National Taiwan University of Science and Technology Department of Chemical Engineering PhD Thesis ID: D9706815 Computational Design of ZnO-Based Catalysts for Chemical Systems Author: Vo Thanh Cong Advisor: Professor Jyh-Chiang Jiang Spring Semester July 2013     G N O C H N A H T O V s t s y l a t a C d e s a B Os nm Ze fs t o y nS gl i a s c e i Dm le a h nC or i t o a f t u p m o C 博士論文 國立台灣科技大學 化 學 工 程 系 101 Acknowledgments I have benefited from numerous people and many facilities during my graduate study at the National Taiwan University of Science and Technology First, I would like to acknowledge my advisor, Professor Jyh-Chiang Jiang, whose enthusiasm and expertise were greatly appreciated I am very grateful for many current and former members of my lab (T2-513) for their support Of special note are Dr Wang, Dr Ni, and Mr Hung I spent a wonderful student life in the University of Science and Technology-Taiwan with numerous friends They supported lots information for me to understand the good living at school At last, I would like to thank my loving parents, beautiful wife, for their endless support This work has been supported by the National Science Council, Taiwan (NSC 99-2113-M-011-001-MY3) and the National Center of HighPerformance Computing (NCHC) for computer time and facilities Abstract As a semiconductor with a hexagonal wurtzite crystal structure and wide direct band gap, zinc oxide (ZnO) has a wide range of technological uses It is widely known that, in combination with other metal particles (e.g., Na, Cu, Pd and Ti) at the surface, ZnO, in general, has better selectivity and catalytic performance Thus, in this study we investigate the pure ZnO and modification of ZnO (1010) surface by doping and depositing different types of metals in attempt to design better catalysts for several important chemical reactions including; (1) methanol decomposition on ZnO surface; (2) CO oxidation on undoped and Ti doped ZnO surface; and (3) the water gas-shift reaction (WGS) on 2Cu deposited ZnO surface To detail, the adsorption and reaction mechanisms of methanol decomposition, CO oxidation, and water gas shift on pure and modified ZnO(1010) surfaces have been investigated using density functional theory (DFT) slab calculations, respectively In addition, the effect of the adsorption bonding between adsorbates and surfaces are studied also using density of states (DOS) and electron density difference (EDD) contour plots The understanding on these systems will help to shed more light on how to design better catalytic materials for different chemical systems Keywords: Density Functional Theory (DFT); ZnO 1010 ; CO Oxidation, Methanol decomposition, Water Gas-Shift Reaction Table of Contents Abstract 2  Chapter 1: Introduction 15  1.1.  Fuel cell 15  1.1.1.  Background 15  1.1.2.  Fuel cell principle and fuel cell types 16  1.1.3.  The hydrogen generation via methanol 19  1.1.4.  CO oxidation 21  1.1.5.  Water gas shift reaction (WGSR) 22  1.2.  Catalytic performances 23  1.2.1.  Catalyst development 23  1.2.2.  Bravais-miller index - case of ZnO structures 25  1.3.  This research 26  Chapter 2: Computational Details 28  2.1.  Theoretical background 28  2.2.  Method and parameters in this work 34  Chapter 3: Surface Models 37  3.1.  ZnO (1010) surface 37  3.2.  Ti doped ZnO (1010) surface 41  3.3.  Two Cu atoms deposited on ZnO (1010) surface 44  Chapter 4: Methanol Decomposition on ZnO (1010) Surface 49  4.1.  Methanol adsorption 49  4.2.  Electron density difference analysis 51  4.3.  CH3OH decomposition on ZnO (1010) surface 53  4.4.  Conclusions 59  Chapter 5: CO Oxidation on Undoped and Ti Doped on ZnO (1010) Surface 61  5.1.  CO adsorption on undoped ZnO surface 61  5.2.  CO adsorption on Ti doped ZnO surface 68  5.3.  Reaction mechanisms 70  5.3.1.  Undoped ZnO (1010) surface 71  5.3.2.  Ti doped ZnO (1010) surface 74  5.3.3.  Restored surface 75  5.4.  Conclusions 77  Chapter 6: Water Gas Shift Reaction on 2Cu Deposited ZnO (1010) Surface 78  6.1.  Adsorption of reactants, intermediates, and products on the 2Cu/ZnO (1010) surface 79  6.2.  Conversion of CO and H2O on the 2Cu/ZnO (1010) surface 85  6.3.  Desorption products 95  6.4.  Conclusions 98  Index of Figures Figure Chart to summarize the applications and main advantages of fuel cells of different types, and in different applications 17  Figure Wurtzite structure of ZnO 26  Figure The ZnO (1010) surface model: two-dimensional view with 15Å vacuum slab and six layers representation where the first two upper and the four bottom layers in the (1010) direction The blue and red balls are Zn and O atoms, respectively - 35  Figure A side view of relaxation and reconstruction on ZnO (1010) surface In which Figure 4(a) shows six layer structure; Figure 4(b) is an extraction (bordering by the dashed circle) from Figure 4(a) to describe the relaxed surface Compared to its bulk structure, the surface Zn-O bond tilts, marked with parameter θ and the bond linked between the surface zinc and oxygen contracts shown by d 38  Figure The side-view of the (2x2)-ZnOsupercell of Ti-doped systems Only first two outermost layers in the direction (1 010 ) are shown More detailed description about the tilt angle (α) can be found in the work of Wander 43  Figure Local density of state (LDOS) analysis for the surface oxygen bonded to the undoped ZnO surface (grey line) and Ti doped ZnO surface (black line) 44  Figure The most stable site (site E from Table S2, see in appendix) was chosen for calculations in WGSR In which, (a), (b) noted a top view and site view of structural site E The black circle of in each (a) and (b) noted that different positions of Cu-site and IF-site on 2Cu/ZnO (1 010 ) surface - 46  Figure The electronic properties of 2Cu atoms adsorbed on ZnO (1 010 ) surface Figure 8(a) shows DOS analysis of 2Cu adsorption on the surface in which the top box is states of 2Cu atoms before (grey line) and after adsorption (dash line), and the bottom box notes the states of ZnO before (grey line) and after (dash line) adsorption; Figure 8(b) presents EDD contour plot of 2Cu adsorption on the surface, where the dash-line is represented Cu electrons loss, and solid line as electrons accumulation of adsorbate-surface - 48  Figure Configuration of CH3OH adsorbed on ZnO (1010) surface at three different adsorption sites All bond distances (black-arrow line) are in Å Shaded white spheres represent the hydrogen, the grey ones are the carbon atoms 49  Figure 10 Electron density difference contour plot of CH3OH adsorbed on ZnO (1010) surface at (a) site A, (b) site B, and (c) site (C) models In which, the dashlines represents the loss of electron and the solid-lines show electron accumulation 52  Figure 11 In each box, the top and bottom shown transition and final states structures of CH3OH decomposition on ZnO (1010) surface; (a), (b), and (c) present O-H, C-H, and C-O bond breaking of site A, respectively, (d), (e) as O-H-C, and C-O bond breaking of site C - 55  Figure 12 Calculated activation barrier and reaction energies (in eV) characterizing C-O, C-H, and O-H bond breaking of MeOH on ZnO (1010) substrates with respect to the energy calculated for the corresponding MeOH adsorption 56  Figure 13 Reaction pathway of CH3OH decomposition of site A on clean ZnO (1010) surface 57  Figure 14.The most stable structures of CO adsorption on the undopedZnO (1010) surface (top) and the Ti-doped (1010) surface (bottom) - 62  Figure 15 The other less stable structures of CO adsorbed on the undoped ZnO (1010) surface 63  Figure 16 Local density of state (LDOS) analysis of CO adsorption bounded to surface (a) LDOS projected on a CO molecule in gas phase; (b), (c) LDOS projected CO adsorption on undoped ZnO and Ti doped ZnO surface, respectively - 64  promising step for extra hydrogen production This reaction happens as water gas shift reaction CO + H2O  CO2 + H2, ∆H = -41.2 kJ/mol In this study, the mechanism of water gas shift reaction (WGSR) is mainly discussed The WGSR refers to the moderately exothermic reaction between carbon monoxide (CO) with steam (H2O) to form carbon dioxide (CO2) and hydrogen (H2) It is considered a preferable method in CO removal as well as extra H2 production It requires low temperatures to achieve a high equilibrium conversion of CO the equation of WGSR is following: CO + H2O → CO2 + H2, ∆H = -41.2 kJ/mol (6) In fact, WGSR is carried out in two scale of temperature corresponding with catalytic process The first is a high temperature step at 350-500oC with iron-oxide based catalyst promoted with chromium oxide for the favorable reaction kinetics The other which is also this research’s objective refers a low temperature step at 180-250oC with copper-zinc oxidealuminum oxide supported catalyst favoring higher CO conversion Previous kinetic experiments34, 35 indicate that the rate-determining step of WGS process is water dissociation, therefore, we force on water gas shift reaction on different surface in this work 1.2 Catalytic performances 1.2.1 Catalyst development The current technical goals and challenges for fuel processor catalysts are the development of very active, poison-resistant materials that will result in small catalytic volumes 23 and reduced start-up times, durability under steady-state and transient conditions at the required temperatures, cost reductions, and versatility to variations in fuel/feed composition Methanol steam reforming over Cu-based catalysts was originally thought to have involved decomposition followed by water gas shift reaction However, in recent years, there is much evidence to suggest another pathway including a methyl formate intermediate The presence of methanol methoxy reacts to produce methyl formate, which has been shown to be the rate-determining step in methanol steam reforming.36-38 While some agreement exists in describing the existence of formate intermediates, involving a direct CO2 product pathway, there appears to be some discrepancy in explaining the involvement of the decomposition and WGSR.37, 38 Whatever the source of CO, measuring must be taken to minimize it as much as possibly for fuel processing applications where CO is poisonous to the downstream fuel cell Although mechanistic arguments are still in debate, it is generally observed that CO can be minimized by decreasing the contact time and decreasing the temperature, which acts to suppress CO thermodynamically.39 It is generally agreed that the active component on the CuO/ZnO/Al2O3 catalyst for any of the reactions, including methanol synthesis, decomposition, or reforming, is copper A good catalyst formulation contains well dispersed copper crystallites.8 Generally, catalysts with high copper content give higher conversion and selectivities.31 The role of ZnO is regarded to be relatively minimal, but it is needed as a textural support in segregating the Cu, which is highly susceptible to sintering However, promotional effects of ZnO additives on Cu for the steam 24 reforming reaction have been reported.40 The use of alumina creates a high surface area support which serves to increase copper dispersion and decrease the susceptibility to sintering.41 Just as there is dispute concerning the details for the mechanism, there is also dispute concerning the oxidation state of the active components It is generally agreed that there is an optimum balance between metallic Cu0 and oxidized CuI for maximum activity/selectivity and this is a function of not only the catalyst preparation and composition but also the feed and reaction conditions Keeping an optimum oxidization state is an important feature of the commercial CuO/ZnO/Al2O3 This suggestion alone has allowed researchers to question the use of such a catalyst system in a fuel processing environment where changes in oxidizing condition are of concern.42 However, the biggest problem with the Cu-based catalyst is the tendency for copper crystallites to readily sinter at temperatures > 280 °C.43 Hughs gave the following increasing order of stability for metals: Ag < Cu < Au < Pd < Fe < Ni < Co < Pt < Rh < Ru < Ir < Os < Re.43, 44 In this analysis lies the rationale for why copper-based catalysts are more susceptible to thermal sintering This explains why all modern copper catalysts contain one or more metal oxides to minimize thermal sintering.44 1.2.2 Bravais-miller index - case of ZnO structures 25 Figure Wurtzite structure of ZnO.45 As shown in Figure 2, ZnO is hexagonal structure Miller indices are a notation system in crystallography for planes and directions in crystal lattices In particular, a family of lattice planes is determined by three intergers ℓ, m, and n, the Miller indices They are written (hkl), and each index denotes a plane orthogonal to a direction (h, k, l) in the basis of the reciprocal lattice vectors With hexagonal, it is possible to use the Bravais-Miller index which has numbers (h k i l), i = - (h + k) (7) Here, h, k and l are identical to the Miller index, and i is a redundant index This four-index scheme for labeling planes in a hexagonal lattice makes permutation symmetries apparent For example: (110) = (1120) and (120) = (1210) (8) 1.3 This research Our goal in this thesis, therefore, is simply to understand ZnO-catalyzed and modified ZnO-catalyzed systems, for example such as doping or depositing From the theoretical simulations, the adsorption and reaction mechanisms of methanol decomposition, CO oxidation, and water gas shift on pure and modified ZnO(1010)–catalyzed systems have been investigated using density functional theory (DFT) slab calculations, respectively This research is organized as following: the first, we give a brief description of the theoretical background, then, introduce 26 computational details in chapter The secondly, in order to pure ZnO surface, there are two types of 2Cu/ZnO(1010) and Ti doped ZnO(1010) surfaces are examined, respectively in chapter The finally, the results, discussion, and conclusions of reaction in each chapter on surfaces are detailed, for example: in Chapter for methanol decomposition, chapter for CO oxidation, and chapter for water gas shift reaction Specially, we also characterized the details of the interactions between adsorbates and the surface by carrying out the analysis of the electron density difference (EDD) and density of states (DOS) for all of adsorption reaction on surfaces 27 Chapter 2: Computational Details 2.1 Theoretical background46-48 Ab initio is Latin for “from the beginning” and indicates a calculation based on fundamental principles An ab initio (or first principles) calculation uses the correct Hamiltonian and does not use experimental data other than the values of the fundamental physical constant A Hartree-Fock SCF calculation seeks the anti-symmetrized product  of one-lectron functions that minimizes ˆ d   *H , where Hˆ is the true Hamiltonian, and is thus an ab initio calculation The term ab initio should not be interpreted to mean “100% correct.” An ab initio SCF MO calculation uses the approximation of taking  as an anti-symmetrized product of oneelectron spin-orbitals and uses a finite basis set In contrast, Semi-empirical molecular quantummechanical methods use a simpler Hamiltonian than the correct molecular Hamiltonian and use parameters whose values are adjusted to fit experimental data or the results of ab initio calculations An example is the Hückel MO treatment of conjugated hydrocarbons, which uses a one-electron Hamiltonian and takes the bond integrals as adjustable parameters rather than quantities to be calculated theoretically The molecular-mechanics method is not a quantummechanical method and does not use a molecular Hamiltonian operator or wave function Instead, it views the molecule as collection of atoms held together by bonds and expresses the molecular energy in terms of force constants for bond bending and stretching and other parameters The density-functional method does not attempt to calculate the molecular wave 28 function but calculates the molecular electron probability density  and calculates the molecular electron energy from  In 1964, Pierre Hohenberg and Walter Kohn proved that for molecules with a nondegenerate ground state, the ground state molecular energy, wave function, and all other molecular electronic properties are uniquely determined by the ground-state electron probability density 0(x, y, z), a function of only three variables.49 This called the Hohenberg—Kohn theorem One says that the ground-state electronic energy E0 is a functional relation Densityfunctional theory (DFT) attempts to calculate E0 and other ground-state molecular properties from the ground-state electron density 0 To turn the Hohenberg—Kohn theorem to a practical tool, the Hohenberg—Kohn variational theorem was proved Hohenberg and Kohn proved that for every trial density function tr(r) that satisfies  tr (r )dr  n and tr(r) ≥ for all r, the following inequality holds: E0 ≤ Ev(tr) Since E0 = E0(tr), where 0 is the true ground-state electron density, the true ground-state electron density minimizes the energy functional Ev(tr) If we know the ground-state electron density 0(r), the Hohenberg-Kohn theorem tells us that it is possible in principle to calculate all the ground-state molecular properties from 0, without having to finding the molecular wave function In the traditional quantum-mechanical approach, one first finds the wave function and then finds  by integration The Hohenberg-Kohn theorem does not tell us how to calculate E0 from 0, nor does it tell us how to find 0 without first finding the wave function A key step toward these goals was taken in 1965 when Kohn and Sham devised a practical method for finding 0 and for finding E0 from 0.50 The Kohn-Sham 29 (KS) method is capable, in principle, of yielding exact results, but because the equations of this method contain an unknown functional that must be approximated, the KS formulation of DFT yields approximate results Hohenberg and Kohn showed that if  varies extremely slowly with position, then Exc[r] is accurately given by E xcLDA [  ]    (r ) xc (  ) dr Taking the functional derivative of E xcLDA , one finds the exchange-correlation potential  xcLDA  xcLDA  δE xcLDA  (  )   xc (  (r ))   (r ) xc δ  Kohn and Sham suggested the use of these two equations as approximations to Exc and xc, a procedure that is called the local density approximation (LDA) For open-shell molecules and molecular geometries near dissociation, the local-spin-density approximation (LSDA) gives better results than the LDA Whereas in the LDA, electrons with opposite spins paired with each other have the same spatial KS orbital, the LSDA allows such electrons to have different spatial KS orbitals Despite the fact that  in a molecule is not a slowly varying function of position, the LSDA works surprisingly well for calculating molecular equilibrium geometries, vibrational frequencies, and dipole moments, even for transition-metal compounds However, calculated LSDA molecular atomization energies are very inaccurate Accurate dissociation energies require functionals that beyond LSDA 30 The LDA and LSDA are based on the uniform-electron-gas model, which is appropriate for a system where  varies slowly with position The integrand in the expression for E xcLDA is a function of only , and the integrand in E xcLSDA is a function of only  and  Functionals that go beyond the LSDA aim to correct the LSDA for the variation of electron density with position This can be done by including the gradients of  and  in the integrand Thus E xcGGA [   ,   ]   f (   (r ),   (r ),   (r ),   (r ))dr where f is some function of the spin densities and their gradient The letters GGA stand for generalized-gradient approximation In all of the calculations in this work, we will use the plane wave pseudopotential approach to solving the Kohn-Sham equations This involves using a plane wave basis set to represent the orbitals, and pseudopotentials to represent the nuclei and core electrons In realistic systems, there are around 1020 atoms in a cubic millimeter, and there are no any numerical method could treat a system with so many atoms However, at this scale, the systems are often repeating; that is, the systems are infinite but periodic One can choose periodic boundary conditions to reduce this problem to the study of a finite system By using the periodic boundary condition, the system is contained within a supercell which is then replicated periodically throughout space A plane wave basis set must be used in conjunction with periodic boundary conditions For any continuous periodic orbital with Bloch wave vector k, the plane wave basis function may be written as 31 k (r )   c k (G )e i( k G ) r G where the functions e i( k G ) r are plane waves, and the pre-factor of  preserves the normalization of the wavefunction The ck(G) are a set of complex of coefficients that constitute the reciprocal space representation of orbital, and the G are reciprocal lattice vectors The coefficients ck(G) can be obtained by a Fourier transform of an orbital from real to reciprocal space, that is ck (G )  FT[k (r)] The coefficients ck(G) for the lowest eigenvectors decreases exponentially with the kinetic energy (k+G)2/2 The selection of plane waves is determined by cut-off energy (Ecut): (k  G )  E cut One of the main advantages of using a plane wave basis set is that its accuracy can be easily controlled This is related to the fact that, when using such a basis set, we are making no assumptions about the final shape of the orbitals, other than that there is some scale below which they become smoothly varying However, this also leads to a major disadvantage of using a plane wave basis set, which is that the size of the basis set required for a given system is often far larger than would be required with a localized basis set This is because, in condensed matter systems, the orbitals tend to oscillate very rapidly in the vicinity of atomic nuclei, and are much more smoothly varying elsewhere In order to describe this rapid oscillation we must set very 32 large cut-off energy, so that we include plane waves with very short wavelengths But, since most of the space in the cell does not contain rapidly oscillating orbitals, most of the computational expense associated with all these plane waves effectively goes to waste A localized basis set can be tailored such that the basis functions themselves are rapidly oscillating in the vicinity of atomic nuclei and more smoothly oscillating elsewhere, so that the total number of basis functions required for the system is far smaller The use of pseudopotentials, in conjunction with plane waves, can dramatically reduce the magnitude of this problem Lower energy orbitals can often be considered represent core electrons These are electrons that are well localized around an atomic nucleus and whose properties not change significantly with the atom's “chemical environment” Orbitals representing electrons that are not core electrons oscillate very rapidly in the vicinity of atomic nuclei, but most of this oscillation can be put down to the fact that they have to be orthogonal to the core electrons The core electrons, and the potential due to the bare nuclear charge, are replaced by a fictitious potential that is defined such that the behavior of the valence electrons is not affected outside of some cutoff radius from the nucleus The pseudopotentials work by effectively removing core electrons from the calculation, the number of Kohn-Sham orbitals is reduced This reduces the memory required to store the orbitals, the time required to evaluate orbital-dependant quantities, and the time required to orthonormalise a set of orbitals In addition, after applying the pseudopotentials, there are no core-electrons to which valence electrons must be orthogonal This means that a lower cut-off energy can be used to represent the orbitals, and this lowering of the cut-off energy is typically a few orders of magnitude resulting in massive gains in efficiency 33 2.2 Method and parameters in this work In this thesis, all DFT calculations were performed with the Vienna ab initio Simulation Package (VASP).51-53 The generalized gradient approximation (GGA) was used with the functional described by Perdew.54, 55 Electron-ion interactions were investigated with the use of the projector augmented wave method and we performed non spin-polarized calculations for all of the structural optimizations.56, 57 Normal-mode analysis was performed to verify the nature of each of these stationary points The ZnO (1010) surface was modeled as a two-dimensional slab in a three-dimensional periodic cell A (2×2) surface slab has thickness of six layers in which the top three layers were relaxed and the bottom three layers were fixed during the calculations (as shown in Figure 3) In order to curtail the interaction between the slabs, we introduced a 15 Å of vacuum space in the z direction The calculations were carried out using the Brillouim zone sampled with (4×4×4) and (4×2×1) Monkhorst-Pack58 mesh k-points grid for ZnO bulk and all of ZnO (1010) surface calculations, respectively 34 Figure The ZnO (1010) surface model: two-dimensional view with 15Å vacuum slab and six layers representation where the first two upper and the four bottom layers in the (1010) direction The blue and red balls are Zn and O atoms, respectively The adsorption energy, Eads , was calculated as follows 35 Eads  Eadsorbate / ZnO  Eadsorbate  EZnO (9) Where, Eads is the adsorption energy of the adsorbate, CH3OH, on (1010) ZnO surface; Eadsorbate / ZnO is the total energy of the optimized adsorbate on ZnO, Eadsorbate is the total energy of the adsorbate, and EZnO is the total energy of the clean ZnO (1010) slab The nudged elastic band (NEB) method59-61 was applied to locate the each transition state (TS), and all the TSs had been characterized as real first-order saddle points 36 Chapter 3: Surface Models 3.1 ZnO surface Metal-oxide catalysts are attracted considerable attention owing to their potential applications in industry and academic research In the last years, several experimental and theoretical studies on metal oxides have been performed to increase their performance and reduce the energy required for their practical use.24, 62, 63Among the oxide catalysts, zinc oxide (ZnO) has become a frequently studied material in surface science because of its wide range of applications such as a basic material for varistors, as a chemical sensor in gas-detecting systems and as a catalyst for many important chemical processes on the industrial scale (e.g., de/hydrogenation and methanol synthesis/conversion).64-66 Among the four main ZnO surfaces, the non-polar 1010 surface has been the focus of several experimental and theoretical studiesdue to itsmostabundant and stable nature.65-67 The surface is obtained by cutting the crystal perpendicular to the hexagonal Zn and O layers The atomic planes perpendicular to the 1010 direction consist of equal numbers of zinc and oxygen ions and can be formed as rows of zincoxygen dimmers These dimmers are then bonded to dimmers in the next layer.67, 68 37 ... electronic systems of low power, running for long times Large numbers of 200-kW CHP systems in use Suitable for medium- to large-scale CHP systems, up to MW capacity Suitable for all sizes of CHP systems, ... investigate the pure ZnO and modification of ZnO (1010) surface by doping and depositing different types of metals in attempt to design better catalysts for several important chemical reactions... CO adsorption on undoped ZnO and Ti doped ZnO surface, respectively - 64  Figure 17.The analysis of partial density of state (PDOS) of s and p orbitals of CO before adsorption (clean) and