)O 2 or Cu 3 TiO 4 , that could improve the interfaci al strength were identified via our computational search. In order to form these oxide layers, the complex oxide could either be pre - sputtered or coated on the YSZ surface before brazing or be directly added to Ag, as similar methods have been used for Ag - CuO system. 145,191 Alternatively, alloying Ag with a mixture of Cu and Al (or Ga, Ti) and fine tuning the brazing gas environment and temperature conditions to intentionally maximize th e formation of CuAl(or Ga)O 2 or Cu 3 TiO 4 could also improve the interfacial strength. It should be noted that, according to the original definition, in order to be classified as a RAB component, the oxides formed by the secondary metals in air must be at l east be partially soluble in Ag 192 in order to avoid forming thick oxide scale on the liquid surfa ce to inhibit wetting. The oxide solubility, the impact of these alloy or oxide additions on the melting temperature, can be considered in follow - up studies. Nevertheless, these examples suggest that ternary alloy systems may offer many more possibilities. The discovery of potential oxides can be expanded beyond the examples given here. 56 Figure 2 . 6 . The relaxed structure of the Ag/CuAlO 2 by LDA functional showing an interface gap - shift in the oxygen's p - states. This indicates strong interactions between Ag and CuAlO 2 (b). 2 .7 Conclusion The RAB process is a promising solution to solve wetting and adhesion problem of Ag on an YSZ surface for sealing applications as well as providing practicality for industrial settings. As the current Ag - CuO system has a short - coming due to pore formation from re duction of CuO to Cu, there are alternative approaches of other oxide interlayer design that may be effective. Using DFT calculations and interface models, the present paper has identified the two primary mechanisms that enabled Ag to wet on YSZ surfaces, the Ag/YSZ interface from the oxygen in solution ( Mechanism I ), and 2) enhanced by formation of a CuO scale on top of the YSZ ( Mechanism II ). Mechanism I reduces the wetting 57 - wetting to wetting. The electronic analysis of the involved interfaces shows that interactions between Ag and oxygen atoms of the oxide dominates the interfacial bonding. With this knowledge, a simple descriptor for calculation of interface energy was developed to reduce the dependence on expensive interface calculations. With known sur face energy, which is much more easily calculated, the descriptor can also be used for prediction of . Several single cation metal oxides were considered, but the number of those with comparable to CuO and tolerant of a reducing environment were few. For example, none of the investigated metal oxides, with crystalline structures of rock salt, fluorite, and anti - fluorite, showed high . However, by expanding the search to multi - cation oxides, several promising candidates in oxides with delafos site structure, such as CuAlO 2 , CuGaO 2 , and Cu 3 TiO 4 , were discovered to double the work of adhesion of CuO with Ag. The study provides insight to geometry and bonding energy of metal - oxide interfaces with implications beyond just Ag and YSZ interfaces. The convenient search also opens new pathways to develop potential ternary Ag - oxide braze alloys with enhance d adhesion and wetting on YSZ. 58 CHAPTER 3 Experimental Identification of the Yields Determining Functional Group s in Lignin s 3 .1 Summary Lignin depolymerization to aromatic monomers with high yields and selectivity is essential for the economic feasibility of many lignin - valorization strategies within integrated biorefining process es. Importantly, the functional groups (henceforth referred to as p roperties) of the lignin source play an essential role in impacting the conversion chemistry, yet this relationship between lignin properties and lignin susceptibility to depolymerization is not well established. In this study, we quantitatively demonstrat e how the detrimental effect of a pretreatment process on the O - 4 content, limit high yields of aromatic monomers using three lignin depolymerization approaches: thioacidolysis, hydrogenolysis, and oxidation. Through p H - based fractionation of alkali - solubilized lignin from hybrid poplar, this study demonstrates that O - 4 linkages, phenolic hydroxyl groups, molecular weight, and S/G ratios exhibit strong correlations with each other even after pretreatment. Furthermore, the differences in these properties lead to discernible trends in aromatic monomer yields using the three depolymerization techniques. Based on the interdependency of alkali lignin properties and its susceptibility to depo lymerization, a model for the prediction of monomer yields was developed and validated for depolymerization by quantitative thioacidolysis. These results highlight the importance of the lignin properties for their suitability for an ether - cleaving depolyme rization O - 4 content. Therefore, this research encourages and provides a reference tool for future studies to identify new methods for lignin - first biomass pretreatment and lignin valorization that emphasize 59 preservation of lignin qualities, apart from focusing on optimization of reaction conditions and catalyst selection. 3 .2 Introduction Technologies for the conversion of renewable, non - food plant biomass to fuels, che micals, polymers, and materials have the potential to displace fossil sources of carbon while simultaneously contributing to rural agricultural economies and lowering CO 2 emissions. 193 While technologies for the production of cellulosic biofuels are beginning to be commercialized in the U.S., Europe, and Brazil, substantial technical, logistical, and economic challenges still remain. 194 In particular, one economic challenge is that current processes produce low - value, high - volume biofuels that require substantially higher capital costs per unit of biofuel produced relative to starch - or sucrose - derived biofuels. This is a consequence of plant cell wall recalcitrance, whereby the polysaccharides that may serve as the feedstock for biological conversion to biofuels are embedded in the plant cell wall matrix. This cell wall matrix can be considered as a composite material comp rising polysaccharides (cellulose and hemicelluloses) and lignin that presents a 195 Diversifying the product portfolio of lignocellulose - to - biofuels processes to include renewable, bio - based chemicals is widely recognized as an opportunity both to improve the economics of these processes and to buffer against market fluctuations. Thus, th ere is an obvious and compelling need to develop integrated processing approaches that enable multi - product biorefineries. Despite comprising a substantial portion of plant cell walls (as much as one third of the mass of lignocellulosic biomass) and repres enting the second - most abundant naturally occurring biopolymer after cellulose, the chemical functionality in lignins remains largely unutilized as a source of renewable carbon for the production of bio - based chemicals, polymers, 60 and materials. As such, th ere is an opportunity for technologies that can be integrated with conversion processes employing a biomass pretreatment/fractionation strategy, whereby it may be possible to utilize all or a fraction of the process - modified lignins for purposes other than as a low - value solid fuel. 196 Commercialization of co - products from wood lignins derived from Kraft pulping have long proved challenging, but it should be understood that lignin origin ( e.g. , hardwood, softwood, or grass) and processing history ( e.g. , alkaline pulping, organosolv pulping, sulfite pulping, dilute - acid pretreatment, etc . ) have an enormous impact on the structure, properties, and suitability of the lignin for a target application, which this work will investigate. Although lignin valorization has increasingly been the subject of substantial research, significant challenges still remain. 197,198 General strategies for lignin utilization include thermochemical conversion to heat and power or syngas, utilization of (potentially modified) polymeric lignins in applications such as resins for application as adhesives or coatings, 199 and lignin depolymerization to generate sets of aromatic compounds that can serve as bio - based alternatives to petroleum - derived aromatic platform chemicals. 200 Notably, a recent DOE report has highlighted the critical need for utilizing lignin - derived compounds as potential platform chemicals as well as the challenge in achieving this due to the difficulty of genera ting a limited range of aromatic compounds from lignin at high yields and selectivities that can act as renewable aromatic monomer cognates to existing petrochemicals. 201 This challenge is derived both from the random nature of the lignin polymer as a consequence of its synthesis via free radical - mediated oxidation and its propensity for modification during conversion or extraction processes. Strategies for the depolymerization of lignin include catalytic cra cking/pyrolysis, 202,203 oxidation 204 and hydrogenolysis - hydrodeoxyg enation, 205,206 and base - catalyzed depolymerization. 207 As examples, the depolymerization of lignin using solid acid - catalyzed hydrolysis has been reported to yield 61 primarily 4 - propenyl guaiacol/syringol products, while base - catalyzed depolymerization can yield additional aromatic products. 207,208 Oxidative cleavage of lignin in many studies aims to achieve a defined set of products including vanillin, vanillic acids, and their syringyl counterparts. 209 211 A number of studies have investigated molecular oxygen or hydrogen peroxide as an oxidant in the presence of a transition metal - based homogeneous catalysts including Cu, Fe, or V. 212 215 Other oxidative approaches include using a combination of formic acid and TEMPO catalysts, or polyoxometalates. 216,217 Lignosulfonates derived from sulfite pulping of wood have been utilized s using catalyzed and/or uncatalyzed alkaline oxidation with vanillin yields of up to 15% reported. 204,218,219 Vanillin production from Kraft lignin using these approaches has been extensively explored, although yields are typically lower compared to lignosulfonate s since Kraft lignins may be more condensed. 220 Approaches for reductive cleavage, particularly hydrogenolysis, have received substantial recent attention, and a number of heterogeneous catalysts have been proposed including ligan d - less Ni, 221 Ni alloys, 222 NiAu, 223 Ni/SiO 2 , 224 or Pd/C. 225 The products differ depending on reaction conditions. A mild liquid reaction of range ~ 200 °C will predominantly produce guaiacol or syringol with the alkyl side chain intact at the 4 - position. The amount of substitution of this alkyl side chain with hydroxyl groups can vary depending on reaction conditions and catalyst. 226,227 Under more severe conditions, ring hydrogenation and hydrodemethoxylation can occur to yield saturated compounds such as 4 - propylcyclohexanol or 4 - propylcyclohexane, which are suitable for fuel applications. 228,229 For many depolymerization approaches, enormous effort has been invested in the study of reaction conditions, particularly selection of catalysts and solvents. 40,230 232 More importantly, these conditions were also often used as a benchmark to determine the effectiveness of the 62 depolymerization approach. On the other hand, the structural features of lignin used in the reaction are often n ot considered, despite several studies reporting significantly lower yields when the same methods performed on lignin dimer models were applied to real lignin. 222,233 Specifically, ther e has not been conclusive evidence demonstrating how these structural features are fundamentally related, especially after lignin - modifying extraction or pretreatment, and how it would affect es have suggested that native lignin in biomass with higher syringyl/guaiacyl unit ratios (S/G ratios) give higher monomer yields O - 4 content. 234,235 Many lignin - modifying pretreatments have been investigated, 236 238 wit h each of these imparting changes to the physical and chemical properties of lignin through different mechanisms (for example by the cleavage of C - O bonds). However, many of these pretreatments can also result in the formation of new C - C linkages which hin der approaches to generate aromatic monomers. 239 241 Knowledge ity to depolymerization could help in the integration of strategies for lignin conversion into biorefining processes. Therefore, the objective of this study is to develop a better understanding of the relationship between lignin properties and to model th eir effect on lignin susceptibility to depolymerization through cleavage of ether linkages. This will be achieved through solubility - based fractionation of a hardwood lignin from an alkaline pretreatment liquor in order to generate lignin fractions exhibit ing diverse properties. These fractions will be characterized and subjected to three depolymerization approaches. Correlations amongst the lignin properties and between obtained yields will be established in order to develop a methodology for assessing lig nin quality for its suitability for depolymerization. 63 3 .3 Methods 3 .3.1 Lignin Fractionation Pretreatment was performed in an M/K Systems, Inc. (Peabody, MA, USA) single vessel pulp digester with a capacity of 10 L using debarked wood chips of 20 - year old hybrid poplar ( Populus nigra x maximowiczii cv. NM6) obtained from Dr. Raymond Miller (Michigan State University Forest Biomass Innovation Center, Escanaba, Michigan). Pretreatment was performed at a liquor - to - wood ratio of 4:1 using conditions identical t o those in our previous work for an H - factor of 166. 242 These mild pretreatment conditions wer e utilized to preserve solubilized lignin quality. Lignin was fractionally precipitated from this liquor through two sequential pH reduction approaches. The first one employed sequential acidification and precipitation at using CO 2 at room temperature and atmospheric pressure. For room temperature fractionation, pH was decreased stepwise as shown in Table 3. 1 until a pH of 9.0 was achieved and yielding 4 fractions. A final fraction was generated by further acidification to a pH of 2 by addition of 1 M H 2 SO 4 . The second lignin fractionation approach employed the Sequential Liquid - Lignin Recovery and Purification (SLRP) process using CO 2 acidification at elevated temperature (115 °C) and pressure (6.2 bar) as reported in our previous work. 243 For this, the SLRP fractions represent the lignin recovered by acidification between a pH of 11.0 of 10.5 (SLRP - 1) and the pH increment of 10.5 to 10.0 (SLRP - 2) as shown in Table 3. 1. The Klason lignin content of each pH fractions, as determin ed by NREL standard procedure (NREL/TP - 510 - 42618), was used as a basis for subsequent characterization and monomer yield determination. 3 .3.2 Lignin Extraction from Birch acted from debarked silver birch ( Betula pendula Roth.) grown in northern Sweden and supplied Curt 64 Lindström (Smurfit - Kappa Kraftliner AB, Piteå, Sweden). The lignin extraction was performed according to the method proposed by Gu et al. 244 In short, the biomass was cryogenically ball - milled in a TissueLyser II (Qiagen, Hilden, Germany) for a total of 4 hr with 15 min interval for cooling in liquid nitrogen. The ball - milled sample was dissolved in 8% LiCl/DMSO at a concentration of 5% by wei ght and stirred at 25 °C for 48 hr then at 50 °C for 24 hr. The biomass was precipitated as a gel by dropwise addition of the LiCl/DMSO solution into deionized water. This gel was thoroughly washed with deionized water until no Cl remained in the gel as de termined by addition of AgNO 3 . The decrystallized biomass sample was then lyophilized before undergoing enzymatic hydrolysis for 72 hrs (20 mg protein / g biomass; CTec II and HTec II from Novozymes in a 2:1 ratio at 50°C with in pH 5.25, using 0.05 M Na - c itrate buffer). Next, the sample was washed with excess water and lyophilized again. The solid residue was then subjected to dioxane extraction for 6 hr with 50 mL 96% aqueous (v/v) dioxane. 3. 3.3 Lignin Acetylation Acetylated lignin samples were used for 1 H NMR and GPC analyses. The procedure was adopted from previous study with slight modification. 245 Essentially, 12.5 g/L of lignin solution in 1:1 pyridine/acetic anhydride was prepared. Fractions that were harder to solubilize were subjected to sonication for 10 - 20 min. The completely dissolved solutions were left in an incubator at room t emperature and 180 rpm for 24 hr before applying washing steps as in the previous study. 3 .3.4 NMR Analysis of Lignin 1 H - NMR spectra were acquired on a 500 MHz NMR spectrometer (Agilent 500/54) equipped with a OneNMR probe. Acetylated lignin was dissolved in CHCl 3 - d 6 at a concentration of 44 mg/mL. Pentafluorobenzaldehyde (PFB) was used as an internal standard with tetramethylsilane (TMS) as a peak reference. The spectra were acquired from - 2 to 14 ppm with a 65 30° pulse, a recycle delay of 4 s, and a total of 800 scans. Aliphatic and aromatic hydroxyl content were determined on a per aromatic basis (mol/mol) by using the methoxyl content corrected for the thioacidolysis - determined S/G ratio. 13 C - NMR spectra were acquired on a 500 MHz NMR spectrometer (Varian Inova) equipped with a double - resonance broadband probe. Proton decoupling was applied only during acquisition period ( i.e ., decoupling - NOE). A sample of lignin (300 mg) was dissolved in 0.6 mL DMSO - d 6 along with 2.0 mg chromium(III) acetylacetonate as a relaxation reagent. Sonication was used to facilitate dissolution. The spectra were acquired from - 15 to 235 ppm with a 90° pulse, a recycle delay of 1.7 s, and an acquisition time of 1.2 s. A total of 10,000 scans were collected. Peak assignments were based on the previous literature. 246,247 The - O - 4 content was determined on a per 100 aromatic monomer basis (mol/mol) by using the methoxyl content corrected for the thioacidolysis - determined S/G ratio. The gradient 2D 13 C - 1 H - - sensitive ge - 2D multiplicity - edited HSQC using PEP and adiabatic inversion and refocusing pulses with gradients in back - inept) were performed on a 900 MHz NMR spectrometer (Bruker Avance) equipped with a TCI triple - resonance inverse detection cryoprobe. A sample of lignin (100 mg) was dissolved in 0.6 mL DMSO - d 6 C H 2.49 ppm) is used as an internal reference. The spectra were acquired from 0 to 9 ppm in F 2 ( 1 H) dimension by using 3234 data points for an acquisition time (AQ) of 200 ms, and from - 15 to 155 ppm in F 1 ( 13 C) dimension by using 512 increments (AQ of 6.66 ms). The number of scans per t1 increment was 32, and the recycle dela y was 1 s. The total experimental time was 337 min. Peak assignments were based on the prior literature. 217,248 251 25,87 90 66 3 .3.5 Gel Permeation Chromatography Gel permeation chromatography (GPC) was performed as in our previous work. 245 Briefly, acetylated ligni n (2 mg/mL) was dissolved in THF. The samples were analyzed by HPLC (Agilent 1100) equipped with a Waters Styragel HR 4 column (Milford, MA, USA) at a flow rate of 0.5 mL/min, temperature of 40 °C, and detection by UV absorbance at 280 nm. Polystyrene stan dards of molecular weight 1, 10, 50, 100, 200 kDa were used for calibration. 3 .3.6 Thioacidolysis Depolymerization of Lignin Quantitative thioacidolysis utilizing derivatized monolignol standards was performed as described in our previous work. 234 For this, three 2 - mg replicates of dried and isolated lignin sample were weighed into glass vials and heated with a mixture of dioxane, ethanethiol, and boron trif luoride diethyl etherate to liberate the lignin monomers. The extracted thioether derivatized monomers were subsequently silylated with N,O - bis - trimethylsilyl - acetamide ( BSA ) and quantitated using GC - MS analysis (Agilent 7890A / 5975C MS) . 3. 3.7 Catalytic Oxidati ve Cleavage of Lignin The catalytic oxidation method was adapted from the literature. 209 For this, 10 mg of each lignin sample and 5 % (w/w) CuSO 4 were dissolved in 5 mL of 2 M NaOH. Samples were sonicated for 10 min to ensure complete dissolution. The solution was then poured into a 50 - mL passivated stainless steel reactor and pressurized to 10 bars with O 2 gas (>99.99% purity, Airgas) following evacuation of the reactor headspace by vacuum. The reactors were submerged in a fluidized sand bath (Techne, Cole - Palmer) preheated to 160 °C for a specific period of time. When the reaction was complete, the reactor was submerged in an ice bath for 10 min. The pH of the final solution was reduced to 2.0 by addition of 6 M HCl and centrifuged to isolate the solid residue. The products in the liquid fraction were extracted four times by chloroform. The solution 67 was then evaporated under vacuum at 40 °C and re - dissolved in 10 mL of 1:1 methanol/water with 1% (v/v) formic acid. Quantification of the products was performed by HPLC (Agilent 1100) using 1:4 methanol/water with 1% formic acid as a mobile phase and a C18 column (Atlantis T3 100 Å, 3 µ m, 3 .0 mm x 150 mm; Waters). Standards for product quantification included vanillin, vanillic acid, acetovanillone, syringaldehyde, syringic acid, acetosyringone, 4 - hydroxybenzaldehyde, and 4 - hydroxybenzoic acid. All standards were purchased from Sigma Aldrich and monomer yields were determined on a per mass Klason lignin basis. 3 .3.8 Catalytic Hydrogenolysis of Lignin Catalytic hydrogenolysis of lignin was performed according to previous work. 230 In short, 1 0% metal loading Ni/C was synthesized by incipient wetness. Typically, 2.76 g of Ni(NO 3 ) 2 6H 2 O (Sigma - Aldrich) was added to 8 g deionized water and stirred until well mixed. The solution was then added dropwise to 5 g of activated carbon (Darco ® G - 60, Aldr ich, water absorption capacity of 2.1 mL/g). The mixture was kept wet at room temperature for 24 h before drying overnight at 100 °C. Then it was calcined at 400 °C under nitrogen atmosphere for 4 hr, and reduced at 400 °C under 10% H 2 in an Ar atmosphere, both of which used at a temperature ramp of 6.25 °C/min. The finished catalysts were stored and weighed in an anaerobic glove box. For the hydrogenolysis reaction, 40 mg of lignin, 30 mg of catalyst, and 4 mL of ethanol were charged to a 10 mL passivated stainless steel reactor while in the anaerobic glove box. Reactors were submerged into a fluidized sand bath set at 220 °C for 6 hr. The resulting liquid was collected, filtered, and injected into GC/MS (Agilent 7890A / 5975C) using Agilent DB - WAXETR colum n 60 °C for 3 min before ramping up at 40 °C/min to 260 °C and held for 2.5 min. Standards were run for quantification. Yields were determined as mass products per mass Klason lignin using 68 standard curves constructed from pure monomers including 4 - propyl guaiacol, 4 - propyl syringol, 4 - ethyl guaiacol, 4 - ethyl syringol, and syringol. All standards were purchased from Sigma Aldrich except 4 - propyl syringol, which was sy nthesized according to Parsell et al., 37 and 4 - ethyl syringol, which was estimated using the same response factor as 4 - propyl syringol. These products comprise all peaks that were detected. 3 .3.9 Yield Prediction Model Development Computational prediction of monomer yields was performed in MATLAB (MathWorks, Inc., Natick, MA). For this approach, populations of lignin polymers of length n with n - 1 bonds between the n monomers were generated over a range of expec te - O - 4 contents (See Section 3.4.5 ). The assignment of the identity of each of the n - - aryl - O - 4 content for a given population function in MATLAB. This was done to generate - O - 4 bonds at - O - 4 bonds. This methodology was applied to determine monomer yields from populations of 5000 polymers of desired chain length n over a range of 0 to 100% over 2500 - O - 4 bond, where a monomer can be generated - O - - O - 4 bond at a chain end. This methodology assumes that lignin is a linear polymer and that no cross - linking or aryl - aryl bonds are present in the polymer. For comparing calculated theoretical yields to quantified yields, the experimental yields by thioacidolysis, h ydrogenolysis, and oxidation were converted to a mol/mol yield by assuming a molar mass of 196 g/mol for a guaiacyl monomer and 226 g/mol syringyl monomer within the lignin polymer. 69 3 .4. Lignin Fractionation Process and Recovery The lignins used in the f irst part of this study were obtained from the fractional precipitation of pretreatment liquors following the mild alkali pretreatment of hybrid poplar at a final temperature of 150 °C at a total time of 3 hr as performed in our previous work in order to m inimize lignin modification and to maximize xylan retention in the pretreated biomass. 242 Both native and process - modified lignin are highly heterogeneous which hinders accurate characterization and may limit their value for applications requiring a homogeneous, well - defined starting material. For this work, alkali - solubilized lignins are subjected to fractionation in order to both obtain more homogeneous lignin fractions and to enrich or deplete these fractions in certain properties that can yield insight into property differences that impact the conversion process chemistry. A number of approaches for the recovery and fractionation of lignin from alkaline pulping or pretreatment liquors have been explored in the past. These approaches include ultrafiltration or approaches based on precipitation due to altering the properties of the solven t by acidification with a weak or strong acid and the use of organic solvents. 252 255 Of these approaches, CO 2 acidification has been employed in a number of recent commercial or pilot studies, including LignoBoost TM , LignoForce TM , and of particular interest to this study, Sequential Liquid Recovery and Purification TM (SLRP TM ) process. 256 258 This pH - based fractionation by CO 2 has previously shown the ability to generate lignin fractions with different properties from Kraft black liquor of softwood biomass. 245 For the present work, lignin fractionation was performed using two approaches to sequentially acidify and fractionally recover lign ins from this alkaline pretreatment liquor. The first approach utilized sequential acidification and precipitation with CO 2 at room temperature and atmospheric pressure, while the other ( i.e., the SLRP TM process) was performed at elevated temperature and p ressure as in our previous work. 245 The mass yields for the lignin fractions 70 obtained from these processes are presented in Table 3. 1. This shows that for recovery at the same pH range, the room temperature recoveries were more than double from the SLRP TM process. This may be due to the higher solubility of the lignin at higher tem perature. These fractions, hereafter, will be referred to as named in Table 3. 1 throughout the study. Furthermore, it should be understood that these fractions do not represent 100% of the recoverable lignin, but were generated in order to better understan d how diversity in lignin properties impact depolymerization aromatic monomer yields. Table 3 . 1 . Percent recovery of each fraction through CO 2 acidification at room temperature and at 115 °C (SLRP TM ) . The pH 2 fraction was obta ined by acidification with H 2 SO 4 . Room Temperature (RT) Fractionation Precipitation pH Fraction name Percent Recovery 11 RT - 1 8.42 10.6 RT - 2 20.79 10 RT - 3 1 3.95 9 RT - 4 12.59 2 RT - 5 26.07 Sequential Liquid - Lignin Recovery and Purification (SLRP) 10.5 SLRP - 1 4.05 10 SLRP - 2 7.95 3 .5 Trends in Properties of Fractionated Lignins Next, we characterized the properties of the different lignin fractions by applying a number of characterization techniques. These techniques included determination of S/G ratios as measured by both 2D 1 H - 13 C - correlated HSQC NMR and quantitative thioacidolysis, the fraction of linkages O - 4 contents by quantitative 13 C - NMR, the phenolic and aliphatic hydroxyl content as determined by 1 H - NMR, and the molecular weight using gel 71 permeation chromatography (GPC). A ll results are available in the Appendix B . It was found that all of these properties, with the exception of the aliphatic hydroxyl content, exhibited obvious trends across the various fractions obtained by pH - based lignin fractionation. The HSQC spectra o f fraction SLRP - 1 is shown in Figure 3. 1 as a representative spectrum. From this, it can be observed that in the aromatic region a minor fraction (3%) of the syringyl units n the peak corresponding to p - hydroxyphenyl, which implies that there was no or negligible demethoxylation occurring during the alkali pretreatment and fractionation process. The S/G ratios from the HSQC spectra were determined as the relative integration areas of the S 2,6 2,6 peaks to the G 2 peak. The results ( Figure B 1 in appe n dix ) indicate a distinct trend that lignin precipitated at higher pH, particularly from the SLRP process, showed higher values of S/G ratios with the highest (S/G = 1.99) from fraction SLRP - 1 and lowest (S/G = 1.80) at pH 2 (RT - 5). It should be noted that the SLRP fractions exhibit higher S/G ratios relative to room - temperature fractions. This trend is in a good agreement with S/G ratios obtained from thioacidolysis, which are all slightly lower than those from HSQC, ranging from 1.95 in SLRP - 1 fraction down t o 1.73 in sample RT - 5. Three types of linkages in the aliphatic region from HSQC spectra were detected, including O - Figure 3. 1). From both regions of the spectra, no other peaks corresponding to possible linkages from alkali pretreatment , such as stilbene and diarylmethane, were detected. It should be noted that there could be some linkages as a result of condensation that would not be detectable through HSQC NMR such as 5 - HSQC NMR wa s semi - quantitative, i.e. percentage of detected linkages. The percentage trend of linkages is obvious and is shown in Figure B 3, which indicates that lignin at higher pH fractions, O - 4 an 5 linkages, while 72 that adding standards for absolute quantification was not performed due to several prohibitive factors that originate from the lignin; pr imarily the error caused by varying T 2 relaxation and resonance offsets in HSQC NMR which affect the signal intensity. 259 Instead, quantitative 13 C - O - 4 linkages per 100 aryl units. A similar trend was observed, in which the fractions O - 4, and the increase was particularly pronounced in SLRP fractions ( Figure B 1). Notably the fraction with O - 4 content, SLRP - O - 4 content relative to the lowest . Figure 3 . 1 . 2D 1 H - 13 C HSQC NMR spectra showing lignin fraction SLRP - 1 composition and linkages. The contours in aliphatic region (left) are used to quantify the fraction of inter - unit linkages, while those in the aromatic region (right) are used for S/G ratios. Next, we investigated the amount of phenolic and aliphatic hydroxyl groups by application of 1 H - NMR of the acetylated lignin fractions. The results indicated that the amounts of aliphatic hydroxyl groups rem ain nearly constant across all fractions with the exception of the pH 2 fraction ( Figure B 2). The amount of phenolic hydroxyl groups , on the other hand, demonstrated a clear trend in which higher pH fractions, especially the SLRP fractions, had a lower con tent of phenolic 73 hydroxyl groups . We have previously identified this same trend in SLRP - fractionated lignins derived from a softwood Kraft black liquor. 242 Lastly, the results from GPC for molecular weight also demonstrated a trend across the fractions with the high - pH fractions exhibiting the highest molecular weights ( Figure B 3). In summary, we i dentified trends in almost all properties of lignin except aliphatic hydroxyl O - 4 content, S/G ratios, and molecular weight were observed in higher pH fractions with the effect more pronounced in SLRP fractions. The phenolic hydr oxyl group showed an opposite trend where its amount is lower in higher pH fraction and increases as pH fraction decreases. Apart from S/G ratios, these trends from the hardwood used in this study are similar to those reported earlier from softwood by Stok losa et al. 245 3 .6 Trends in Monomer Yields from Depolymerization of Fractionated Lignins As it was shown that there are trends in structural features corresponding to precipitated pH, this section aims to investigate how the difference in these features affects their susceptibility to depolymerization. This was examined by quantifying yields of monomers generated from lignin following depolymerization utilizing three different conversion chemistries. These included catalytic oxidation utilizing molecular oxygen and a homogeneous copper catalyst under alkaline conditions and elevated tem perature. This approach is comparable to what has been employed commercially in the past to produce vanillin from lignosulfonates. 260 The second approach employed hydrogenolysis using ethanol as both a solvent and a hydrogen source catalyze d by Ni on an inert support at elevated temperature and pressure. The third approach employed quantitative thioacidolysis. Although each approach shows different monomer - product distributions, we chose to focus on the total yields ( Figure 3. 2). Similar to the lignin properties across the different fractions, clear trends were observed for monomer yields following depolymerization. We must 74 stress that the yields, by necessity, must be based on the total content of lignin. There is not a clear consensus on ho w aromatic monomer yields from lignin should be defined, especially lignin that has been chemically modified where its composition may be altered beyond definition of lignin. 261 Therefore in order to be consistent, the present study has reported all yields from these reac tions on a basis of Klason lignin, which does not consider acid - soluble lignin and may contain contaminants such as proteins. 262 Figure 3 . 2 . Trends in m onomer yields from different depolymerization techniques, including (A) Cu - catalyzed oxidation, (B) quantitative thioacidolysis, and (C) Ni/C - catalyzed hydrogenolysis . Prior to performing oxidation on the different lignin fractions, a kinetic study of monomer yields and selectivities was performed on SLRP - 2 ( Figure B 4) in order to determine optimal reaction conditions since oxidation products have been reported to repolymerize under these conditions. 263 Although optimal yields may differ from fraction to fraction, the optimal yields for sample SLRP - 2 were obtained at 7 min which was chosen as the reaction time used for all other pH fractions. The results show that mo nomer yields from oxidation generally increased as the recovery pH increased ( Figure 3. 2A). It should be noted that previous work using alkali poplar lignins for Cu - catalyzed oxidation were able to achieve yields of around 10%, which is roughly an average of all the RT samples combined. 213 The differences in yields were not apparent, especially in fractions recovered at higher pH. There was also a discrepancy in sample RT - 1 and RT - 2, which 75 fell out of trend. The ambiguity in trends is likely due to the intrinsic mechanism of oxidat ive cleavage in alkali solution, wherein depolymerization and condensation reactions share similar intermediates. 264 Moreover, the desired product itself, vanillin, was previously sh own to be able to condense to form dimers. 263 The concurrence of these reactions makes the trend of lignin - monomer yi elds through oxidation not as obvious as we would expect compared to the trend from quantified properties. We next employed quantitative thioacidolysis to assess the yields across the lignin fractions. Thioacidolysis cleaves exclusively ether linkages wi thin lignin and yields stable products. The results from quantitative thioacidolysis show a clear trend whereby fractions recovered at higher pH demonstrated higher yields ( Figure 3. 2B). Similar to the trends observed for lignin properties, the results fro m the SLRP fractions are particularly notable when compared to the fractions recovered at ambient temperatures and pressures. Additionally, lower yields were observed for these fractions than were obtained from oxidation, which may be attributed to the les s selective nature of oxidation reaction, which was shown to also cleave certain C - C linkages, in addition to ether linkages in presence of Cu 2+ or other catalysts to yield aldehydes and ketones. 214,215,265 The results were not entirely unexpected as both 13 C - NMR and HSQC NMR show that higher pH - O - 4 linkages, which is the target for thioacidolysis. However, the - O - 4 content affect lignin depolymerization. From Figure 3. 2B a five - fold difference in yields can be observed between the two extreme fractions that correspond to a two - - O - - O - 4 content can be related to these depolymerization yields will be investigated in subsequ ent sections. Lastly, the lignin fractions were screened for monomer yield by hydrogenolysis catalyzed by carbon - supported Ni. This approach has been used in many recent studies 37,226,230,266 as it has 76 been demonstrated to produce high monomer yields (>40%) with excel lent selectivity. 267 It was also shown by HSQC NMR to selectively cleave only ether linkages at a reaction temperature of 200 °C in ethanol. 227 When subjected to hydr ogenolysis the fractionated alkali lignins, however, showed that the highest yield achieved was 7% ( Figure 3. 2C). Unlike the approach using oxidation, the monomer products are stable under the conditions used for hydrolysis and a clear trend can be observe d within the fractions from alkali - pretreated lignin. The samples from higher pH fractions, particularly SLRP, produced more monomers. We have shown that lignin properties have a strong impact on its susceptibility toward cleavage of ether linkages. Parti cularly, the trends are clear in thioacidolysis and hydrogenolysis as it was not complicated by other side reactions that may have occurred during oxidation. Nonetheless, general trends from any of the depolymerization techniques point toward higher monome r generations in the fractions recovered at higher pH, notably in the SLRP fractions, with significant gaps as high as four times the yield from thioacidolysis between the highest pH SLRP fractions and the lowest pH atmospheric fraction. Therefore, lignin properties can greatly affect the outcome of depolymerization reactions. 3 .7 Correlating Lignin Properties and Monomer Yields for Fractionated Alkali Lignins Several obvious trends were identified among the fractionated alkali lignins, namely that properti es of lignin co - vary and that these properties can be linked to the potential for generation of aromatic monomers from lignin. In this section, these results are correlated in order to gain deeper insight into how these trends are connected. The trends amo ng almost all properties from the previous section imply that the properties of lignin, even after the alkali pretreatment, were still fundamentally related. Correlations among some of the quantified properties, as summarized oefficient (R), are presented in Figure 3. 3. This shows that a number 77 of strong correlations can be observed between many properties. Importantly, S/G ratios, - O - 4 content, were all found to have R values of - O - 4 content, which - O - 4 - O - 4 linkages as the extra methoxyl group in syringyl monomers prevents formation of C - C linkages at the 5 - C position during lignin biosynthesis. 235 In our pretr eated lignin fractions, this same correlation still exists. Figure 3 . 3 . Correlation map for fractionated alkali lignin properties and monomer yields demonstrating strong correlations between most lignin properties and its susceptibility to depolymerization via - O - 4 cleavage . 78 - O - 4 content and that during pretreatment these linkages are broke - O - 4 content signify more - O - 4 linkages then result in the formation of new phenolic hydroxyl groups giving negative correlations between phenolic hydroxy - O - 4 linkages, and molecular weight ( Figure 3. 4A). The difference in the amount of phenolic hydroxyl groups is likely the factor that allows fractionation based on pH in this study to be possible because in alkaline solutions one of the factors that promotes lignin solubility is the deprotonation of phenolic hydroxyl groups. 268 Fig ure 3 . 4 . Selected linear correlation between fractionated lignin properties and monomer yields, - O - 4 content is directly correlated to molecular weight and inversely correlated to phenolic hydroxyl content, that (B) monomer yields from thioacidolysis (g/g) were - O - 4 contents, and that (C) as well as molecular weight and phenolic hydroxyl contents. Similar to the trends observed among the lignin properties, trends can be observed across these lignin fractions subjected to depolymerization with different techniques. Monomer yields from thioacidolysis and hydrogenolysis exhibited strong correlations to a number of lignin properties, while oxidation yields were much less strongly correlated ( Figure 3. 3). Specifically, thioacidolysis yields showed a linear correlation for many of the lignin properties ( Figure 3. 4B & C). However, it should be noted that this linearity could be the result of the small r - O - 4 - O - 4 content 79 as will be shown in later sections. Nonetheless, the results indicate that the diverse structural properties of the lignin are fundamentally co rrelated and affect the depolymerization reaction accordingly. Consequently, a single property can be used as a descriptor of lignin susceptibility to depolymerization. 3 .8 O 4 Content We have established that lignin properties, even after the alkaline pretreatment performed in this work, are still fundamentally inter - dependent and they also correlate well with the yield of monomers generated by three different catalytic depolymerization chemistries. This section aims to further analyze the effect of lignin properties beyond just correlation. We aim to validate a generalizable prediction model based on a quantifiable lignin property. This would yield a tool to assess the potential of lign in to yield aromatic monomers. Since most of these properties are related, this implies that a single property can be selected to represent the other properties, and serve as a descriptor for the depolymerization of lignin utilizing approaches that target inter - unit - O - 4 content as a property to predict lignin susceptibility to depolymerization, i.e. maximum possible yields, because it is relatively easy to establish a rationale for the predicting formula. O 4 linkages, a unit in a triad as shown in Figure 3. O 4 linkages or be at the chain end connected O 4 linkage. For example, if m 2 in Figure 3. 5 were to become a monomer, both d 1 and d 2 O 4 linkages because otherwise a dimer or oligomer would be generated instead if one of the connecting linkages was a C - C bond. Likewise, for a monomer to be generated from m 1 , d 1 must be an ether linkag e. As a lignin polymer is assumed to be synthesized by the free radical mediated oxidative polymerization mechanism, the distribution of linkages throughout the 80 polymer can be assumed to be random. Therefore, the quantified content of a lignin property, su ch - O - 4 content, can be considered to represent the probability that linkage within the polymer is - O - 4. With this assumption, the prior rationale can be applied to yield an expression predicting the probability that any monomer within the lignin po lymer can be releas - O - 4 bond as: P ( A B ) = P( A ) P( B ) Eqn. ( 3 . 1 ) w here P ( A ) is the probability of a - O - 4 linkage existing on one end of a triad and P ( B ) is the - O - 4 linkage occurring at the other end of the t riad. Likewise, the probability that a monomer will be released from a polymer chain end can be considered as P ( C ), represents the probability of a - O - 4 linkage occurring as the terminal linkage at the end of a polymer chain. This methodology is based on the simplification that lig nin is a linear polymer without branches or crosslinks between polymers. Recent work has provided evidence that at least some native lignins may be prim arily linear oligomers. 269,270 Based on this rationale and assuming that any of the monomer - yielding events are mutually exclusive, the monomer yield can be estimated based on the expected values - O - 4 content, which is: Eqn. ( 3 . 2 ) Figure 3 . 5 . Lignin chain model of chain length n used for formulating prediction rationale. The i th monomer unit, n th linkages of lignin are represented by m i and d n . A triad is composed of three monomer units. 81 - O - 4 content is randomly distributed throughout the polymer chain (i.e., the pro - O - 4 bond are equal) , this expression simplifies to: Eqn. ( 3 . 3 ) w here n is the number of monomers within the chain. The general form of this same equation has been previously derived for predicting distributions of polymer depolymerization products by Montroll and Simha in 1940 . 271 For the case where the lignin chain is very long (i.e., n 100), E qn . 3 .3 can also be truncated to: Eqn. ( 3 . 4 ) - O - 4 content to monomer yields was first proposed in the work of Yan et al. 267 Based on Eqn. 3 . 3 O 4 content for a lignin O 4 content exhibiting a Gaussian distribution were generated ( Figure 3. 6). It can be observed that the trends do not vary significantly within this window of chain length indicating that the difference in number of end units is still relatively insignificant. 272 In Figure 3. 6, the red circles represent thioacidolysis yields from different alkali - pretreated fractions. These yields followed the predicted trend very well, where twice the amount O 4 co ntent in the fraction SLRP - 1 produced almost three times the monomer amount compared to RT - O 4 content within these fractions are still limited. Therefore, a lignin from dioxane extraction was generated with a 38% extraction yiel d. Application of 13 C - O 4 content of 65% and 38% monomer yield ( Figure 3. 6, green square). The results conformed to the predicted trend as well as 82 asserting the importance of obtaining a lignin - O - 4 content if the goal is to generate high monomer yields. O 4 content is reduced during alkali pretreatment, this should not cause its thioacidolysis yields to be inferior to those from extracted lignin as the loss i n ether linkages would be compensated by an increase in end - chain units. However, the fact that the monomer yields from alkali lignin fractions are significantly lower could be derived from two factors. Namely, the potential formation of C - C bonds by conde nsation reactions during the O 4 units cleaved during pretreatment yield monomers and short chain oligomers that were not recovered during lignin recovery by pH fractionation. Figure 3 . 6 . - O - 4 content for lignin polymer populations of chain length of 100 monomers (left) and 25 monomers (right) with results from quantitative thioacidolysis of fractionated poplar lignin a nd native birch biomass included for validation. A closer look to the prediction results further emphasizes the significance of the quality of O 4 contents in the pretreated biomass, regardless of reaction optimization or catalyst selection, high monomer yields are not achievable from these lignins. The correlation in Eqn. 3. 4 not only explains the low yields from the alkali lignin fractions in thi s study, it also provides justification for performing lignin depolymerization prior to significant lignin 83 modification. For example, certain studies employed catalytic reductive cleavage of lignin using whole biomass as a substrate with the reaction condi tions and solvents chosen to limit undesirable condensation reactions. 273,274 With high quality lignin substrate and limited condensation, yields have been demonstrated in the range 2 0 - 50% from a variety of catalysts and biomass sources for - O - 4 contents that were expected to range from 50 70% , 226,266,275,276 while a high - syringyl content O 4 bonds could achieve monomer yields as high as 78% monomers. 277 Likewise, high monomer yields were also observed in oxidative cleavage approaches that used i solated lignin and mild reaction conditions. 30 Therefore, from the predicted results and mentioned examples, it was clear that lignin quality was crucial for achieving high monomer yields. This can be c ontrasted to the hydrogenolysis yields obtained from alkali hardwood lignin in this study, which used similar reaction conditions ( Figure 3. 2). Figure 3 . 7 . Model predictions of monomer yields from lignin polymer populations of different chain lengths versus monomer yields from thioacidolysis, oxidation, and hydrogenolysis of fractionated lignin . Monomer yields of pretreatment - derived lignins from oxidation an d hydrogenolysis are also plotted against the prediction model ( Figure 3. 7) and show that unlike the yields from thioacidolysis that show good agreement with the model prediction, the yields from the oxidation 84 reaction do not. As the model prediction is ba O 4 bonds, this discrepancy can be hypothesized to be due to the less selective nature of the oxidation reaction as well as potentially condensation of the monomer products. The yields for hydrogenolysis, on the other hand, seem to follow the trend. This should be expected as previous work has shown that only peaks corresponding to ether regions on HSQC NMR disappear after the hydrogenolysis 5 dimer did not produce any monomers. 227 However, it can be seen in Figure 3. 7 that although the trend follows the prediction line, the values are lower than what is p redicted for thioacidolysis. This could be due to a number of factors such as incomplete conversion for the reaction time used, or incomplete solubilization of the pretreatment - derived lignins in the solvent. Either of these assertions is supported by the finding that a considerable amount of lignin residue remained after the reaction; however, accurate measurement of the leftover residue was prohibited due to limited amount of available samples. Figure 3 . 8 . Predicted yield distributions for monomers, dimers, and oligomers from the cleavage of - O - 4 in a population model of lignin polymers of n - O - 4 contents. Besides monomer yields, this approach can be extended to predict the yield of o ligomeric products by employing a similar rationale to that used for monomer generation. As an example, to 85 yield a dimer, three monomers must be linked by two non - O 4 linkages. These results, as shown in Figure 3. 8 presented as cumulative yields O 4 content of 50%, O 4 linkages would be predicted to yield approximately 10% tetramers, 17% trimers, 25% dimers, 27% monomers, and 21% all other oligomeric products. From this plot it can O 4 content increases above ca. 55%, the yield of monomers dominates over other oligomeric products ( Figure 3. 8 ). Again, this emphasizes the importance of enhancing and O 4 content of lignin, preferably above 55% or more, if high monomer yields ar e the goal. In fact, recent studies tried to achieve this through so called lignin - first biomass pretreatment and isolation processes. 33,278,279 3 .9 Conclusion The present work has demonstrated through characterization of hardwood lignins generated by pH - based fractionation of alkaline pretreatment liquors that many lignin properties exhibited strong correlations with each other. Specifically, lignin fractions wi th higher S/G ratios O 4 contents and molecular weights with lower phenolic hydroxyl O 4 content, were shown to greatly impact the a chievable monomer yields through three different lignin depolymerization approaches. Furthermore, a model was developed to explain the O 4 content and monomer yield. The prediction model was validated experimentally by 13 C - NMR O 4 content and quantitative thioacidolysis yields from O 4 content squared, it highlights the importance of O 4 contents in the lignins during pretreatment or isolation. 86 CHAPTER 4 Adsorption of the Yield Determining Lignin Dimer on Catalyst Surfaces 4.1 Summary Lignin hydrogenolysis has recently been studied extensively as it was shown to result i n high monomer yields. Most of these reactions were conducted in the liquid so lvent, which has been shown to have large impacts on product types and yields. Since adsorption is the first step to any heterogeneous catalyst reactions, this work aims to under stand how solvent affects lignin adsorption on Ni(111) and Cu(111) surfaces. To achieve so, density functional theory (DFT) - O - 4 lignin dimer (a model compound) adsorption conformations in both vacuum and l iquid ethanol. In the vacuum , it was found that lignin prefers to adsorb strongly on Ni(111) and weakly on Cu(111) with both aromatic rings parallel to the surface. Solvated adsorption was modeled using both implicit and explicit models. It was found that an exp licit model is required to accurately describe the lignin - solvent interactions. With the explicit solvation model, it was found that lignin dimer adsorbs on Ni(111) surface but not on Cu(111). Furthermore, to circumvent the expensive liquid interface calcu lations, a thermodynamic cycle method was developed to quickly estimate the solvated lignin dimer adsorption energy from the adsorption energy in a vacuum and the so lv ation energies. This model quantifies the effects from solvent on lignin dimer adsorption , including the contributions from the lignin - solvent and the solvent - metals interactions, and suggests how to design catalyst and solvent to tune lignin adsorption . 87 4.2 Introduction Lignin is an aromatic polymer that can comprise a significant fraction of the mass (up to 30%) of lignocellulosic biomass, 13 b ut is currently underutilized. Its valorization is important for the economic feasibility of lignocellulosic biorefineries. Lignin is composed of primarily of 2 types of aromatic moieties syringyl (S) and guaiacyl (G), that differ in the number of methoxyl substitutions on the aromatic ring. These monomers are linked by various types of C - O and C - C bonds 280 with the - aryl ether linkage ( i.e ., the - O - 4 linkage) comprising the most abundant type of linkage at 50 - 80% of the inter - unit linkages in native lignins. 279 Lignin depolymerization to aromatic monomers and low molar mass oligomers is one route to generate renewable aromatic - O - 4 linkage offer the potential for achieving high yields and selectivities for a limited set of aromatic monomers. 5,19,281 Lignin hydrogenolysis and partial hydrodeoxygenation is one of these promising approaches to lignin depolymerization and has been reported using diverse catalysts s uch as NiAu, 223 Ni/ C , 226 or Pd/C 225 in solvents that include short chain alcohols, dioxane, or alkali water. 225,233,282,283 As will be discussed in more detail, the selection of solvent was shown to have significant impacts o n products yields and selectivities. However, how these solvents affect the lignin - metal interactions at the catalyst surface is still not well - understood. In fact, theoretical studies on how lignin model compounds ( e.g - O - 4 lignin dimer) are adsorbed onto catalyst surfaces, even in the vacuum state, is scarce and still inconclusive. The few existing studies only briefly explore the potential 31,44 More importantly, it has been reported that different isomers of the model yielded different adsorption conformation s on the Ni surface, which subseque ntly lead to different reaction pathways . 44 This stresses the significance of adsorption conformation as it is the first step in any 88 reaction. Therefore, this study aims to clarify how lignin adsorbs on metal surfaces and to characterize the impact of solvent on this adsorption. Experimentally, the choice of solvent, among others, has been shown to significantly impact hydrogenolysis product selectivities and yields. One study assessed various solvents using Raney Ni as a catalyst and demonstrated that in a pre sence of H 2 gas, aprotic non - polar solvents, such as decalin and n - heptane, gave the highest yields at low temperature (90 °C); h owever, at high temperature ( 300 °C ) , alcohol solvents demonstrated comparable performance while also exhibiting high selectivi ty. 40 Furthermore, studies also point ou t that ethanol could give higher yields than methanol as it can act as a capping agent by alkylation and este phenolic hydroxyl group, essentially limiting repolymerization . 28,27 4 Given this knowledge, it is not uncommon to encounter studies in literature where similar reaction conditions and catalysts were used , except for solvent choice, but different products and selectivities between products were obtained. On Ni/C, studies using dioxane as a solvent resulted in phenolic monomers with hydroxylated propyl side chains ( i.e ., dihydrosinapyl and dihydroconiferyl alcohol ), 32,284 whereas others using methanol as a solvent reported higher selectivities for products with deoxygenated side chains ( i.e. , 4 - propylsyringol and 4 - propylguaiacol ). 226,285 These are in agreement to a recent study suggesting that methanol can limit the product defunctionalization when compared to other alcohol solvents. 34 The effect was claimed to be due to the stronger che misorption of methanol on the Ni surface. For other catalysts, such as Pd/C, a similar case can be found when comparing pure methanol versus methanol mixed with water as the solvent. 35,36 All of these examples strongly suggest the important role played by the solvent on the reaction pathway followed. However, the understanding of solvent impacts at surfaces, which could improve the process of catalyst an d solvent selection, is still very limited. To address this shortcoming, density functional theory 89 (DFT) calculation is selected as the tool because, in addition to providing insights into adsorption energetics, it can yield insights into molecular geometr - O - 4 bond within lignin is selected as the key feature of the model system as it is the prime target for hydrogenolysis and previous studies have shown it to be the primary yield - determining property in lignin. 286 Computational methods have been used in the past to evaluate the impact of solvents on lignin solubility and conformation. In an ionic liquid, charge analysis also s howed that dissolution of lignin dimer is facilitated primarily by H - - stacking interactions. 287,288 For H - - OH position of the dimer, while cations had the strongest in - OH position, whereas the cation contribute d - stacking interactions. The strength of the inter action could lead to changes in the structure of lignin. This was studied in detail for an aqueous solution using classical molecular dynamics (MD), where polymeric lignin was shown to form a compact globular structure minimizing the exposure of the surfac e to the solvent due to its natural hydrophobic interactions . 282,289 H owever, it was found that the structure exhibited more open conformations with increased temperature. 290 Interestingly, using DFT calculations combined with an implicit solvation model, such as the conductor - like screening model (COSMO ) 121 to represent alkali water, it was found that the compact globular structure of lignin could also be opened up by oxida tion of benzylic alcohol to form ketone . 282 The more open, accessible structure of the poly meric lignin was speculated to be the reason for higher monomer yields from oxidized lignin relative to natural lignin. This example demonstrates the importance of understanding the interfacial behavior of lignin. 90 To address the limited understanding of th e adsorption behavior of model lignins on surfaces, particularly when the solvent is involved, the present work will conduct a thorough study of lignin erythro and threo environments. The goals of the study are to identify the most likely stab le adsorption of lignin dimers and to identify the relationship between gas phase and liquid phase adsorption. Ultimately, we hope that the relationship can be leveraged for a quick estimation of adsorption ene rgy in the solvated state. The selected metal s urfaces are Ni (111) and Cu (111) to represent metals with strong and weak adsorption s . The solvent selected for the explicit model is ethanol as it is commonly used in hydrogenolysis reactions. 28,226 Particularly, the solvent will be modeled at a density corresponding to the temperature of 50 °C to accommodate experiments at low temperatures, 222,291,292 and therefore, the interactions at higher temperatures should lie between this and the simulation results for in a vacuum . - O - 4 dimer is important. Previous theoretical studies employed dimers with v - O - 4 linkages within a lignin polymer. Linin dimers composed of all fu nctional groups present in the polymer are known to have a bond dissociation energy in a range of 2.6 to 3.0 eV. 293 However, simplifying the dimer model by eliminating either the phenolic hydroxyl or the aliphatic hydroxymethyl group can affect the bond breaking energy by approximately 10%, while the effect from methoxyl group could be negligible. 294,295 Considering that these functional groups can also affect adsorption geometries, the present study will use guaiacylglycerol - beta - guaiacyl ether as a - O - 4 dimer as it contains all functional groups an a ct ual lignin polymer would have as well as allowing us to study the effect of lignin side chain chirality. 91 To achieve the aforementioned goals, the present manuscript will 1) validat e the computational setup and suitability of the chosen lignin dimer as repr esentative of polymeric lignin conformation in the gas phase, 3) demonstrate the effects of solvation on adsorption , and 4) deriv e and validat e the relationship between gas phase and liquid phase adsorption . 4.3 Computational Methods 4.3.1 Adsorption of Lignin Dimer and Monomers from DFT Calculations Calculations of dimer adsorption in this study were performed through the implementation of plane wave DFT in Vienna Ab Initio Simulation Package (VASP). The exchan ge correlations were treated within the generalized gradient approximation (GGA) by Purdew - Burke - Ernzerhorf (PBE) functional. 161 Projector augmented wave (PAW) potentials were used to represent the effective core radii. 163 The electronic optimization was achieved self - consistently with a cutoff energy of 500 eV. The electronic energy convergence was set at 10 - 4 eV due to convergence difficulty of th e surface when included van der Waals, the same criteria w ere used in previous studies with results validated with experimental data. 48 The convergence for forces during structural relaxation is set at 0.03 eVÅ - 1 . A Monkhorst - Pack grid with 14 × 14 × 1 k - point sampling per (1×1) cell was used for the calculation. A spin - polarized calculation was performed with dipole moment correction along the perpendicular direction. Methfessel - Paxton smearing was employed with a sigma of 0.20. The ef fect of van der Waals interaction was accounted for by introducing - D3) 29 6,297 or a non - local correlation functional through optB88 - vdW. 298,299 The computed bulk lattice constants for Cu and Ni as relaxed by PBE functional with optB88 - vdW are 3.635, 3.517 Å, respectively. 92 - O - 4 model dimer selected for the present study is guaiacylglycerol - beta - guaiacyl ether as it has functionali ties and stereochemistry representative of real lignin. The adsorption energy of the dimer in the gas phase ( is calculated with simulation cells in vacuum by: Eqn. ( 4 . 1 ) where , , and correspond to the energy of a dimer adsorbed on a metal slab, the metal slab only , and the dimer in gas phase. The [Metal] structure refers to a slab model with 4 layers of close - packed Cu(111) or Ni(111) at (7 × 7) unit cells with a vacuum of 16 Å on top of the surface. The bottom two layers of the metal slab were constrained in all direction to emulate bulk characteristics. A negative denotes a gain in energy of the system (ex othermic process) . T he adsorption energy in the liquid phase calculated with the implicit model, where the solvent was represented as a homogeneous constant dielectric constant, was implemented with a polarizable continuum model (PCM) developed for VASP ( V ASPs ol ) by Mathew et al . 300,301 The model incorporates the dielectric screening due to permittivity of the solvent and the electrostatic shielding due to mobile ions in the electrolyte. The adsorption energy is calculated as: Eqn. ( 4 . 2 ) where each simulation cell in the calculation was filled with solvent using an appropriate dielectric constant value ( . A The adsorption of a lignin dimer in liquid is also simulated using an explicit solvation model with the adsorption ener gy from the model calculated as: Eqn. ( 4 . 3 ) 93 where is the energy of a lignin dimer in an adsorbed state on a metal surface with solvent molecules surrounding it ( Figure 4. 1B) and is the energy where the dimer is fully solvated in the middle of the solvent slab, away from the metal surface ( Figure 4. 1A). For both Cu and Ni, 37 molecules of ethanol and a dimer were packed into a 4152.91 Å 3 volume ( corresponding to 0.80 g/cm 3 , See Se ction 2.2 for calculation detail s ) between metal slabs. Noted that the height in z - direction of the metal slab surface ends at a distance corresponds to van der Waals radius of Cu (2.09 Å) and Ni (1.97 Å) from the atom center position. 302 This corresponds to solution density at a temperature of 50 °C, and lignin dimer concentration of 0.13 g/cm 3 . The concentration is slightly higher than a typical range of 0.01 to 0.1 g/ cm 3 used in lig nin depolymerization studies, but is not expected to affect the conclusion. 37,226 Figure 4 . 1 . Thermodynamic adsorption energy cycle of lignin dimer (lig) in solvated (A and B) with ethanol (EtOH) and vacuum (C and D) state. It should be noted that, by definition, the from state A to B is intended to encompass both partial loss in lignin solvation energy and ethanol adsorptio n energy. The is fraction of solvation energy retained when dimer adsorbs relative to full solvation. The is fraction of the surface area covered by adsorbed dimer. Alternatively, the liquid state adsorption energy could also be calculated using the gas phase adsorption energy and the solvation energy via a thermodynamic cycle shown in Figure 4. 1. 94 Instead of going through pathway A to B as done in Eqn. 4. 3, one could go from A to D to C to B to complete the thermodynamic cycle. Therefore, following the energy terms associated with the path, we can define that the liquid state adsorption energy of dimer from a thermodynamic cycle as : Eqn. ( 4 . 4 ) where is the solvation energy of lignin dimer (in EtOH for this study) and is the adsorption energy of the bulk solvent on the metal surface. The term is the estimated fraction (from 0 to 1) of solvation area retained after adsorption in the liquid phase compared to fully solvated in the solvent, which could be approximated in several ways as detailed in Section 3.2b. The term is the fraction from 0 t o 1 of the area of the surface projected on to it by adsorbed lignin dimer ( with perimeter drawn by d surfaces (including the area from the top and bottom of metal surfaces due to periodic bound ary condition) . The is calculated via an explicit interface model as: Eqn. ( 4 . 5 ) where is the energy of the cell of 37 ethanol molecules packed into a 3735.55 Å 3 volume ( corresponding to 0. 76 g/cm 3 (s ee Section 2.2 for calculation detail s ) between metal slabs with van der Waals radii taken into account, is the energy of the metal slab, and is the energy of the ethanol slab by removing metals from without further relaxation to retain the same density. 4.3.2 Solvation Energy of Lignin Dimer by DFT Implicit and Explicit Model The liquid properties are more sensitive to the temperature change than those of solids. Although the DFT models in general ignored the entropy contribution from translation and 95 rotational modes , to capture the temperature effect on the solvent, the corresponding ethanol density and dielectric constants at the typical experimental temperature were considered for the following calculation setup. Equation 4 .4 requires the knowledge of the solvation energy, therefore it is computed from both implicit and explicit models. For an implicit model, the effect of solvation was simulated using VASPsol 303,304 From which, the is obtained from: Eqn. ( 4 . 6 ) where is the energy of the dimer in vacuum. For ethanol at 50 °C, the 305 In an explicit model of solvatio n, the energy is calculated by: Eqn . ( 4 . 7 ) where an energy in each term was obtained from averaging DFT relaxed 5 initial structures generated from classical MD ran at 50 °C (Section 2.3) . For the ethanol only cell, t o achieve an ethanol density of 0.76 g/cm 3 (at 50 °C), 37 molecules of ethanol were packed into a cell volume of 3735.55 Å 3 . For a cell of ethanol with a dimer solute, the cell volume was increased to 4152.91 Å 3 to accommodate for the molar volume of lignin dimer. This approach allow s the density of the bulk liquid phase to be kept approximately constant (0.04 g/cm 3 change). Due to the lack of experimental values, the density of the lignin dimer was approximated to be 1.28 g/cm 3 using COMPASS II force field isothermal isobaric ensembl e (NPT) MD. The calculation detail and result validations with other similar known density molecules, which show 3 - 7% error from experiments, can be found at Table C 1 in the Appendix C . It is acknowledged that the molar 96 volume of the dimer in ethanol could change; the molar volume of phenol in water decreases by 7%. 306 Therefore, additional tests for on several molecules were performed to assure sufficiently accurate results (Section 4.5 .1). 4.3.3 Additional Classical Molecular Dynamics The liquid structures are all first generated by classical MD methods. The same calculation approach is also used to evaluate the adsorption of a lignin polymer to confirm whether the lignin dimer selected for the present study is representative of polymeric lignin. These classical MD simulations were carried out in the Material Studio software package using the COMPASS II force field ( type c3a, c43o, c4o, h1, h1o, o2e, and o2h ) commonly used for study of o rganic polymers. 178,307 The simulation was performed at the required temperature for a duration of 200 ps (100 ps equilibration, 100 ps production) at a time step of 1 fs. Tempera ture coupling was performed with the Nose algorithm (Q = 0.01). An accuracy of 10 - 4 kcal/mol Edwald summation method was used for neighboring electrostatic calculations. The van der Waals interaction switch function was set at a cutoff distance of 15.5 Å. 4.4. Development of Computational Approaches The model lignin dimer is a complicated molecule as it involves several degrees of freedom along the side chain. Moreover, there is currently no experimental data and very limited computational data for benchma rking purposes. Therefore, to ensure that the selected calculation methodologies are suitable and reliable, simpl er aromatic and polymer models were first calculated and benchmarked in this section. 4.4.1 T he Significance of van der Waals C orrection on A dsorption of A romatics Although contributions from van der Waals interaction by definition should be relatively minimal, this was not the case for - bond - metal interaction that dominated adsorptions. 97 The calculation of adsorption of smaller aromatics in this section was performed using a smaller (4 × 4) Ni super cell. Benzene was selected as a model to investigate s elected choice of calculation methodology, particularly the importance of inclusion of appropriate van der Waals corrections, based on the abundant literature 48,308 for benzene and superficial structural resemblance to the lignin dimer. Van der Waals interactions were accounted for using optB88 - vdW 48,308 and DFT - D3 31,309 functionals. Table 4 . 1 . Calculated a dsorption e nergies (eV) of benzene on close - packed metal surfaces with different calculation methodology and comparison to the existing literature. Cu(111) Ni (111) PBE/PAW - 0.03 - 1.02 PBE + DFT - D3 - 0.99 - 2.13 PBE + optB88 - vdW - 0.72 - 1.80 Theoretical - 0.68, 48 , a - 0.79, 48 , b - 0.53 308 , c - 1.79, 308 , a Experiment - 0.62, 310 - 0.58, 311 - 0.68 to - 0.81 312 - 1.95 50 References employing a PBE + optB88 - vdW, b PBE + vdW surf . The results in Table 4. 1 show adsorption energies of benzene lying horizontally on Ni and Cu(111) bridge site, which was previously shown to be the most preferable by various metals. 48 The results indicate a systematic tren d among investigated metals, wherein PBE alone substantially underestimates the adsorption energies, as the underbinding nature of GGA is well - established 313 . This can be addressed by incorporating van der Waals cor rections. By correcting this using DFT - D3, the adsorption energies were all slightly over - compensated compared to experiments. This is in agreement with a previous study of benzene on Ag(111). 314 Using o ptB88 - vdW the results are in closer agreement with experiment results on a Cu(111) and slightly underestimate for the case of Ni(111). 315,316 These indicate that optB88 - - bond interaction with d - electrons of the transition metal. Since coupling PBE with optB88 - vdW results in the most accurate energy for this system, this combination will be employed throughout the pr esent study. 98 4 . 4. 2 Determining P referable A dsorption S ite u sing G uaiacol Where the lignin dimer adsorbs on the surface can significantly impact the adsorption energy. To locate the most preferable site, we employed guaiacol, which is structurally equivalen t to a unit of the dimer. Figure 4. 2 shows the adsorption of guaiacol at the different sites of Ni(111) The results indicate that guaiacol oriented as 30° to the bridge site is the most stable site with a of - 2.00 eV ( Figure 4. 2). This site and angle preference were in agreement with adsorption of phenol on Pd(111). 31 Therefore, adsorption of both aromatic rings of the lignin will be positioned to best fit the bridge sites. Figure 4 . 2 . Adsorption energy ( ) of guaiacol at four different sites, top, bridge, FCC, and HCP, with two orientations, 0° and 30° with respect to the hexagonal ring formed by of nickel atoms on the top layer of the slab. 99 It should be noted that the of guaiacol, which contains two more oxygenated functional group than benzene, on Ni only shows 0.08 eV higher . On a Cu(111), guaiacol is - 1.04 eV (result not shown), 0.32 eV higher than benzene. These indicate that the functional groups could have different contributions to the on different metals. However, the major contribution to the - bonds on the benzene ring with the metal surface. Hence, although the dimer model used in this study is a G unit of lignin, the results are expected to be extendable to S units of lignin as well. 4 .4 . 3 Representing Polymeric Lignin S tructure with a D imer Figure 4 . 3 . Top view and side view of relaxed hypothetical lignin of 10 G - - O - 4 linkages and possible adsorption conformations of the polymer on metal surfaces ( A ). The approximated adsorption energy of the polymer ( ) for a given nu mber of units adsorbed on Ni and Cu and conformation energy ( ) for a given number of unit fixed to lie horizontal (B). I n a natural state , the fully relaxed polymeric lignin is highly twisted and tend to self - aggregate ( Figure 4. 3A top ). In contrast, the adsorption of lignin is assumed to require the aromatic ring s to be flat and parallel to the - bond - metal interaction. This would therefore exert constraints on parts of the lignin dimer. The energy cost to constrain the dimer is defined here as conformational energy ( ). Unlike the lignin dimer that only has two units, the adsorbed conformation of a polymer chain can vary from an end - of - the - chain adsorption 317 to 100 a completely flat configuration ( Figure 4. 3A), 318 which is a tradeoff between the gain from adsorption energy and cost due to conformation energy. Therefor e, t he objective of this subsection is to determine the adsorb ed conformation of the polymer in vacuum and evaluate if that conformation c ould be modeled using a dimer. This is accomplished by first calculating the of a hypothetical 10 - unit polymer comprised of only - O - 4 linkages using classical MD. The is the energy cost by constraining a certain number horizontally parallel to an imaginary plane (as if they were in adsorbed state), which can be expr essed by : Eqn. ( 4 . 8 ) where is the average energy (from 3 relaxed frames of NVT MD) of the polymer with a set number of monomer units constrai ned, and is that of a fully relaxed polymer with no units constrained. The resulting is shown in Figure 4. 3 B . The increases from 0 eV up to 4 eV as more units in the polymer were constrained. To compare of the polymer was estimated by : Eqn. ( 4 . 9 ) where is the DFT - calc ulated per unit. For the aromatic unit adsorption, the is - 0 .72 eV for Cu and - 1.80 for Ni. In both of these estimations, the outweighed the significantly, indicating that the lignin polymer preferred to adsorb entirely flat on Ni (111) and Cu - O - 4 lignin dimer , presumably with both rings adsorb on the surface, can be used to represent the adsorption of the lignin polymer. 101 4. 4. 4 Identifying the m ost Stable Dimer Adsorption Conformation i n Gas Phase Because a lignin dimer can exist in multiple conformations, the first task for this section is to identify the stable free lignin conformation. To achieve this , 15 conformations of a threo - lignin dimer were generated and compared with respect to energy. The image of the most stable conformation is shown in Figure 4. 4 (free dimer) . T he relative energy difference between the least and the most stable conformation studied was 0.33 e V. The stability of the lignin conformation was found to be partially affected by the intramolecular hydrogen bonds. Focusing on just the propyl chain, stable conformations (~0.15 eV difference from the most stable one in this study) - O - the conformat ion (3 was the max imum ). This is in agreement with previous reports suggesting that intramolecular hydrogen bond ing play an important role for lignin conformation . 319,320 In addition, it should be noted that several other conformations exhibited energies close to the most stable one (~0.02 eV), suggesting there could be more than one solution. However, a common feature among all the studied free dimers is that the angle between the aromatic ring and the propyl chain atoms next to it is about 180°, which is to be e xpected of a free dimer. Upon adsorption with aromatic rings completely parallel to the Ni(111), the angle of the threo - dimer was no longer 180° as in the case for the relaxed two - ring adsorbed conformation (conformation A and B) in Figure 4. 4A and 4B. Th is supposedly results in an increase in conformation energy. Therefore, to ensure that the most likely stable conformation is identified, other conformations that are expected to exhibit less constraints were also investigated. These conformations are call ed one - ring conformations and include conformation C, D, and E in Figure 102 4. 4. The common trait among these three conformations is that they allow one of the aromatic rings to be relatively free, while the other aromatic ring adsorbs on to the surface. The resulting for each of these dimer s is presented in Table 5.2 . As the results show higher from both conformation A and B, it is clear that the interaction between the aromatic ring and the metal outweighs the conformation energy by having two rings completely adsorb on the surface. Furthermore, conformation A was shown to be statistically more stable than B. The bond lengths between these two are shown in Fig. 5.4 and are mostly similar. The main difference - O - 4 bond where, in one conformation, this oxygen atom is away from the surface (2.91 Å away, Fig. 5.4A), and the ot her closer to the surface (1.94 Å away, Fig. 5.4B). Conformation A is analogous to a previous study using 2 - phenoxy - 1 - phenylethanol as a model dimer adsorbing on Pd(111). 14 On the other hand, conformation B is similar to what was used in a recent DFT study on reaction pathways during lignin hydrogenolysis. 13 Taking their reported adsorbing structure an relaxed using our setup resulted in of - 3.33 eV , a similar value to the result for conformation B in the present study ( - 3.50 eV from Table 5.2) . As conformation A is shown here to b e more stable, this may affect subsequent reaction path ways . It should be noted that we have investigate d changes within each conformation using at least 5 additional images , each with minor changes such as adjusting dihedral angles and alternating position of the methoxyl group, to determine how these perturbations impact . The results are presented as an average ( ) with standard deviation shown in Table 5.2 . The image shown in Fig. 5. 4 is of the most stable conformation with the atomic position of conformation A and B provided in the supporting materials . 103 Lastly, f or lig nin dimer adsorption on a Cu (111) surface, only conformation s A and D were selected for comparison between two - ring and one - ring adsorption. Although the interaction between Cu and the dimer was weaker than that of Ni, as observed by the significantly lowe r (Table 5.2 ), the adsorption conformation A is still more favorable . Figure 4 . 4 . Relaxed adsorption conformations of lignin threo - dimer where in A) and B) both rings adsorb to the surface with both aryl rings completely planar, C) only one ring is completely planar, D) and E ) only one ring adsorbs. The change in angles between the aromatic ring to the atoms next to it in conformation A and B compared to relaxed free dimer is shown in the enlarged ima ges. The d istances ( Å) from the surface to all oxygen atoms in the molecule are also shown for conformation A and B. To supplement the results on adsorption of the threo - isomer the adsorption of the erythro - isomer on Ni (111) using two conformations is next investigated. These two conformation are derivative s from threo - isomer in the Fig s . 4. 4A and 4. 4 B structures . To reserve conformation A 104 and B for threo - isomer and prevent confusion, these derived structured are called conformation I and II shown in Figure 4. 5I and II, respectively. Unlike the threo - isomer, the result s show that conformati - O - 4 is close to the surface, is now the most stable conformation (Table 3) , and is in agreement with previous ly published literature . 44 Th e reason that the conformation in Fig. 4. 5I is not the most stable could be due to the strain at its ether oxygen a tom that need s to b e as high as 3.14 Å away from the surface compared to 2.91 Å for the conformation in Fig. 4. 4A to remain connected with the propyl chain of this isomer. This resulted - O - 4 bond angle of 130° , compared to 122° in both Figure 4. 4A for the threo - isomer and for the free lignin ( data not shown). Overall, the observed difference s in adsorption conformation implies that erythro - and threo - isomer could undergo different reaction paths. Table 4 . 2 . Adsorption energy in eV and average aryl carbon atoms distance to metal surface ( d ) in Å for different adsorption conformations of the threo - dimer (shown in Figure 4.4 ) and conditions on Ni(111) and Cu(111) surfaces. Metal Ni(111) Cu(111) Conf. A B C D E A D Vacuum - 3.59 - 3.50 - 3.10 - 2.48 - 2.26 - 1.99 - 1.55 - 3.53 ± 0.03 - 3.33 ± 0.10 - 3.06 ± 0.04 - - - - d 1.98 2.00 1.98 1.97 1.99 2.97 2.77 r = 21.04) by Eqn. 4. 2 - 3.55 - - - - - - d 1.97 - - - - - - r = 80.00) by Eqn. 2 - 3.52 - - - - - - d 1.98 - - - - - - 3 ) by Eqn. 3 (explicit) - 0.58 ± 0.55 - - 0.46 ± 0.54 - 0.35 ± 1.16 0.76 ± 1.05 d 1.99 - - 2.00 - 2.90 3.08 Thermodynamics cycle by Eqn. 4 .4 (a) (b) (c) - 0.58 ± 0.20 - 1.09 ± 0.50 - 1.21 ± 0.53 - - 0.24 ± 0.18 0.03 ± 0.37 - 0.05 ± 0.36 - 0.46 ± 0.14 0.40 ± 0.33 0.53 ± 0.26 0.69 ± 0.1 0 0.60 ± 0.44 0.72 ± 0.32 (a) The in Eqn. 4 . 4 was calculated based on solvent accessible surface area (SASA) using probe radius of 1.25 Å , (b) by the number of neighbor atoms within 2.5 Å distanc e , and (c) by number of bonds within the 2.5 Å distance. 105 Figure 4 . 5 . Relaxed adsorption conformations of lignin erythro - dimer where both rings adsorbed on the Ni(111) with one conformation yielding an ethyl oxygen atom away from the surface (I) and the other adsorbed onto the surface (II) with d istances from the surfaces to all oxygen atoms in the molecule and some bond distances within the molecule. Table 4 . 3 . Adsorption energy ( ) in eV and average aryl carbon atoms distance to Ni(111) surface ( d ) in Å for the two different adsorption conformations of the erythro - dimer (shown in Figure 4.5 ) . Metal Ni(111) Conf. I II - 3.1 6 - 3.7 3 - 3.1 6 ± 0.00 - 3.6 3 ± 0.10 d 1.98 1.96 4.5 Adsorption Energy of Lignin in Solvated Environments 4.5 .1 Compar ison of the I mplicit and E xplicit M odel for Solvated Lignin Adsorption The adsorption energies of a solvated lignin dimer from both implicit and explicit models are shown in Table 5.2, where t two - ring and one - ring adsorption was taken from the most stable threo - isomer adsorption in vacuum ( i.e. the A and D conformation ) . The explicit model and implicit models gave very different adsorption energies. In the implicit model, , i 106 at 50 °C) the adsorption energy change very slightly compared to that of the vacuum state. To t the change in remained minimal. In contrast , the explicit solvation model yielded an average of - 0.58 eV for two - ring adsorption on a Ni(111), a 2.95 eV reduction compared to its vacuum counterpart (Table 5.2). Furthermore, on a Cu(111) the adsorption decreased so drastically that the is positive for both the one - ring and two - ring conformation, suggesting that the dimer no longer pre fers to adsorb on the surface at this liquid density. It should be noted that the large standard deviations of the explicit model results are associated with the fluctuations in NVT MD from which the images were sampled, and are expected of liquid phase si mulations. A test run using ab initio MD for 3 ps production run also yielded similar fluctuations (results not shown). Apart from the energy, there are also differences in the relaxed geometry between the implicit and explicit model. In the results using the implicit model, there was virtually no change in either bond length or the distance of the dimer from the surface compared to the gas phase conformations (within 0.03 Å ). However, noticeable structural changes can be observed from the explicit model. F or the case of the two - ring adsorption conformation on a Ni(111), the entire hydroxymethyl group shifted away from the surface and closer to the solvent phase primarily due - hydroxyl group with ethanol. For example, the oxyg en atom of - hydroxyl became 3.73 ± 0.26 Å away from the surface compared to 3.29 Å in the vacuum state as shown in Figure 4. - O - 4 ether oxygen atom, which drifted 0.10 ± 0.04 Å away from the s urface compared to its vacuum state. - hydroxyl group due to its structural - O - - hydroxyl group to form 107 intramolecular hydrogen bon ds with these ether groups and less likely to interact with ethanol. Apart from these, no significant changes were observed among the aromatic atoms. On the Cu(111), similar changes were observed upon explicit solvation. Table 4 . 4 . Comparison of solvation energy ( ) of similar aromatic molecules in alcohol solvents using different implicit models, an explicit model, and from experiments. Conditions Solvation Energy Solute Solvent T° Density (g/cm 3 ) VASP sol COSMO Explicit Exp Phenyl ethanol EG 25 °C 1.11 40.00 - 0.20 - 0.36 - 0.63 ± 0.38 - 0.68 321 1,3 - Dimethoxy benzene 2 - PrOH 25 °C 0.79 20.18 - 0.09 - 0.30 - 0.60 ± 0.52 - 0.58 322 Benzoic acid 2 - PrOH 25 °C 0.79 20.18 - 0.20 - 0.35 - 0.78 ± 0.28 - 0.76 322 Benzyl alcohol EtOH 25 °C 0.79 24.50 - 0.16 - 0.35 - 0.81 ± 1.04 - 0.66 322 Dimer * EtOH 50 °C 0.76 21.04 - 0.63 - 0.78 - 1.53 ± 0.74 - * threo - guaiacylglycerol - beta - guaiacyl ether To ensure that the description of the solute - solvent interactions by an explicit model is dependable, solvation energies of similar aromatic molecules in alcohols were calculated using implicit models and an explicit model and compared to experimentally determined values (Table 4). It should be noted that for this, another implicit model (COSMO) was also u sed for comparison purposes. The differences between VASPsol and COSMO include, but are not limited to, the method used for calculating cavity size and electrostatic interactions. 119 From the results, we can see that both implicit models consistently underestimated the solvation energy. This could be due to their limitations in describing site - specific interactions between solute and solvents ( e.g. hydrogen bonding) as 323 In contrast, the explicit model showed promising results compared to experiments and support that the results from explicit solvation model are the most reliable in the present study. These results suggest the adsorption 108 energy difference between the implicit and explicit model is mainly due to the inaccurate description of lignin - solvent inte raction in the implicit model. 4.5.2 Calculation of Solvated Lignin Adso rption Energy from a Thermodynamic Cycle Figure 4 . 6 . The values for solvent accessible surface area (SASA) using a 1.25 Å probe radius (grey mesh) , the number of neighboring atoms (enlarged solvent atoms), and the number of bonds within a 2.5 Å radius before and after adsorption for relaxed A and D conformation from Table 4. 2 in ethanol. The average value is shown in parentheses. Calculation of using Eqn. 4. 3 is an expensive calculation and further scaling up of the cell size would be computationally prohibitive for most cases. In this section, we propose strategies to approximate adsorption energy in the solvated state using a thermodynamic cycle ( ). As shown in Figure 4. 1 and Eqn. 4. 4, the can be calculated by leveraging known with calculated and bulk solvent adsorption energy 109 ( ), all of which can be simulated with les s expense than direct calculations. For example, does not include metal atoms and can be calculated in smaller cell and scaled up as necessary using the work of adhesion concept, 111 i.e. adsorption per area. The fraction of the metal surface area covered by lignin dimer ( ) was assumed to be the same as in gas state. This leaves only the parameter to be calculated, which is the fraction of solvation su rface of the dimer in the solvated, adsorbed state ( Figure 4. 1C) relative to the fully solvated state ( Figure 4. 1B). Here, we investigated three approaches to estimate the , including surface accessible surface area (SASA), the number of bonds, and the number of neighboring atoms. The SASA is defined as the surface area of a molecule that is accessible to a solvent , and is typically used in biomolecule investigations . For these, we employed a cutoff radius of 2.5 Å for bond length and number of neighbori ng atoms . I t should be noted that the number of neighbors is not equal to number of bonds because some neighbor atoms may form bonds with multiple dimer atoms. Correspondingly, the probe radius of the SASA is 1.25 Å. The cutoff was chosen based on the assu mption that the solvation energy was mostly influenced by neighbor atoms and that 2.5 Å al so represents a strong hydrogen bond distance. The values were calculated using Python scripts coupled with the PyMOL package. 324 Figure 4.6 shows the results of these simulations for the SASA, the number of neighbor atoms, and the number of bonds. The SASA is represented with a grey mesh around the dimer whereas neighboring atoms are enlarged and translucent. Taking SASA as an example, from t hese we can see that in the bulk solvent, the lignin dimer has ~550 Å 2 of SASA. As expected, the accessible area decreased upon adsorption to ~300 Å 2 for Ni(111). The trend is similar using other criteria. It should be noted that there is not a significant difference between the SASA of the two - ring ( Figure 4. 6A) and the one - ring conformations ( Figure 4. 6D). Originally, it was anticipated 110 that the one - ring conformation may allow more interactions between the dimer and ethanol molecules. However, all three c riteria suggested equivalent, or even less, interactions because the folded structure of the one - ring conformation limited the exposure of part of the dimer to solvent. Given these values for each criteria, the value for each conformation is the ratio o f the values from a selected criterion of that conformation to that of the dimer in the bulk solvent. For example, the value for the adsorption conformation presented in image A1 of Figure 4. 6B calculated by SASA is 0.53 and was obtained from 288 divide d by 545. All Figure 4. 6. Table 4.2 presents the results for the determined using the calculated values in Eqn. 4.4. The results indicate that using the SASA for yielded the closest approximation to , which typically show only ~0.1 eV error and correctly predict that the dimer would not adsorb on Cu. More importantly, the SASA of vacuum and solvated states are almost identical. Calculating the SA SA of two - ring adsorbed dimer in vacuum yielded 285 Å 2 , compared to 288 Å 2 for the solvated state. On a Cu(111), the SASA in vacuum state is 315 Å 2 , compared to 319 Å 2 in the solvent. This means that users would not be required to generate solvated structu res, which could inherently be dependent on the type of methods used. Since the SASA is highlighted, it should be noted that SASA is also not sensitive to the chosen probe radius. As shown in Figure S1, the value only changed by 0.1 in response to chan ging the probe radius from 0.5 to 1.5 Å. For Cu, similar to the results from , estimation using with SASA indicates that the dimer does not prefer to adsorb at this density. Equation 4. 4 implies that the use of a Cu catalyst with alcohol solvent for depolymerization of a lignin dimer would require high temperature to mitigate the solvation effect and allow lignin to adsorb. This is in agreement with the observed experimental conditions w here Cu - based catalysts were typically found to be 111 used with these elevated temperatures. 228,325 On the other hand, it also explains why a majority of the successful depolymerization experimen ts, particularly those using more mild conditions, employ Ni - based catalysts which can maintain adsorption of dimer even in solvated conditions. 222,226,291 Equation 4. 4 highlights the effect of solvents on adsorption ener gy and suggests strategies to improve adsorption via either choosing solvent that has less interaction with the dimer (lower ), or choosing a solvent that has less interaction with the metal surface (lower ). Investigating to the con tribution from each term in Eqn. 4.4 shows that for both Ni(111) and Cu(111) Given that the former is harder to tune (as solvation of lignin is required), it may be easier to find solvent that can dissolve lignin but has less interaction with the metal sur face. This is in agreement with experiments which investigate the effect of solvents on hydrogenolysis of a lignin 4 - O - 5 model dimer and found that the highest yields were obtained in decaline and methylcyclohexane, both of which are expected to yield low , due to being relatively large molecules and completely hydrogenated. 40 Therefore, the use of Eqn. 4. 4 can help approximate adsorption energy of dimer and aid in solvent selection for a particular catalyst. 4.6 Conclusion The present study has provided a systematic investigation and analysis of the adsorption conformation of lignin dimer s representing a vacuum and solvated state . First, it was demonstrated - O - 4 dimer exhibited similar adsorption behav ior to its polymer counterpart. The prefe rred adsorption conformation for both a erythro - and threo - isomer of the dimer was found to be with two aryl rings adsorbed on the surface. However, the erythro - isomer prefers the 112 conformation where its ether oxygen on the alkyl chain is close to the surfa ce, whereas the threo - isomer favors the conformation where the ether oxygen is away from the surface . Adsorption energy of the dimer in vacuum on Ni(111) was also shown to be significantly high er than that on Cu(111) surface . Upon solvation with ethanol using an explicit representation of the ethanol at a density corresponding to 50 °C, the adsorption energy was reduced significantly for all surfaces. While the dimer prefers adsorbing on Ni(111), the reduction in adsorption energy leads to endothermic ads orption for Cu(111). As implicit models were found to not be able to sufficiently describe solvent - solute interactions, and the explicit model is quite expensive computationally, an alternative, more computationally efficient approach to calculate the solv ated adsorption energy was developed based on a thermodynamic cycle. The model resulted in errors on the scale of only ~0.1 eV relative to the explicit model and correctly predicted that the dimer would not adsorb on Cu in the studied condition. This findi ng would allow for future studies to be able to select the solvent for a particular metal catalyst more effectively relative to explicit solvation method. Moreover, these results emphasize the critical impact that the solvent plays on lignin adsorption dur ing processes employing heterogeneous catalysts for lignin upgrading. 113 CHAPTER 5 Catalytic Hydrogenolysis of Lignin to Aromatic Monomers 5.1 Summary Lignin hydrogenolysis has been recently been studied extensively as it shows high yields and selectivities with limited condensation reaction. Moreover, the reaction can also exploit other hydrogen source apart from H 2 such as from ethanol, i.e. catalytic transfer hydrogenolysis. A number of studies have shown that several supported metal catalysts can be used for t he reaction using biomass as a substrate. The reaction does not only serve as a lignin valorization method, but also biomass pretreatment. However, due to variable quality of lignin and biomass substrate , it is unknown how do these catalysts compare both i n term of monomer yields, and the quality of pretreated biomass. In this study, a series of supported metal catalysts (metal/C) is used for catalytic transfer hydrogenolysis using birch biomass as a substrate and ethanol as a solvent at 200 and 220 °C . The catalysts were chosen based on previous report that it has been used for ethanol reforming . The results revealed that Co/C, Ni/C, Ru/C, and Pd/C can give high yields and selectivities, in a respective order. Particularly, Co/C has not been used previously for catalytic hydrogenolysis transfer of lignin before. It was also showed that, for this method, the amount of - O - 4 contents is the most critical functional group the reaction. In term of pretreatment efficacy, biomass resulted from these catalysts show ed significant higher removal of xylan and lignin, compared to catalysts that give low yields and no catalyst at all. Moreover, in all cases glucan was almost perfectly retained. Significant improvement in glucose yields were observed from reactions with c atalysts that gave high yields, particularly it was shown that the glucose yields were correlated to lignin removal. These show that catalytic hydrogen transfer, with an appropriate catalyst, is an effective method for both lignin monomer generation from l ignin and pretreating biomass. 114 5.2 Introduction The plant cell walls of lignocellulosic biomass are comprised primarily of the structural polymers cellulose, hemicellulose, and lignin. Among these, lignin is the most underutilized component where it is ch eaply burned at site. Considering that it comprises 17 - 33% of the biomass, 13 there has been a sig nificant effort in valorization of lignin. One of the methods is depolymerization of lignin to valuable products through hydrogenolysis, in which the products are typically followed by hydrodeoxygenation in the same reaction pot. 5,281 The methods have drawn significant attention in the past several years due to its abil ity to apply to whole biomass, i.e. the lignin valorization step has become pretreatment step in itself, showing limited condensation in certain conditions, and the compatibility with hydrogen transfer reactions, which allows it to leverage other hydrogen sources than pressurized gas. 274,326 A number of catalysts for the reaction has been proposed and shown to give high yields, such as Ni, Pd, and Ru - based catalysts. 19 However, different studies using different lignin sources, conditions , and catalysts have resulted in not just monomer yields and, sometimes, product types that are not comparable, but also the impact on the pretreated biomass. The lack of benchmarking standard hinders the process of catalyst selection and design. Therefore , the present study aims to fill this gap by determining the efficiency of different supported metal catalysts on both monomer yields and quality of the pretreated hardwood biomass by a mean of enzymatic hydrolysis. The reaction will be performed using hyd rogen transfer from the selected solvent, which is ethanol in this work. Early dedicated lignin hydrogenolysis studies were performed by Pepper et al ., where it was shown that using whole softwood biomass as a substrate, monomer yields as high as 22% coul d be achieved by using Raney Ni catalyst with water/dioxane (50:50) solvent. 327,328 Since then, there has been significant development in the process including identifying new catalysts, s olvents, 115 and new hydrogen sources. Arguably, the reaction gains the most traction when it was demonstrated that using Ni/C with hardwood and alcohol solvents, lignin monomer as high as 50% were achieved with >90% selectivity. 226 Many other studies using similar conditions with Ni/C have shown similarly high yields ranging from 26% to similarly high depending on biomass type. 275,283 285 Other supported transition metal catalysts a lso have shown promising results, for example Pd/C, Ru/C, Pt/C, and many other transition metal alloys. 35 37,223,233 These indicate that there are many possible transition metals, possi bly more that has not been investigated, that could efficiently perform lignin hydrogenolysis. Many of the hydrogenolysis reactions did not use H 2 as a hydrogen source, but instead relies on the hydrogen transfer mechanism (catalytic transfer hydrogenolys is), such as from alcohols, 200,226,229,326,329 formic acid, 330 and dioxane, 233 each of which with different donating pressure - rating vessels as one would do for molecular hydrogen. However, this also indicates that the efficacy of the reactio n would also partly depend on the alcohol reforming activity of the catalyst. As alcohols are the most commonly used hydrogen donors and ethanol will be used in the present study, it is worth discussing about the capability of ethanol reforming of some of transition metal catalysts. Ethanol reforming is a process that convert ethanol to methane and other syngas and is typically conducted at very high temperature (600 - 800 °C). 331,332 Adsorbed hy drogen atom from the process could be the source for lignin hydrogenolysis. Mechanistically, ethanol reforming is a complex reaction process that involve several molecular decompositions, water activation, and oxidation of the adsorbed carbon species. 331 Metals that are typically used include, Ni, Co, Fe, Cu, Pt, Pd, and Rh. 331 334 Therefore, these metals could have potential for lignin hydrogenolysis. Particularly, to our knowledge, supported Co and Rh catalyst have not been previously used for 116 catalytic transfer hydrogenolysis. Nevert heless, it should be noted that some of the metals have different decomposition mechanism compared to others. For example, on Ni and Co, ethanol first loses the hydroxyl hydrogen, resulting in adsorbing species, whereas on Pt and Pd, hydrogen abstraction first occurs at the - carbon position, resulting in adsorbing species. 330 These adsorbing species ar e likely to have different affinities to the metal surface. Previous studies have proposed that adsorbing species, particularly one with stronger interactions with the catalyst, could block lignins from getting to the surface and hence results in lower hyd rogenolysis yields. 40 Therefore, the low /high final monomer yields could be partly attributed Regardless of catalysts and conditions, a common trait among all the examples that resulted in relatively high monomer yields is that a high - quality lignin, i.e. high amount of - O - 4, was used as a substrate. This is in agreement with studies indicating that C - O ether bonds in lignin are significantly more fragile than C - C bonds; 295 additionally, investigation of the reaction pathways from some of the catalysts (using moderate temperatures around 200 °C) point out that the products are primarily from cleavage of the - O - 4 bond. 19,226 High - quality lignins are mostly obtained in situ during the hydrogenolysis rea ction of biomass, where lignin is extracted out by the solvents. This is because lignins isolated from biomass pretreated by traditional methods, such as dilute acid or kraft, typically has a degraded structure where - O - 4 amount is low and significant amo unt of C - C bonds was formed during the pretreatment. 335,336 However, it should be noted that with recent advances in pretreatment processes, high quality lignin could also be obtained from nov el pretreatments such as Cu - catalyzed alkaline - oxidative (Cu - AHP) 249,337,338 and - valerolactone pretreatment, 339 where studies have shown that lignin retain high amount of its native - O - 4 117 contents. 340 In the present study, to ensure high quality of lignin, whole biomass will be used as a reaction substr ate to maximize monomer yields. Using biomass as a substrate implies that lignin valorization process has also become a pretreatment process in itself, which is also commonly known as lignin - first biorefinery. 33,341,342 As such, apart from the monomer yields, the process should also be subjected to criteria that are traditionally used to evaluate the efficacy of a pretreatment process. One of the common method is to measure how much the pretreatment has increase the acce ssibility to the carbohydrates. This is typically performed by enzymatic hydrolysis to produce sugar monomers. Some of the factors that could impact enzymatic hydrolysis yields include lignin content, hemicellulose content, and inhibits enzyme accessibility and its removal is well known to be vital for enzymatic hydrolysis. 343 However, it should be noted that it is not in a linear relationship with sugar yields as cer tain biomass and pretreatment do not require high removals to obtain high sugars. 344 Similarly, hemicellulose was also reported to inhibit accessibility, 344 although this could be addressed by addition of xylanases. 345 Lignin re - deposition and other s tructural changes that affect sugar yields. 346,347 Without addition of cata lysts, the lignin valorization process using biomass substrate would be similar to an organosolv pretreatment. Previous ethanol - based organosolv pretreatment is typically mixed with water and slight addition of acids to help remove hemicellulose. 348,349 The pretreatment were able to remove lignin down to just 5% depending on the severity. In the present thesis, no acid will be added, and hence the effect from catalyst on this process in term of enzymatic hydrolysis would be demonstrated. In summary, previous studies have suggested that several transition metal catalysts can be used to perform lignin hydrogenolysis using whole biomass as a substrate, particularly through 118 hydrogen transfer reactions. However , due to diversity of lignin sources, pretreatments, and reaction conditions, it has been difficult in determining the efficiency of different supported metals catalysts, both on the criteria of monomer yields and as a pretreatment. Therefore, in the prese nt study different metal/C catalysts, particularly those that have been shown to possess ethanol reforming capability, will be compared using the same source of lignin, from which monomer yields, characteristics of pretreated biomass, and enzymatic hydroly sis yields will be determined. The metals include Fe, Co, Ni, Cu, Zn, Ru, Pd, and Ag. Additionally, the same treatment will be performed on lignins obtained after alkali pretreatment to demonstrate the importance of the quality of the substrate. 5.3 Experi mental Methods 5.3.1 Biomass and Lignin Compositions The biomass used for the present thesis is silver birch ( Betula pendula Roth.) grown in northern Sweden and supplied by Curt Lindström (Smurfit - Kappa Kraftliner AB, Piteå, Sweden). The biomass was charac terized with S/G ratios and - o - 4 content of 3.2 . The biomass was Wiley milled to 5 mm size. Following air - drying, all samples had moisture contents of 3 - 5%, and the composition for each feedstock was determined by NREL/TP 510 - 42618 with minor modification s. 350 The lignin was obtained from alkaline pretreatment of debarked wood chips of 20 - year old hybrid poplar ( Populus nigra x maximowiczii cv. NM6) obtained from Dr. Raymond Miller (Michigan State University Forest Biomass Innovation Center, Escanaba, Michigan) at a liquor - to - wood ratio of 4:1 using conditions identical to those in our previous work for an H - factor of 166. 242 Lignin was obtained by fractionally precipitated from this liquor through two sequential pH reduction approaches. The first one employed sequential acidification and precipitation at using 119 CO 2 at room temperature and atmospheric pressure to stepwise decrease the pH. The lignin was collected at pH 11, 10.6, 10, 9, and by adding 1 M H 2 SO 4 , also at the pH 2. These samples are named RT - 1, RT - 2, RT - 3, RT - 4, and RT - 5, respectively. The second lignin fractionation approach employed the Sequential Liquid - Lignin Recovery and Purification (SLRP) process using CO 2 acidification at elevated temperature (115 °C) and pressure (6.2 bar) as reported in our previous work. 243 For the SLRP fractions, the lignin was collected at pH 11 and 10.5 and named SLRP - 1 and SLRP - 2, respectively. The percent recovery from each pH fraction can be found in our previous work. 286 For both biomass and lignin, the Klaso n lignin content as determined by NREL standard procedure (NREL/TP - 510 - 42618) was used as a basis for subsequent characterization and monomer yield determination. 5.3.2 Catalytic Hydrogenolysis of Lignin and Biomass Catalytic hydrogenolysis of lignin was p erformed according to previous work. 230 In short, 10% metal loading metals/C was synthesized by incipient wetness where the metals were sourced from Fe(NO 3 ) 3 9H 2 O, Co(NO 3 ) 2 6H 2 O, Ni(NO 3 ) 2 6H 2 O, Cu(NO 3 ) 2 3H 2 O, Zn(NO 3 ) 2 6H 2 O, and Ag(NO) 3 , and RuCl 3 xH 2 O, where x is estimated as 0.5, which is an average degree of hydration reported for the product (Sigma - Aldrich). For example, 2.76 g of Ni(NO 3 ) 2 6H 2 O was added to 8 g deionized water and stirred until well mixed. The solution was then added dropwise to 5 g of activated carbon (Darco ® G - 60, Aldrich, water absorption capacity of 2.1 mL/g). The mixture was kept wet at room temperature for 24 h before drying overnight at 100 °C. Then it was calcined at 400 °C under nitrogen atmosphere for 4 hr, and reduced at 400 °C under 1 0% H 2 in an Ar atmosphere, both of which used at a temperature ramp of 6.25 °C/min. For Pd/C, it was purchased 120 from commercially available product (Sigma - Aldrich 205699). The finished catalysts were stored and weighed in an anaerobic glove box. For the hy drogenolysis reaction, 200 mg of biomass (or 40 mg of alkali lignin), 30 mg of catalyst, and 4 mL of ethanol were charged to a 10 mL passivated stainless steel reactor while in the anaerobic glove box. Reactors were submerged into a fluidized sand bath set at 200 or 220 °C for 6 hr. The resulting liquid was collected, filtered, and injected into GC/MS (Agilent 7890A / 5975C) using Agilent DB - WAXETR column (30 m x 0.250 mm x 0.25 m) with 1:1000 split for product. The temperature program holds at 60 °C for 3 min before ramping up at 40 °C/min to 260 °C and held for 2.5 min. Standards were run for quantification. Yields were determined as mass products per mass Klason lignin using standard curves constructed from pure monomers including 4 - propyl guaiacol, 4 - pr opyl syringol, 4 - ethyl guaiacol, 4 - ethyl syringol, and syringol. All standards were purchased from Sigma Aldrich except 4 - propyl syringol, which was synthesized according to Parsell et al., 37 and 4 - ethyl syringol , which was estimated using the same response factor as 4 - propyl syringol. After collecting the liquid, the leftover biomass was washed with excessing amount of water and left to dry at room temperature. 5.3 .3 Enzymatic Hydrolysis Approximately 10 ±0. 3 mg of the pretreated biomass was added to 1.5 - mL Eppendorf ® Safe - Lock microcentrifuge tubes . The hydrolysis was performed at a 5% (w/v) solid loading with the citric acid buffer at a concentration of 50 mM and pH around 5.5 was added based on the finding s of Lan et al . 351 . To prevent bacterial contamination during hydrolysis, tetracycline and cycloheximide cocktails consisted of Cellic CTec2 and HTec2 (Novozymes A/S, Bagsævard, Denmark) and were added at a rati o of 2:1, respectively. For all experiments, the total enzyme loading was 20 mg total 121 proteins per gram of glucan in the biomass sample. The reaction was performed in a shaking incubator at 50 °C at 160 rpm. After 72 hours, samples were collected and analy zed for glucose and xylose using HPLC with a method reported previously. 352 The amount of sugar yields are based on the fraction of glucose and xylose determined from composition analysis of the whole biomass from NREL standard method. 5.4. Monomer Yields from Native Lignin in Biomass Using Different Catalys ts Figure 5 . 1 . Monomer yields from hydrogenolysis reaction at 200 °C (patterned) and 220 °C using different metal/C catalysts with whole biomass. The results are arranged according to group and period in the from left to right, 4 - propylguaiacol (4 - PG), 4 - propylsyringol (4 - PS), 4 - propenylguaiacol (4 - PnylG), 4 - propenylsyringol (4 - PnylS), 4 - ethylguaiacol (4 - EG), 4 - ethylsyringol (4 - ES). The very small green color in Pd corresponds to syringol. 122 From a general observation of Figure 5. 1, it can be seen that the highest monomer yields obtained by using whole biomass as a substrate were obtained from Co and Ni catalysts at ~50% with Ru a nd Pd also resulting in relatively high yields. Not using catalyst resulted in virtually zero monomers. Notably, there is a large difference between the temperature of 200 and 220 °C. This is likely due to that at 220 °C resulted in significantly higher et hanol reforming activity as there was noticeable more bubbles and pressure upon opening the reactors. Noted that by arranging the results according to groups in periodic table ( Figure 5. 1), a volcano - plot like shape could be observed, particularly for peri od 1 where a sufficient number of metals was investigated. This suggests that the energetics related to adsorption and desorption of species on Co and Ni resulted in optimal reactivity. Interestingly, Co - based catalyst was also shown to yield highest amoun t of H 2 during low temperature ethanol reforming (400 °C). 332 From each of the catalyst in Figure 5. 1, we can see that not all of the catalysts that were previously reported for eth anol reforming (all except Zn and Ag) are working well, including Fe and Cu. For Fe, the low yields were likely to be caused by the high affinity of Fe with oxygen. From a previous theoretical study on ethanol reforming mechanism at, it is likely that etho xy adsorbing species was formed at the interface. At the temperatures in the present study (compared to 800°C, where H 2 formation from Fe was comparable to Co and Ni), 332 the adsorbi ng species are likely to stick to the Fe surface. On the other side of the volcano, the reason why Cu does not work is likely to be opposite that of Fe. Studies have suggested low adsorption energy of ethanol on Cu surfaces. Therefore, higher temperatures than the current study (800°C) were typically used for ethanol reforming with Cu to produce high amount of H 2 likely by improving kinetics and lowering the solvent - solvent interactions. 332 Apart from this, the interaction between lignin and the Cu itself was also low. It was even showed that lignin dimer solvated in ethanol at 50 °C did not adsorb on 123 Cu surface anymore. 353 Therefore, higher temperature should be used for Cu to reduce the solvent - solute interactions. In agreement w ith this analyses, in previous studies on catalytic transfer hydrogenolysis of lignin using Cu - based catalysts, studies have shown that a temperature of 300 °C was sufficient for Cu to generate significant H from methanol, resulting in almost complete conv ersion. 228,354 However, due to high temperatures, the products were also completely hydrogenated. Apart from metals that were typically used for ethanol reforming, Zn and Ag were also invest igated. It can be seen that both of which did performed hydrogenolysis as compared to no catalysts, they both resulted in ~7% monomer yields at 220 °C, similar to Cu ( Figure 5. 1). For Zn, it has been previously used as a co catalyst with Pd, where it was f ound through multiple OH sites, where it was proposed to facilitate reaction either by help binding with hydroxyl groups of lignin or it could desorb from the surface as Zn 2+ ions from the reaction heat and deposited back upon cooling down. Nevertheless, in the present study, it is clear that not sufficient hydrogen was generated for both of these catalysts as evidenced by the low yields, and the products were virtually al l composed of unsaturated propenyl side chain. In the context of product types and distribution, it can be seen from Figure 5. 1 that at 220 °C, the metal Co, Ni, Ru, and Pd resulted in mostly 4 - propylguaiacol (4 - PG) and 4 - propylsyringol (4 - PS). The cataly st Cu and Ni gave the highest amount of saturated products and units (>96 % for both catalysts). In the native lignin, there are primarily two units, i.e. syringyl and guaiacyl, where the former has two methoxyl groups on its aromatic ring and the latter has one. The monomer products from each of these groups are expected to retain its original methoxyl group(s). Given that the S/G ratio of the 124 birch used in this study is 3.2 but the S/G ratio of the products from Co and Ni catalysts is 4.83 on 4 % products from guaiacyl units were lacking assuming that monomers from syringyl have reached maximum. This could b e due to several reasons, for example, the guaiacyl units of lignin could be less susceptible to hydrogenolysis, or more likely, higher number of the guaiacyl units are not linked by - O - 4 from both sides. As the reaction primarily aims at cleaving - O - 4, to generate monomer it is required that both ends of the unit are - O - 4 linkage. This could be supported by well - studied topic that during lignin biosynthesis, more guaiacyl units leads to less - O - 4 bonds. 235 Apart from these, a major portion of monomers from catalysts that resulted in low yield is contributed to unsaturated products, including 4 - propenylguaiacol (4 - PnylS) and 4 - propenylsyring ol (4 - PnylG). The same is applied for reaction at 200 °C. Both of these suggest the insufficient generation of hydroge n adatoms in those conditions. Figure 5 . 2 . Monomer yields from hydrogenolysis reactions o f lignin dimer ( guaia cylglycerol - beta - guaiacyl ether) using Pd/C in ethanol with and without presence of H 2 gas. The 4 - vinylG and 4 - PnolG refer to 4 - vinylguaiacol and 4 - propenylguaiacol, respectively. Among all of the catalysts, Pd is the only catalyst that gave unique products including 4 - ethylguaiacol (4 - EG), 4 - ethylsyringol (4 - ES), and a very tiny amount of syringol (result not shown in Figure 5. 1 as it is visually too small). Additionally, unlike ot her catalysts that were reported to yield similar products whether there is with or without H 2 added such as Ni/C. The product 4 - EG 125 and 4 - ES were not reported in previous studies using that use Pd/C with H 2 gas. To ensure that our reactions and product ide ntification is accurate, a lignin dimer ( guaia cylglycerol - beta - guaiacyl ether) was employed for further investigation. Lignin dimer was used to eradicate any other possible side reactions or contributions from other linkages in real lignin. Figure 2 shows the different products obtained from lignin dimer hydrogenolysis on Pd/C with and without presence of H 2 gas. Without hydrogen gas, a similar product profile to lignin was generated. However, addition of H 2 gas resulted in products (4 - propanolguaiacol, 4 - P nolG) that are similar to previous studies. 266 Specifically, if we look at difference H 2 pressure in Figure 5. 2, we could see that at lower pressure, 4 - PG was produced (possibly more if the pressure was ever lower). However, it totally disappears at high pressure. These suggest that lignin on Pd/C undergoes different r eaction path in presence of H 2 , and was likely dependent of the hydrogen coverage at the metal - liquid interface. Future studies to probe into the effect of hydrogen coverage on different catalyst would make an interesting future study. Note that the result s in Figure 5. 2 are estimated from peak areas and were not quantified, therefore, yields were not reported. Additionally, the temperature of the one with H 2 gas was slightly lower due to reactor limits, while the other one needs to be at higher temperature s to ensure sufficient hydrogen is produced. The longer reaction time was used for the one with H 2 to see if the product would eventually change. 5.5 Lower Monomer Yields from Low Quality Lignin To elucidate the importance of using high quality lignin for hydrogenolysis, a similar reaction condition at 220 °C using Ni/C catalyst was performed lignin obtained from alkali pretreated black liquor and precipitated through CO 2 sequential acidification. Noted that it was shown in our previous work that fraction precipitated at higher pH, e.g. RT - 1, and temperature, 126 e . g . SLRPs, tend to have higher - O - 4 content than those precipitated at lower pH. 286 In the pres ent samples, the - O - 4 of alkali lignin ranges from 15 to 31%. Figure 5 . 3 . Monomer yields from hydrogenolysis reaction using Ni/C and ethanol at 220 °C of alkali lignin that was sequentially precipitated through CO 2 acidification. The S refers to syringol. The hydrogenolysis yields from different fractions of alkali pretreatment show signific antly lower monomer yields ( Figure 5. 3). Even the fraction precipitated at higher temperatures only resulted in ~7% yields compared to 50% when whole biomass is used. The 7% is close to yield approximation using - O - 4 content, i.e. 0.3 2 = 0.09. This indicate that, similar to whole biomass, the hydrogenolysis reaction results in almost complete cleavage of - O - 4 linkage s. However, the monomers from alkali lignin are different in that there are 4 - EG and 4 - ES that was only found from Pd/C previously. More importantly, significant amount of syringol was produced. These indicate the capability of C - C cleavage of Ni/C that co uld be facilitated by certain functionality of alkali lignin that was induced from the acid precipitation, particularly SLRP. However, this is not within a scope of the present work and should be subject to future studies. 127 5.6 Characteristics of the Pretre ated Biomass and Enzymatic Hydrolysis Figure 5 . 4 . Images of raw birch biomass (A), and after pretreatment at 220 °C in ethanol with no catalyst ( B), and with Co/C catalyst (C). After partial removal and hydrogenolysis, the biomass could now be considered pretreated. The physical color before and after pretreatment is shown in Figure 5. 4. Comparing between before pretreatment ( Figure 5. 4A) and after pretreatment without catalysts ( Figure 5. 4B), it can be seen that biomass has become significantly darker, despite compositional analysis indicating that pretreatment without catalyst removes 42% of the lignin ( Figure 5. 5). This indicates that lignin was melted during the process and r edeposited on biomass surface, which is a common phenomenon observed previously in dilute acid pretreatment and is expected to retard cellulose accessibility. 355 With addition of Co/C catalyst it is clear that the color be comes lighter ( Figure 5. 4C), suggesting of more lignin removal. 128 Figure 5 . 5 . Percent recovery of glucan, xylan, and lignin after pretreatment at 200 (dotted line) and 220 °C in ethanol with different metal/C c atalysts (A). The correlation between total monomeric yields and lignin removal (100 lignin recovery) (B). A summary of percent recovery of each main biomass component after pretreatment is shown in Figure 5. 5A. Noted that untreated birch was characteri zed with 41 % glucan, 20% xylan, and 20 % lignin. Overall, it was shown that cellulose was successfully reserved even after the reaction at 220 °C. However, hemicellulose was partially lost particularly at the higher temperature. Catalysts seem to have some effect on the loss of hemicellulose, most likely through its hydrogenolysis. Hydrogenolysis of hemicellulose at similar conditions was previously reported to result in soluble sugars without further hydrogenations , 37,226 which should theoretically be recoverable from the liquid fraction. More noticeably, for lignin recovery there seems to be an inverse - shaped volcano plot, indicating that lignin removal could be correlated to monomer yields. The correlation is shown in Fig ure 5. 5B, where it can be seen that the relationship is not entirely linear as it seems to flatten out in for higher monomer yields. This could be due to couple of reasons. First, the maximum amount of lignin removal was reached. Second, the dimers and oth er ethanol soluble products from some catalysts were not taken into consideration. These products 129 are not likely to redeposit on biomass, and hence higher lignin removal even though monomer yields were not as high. Figure 5 . 6 . Enzymatic hydrolysis yields of the biomass after the reaction at 200 (dotted line) and 220 °C in ethanol with different metal/C catalysts (A). The sugar yields are based on the untreated biomass compositions. The correlation between lig nin removal and glucose yields (B). Lastly, effects from pretreatment on enzymatic hydrogenolysis is evaluated. Noticeably, there is also a volcano - plot like for the glucose yields (except from Cu to Zn , Figure 5.6 ). This reflects the effect from lignin a nd hemicellulose removals on hydrolysis yields. Without any catalyst, the glucose yields were only 31%, whereas the glucose yields >75% could be obtained from reactions that added Co/C, Ni/C, Ru/C, and Pd/C using a temperature of 220 °C. Noted that the sli ghtly higher glucose yields from Pd and Ru catalysts was because they preserved slightly higher amount of glucose. For xylose yield, it does not exhibit a strong trend because those that could yield higher xylose, such as Co/C, also has higher loss of xyla n after pretreatment. Overall, this indicates that adding catalys ts, such as Co, Ni, Ru, and Pd catalysts did not only result in high monomers from lignin hydrogenolysis, but also significantly promote accessibility of enzymes to cellulose and hemicellulos e. 130 5.7 Conclusion The present study has conducted a comparison of lignin catalytic transfer hydrogenolysis from ethanol using different supported metal catalysts with the goal of demonstrating their effects on both aromatic monomer yields and pretreatmen t efficacy. Selecting catalysts that were previously used for study of ethanol reforming, we found that in term of total monomer yields at 220 °C, Co/C gave the highest amounts (52%) followed by Ni/C, Ru/C, and Pd/C. All of these catalysts also exhibited h igh selectivity, particularly for Co/C and Ni/C. Interestingly, Co/C is an promising candidate that has not been used for this purpose previously. It was also affirmed that hydrogenolysis at this condition relies heavily on the quality of the substrate, i.e. - O - 4 content, as evidenced by significantly lower yields (~7%) when using alkali. Apart from lignin valorization, the reaction is also a pretreatment in itself. It was shown that after the reaction, cellulose was essentially preserved in all cases, whereas hemicellulose was lost. The loss tends to be higher with catalysts that generate higher monomer yields. This is similar to lignin, although more lignin is removed than hemicellulose (down to as low as 20% of the original). The evaluation of its eff icacy as a pretreatment was determined by sugar yields from enzymatic hydrolysis. Similarly, it was found that catalysts that resulted in high monomer yields tend to also result in more glucose yields. This was likely because of higher lignin removal from those catalysts. Therefore, through comparison of different catalysts, our results suggest that addition of catalysts have a large impact not on just monomer yields, but also pretreatment aspect of the biomass. Therefore, selecting a proper catalyst for li gnin - first approach is vital and further development, such as alloying or support modification, is needed for a robust and cost - effective catalyst. 131 CHAPTER 6 Summary and Outlook In the present thesis, a complete application of DFT has been used to get adhesion va lues, fundamental understanding of the solid/solid interface, predictive model s , and recommendations for experiments. oxides, CuO has a strong interaction with Ag, and is alm ost 5 times stronger than that of YSZ. Moreover, during the brazing of the Ag in air, diffused oxygen in Ag also partially promote s the interfacial interactions. With these, the Ag - CuO system was able to wet and adhere strongly to the YSZ surface. The stro ng Ag/CuO adhesion was identified to be due to the CuO chemical properties and uni que surface structure. Using this knowledge, a descriptor was formulated based on a combination of these factors and validated with a number of difference oxides. Using this descriptor which allows quicker estimation of Ag/oxides adhesion for materials search, new multivalent oxides that will be stable in SOFC operati ng conditions while also providing strong adhesion with Ag were identified, such as CuAlO 2 and Cu 3 TiO 4 . The des criptor can have wider impacts for materials search in Ag/oxide interfaces, such as thin film growth or coating applications. Due to the complexity of the problem, in the liquid/solid interface, DFT was only employed to draw insight from a solvated lignin dimer on catalyst surfaces, and not for development of screening models. To reduce the complexity of lignin and render it small enough for DFT simulations, spectroscopic and wet chemistry methods showed that despite the variability and complexity of lignin polymer, - O - 4 linkage emerges as the primary functional group that determines the amount of monomer yields for hydrogenolysis . A model was also developed and validated to show that using this functionality alone was sufficient to predict theoretical maxi mum monomer yields. Therefore, a dimer containing just this linkage was used for DFT simulation. The 132 simulation of a - O - 4 dimer adsorption on Ni(111) and Cu(111) surface shows that, in a vacuum, the dimer prefers to adsorb on the metal surface s with both aromatic rings parallel to the surface s , where Ni(111) shows significantly higher adsorption energy. Upon solvation with an explicit model that represent s ethanol with density at 50 °C, it was found that the lignin dimer still adsorbs on the Ni(111) surfac e with a similar conformation. However, it no longer adsorbs on the Cu(111) surface. This implies that Cu(111) may not be a n effective catalyst, particularly at lower temperatures. It should be n oted that representing solvation using implicit models could not describe the solvent - lignin interactions accurately. To circumvent the large liquid/ solid interface calculations, a model based on a thermodynamic cycle was developed for prediction of dimer adsorption energy for a given solvent and catalyst. Although the model is not suitable for large screening, it facilitates understanding for solvent selection by providing quantification to the solvent effect . Performing catalytic hydrogen transfer reactions on lignin shows that using Ni - based catalysts did result i n much larger yields than a Cu - based catalyst. This is partly due to low lignin - Cu interactions; however, it is also contributed by other factors, and among those is lower amount of hydrogen produced from ethanol in Cu(111). Figure 6 . 1 . Lignin dimer adsorption on Pd (111) with 0 and 0.33 ML hydrogen coverage. Although the present thesis has shed light the effect of solvent at the catalyst surface for lignin reactions, there remains many other pheno mena at the interface that need to be explored in 133 order to fully comprehend the reaction and for catalyst design. Naturally , one of those is the solvent effect beyond adsorption, i.e. on the reaction pathways. Stabilization of the intermediates from different solvents may fav or different reaction paths and different energy barriers. Figure 6 . 2 . Ab intio shows alcohol decomposition leads to H and ethoxy adsorbing species on Ni(111). Other phenomena that would affect lignin hydrogenolysis are the effect s from other adsorb ing species. For example, one of those is the effect of hydrogen coverage. As we have briefly discussed in Chapter 5 , different hydrogen pressure could result in different products. Preliminary DFT calculations support the hypothesis by showing that on the Pd(111) surface, a hydrogen coverage of 0.33 ML on the surface leads to much weaker adsorption as shown by the distance between the dimer and surface shown in Figure 6.2. As many liquid and ga s phase experiments used excessive H 2 gas, study on this topic may reveal that having higher H 2 does not necessarily always lead to higher catalytic activity. More importantly, it would provide valuable insight on how hydrogen coverage affect s reaction ene rgy barriers as well as reaction pathway s . The other type of surface coverage is that, instead of pressurizing with H 2 gas, it could be covered with other intermediates from solvents , such as when hydrogen is obtained through 134 catalytic hydrogen transfer fr om ethanol. This re presents challenging simulations as it involves two concurrent reactions ethanol decomposition and lignin hydrogenolysis. As detected in the preliminary ab intio molecular dynamics shown in Figure 6.1, ethanol decomposed to an ethoxy a nd hydrogen adatoms. While the hydrogen will be used for hydrogenolysis and its coverage is expected to be low, the ethoxy adatom could affect the reaction rate depending on its interaction with the catalyst . These topics will make an interesting subject for future studies as the understanding could have a wider impact on more than just lignin hydrogenolysis, particularly on other reductive reactions using different hydrogen sour ces and pressures. 135 APPENDICES 136 APPENDIX A: Supplementary Materials for Chapter 2 Table A 1 . List of al l the studied metals and oxides, including their crystal structures, lattice constants, surface energies from the present work, along with comparisons to experiment values, and the calculated values for the descriptor ( ) as shown in Eqn. 2. 4. All computational values are based on LDA functional unless otherwise stated. Structural Info (ev) Metallic Fm - 3m (111) Ag 4.1474 PBE 4.0857 0.77 PBE 0.75 356 , PBE 1.24 182 4.0035 1.10 1.15 357 Ag( l ) 0.77 ±0.01 PBE 0.93 358 Ag( l ) with O at surface 0 .67 ±0.01 PBE Ni 3.5161 PBE 3.5240 1.94 PBE 2.02 359 , PBE 2.37 358 3.4230 2.68 2.56 360 Corundum R - 3c (0001) M - Al 2 O 3 4.7333, 4.7533, 12.9177 4.7590, 4.7590, 12.9910 2.08 2.13 361 1.69 362 - 17.42 6, 3 - 1.45 19.40 0.855 - Cr 2 O 3 4.9388, 4.9930, 25.9979 4.9589, 4.9589, 13.5931 2.04 1.60 363 , PBE 1.60 364 - 14.27 6, 3 - 1.19 21.43 0.627 - Fe 2 O 3 5.0083, 5.0083, 13.5404 5.0350, 5.0350, 13.7200 1.52 1.60 365 , 1.90 366 - 9.81 3, 3 - 0.82 21.73 0.094 In 2 O 3 5.5042, 5.5042, 14.5231 5.4870, 5.4870, 14.5100 1.58 - 9.78 6, 3 - 0.81 26.24 0.341 Rh 2 O 3 5.0524, 5.0524, 13.6446 5.0529, 5.0529, 13.6400 2.11 1.52 367 , PBE - 6.53 6, 3 - 054 22.11 0.258 Rock salt Fm - 3m {100} BaO 5.4803 5.5230 0.46 0.53 368 - 5.64 2, 1 - 0.94 15.02 0.116 CaO 4.7203 4.8105 0.85 0.88 368 1.31 369 - 6.78 2, 1 - 1.13 11.14 0.190 CdO 4.6546 4.6953 0.82 - 2.57 2, 1 - 0.43 11.01 0.065 MgO 4.1457 4.2112 1.26 1.20 368 1.20 370 - 6.40 2, 1 - 1.07 8.50 0.235 NiO 4.0600 4.1684 1.28 1.14 371 1.38 364 - 3.48 2, 1 - 0.58 8.24 0.124 Fluorite Fm - 3m (111) O c - CeO 2 5.4151 5.4110 1.01 1.00 372 1.20 373 - 11.54 2, 1 - 1.44 12.70 0.216 c - ZrO 2 5.1175 PBE 5.0700 0.81 PBE 0.79 374 , PBE - 19.09 2, 1 - 2.39 10.96 0.423 137 5.0320 1.18 c - YSZ 7.2373, 7.2373, 15.3526 0.70 PBE 1.04 375 , 1.23 94 - 421.77 8, 4 - 2.29 43.86 0.405 7.1163, 7.1163, 15.0960 1.42 Antifluorite Fm - 3m (111) M Rb 2 O 6.5600 6.7400 0.25 - 3.72 2, 0 - 0.46 18.62 0.025 Tenorite C2/c (111) CuO 4.6939, 3.4318, 5.1426 PBE 4.6530, 3.4100, 5.1080 0.72 PBE 0. 74 148 , PBE - 1.94 2, 1 - 0.65 17.52 0.066 4.5359, 3.3422, 5.0054 1.16 Tenorite C2/c (111) Cu 2 O 4.2772 4.2696 0.65 PBE 0.67 376 , PBE - 2.06 2, 1 - 0.51 29.50 0.030 4.1271 0.94 Zincite P6 3 mc (0 0) ZnO 3.2044, 3.2044, 5.1373 3.2493, 3.2493, 5.2054 1.10 - 3.45 2, 1 - 0.86 28.51 0.097 Rutile P4 2 /mnm (110) SnO 2 4.7309, 4.7309, 3.1958 4.7373, 4.7373, 3.1864 1.46 1.29 377 - 6.33 3,2 - 1.05 21.38 0.135 TiO 2 4.5716, 4.5716, 2.9269 4.5940, 4.5940, 2.9590 0.97 0.90 378 - 10.03 3,2 - 1.67 18.92 0.250 Anatase I4 1 /amd (101) TiO 2 3.7652, 3.7652, 9.4766 3.7760, 3.7760, 9.4860 0.82 0.52 379 , PBE - 8.62 4, 2 - 1.44 19.27 0.284 Delafossite P 63/mmc (for CuAlO 2 ) or R - 3m (0001) O CuAlO 2 2.8218, 2.8218, 11.1750 2.858, 2.858, 11.293 380 3.02 3.47 187 , HF - 9.69 1, 1 - 1.21 6.88 0.156 CuCrO 2 2.9393, 2.9393, 16.7759 2.9734, 2.9734, 11.400 381 1.40 - 10.11 1, 1 - 1.26 7.48 0.150 CuFeO 2 2.7498, 2.7498, 17.1314 3.035, 3.035, 11.449 382 1.08 - 8.51 1, 1 - 1.06 6.54 0.141 138 * lattice constants were taken from the default structures in Mate rial Studio unless cited, PBE values from PBE functionals, values from GGA with full - potential linearized augmented plane wave, HF values from Hatree - Fock, lattice constants were supercell of ZrO 2 and substituted with Y to obtain 9% - doped yttria, 12.5% - doped yttria, is defined by , where is the energy of the oxide cell per formula, / is the numbe r of metal/oxygen in the oxide structure, is the energy of metal per atom taken from its bulk structure, and is the energy of atomic oxygen tak en from half of the oxygen gas. A.1 Surface Energy Calculation The surface energies for stoichiometric symmetric slabs were calculated from the formula, , where is the surface energy, is the calculated energy of the slab, is the molecular ratio between the slab to the bulk formula , is energy of the bulk str ucture, and A is the surface area of the slab. For stoichiometric asymmetric slabs, including in the delafossite and Cu 3 TiO 4 , the surface energies were estimated based on an average between both surface terminations as follow, CuGaO 2 2.9510, 2.9510, 17.0000 3.011, 3.011, 11.441 383 2.82 - 6.76 1, 1 - 0.85 7.54 0.094 CuInO 2 3.3052, 3.3052, 17.1665 3.3831, 3.3831, 11.551 383 2.54 - 5.70 1, 1 - 0.71 9.46 0.061 Cu 3 Ti O 4 P2 1 /c (010) O Cu 3 Ti O 4 2.9188, 11.1783, 6.6167 3.090, 5.272, 11.552 384 1.76 - 13.12 2, 2 - 0.82 1 5.86 0. 0 86 139 , where and are the energy of the non - stoichiometric symmetric slabs of the two terminations present in the stoichiometric slab, and is the molecular ratio between the combined slabs and the bulk formula. Figure A 1 . The relaxed bulk structure of CuAlO 2 (a) and its (0001) O surface (b). The relaxed bulk structure of TiCu 3 O 4 (c) and its (010) O surface (d). A.2 Energetics and Orientations of the Individual Surfaces In the present study, two metal s and six types of oxide structure were investigated. The surface energies of these components are listed in Table A 1 with some obtained by both PBE and LDA functionals for benchmarking purposes throughout the study. For individual surface energies, it was found that results from LDA functional, as obtained from metals, ZrO 2 , and CuO, conformed better with experiments. Similar preference for LDA was also reported in a recent study where PAW potentials and GGA functional, which is typically regarded as a mor e superior one, were used for calculation of work functions, but particularly for surface energy, LDA was chosen for the reason that it yielded results closer to experiments . 357 Therefore, most of the results presented 140 in this study, including Table 2. 1, will be based on LDA functional. Nonetheless, we expected that the trends of surface energy among oxides and metals will be the same for both functionals. Hereafter, mentions of metals/oxides surface in the present study are referred to the phase and su rface presented in Table A 1 unless otherwise stated. For oxides of corundum structure, the stoichiometric surface (0001) with a single metal atom per unit cell at the surface termination was chosen for this study ( Figure A 5 ). It should be noted that in pra ctice the surfaces of many corundum oxides could coexist in two forms, (0001) and ( ) due to extremely close surface energies . 367,385,386 Previous studies have reported that for (0001) surface of Al 2 O 3 , Cr 2 O 3 , Fe 2 O 3 , and Rh 2 O 3 , the single metal atom termination was the most stable over a wide range of oxygen chemical potential . 367,387 389 No information was found for corundum In 2 O 3 , but it was expected to be the same. The (0001) surface used in this study was rel axed with more stringent criteria (force < 0.015 eV and broken symmetry) because of potential present study exhibited substantial change, wherein the top metal atom exhibited an inward re laxation toward the bulk ( Figure A 5 ). In relative to the bulk structure the position moved inward by 76, 58, 97, and 45% for Al 2 O 3 , Cr 2 O 3 , In 2 O 3 , Rh 2 O 3 . Previous study on Al 2 O 3 using GGA functional and experiments also reported ca . 80 % inward movement of an Al atom. 390 392 The inward movement of the top metallic atom was likely in order to minimize the energy from a numbers of dangling bonds at the metal atom. To an even greater extent, the inward movement of the surface Fe in Fe 2 O 3 was 163%, meaning it subm erged into the sublayer ( Figure A 5 d). This inward movement of the atom for Fe 2 O 3 was tested in both ferromagnetic and antiferromagnetic state, both of which yielded the same configuration. These changes were expected to affect ad hesion properties, particularly in case of Fe 2 O 3 . The drastic change in the position of the metal 141 atom in Fe 2 O 3 and In 2 O 3 could contribute to their lower surface energy in relative to most other corundum oxides in Table A 1. Within fluorite structure, the s urface energy of cubic ZrO 2 was obtained by both LDA and PBE functionals as it represented a YSZ surface, which is one of the main components in SOFC - related interface that will be investigated in more detail. The results indicated that using PAW potential with PBE functional, like in cases of CuO and metals, tends to underestimate the surface energy. It should be noted that the selection of (111) surface was not only supported by it being the lowest surface energy, but also adhesion of YSZ with metal inter faces were shown to occur on this surface. 91 - of - a - kind due to its unpaired d electron inducing Cu magnetic moments and a cubic - to - monoclinic structural distortion . 393 surface was reportedly to be (111). 148 electronic structure, and catalysis, employed lower U eff of 4 - 7 eV with regards to PBE functionals . 148,149,394,395 tice constants was significant when U eff was 7 eV. Since lattice constants would be more important given the context of this study, we employed a lesser used U eff of 8.50 for Cu. The discrepancy between our results and previous computational studies is lik whereas previous studies used 400 eV or less. Additionally, similar reasonable lattice constants were achieved if we used the cutoff energy of 400 eV. For cuprite structure, previous computationa l study reported that (111) surface exhibited the lowest energy, with a caveat that symmetry must be broken during surface relaxation to facilitate its reconstruction, and hence, lower surface energy. 376,396 As the study was performed in 142 PBE functional, Table 2. 1 also includes results from PBE functional for Cu 2 O to ensure the present computational setup enabled us to achieve the lowest surface energy possible. For zincite structure, it was reported to mostly exposed (2 0) and (0 0) surfaces. 397 However, computationally, our calculated surface energy showed slightly higher surfac e energy of the (2 0) surface (1.21 J/m 2 ) than the (0 0) surface (1.10 J/m 2 ). Therefore, the prediction was based on (0 0) surface, although it should be noted that the W adh from the prediction based on (2 rutile phase, many studies have demonstrated that the (110) surface is the most stable ones for both SnO 2 and TiO 2 . 377,378 In case of TiO 2 , the other morphology that was investigated was that of anatase structure. Anatase TiO 2 was also previously sh own computationally to have the lowest surface energy being (101), which was also one of the most exposed surface s on its synthesized thin films. 398,399 Lastly, in delafossite structure there was limited literature on their surfaces. However, detailed investigation was conducted on surfaces of CuAlO 2 . By using XPS, studies suggested that t hin film of CuAlO 2 on alumina exposed almost solely (0001) surface 185,186 . The surface was also shown co mputationally to be O - terminated. 187 As the struct ure of Cu 3 TiO 4 is similar to delafossite oxides, and its (101) is equivalent to the (0001) surface of delafossite. Therefore, the (101) surface of Cu 3 TiO 4 was used in this study. 143 A.3 AIMD S imulation of the L iquid Ag/YSZ I nterface Figure A 2 . Pair radial distribution function of Ag - Ag in liquid Ag with 2 O atoms dissolved showing the nearest Ag - Ag distance of 2.8 Å, which is the same to the experimental value of pure silver measured in reducing atmosphere. A.4 Comparison to P revious C omputational S tudy of Ni/ZrO 2 The importance of using c - YSZ instead of c - ZrO 2 due to stability concern s was highlighted when compari ng the with prior computational studies. Previous DFT studies of the Ni(111) / ZrO 2 (111) interface using the PBE functional with an ultrasoft pseudopotential indicated a of 1.00 J/m 2 . 400 Other studies using c - ZrO 2 also reported similar values at 1.12 and 1.28 J/m 2 when using the PBE functional with norm - conserving pseudopotential a nd linear combination of atomic orbital (LCAO) DFT . 105,401 All of these values are larger than the results of 0.49 J/m 2 from GGA - PBE functional in the present study on Ni(111)/YSZ interface . This is likely due to the phase instability of cubic ZrO 2 . P ure zirconia exist s in a monoclinic phase at low temperature, tetragonal at 927 ° C , and cubic at 2 377 ° C until its melting point . 402,403 To stabilize the cubic phase, 8% - 40% yttria is necessary. 153 C alculations on Ni(111) / c - ZrO 2 (111) interface calculations were performed with both GGA - PBE and LDA functionals. Up on relaxation, the cubic to tetragonal phase transition started from the interface region and propagated through the ZrO 2 slab. This led to an energy reduction of 144 the slab by 1.07 and 0.22 eV (54 and 11 meV per ZrO 2 formula unit ) for the GGA - PBE and LDA functional s , respectively. However, the cubic ZrO 2 slab for surface energy calculatio ns remained meta - stable cubic phase after minimization. Thus, a ccording to Eqn. 2. 1, the energy reduction of the interface slabs due to phase transformation would proportionally reduce the interface cell energy ( ), and consequently falsely increase the calculated . This might be the reason for the high at Ni(111) /c - ZrO 2 (111) in terface in previous DFT studies. 105,400,401 A.5 Formation of Metal - Oxygen Bonds at the Interface Figure A 3 . Projected density of states (PDOS) of interface atoms in Ni/YSZ, Ag/YSZ, and Ag/Cu O in their no - contact positions (slabs were 10 Å apart) and in their interface position as calculated by LDA functional. To obtain deeper insight into the nature of metal/oxide interface bonding, projected density of states (PDOS) calculations were generat ed for the interfaces shown in Figure A 3 . Firstly, before the slabs were put together to make the interfaces, PDOS calculations were performed on a no - contact configuration, where the metal and oxide slabs were separated by 10 Å. This serves as a 145 reference point as well as providing initial characteristics of individual slabs. Comparing the Figure A 4, the range of energies of d - orbitals in Ni was wider than Ag and there a higher density of incompletely filled states evidenced by spilling acr oss the Fermi level. This suggests an easier distribution of charges in Ni, and hence more polarizable than Ag. In the YSZ slab, a band gap in the O of 2.88 eV was observed. However, for CuO, the band gap was not very clear in the no - contact PDOS. This min or error was likely due to the disturbance of the CuO antiferromagnetic spins by the presence of Ag in spite of the 10 Å vacuum in between. A 0.28 eV band gap was obtained in ( Figure A 4 ) by removing the Ag slab and keeping the CuO slab in the same cell dim ensions. Figure A 4 . Projected density of states (PDOS) of CuO(111) surface with stretched lattice parameters to match that of the metals in the interface. It exhibited a bandgap of 0.25 eV. Upon contact with the oxide surface, the noticeable feature of interface bonding observed in all three interfaces was the previously developed metal - induced gap states (MIGS) proposition, wherein the oxygen p - states spilled over into the band gap region ( F igure A 3 ) . 404 This stemmed from the charge transfer at the interface in order to reach electrochemical potential equilibrium for electrons. Although the charge transfer, presumably from metals to the oxide oxygen atoms, would theoretically increase the energy o f the oxide slabs, but the energy of the overall system was 146 reduced as a results of the charge transfer. This was manifested in the shift of the PDOS for both the metal and oxide toward lower energy states ( Figure A 4). Principally, the strength of the bon ds at the interface was attributed mostly by the appearance of bonding and anti - bonding states resulting from the coupling of the metal d - states and the oxygen p - states in the oxide, which are located below the oxide valence bands and above the metal d - ban ds, respectively. 405 It should be noted that non - bonding states could appear and mingle i n between these two states. However, a definitive criterion separating these states cannot be drawn from PDOS diagrams due to complex deformation of dispersive bands. 406 Therefore, as an estimate, any occupied states whose energy was lowered corresponded to bond formation, whereas the opposite was observed for the anti - bonding. The shifts in density of states shown in Figure A 3 correspond well to the higher adhesion energy of Ni/YSZ than Ag/YSZ. In the Ni/YSZ interface, the formation of the bonding p - states below the metal d - states was noticeab le as the majority of the p states at interface oxygen atoms shifted to under - 5 eV, well below the oxygen valence bands in the no - contact states. The same phenomenon was observed in the Ag/YSZ interface, but to a much lesser extent. Lastly, for the Ag/CuO interface, the shift in p - states of the oxygen atoms was not as dramatic as those in YSZ structure. However, the trend was similar in that there is a lesser amount of the p - state in the anti - bonding region (~0 to - 3 eV) and a higher population in the lowe r energy range of - 4 to - 8 eV. Overall, the PDOS showed several indications of the interaction between interface metal atoms and the oxygen atoms of the oxides. 147 A.6 The I nterface E nergy and W ork of A dhesion D escriptor Figure A 5 . Surface structures showing the difference between the position of the terminated metallic atom of most corundum oxides (represented by Al 2 O 3 ) and the Fe 2 O 3 . These are (0001) surface of initial Al 2 O 3 slab (a), relaxed Al 2 O 3 slab (b), initial Fe 2 O 3 slab (c), relaxed Fe 2 O 3 slab (d), relaxed Ag(111)||Al 2 O 3 (0001) interface (e), and relaxed Ag(111)||Fe 2 O 3 (0001) interface (f). Figure A 6 . The correlation between W adh and the oxide formation energy of (0001) surface corundum type oxide and {001} rock salt type oxide. Note that Fe 2 O 3 was not included for the reason that its geometry was unlike other corundum oxides after surface relaxation. 148 APPENDIX B: Supplementary Materials for Chapter 3 Table B 1 . The Summary of quantified lignin properties and monomer yields . Fraction Klason Lignin - O - 4 - O - 4 - 5 - S/G S/G Mn Mw PhOH AlOH 13 CNMR HSQC HSQC HSQC HSQC Thio * RT - 5 57.80 15.23 58.08 3.49 38.43 1.8 1.73 1825 7161 0.82 1.59 RT - 4 69.46 20 58.27 4.98 36.75 1.8 1.70 2811 21558 0.73 1.27 RT - 3 86.61 21.91 59.56 4.77 35.67 1.84 1.73 3461 28066 0.71 1.26 RT - 2 68.25 21.92 59.88 5.48 34.64 1.86 1.76 3765 33923 0.71 1.27 RT - 1 86.45 22.88 64.62 5.72 29.66 1.86 1.76 4161 30883 0.68 1.27 SLRP - 2 78.50 28.66 67.67 8.53 23.79 1.95 1.92 6269 71216 0.52 1.25 SLRP - 1 70.74 31.25 73.02 7.94 19.04 1.99 1.95 6960 15400 0.47 1.25 Fraction Ni - catalyzed hydrogenolysis yields (%) Quantitative thio * yields (%) 4 - PG 4 - PS 4 - VG 4 - VS Sy Total S unit G unit H unit Total RT - 5 0.44 1.2 0 0 0 1.64 2.99 1.52 0.07 4.58 RT - 4 0.62 1.62 0 0 0.6 2.84 5.12 2.64 0.07 7.83 RT - 3 0.98 2.31 0 0 0 3.29 5.59 2.84 0.05 8.48 RT - 2 0 0 0 0 0 0 5.50 2.75 0.06 8.31 RT - 1 1.32 3.17 0 0 0 4.49 6.14 3.05 0.06 9.25 SLRP - 2 0.76 2.65 0.95 1.48 1.1 6.92 8.88 4.06 0.06 13.00 SLRP - 1 0.77 2.51 0.92 1.24 0.78 6.23 10.70 4.81 0.06 15.57 Fraction Cu - catalyzed oxidation yields (%) V S VA SA AceV AceS Total RT - 5 2.97 ± 0.05 1.23 ± 0.14 0.51 ± 0.08 0.08 ± 0.04 0.57 ± 0.04 0.00 ± 0.00 5.40 ± 0.18 RT - 4 4.63 ± 0.30 6.29 ± 0.26 0.78 ± 0.01 0.01 ± 0.19 0.70 ± 0.02 0.91 ± 0.05 13.41 ± 0.40 RT - 3 4.88 ± 0.44 6.83 ± 0.51 0.85 ± 0.12 0.12 ± 0.06 0.71 ± 0.06 1.04 ± 0.16 14.42 ± 0.73 RT - 2 4.87 ± 0.05 6.62 ± 0.13 0.84 ± 0.05 0.05 ± 0.06 0.75 ± 0.02 0.91 ± 0.11 14.11 ± 0.19 RT - 1 4.14 ± 0.35 6.34 ± 0.21 0.77 ± 0.03 0.03 ± 0.06 0.63 ± 0.02 0.81 ± 0.09 12.80 ± 0.42 SLRP - 2 5.58 ± 0.09 7.85 ± 1.06 0.68 ± 0.01 0.01 ± 0.06 0.63 ± 0.06 0.33 ± 0.04 15.21 ± 1.07 SLRP - 1 5.91 ± 0.08 7.76 ± 0.12 0.72 ± 0.03 0.03 ± 0.11 0.62 ± 0.01 0.44 ± 0.08 15.57 ± 0.20 * thio = thioacidolysis - PG = 4 - propylguaiacol, 4 - PS = 4 - propylsyringol, 4 - VG = 4 - vinylguaiacol, 4 - SG = 4 - vinylsyringol, Sy = Syringol V = Vanillin, S = Syringaldehyde, VA = Vanillic Acid, SA = Syringic Acid, AceV = Acetovanillone, AceS = Acetosyringone 149 Figure B 1 . Trends of phenolic hydroxyl contents (black) in fractionated lignin showing lower values in fractions precipi tated at higher pH, while aliphatic hydroxyl content (red) do not show any specific trend. Figure B 2 . Trends of S/G ratios in fractionated lignin by HSQC NMR (red) and thioacidolysis (black) showing higher S/G values in fractions precipitated at higher pH and SLRP sam ples Figure B 3 . - O - 4 content (black) as measured by 13 C - NMR and molecular weight among fractionated lignin showing similar increasing trend in fractions precipitated at higher pH and SLRP samples. 150 Figure B 4 . K inetics of Cu - catalyzed oxidat ion reactions of lignin sample SLRP - 2 showing time for optimal yield at the study condition is at 7 minutes . 151 APPENDIX C: Supplementary Materials for Chapter 4 The isothermal isobaric ensemble (NPT) molecular dynamics was performed using COMPASS II force field with types similar to that described in the main manuscript. For each compound, 10 molecules of it was randomly packed in a cell with a volume of 4000 Å 3 . The cell was then relaxed and underwent NPT MD with 1000 ps equilibration time and 1000 ps production time, from which the cell volume was averaged over. Temperature coupling was performed with the Nose algorithm (Q = 0.01) and Berendsen barostat was employed w ith a decay constant of 0.1 ps. Table C 1 . Comparison between density calculated from NPT dynamics using COMPASS II force field and experimental density of compounds similar to the dimer. MD (g/cm 3 ) Exp. (g/cm 3 ) Error (%) Benzoic acid 1.23 1.27 - 3.15 1,3 Methoxy benzene 1.06 1.01 4.95 Phenyl Ethanol 0.99 1.02 - 2.94 Guaiacol 1.19 1.11 7.21 Glucose 1.45 1.54 - 5.84 guaiacylglycerol - beta - guaiacyl ether 1.27 - - Figure C 1 . The change in fraction of surface accessible surface area (SASA) of the dimer in solvated adsorb in reference to fully solvated state ( ) as estimated by SASA as a function of probe radius. 152 BIBLIOGRAPHY 153 BIBLIOGRAPHY 1. Generation of renewable energy in the U.S. from 2008 to 2017, by technology (in terawatt hours). (Bloomberg New Energy Finance. n.d.). 2. Ellabban, O., Abu - Rub, H. & Blaabjerg, F. Renewable energy resources: Current status, future prospects and their enabling technology. Renew. Sustain. Energy Rev. 39, 748 764 (2014). 3. Wachsman, E. D. & Lee, K. T. Lowering the temperature of solid oxide fu el cells. Science 334, 935 939 (2011). 4. Wachsman, E. D., Marlowe, C. A. & Lee, K. T. Role of solid oxide fuel cells in a balanced energy strategy. Energy Environ. Sci. 5, 5498 5509 (2012). 5. Schutyser, W. et al. Chemicals from lignin: an interplay of li gnocellulose fractionation, depolymerisation, and upgrading. Chem. Soc. Rev. 47, 852 908 (2018). 6. Voelcker, J. Nissan takes a different approach to fuel cells: ethanol. Green Car Reports Available at: https://www.greencarreports.com/news/1104467_nissan - t akes - a - different - approach - to - fuel - cells - ethanol. (Accessed: 10th September 2018) 7. - oxide fuel cell e - NV200. Available at: https://insideevs.com/nissan - unveils - worlds - first - solid - oxide - fuel - cell - vehicle - e - nv200 - sofc/. (Ac cessed: 19th September 2018) 8. Nissan SOFC powered vehicle system runs on bioethanol. Fuel Cells Bull. 2016, 2 3 (2016). 9. Ahamer, C., Opitz, A. K., Rupp, G. M. & Fleig, J. Revisiting the temperature dependent ionic conductivity of yttria stabilized zirc onia (YSZ). J. Electrochem. Soc. 164, F790 F803 (2017). 10. Asadikiya, M. & Zhong, Y. Oxygen ion mobility and conductivity prediction in cubic yttria - stabilized zirconia single crystals. J. Mater. Sci. 53, 1699 1709 (2018). 11. Mahato, N., Banerjee, A., Gu pta, A., Omar, S. & Balani, K. Progress in material selection for solid oxide fuel cell technology: A review. Prog. Mater. Sci. 72, 141 337 (2015). 12. Solid Oxide Fuel Cells (SOFCs). Forschungszentrum Jülich 13. Heather L. Trajano & Charles E. Wyman. Fund amentals o f biomass pretreatment at low pH . in A queous pretreatment of plant biomass for biological and chemical conversion to fuels and chemicals (John Wiley & Sons Ltd). 14 . D. Kärkäs, M., S. Matsuura, B., M. Monos, T., Magallanes, G. & J. Stephenson, C. R. Transition - metal catalyzed valorization of lignin: the key to a sustainable carbon - neutral future. Org. Biomol. Chem. 14, 1853 1914 (2016). 154 15 . Ammal, S. C. & Heyden, A. Co mbined DFT and microkinetic modeling study of hydrogen oxidation at the Ni/YSZ anode of solid oxide fuel cells. J. Phys. Chem. Lett. 3, 2767 2772 (2012). 16 . Cui, J. et al. Conversion of xylose to furfuryl alcohol and 2 - methylfuran in a continuous fixed - be d reactor. ChemSusChem 9, 1259 1262 (2016). 17 . Zhang, L., Xi, G., Yu, K., Yu, H. & Wang, X. Furfural production from biomass derived carbohydrates and lignocellulosic residues via heterogeneous acid catalysts. Ind. Crops Prod. 98, 68 75 (2017). 18 . Cui, J . et al. Conversion of carbohydrates to furfural via selective cleavage of the carbon carbon bond: the cooperative effects of zeolite and solvent. Green Chem. 18, 1619 1624 (2016). 19 . Li, C., Zhao, X., Wang, A., Huber, G. W. & Zhang, T. Catalytic transfor mation of lignin for the production of chemicals and fuels. Chem. Rev. 115, 11559 11624 (2015). 20. DeRose, A. J. et al. Inhibitor for prevention of braze migration in solid oxide fuel cells. (2010). 2 1 . Smithells, C., Gale, W. & Totemeier, T. Smithells me tals reference book. ASM Mater. Inf. Soc. (2004). 22 . Kim, J. Y., Hardy, J. S. & Scott Weil, K. Effects of CuO Content on the wetting behavior and mechanical properties of a Ag CuO braze for ceramic joining. J. Am. Ceram. Soc. 88, 2521 2527 (2005). 23 . Nik olopoulos, P. & Sotiropoulou, D. Wettability between zirconia ceramics and the liquid metals copper, nickel and cobalt. J. Mater. Sci. Lett. 6, 1429 1430 (1987). 24 . Hardy, J. S., Kim, J. Y. & Weil, K. S. Joining mixed conducting oxides using an air - fired electrically conductive braze. J. Electrochem. Soc. 151, J43 J49 (2004). 25 . Weil, K. S., Kim, J. Y. & Hardy, J. S. Reactive air brazing: A novel method of sealing sofcs and other solid - state electrochemical devices. Electrochem. Solid - State Lett. 8, A133 A136 (2005). 26 . Phongpreecha, T., Nicholas, J. D., Bieler, T. R. & Qi, Y. Computational design of metal oxides to enhance the wetting and adhesion of silver - based brazes on yttria - stabilized - zirconia. Acta Mater. 152, 229 238 (2018). 27 . Adler, E. L ignin chemistry past, present and future. Wood Sci. Technol. 11, 169 218 (1977). 28 . Huang, X., Korányi, T. I., Boot, M. D. & Hensen, E. J. M. Catalytic depolymerization of lignin in supercritical ethanol. ChemSusChem 7, 2276 2288 (2014). 155 29. Bouxin, F. P. et al. Catalytic depolymerisation of isolated lignins to fine chemicals using a Pt/alumina catalyst: part 1 impact of the lignin structure. Green Chem. 17, 1235 1242 (2015). 30 . Rahimi, A., Ulbrich, A., Coon, J. J. & Stahl, S. S. Formic - acid - induced depolymerization of oxidized lignin to aromatics. Nature 515, 249 252 (2014). 31 . Lu, J., Wang, M., Zhang, X., Heyden, A. & Wang, F. Cleavage mechanism f or lignin model compounds over P d catalysts identified by combination of first - principles calculations and experiments. ACS Catal. 6, 5589 5598 (2016). 32. Si, X. et al. A strategy for generating high - quality cellulose and lignin simultaneously from woody biomass. Green Chem. 19, 4849 4 857 (2017). 33. Bosch, S. V. den et al. Integrating lignin valorization and bio - ethanol production: on the role of Ni - Al 2 O 3 catalyst pellets during lignin - first fractionation. Green Chem. 19, 3313 3326 (2017). 34. Ferrini, P., Chesi, C., Parkin, N. & Rinal di, R. Effect of methanol in controlling defunctionalization of the propyl side chain of phenolics from catalytic upstream biorefining. Faraday Discuss. 202, 403 413 (2017). 35. Klein, I., Marcum, C., Kenttämaa, H. & Abu - Omar, M. M. Mechanistic investigati on of the Zn/Pd/C catalyzed cleavage and hydrodeoxygenation of lignin. Green Chem. 18, 2399 2405 (2016). 36. Ye, Y., Zhang, Y., Fan, J. & Chang, J. Selective production of 4 - ethylphenolics from lignin via mild hydrogenolysis. Bioresour. Technol. 118, 648 6 51 (2012). 37. Parsell, T. et al. A synergistic biorefinery based on catalytic conversion of lignin prior to cellulose starting from lignocellulosic biomass. Green Chem. 17, 1492 1499 (2015). 38. Konnerth, H., Zhang, J., Ma, D., Prechtl, M. H. G. & Yan, N. Base promoted hydrogenolysis of lignin model compounds and organosolv lignin over metal catalysts in water. Chem. Eng. Sci. 123, 155 163 (2015). 39. Barta, K., Warner, G. R., Beach, E. S. & Anast as, P. T. Depolymerization of organosolv lignin to aromatic compounds over Cu - doped porous metal oxides. Green Chem. 16, 191 196 (2013). 40. Wang, X. & Rinaldi, R. Solvent effects on the hydroge nolysis of diphenyl ether with R aney nickel and their implicat ions for the conversion of lignin. ChemSusChem 5, 1455 1466 (2012). 41. Gasson, J. R. et al. Modeling the lignin degradation kinetics in an ethanol/formic acid solvolysis approach. Part 1. Kinetic model development. Ind. Eng. Chem. Res. 51, 10595 10606 (20 12). 156 42. Forchheim, D., Gasson, J. R., Hornung, U., Kruse, A. & Barth, T. Modeling the lignin degradation kinetics in a ethanol/formic acid solvolysis approach. Part 2. Validation and transfer to variable conditions. Ind. Eng. Chem. Res. 51, 15053 15063 (2 012). 43. Wanmolee, W. et al. Catalytic conversion of organosolv lignins to phenolic monomers in different organic solvents and effect of operating conditions on yield with methyl isobutyl ketone. ACS Sustain. Chem. Eng. 6, 3010 3018 (2018). 44. Li, Q. & L opez, N. Chirality, rigidity and conjugation: a first principles study on the key molecular aspects in lignin depolymerization on Ni - based catalysts. ACS Catal. 8 , 4230 4240 (2018). 45. Horstemeyer, M. F. Multiscale modeling: A review. in Practical aspects of computational chemistry: Methods, concepts and applications (eds. Leszczynski, J. & Shukla, M. K.) 87 135 (Springer Netherlands, 2010). doi:10.1007/978 - 90 - 481 - 2687 - 3_4 46. Jancar, J. et al. Current issues in research on structure property relationships in polymer nanocomposites. Polymer 51, 3321 3343 (2010). 47. Campbell, C. T. & Sellers, J. R. V. The entropies of adsorbed molecules. J. Am. Chem. Soc. 134, 18109 18115 (2012). 48. Carrasco, J., Liu, W., Michaelides, A. & Tkatchenko, A. Insight into the d escription of van der Waals forces for benzene adsorption on transition metal (111) surfaces. J. Chem. Phys. 140, 084704 (2014). 49. Liu, W. et al. Benzene adsorbed on metals: Concerted effect of covalency and van der Waals bonding. Phys. Rev. B 86, 245405 (2012). 50. Carey, S. J., Zhao, W. & Campbell, C. T. Energetics of adsorbed benzene on Ni(111) and Pt(111) by calorimetry. Surf. Sci. 676, 9 16 (2018). 51. Abild - Pedersen, F. et al. Scaling properties of adsorption energies for hydrogen - containing molecules on transition - metal surfaces. Phys. Rev. Lett. 99, 016105 (2007). 52. Jacobsen, C. J. H. et al. ammonia synthesis catalysts. J. Am. Chem. Soc. 123, 8404 8405 (2001). 53. Vojvodic , A. et al. Exploring the limits: A low - pressure, low - temperature Haber Bosch process. Chem. Phys. Lett. 598, 108 112 (2014). 54. Medford, A. J. et al. Activity and selectivity trends in synthesis gas conversion to higher alcohols. Top. Catal. 57, 135 142 (2013). 55. Andersson, M. P. et al. Toward computational screening in heterogeneous catalysis: Pareto - optimal methanation catalysts. J. Catal. 239, 501 506 (2006). 157 56. Greeley, J., Jaramillo, T. F., Bonde, J., Chorkendorff, I. & Nørskov, J. K. Computationa l high - throughput screening of electrocatalytic materials for hydrogen evolution. Nat. Mater. 5, 909 913 (2006). 57. Studt, F. et al. Identification of non - precious metal alloy catalysts for selective hydrogenation of acetylene. Science 320, 1320 1322 (200 8). 58. Ferrin, P. et al. Modeling ethanol decomposition on transition metals: A combined J. Am. Chem. Soc. 131, 5809 5815 (2009). 59. Montemore, M. M. & Medlin, J. W. Scaling relations between adsorption energies for computational screening and design of catalysts. Catal. Sci. Technol. 4, 3748 3761 (2014). 60. Nørskov, J. K., Bligaard, T., Rossmeisl, J. & Christensen, C. H. Towards the c omputational design of solid catalysts. Nat. Chem. 1, 37 46 (2009). 61. Nørskov, J. K., Studt, F., Abild - Pedersen, F. & Bligaard, T. Activity and selectivity maps. in Fundamental concepts in heterogeneous catalysis 97 113 (John Wiley & Sons, Inc, 2014). 62 . Bull, S. J. & Berasetegui, E. G. An overview of the potential of quantitative coating adhesion measurement by scratch testing. Tribol. Int. 39, 99 114 (2006). 63. Mittal, K. L. Adhesion measurement of thin films. active and passive electronic components (1976). doi:10.1155/APEC.3.21 64. Nahor, H., Meltzman, H. & Kaplan, W. D. Ni YSZ(111) solid solid interfacial energy. J. Mater. Sci. 49, 3943 3950 (2014). 65. Li, H. - T. et al. - Al 2 O 3 Interfaces. J. Am. Ceram. Soc. 94, s154 s159 66. Meltzman, H., Mordehai, D. & Kaplan, W. D. Solid solid interface reconstruction at equilibrated Ni Al 2 O 3 interfaces. Acta Mater. 60, 4359 4369 (2012). 67. Zhang, Z., Li, L. & Yang, J. C. Adhesion of Pt Nanoparticles Supported - Al 2 O 3 Single Crystal. J. Phys. Chem. C 117, 21407 21412 (2013). 68. Si - Qi, S., Shingo, T. & Masanori, K. Effect of the stoichiometry on the electronic st ructure - Al 2 O 3 (0001) interface: a first - principles investigation. Chin. Phys. B 1 7, 2655 (2008). 69. Rose, J. H., Smith, J. R. & Ferrante, J. Universal features of bonding in metals. Phys. Rev. B 28, 1835 1845 (1983). 70. Rose, J. H., Ferrante, J. & Smith, J. R. Universal binding energy curves for metals and bimetallic interfaces. Phys . Rev. Lett. 47, 675 678 (1981). 158 71. Banerjea, A. & Smith, J. R. Origins of the universal binding - energy relation. Phys. Rev. B 37, 6632 6645 (1988). 72. Rose, J. H., Smith, J. R., Guinea, F. & Ferrante, J. Universal features of the equation of state of me tals. Phys. Rev. B 29, 2963 2969 (1984). 73. Chabanet, O. et al. Predicting crack growth resistance of aluminium sheets. Comput. Mater. Sci. 26, 1 12 (2003). 74. Elzas, A. & Thijsse, B. Cohesive law describing crack growth at iron/precipitate interfaces. C omput. Mater. Sci. 134, 214 224 (2017). 75. Schlosser, H. & Ferrante, J. Universality relationships in condensed matter: Bulk modulus and sound velocity. Phys. Rev. B 37, 4351 4357 (1988). 76. Guinea, F., Rose, J. H., Smith, J. R. & Ferrante, J. Scaling re lations in the equation of state, thermal expansion, and melting of metals. Appl. Phys. Lett. 44, 53 55 (1984). 77. Li, Y. J., Qi, W. H., Huang, B. Y., Wang, M. P. & Xiong, S. Y. Modeling the melting temperature of metallic nanowires. Mod. Phys. Lett. B 24 , 2345 2356 (2010). 78. Shu, G. G., Xu, Q. & Wu, P. Study of wetting on chemically soften interfaces by using combined solution thermodynamics and DFT calculations: Forecasting effective softening elements. ACS Appl. Mater. Interfaces 7, 7576 7583 (2015). 79. Alfredsson, M., Richard, C. & Catlow, A. Predicting the metal growth mode and wetting of noble metals supported on c - ZrO 2 . Surf. Sci. 561, 43 56 (2004). 80. Malijevský, A. Does surface roughness amplify wetting? J. Chem. Phys. 141, 184703 (2014). 81. Durov, A. V., Naidich, Y. V. & Kostyuk, B. D. Investigation of interaction of metal melts and zirconia. J. Mater. Sci. 40, 2173 2178 (2005). 82. Peereboom, L., Koenigsknecht, B., Hunter, M., Jackson, J. E. & Miller, D. J. Aqueous - phase adsorption of glycer ol and propylene glycol onto activated carbon. Carbon 45, 579 586 (2007). 83. Zuo, Z., Huang, W., Han, P. & Li, Z. Solvent effects for CO and H 2 adsorption on Cu 2 O (111) surface: A density functional theory study. Appl. Surf. Sci. 256, 2357 2362 (2010). 84 . Zhang, R., Ling, L., Li, Z. & Wang, B. Solvent effects on Cu 2 O(111) surface properties and CO adsorption on Cu 2 O(111) surface: A DFT study. Appl. Catal. Gen. 400, 142 147 (2011). 85. Garcia - Ratés, M. & López, N. Multigrid - based methodology for implicit s olvation models in periodic DFT. J. Chem. Theory Comput. 12, 1331 1341 (2016). 159 86. Steinmann, S. N., Sautet, P. & Michel, C. Solvation free energies for periodic surfaces: comparison of implicit and explicit solvation models. Phys. Chem. Chem. Phys. 18, 31850 31861 (2016). 87. Nikolopoulos, P., Ondracek, G. & Sotiropoulou, D. Wettability and interfacial energies between zirconia ceramic and liquid metals. Ceram. Int. 15, 201 206 (1989). 88. Laosiripojana, N. & Assabumrungrat, S. Catalytic steam reforming of methane, methanol, and ethanol over Ni/YSZ: The possible use of these fuels in internal reforming SOFC. J. Power Sources 163, 943 951 (2007). 89. Khor, K. A., Dong, Z. L. & Gu, Y. W. Influence of oxide mixtures on mechanical properties of plasma spraye d functionally graded coating. Thin Solid Films 368, 86 92 (2000). 90. Hong, H. S., Chae, U. - S., Choo, S. - T. & Lee, K. S. Microstructure and electrical conductivity of Ni/YSZ and NiO/YSZ composites for high - temperature electrolysis prepared by mechanical a lloying. J. Power Sources 149, 84 89 (2005). 91. Sasaki, T. et al. Atomic and electronic structure s of Ni/YSZ(111) interface. Mater. Trans. 45, 2137 2143 (2004). 92. Winterbottom, W. L. Equilibrium shape of a small particle in contact with a foreign substr ate. Acta Metall. 15, 303 310 (1967). 93. Meltzman, H., Chatain, D., Avizemer, D., Besmann, T. M. & Kaplan, W. D. The equilibrium crystal shape of nickel. Acta Mater. 59, 3473 3483 (2011). 94. Tsoga, A. & Nikolopoulos, P. Surface and grain - boundary energie s in yttria - stabilized zirconia (YSZ - 8 mol%). J. Mater. Sci. 31, 5409 5413 (1996). 95. Matsunaga, K. et al. Bonding nature of metal/oxide incoherent interfaces by first - principles calculations. Phys. Rev. B 74, 125423 (2006). 96. Sawata, A., Tsuneyoshi, K. , Mizusaki, J. & Tagawa, H. Oxygen chemical potential profile in a solid oxide fuel cell and simulation of electrochemical performance. Solid State Ion. 40 , 415 420 (1990). 97. Eremeev, S. V., Nemirovich - vacancies on adhesion at the Nb/Al 2 O 3 and Ni/ZrO 2 interfaces. Phys. Solid State 50, 543 552 (2008). 98. Eremeev, S. V., Schmauder, S., Hocker, S. & Kulkova, S. E. Ab - initio investigation of Ni(Fe)/ZrO 2 (0 0 1) and Ni Fe/ZrO 2 (0 0 1) interfaces. Surf. Sci. 60 3, 2218 2225 (2009). 99. Kasamatsu, S., Tada, T. & Watanabe, S. First principles study of oxygen vacancies near nickel/zirconia interface. E - J. Surf. Sci. Nanotechnol. 8, 93 100 (2010). 160 100. Shishkin, M. & Ziegler, T. The oxidation of H 2 and CH 4 on an oxyg en - enriched yttria - stabilized zirconia surface: A theoretical study based on density functional theory. J. Phys. Chem. C 112, 19662 19669 (2008). 101. Hemmingson, S. L. & Campbell, C. T. Trends in adhesion energies of metal nanoparticles on oxide surfaces: Understanding support effects in catalysis and nanotechnology. ACS Nano 11, 1196 1203 (2017). 102. Didier, F. & Jupille, J. The van der Waals contribution to the adhesion energy at metal - oxide interfaces. Surf. Sci. 314, 378 384 (1994). 103. Li, J. - G. Wet ting and Interfacial Bonding of Metals with Ionocovalent Oxides. J. Am. Ceram. Soc. 75, 3118 3126 (1992). 104. Miedema, A. R. & Boom, R. Surface tension and electron density of pure liquid metals. Z. Fuer Met. 69, 183 190 (1978). 105. Beltrán, J. I. & Muño z, M. C. Ab - initio study of decohesion properties in oxide/metal systems. Phys. Rev. B 78, 245417 (2008). 106. Chatain, D., Rivollet, I. & Eustathopoulos, N. Adhésion thermodynamique dans les systèmes non - réactifs métal liquide - alumine. J. Chim. Phys. 83, 561 567 (1986). 107. Chatain, D., Coudurier, L. & Eustathopoulos, N. Wetting and interfacial bonding in ionocovalent oxide - liquid metal systems. Rev. Phys. Appliquée 23, 1055 1064 (1988). 108. Peden, C. H. F., Kidd, K. B. & Shinn, N. D. Metal/metal oxide interfaces: A surface science approach to the study of adhesion. J. Vac. Sci. Technol. Vac. Surf. Films 9, 1518 1524 (1991). 109. - Fe 2 O 3 (0001) surface structure. Surf. Sci. 432, L627 L634 (1999). 110. Dai, Y., Lu, P., Cao, Z., Campbell, C. T. & Xia, Y. The physical chemistry and materials science behind sinter - resistant catalysts. Chem. Soc. Rev. 47, 4314 4331 (2018). 111. Gerberich, W. W. & Cordill, M. J. Physics of adhesion. Rep. Pr og. Phys. 69, 2157 (2006). 112. Zhang, Z., Kawamura, M., Abe, Y. & Kim, K. H. Thermal stability of Ag films with various interface layers. Jpn. J. Appl. Phys. 53, 048003 (2014). 113. Kriese, M. D., Moody, N. R. & Gerberich, W. W. Quantitative measurement o f the effect of annealing on the adhesion of thin copper films using superlayers. MRS Online Proc. Libr. Arch. 505, (1997). 114. Moody, N. R., Adams, D. P., Cordill, M. J., Yang, N. & Bahr, D. F. The effects of copper on the interfacial failure of gold films. MRS Online Proc. Libr. Arch. 695, (2001). 161 115. Baron, C. H., Raabe, D., Korte - Kerzel, S. & Scheu, C. On the design of alloys and synthesis for composite steels. (RWTH Aachen Aachen, Germany, 2017). 116. Yoon, Y., Rousseau, R., Weber, R. S., Mei, D. & Lercher, J. A. First - principles study of phenol hydrogenation on Pt and Ni catalysts in aqueous phase. J. Am. Chem. Soc. 136, 1028 7 10298 (2014). 117. Hensley, A. J. R., Wang, Y., Mei, D. & McEwen, J. - S. Mechanistic effects of water on the Fe - catalyzed hydrodeoxygenation of phenol. The role of Brønsted acid sites. ACS Catal. 8, 2200 2208 (2018). 118. Garcia - Ratés, M., García - Muelas, R. & López, N. Solvation effects on methanol decomposition on Pd(111), Pt(111), and Ru(0001). J. Phys. Chem. C 121, 13803 13809 (2017). 119. Kulig, W. Introduction to solvation models for theoretical chemistry. (2016). 120. Mennucci, B. & Tomasi, J. Contin uum solvation models: A new approach to the problem of J. Chem. Phys. 106, 5151 5158 (1997). 121. Dynamics. C hem. Rev. 99, 2161 2200 (1999). 122. Sundararaman, R. & Schwarz, K. Evaluating continuum solvation models for the electrode - electrolyte interface: Challenges and strategies for improvement. J. Chem. Phys. 146, 084111 (2017). 123. Gunceler, D., Letchworth - W eaver, K., Sundararaman, R., Schwarz, K. A. & Arias, T. A. The importance of nonlinear fluid response in joint density - functional theory studies of battery systems. Model. Simul. Mater. Sci. Eng. 21, 074005 (2013). 124. Le, J., Cuesta, A. & Cheng, J. The s tructure of metal - water interface at the potential of zero charge from density functional theory - based molecular dynamics. J. Electroanal. Chem. 819, 87 94 (2018). 125. Le, J., Fan, Q., Perez - Martinez, L., Cuesta, A. & Cheng, J. Theoretical insight into th e vibrational spectra of metal water interfaces from density functional theory based molecular dynamics. Phys. Chem. Chem. Phys. 20, 11554 11558 (2018). 126. Iyemperumal, S. K. & Deskins, N. A. Evaluating solvent effects at the aqueous/Pt(111) interface. C hemPhysChem 18, 2171 2190 (2017). 127. Saiz, E., Cannon, R. M. & Tomsia, A. P. High - temperature wetting and the work of adhesion in metal/oxide systems. Annu. Rev. Mater. Res. 38, 197 226 (2008). 128. Fergus, J. W. Sealants for solid oxide fuel cells. J. Power Sources 147, 46 57 (2005). 129. Braze alloy containing particulate material. (2005). 162 130. Method and apparatus for gasketing a fuel cell. (2004). 131. Le, S. et al. Effective Ag CuO sealant for planar solid oxide fuel cells. J. Alloys Compd. 496, 96 99 (2010). 132. Zhou, Q., Bieler, T. R. & Nicholas, J. D. Transient porous nickel interlayers for improv ed silver - based solid oxide fuel cell brazes . Acta Mater. 148, 156 162 (2018). 133. Ueki, M., Naka, M. & Okamoto, I. Wettability of some metals against zirconia ceramics. J. Mater. Sci. Lett. 5, 1261 1262 (1986). 134. Achievements in Reactive Air Brazing. Adv. Eng. Mater. 16, 1413 1414 (2014). 135. Meier, A. M., Chidambaram, P. & Edwards, G. R. A comparison of the wettability of copper - copper oxide and s ilver - copper oxide on polycrystalline alumina. J. Mater. Sci. 30, 4781 4786 (1995). 136. Kim, J. Y., Hardy, J. S. & Weil, K. S. Silver - copper oxide based reactive air braze for joining yttria - stabilized zirconia. J. Mater. Res. 20, 636 643 (2005). 137. Che n, H. et al. Reactive air brazing for sealing mixed ionic electronic conducting hollow fibre membranes. Acta Mater. 88, 74 82 (2015). 138. Hashimoto, S. et al. Oxygen nonstoichiometry and thermo - chemical stability of La 0.6 Sr 0.4 Co 1 - y Fe y O 3 - (y=0.2, 0.4, 0.6, 0.8). Solid State Ion. 181, 1713 1719 (2010). 139. Chen, Yu & Adler, S. B. Thermal and chemical expansion of S r - doped lanthanum cobalt oxide (La 1 - x Sr x CoO 3 - ). Chem. Mater. 17, 4537 4546 (2005). 140. Bishop, S. R., Duncan, K. L. & Wachsma n, E. D. Thermo - chemical expansion in strontium - doped lanthanum cobalt iron oxide. J. Am. Ceram. Soc. 93, 4115 4121 (2010). 141. Pfaff, E. M., Kaletsch, A. & Broeckmann, C. Design of a mixed ionic/electronic conducting oxygen transport membrane pilot modul e. Chem. Eng. Technol. 35, 455 463 (2012). 142. Cao, J., Si, X., Li, W., Song, X. & Feng, J. Reactive air brazing of YSZ - electrolyte and Al2O3 - substrate for gas sensor sealing: Interfacial microstructure and mechanical properties. Int. J. Hydrog. Energy 42, 10683 10694 (2017). 143. Weil, K. S., Hardy, J. S. & Kim, J. Y. Oxidation ceramic to metal braze seals for applications in high temperature electrochemical devices and method of making. (2006). 144. Middelkoop, V. Reactive air brazing (RAB) for gas sep aration membranes. in Encyclopedia of Membranes (eds. Drioli, E. & Giorno, L.) 1710 1713 (Springer Berlin Heidelberg, 2016). doi:10.1007/978 - 3 - 662 - 44324 - 8_1773 163 145. Chatzimichail, R., Triantafyllou, G., Tietz, F. & Nikolopoulos, P. Interfacial properties o f (Ag + CuO) brazes used as sealing materials in SOFC stacks. J. Mater. Sci. 49, 300 313 (2014). 146. Barin, I. Compilation of thermochemical data. in Thermochemical Data of Pure Substances 33 34 (Wiley - VCH Verlag GmbH, 1995). doi:10.1002/9783527619825.ch3 147. Zhou, Q., Bieler, T. R. & Nicholas, J. D. Reactive - element - free, silver - based brazes for SOFC applications. ECS Trans. 78, 1759 1767 (2017). 148. Hu, J., Li, D., Lu, J. G. & Wu, R. Effects on electronic properties of molecule adsorption on CuO surfac es and nanowires. J. Phys. Chem. C 114, 17120 17126 (2010). 149. Maimaiti, Y., Nolan, M. & Elliott, S. D. Reduction mechanisms of the CuO(111) surface through surface oxygen vacancy formation and hydrogen adsorption. Phys. Chem. Chem. Phys. 16, 3036 3046 ( 2014). 150. Eichler, A. & Kresse, G. First - principles calculations for the surface termination of pure and yttria - doped zirconia surfaces. Phys. Rev. B 69, 045402 (2004). 151. Takemoto, S. & Tada, T. Zirconium - peroxo embedded in non - stoichiometric yttria s tabilized zirconia (110) from first - principles. Solid State Ion. 285, 215 221 (2016). 152. Ding, H., Virkar, A. V. & Liu, F. Defect configuration and phase stability of cubic versus tetragonal yttria - stabilized zirconia. Solid State Ion. 215, 16 23 (2012). 153. Pascual, C. & Durán, P. Subsolidus phase equilibria and ordering in the system ZrO 2 - Y 2 O 3 . J. Am. Ceram. Soc. 66, 23 27 (1983). 154. Beltrán, J. I., Gallego, S., Cerdá, J., Moya, J. S. & Muñoz, M. C. Internal ceramic 2 adhesion. J. Phys. Chem. B 108, 15439 15442 (2004). 155. at the metal - zirconium dioxide interfaces. Tech. Phys. 58, 325 334 (2013). 156. Malyi, O. I., Kulish, V. V., Bai, K., Wu, P. & Chen, Z. A computational study of the effect of alloying additions on the stability of Ni/c - ZrO 2 interfaces. Surf. Sci. 611, 5 9 (2013). 157. Scott Weil, K., Coyle, C. A., Darsell, J. T., Xia, G. G. & Hardy, J. S. Effects of thermal cycli ng and thermal aging on the hermeticity and strength of silver copper oxide air - brazed seals. J. Power Sources 152, 97 104 (2005). 158. Skiera, E. et al. Mechanical properties of reactive air brazed (RAB) metal/ceramic joints. Part 1: Visco - plastic deforma tion of silver - based reactive air brazes. Adv. Eng. Mater. 16, 1462 1467 (2014). 159. Kim, J. Y., Hardy, J. S. & Weil, S. Dual - atmosphere tolerance of Ag CuO - based air braze. Int. J. Hydrog. Energy 32, 3655 3663 (2007). 164 160. Pönicke, A., Schilm, J., Kusnez off, M. & Michaelis, A. Aging behavior of reactive air brazed seals for SOFC. Fuel Cells 15, 735 741 (2015). 161. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple. Phys. Rev. Lett. 77, 3865 3868 (1996). 162. Becke, A. D. Density - functional exchange - energy approximation with correct asymptotic behavior. Phys. Rev. A 38, 3098 3100 (1988). 163. Kresse, G. & Joubert, D. From ultrasoft pseudopotentials to the projector augmented - wave method. Phys. Rev. B 59, 1758 1775 (1999 ). 164. Dudarev, S. L., Botton, G. A., Savrasov, S. Y., Humphreys, C. J. & Sutton, A. P. Electron - energy - loss spectra and the structural stability of nickel oxide: An LSDA+U study. Phys. Rev. B 57, 1505 1509 (1998). 165. Mosey, N. J., Liao, P. & Carter, E . A. Rotationally invariant ab initio evaluation of Coulomb and exchange parameters for DFT+U calculations. J. Chem. Phys. 129, 014103 (2008). 166. Mansourian - Hadavi, N. et al. Transport and band st ructure studies of crystalline ZnRh 2 O 4 . Phys. Rev. B 81, 0 75112 (2010). 167. Bengone, O., Alouani, M., Blöchl, P. & Hugel, J. Implementation of the projector augmented - wave LDA+U method: Application to the electronic structure of NiO. Phys. Rev. B 62, 16392 16401 (2000). 168. Andersson, D. A., Simak, S. I., Johan sson, B., Abrikosov, I. A. & Skorodumova, N. V. Modeling of CeO 2 , Ce 2 O 3 , and CeO 2 - x in the LDA+U formalism . Phys. Rev. B 75, 035109 (2007). 169. Density functional theory analysis of dopants in cupric oxide. J. Appl. Phys. 111, 103708 (2012). 170. Zhao, Z. et al. First - principles study on the doping effects of nitrogen on the electronic structure and optical properties of Cu 2 O. RSC Adv. 3, 84 90 (2012). 171. Chevalier, J., Gremillard, L., Virkar, A. V. & Clarke, D. R. The tetragonal - monoclinic transformation in zirconia: Lessons learned and future trends. J. Am. Ceram. Soc. 92, 1901 1920 (2009). 172. Idris, M. A., Bak, T., Li, S. & Nowotny, J. Effect of segreg ation on surface and near - surface chemistry of yttria - stabilized zirconia. J. Phys. Chem. C 116, 10950 10958 (2012). 173. Winnubst, A. J. A., Kroot, P. J. M. & Burggraaf, A. J. AES/STEM grain boundary analysis of stabilized zirconia ceramics. J. Phys. Chem . Solids 44, 955 960 (1983). 174. Kumikov, V. K. & Khokonov, K. B. On the measurement of surface free energy and surface tension of solid metals. J. Appl. Phys. 54, 1346 1350 (1983). 165 175. Donnan, F. G. & Shaw, T. W. A. The solubility of oxygen in molten si lver. J. Soc. Chem. Ind. 29, 987 989 (1910). 176. Kirshenbaum, A. D., Cahill, J. A. & Grosse, A. V. The density of liquid silver from its melting point to its normal boiling point 2450°K. J. Inorg. Nucl. Chem. 24, 333 336 (1962). 177. (Accelrys Inc., 2011). 178. Sun, H. et al. COMPASS II: extended coverage for polymer and drug - like molecule databases. J. Mol. Model. 22, 47 (2016). 179. Zhang, W. X. & He, C. Melting of cu nanowires: A study using molecular dynamics sim ulation. J. Phys. Chem. C 114, 8717 8720 (2010). 180. Wagner, C. N. J., Ocken, H. & Joshi, M. L. Interference and radial distribution functions of liquid copper, silver, tin, and mercury. Z. Für Naturforschung A 20, 325 335 (2014). 181. Darsell, J. T. & We il, K. S. Experimental determination of phase equilibria in the silver copper oxide system at high temperature. Scr. Mater. 56, 1111 1114 (2007). 182. Tyson, W. R. & Miller, W. A. Surface free energies of solid metals: Estimation from liquid surface tensio n measurements. Surf. Sci. 62, 267 276 (1977). 183. Sotiropoulou, D. & Nikolopoulos, P. Work of adhesion in ZrO 2 - liquid metal systems. J. Mater. Sci. 28, 356 360 (1993). 184. Zhang, W. & Smith, J. R. Stoichiometry and adhesion of Nb /A l 2 O 3 . Phys. Rev. B 61, 16883 16889 (2000). 185. Characterization of CuAlO 2 thin films prepared on sapphire substrates by reactive sputtering and annealing. Jpn. J. Appl. Phys. 47, 592 (2008). 186. Shih, C. H. & Tseng, B. H. Formation mechanism of CuAlO 2 prepared by rapid thermal annealing of Al 2 O 3 /Cu 2 O/sapphire sandwich structure. Phys. Procedia 32, 395 400 (2012). 187. Liu, Q. - J. & Liu, Z. - T. Theoretical insights into structural electronic relationships and relative stability of the Cu - , Al - and O - termi nated CuAlO 2 (0001) surfaces. Vacuum 107, 90 98 (2014). 188. Hennings, D. Phase equilibria and thermodynamics of the double oxide phase Cu3TiO4. J. Solid State Chem. 31, 275 279 (1980). 189. Derling, S., Müller, T., Abicht, H. - P., Felgner, K. - H. & Langhamme r, H. T. Copper oxide as a sintering agent for barium titanate based ceramics. J. Mater. Sci. 36, 1425 1431 (2001). 190. Jacob, K. T. & Alcock, C. B. Thermodynamics of CuAlO 2 and CuAl 2 O 4 and phase equilibria in the system Cu 2 O - CuO - Al 2 O 3 . J. Am. Ceram. Soc. 58, 192 195 (1975). 166 191. Erskine, K. M., Meier, A. M. & Pilgrim, S. M. Brazing perovskite ceramics with silver/copper oxide braze alloys. J. Mater. Sci. 37, 1705 1709 (2002). 192. Weil, K. S., Hardy, J. S., Rice, J. P. & Kim, J. Y. Brazing as a means of s ealing ceramic membranes for use in advanced coal gasification processes. Fuel 85, 156 162 (2006). 193. Souza, G. M., Victoria, R., Joly, C. & Verdade, L. Bioenergy & Sustainability: Bridging the gaps . (SCOPE Bioenergy, 2015). 194. Davison, B. H. et al. Th e impact of biotechnological advances on the future of US bioenergy. Biofuels Bioprod. Biorefining 9, 454 467 (2015). 195. Himmel, M. E. et al. Biomass recalcitrance: Engineering plants and enzymes for biofuels production. Science 315, 804 807 (2007). 196. Bruijnincx, P. C. A. & Weckhuysen, B. M. Biomass conversion: Lignin up for break - down. Nat. Chem. 6, 1035 1036 (2014). 197. Ragauskas, A. J. et al. Lignin valorization: Improving lignin processing in the biorefinery. Science 344, 1246843 (2014). 198. Lora, J. H. & Glasser, W. G. Recent industrial applications of lignin: A sustainable alternative to nonrenewable materials. J. Polym. Environ. 10, 39 48 (2002). 199. Mansouri, N. - E. E. & Salvadó, J. Structural characterization of technical lignins for the production of adhesives: Application to lignosulfonate, kraft, soda - anthraquinone, organosolv and ethanol process lignins. Ind. Crops Prod. 24, 8 16 (2006). 200. Zakzeski, J., Jongerius, A. L., Bruijnincx, P. C. A. & Weckhuysen, B. M. Catalytic lignin valo rization process for the production of aromatic chemicals and hydrogen. ChemSusChem 5, 1602 1609 (2012). 201. Holladay, J. E., White, J. F., Bozell, J. J. & Johnson, D. Top value - added chemicals from biomass - Volume II Results of screening for potential candidates from biorefinery lignin . (2007). 202. Zheng, Y., Chen, D. & Zhu, X. Aromatic hydrocarbon production by the online catalytic cracking of lignin fast pyrolysis vapors using Mo 2 - Al 2 O 3 . J. Anal. Appl. Pyrolysis 104, 514 520 (2013). 203. Yoshikawa , T. et al. Production of phenols from lignin via depolymerization and catalytic cracking. Fuel Process. Technol. 108, 69 75 (2013). 204. Werhan, H., Mir, J. M., Voitl, T. & Rudolf von Rohr, P. Acidic oxidation of kraft lignin into aromatic monomers cataly zed by transition metal salts. Holzforschung 65, 703 709 (2011). 167 205. Kelley, P., Lin, S., Edouard, G., Day, M. W. & Agapie, T. Nickel - mediated hydrogenolysis of C O bonds of aryl ethers: What Is the Source of the Hydrogen? J. Am. Chem. Soc. 134, 5480 5483 (2012). 206. Parsell, T. H. et al. Cleavage and hydrodeoxygenation (HDO) of C O bonds relevant to lignin conversion using Pd/Zn synergistic catalysis. Chem. Sci. 4, 806 813 (2013). 207. Sturgeon, M. R. et al. Lignin depolymerisation by nickel supported la yered - double hydroxide catalysts. Green Chem. 16, 824 835 (2014). 208. - O - 4 bonds of lignin model compounds in an acidic ionic liquid, 1 - H - 3 - methylimidazolium chloride: An optional st rategy for the degradation of lignin. ChemSusChem 3, 1078 1084 (2010). 209. Wu, G., Heitz, M. & Chornet, E. Improved Alkaline oxidation process for the production of aldehydes (vanillin and syringaldehyde) from steam - explosion hardwood lignin. Ind. Eng. Ch em. Res. 33, 718 723 (1994). 210. Behling, R., Valange, S. & Chatel, G. Heterogeneous catalytic oxidation for lignin valorization into valuable chemicals: what results? What limitations? What trends? Green Chem. 18, 1839 1854 (2016). 211. Azarpira, A., Ral ph, J. & Lu, F. Catalytic alkaline oxidation of lignin and its model compoun ds: A pathway to aromatic biochemicals. BioEnergy Res. 7, 78 86 (2014). 212. Stärk, K., Taccardi, N., Bösmann, A. & Wasserscheid, P. Oxidative depolymerization of lignin in ionic l iquids. ChemSusChem 3, 719 723 (2010). 213. Wu, G. & Heitz, M. Catalytic mechanism of Cu 2+ and Fe 3+ in alkaline O 2 oxidation of lignin. J. Wood Chem. Technol. 15, 189 202 (1995). 214. Jiang, Y. - Y., Yan, L., Yu, H. - Z., Zhang, Q. & Fu, Y. Mechanism of vanadium - catalyzed selective C O and C C cleavage of lignin model compound. ACS Catal. 6, 4399 4410 (2016). 215. Sedai, B. et al. Comparison of copper and vanadium homogeneous catalysts for aerobic oxidation of lignin models. ACS Catal. 1, 794 804 (2011). 216. Hoover, J. M. & Stahl, S. S. Highly practical copper(I)/TEMPO catalyst system for chemoselective aerobic oxidation of primary alcohols. J. Am. Chem. Soc. 133, 16901 16910 (2011). 217. B ujanovic, B., Reiner, R. S., Ralph, S. A. & Atalla, R. H. Polyoxometalate delignification of birch kraft pulp and effect on residual lignin. J. Wood Chem. Technol. 31, 121 141 (2011). 218. Hocking, M. B. Vanillin: Synthetic flavoring from spent sulfite liq uor. J. Chem. Educ. 74, 1055 (1997). 168 219. Dignum, M. J. W., Kerler, J. & Verpoorte, R. Vanilla production: technological, chemical, and biosynthetic aspects. Food Rev. Int. 17, 119 120 (2001). 220. Fargues, C., Mathias, Á. & Rodrigues, A. Kinetics of vanil lin production from kraft lignin oxidation. Ind. Eng. Chem. Res. 35, 28 36 (1996). 221. Sergeev, A. G., Webb, J. D. & Hartwig, J. F. A Heterogeneous nickel catalyst for the hydrogenolysis of aryl ethers without arene hydrogenation. J. Am. Chem. Soc. 134, 2 0226 20229 (2012). 222. Zhang, J. et al. A Series of NiM (M = Ru, Rh, and Pd) Bimetallic catalysts for effective lignin hydrogenolysis in water. ACS Catal. 4, 1574 1583 (2014). 223. Zhang, J. et al. Highly efficient, NiAu - catalyzed hydrogenolysis of lignin into phenolic chemicals. Green Chem. 16, 2432 2437 (2014). 224. He, J., Zhao, C. & Lercher, J. A. Ni - catalyzed cleavage of aryl ethers in the aqueous phase. J. Am. Chem. Soc. 134, 20768 20775 (2012). 225. Onwudili, J. A. & Williams, P. T. Catalytic depoly merization of alkali lignin in subcritical water: influence of formic acid and Pd/C catalyst on the yields of liquid monomeric aromatic products. Green Chem. 16, 4740 4748 (2014). 226. Song, Q. et al. Lignin depolymerization (LDP) in alcohol over nickel - based catalysts via a fragmentation hydrogenolysis process. Energy Environ. Sci. 6, 994 1007 (2013). 227. Bosch, S. V. den et al. Reductive lignocellulose fractionation into soluble lignin - derived phenolic monomers and dimers and processable carbohydrate pulps. Energy Environ. Sci. 8, 1748 1763 (2015). 228. Barta, K. & Ford, P. C. Catalytic conversion of nonfood woody biomass solids to organic liquids. Acc. Chem. Res. 47, 1503 1512 (2014). 229. Barta, K. et al. Catalytic disassembly of an organosolv lignin via hydrogen transfer from supercritical methanol. Green Chem. 12, 1640 1647 (2010). 230. Song, Q., Wang, F. & Xu, J. Hydrogenolys is of lignosulfonate into phenols over heterogeneous nickel catalysts. Chem. Commun. 48, 7019 7021 (2012). 231. Sun, S. - N., Li, M. - F., Yuan, T. - Q., Xu, F. & Sun, R. - C. Effect of ionic liquid/organic solvent pretreatment on the enzymatic hydrolysis of cornc ob for bioethanol production. Part 1: Structural characterization of the lignins. Ind. Crops Prod. 43, 570 577 (2013). 232. Schutyser, W. et al. Influence of bio - based solvents on the catalytic reductive fractionation of birch wood. Green Chem. 17, 5035 50 45 (2015). 169 233. Gao, F., Webb, J. D., Sorek, H., Wemmer, D. E. & Hartwig, J. F. Fragmentation of lignin samples with commercial Pd/C under ambient pressure of hydrogen. ACS Catal. 6, 7385 7392 (2016). 234. Yue, F., Lu, F., Sun, R. - C. & Ralph, J. Syntheses of lignin - derived thioacidolysis monomers and their uses as quantitation standards. J. Agric. Food Chem. 60, 922 928 (2012). 235. Boerjan, W., Ralph, J. & Baucher, M. Lignin biosynthesis. Annu. Rev. Plant Biol. 54, 519 546 (2003). 236. Aguayo, M. G., Ferra z, A., Elissetche, J. P., Masarin, F. & Mendonça, R. T. Lignin chemistry and topochemistry during kraft delignification of eucalyptus globulus genotypes with contrasting pulpwood characteristics. Holzforschung 68, 623 629 (2013). 237. Alvira, P., Tomás - Pej ó, E., Ballesteros, M. & Negro, M. J. Pretreatment technologies for an efficient bioethanol production process based on enzymatic hydrolysis: A review. Bioresour. Technol. 101, 4851 4861 (2010). 238. Brodeur, G. et al. Chemical and Physicochemical Pretreat ment of Lignocellulosic Biomass: A Review. Enzyme Res. 2011, e787532 (2011). 239. Gellerstedt, G., Majtnerova, A. & Zhang, L. Towards a new concept of lignin condensation in kraft pulping. Initial results. C. R. Biol. 327, 817 826 (2004). 240. Nakamura, T. , Kawamoto, H. & Saka, S. Condensation reactions of some lignin related compounds at relatively low pyrolysis temperature. J. Wood Chem. Technol. 27, 121 133 (2007). 241. Shimada, K., Hosoya, S. & Ikeda, T. Condensation reactions of softwood and hardwood l ignin model compounds under organic acid cooking conditions. J. Wood Chem. Technol. 17, 57 72 (1997). 242. Stoklosa, R. J. & Hodge, D. B. Fractionation and improved enzymatic deconstruction of hardwoods with alkaline delignification. BioEnergy Res. 8, 1224 1234 (2015). 243. Velez, J. & Thies, M. C. Solvated liquid - lignin fractions from a Kraft black liquor. Bioresour. Technol. 148, 586 590 (2013). 244. Gu, F. et al. Effect of complete dissolution in LiCl/DMSO on the isolation and characteristics of lignin f rom wheat straw internode. Ind. Crops Prod. 74, 703 711 (2015). 245. Stoklosa, R. J. et al. Correlating lignin structural features to phase partitioning behavior in a novel aqueous fractionation of softwood Kraft black liquor. Green Chem. 15, 2904 2912 (2013). 246. Balakshin, M. & Capanema, E. On the quantification of lignin hydroxyl groups with 31P and 13C NMR spectroscopy. J. Wood Chem. Technol. 35, 220 237 (2015). 170 247. El Hage, R. et al. Characterization of milled wood lignin and ethanol or ganosolv lignin from miscanthus. Polym. Degrad. Stab. 94, 1632 1638 (2009). 248. Kim, H. & Ralph, J. Solution - state 2D NMR of ball - milled plant cell wall gels in DMSO - d6/pyridine - d5. Org. Biomol. Chem. 8, 576 591 (2010). 249. Li, Z. et al. Chemical and str uctural changes associated with Cu - catalyzed alkaline - oxidative delignification of hybrid poplar. Biotechnol. Biofuels 8, 1 12 (2015). 250. Ralph, S. A. & Ralph, J. NMR database of lignin and cell wall model compounds. 251. Liitiä, T. M., Maunu, S. L., Hor tling, B., Toikka, M. & Kilpeläinen, I. Analysis of technical lignins by two - and three - dimensional NMR spectroscopy. J. Agric. Food Chem. 51, 2136 2143 (2003). 252. Toledano, A., Serrano, L., Garcia, A., Mondragon, I. & Labidi, J. Comparative study of lig nin fractionation by ultrafiltration and selective precipitation. Chem. Eng. J. 157, 93 99 (2010). 253. black liquor. Carbohydr. Polym. 70, 218 223 (2007). 254. Klett, A. S., Chappell, P. V. & Thies, M. C. Recovering ultraclean lignins of controlled molecular weight from Kraft black - liquor lignins. Chem. Commun. 51, 12855 12858 (2015). 255. Saito, T. et al. Methanol fractionation of softwood kraft lignin: Impact o n the lignin properties. ChemSusChem 7, 221 228 (2014). 256. Lake, M. A. & Blackburn, J. C. Process for recovering lignin. (2011). 257. Tomani, P. The lignoboost process. Cellul. Chem. Technol. 44, 53 58 258. Kouisni, L., Holt - hindle, P., Maki, K. & Paleol ogou, M. The lignoforce system TM : A new process for the production of high - quality lignin from black liquor. J. Sci. Technol. For. Prod. Process. 2, 6 10 (2012). 259. Zhang, L. & Gellerstedt, G. Quantitative 2D HSQC NMR determination of polymer structures by selecting suitable internal standard references. Magn. Reson. Chem. 45, 37 45 (2007). 260. Faith, W. L., Keyes, D. B., Clark, R. L., Lowenheim, F. A. & Moran, M. K. Vanillin. in Industrial chemicals (Wiley New York, 1957). 261. Brunow, G., Lundquist, K. & Gellerstedt, G. Lignin. in Analytical methods in wood chemistry, pulping, and papermaking (eds. Sjöström, P. D. E. & Alén, P. D. R.) 77 124 (Springer Berlin Heidelberg, 1999). doi:10.1007/978 - 3 - 662 - 03898 - 7_4 262. Hatfield, R. & Fukushima, R. S. Can lignin be accurately measured? Crop Sci. 45, 832 (2005). 171 263. Constant, S., Robitzer, M., Quignard, F. & Di Renzo, F. Vanillin oligomerization as a model of side reactions in lignin fragmentation. Catal. Today 189, 123 1 28 (2012). 264. Tarabanko, V. E., Petukhov, D. V. & Selyutin, G. E. New mechanism for the catalytic oxidation of lignin to vanillin. Kinet. Catal. 45, 569 577 (2004). 265. Sedai, B. et al. - 1 lignin model compounds with copper and oxo vanadium catalysts. ACS Catal. 3, 3111 3122 (2013). 266. Bosch, S. V. den et al. Tuning the lignin oil OH - content with Ru and Pd catalysts during lignin hydrogenolysis on birch wood. Chem. Commun. 51, 13158 13161 (2015). 267. Yan, N. et al. Selective degra dation of wood lignin over noble - metal catalysts in a two - step process. ChemSusChem 1, 626 629 (2008). 268. Evstigneev, E. I. Factors affecting lignin solubility. Russ. J. Appl. Chem. 84, 1040 1045 (2011). 269. Crestini, C., Melone, F., Sette, M. & Saladin o, R. Milled wood lignin: A linear oligomer. Biomacromolecules 12, 3928 3935 (2011). 270. Yue, F., Lu, F., Ralph, S. & Ra lph, J. Identification of 4 O 5 units in softwood lignins via definitive lignin models and NMR. Biomacromolecules 17, 1909 1920 (2016). 271. Montroll, E. W. & Simha, R. Theory of depolymerization of long chain molecules. J. Chem. Phys. 8, 721 726 (1940). 272. Feghali, E., Carrot, G., Thuéry, P., Genre, C. & Cantat, T. Convergent reductive depolymerization of wood lignin to isolated phenol derivatives by metal - free catalytic hydrosilylation. Energy Environ. Sci. 8, 2734 2743 (2015). 273. Voitl, T. & Rudolf von Rohr, P. Oxidation of lignin using aqueous polyoxometalates in the presence of alcohols. ChemSusChem 1, 763 769 (2008). 274. Huang, X., Korányi, T. I., Boot, M. D. & Hensen, E. J. M. Ethanol as capping agent and formaldehyde scavenger for efficient depolymerization of lignin to aromatics. Green Chem. 17, 4941 4950 (2015). 275. Anderson, E. M. et al. Reductive catalytic fractionation of corn stover lignin. ACS Sustain. Chem. Eng. 4, 6940 6950 (2016). 276. Galkin, M. V. & Samec, J. S. M. Selective route to 2 - Propenyl aryls directly from wood by a tandem organosolv and palladium - catalysed transfer hydrogenolysis. ChemSusChem 7, 2154 2158 (2014). 277. Shuai, L. et al. Formaldehyde stabilization facilitates lignin monomer production during biomass depolymerization. Science 354, 329 333 (2016). 172 278. Luterbacher, J. S. et al. - valerolac tone sugar platform. Energy Environ. Sci. 8, 2657 2663 (2015). 279. Rinaldi, R. et al. Paving the way for lignin valorisation: Recent advances in bioengineering, biorefining and catalysis. Angew. Chem. Int. Ed. 55, 8164 8215 (2016). 280. Vanholme, R., Deme dts, B., Morreel, K., Ralph, J. & Boerjan, W. Lignin biosynthesis and structure. Plant Physiol. 153, 895 905 (2010). 281. Zakzeski, J., Bruijnincx, P. C. A., Jongerius, A. L. & Weckhuysen, B. M. The catalytic valorization of lignin for the production of re newable chemicals. Chem. Rev. 110, 3552 3599 (2010). 282. Qi, S. - C., Hayashi, J., Kudo, S. & Zhang, L. Catalytic hydrogenolysis of kraft lignin to monomers at high yield in alkaline water. Green Chem. 19, 2636 2645 (2017). 283. Klein, I., Saha, B. & Abu - Omar, M. M. Lignin depolymerization over Ni/C catalyst in methanol, a continuation: effect of substrate and catalyst loading. Catal. Sci. Technol. 5, 3242 3245 (2015). 284. Chen, J. et al. High yield production of natural phenolic alcohols from woody biomass using a nickel - based catalyst. ChemSusChem 9, 3353 3360 (2016). 285. Anderson, E. M. et al. Flowthrough reductive catalytic fractionation of biomass. Joule 1, 613 622 (2017). 286. Phongpreecha, T. et al. Predicting lignin depol ymerization yields from quantifiable properties using fractionated biorefinery lignins. Green Chem. 19, 5131 5143 (2017). 287. Zhang, Y., He, H., Dong, K., Fan, M. & Zhang, S. A DFT study on lignin dissolution in imidazolium - based ionic liquids. RSC Adv. 7 , 12670 12681 (2017). 288. O 4 type lignin model compound and 1 allyl 3 methylimidazolium chloride ionic liquid. Biopolymers 107, (2017). 289. Silveira, R. L., Stoyanov, S. R., Gusarov, S., Skaf, M. S. & Kovalenko, A. Supramolecular interactions in secondary plant cell walls: effect of lignin chemical composition revealed with the molecular theory of solvation. J. Phys. Chem. Lett. 6, 206 211 (20 15). 290. Petridis, L., Schulz, R. & Smith, J. C. Simulation analysis of the temperature dependence of lignin structure and dynamics. J. Am. Chem. Soc. 133, 20277 20287 (2011). 291. Galkin, M. V., Sawadjoon, S., Rohde, V., Dawange, M. & Samec, J. S. M. Mil d heterogeneous palladium - - O - - ether linkages of lignin model compounds and native lignin in air. ChemCatChem 6, 179 184 (2014). 173 292. Galkin, M. V., Dahlstrand, C. & Samec, J. S. M. Mild and robust redox - neutral Pd/C - catalyzed ligno - O - - energy - barrier pathway. ChemSusChem 8, 2187 2192 (2015). 293. Parthasarathi, R., Romero, R. A., Redondo, A. & Gnanakaran, S. Theoretical study of the remarkably diverse linkages in lignin. J. Phys. Chem. Lett. 2, 2660 2 666 (2011). 294. Beste, A. & Buchanan, A. C. Computational study of bond dissociation enthalpies for lignin model compounds. substituent effects in phenethyl phenyl ethers. J. Org. Chem. 74, 2837 2841 (2009). 295. Kim, S. et al. Computational study of bond dissociation enthalpies for a large range of native and modified lignins. J. Phys. Chem. Lett. 2, 2846 2852 (2011). 296. Grimme, S., Antony, J., Ehrlich, S. & Krieg, H. A consistent and accurate ab initio parametrization of density functional dispersion c orrection (DFT - D) for the 94 elements H - Pu. J. Chem. Phys. 132, 154104 (2010). 297. Grimme, S., Ehrlich, S. & Goerigk, L. Effect of the damping function in dispersion corrected density functional theory. J. Comput. Chem. 32, 1456 1465 (2011). 298. Dion, M. , Rydberg, H., Schröder, E., Langreth, D. C. & Lundqvist, B. I. Van der Waals density functional for general geometries. Phys. Rev. Lett. 92, 246401 (2004). 299. fu nctional. J. Phys. Condens. Matter 22, 022201 (2010). 300. Mathew, K. & Hennig, R. G. Implicit self - consistent description of electrolyte in plane - wave density - functional theory. ArXiv160103346 Cond - Mat (2016). 301. Mathew, K., Sundararaman, R., Letchworth - Weaver, K., Arias, T. A. & Hennig, R. G. Implicit solvation model for density - functional study of nanocrystal surfaces and reaction pathways. J. Chem. Phys. 140, 084106 (2014). 302. Batsanov, S. S. Van der Waals Radii of Elements. Inorg. Mater. 37, 871 88 5 (2001). 303. Fishman, M., Zhuang, H. L., Mathew, K., Dirschka, W. & Hennig, R. G. Accuracy of exchange - correlation functionals and effect of solvation on the surface energy of copper. Phys. Rev. B 87, 245402 (2013). 304. Implicit solvation model for dens ity - functional study of nanocrystal surfaces and reaction pathways. J. Chem. Phys. 140, 084106 (2014). 305. Dannhauser, W. & Bahe, L. W. Dielectric Constant of Hydrogen Bonded Liquids. III. Superheated Alcohols. J. Chem. Phys. 40, 3058 (2004). 306. Hinz, H . - J. Thermodynamic Data for Biochemistry and Biotechnology . (Springer Science & Business Media, 2012). 174 307. Rigby, D., Sun, H. & Eichinger, B. E. Computer simulations of poly(ethylene oxide): force field, pvt diagram and cyclization behaviour. Polym. Int. 44, 311 330 (1997). 308. Yildirim, H., Greber, T. & Kara, A. Trends in adsorption characteristics of benzene on transition metal surfaces: Role of surface chemistry and van der waals interactions. J. Phys. Chem. C 117, 20572 20583 (2013). 309. Lu, J. & Hey den, A. Theoretical investigation of the reaction mechanism of the hydrodeoxygenat ion of guaiacol over a Ru(000 1) model surface. J. Catal. 321, 39 50 (2015). 310. Benzene adsorption on Cu(111): Formation of a stable bilayer. J. Chem. Phys. 101, 9122 9131 (1994). 311. Koschel, H., Held, G. & Steinrück, H. - P. The orientation of benzene on bimetallic surfaces. Surf. Rev. Lett. 06, 893 901 (1999). 312. Ruiz, V. G., Liu, W., Zojer, E., Scheffler, M. & Tkatchenko, A. Density - functional theory with scr eened van der Waals interactions for the modeling of hybrid inorganic - organic systems. Phys. Rev. Lett. 108, 146103 (2012). 313. Lee, K., Morikawa, Y. & Langreth, D. C. Adsorption of n - butane on Cu(100), Cu(111), Au(111), and Pt(111): Van der Waals density - functional study. Phys. Rev. B 82, 155461 (2010). 314. Liu, W. et al. Structure and energetics of benzene adsorbed on transition - metal surfaces: density - functional theory with van der Waals interactions including collective substrate response. New J. Phys . 15, 053046 (2013). 315. Mittendorfer, F. et al. Graphene on Ni(111): Strong interaction and weak adsorption. Phys. Rev. B 84, 201401 (2011). 316. Lebègue, S. et al. Cohesive properties and asymptotics of the dispersion interaction in graphite by the rand om phase approximation. Phys. Rev. Lett. 105, 196401 (2010). 317. Delle Site, L., Abrams, C. F., Alavi, A. & Kremer, K. Polymers near metal surfaces: selective adsorption and global conformations. Phys. Rev. Lett. 89, 156103 (2002). 318. Xu, L. & Yang, X. Molecular dynamics simulation of adsorption of pyrene polyethylene glycol onto graphene. J. Colloid Interface Sci. 418, 66 73 (2014). 319. Besombes, S., Robert, D., Utille, J. - P., Taravel, F. R. & Mazeau, K. Molecular modeling of syringyl and p - hydroxyphen - O - 4 dimers. Comparative study of the computed and - O - 4 model compounds. J. Agric. Food Chem. 51, 34 42 (2003). 320. Simon, J. P. & Eriksson, K. - E. L. The significance of intra - molecular hydrogen bondin g in the ß - O - 4 linkage of lignin. J. Mol. Struct. 384, 1 7 (1996). 175 321. Stolov, M. A., Zaitseva, K. V., Varfolomeev, M. A. & Acree, W. E. Enthalpies of solution and enthalpies of solvation of organic solutes in ethylene glycol at 298.15K: Prediction and an alysis of intermolecular interaction contributions. Thermochim. Acta 648, 91 99 (2017). 322. Stephens, T. W. et al. Enthalpy of solvation correlations for organic solutes and gases dissolved in 2 - propanol, 2 - butanol, 2 - methyl - 1 - propanol and ethanol. Thermo chim. Acta 523, 214 220 (2011). 323. Saleheen, M. & Heyden, A. Liquid - phase modeling in heterogeneous catalysis. ACS Catal. 8, 2188 2194 (2018). 324. DeLano, W. L. PyMOL . (2002). 325. Bernt, C. M., Manesewan, H., Chui, M., Boscolo, M. & Ford, P. C. Tempera ture tuning the catalytic reactivity of Cu - Doped porous metal oxides with lignin models. ACS Sustain. Chem. Eng. 6, 2510 2516 (2018). 326. Besse, X., Schuurman, Y. & Guilhaume, N. Reactivity of lignin model compounds through hydrogen transfer catalysis in ethanol/water mixtures. Appl. Catal. B Environ. 209, 265 272 (2017). 327. Pepper, J. M. & Lee, Y. W. Lignin and related compounds. I. A comparative study of catalysts for lignin hydrogenolysis. Can. J. Chem. 47, 723 727 (1969). 328. Pepper, J. M. & Lee, Y. W. Lignin and related compounds. II. Studies using ruthenium and Raney nickel as catalysts for lignin hydrogenolysis. Can. J. Chem. 48, 477 479 (1970). 329. Singh, S. K. & Ekhe, J. D. Cu Mo doped zeolite ZSM - 5 catalyzed conversion o f lignin to alkyl phenols with high selectivity. Catal. Sci. Technol. 5, 2117 2124 (2015). 330. Oregui Bengoechea, M., Gandarias, I., Arias, P. L. & Barth, T. Unraveling the role of formic acid and the type of solvent in the catalytic conversion of lignin : A holistic approach. ChemSusChem 10, 754 766 (2017). 331. Zanchet, D., Santos, J. B. O., Damyanova, S., Gallo, J. M. R. & C. Bueno, J. M. Toward understanding metal - catalyzed ethanol reforming. ACS Catal. 5, 3841 3863 (2015). 332. Konsolakis, M. et al. H ydrogen production by ethanol steam reforming (ESR) over CeO 2 supported transition metal (Fe, Co, Ni, Cu) catalysts: Insight into the structure - activity relationship. Catalysts 6, 39 (2016). 333. Hou, T., Zhang, S., Chen, Y., Wang, D. & Cai, W. Hydrogen pr oduction from ethanol reforming: Catalysts and reaction mechanism. Renew. Sustain. Energy Rev. 44, 132 148 (2015). 334. Li, W., Wang, H., Ren, Z., Wang, G. & Bai, J. Co - production of hydrogen and multi - wall carbon nanotubes from ethanol decomposition over Fe/Al 2 O 3 catalysts. Appl. Catal. B Environ. 84, 433 439 (2008). 176 335. Samuel, R., Pu, Y., Raman, B. & Ragauskas, A. J. Structural characterization and comparison of switchgrass ball - milled lignin before and after dilute acid pretreatment. Appl. Biochem. Bio technol. 162, 62 74 (2010). 336. Jiang, X. et al. Fractionation and characterization of kraft lignin by sequential precipitation with various organic solvents. ACS Sustain. Chem. Eng. 5, 835 842 (2017). 337. Bhalla, A. et al. Engineered lignin in poplar bi omass facilitates Cu - catalyzed alkaline - oxidative pretreatment. ACS Sustain. Chem. Eng. 6, 2932 2941 (2018). 338. Li, Z. et al. Catalysis with CuII(bpy) improves alkaline hydrogen peroxide pretreatment. Biotechnol. Bioeng. 110, 1078 1086 (2013). 339. Luterbacher, J. S. et al. Nonenzymatic Sugar Production from Biomass Using Biomass - - Valerolactone. Science 343, 277 280 (2014). 340. Das, A. et al. Lignin conversion to low - molecular - weight aromatics via an aerobic oxidation - hydrolysis sequence: C omparison of different lignin sources. ACS Sustain. Chem. Eng. 6, 3367 3374 (2018). 341. Rinaldi, R. A Tandem for lignin - first biorefinery. Joule 1, 427 428 (2017). 342. Huang, X., Zhu, J., Korányi, T. I., Boot, M. D. & Hensen, E. J. M. Effective release o f lignin fragments from lignocellulose by lewis acid metal triflates in the lignin - first approach. ChemSusChem 9, 3262 3267 (2016). 343. Zeng, Y., Zhao, S., Yang, S. & Ding, S. - Y. Lignin plays a negative role in the biochemical process for producing lignoc ellulosic biofuels. Curr. Opin. Biotechnol. 27, 38 45 (2014). 344. Yang, B. & Wyman, C. E. Effect of xylan and lignin removal by batch and flowthrough pretreatment on the enzymatic digestibility of corn stover cellulose. Biotechnol. Bioeng. 86, 88 98 (2004 ). 345. Öhgren, K., Bura, R., Saddler, J. & Zacchi, G. Effect of hemicellulose and lignin removal on enzymatic hydrolysis of steam pretreated corn stover. Bioresour. Technol. 98, 2503 2510 (2007). 346. Crowe, J. D., Zarger, R. A. & Hodge, D. B. Relating na noscale accessibility within plant cell walls to improved enzyme hydrolysis yields in corn stover subjected to diverse pretreatments. J. Agric. Food Chem. 65, 8652 8662 (2017). 347. Williams, D. L., Crowe, J. D., Ong, R. G. & Hodge, D. B. Water sorption in pretreated grasses as a predictor of enzymatic hydrolysis yields. Bioresour. Technol. 245, 242 249 (2017). 348. Bär, J. et al. Deconstruction of hybrid poplar to monomeric sugars and aromatics using ethanol organosolv fractionation. Biomass Convers. Biore finery 1, 1 - 12 (2018). 177 349. Pan, X. et al. Biorefining of softwoods using ethanol organosolv pulping: Preliminary evaluation of process streams for manufacture of fuel - grade ethanol and co - products. Biotechnol. Bioeng. 90, 473 481 (2005). 350. Hames, Bonn ie, et al. "Preparation of samples for compositional analysis." Laboratory Analytical Procedure (LAP) 1617 (2008). 351. Lan, T. Q., Lou, H. & Zhu, J. Y. Enzymatic saccharification of lignocelluloses should be conducted at elevated pH 5.2 6.2. BioEnergy Res . 6, 476 485 (2013). 352. Williams, D. L. & Hodge, D. B. Impacts of delignification and hot water pretreatment on the water induced cell wall swelling behavior of grasses and its relation to cellulolytic enzyme hydrolysis and binding. Cellulose 21, 221 235 (2014). 353. - O - 4 dimers on metal surfaces in vacuum and solvated environments. In preparation. 354. Macala, G. S. et al. Hydrogen transfer from supercritical methanol over a solid base catalyst: A Model for lignin depolymerization. ChemSusChem 2, 215 217 (2009). 355. Selig, M. J. et al. Deposition of lignin droplets produced during dilute acid pretreatment of maize stems re tards enzymatic hydrolysis of cellulose. Biotechnol. Prog. 23, 1333 1339 (2007). 356. Piccinin, S., Stampfl, C. & Scheffler, M. First - principles investigation of Ag - Cu alloy surfaces in an oxidizing environment. Phys. Rev. B 77, 075426 (2008). 357. Wang, J . & Wang, S. - Q. Surface energy and work function of fcc and bcc crystals: Density functional study. Surf. Sci. 630, 216 224 (2014). 358. Mills, K. C. & Su, Y. C. Review of surface tension data for metallic elements and alloys: Part 1 Pure metals. Int. Ma ter. Rev. 51, 329 351 (2006). 359. Hong, S., Shin, Y. - H. & Ihm, J. Crystal shape of a nickel particle related to carbon nanotube growth. Jpn. J. Appl. Phys. 41, 6142 (2002). 360. Abrikosov, I. A., Ruban, A. V., Skriver, H. L. & Johansson, B. Calculated ori entation dependence of surface segregations in Pt 50 Ni 50 . Phys. Rev. B 50, 2039 2042 (1994). 361. Tepesch, P. d. & Quong, A. a. First - - alumina (0001) surfaces energies with and without hydrogen. Phys. Status Solidi B 217, 377 387 (2000). 362. McHale, J. M., Navrotsky, A. & Perrotta, A. J. Effects of increased surface area and chemisorbed H 2 - Al 2 O 3 - Al 2 O 3 . J. Phys. Chem. B 101, 603 613 (1997). 178 363. Costa, D., Sharkas, K., Islam, M . M. & Marcus, P. Ab initio study of the chemical states of water on Cr 2 O 3 (0001): From the isolated molecule to saturation coverage. Surf. Sci. 603, 2484 2493 (2009). 364. Rohr, F. et al. Strong relaxations at the Cr 2 O 3 (0001) surface as determined via low - energy electron diffraction and molecular dynamics simulations. Surf. Sci. 372, L291 L297 (1997). 365. Wang, X. - G. et al. The hematite - Fe 2 O 3 (0001) surface: evidence for domains of distinct chemistry. Phys. Rev. Lett. 81, 1038 1041 (1998). 366. Mazeina, L. & Navrotsky, A. Enthalpy of water adsorption and surface enthalpy of goethite - - Fe 2 O 3 ). Chem. Mater. 19, 825 833 (2007). 367. Scherson, Y. D., Aboud, S. J., Wilcox, J. & Cantwell, B. J. Surface structure and reactivity of rhodium oxide. J. Phys. Chem. C 115, 11036 11044 (2011). 368. Broqvist, P., Grönbeck, H. & Panas, I. Surface properties of alkaline earth metal oxides. Surf. Sc i. 554, 262 271 (2004). 369. Brunauer, S., Kantro, D. L. & Weise, C. H. The surface energies of calcium oxide and calcium hydroxide. Can. J. Chem. 34, 729 742 (1956). 370. Gilman, J. J. Direct measurements of the surface energies of crystals. J. Appl. Phys . 31, 2208 2218 (1960). 371. Rohrbach, A., Hafner, J. & Kresse, G. Molecular adsorption on the surface of strongly correlated transition - metal oxides: A case study for CO/NiO(100). Phys. Rev. B 69, 075413 (2004). 372. Skorodumova, N. V., Baudin, M. & Herma nsson, K. Surface properties of CeO 2 from first principles. Phys. Rev. B 69, 075401 (2004). 373. Hayun, S., Ushakov, S. V. & Navrotsky, A. Direct measurement of surface energy of CeO 2 by differential scanning calorimetry. J. Am. Ceram. Soc. 94, 3679 3682 ( 2011). 374. Wang, G. et al. Density functional studies of zirconia with different crystal phases for oxygen reduction reaction. RSC Adv. 5, 85122 85127 (2015). 375. Ballabio, G., Bernasconi, M., Pietrucci, F. & Serra, S. Ab initio study of yttria - stabilized cubic zirconia surfaces. Phys. Rev. B 70, 075417 (2004). 376. Bendavid, L. I. & Carter, E. A. First - Principles predictions of the structure, stability, and photocatalytic potential of Cu 2 O Surfaces. J. Phys. Chem. B 117, 15750 1 5760 (2013). 377. Evarestov, R. A., Bandura, A. V. & Proskurov, E. V. Plain DFT and hybrid HF - DFT LCAO calculations of SnO 2 (110) and (100) bare and hydroxylated surfaces. Phys. Status Solidi B 243, 1823 1834 (2006). 179 378. Perron, H. et al. Optimisation of accurate rutile TiO 2 (110), (100), (101) and (001) surface models from periodic DFT calculations. Theor. Chem. Acc. 117, 565 574 (2007). 379. Vittadini, A., Selloni, A., Rotzinger, F. P. & Grätzel, M. Structure and energetics of water adsorbed at TiO 2 Anata se (101 ) and (001 ) surfaces. Phys. Rev. Lett. 81, 2954 2957 (1998). 380. Synthesis and crystal structure of 2H CuAlO 2 : Zeitschrift für Kristallographie - Crystalline Materials. (1983). Available at: https://www - degruyter - com.proxy2.cl.msu.edu/view/j/zkri. 1983.165.issue - 1 - 4/zkri.1983.165.14.313/zkri.1983.165.14.313.xml. (Accessed: 28th July 2017) 381. Crystal structure of copper(I) chromium(III) oxide, 2H - CuCrO 2 : Zeitschrift für Kristallographie - Crystalline Materials. (1996). Available at: https://www - de gruyter - com.proxy2.cl.msu.edu/view/j/zkri.1996.211.issue - 7/zkri.1996.211.7.481/zkri.1996.211.7.481.xml. (Accessed: 28th July 2017) 382. Effenberger, H. Structure of hexagonal copper(I) ferrite. Acta Crystallogr. C 47, 2644 2646 (1991). 383. Kandpal, H. C. & Seshadri, R. First - P rinciples electronic structure of the delafossites ABO 2 (A=Cu, Ag, Au; B=Al, Ga, Sc, In, Y): evolution of d10 d10 interactions. Solid State Sci. 4, 1045 1052 (2002). 384. Jain, A. et al. Commentary: The Materials Project: A materials genome approach to accelerating materials innovation. APL Mater. 1, 011002 (2013). 385. Mason, S. E., Iceman, C. R., Trainor, T. P. & Chaka, A. M. Density functional theory study of clean, hydr ated, and defective alumina (1 02) surfaces. Phys. Rev. B 81, 1 25423 (2010). 386. Lo, C. S., Tanwar, K. S., Chaka, A. M. & Trainor, T. P. Density functional theory study of the clean and hydrated hematite (1 02) surfaces. Phys. Rev. B 75, 075425 (2007). 387. Batyrev, I., Alavi, A. & Finnis, M. W. A b initio calculations on the Al 2 O 3 (0001) surface. Faraday Discuss. 114, 33 43 (1999). 388. Rohrbach, A., Hafner, J. & Kresse, G. Ab initio study of the (0001) surfaces of hematite and chromia: Influence of strong electronic correlations. Phys. Rev. B 70, 125426 (2 004). 389. Barbier, A., Stierle, A., Kasper, N., Guittet, M. - J. & Jupille, J. Surface termination of hematite at environmental oxygen pressures: Experimental surface phase diagram. Phys. Rev. B 75, 233406 (2007). 390. Batyrev, I. G. & Kleinman, L. In - plane relaxation of Cu(111) and Al(111)/ - Al 2 O 3 (0001) interfaces. Phys. Rev. B 64, 033410 (2001). 391. Guenard, P., Renaud, G., Barbier, A. & Gautier - - Al 2 O 3 (0001) surface relaxation and termination by measurements of crystal tru ncation rods. MRS Online Proc. Libr. Arch. 437, (1996). 180 392. Wang, X. - G., Chaka, A. & Scheffler, M. Effect of the environment on - Al 2 O 3 (0001) surface structures. Phys. Rev. Lett. 84, 3650 3653 (2000). 393. Filippetti, A. & Fiorentini, V. Magnetic ordering in CuO from first principles: A cuprate antiferromagnet with fully three - dimensional exchange interactions. Phys. Rev. Lett. 95, (2005). 394. Song, Y. - Y. & Wang, G. - C. A DFT Study and microkinetic simulation of propylene partial oxidation on CuO (111) and CuO (100) surfaces. J. Phys. Chem. C 120, 27430 27442 (2016). 395. Wang, Y. et al. Electronic structures of Cu 2 O, Cu 4 O 3 , and CuO: A joint exp erimental and theoretical study. Phys. Rev. B 94, 245418 (2016). 396. Soon, A., Söhnel, T. & Idriss, H. Plane - wave pseudopotential density functional theory periodic slab calculations of CO adsorption on Cu 2 O(111) surface. Surf. Sci. 579, 131 140 (2005). 397. Wang, Z. l. Nanobelts, nanowires, and nanodiskettes of semiconducting oxides from materials to nanodevices. Adv. Mater. 15, 432 436 (2003). 398. Burnside, S. D. et al. Self - Organization of TiO 2 nanoparticles in thin films. Chem. Mater. 10, 2419 2425 ( 1998). 399. Vittadini, A., Casarin, M. & Selloni, A. Chemistry of and on TiO 2 - anatase surfaces by DFT calculations: a partial review. Theor. Chem. Acc. 117, 663 671 (2007). 400. Christensen, A. & Carter, E. A. Adhesion of ultrathin ZrO 2 (111) films on Ni(11 1) from first principles. J. Chem. Phys. 114, 5816 5831 (2001). 401. Muñoz, M. C., Gallego, S., Beltrán, J. I. & Cerdá, J. Adhesion at metal ZrO 2 interfaces. Surf. Sci. Rep. 61, 303 344 (2006). 402. Teufer, G. The crystal structure of tetragonal ZrO2. Acta Crystallogr. 15, 1187 1187 (1962). 403. Ackermann, R. J., Rauh, E. G. & Alexander, C. A. Thermodynamic properties of ZrO 2 (g). High Temp. Sci. 7, 304 316 (1975). 404. Bordier, G. & Noguera, C. Electronic structure of a metal - insulator interface: Towards a theory of nonreactive adhesion. Phys. Rev. B 44, 6361 6371 (1991). 405. Nørskov, J. K., Studt, F., Abild - Pedersen, F. & Bligaard, T. The electronic factor in heterogeneous catalysis. in Fundamental concepts in heterogeneous catalysis 114 137 (John Wiley & Sons, Inc, 2014). 406. Goniakowski, J., Mottet, C. & Noguera, C. Non - reactive metal/oxide interfaces: from model calculations towards realistic simulations. Phys. Status Solidi B 243, 2516 2532 (2006).