Transition Metals Thermal Crystal Physics: Cu-Sb-S, Cu-Li-Mg, Bi-Sn-Zn and Al-Fe-Ti

Transition Metals Thermal Crystal Physics: Cu-Sb-S, Cu-Li-Mg, Bi-Sn-Zn and Al-Fe-Ti José Jorge do Amaral Ferreira Dissertation presented to obtain the degree of DOCTOR IN ENGINEERING PHYSICS by the UNIVERSITY OF PORTO SUPERVISOR: Maria Helena Sousa Soares de Oliveira Braga, Assistant Professor 2012 1 2 ACKNOWLEDGEMENTS First of all, I would like to express here my gratitude to my supervisor, Prof. Helena Braga, for her endless support, guidance and advice throughout the time of my thesis. I would also like to acknowledge the discussions with Prof. Helena Braga; her expertise on Materials Physics was very helpful. I would like to acknowledge Prof. Mario Machado Leite, Director of my laboratory for providing much of the material support for this work and for his enormous motivation and encouragement. I would like to thank the members of the jury for their interest in this present work and for their in-depth reading of the manuscript. I thank my colleagues of UCTM-Lab LNEG for their warm support. I would like to acknowledge FCT – Portugal and the FEDER - EU, for the PTDC/CTM/099461/2008 project Hydrogen Storage on New Metal Hydrides Based on the Cu-Li-Mg system 3 4 CONTENTS Abstract 7 Resumo 10 Acronyms 13 Chapter 1. INTRODUCTION 17 1.1. Transition metals and their applications 19 1.2. Theoretical tools supporting the materials study 23 1.2.1. The Kohn-Sham equations 23 1.2.2. Phonon calculations, the PHONON code 35 1.2.3. Planewave pseudopotential method, the VASP code 36 1.3. References 38 Chapter 2. Cu-Sb-S SYSTEM 45 2.1. Motivations 47 2.2. The Cu-Sb-S system and the Tetrahedrite, Cu12Sb4S13 50 2.3. References 57 Papers on the Cu-Sb-S system 61 2.4. First Principles Study of Copper Sulfides (for applications as photoconductors) 63 2.5. Phase transitions in the Cu-Sb-S System 69 Chapter 3. Cu-Li-Mg (H, D) SYSTEM 75 3.1. Introduction to Cu-Li-Mg system 77 3.2. Introduction to hydrogen storage 77 3.3. Hydrogen storage: a brief overview 78 3.4. Magnesium hydride 79 3.5. Cu-Mg, Ni-Mg and other MgH2 destabilizing systems 82 3.6. Lithium hydride 86 3.7. Neutron techniques associated with hydrogen solid storage 87 3.8. Hydrides of Cu and Mg intermetallic systems 91 3.9. The CuLi0.08Mg1.92 compound 92 3.10. Hydrogen storage in the Cu-Li-Mg-H(D) system 94 3.11. Conclusion 99 3.12. References 100 5 Papers on the Cu-Li-Mg system 109 3.13. A ternary phase in Cu-Li-Mg system 111 3.14. HT-XRD in the study of Cu-Li-Mg 119 3.15. Neutron Powder Diffraction and First-Principles Computational Studies of CuLixMg2-x (x  0.08), CuMg2, and Cu2Mg 125 3.16. Study of the Cu–Li–Mg–H system by thermal analysis 135 3.17. First Principles Calculations and Experiments to Determine the Hydrogenation Process of Cu-Li-Mg 143 Chapter 4. Bi-Sn-Zn SYSTEM 149 4.1. Introduction 151 4.2. References 153 Papers on the Bi-Sn-Zn system 155 4.3. Experimental Phase Diagram of the Ternary Bi-Sn-Zn 157 4.4. The experimental study of the Bi-Sn, Bi-Zn, and Bi-Sn-Zn systems 163 4.5. Thermodynamic assessment of the Bi–Sn–Zn System 175 4.6. The Behaviour of the Lattice Parameters in the Bi-Sn-Zn System 187 4.7. Phase field simulations in miscibility gaps 197 Chapter 5. Al-Fe-Ti SYSTEM 205 5.1. Introduction 207 5.2. Al-Fe-Ti system 208 5.3. References 213 Chapter 6. FINAL CONCLUSIONS AND FUTURE WORK 215 6.1. Final Conclusions 217 6.2. Future Work 219 6.3. Cu-(Fe,Sn)-S system 219 6.4. Cu-Li-Mg (H,D) system 222 6.5. Al-Fe-Ti system 223 6.6. References 224 6 ABSTRACT This thesis focuses four different systems containing at least one transition element. A transition element is an element whose atom it has an incomplete d sub-shell, or which can give rise to cations with an incomplete d sub-shell. This characteristic gives rise to some distinctive properties that were object of this study. All the following systems Cu-Sb-S, Cu-Li-Mg, Bi-Sn-Zn, and Al-Ti-Fe, were studied by means of experimental and/or theoretical techniques. Differential Scanning Calorimetry (DSC), Scanning Electron Microscopy (SEM) or, alternatively, Electron Probe Micro Analysis (EPMA) and X-ray diffraction (XRD). Additionally, X-ray Absorption (XANES) experiments were performed for Tetrahedrite structures of the Cu-Sb-S system, and Neutron Scattering (diffraction and spectroscopy) and hydrogen absorption studies were performed for the Cu-Li-Mg (H, D) system. The theoretical studies are of paramount importance, not only as a auxiliary technique used in the analysis of experimental data, but also to predict worth doing experiments and to tailor new materials. Density Function Theory (DFT), as implemented in VASP codes, was used to optimize the crystal structure of the phases in study. Electrical, mechanical, and thermodynamic properties were also studied. For the latter PHONON code was additionally used. Phase diagrams were furthermore assessed using Thermo-Calc code. Cu-Sb-S Tetrahedrite, Cu12Sb4S13 is one of the ternary phases of the Cu-Sb-S system. This phase is interesting from the mineralogical point of view since Cu can be replaced by impurities like Zn, Fe, Ag, Cd, Mn, In, Tl and Ge and by that being able to store these elements. Bond distortion is the result of these replacements. In this thesis, I’ve initiated the experimental and theoretical work leading to an understanding of the structural features conducing to this effect that is not verified in other phases like, for example, Chalcopyrite, CuFeS2. Phase diagram experimental studies (including XRD, DSC and EPMA) were performed to characterize the Cu-Sb-S system. 7 Cu-Li-Mg (H, D) The increasing need for materials that optimize the performance of fuel cells and batteries is becoming one of the major trends of Materials Research. In this work, I’ve contributed for the experimental and theoretical study of CuLi0.08Mg1.92H5 as a hydrogen storage material with applications in both fuel cells and Li batteries conversion electrodes. I was specially focused in studying the parent CuLi0.08Mg1.92 phase by means of XRD and presently I am complementing the neutron scattering studies on the Cu-Li-Mg-H system with XRD experiments. Furthermore, I am involved in the DFT and PHONON calculations developed for some of the phases of the mentioned system. Van’t Hoff plots were obtained by means of experimental hydrogen absorption experiments, the CuLi0.08Mg1.92H5 phase was characterized mostly by neutron diffraction and spectroscopy and DSC in synchrony with thermogravimetric analysis. The catalytic effect of the Cu-Li-Mg-H system over Li/LiH, Mg/MgH2 and Ti/TiH2 was moreover studied. The first part of this work was devoted to the study of the CuLi0.08Mg1.92 compound and its thermodynamic and structural characteristics. Bi-Sn-Zn Lead and lead-containing compounds are considered toxic substances due to their detrimental effect to the well-being of humans and the environment. Suitable development policy has been implemented by many countries around the world and, in order to protect the environment, the restriction of lead used in industry has been strongly promoted. Sn- Zn is one of the systems that has been used as solder as a substitute of the traditional Sn- Pb solder. We have studied the Sn-Zn and Bi-Sn-Zn systems experimentally. My main participation in these studies was related with the XRD studies at room and elevated temperatures. The Sn-Zn and Bi-Sn-Zn systems were furthermore studied by means of DSC and SEM. The system was reassessed and liquid miscibility gap of the Bi-Zn system was moreover studied using phase field. Microstructures were determined as a function of temperature and composition and phase diagram assessed parameters. 8 Al-Fe-Ti Both the Al-Fe and Al-Fe-Ti systems were experimentally studied. Nonetheless, I only present some of the results obtained for the Al-Fe system since the study of the Al-Fe-Ti system is still insipient and only contemplates experimental results (XRD, DSC and EPMA). With respect to Al-Fe, I have mainly contributed with XRD and EPMA data and/or analysis and first principles calculations, including phonon calculations. Even if we do not present the complete study in this thesis, the study of the Al-Fe system, centered on the ε high temperature phase, is becoming mature and we just need a final reflection over the whole data (including the immediate neighbor phases such as AlFe and AlFe2) to publish this work. 9 16 Chapter1-INTRODUCTION 17 18 Chapter1. INTRODUCTION 1.1. Transition metals and their applications The definition of the International Union of Pure and Applied Chemistry (IUPAC) for what a transition element is: "an element whose atom has an incomplete d subshell, or which can give rise to cations with an incomplete d sub-shell" (McNaughton & Wilkinson, 1997). Fig. 1.1 Transition metals in the periodic table of elements (Glenbard West AP Chemistry Wiki, 2012). The main characteristic of the transition elements, which can be referred as transition metals as well, is having a partially filled d sub-shell or to give rise to cations with an incomplete d sub-shell (e.g. Cu2+). The transition elements belong to three rows in the periodic table which correspond to the progressive filling of 3d, 4d, and 5d states. The presence of the d electrons changes the “picture” of bonding in these metals considerably. The transition metals usually have high melting points and several oxidation states; they usually form colored compounds and are often paramagnetic (McNaughton & Wilkinson, 1997). Fig. 1.2 Transition metals oxidation states. The most common is represented in bigger dots (Glenbard West AP Chemistry Wiki, 2012). 19 There is an extensive variety of materials that include transition metals. Such compounds, with transition metals can have different electronic behaviors. They can be conductors or semiconductors and more rarely insulators or superconductors. Materials having transition metals may have interesting magnetic, ferroelectric, antiferroelectric, and piezoelectric properties (Vaughan & Rosso, 2006), (Cramer & Truhlar, 2009). When an anion of sulfur in its lowest oxidation state of -2 is bonded to a transition metal, a transition-metal sulfide is formed. Transition-metal sulfides constitute an important class of inorganic compounds with diverse applications in industry, ranging from catalysis to lubrication, corrosion protection and photoconductivity. In important minerals containing sulfur, such as Chalcopyrite, Tetrahedrite (Fig. 1.3) and Blend, it is common to find some of latter properties. These minerals can be doped with other elements enhancing the above mentioned properties (Rignanese, 2005). Sulfides are also components of many thin-film devices and have been extensively investigated as part of the nanotechnology revolution. Fig. 1.3 Photograph of a Tetrahedrite mineral (School of Geology - University of Aristotle’s – Greece, 2012). 20 A cross-section of a ZnO-CdS-CIGS solar cell is shown in Fig. 1.4. From bottom to top: Kapton foil, layer of Mo with 0.5-1.5 μm thick working as positive electrode, roughly 1-2 μm of CIGS functionalized as the positive semiconductor, layer of 3-5μm CdS functionalized as the negative semiconductor, and a layer with 0.5-1.5 μm of ZnO working as negative electrode. In the top cell, two wire conductors based on Ni/Al alloy. Fig. 1.5 Photograph of park of thin film solar cell panels, (CNET, 2010) Since the beginning of the 21st century the increasing demand for renewable energies, due to the limited availability of fossil fuels and to environmental problems, resulted in an extraordinary request for photovoltaic materials. Fig. 1.4 Schematic representation of a sulfide thin-film solar cell deposited on Kapton substrate (Hepp et al, 2000) 21 Currently, the high cost of the materials applied in photovoltaic solar cells, (e.g. Ga and In) Fig. 1.5, is one of the main obstacles for their production and application to a larger scale. Ternary Chalcopyrite’s with the general formula AIBIIIX2 (A = Li, Na, Cu, Ag; B = Fe, Al, Ga, Ge, In; X = S, Se, Te; I is a +1 cation and III a +3 cation) are of considerable interest because of their potential optoelectronic applications as solar energy converters, nonlinear optical devices, Light Emitting Diodes (LED), and detectors. Chalcopyrite is a semiconductor. Polycrystalline solar cells with Cu(In,Ga)Se2 - Chalcopyrite (Shafarman & Zhu, 2000) absorber are reach up to 18.8 % efficiency. The importance and application of compounds based on transition-metal sulfides have many technological applications, ranging from lubrication to the use as catalysts in sulforeductive hydrotreating processes in the petroleum refining industry (Harris & Chianelli, 1984), (Topsoe et al., 1996), (Hobbs & Haffner, 1999) In addition, transition-metal sulfides have also a fundamental scientific interest. The sulfides of the transition metals are intermediate between the transition-metal oxides, whose properties are determined by strong electronic correlation effects (Ansisimov et al., 1991), and the transition-metal selenides showing a variety of electronically induced structural phase transitions, including the formation of incommensurate phases (di Salvo & McMillan, 1977). The combined scientific and technological importance of these materials has motivated a substantial research effort directed towards an understanding of their properties at an atomistic level. The electromagnetic properties of sulfide minerals are furthermore responsible for their contribution to geomagnetism and paleo-magnetism, since they become geophysical prospectors to be used as exploration tools in metalliferous ore deposits (Fig. 1.6). To the mineral technologist, these same properties provide methods for the separation of the metal-bearing sulfides from the associated waste minerals after mining and milling and before extraction of the metal by pyrometallurgical or hydrometallurgical treatment (Pearce et al., 2006). 22 Fig. 1.6 Chuquicamata sulfide ore deposit. Open pit in Andean Chile (CODELCO National Copper Corporation of Chile, 2012) In this doctoral thesis we will explore several systems including transition metals, from minerals to light weight hydrogen storage, passing through lead free solders. 1.2. Theoretical tools supporting the materials study A great approach for studying structural and electronic properties is the Density- Functional Theory (DFT). DFT is the most popular and robust theoretical approach currently available for solving the electronic structure of materials. Although far from an answer for all physical problems in this domain, no other theoretical approach has provided as much understanding of the electronic properties. DFT has proven to be capable of computing a multitude of properties of condensed matter with a reasonable accuracy (Michaelides & Scheffler 2012). 1.2.1 The Kohn-Sham e qua ti ons Hohenberg and Kohn, 1964, and Kohn and Sham, 1965, formulated a rather outstanding theorem which states that the total energy of a system such as a solid, surface or molecule depends only on the electron density of its ground state. In other words, one can express the total energy, E, of an atomistic system as a functional of its electron density, E [ ρ ]. E = E [ ρ ] (1.1) The idea of using the electron density as the fundamental entity of a quantum mechanical theory for describing matter was developed in the early days of quantum mechanics, 23 particularly by the work of Fermi, 1928 and Tho ma s, 1929, (Slater, 1951) . In the following decades, it was rather the Hartree-Fock approach Hartree, 1928; Fock, 1930; Fock, 1934 (Slater, 1951) which was developed and applied to small molecular systems. Calculations on realistic solid state systems were then out of reach. Later Slater, 1951, used ideas from the electron gas with the intention to simplify the Hartree-Fock theory to a point in which electronic structure calculations on solids became feasible. Slater's work, which led to the so-called Xα-method (Slater, 1974) has contributed enormously to the progress of electronic structure calculations. The Xα-approach, combined with other simplifications in the entitled Scattered-Wave method have driven heated disputes about the “ab initio” character of this approach, which was carried over into the comparison of density functional theory vs. Hartree-Fock based methods. Today’s density functional methods can be considered “ab initio” (Wimmer & Freeman, 2000). In solid-state systems, molecules, and atoms, the electron density,, is a scalar function defined at each point, r, in real space, ρ = ρ (r) (1.2) The electron density and the total energy, E, depend on the type and arrangements of the atomic nuclei. Therefore, E, can be expressed by, E = E [ ρ ( r ) , (R α )] (1.3) Conventionally ( Rα), represents the positions of all atoms of the element, α, in the system under consideration. Equation (1.2) is the key to the atomic-scale understanding of electronic, structural, and dynamic properties of matter. If (1.2) can be evaluated, then the equilibrium structure of a solid, the reconstruction of a surface, and the equilibrium geometry of molecules adsorbed on surfaces, can be predicted (Wimmer & Freeman, 2000). Moreover, the derivative of the total energy (1.3), with respect to the nuclear position, of an atom, is the absolute value of the force acting on that atom. This enables the efficient search for stable structures and, perhaps more importantly, the study of dynamical processes such as diffusion or the reaction of molecules on surfaces. 24 Most of the considerations presented here are based on the Born-Oppenheimer approximation in which it is assumed that the motions of the electrons are infinitely faster than those of the nuclei. This approximation rests on the fact that the nuclei are much more massive than the electrons, which allows us to say that the nuclei are nearly fixed with respect to electron motion (Sherrill, 2005). This means that the electronic structure is calculated for a fixed atomic arrangement and the atoms are then moved according to classical mechanics. In density functional theory, the total energy is decomposed into three parts: a kinetic energy, T0, an electrostatic or Coulomb energy, U, and exchange-correlation energy, Exc , E = T0 + U + Exc (1.4) The most straightforward term is the Coulomb energy, U. It is purely classical and contains the electrostatic energy arising from the Coulomb attraction between electrons and nuclei, the repulsion between all electronic charges, and the repulsion between nuclei. U = U en + U ee + U nn (1.5) With  =−∑  ∫() |  |  (1.6)  =∫∫()(󰆓) |󰆓|󰆒 (1.7)  =∑    󰆓     󰆓  (1.8) Where, e, is the elementary charge of a proton, and, Zα ,is the atomic number of the atom of α. The summations extend over all atoms and the integrations over all space. Once the electron density and the atomic numbers and positions of all atoms are known, equations (1.6)-(1.8) can be evaluated by using the techniques of classical electrostatics, were , , is the the potential energy between the electrons and nuclei, , is the potential energy 25 fitted to the exchange energy of inert gases. The explicit form of the functions f and g in Perdew's expression for the correlation energy is given in the original paper (Perdew, 1986). Modern DFT provides an extremely valuable tool for predicting structures, thermodynamic, mechanical and electronic properties of new materials for both finite and periodic systems. In theory, the electron affinity and ionization potential can be obtained exactly from the ground-state energy differences of the N+1 and N-1 electron systems versus the N electron system being N the number of electrons of the atom. In practice, for periodic structures, the band gap, is the difference between the ionization potential and the electron affinity, and it is obtained from the difference in the orbital energies of the Lowest Occupied Molecular Orbital (LUMO) corresponding to the conduction band minimum - and the Highest Occupied Molecular Orbital (HOMO) corresponding to the valence band maximum from a single calculation. It is well documented in (Xiao et al., 2011) that standard DFT exchange-correlation (XC) functionals, including various Generalized Gradient Approximations (GGA), dramatically underestimate the band gaps, Eg, for insulators due to the existence of a derivative discontinuity of the energy with respect to the number of electrons. The study of the materials based on transition-metal sulfides, using DFT, began with the work of (Raybaud et al., 1997), in which a complete research of the structural, cohesive and electronic properties of transition-metal sulfides in the Local Density Approximation (LDA) is made. Raybaud et al., 1997, demonstrated the overbinding tendency of the LDA (prediction of too small atomic volumes and too large cohesive energies) pronounced, in particularly, for the transition-metal sulfides. For the 4d and 5d orbitals, in transitionmetal sulfides, the overbinding is largely corrected by including non-local corrections in the form of a generalized gradient approximation (GGA) (Hobbs & Hafner, 1999), (Hafner, 2008). An approximate description of strong intra-atomic correlation effects is provided by the DFT+U. The physical idea behind the LDA+U or GGA+U schemes comes from the Hubbard Hamiltonian. In practical implementations the on-site two-electron integrals, which would appear in Hartree-Fock, are expressed in terms of two parameters. These are the Hubbard parameter U, which reflects the strength of the on-site Coulomb interaction, and the parameter J, which adjusts the strength of the exchange interaction. In the somewhat simplified, yet rotationally invariant method of (Dudarev et al., 1998) these two 32 parameters are combined into a single parameter Ueff = U–J (Loschen et al., 2007) approach in which on-site Coulomb and exchange interactions, described in an unrestricted Hartree–Fock approximation, are added to the DFT Hamiltonian (Lichtenstein et al., 1995). The DFT+U as formulated by (Dudarev et al., 1998), allows investigate the structural and electronic properties of several 3d transition-metal sulfides to be calculated using both LDA and GGA as a starting point. Using DFT+U, it is possible to deal more accurately, with electron correlations in transition metal and rare earth compounds. Its implementation within a PAW framework was developed by (Bengone et al., 2004). The main important concept of DFT+U is to address the on-site Coulomb interactions in the localized d or f orbitals with an additional Hubbard-type term (Hafner, 2009). When DFT fails to account for the strong Coulomb repulsion between electrons occupying narrow bands, which leads to an enhanced exchange splitting between occupied and empty eigen states, The DFT+U method attempts to resolve this defect by adding a Hubbard-type Coulomb repulsion to the DFT Hamiltonian (Hafner, 2009). The LDA functional already provides a good approximation for the calculation of properties in transition metal solids (Hathaway et al., 1985), and the GGA functional does not always necessarily improve the LDA results (Garcia et al., 1992). Some testes are made to evaluate the accuracy of the GGA PBE local functionals for the calculation of the unit cell volumes of Cu, Pd, W, Pt, and Au. The Mean Unsigned Errors (MUE) are 1.0% for LDA, 1.3– 2.9% for GGA-PBE (Kurth et al., 1999). In summary, according to (Kurth et al., 1999), LDA underestimates the volumes (over binding) whereas the GGA usually overestimate them (under binding). A similar effect was found for Cu, Rh, Pd, Ag, Ir, Pt, and Au with LDA and GGA. According to (Grabowski et al., 2007), the mean errors for the calculation of the bulk moduli are 19% for LDA and 8-34% for GGA (Kurth et al., 1999). Relevant is also the work of (Heyd et al., 2005), in which they show that accurate lattice constants with the screened Coulomb hybrid functional Heyd-Scuseria-Ernzerhof (HSE) can be calculated. For example, the lattice constants of ZnS, ZnSe, and ZnTe are overestimated by HSE by an average of 0.7%, (Heyd et al., 2005), whereas the local GGAPBE functional overestimates the same lattice constants by an average of 1.4% (Heyd et 33 al., 2005) . This is important because HSE predicts much more accurate band structures than do local functionals. Chevier et al., 2010, compare the accuracy of conventional semilocal DFT, the DFT+U method and the Heyd-Scuseria-Ernzerhof (HSE06) (Heyd et al., 2005) hybrid functional for structural parameters, redox reaction energies, and formation energies of transition metal compounds. As an example, (Sun et al., 2008) use the LDA with PBE to calculate the equation of state of Pt. The Helmholtz energy was determined as a function of the lattice constants and temperature by setting it equal to a sum of three terms: the ground-state electronic energy, the free energy of thermally excited electronic states and the vibrational free energy of the lattice (Li et al., 2007), (Sun et al., 2008). Presently hybrids functionals like PBE0 (Perdew et al., 1996) and HSE (Heyd et al., 2003) can give more accurate band gap values, using the orbital-energy gap expression instead of using local functionals like PBE. For corroborating this assumption, (Heyd et al., 2005) presents an example, for ZnS, ZnSe, and ZnTe, in which HSE orbital energies underestimate the band gaps by an average of 0.27 eV while for the same three materials, the local PBE functional underestimates the band gaps by an average of 1.46 eV . (Heyd et al., 2005). The ability to calculate band gaps more accurately allows us to study interesting problems such as the charge states of interstitial hydrogens in oxides like TiO2, SrTiO3, PbTiO3, ZrO2, SrZrO3, ZrSiO4, HfO2, ZnO and CdO (Grabowski et al., 2007), (Cramer & Truhlar, 2009). Theoretical studies are especially useful for studying materials under conditions that are hard to reproduce experimentally, for example, high-pressure conditions in the earth’s mantle, and the GGA-PBE density functional was used to calculate the electronic energy, including core repulsion, as a function of volume at high pressure for three phases of FeH (Isaev et al., 2009). Moreover, they extended their calculations to free energy as a function of pressure by adding a quasiharmonic vibrational (i.e., Phonon) contribution to the ground-state electronic energy. Studies performed by (Umemoto et al., 2008) about the spin state of Fe in iron bearing magnesium silicate Perovskite, (Mg,Fe)SiO3, under lower-mantle conditions, show that state is a crucial parameter for determining key mantle properties such as elastic and 34 seismic wave velocities, the post-Perovskite transition pressure, and electrical and thermal conductivities. For calculation of phonon frequencies of solid materials with transition-metals, using full potentials or pseudo-potentials, LDA usually overestimates, and GGA-PBE usually underestimates (Grabowski et al., 2007). This author also shows that, for the calculation of thermal expansion, heat capacity, and free energy, LDA is usually more accurate than GGAPBE (Grabowski et al 2007). For phonon dispersion curves, the experimental results generally lie between those calculated by LDA and GGA-PBE. These same authors suggest that for improving the accuracy of calculated thermodynamic data of solids with transition metals, pure DFT data should be combined with temperature dependent data calculated with Phonons. The Equation of State (EOS) of a material is determined by its Helmholtz free energy F(V,T), which consists of three parts: F(V,T) = U(V) + Fvib(V,T) + Fele(V,T) (1.20) (,) = ∑ ,󰇣2 sinh 󰇡 󰇢󰇤 (1.21) where U(V) is the static energy of the lattice, Fvib(V,T), is the vibrational free energy, and, Fele(V,T), accounts for the thermal excitation of the electrons, and, kB, is the Boltzman constant, nq, is the frequency of the phonon mode for wave vector, q, and volume, V, (Isaev et al. 2007). 1.2.2 Phonon calculations, the PHONON code The software PHONON was used to perform the Phonons calculations. PHONON (Parlinski, 1999), is a code for calculating phonon dispersion curves, and phonon density spectra of crystals, surfaces and adsorbed atoms on surfaces. From either a set of force constants or from a set of Hellmann-Feynman forces calculated within an Ab-initio code as Vienna Abinitio Simulation Package (VASP), (Kresse & Hafner, 1993), (Kresse et al., 1994), (Kresse & Furthmüller, 1996), (Kresse & Furthmüller, 1996a). This code is capable to optimize a 35 supercell and calculate the Hellmann-Feynman forces. PHONON builds a crystal structure, using one of the 230 crystallographic space groups, finds the force constant from the Hellmann-Feynman forces, builds the dynamical matrix, diagonalizes it, and calculates the phonon dispersion relations, and their intensities. PHONON finds the polarization vectors, and the irreducible representations, Gamma point of phonon modes, and calculates the total and partial phonon density of states. It plots the internal energy, free energy, entropy, heat capacity and tensor of mean square displacements, Debye-Waller factores (Parlinski, 1999). PHONON finds the dynamical structure factor for the coherent inelastic neutron scattering and the incoherent doubly differential scattering cross section for a single crystal and polycrystal. A detailed knowledge of lattice vibrations is critical for the understanding and quantitative prediction of a wide variety of temperature dependent physical properties of solids. The fundamental thermodynamic functions of internal and free energy, entropy, heat capacity as well as non-linear properties such as thermal expansion and heat conduction are to a considerable extent determined by the vibrations of the constituent atoms in the lattice. Fortunately, the quantum theory of lattice dynamics is well developed and has proven to be one of the most successful theories of solid state physics (Müller et al., 2008). 1.2.3 Planewave pseudopotential method, the VASP code The ab initio calculations are performed using the planewave pseudopotential code, the Vienna Ab-initio Simulation Package (VASP), (Kresse & Hafner, 1993), (Kresse et al., 1994), (Kresse & Furthmüller, 1996), (Kresse & Furthmüller, 1996a). The VASP codes have been used to determine the equilibrium lattice parameters, heats of formation, elastic constants. The Phonon dispersions are calculated together with PHONON code (Parlinski, 1999). The VASP approach, is based on a finite-temperature local-density approximation, with the free energy as variational quantity, and an exact evaluation of the instantaneous electronic ground state at each calculation step using efficient matrix diagonalization schemes (Kresse & Furthmüller, 1996). Planewaves (PAW) (Blöchl, 1994) and pseudopotentials form a natural alliance in this method, and they are fundamental when DFT calculations are used. In planewave pseudopotential method, the model system is constructed in a tridimensional periodic supercell which allows Bloch’s theorem to be applied to the electron wavefunctions. 36 ,()=,()exp(.) (1.22) The function u(r) has the periodicity of a supercell. It can be of any appropriate mathematical form and habitually one chooses a series expansion in terms of a set of basis function. In planewaves pseudopotential, planewaves are used for this expansion, so that each single-electron wavefuntion ψn,k is written as (Segall et al., 2002). ,()=∑,()exp((+).)  (1.23) The n,k are the expansion coefficients. The wavevectors G are such that the planewaves are commensurate with the supercell. Both the number of G vectors in the sum and the number of k considered should in principle be infinite (Kresse & Furthmüller, 1996a). The exponential term is a planewave of wavevector k which must be commensurate with the entire system not just the periodically-replicated cell. For an infinite system there is an infinite number of k vectors, at each of which solutions for ψn,k exist. This simply reflects the fact that the number of electrons is infinite (Segall et al., 2002). 37 1.3 References Ansisimov, V., Zaanen, J., & Andersen, O., (1991.) Band theory and Mott insulators: Hubbard U instead of Stoner I Physical Review B, Vol.44, No.3 pp. 943-954, ISSN 1098- 0121 Andzelm, J., & Wimmer, E., (1992). Density functional Gaussian‐type‐orbital approach to molecular geometries, vibrations, and reaction energies, Journal of Chemical Physics. Vol. 96 No. 2 , pp. 1280-1304, ISSN 0021-9606 Becke, A., (1988). Density-functional exchange-energy approximation with correct asymptotic behavior, Physical Review A, Vol.38, No. 6, pp. 3096–3100, ISSN 1050-2947 Bengone, O., Alouani, M., Blochl, P., & Hugel, J., (2000). Implementation of the projector augmented-wave LDA+U method: Application to the electronic structure of NiO. Physical Review B, Vol. 62, No.24, pp. 16392-16401, ISSN 1098-0121 Blöchl, P., (1994). Projector augmented-wave method, Physical Review B, Vol. 50, Vol.24, pp. 17953-17979, ISSN 1098-0121 Cramer, C., & Truhlar, D., (2009). Density Functional Theory for Transition Metals and Transition Metal Chemistry, Physical Chemistry Chemical Physics, Vol.11, No.44, pp. 10757-10816, ISSN 1463-9076 CODELCO National Copper Corporation of Chile, (2012). http://www.joeskitchen.com/chile/photos/norte/chuqui.htm di Salvo F. & McMillan, W,. (1977). Electron–Phonon Interaction and Phase Diagrams, Ed. Risk, T., New York: Plenum, pp. 107 Dudarev, S., Botton, G., Savrasov, S., Humphreys, C., & Sutton, A. (1998). Electron-energy- loss spectra and the structural stability of nickel oxide: An LSDA+U study, Physical Review B, Vol. 57, No.3, pp. 1505-1509, ISSN 1098-0121 Eschrig, H., (2003) The Fundamentals of Density Functional Theory (revised and extended version), Institute for Solid State and Materials Research Dresden, University of Technology Dresden EAGLE Ed 2nd edition. pp224 Fermi, E., (1927). Un Metodo Statistico per la Determinazione di alcune Prioprietà dell'Atomo, Atti della Accademia Nazionale dei Lincei, Vol.6, pp.602–607, ISSN 0001-4435 38 Fock, V., (1930). Näherungsmethode zur Lösung des quantenmechanischen Mehrkörperproblems, Zeitschrift für Phisik Hadrons and Nuclei, Vol. 61, No. 1-2, pp. 126- 148, ISSN 0939-7922 Garcia, A., Elsasser, C., Zhu, X.,. Louie, S., & Cohen, M., (1992). Use of gradient-corrected functionals in total-energy calculations for solids. Physical Review B, Vol. 46, No.15, pp. 9829-9832, ISSN 1098-0121 Glenbard, W., (2012) AP Chemistry Wiki, http://gwapchem.wikispaces.com/23.7+Transition+Metals Grabowski, B., Hickel, T. & Neugebauer, J., (2007). Ab initio study of the thermodynamic properties of nonmagnetic elementary fcc metals: Exange –correlation-related error bars and chemical trends, Physical Review B, Vol. 76, Vol.2, pp. 024309-16, ISSN 1098-0121 Grabowski, B., Ismer, L., Hickel, T., & Neugebauer J., (2009). Ab initio up to the melting point: Anharmonicity and vacancies in aluminum, Physical Review B, Vol. 79, No.13, pp. 134106-16, ISSN 1098-0121 Hafner, J., (2008). Ab-initio simulations of materials using VASP: Density-functional theory and beyond, Journal of Computational Chemistry, Vol. 29, No.13, pp. 2044-2078, ISSN 1096-987X Hai, X., Tahir-Kheli, J., & Goddard, W., (2011). Accurate Band Gaps for Semiconductors from Density Functional Theory. Journal of Physical Chemistry Letters, Vol.2, No. 3, pp. 212-217. ISSN 1948-7185 Harris, S., & Chianelli R., (1984). Catalysis by transition metal sulfide: the relation between calculated electronic trends abd HDS activity, Journal of Catalysis. Vol.86, No- 2, pp. 400-412, ISSN: 0021-9517 Hartree, D., (1928). The Wave Mechanics of an Atom with a Non-Coulomb Central Field. Part I. Theory and Methods, Mathematical Proceedings of the Cambridge Philosophical Society Vol.24 pp. 89-110, ISSN 0068-6735 Hathaway, K., Jansen, H,. & Freeman, A., (1985). Total energy local spin density approach to structural and electronic properties of ferromagnetic iron. Physical Review B, Vol. 31, No.12, pp. 7603-7611, ISSN 1098-0121 Hedin, L., & Lundqvist, B., (1971). Explicit local exchange-correlation potentials, Journal of Physics C: Solid State Physics, Vol. 4, No. 14, ISSN 0953-8984 39 Hepp, A., Rybicki, G., Raffaelle, R., Harris, J., Hehemann, D.,Junek, W., Gorse, J., Thompson, T., Hollingsworth, J., & Buhro, W., (2000). Chemical Fabrication Used to Produce Thin-Film Materials for High Power-to-Weight-Ratio Space Photovoltaic Arrays, NASA, in, http:// www.grc.nasa.gov/ WWW/RT/RT1999/5000/5410hepp.html Heyd, J., Peralta, J., Scuseria, G., & Martin, R., (2005). Energy band gaps and lattice parameters evaluated with the Heyd-Scuseria-Ernzerhof screened hybrid functional, Journal of Chemical Physics, Vol. 123, No. 17, pp.174101-8, ISSN 0021-9606 Heyd, J., Scuseria, G., & Ernzerhof, M., (2003). Hybrid functionals based on a screened Coulomb potential, Journal of Chemical Physics, Vol. 118, No.18, pp. 8207-8216, ISSN 0021-9606 Hobbs, D., & Hafner, J., (1999). Magnetism and magneto-structural effects in transitionmetal sulphides, Journal of Physics: Condensed Matter, Vol. 11, No.42, pp. 8197-8222, ISSN 0953-8984 Hohenberg, P., & Kohn, W., (1964). Inhomogeneous Electron Gas, Physical Review, Vol. 136, No. 38, pp.864-871, ISSN 0031-899X Hutter, J., (2002).Lecture Notes - Introduction to Ab Initio Molecular Dynamics, Physical Chemistry Institute of University of Zurich. pp 119 Isaev, E., Skorodumova, N., Ahuja, R., Vekilov, Y., Johansson, B., (2007) Dynamical stability of Fe-H in the Earth’s mantle and core regions. Proceedings of the National Academy of Sciences, Vol. 104, No. 22, pp.9168-9171, ISSN 0027-8424 Kohn, W., & Sham, L., (1965). Self-Consistent Equations Including Exchange and Correlation Effects, Physical Review, Vol.140, pp.1133–1138, ISSN 0031-89 Kresse, G., & Hafner, J., (1993). Ab initio molecular dynamics for liquid metals, Physical Review B, Vol. 47, Vol.1, pp. 558-561, ISSN 1098-0121 Kresse, G., Furthmüller, J., & Hafner, J., (1994). Theory of the crystal structures of selenium and tellurium: The effect of generalized-gradient corrections to the local-density approximation, Physical Review B, Vol. 50, Vol.18, pp. 13181-13185, ISSN 1098-0121 Kresse, G., & Furthmüller, J., (1996). Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set, Physical Review B, Vol. 54, Vol.16, pp. 11169- 11186, ISSN 1098-0121 40 Kresse, G., & Furthmüller, J., (1996a). Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set, Computational Materials Science, Vol. 6, No.1, pp. 15-50, ISSN 0927-02569X Kurth, S., Perdew, J., & Blaha P., (1999). Molecular and solid-state tests of density functional approximations: LSD, GGAs, and meta-GGAs. International Journal of Quantum Chemistry, Vol. 75, No. 3, pp. 889-909, ISSN: 1097-461X Li, Z., Bhatt, D., Schultz, N., Siepmann, J., & Truhlar, D., (2007). Free Energies of Formation of Metal Clusters and Nanoparticles from Molecular Simulations: Aln with n = 2-6, Journal of Physical Chemistry C, Vol. 111, No.44, pp. 16227-16242, ISSN 1932-7447 Liechtenstein, A., Anisimov, V., & Zaanen, J., (1995). Density-functional theory and strong interactions: Orbital ordering in Mott-Hubbard insulators. Physical Review B, Vol. 52, No.8, pp. 5467-5470, ISSN 1098-0121 Loschen, C., Carrasco, J., Neyman, K., & Illas, F. (2007). First-principles LDA+U and GGA+U study of cerium oxides: Dependence on the effective U parameter. Physical Review B, Vol. 75, No.3, pp. 035115-8, ISSN 1098-0121 McNaught, A. & Wilkinson, A., (1997) Compendium of Chemical Terminology, International Union of Pure and Applied Chemistry (IUPAC), 2nd ed. Blackwell Scientific Publications, Oxford, ISBN 0-86542-684-8 Michaelides, A., & Scheffler, M., (2012). An Introduction to the Theory of Crystalline Elemental Solids and their Surfaces. In Surface and Interface Science edited by K. Wandelt, Vol.1, pp. 13-72, ISBN 9783527411566 Müller S., Wolf, W., & Podloucky, R., (2008) Ab-initio methods and applications, Alloy Physics – A comprehensive Reference , Ed. Pfeiler, W., Wiley - Technology & Engineering – pp. 589-642, ISBN 978-3-527-31321-1 Parlinski, K., (1999). Calculation of phonon dispersion curves by the direct method, American Institute of Physics, Neutrons and numerical methods, AIP Conference Proceedings, ed. Johnson, M., Kearley, G, .& Buttner H., Vol. 479, pp. 121-126, ISBN 156396838X Pearce, C., Pattrick, R., & Vaughan, D., (2006) Electrical and Magnetic Properties of Sulfides, Reviews in Mineralogy and Geochemistry. Vol.61, No. 1, pp.127-180, ISSN 1529- 6466 41 Recurrently, there are no experimental data available in the published literature or in databases. Therefore, we had to proceed with the experiments to obtain data. We have initiated the synthesis process of new samples with targeted compositions. By relating specific properties of the elements with their position in the periodic table, we selectively manipulate the composition to obtain the optimal configuration for a particular physical property. The electronic properties (e.g. band structure) are crucial to characterize photosensitive behaviors . For some different compositions it can be demonstrated that different elements, occupying certain crystallographic sites, will make a difference in what concerns the photosensitive behaviors of Tetrahedrites. The latter was taken into consideration and ab initio calculations were performed to show that the electronic properties of these compounds make them promising candidates as solar cells photovoltaic materials, since they possess a direct band gap and a band gap between 1.1 eV and 2.2eV. These properties can be further optimized by doping and substituting certain ions in the structure. Fig 2.1 Energy vs. crystal momentum for a semiconductor with an indirect band gap (Adapted from: Wikipedia, 2012). In Fig 2.1 it is illustrated the energy vs. crystal momentum for a semiconductor with an indirect band gap, showing that an electron cannot be removed from the lowest-energy state in the conduction band (green) to the highest-energy state in the valence band (yellow) without a change in momentum. Here, almost all the energy comes from 48 a photon (vertical red arrow), while almost all the momentum comes from a phonon (horizontal blue arrow) (Wikipedia, 2012) Fig 2.2 Energy vs. crystal momentum for a semiconductor with a direct band gap (Adapted from: Wikipedia, 2012). In Fig 2.2 it is illustrated the Energy vs. Crystal Momentum for a semiconductor with a direct band gap, showing that an electron can shifted from the lowest-energy state in the conduction band (green) to the highest energy state in the valence band (yellow) without a change in crystal momentum. Depicted is a transition in which a photon excites an electron from the valence band to the conduction band (Wikipedia, 2012). . This fact is very important for photovoltaics solar cells. Silicon is the most common solarcell material, despite the fact that it has indirect-gap and therefore it does not absorb light very well. Silicon solar cells are typically hundreds of micrometres thick; if it was much thinner, much of the light, particularly in the infrared would simply pass through. On the other hand, thin-film solar cells are made of direct band gap materials such CIGS or CZTS, which absorb the light even if much thinner, and consequently can be made with a very thin active layer, often less than 1 micrometre thick (Miles et al. 2005). The absorption spectrum of an indirect-band gap material usually depends more on temperature than that of a direct material, because at low temperatures there are fewer phonons, and therefore it is less likely that a photon and phonon can be simultaneously 49 absorbed to create an indirect transition. For example, silicon is opaque to visible light at room temperature, but transparent to red light at liquid helium temperatures, because red photons can only be absorbed in an indirect transition (Dresselhaus et al. 2008). Although Cu12Sb4S13 presents an electronic band gap of 1.24 eV, verifying the band gap condition, we want to replace the antimony in the Tetrahedrite structure since this element and its compounds are toxic. Chalcopyrite (Cpy), CuFeS2 which crystallizes in a tetragonal scalenohedral P-42m structure was furthermore studied. Based on our calculations, the band gap of this compound is 0.89 eV; but by doping and by high-pressure synthesis this value can be changed to be within the optimal band gap energy range (Hossain, 2012). The most common elements used to dope Fe, are In or Ga and to dope S, is Se. In reality, Cu(In,Ga)Se2 (CIGS) is one of the most promising thin-film solar cell materials, demonstrating an efficiency of about 20% (Contreras et al., 2005; Chen et al., 2009). Nonetheless, In and Ga are expensive and toxic elements and the band gap is usually not optimal for high efficiency CIGS solar cells. We want to design and synthesize new, highefficient and low cost solar absorbers to replace CIGS. The Enargite-Famatinite (Enr-Fam) Cu3AsS4-Cu3SbS4 is inexpensive; Enargite presents photosensitive properties (Pauporté & Lincot, 1995). These systems are constituted by the toxic antimony, Sb, and arsenic, As, elements. Nonetheless, these systems are interesting to be used as control for the structural and electronic properties. Mechanical properties were additionally calculated for all compounds and were compared. It is clear, from the recent published literature (Chirilă et al., 2010), the continuing interest in the study of Chalcogenide materials constituted by multielement and coordinated by sulfur. 2.2. The Cu-Sb-S system and the Tetrahedrite, Cu12Sb4S13. Sulfur is one of the most abundant elements in the universe and the 14th most abundant element in the Earth’s crust. It plays a critical role in biology, atmospheric and ocean chemistry. Sulfur is the protagonist in multiple processes in different geochemical settings (Mandeville, 2010). 50 Sulfur is distributed throughout the mantle and crust mainly in sulfide or sulfate minerals and it finds its way into deep geologic fluids when these mineral reservoirs are perturbed by tectonic processes. The study of the sulfur behavior in the geochemical cycle provides the understanding of the origins of these mineral phases and their different formation conditions (Wallace & Carmichael, 1994). The latter variations of the formation conditions are due to temperature differences, or more importantly due to sulfur oxidation or reduction reactions. The oxidation or reduction reactions can occur at high temperature, such as in igneous systems, at intermediate temperatures, such as in hydrothermal systems and at low temperature during sedimentary digenesis. At high temperatures, the reactions tend to occur under equilibrium conditions, whereas at low temperatures, non-equilibrium is prevalent. The chemical speciation of sulfur in geological fluids is a controlling factor of the geochemical processes. The two major chemical forms of sulfur in crustal fluids, over a wide range of temperature and pressure are believed to be sulfate (SO42-) and sulfide (S2-). Using Raman spectroscopy it is possible to show that the dominant stable form of sulfur in aqueous solution above 250°C and 0.5 GPa is the trisulfur ion (S3-) (Pokrovski, & Dubrovinsky, 2011). The large stability range of (S3-) enables efficient transport and concentration of sulfur by geological fluids in deep metamorphic and subduction-zone settings. Understanding the solubility of sulfur in molten silicate masses, together with the solubility of other volatile components, is a key factor for understanding the oxidation state of magma and recycling of global elements in subduction zones. It can contribute to the interpretation of the genesis of the deposits of metallic sulfides and allows evaluating the influence of volcanic sulfur emissions on the climate (Carroll & Webster, 1994). It is known that sulfur plays a role in diagenesis and preservation of the organic matter and its conversion into oil fields (Vairavamurthy et al., 1993). Sulfur occurs in silicate liquids in extreme states of formal valence (Connolly & Haughton, 1972; Carroll & Rutherford, 1988); oxidized as sulfate S6+, and reduced in the form of sulfide S2-. The absorption spectroscopy using X-ray synchrotron radiation, XANES (X-ray Absorption Near Edge Structure), allows us to obtain the discontinuities in the absorption. A XANES edge discontinuity occurs when a core electron absorbs energy that is equal to or greater 51 than its binding energy. Edges are labeled according to the shell from which the electron was excited. Sulfur K-edge correspondent energy range is 2.45 to 2.53 keV. XANES spectroscopy is a non-destructive technique and additionally it is strongly sensitive to the chemistry (formal oxidation state and geometry) of the absorbing atom; this allows us to make inferences about the coordination state of the environment of the absorbing atom (Li et al., 1995; Morra et al., 1997; Paris et al., 2001). The microbeam XANES consents furthermore to realize studies in fluid inclusions in crystals of Olivine and Pyroxene (Gurenko & Schmincke, 2000; Bonin-Mosbach et al., 2002) which have shown the presence of intermediate oxidation states, namely S4+ in basaltic magmas rich in water (Métrich et al., 2002). A spectroscopic study in order to characterize the sulfur state of speciation was performed but not yet published. In this work we present the first results of the analysis of the fine structure of the S K-edge absorption discontinuity of this element by XANES. X-ray absorption experiments at ESRF (European Synchrotron Research Facility) had the purpose of studying the influence of the environment on sulfur ions and characterizing the geometry of metal ions coordinating sulfur and their bonding effects (e.g. the local order). These goals are achieved by comparing the observed sulfur K-edge profile of the minerals in study with the correspondent profile of well-known structures, which constitute model compounds for different binding situations. Native sulfur was used to characterize the molecule S = S (double bond) in S8, Fig. 2.1; Pyrite was used to characterize the dimer S22-, Fig. 2.2, and Chalcopyrite to characterize the sulfide ion S2-, Fig. 2.3. The following sulfide minerals were also studied as model compounds: Galena, PbS, Fig. 2.4, Blend, ZnS, Fig. 2.5, Chalcopyrite, CuFeS2, Fig. 2.3, and Tetrahedrites Cu12Sb4S13, Fig. 2.6. These phases have a very well defined structure resulting in characteristic absorption spectra. The comparison of the latter with the samples’ spectra permits a better interpretation of these spectra. In this particular case we can understand the conjugated influence of the nature of the coordinating metal atoms and the geometry of their arrangement over the electronic configuration of sulfur. The goal of our study is to determine how impurities - replacing Cu - in the Tetrahedrite structure, will induce changes in the ideal bond lengths and geometries when compared to the ideal. Tetrahedrite, conversely to other structures like for example the Chalcopyrite, can hold a big variety of impurities that replace Cu, such as: Zn, Fe, Ag, Cd, Mn, In, and Ge. We want to 52 understand why this phenomenon occurs in Tetrahedrite making it so different with this respect. Fig. 2.1 Crystal structure of the molecular sulfur. Space group: Fddd. Fig. 2.2 Crystal structure of the Pyrite, FeS2. Space group: Pa3. Fig. 2.3 Crystal structure of the Chalcopyrite, CuFeS2. Space group: I-42d. Fig. 2.4 Crystal structure of the Blend, ZnS. Space group: F-43m. Fig. 2.5 Crystal structure of the Tetrahedrite, Cu12Sb4S13. Space group:I-43m. Fig. 2.6 Crystal structure of the Galena, PbS. Space group: Fm3m. 53 Scans between 2.45 and 2.53 KeV were performed to collect S K-edge XANES spectra. Fig. 2.7 illustrates the general trend of S K-edge XANES spectra of natural, synthetic and simulated Tetrahedrite. The white line shown in the spectra reproduced in Fig. 2.7 (corresponding to the maximum absorption energy which in the Tetrahedrite XANES spectra is due to the Cu-S bond - an overlap of the tetrahedral and octahedral coordination contributions), does not seem to occur with similar intensity in the spectra collected from pure synthetic Tetrahedrite. This is eventually due to the presence of non-ideal structures in which there are slightly different bond lengths. Furthermore, the presence of a small absorption peak (in the pre-edge region of the natural Tetrahedrite) is indicative of the presence of an element that replaces Cu in Cu-S bonds. This impurity element is likely to be zinc in natural Tetrahedrites. 2460 2480 2500 2520 2540 2560 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Tetrahedrite natural (measured) Tetrahedrite synthetic (measured) Tetrahedrite synthetic (simulated) Energy (eV) Intensity (a.u.) white line pre - edge post - edge XANES EXAFS Fig. 2.7 Absorption s pectra obtained in XANES region for natural, synthetic and simulated Tetrahedrite. 54 The cubic crystal structure of Tetrahedrite has two formula units per cell, being derived from the Blend structure by replacing a tetrahedral cluster of packing vacancies (4[vac.]) in each 16 packing positions by one single S-atom according to the sequence (Figueiredo & Ferreira, 2002). [St 16]  [S12 [vac.] 4]  {[St 12] [So]} The general crystal chemical formula of Tetrahedrite may then be written (t, tetrahedral; tr, triangular; π, pyramidal; o, octahedral), clearly specifying the particular situation of one out of thirteen sulphur atoms in the crystal structure (Figueiredo & Ferreira, 2002). (Cu,Zn,Ag,Fe,Hg,Cd)t 6(Cu,Ag)tr 6(Sb,As,Bi,Te)π 4{[St 12][So]} This process provides pyramidal and triangular coordinations respectively to four As/Sb/Bi and to six Cu/Ag ions hosted in former tetrahedral sites within the original nonlacunar closest packing; the remaining six Cu ions are hosted in true tetrahedral interstices within the closest packing of sulfide anions, Fig. 2.8. Fig. 2.8 Tetrahedral and octahedral arrangements in the Tetrahedrite structure. Cu S Sb 55 Theoretical calculations and spectra modeling were performed with the codes FEFF 8.40 (Ankudinov et al., 1998) and FEFF 9.05 codes (Rehr et al., 2010). We want to compare FEFF simulations on DFT optimized impure Tetrahedrites with synthetic and then natural samples measurements. Spectroscopic data for Sb and Ag atoms were also recorded using XANES spectra and will be the subject of our near future study. 56 2.3. References Ankudinov, A., Ravel, B., Rehr, J., & Conradson, S., (1998). Real-space multiple-scattering calculation and interpretation of x-ray-absorption near-edge structure, Physical Review B Vol. 58, No. 12, pp. 7565-7576, ISSN 1098-0121 Bonin-Mosbach, M., Métrich, N., Susini, J., Salomé, M., Massare, D., & Menez, B., (2002). Micro X-ray absorption near-edge structure at the sulfur and iron K–edges in natural silicate glasses, Spectrochimica Acta, Part B, Vol. 57, No. 4, pp. 711-725, ISSN 0584-8547 Braga, M., Ferreira J., & Malheiros L., (2008). Phase transitions in the Cu-Sb-S system, Materials Science Forum, Vol. 587-588, pp. 435-439, ISSN 0255-5476 Carroll, M., & Rutherford, M., (1988). Sulfur speciation in hydrous experimental glasses of varying oxidation state: results from measured wavelength shifts of sulfur X-rays, American Mineralogist, Vol. 73, No. 7-8, pp. 845-849, ISSN 1945-3027 Carroll, M. & Webster, J. (1994). Solubilities of sulfur, noble gases, nitrogen, chlorine and fluorine in magmas. In Volatiles in Magmas, Edts. Carroll, M., & Holloway, J., Reviews in Mineralogy, Vol. 30, No. 1, pp. 231-27, ISSN 1529-6466 Chen, S., Gong, G., Walsh, A., Wei, S., (2009). Crystal and Electronic band structure of Cu2ZnSnX4 (X=S and Se) photovoltaic absorbers: First-principles insights, Applied Physics Letters, Vol. 94, No.4, pp. 041903-3, ISSN 1077-3118 Chirilă, A., Buecheler, S., Pianezzi, F., Bloesch, P., Gretener,C., Alexander, Uhl, A., Fella, C., Kranz, L., Perrenoud, J., Seyrling, S., Verma, R., Nishiwaki, S., Romanyuk, Y., Bilger, G., & Tiwari, A., (2011). Highly efficient Cu(In,Ga)Se2 solar cells grown on flexible polymer films, Nature Materials, Vol.10, No. 11, pp 857-861, ISSN 1476-1122 Connolly, J., & Haughton, D., (1972). The valence of sulfur in glass of basaltic composition formed under conditions of low oxidation potential, American Mineralogist, Vol. 57, No.9, pp. 1515-1517, No. 9, ISSN 1945-3027 Contreras, M., Ramanathan, K., Shama, J., Hasoon, F., Young, D., Egass B., & Noufi, R., (2005). Diode characteristics in state-of-the-art ZnO/CdS/Cu(In1−xGax)Se2 solar cells, Progress in Photovoltaics: Research and Applications, Vol. 13, No. 3, pp. 209-216, ISSN 1099-159X Dresselhaus, M., Dresselhaus, G., & Jori A., (2008). Group Theory: Application to the Physics of Condensed Matter, Springer, pp. 598, ISBN-10: 3540328971 57 This process provides pyramidal and triangular coordination, respectively, to four Sb and to six Cu+ cations hosted in former tetrahedral sites within the original non-lacunar closest packing; the remaining six Cu2+ cations are hosted in true tetrahedral interstices within the closest packing of sulfide anions (Fig. 2). The general crystal chemical formula of the Tetrahedrite may then be written (tr, triangular; t, tetrahedral; π, pyramidal; o, octahedral), as in the following expression: (Cu+)tr6(Cu2+)t6(Sb)π4[(S)t12(S)o] (2) The copper tetrahedral sites accommodate Cu2+ (Fig. 2 – bellow middle) as well as all tetrahedral coordinated divalent cations that can eventually substitute Cu2+ in the center of the tetrahedra. Cu+ occupies the center of a triangle (Fig. 2 – bellow left) of sulfur atoms and additionally, in a more distant position, in pyramidal coordination and oriented towards the central octahedra, is the Sb cation (Fig. 2 – bellow right). The (Cu+)tr sites share a single So atom belonging to the central octahedra (Fig. 2 – bellow left) and each (Cu+)tr is coordinated with two additional St atoms belonging to the tetrahedral framework. Such crystal chemical insight illustrates the decisive role that ab initio simulations have in providing a comprehensive interpretation of the Tetrahedrite behavior since substitutions in different sites will perturb different arrangements; furthermore since the panoply of substituting ions is comprehensive, ab initio calculations deliver valuable information on the properties of the various Tetrahedrites. Crystal engineering will be performed by substituting ions with the same valence in Tetrahedrites and eventually by creating vacancies in order to optimize electrical properties with the ultimate motivation of developing suitable photovoltaic materials. The 89,000TW of sunlight reaching the Earth's surface is plentiful – almost 6,000 times more than the 15TW equivalent of average power consumed by humans [4,5]. Additionally, solar electric generation has the highest power density (global mean of 170 W/m²) [4,6] among renewable energies. Photovoltaic materials are semiconductors used to convert solar radiation into direct current electricity. Solar photons colliding with electrons will make them jump to a higher energy state (an excited state); creating positively charged holes that in turn generate electrical current. An ideal solar cell absorber material should have a direct electronic band gap around 1.0 - 2.2 eV (in a direct band gap semiconductor, the top of the valence band and the bottom of the conduction band occur at the same value of momentum) with abundant, inexpensive and nontoxic elements. Materials Fig 2: (left) Crystal structure of the Tetrahedrite. Tetrahedral structures centered by St atoms, and the octahedra centered by So. (below left) triangular coordination for Cu+ cations. (below middle) tetrahedral coordination for Cu2+ cations. (bellow right) piramidal coordination for Sb. (Cu + ) tr (Cu 2+ ) t Sb Cu S Sb S o Sb (Cu 2+ ) t (Cu + ) tr S t 112 Advanced Materials Forum VI 64 presently used for photovoltaics include monocrystalline, polycrystalline and amorphous silicon, cadmium telluride, and copper indium selenide/sulfide [7]. Among these materials, Cu2ZnSnS4 (CZTS) is one of the most promising materials as it consists of abundant and relatively cheap elements. This semiconductor combines the properties for an ideal absorber layer for photovoltaic applications, like a direct band gap of about 1.4-1.5 eV and an optical absorption coefficient higher than 104 cm-1 [7,8]. In this work we are going to calculate the structural and electronic properties of Tetrahedrite, Cu12Sb4S13, and compare them with those of Ag6Cu6Sb4S13 to prove the viability of crystal engineering towards photovoltaics optimization. Calculation Methods The electronic band structures for the cubic crystalline Tetrahedrite, I -43m, were calculated using ab initio Density Functional Theory (DFT) and exchange correlation functionals. The calculations were performed using Vienna Ab initio Simulation Package code, VASP 5.2. [9] based on the DFT approach employing the Projector-Augmented Wave (PAW) method [10,11] within the Generalized Gradient Approximation and the Perdew-Burke-Ernzerhof (GGA-PBE) [12] functional. The PBE functional presents a simple derivation of the GGA, in which all parameters are fundamental constants, and is usually applied for calculation under periodic boundary conditions. The wave functions were expanded in the plane waves up to the kinetic energy cutoff of 500 eV, which ensures the convergence with respect to basis set for this structure. The Monkhorst– Pack [13] k-point generation scheme was used with a grid of 5×5×5 points in the irreducible part of the Brillouin zone. All calculations were performed in the 1×1×1 cell containing 58 atoms. The model structure was fully optimized (atomic and cell parameters) at hydrostatic pressures equal to 0 GPa. The electronic optimizations were continued until the energy differences between the successive electronic and ion cycles were less than 10−5 eV. Electronic Properties The band gap is a critical property for understanding the optical and electrical properties of the materials and for the design of semiconductor devices. The accurate calculation of band gaps is an active and important research area in Solid-State Physics and Theoretical Chemistry. Although Kohn–Sham DFT has been very successful in Theoretical Physics and Quantum Chemistry [14], the local density functionals such as the generalized gradient approximations GGA-PBE must be engaged in order to perform the calculation of electronic properties with greater accuracy. Calculation results The energy bands of electrons in semiconductors consist of a valence band fully occupied by electrons at low energy levels, a vacant conduction band at high energy levels and a forbidden band called - band gap - which separates the former two bands and which is not so high as that of an insulator. Fig. 2 shows the high symmetry lines within the first Brillouin zone of a cubic body centered lattice. The band gap of Cu12Sb4S13 was calculated to be 1.24 eV and 1.20 eV for Ag6Cu6Sb4S13 (Fig. 3). Generally, in Cu12Sb4S13 and Ag6Cu6Sb4S13 the 3d orbitals of Cu (or Ag, respectively) and the 3p of S occupy the upper region of the valence band denoting that there is bonding between Cu (or Ag) and S atoms. The orbitals 3s and 3p of S compose the bottom of the conduction band. The peaks in the interval -7 eV to -2 eV are due to the S 3p orbital. The Cu 3d orbital is located at -7 to 0 eV. The S 1s orbital on the lower region of the valence band is divided into two intervals from -12.5 eV to - 12 eV and from -16 eV to 13 eV. Materials Science Forum Vols. 730-732 113 65 Most of the states in the valence band, which is located from -5 eV to 0 eV, are mainly occupied by the Cu 3d orbital and in a small percentage by S 3p orbitals (Fig. 4b). Fig 4a, 4b: The partial Density of States (pDOS) was calculated for each atom occupying a different site in Cu12Sb4S13 and Ag6Cu6Sb4S13 crystal structures and for s, p and d orbitals. Following the previous analysis but now focusing only on the use of the partial density of states (pDOS), we will be able to understand the orbital contribution for the promotion of electrons to the conduction band thus facilitating the occurrence of the photovoltaic effect. Essentially, from the analysis of the partial density of states (pDOS), one can infer the contribution of the electronic valence orbitals of the atoms occupying a given crystal position as well as all its equivalent positions existing in the unit cell. By observing the data in the Fig 4a), it can be understood that both the 3d orbitals of copper that is in tetrahedral coordination, which are formally in the divalent state (Cu2+) (2), as the electrons of the 3d orbitals of copper in triangular coordination, which are in the monovalent state (Cu+) (2), Fig 2: High symmetry lines within the first Brillouin zone of a cubic body centered lattice , as t he Tetrahedrite structure. Point coordinates: (g1,g2,g3) Γ (0,0,0) H (1/2,-1/2,1/2) P (1/4,1/4,1/4) N (0,0,1/2) Fig 3: The calculated band gap is 1.24 eV for Cu12Sb4S13 (left) and 1.20 eV for Ag6Cu6Sb4S13 (right) and can be both considered direct band gaps. In vertical, at each right side, the total Density of States (tDOS) spectra. The Fermi energy corresponds to E = 0 eV. 4a) 4b) 114 Advanced Materials Forum VI 66 have equivalent probability to contribute to the promotion of these electrons to the conduction band, although the electrons of the 3d orbitals, which arise from the triangular positions, have a slightly more favorable electronic energy to promote the electrons to the conduction band. The 3p and 3d valence orbitals of the Cu2+ cation (2) are filled with fifteen electrons in the following way: 3p6 3d9, while the 3p and 3d valence orbitals of Cu+ cation (2), are filled with sixteen electrons in the following way: 3p6 3d10. As it can be perceived, the difference in one electron, which induces a decoupling in a 3d orbital, facilitates the photo excitation. In Ag6Cu6Sb4S13, there is a substitution of the Cu+ cation by Ag+ in the positions corresponding to the triangular coordination (2). In Fig 4b) it can be observed that the 3d orbitals of copper in the divalent state (Cu2+), in tetrahedral coordination, are much more likely to contribute with electrons to the conduction band. The promotion of the monovalent silver (Ag+) electrons to the conduction band is obtained from the 4d orbitals. In the 3d orbitals of Cu2+, the electronic filling of the valence orbitals is made as following: 3p6 3d9, while for the 4d orbitals of Ag+, the electronic filling of the valence orbitals is obtained by: 4p6 4d10. It is clear from the analysis of the Fig 4b) that the Ag+ valence electrons correspond to orbitals that will contribute less to conduction, since they have lower energy. What makes them not so favorable to lose electrons capable of reaching the conduction band is, possibly, the different lengths of chemical bonds with the neighboring atoms. As shown in Table 1, the bond distances between copper and sulfur are smaller for both St and So. S t S o (Cu + ) tr 2.254 Ǻ 2.291 Ǻ (Ag + ) tr 2.264 Ǻ 2.304 Ǻ First principles calculations allow the characterization of band gaps as direct / indirect which is of paramount importance for photovoltaic applications. Debye model Elastic properties of solids are very important due to the fundamental solid-state phenomena such as equation of state, phonon spectra and atomic potentials. Elastic rigidity coefficients are essential for achieving the comprehension of the mechanical properties of a solid; such as internal strain, thermoelastic stress and load deflection. Measurements of the deformations of a solid are dependent of the coefficients of elastic rigidity (Table 2). Table 2: Lattice parameter, enthalpy of formation at 0 K (with respect to the atomic state), elasticity tensor components Cij, and bulk, shear, and young’s modulus calculated using Voigt, Reuss and Hill approximations, obtained from the VASP optimization of the experimental structure. Tetrahedrite P (GPa) a(Å) (0 K) Hformation (kJ/mol; 0 K) C11 (GPa) C12 (GPa) C44 (GPa) Modulus (GPa) Voigt Reuss Hill Cu12Sb4S13 0 10.340 -425.9 80.46 52.99 19.61 Bulk 62.1 62.1 62.1 Shear 17.3 16.7 17.0 Young’s 47.4 46.1 46.7 Longitudinal -- -- 84.8 Ag6Cu6Sb4S13 0 10.374 -172.0 106.3 87.4 18.9 Bulk 93.7 93.7 93.7 Shear 15.1 13.5 14.3 Young’s 43.1 38.6 40.8 Longitudinal -- -- 112.8 Table 1. Bond distances of Cu + -S t , Cu + -S o , Ag + -S t and Ag + -S o when the cations Cu and Ag are in triangular coordination and the anion S in tetrahedral and octahedral positions. Materials Science Forum Vols. 730-732 115 67 It can be observed in Table 2. that the enthalpy of formation at 0 K is considerably different for each Tetrahedrite; Ag6Cu6Sb4S13 is much less stable than Cu12Sb4S13. Conversely, the bulk modulus which measures the substance’s resistance to uniform pressure is much higher in Ag6Cu6Sb4S13 although Cu (140 GPa), as a pure metal, has higher bulk modulus than Ag (100 GPa), in the same conditions [15]. The latter properties and differences highlight the necessity of first principles calculations to determine the properties of photovoltaic materials candidates, prior to synthesis and measurements. Furthermore, because both electronic band gap and mechanical properties change with the external pressure, it is important to correlate these properties if pressure is to be used as an additional tool to develop crystal engineering. Conclusions In this study, we present ab initio calculations using DFT and GGA-PBE to prove the usefulness of the latter in the selection of a particular substituting/doping element/ion and obtain new and optimized photovoltaic materials. Using the electronic band structures, density of states and correspondent orbital energies, the probability of certain electrons to be excited from the valence to the conduction band can be inferred. It can also be determined the energy cost of such a jump. Our results for the band structure and DOS for Cu12Sb4S13 and Ag6Cu6Sb4S13 show that both the compounds have similar band gaps even if the electrons that will jump to the conduction band, are formerly from atoms occupying different sites. The energy to overcome the band gap of these semiconductors – Tetrahedrites – falls within the photovoltaic operational range of 1.0 - 2.2 eV and the band gaps can be considered direct, as required. References [1] B.J. Wuensch, The crystal structure of Tetrahedrite, Cu12Sb4S13. Z. Kristallogr. 119, (1964) 437- 453 [2] K.Tatsuka, N. Morimoto, Tetrahedrite stability in the Cu-Fe-Sb-S system, Am. Min. 62 (1977) 1101-1109. [3] C. An, Y. Jin, K. Tang, Y. Qian, Selective Synthesis and Characterization of Famatinite Nanofibers and Tetrahedrite Nanoflakes, J. Mater. Chem., 13 (2003) 301-303. [4] V. Smil, Energy at the Cross Roads, Global Science Conference on Scientific Challenges for Energy Research, Paris, May 17-18, 2006. [5] U.S. Climate Change Technology Program – Technology Options for the Near and Long Term, November 2003 – pg. 35 [6] M. Jacobson, Review of Solutions to Global Warming, Air Pollution, and Energy Security, Ener. Environ. Sci., 2 (2009) 148-173. [7] S. Chen, X. G. Gong, A. Walsh, S.-H. Wei, Crystal and Electronic band structure of Cu2ZnSnX4 (X=S and Se) photovoltaic absorbers: First-principles insights, Appl. Phys. Let., 94 041903 (2009). [8] Th. Pauporté, D. Lincot, Electrical, Optical and Photoelectrochemical Properties of Natural Enargite, Cu3AsS4, Adv. Mat. Opt. Elect., 5 (1995) 289-298. [9] G. Kresse, J. Furthnüller, Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set, Comp. Mater. Sci. 6 (1996) 15-50. [10] P. E. Blöchl, Projector augmented-wave method, Phys. Rev. B 50 (1994) 17953-17979. [11] G. Kresse, D. Joubert, From ultrasoft pseudopotentials to the projector augmented-wave method, Phys. Rev. B 59 (1999) 1758-1775. [12] J. P. Perdew, K. Burke, M. Ernzerhof, Generalized gradient approximation made simple, Phys. Rev. Lett. 77 (1996) 3865-3868. [13] H. J. Monkhorst, J. D. Pack, Special points for Brillouin-Zone integrations Phys. Rev. B (1976) 5188-5192. [14] W. Kohn, L.J. Sham, Self-Consistent Equations Including Exchange and Correlation Effects, Phys. Rev.140 (1965) 1133-1138. [15] Information on: Wolfram research, http://periodictable.com/Elements/029/data.html 116 Advanced Materials Forum VI 68 Phase Transitions in the Cu-Sb-S System M. H. Braga1,a, J. A. Ferreira2,b, C. Lopes3,c and L. F. Malheiros3,d 1GMM-IMAT, Dep. of Physics, Faculty of Engineering of the University of Porto, Portugal 2INETI Laboratory, S. Mamede Infesta, Portugal 3GMM-IMAT, Dep. of Metallurgical and Materials Engineering, FEUP, Portugal [email protected], bjorge.f[email protected], c[email protected].pt [email protected], Keywords: Cu-Sb-S, Tetrahedrite, Phase diagrams, XRD, DTA/DSC, EPMA Abstract. The aim of this study is to determine the thermodynamic influence of the presence of Cu3SbS3, Cu2S and Sb2S3 in the formation of Cu12Sb4S13 - equivalent to mineral tetrahedrite. Thus, the Cu-Sb-S system is studied. Twenty three samples, with compositions that essentially lay along the vertical section Cu2S-Sb2S3, were prepared. The samples were analyzed by Electron Microprobe (EPMA) in order to determine the room temperature phases' composition. Samples were also analyzed by Differential Thermal Analysis (DTA/DSC) in order to establish thermal transitions, and by X-Ray Diffraction (XRD), at room and high temperatures, so as to determine the phases that are present in the equilibria at certain temperatures. The experimental phase diagram was established and the results were compared with those available in the literature. Introduction This study aims to improve the knowledge about the Cu-Sb-S system by comparing synthetic with natural materials. The compositional variation and thermodynamic properties of the synthetic phases can help to establish relations with the natural equivalent composition. The synthetic tetrahedrite type structure, in its simplest elementary composition, can be studied with the Cu-Sb-S system. This ternary system encloses, as well, synthetics with the same composition as a large variety of natural copper antimony chalcogenides. In nature, one of these phases is a very common sulphide, and crystallises in cubic system space group (I43m) [1], with nominal composition M12X4S13 M = Cu, Ag, Bi, Hg, Fe, Zn, Pb X = Sb, As. This phase is very important by economic reasons because it can carry in its structure useful minor elements. It can show, as well, an interesting semiconductive electronic behaviour [2]. The comprehension of the mobility and accommodation of minor elements, like Ag, Bi, Hg, Fe, Zn, Pb,…, that can replace the copper positions or form an independent nanostructure in tetrahedrite lattice, will permit to explore the applicability of this structure. Experimental For studying the thermodynamic behaviour of the phases, belonging to the system Cu-Sb-S, synthetic samples were prepared: analytical powder (99.9995%) pure elements of Cu, Sb and S, from Alfa Aesar Puratronic, were weighed to obtain charges of 1 g. Twenty three samples, with compositions that lay essentially along the vertical section Cu2S- Sb2S3, were sealed in evacuated silica glass containers and melted at 630±2ºC for 6 h after a slow heating that took 4h. After melting, the temperature slowly dropped to room temperature in 4h in order to allow samples' homogenisation. The samples were mounted in epoxy resin and successively observed in reflected-light optical microscopy and analysed in an Electron Microprobe (EPMA) JEOL JXA 8500F with WDS (Wavelength-Dispersive crystal Spectrometer). Samples were also analyzed by Differential Thermal Analysis (DTA) / Differential Scanning Calorimeter (DSC) in a Setaram Labsys TG/DTA/DSC apparatus, in order to determine thermal transitions. Room (RT - XRD) and high temperature X-ray Materials Science Forum Vols. 587-588 (2008) pp 435-439 online at http://www.scientific.net © (2008) Trans Tech Publications, Switzerland All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of the publisher: Trans Tech Publications Ltd, Switzerland, www.ttp.net. (ID: 193.137.43.138-13/06/08,16:13:07) 69 Diffraction (HT - XRD) were performed in a Panalytical X'Pert Pro MPD with an Anton Parr chamber apparatus in order to determine the phases present in equilibrium at certain temperatures. 2323 100 µm Results and discussion EPMA/WDS. Almost every sample was analysed by means of EPMA and the composition of each phase was determined by WDS (using pure Cu, Sb and ZnS as standards). In Fig. 2 it can be seen three of the photomicrographs taken. It was observed that samples 1, 3, 4, 5, 7, 10, 14, 15, 16, 21, 22, and 23 presented the expected phases, at room temperature, after the work of [3] (see Table 1), although in the latter work samples were obtained by solid state synthesis. Samples 8, 9, 17, 18, 19 and 20 showed phases that were not expected to be found, at room temperature. This discrepancies are, in the case of samples 8 and 9, may be due to a composition shift caused by a small loss of sulphur and, in the case of sample 9, also due to a certain heterogeneity (Sb was not completely dissolved in the sample). For the other samples, the latter discrepancies are maybe due to the fact that the samples melt at higher temperatures than 630 ± 2 ºC (they have compositions close to those of the binary Cu-S whose compounds melt between ~813 and ~1127 ºC). Hence, as S melts at 115.2 ºC and Sb at 630.6 ºC, the first compounds to be formed are richer than expected in Sb. The remaining sulphur was probably accumulated at the container walls above 444.7 ºC (boiling point of S) Although in the literature, almost all synthesis, concerning the Cu-Sb-S system, were performed in the solid state (because solid state synthesis is closer to the natural process of formation of the chalcogenides), the homogenization was not sufficient for this kind of process. For the other samples, this problem could not be detected, since the samples reach the liquid state. Fig. 1. (left) Cu-Sb-S system at 300 ºC, focussed essentially on the pseudobinary Cu2S - Sb2S3, as it was proposed by Skinner et al. [2]. Bullets represent the composition of the prepared samples and number its reference. Fig. 2. (below) Photomicrographs of the samples, 23, 21 and 10. Sample 23: light grey - (Sb), dark grey - Cu2S, and medium grey - Cu12.3Sb4.3S13. Sample 21: light grey - CuSbS2, dark grey - Cu11.5Sb4S13 and medium grey - Cu3SbS4 Sample 10: dark grey - CuSbS2, and medium grey - Sb2S3 436 Advanced Materials Forum IV 70 Table 1. The samples and its constituent phase's compositions, at room temperature, after EPMA/WDS and RT-XRD measurements. x(Cu) x(Sb) x(S) (%) EPMA / WDS phases at room temperature ±1.0 at. (Cu); ±1.2 at. (Sb); ±0.9 at. (S) 1. 64.9 0.8 34.3 Cu64.6Sb0.0S35.4 (Cu2S) Cu43.0Sb13.8S43.2 (Cu3SbS3) 3. 54.9 6.1 39.0 Cu64.4Sb0.0S35.6 (Cu2S) Cu41.9Sb14.3S43.8 (Cu3SbS3) 4. 47.8 10.4 41.8 Cu63.5Sb1.0S35.5 (Cu2S) * Cu0.4Sb99.5S0.1 (Sb) 5. 41.3 14.7 44.0 Cu41.6Sb14.2S44.2 (Cu3SbS3) Cu24.5Sb24.6S50.9 (CuSbS2) ** probably (Sb) 6. 33.5 19.2 47.3 Cu42.2Sb14.3S43.5 (Cu3SbS3) Cu24.5Sb24.9S50.6 (CuSbS2) Cu0.3Sb99.7S0.0 (Sb) 7. 24.6 25.1 50.3 Cu36.8Sb12.3S50.9 (Cu3SbS4) Cu24.0Sb24.9S51.1 (CuSbS2) Cu0.7Sb38.7S60.6 (Sb2S3) 8. 17.4 29.7 52.9 Cu24.2Sb25.0S50.8 (CuSbS2) Cu0.3Sb99.7S0.0 (Sb) 9. 11.1 33.8 55.1 Cu24.8Sb25.0S50.2 (CuSbS2) Cu0.3Sb99.7S0.0 (Sb) 10. 6.4 36.0 57.6 Cu24.8Sb25.0S50.2 (CuSbS2) Cu0.2Sb39.9S59.9 (Sb2S3) 11. 1.9 38.3 59.8 ** ** ** 12. 22.4 22.8 54.8 ** ** ** 13. 23.6 24.1 52.3 ** ** ** 14. 35.2 16.4 48.4 Cu40.9Sb14.0S45.1 (Cu11.8Sb4S13) Cu37.1Sb12.5S50.4 (Cu3SbS4) Cu25.0Sb25.1S49.9 (CuSbS2) 15. 30.4 19.0 50.6 Cu36.2Sb12.2S51.6 (Cu3SbS4) Cu24.5Sb24.5S51.0 (CuSbS2) Cu2.1Sb36.1S61.8 (Sb2S3) 17. 48.1 7.1 44.8 Cu62.6Sb0.0S37.4 (Cu2S) Cu36.4Sb12.6S51.0 (Cu3SbS4) Cu24.9Sb24.6S50.5 (CuSbS2) 18. 47.7 5.2 47.1 ** ** ** 19. 44.2 5.1 50.7 Cu42.5Sb14.6S42.9 (Cu3SbS3) Cu25.1Sb25.1S49.8 (CuSbS2) Cu1.1Sb98.4S0.5 (Sb) 20. 32.7 11.3 56.0 Cu39.8Sb14.1S46.1 (Cu11.2Sb4.1S13) Cu36.3Sb12.4S51.3 (Cu3SbS4) Cu24.5Sb24.6S50.9 (CuSbS2) 21. 37.2 12.4 50.4 Cu40.4Sb14.0S45.6 (Cu11.5Sb4S13) Cu36.5Sb12.3S51.2 (Cu3SbS4) Cu24.7Sb24.5S50.8 (CuSbS2) 22. 24.3 25.3 50.3 Cu42.3Sb14.5S43.2 (Cu3SbS3) Cu24.6Sb25.0S50.4 (CuSbS2) Cu0.3Sb99.6S0.1 (Sb) 23. 41.8 14.2 44.0 Cu64.8Sb0.0S35.2 (Cu2S) Cu41.6Sb14.4S44.0 (Cu3SbS3) Cu3.9Sb94.6S1.5 (Sb) * According to XRD measurements, this phase is Cu3SbS3. ** Not measured. DSC/DTA. DTA/DSC measurements were performed on a heat flux instruments with the possibility of determining the latent heat. Alumina crucibles with covers were used and the measurements were performed under flowing argon atmosphere at approx. 2.2 bar and approx. 40 cm3 min-1). The heating rates used were: 2, 5, 10 and 20 ºC/min. At least three different samples of the same composition were studied. A calibration factor: C = b0 + b1*T + b2*R + b3*R2 (T – temperature in ºC, R – heating rate in ºC/min.) was evaluated after performing eleven calibrations with five different elements. Different heating rates (from 2 to 20 ºC/min.) were also used for determining the coefficients b0, b1, b2 and b3. The accuracy of all of the given temperatures is ±1 °C. The temperatures of the invariant phase reactions were taken from the extrapolated onset temperatures on heating. The liquidus surface temperatures were taken from the peaks on heating (Fig. 3). The obtained results are presented in Table 2. Materials Science Forum Vols. 587-588 437 71 Table 2. Expected transitions after [3,4], and DTA/DSC and HT-XRD measurements. N. Expected transitions (ºC) (from [3, 4] and this work) Transition Temperatures (ºC) (from DTA/DSC and HT-DRX) 1 α - Cu2S + Cu1.96S → β - Cu2S + Cu1.96S (90) [4]; Cu1.96S → β - Cu2 S + Cu2-δS (93) [4]; Cu3SbS3' → Cu3SbS3 (122 ± 3) [3]; β - Cu2S + Cu2-δS → Cu2-δS; Cu2-δS + Cu3SbS3 → Liq.2 + Cu2-δS (607 ± 3) [3]; Liq.2+ Cu2-δ S → Liq.2 + Liq.1 (1105) [4]; Liq.1 + Liq. 2 → Liq. [4] 85, 122, 172, 597, 1108, 1111 3 α - Cu2S + Cu1.96S → β - Cu2S (90) [4]; Cu1.96S → β - Cu2S + Cu2-δ (93) [4]; Cu3SbS3' → Cu3SbS3 (122 ± 3) [3]; β - Cu2S + Cu2-δS → Cu2-δS; Cu2- δS + Cu3SbS3 → Liq.2 + Cu2-δS (607 ± 3) [3]; Liq.2 + Cu2-δS → Liq.2 + Cu2-δS + Liq.1; Liq.2 + Cu2-δS + Liq.1 → Liq.2 + Liq.1 90, 123, 142, 604, 611, 911 4 Cu3SbS3' → Cu3SbS3 (122 ± 3) [3]; Cu2-δS + Cu3SbS3 + (Sb) → Cu2-δ S + Cu3SbS3; Cu2-δS + Cu3SbS3 → Liq.2 + Cu2-δS (607 ± 3) [3]; Liq.2 + Cu2-δ S → Liq.2 + Cu2-δS + Liq.1 125, 511, 604, 611 5 Cu3SbS3' → Cu3SbS3 (122 ± 3) [3]; Cu3SbS3 + CuSbS2 + (Sb)→ Cu3SbS3 + (Sb) + Liq.2 (531 ± 2); Cu3SbS3 + (Sb) + Liq.2 → Liq.2 (607 ± 2) 109, 472, 539, 603 6 Cu3SbS3' → Cu3SbS3 (122 ± 3) [3]; Cu3SbS3 + CuSbS2 + (Sb) → Cu3SbS3 + (Sb) + Liq.2 (531 ± 2); Cu3SbS3 + (Sb) + Liq.2 → Cu3SbS3 + Cu2-δS + Liq.2; Cu3SbS3 + Cu2-δS + Liq.2 → Cu2-δS + Liq.2 + Liq.1 (607 ± 2) [3]; Cu2-δS + Liq.2 + Liq.1 → Liq.2 108, 460, 522, 582, 604, 613 7 Cu3SbS4 + CuSbS2 + Sb2S3 → Cu3SbS4 + CuSbS2 + Liq.2 (476.5 ± 2); Cu3SbS4 + CuSbS2 + Liq.2 → Cu3SbS4 + Liq.2 + Liq.1 (553 ± 2); Cu3SbS4 + Liq.2 + Liq.1 → Liq.2 + Liq.1; Liq.2 + Liq.1 → Liq.2 472, 495, 531, 537 8 CuSbS2 + (Sb) → (Sb) + Liq.2 (531 ± 2); (Sb) + Liq.2 → Liq2 + Liq.1 ; Liq2 + Liq.1 → Liq.2 464, 520, 525, 533 9 CuSbS2 + (Sb) → (Sb) + Liq.2 (531 ± 2); (Sb) + Liq.2 → Liq2 + Liq.1 ; Liq.2 + Liq.1 → Liq.2 454, 524, 530, 538 10 CuSbS2 + Sb2S3 → Sb2S3 + Liq.2 (476.5 ± 2); Sb2S3 + Liq.2 → Liq.2 + Liq.1; Liq.2 + Liq.1 → Liq.2 465, 477, 495, 508 14 Cu12Sb4S13 + Cu3SbS4 + CuSbS2 → Cu3SbS4+ Cu3SbS3 + CuSbS2 (522 ± 2); Cu3SbS4+ Cu3SbS3 + CuSbS2 → Cu3SbS4+ Cu3SbS3 + Liq.2 (553 ± 2) ; Cu3SbS4+ Cu3SbS3 + Liq.2 → Cu3SbS4 + Liq.2; Cu3SbS4 + Liq.2 → Liq.2 530, 542, 572 15 Cu3SbS4 + CuSbS2 + Sb2S3 → Cu3SbS4 + CuSbS2 + Liq.2 (476.5 ± 2); Cu3SbS4 + CuSbS2 + Liq.2 → Cu3SbS4 + Liq.2 + Liq.1 (553 ± 2); Cu3SbS4 + Liq.2 + Liq.1 → Liq.1 + Liq.2; Liq.1 + Liq.2 → Liq.2 495, 538, 600, 618 19 The composition is not absolutely accurate, despite the fact that the present phases at room temperature are known 462, 536, 538, 613, 616 20 The composition is not absolutely accurate, despite the fact that the present phases at room temperature are known 63, 72, 541, 549, 601, 603, 605 21 Cu12Sb4S13 + Cu3SbS4 + CuSbS2 → Cu3SbS4+ Cu3SbS3 + CuSbS2 (522 ± 2); Cu3SbS4+ Cu3SbS3 + CuSbS2 → Cu3SbS4+ Cu2-δS + Liq.2 (553 ± 2); Cu3SbS4+ Cu2-δS + Liq.2 → Cu2-δS + Liq.2 + Liq.1; Cu2-δS + Liq.2 + Liq.1 → Liq.2 543, 560, 607, 610 22 Cu3SbS3' → Cu3SbS3 (122 ± 3); Cu3SbS3 + CuSbS2 + (Sb)→ Cu3SbS3 + (Sb) + Liq.2 (553 ± 2); Cu3SbS3 + (Sb) + Liq.2 → (Sb) + Liq.2 (607.5 ± 3); (Sb) + Liq.2 → Liq.2 463, 540, 612, 615 23 α - Cu2S + Cu1.96S → β - Cu2S (90) [4]; Cu1.96S → β - Cu2S + Cu2-δ (93) [4]; Cu3SbS3' → Cu3SbS3 (122 ± 3) [3]; β - Cu2S + Cu2-δS→ Cu2-δS; Cu2- δS + Cu3SbS3 + (Sb) → Cu2-δS + (Sb) + Liq.2 (588 ± 2); Cu2-δS + (Sb) + Liq.2 → Cu2-δS + Liq.2; Cu2-δS + Liq.2 → Liq.2; 90, 109, 155, 586, 595, 600 RT-XRD/HT-XRD. Powder RT-XRD measurements were performed to identify the present phases. For some samples, powder HT-XRD measurements were also performed, under a secondary vacuum of 10-5 mbar. CuKα radiation was used to collect patterns from 10 to 90º (2θ) with steps of 0.008º and counting time of 17 s (Fig. 4). This study complemented the DTA/DSC experiments. 438 Advanced Materials Forum IV 72 Position [°2Theta] 20 30 40 50 60 70 0 5000 10000 Sb2 S3 Cu3 Sb S4 Sb2 S3 Cu3 Sb S4 Sb2 S3 Cu3 Sb S4 Sb2 S3 Sb2 S3 Cu3 Sb S4; Sb2 S3; Cu Sb S2 Sb2 S3 Cu3 Sb S4; Cu Sb S2 Sb2 S3 Cu3 Sb S4; Sb2 S3 Sb2 S3 Sb2 S3 Cu3 Sb S4; Sb2 S3 Cu3 Sb S4 Sb2 S3 Sb2 S3; Cu Sb S2 Sb2 S3 Cu3 Sb S4; Sb2 S3 Sb2 S3 Cu3 Sb S4; Cu Sb S2 Sb2 S3 Cu3 Sb S4; Sb2 S3 Sb2 S3; Cu Sb S2 Sb2 S3 Sb2 S3; Cu Sb S2 Cu3 Sb S4; Sb2 S3 Cu3 Sb S4; Sb2 S3; Cu Sb S2 Sb2 S3 Cu3 Sb S4; Sb2 S3 Cu3 Sb S4; Sb2 S3; Cu Sb S2 Sb2 S3 Cu3 Sb S4; Sb2 S3 Cu3 Sb S4; Sb2 S3; Cu Sb S2 Cu3 Sb S4; Sb2 S3; Cu Sb S2 Cu3 Sb S4; Sb2 S3; Cu Sb S2 Amostra7 Summary Although in the literature all studies concerning the system Cu-Sb-S were obtained after solid synthesis, the present results show that the samples can be studied after being melted and homogenized. The EPMA/WDS and XRD room temperature data for the samples' equilibrium phases are in good agreement with the ternary from [3], at 300 ºC. The DTA/DSC results are also generally in agreement with those from [3, 4]. Nevertheless, some transitions, like that at ~465 ºC (for samples 5, 6, 22, 8, 9 and 10), are still not well determined. References [1] A. Pfitzner, M. Evain and V. Petricek: Acta Crystallogr. B Vol. 53 (1997), p. 337. [2] N. Hisayuki, E. Saburo and I. Taizo: Jpn. J. Appl. Phys. Vol. 8 (1969), p. 443. [3] B. J. Skinner, F. D. Luce and E. Makovicky: Econ. Geol. Vol. 67 (1972), p. 924. [4] D. J. Chakrabarti and D. E. Laughlin: Bull. Alloy Phase Diagrams Vol. 4, N. 3 (1993). Sample 7 Fig. 3. (left) DTA/DSC curve on heating for sample 10, for a heating rate of 2 ºC/min. Four transitions can be observed. The first two seem to correspond to invariant transitions and the last one seems to correspond to the liquidus point. Fig. 4. (below) RT-XRD for Sample 7. The phases identified, using the ICDD PDF 2, 2003, database, are Cu3SbS4 (Famatinite, syn: 00-035-0581), CuSbS2 (Chalcostibite: 00-002-0557) and Sb2S3 (Stibnite, syn: 01-078-1347). 2ºC/min. endo I (a. u.) Materials Science Forum Vols. 587-588 439 73 Magnesium can be transformed in a single step to MgH2 hydride with up to 7.6 wt% of hydrogen with a volumetric storage efficiency of 110g H2/l (Milanese et al., 2010a), according to: Mg (s) + H2 (g)  MgH2 (s) (3.3) Magnesium metal is hexagonal with P63/mmc space group (α-structure) but the absorption of hydrogen induces a structural change into the tetragonal rutile-type structure α-MgH2 (P42/mnm) (Aguey-Zinsou & Ares-Fernández, 2010) (see Fig. 3.1). Fig. 3.1 Crystal structure of magnesium (left) and magnesium hydride (right). At high temperature and pressure, the latter phase undergoes polymorphic transformations to form two modifications: -MgH2 and -MgH2, having an orthorhombic structure and a hexagonal structure, respectively (Schlapbach & Züttel, 2001). Other highpressure metastable phases have also been reported (Cui et al., 2008; Ravindran et al., 2004). The charge density distribution in these materials has also been investigated and revealed a strong ionic character. The charge density determination of MgH2 by means of synchrotron X-ray powder diffraction at room temperature, the maximum entropy method (MEM) and Rietveld refinement revealed that the ionic charge of Mg and H can be expressed by Mg1.91+ and H0.26-, respectively, denoting that Mg in MgH2 is fully ionized, but the H atoms are in a weak ionic state (Noritake et al., 2003). The high strength of these bonds results however in an unacceptably high thermodynamic stability which diminishes the potentialities of using MgH2 in practical applications. The hydrogen desorption temperature is well above 573 K, which is related to its high dissociation enthalpy (75 kJ/mol H2) under standard conditions of pressure (Schlapbach & Züttel, 2001). In addition, 80 the high directionality of the ionic bonds in this system leads to large activation barriers for atomic motion, resulting in slow hydrogen sorption kinetics (Vajo & Olson, 2007). Several solutions were envisaged to circumvent these drawbacks. They can be accomplished to some extent by changing the microstructure of the hydride by ball-milling it (Huot et al., 1999; Zaluski et al., 1997). In this process the material is heavily deformed, and crystal defects such as dislocations, stacking faults, vacancies are introduced combined with an increased number of grain boundaries, which enhance the diffusivity of hydrogen into and out of the material (Suryanarayana, 2008). Alloying the system with other metallic additives, like 3d elements (Ti, Fe, Ni, Cu or Al), or LaNi5, FeTi, Pd, V among others and oxides like V2O5 or Nb2O5 can also be a way of improving kinetic and/or thermodynamic properties by changing the chemical interaction between the atoms (Reule et al., 2000; Rude et al., 2011; Tan et al., 2011a). The use of a proper destabilization or catalyst element/alloy into the system has also been shown to improve adsorption/desorption kinetics and to lower the adsorption temperature (Beattie et al., 2011). Furthermore, substantial improvements in the hydriding-dehydriding properties can be achieved by nanoengineering approaches using nanosized reactants or by nanoconfinement of it (Jeon et al., 2011; Jurczyk et al., 2011; Vajo, 2011; Zaluska et al., 1999a; Zhao-Karger et al., 2010). The latter allows shorter diffusion distances and larger surface area, resulting in faster reaction kinetics. It can also introduce alternative mechanisms to hydrogen exchange modifying the thermodynamic stability of the process. As previously referred, an alternative approach for altering the thermodynamics of hydrogenation-dehydrogenation is achieved by using additives that promote hydride destabilization by alloy or compound formation in the dehydrogenated state. This approach is known as chemical destabilization. The principle underlying this approach is that the additives are capable to form compounds or alloys in the dehydrogenated state that are energetically favourable with respect to the products of the reaction without additives. Destabilization occurs because the system can cycle between the hydride and the additive instead of the elemental metal. A generalized enthalpy diagram illustrating this approach - destabilization of the generic hydride AH2 through alloy formation (ABx) promoted by the presence of the alloying species B - was given by Vajo and Olson (Vajo & Olson, 2007), and is shown in Fig. 3.2 81 Fig. 3.2 Generalized enthalpy diagram illustrating destabilization through alloy formation upon dehydrogenation (adapted from Vajo & Olson, 2007) 3.5 Cu-Mg, Ni-Mg and other MgH2 destabilizing systems The work of Reilly and Wiswall provided the first evidences of this concept (Reilly & Wiswall, 1967, 1968). In their work, they showed that MgH2 can be destabilized by Cu2Mg. The formation of CuMg2 occurs upon dehydrogenation at lower reaction temperatures than those obtained with just pure MgH2. The compound CuMg2 crystallizes in the orthorhombic structure (Braga et al., 2010c) and has a hydrogen capacity of 2.6 wt. % at 573 K (Jurczyk et al., 2007). The hydride formation enthalpy is approximately 5 kJ/mol H2 lower than that of the hydrogenation of MgH2 from Mg and this process obeys to the following scheme (Reilly & Wiswall, 1967): 2CuMg2 (s) +3H2 (g) ⇄ 3MgH2 (s) +Cu2Mg (s) (3.4) The intermetallic cubic compound Cu2Mg does not hydrogenate under conventional hydrogenation conditions and seems to improve dehydrogenation kinetics (as compared to MgH2) due to improve resistance towards oxygen contamination (Andreasen et al., 2006; Kalinichenka et al., 2011; Reilly & Wiswall, 1967). As to the hexagonal intermetallic compound NiMg2, Reilly and Wiswall (Reilly & Wiswall, 1968) established that it reversibly reacts with hydrogen to form a ternary hydride Mg2NiH4, with a hydrogen content of 3.6 wt. %, according to the following scheme: NiMg2Ni (s) +2H2 (g) ⇄ NiMg2H4 (s) (3.5) Results obtained by A. Zaluska and co-workers (Zaluska et al., 1999b) showed that ballmilling the mixtures MgH2 and NiMg2H4 results in a synergetic effect of desorption, allowing the mixture to operate at temperatures as low as 493K – 513K, with good 82 absorption / desorption kinetics and with total hydrogen capacity exceeding 5 wt.%. They point out however that the ball-milled mixtures of the hydrides behave differently from two metal phases that are firstly ball-milled and then hydrogenated. In the latter case volume changes occur during hydrogenation with associated volume expansion of the material, in contrast to what happen in their study in which NiMg2H4 promoted the hydrogen release from an adjacent MgH2 matrix since they undergo a significant volume contraction, which facilitates their dehydrogenation. Many more studies have focused on changes in the hydriding/dehydriding properties of Ni-Mg binary alloys with compositional changes and changes in processing variables. Nonetheless, we highlight the study of C. D. Yim and collaborators (Yim et al., 2007) that showed that the NiMg2 compound acted as a catalyst in the dissociation of the hydrogen molecule, which resulted in a faster nucleation of magnesium hydride compared to pure Mg. It revealed also that the capacity and kinetics of hydriding were larger and faster when the average size of the hydriding phase was smaller and the volume fraction of the phase boundary was larger, since phase boundaries between the eutectic α-Mg and NiMg2 phases acted as a fast diffusion path for atomic hydrogen. In the full hydrogenated state, the NiMg2H4 structure consists of tetrahedral [NiH4]4- complexes in a framework of magnesium ions and two different forms exist, hightemperature (HT) and low-temperature (LT). Under the partial pressure of 0.1 MPa of hydrogen, the HT cubic structure phase transforms into a LT monoclinically distorted structure between 518 and 483 K (Zhang et al., 2009). The LT phase has also two modifications the untwined (LT1) and micro-twinned (LT2), which depend on the thermomechanical history of the sample (Cermak & David, 2011). The hydride formation enthalpy for the NiMg2H4 has been determined experimentally for the HT form, and it is in the range from -64.3 to -69.3 kJ/mol H2, for the LT form this value ranges from -68.6 to -81.0 kJ/mol H2 (Tan et al., 2011b). In the pioneer work of Reilly and Wiswall (Reilly & Wiswall, 1968) it was pointed out the catalytical effect of NiMg2 on the hydrogen desorption characteristics of MgH2. Recently, Cermak and David (Cermak & David, 2011) showed that NiMg2, and more efficiently the LT1 phase of NiMg2H4, were responsible for the catalytic effect of Ni reported in the literature. The fact that NiMg2 is a metal whereas NiMg2H4 behaves like a semiconductor has opened the way to the possibility of using this system also as a switchable mirror upon hydrogenation and dehydrogenation (Setten et al., 2007). A switchable mirror will switch 83 from mirror to transparent material upon hydrogenation. A more detailed study of Ni-Mg- based hydrides can be found in (Orimo & Fujii, 2001). Despite all the interest and extensive research on the above referred systems, a problem still remains; the hydrogen holding capacities of these materials are considerably less than that of MgH2 (Sabitu et al., 2010). A way to overcome this limitation was found by combining MgH2 with LiBH4 (which involves the formation of MgB2 and Li-Mg alloy (Yu et al., 2006)) since pure LiBH4 has high gravimetric and volumetric hydrogen densities, 18.5 wt. % and 121 kg H2/m3, respectively (Bösenberg et al., 2010; Xia et al., 2011). However, although the reaction enthalpy is lowered and the hydrogen storage capacity increases (10.5 wt. %), the sorption and absorption processes occurs at high temperatures with relatively slow kinetic even though more additives are being tested in order to overcome this problem (Fernández et al., 2011; Xia et al., 2011). Alternatively, the study of the destabilization of MgH2 with TiH2 has also been taken experimentally (Choi et al., 2008; Sohn et al., 2011). Observations point to a substantially reduced apparent activation energy of 107-118 kJ/mol and significantly faster kinetics, compared with the 226 kJ/mol for the similarly milled MgH2. The latter system constitutes a promising material to be used in practical applications for hydrogen storage. The combined destabilization effect of Ni-Mg and Cu-Mg intermetallics towards MgH2 was also tested and the Mg-rich ternary Cu-Ni-Mg alloys were recognized to have high potential for solid state hydrogen storage and have attracted many research interests. The study recently reported by Tan and co-workers (Tan et al., 2011b) elucidates about the influence of Cu substitution on the hydrogen sorption properties of magnesium rich Ni-Mg films. This study shows a two-step hydrogen absorption process. The first step is due to the absorption of Mg not alloyed in the form of NiMg2 and/or CuMg2, hereafter denoted as “free Mg” and is very quick, because it is mainly catalyzed by the intermetallic phase, NiMg2. But the second step, due to the hydrogen absorption of intermetallic NiMg2 and/or CuMg2 (“bonded Mg”) is significantly slow. The Cu substitution shows positive effects on desorption kinetics during full capacity hydrogen cycling, but shows strongly negative effects on absorption kinetics, particularly for the second absorption step, due to the segregation of CuMg2 towards the grain boundaries of MgH2, forming a closed shell that traps the hydrogen in MgH2. The authors also reported that the Cu substitution has no thermo-destabilization for MgH2, but since a significant amount can be dissolve in NiMg2, even at elevated temperatures, thermodestabilization of NiMg2H4 and better desorption kinetics are observed. Hong and 84 collaborators (Hong et al., 2011) on their study on the hydrogen storage properties of x wt.% Cu-23.5 wt.% Ni-Mg (x = 2.5, 5 and 7.5) prepared by rapid solidification process and crystallization heat treatment have also reported that the NiMg2 phase has higher hydriding and dehydriding rates than Mg under similar conditions and that the addition of a smaller amount of Cu is considered favourable to the enhancement of the hydriding and dehydriding rates of the sample. The 2.5 wt.% Cu-23.5 wt.% Ni-Mg alloy had the highest hydriding and dehydriding rates. These observations are in line with the ones previously reported by the group of Milanese (Milanese et al., 2010b; 2008), who also observed the high sorption capacity and good sorption performance of Cu-Ni-Mg mixtures and proposed a two steps sorption process with different kinetics. The first step corresponds to the quick hydrogenation of “free Mg”, according to reaction (3.3). After this step, absorption keeps on with a slower rate corresponding to the second step, hydrogenation of the “bonded Mg” phases, NiMg2 and CuMg2, according to reactions (3.4) and (3.5). They also showed that Ni is more effective than Cu in catalyzing the desorption reactions and that NiMg2H4 and Cu2Mg phases destabilized each other with the beneficial effect of decreasing the dissociation temperature of about 50 K in comparison to the MgH2, from “free Mg”. The positive effect of Cu as a catalyst on the hydrogenation and thermodynamic properties of NiMg2 mixed by ball milling technique was also studied and recently reported by Vyas and co-workers (Vyas et al., 2011) showing that hydrogen storage capacity and enthalpy of formation of NiMg2 with 10 wt.% Cu reduces to 1.81 wt.% and 26.69 kJ (mol H)-1 from 3.56 wt.% and 54.24 kJ (mol H)-1 for pure NiMg2 at 573 K, respectively. They attributed the decrement in the absorption capacity to the formation of the intermetallic phase Cu2Mg, which does not absorb the hydrogen but itself behaves like a catalyst. However, in the case of nanocrystalline CuxNi10-xMg20 (x = 0-4) alloys synthesized by meltspinning technique, it was found (Zhang et al., 2010a, 2010b) that the substitution of Ni by Cu does not change the major phase NiMg2 although it leads to a refinement of grains with increased cell volume and the formation of a secondary phase CuMg2. This in turn leads to a decrease of the hydride stability with a clear improve of the hydrogen desorption capacity and kinetics of the alloys. The presence of CuMg2 seems to act as a catalyst for the hydride-dehydride reactions of Mg and Mg-based alloys. Similar behaviour was found in Cu0.25Ni0.75Mg2 and Cu0.4Ni0.6Mg2 alloys that were prepared by mechanical alloying and subsequent thermal treatment (Simičić et al., 2006). The latter effect was also investigated on Cu1-xNixMg2 (x = 0-1) alloys 85 by Hsu and collaborators (Hsu et al., 2010). They observed that by substituting Cu by Ni in CuMg2, the cell volume decreased (since the radius of Cu atom is slightly larger than Ni atom) and with increasing Ni content , the effect of Ni is actually effective in MgH2 and Mg2NiH4 destabilization, leading to a decrease of desorption temperature in these two phases. They also showed that substituted nickel caused the hydriding reaction because absorption kinetics and hydrogen storage capacity increased with the rise of Nisubstitution contents. 3.6 Lithium hydride An alternative route to be considered is to explore other hydrates besides MgH2 for solid hydrogen storage. One of most interesting is lithium hydride, because it contains 12.5 wt.% hydrogen. Nonetheless, the desorption temperature is 1183 K for an equilibrium pressure of 1 bar (Vajo et al., 2004). However, it has been shown (Chen et al., 2003) that when LiH (see Fig. 3.3) reacts with lithium amide (LiNH2) by thoroughly mixing the substances, hydrogen is released at temperatures around 423 K, with formation of lithium imide (Li2NH) or Li-rich imide (LixNH3-X) and lithium nitride (Li3N) depending on the temperature and molar ratio of (LiH/LiNH2) according to the following schemes: - Below 593 K: LiH (s) + LiNH2 (s)  2H2 (g) + Li2NH (s) (3.6) 2LiH (s) +LiNH2 (s)  (x-1) H2 (g) + LixNH3-X (s) + (3-x) LiH(s) (3.7) - At higher temperatures:2LiH (s) + LiNH2 (s)  H2 (g) + Li3N (s) (3.8) From a detailed analysis of high-resolution synchrotron x-ray diffraction data for the lithium amide (LiNH2) - lithium imide (Li2NH) hydrogen storage system (David et al., 2007), the authors were able to propose an alternative mechanism that does not need to have the materials mechanically milled to enhance mixing, as previously recognized by Chen and collaborators (Chen et al., 2003) as essential. The mechanism they propose for the transformation between lithium amide and lithium imide during hydrogen cycling is a bulk reversible reaction that occurs in a nonstoichiometric manner within the cubic anti-fluorite-like Li-N-H structure, based on both Li+ and H+ mobility within the cubic lithium imide. Concluding that increasing the Li+ 86 mobility and/or disorder it is likely to improve the hydrogen cycling in this and related Libased systems. Recently, further systematical evaluation of the decompositions of LiNH2 and Li2NH was carried out by Zhang and Hu (Zhang & Hu, 2011), who also examined the effect of Clanion on the decomposition process. Clis widely employed as a promoter to improve various catalysts. As a result, decomposition mechanisms were established. The decomposition of LiNH2 producing Li2NH and NH3 occurs in two steps at the temperature range of 573-723 K. LiNH2 decomposes into a stable intermediate species (Li1.5NH1.5) and then into Li2NH. Furthermore, Li2NH is decomposed into Li, H2, and N2 without formation of Li3N at the temperature range of 823-1023 K. The introduction of Clcan decrease the decomposition temperature of Li2NH by about 110 K. Fig. 3.3 Crystal structure of lithium hydride. 3.7 Neutron techniques associated with hydrogen solid storage Though some progresses have been made, the state-of-art materials are still far from meeting the aimed targets for hydrogen solid storage material (Churchard et al., 2011). This huge task can be facilitated by employing state-of-the-art techniques like, computational first-principles calculations to evaluate the thermodynamic properties of the potential materials (Alapati et al., 2007; Siegel et al., 2007; Yang et al., 2007). This allows a quick screen of a large number of potential candidates, searching for thermodynamically suitable ones (saving time and money). Once thermodynamic appropriate materials have been found other considerations such as structure and dynamics of the materials during hydrogenation/dehydrogenation will 87 become crucial in order to understand the fundamental properties of hydrogen storage, in realistic conditions and hence design new hydrogen storage materials. Neutron scattering techniques are highly suitable for structure and dynamics studies related to hydrogen in solids and bound on surfaces. The energy distribution of thermal neutrons is nearly ideal for the study of condensed matter in general because it is of the same order of magnitude as most molecular and lattice excitations and the de Broglie wavelengths of thermal neutrons match quite well with interatomic distances in most solids (Squires, 1978). Neutrons have some unique advantages over photons and electrons as scattering media which are of particular use for the analysis of hydrides. For these purposes the two most useful neutron scattering interactions are coherent elastic scattering for Neutron Diffraction (ND) and incoherent inelastic scattering (INS) to measure vibrational density of states. The distinction of coherent and incoherent scattering interactions is important to the unique advantages offered by ND and INS respectively. This is because the relative scattering intensity of a given interaction is dependent highly upon the nucleus involved, and as such is isotope dependant. Each isotope has a different scattering cross section for both coherent (σcoh) and incoherent (σinc) interactions measured in barns (1 barn = 10-28 m2). In general these scattering cross sections do not follow any specific trend regarding nucleus size. INS has numerous advantages to other common techniques of obtaining vibrational spectra such as infrared (IR) and Raman spectroscopy. INS spectroscopy is hyper sensitive to the presence of hydrogen. The protium (1H) nucleus has respective scattering cross sections of σcoh = 1.8 and σinc = 80.2 barns respectively. This means neutron scattering in materials containing natural abundance hydrogen is largely inelastic. Additionally, the incoherent cross section of 1H is one to two orders of magnitude higher than any other isotope (Ross, 2008). This means that in hydrides INS spectra are dominated by vibrational modes of hydrogen almost exclusively. This hyper sensitivity to hydrogen means that hydride phases are detectible even with present in miniscule relative concentration. 88 Another advantage of INS is the complete absence of selection rules for the excitation of vibrational modes. External modes (lattice modes, i.e. phonons) are excited with equal opportunity to molecular vibrations. Because both IR and Raman spectroscopy rely upon different types of charge symmetry interactions, a number of vibrational modes cannot be excited in the majority of solids and molecules. In particular lattice modes are far more easily observable in INS spectra than any other type of vibrational spectroscopy. INS is also more useful for comparison with ab initio calculated density of states because relative excitation amplitudes are simply dependent upon the magnitude of motion and σinc of the excited nucleus (Squires, 1978; Ross, 2008). Free codes, such as a-Climax is available to generate a theoretical INS spectrum from the density of states output files from numerous common ab initio packages such as Gaussian, AbInit and Dmol (Ramirez, 2004). For these reasons INS is extremely useful in identifying the presence of different hydride phases which may not be structurally apparent (for example, due to structural disorder). A good example is the INS study of Schimmel et al. on MgH2 produced from Mg processed by high energy ball milling. Ball milling of Mg to reduce particle size, and introduce fractures, defects, and faults has a beneficial effect of increasing hydride formation rate, and reducing the temperature required for absorption. Comparison of the INS spectra of the MgH2 produced from ball milled Mg with well-ordered MgH2 revealed a partial composition of γ-MgH2, which is metastable and normally exists only at high temperatures (Schimmel et al., 2005). Presence of γ-MgH2 was indicative of internal stress from mechanical processing. However after hydrogen sorption cycling, γ-MgH2 was no longer observable in the INS spectrum of the ball milled material, while the fast kinetics and low sorption temperature remained. In this way INS was indispensable in revealing that the particle size reduction is more significant in the role of lowering temperature and increasing sorption kinetics than the creation of faults and internal stresses after the high energy ball milling of Mg (Schimmel, 2005; Ross, 2008). Neutron diffraction also provides some unique advantages versus more conventional diffraction methods such as X-ray diffraction (XRD). Elastic neutron and X-ray scattering are similar in that both result in interference patterns according to Bragg scattering conditions (Squires, 1978). In XRD the intensity of a given Bragg reflection varies with the atomic number (Z) of the atom at the lattice site. This means that the exact position of hydrogen in a structure is practically impossible to determine with XRD. In ND the relative intensities of reflections 89 The CuLi0.08Mg1.92D5 crystal structure was determined to be monoclinic P121, with a = 1.514 nm., b = 0.688 nm, c = 0.555 nm and  = 91.73° according to the formula CuLi0.08Mg1.92D5 = 0.5(Mg32+.[CuH4]23-.MgD2) corresponding to 4.4 wt% D per formula unit. CuLi0.08Mg1.92D5 is the first deuteride/hydride to be formed. This result is interesting by itself, but the presence of MgD2 in the diffraction pattern, highlights even further the possibilities of applications of this compound. According to these results, it can be obtained MgH(D)2 from a sample that did not contain “free” Mg or CuMg2. Furthermore, the deuteration process occurred at 473 K, which is considerably lower than the hydrogen absorption temperature reported for CuMg2 (3.4) (Reilly & Wiswall, 1967). The experiments with the bulk sample at SMARTS, LANSCE, Los Alamos National Laboratory, show that before MgD2 is observed, CuLi0.08Mg1.92D5 is already distinguishable at the surface even in a sample that initially contained CuMg2 (see Fig. 3.7). Therefore, it seems that CuLi0.08Mg1.92D5 will have a catalytic and destabilizing roll that was additionally observed with the Ti/TiH2 systems (Braga et al., 2010b). Fig. 3.8 ND refined pattern of the center and surface of a bulk cylinder sample initially containing CuLi0.08Mg1.92, Cu2Mg, and CuMg2 obtained in SMARTS during an in situ reaction with D2 at 523 K and ~3.4 MPa. Besides the texture effect that might be present in a bulk sample, it seems that the center initially contained more CuLi0.08Mg1.92 than the surface. Experimental information about the metal–hydrogen interactions can be obtained by measuring lattice vibrations via INS, as previously highlighted in section 3.5. 96 Because of the large difference between the masses of metal and H atoms in transitionmetal–hydrogen systems, the acoustic dispersion branches of the phonon spectra can be attributed to the motion of the metal atoms, the optic branches to the vibrations of the light H atoms relative to the metal lattice. The densities of states of optic phonons typically show a pronounced maximum at the energy of the lattice vibrations at the  point in the center of the phonon Brillouin zone (q = 0) e.g. in (Fukay, 1993). These phonon modes describe the vibration of the undistorted H sublattice relative to the rigid metal sublattice. Hence, they contain the metal–hydrogen interaction only. This is usually stronger than the H–H interaction, which leads to the dispersion of the optic branches. In the limit of very low H concentrations, the H vibrations can be imagined as independent vibrations of local Einstein oscillators at interstitial H sites. For both the lattice vibrations at the  point and the local vibrations, one can observe transitions from the ground state, the quantum-mechanical zero-point vibration of the H atoms, to the first excited states, e.g. by measuring optic phonon excitations, and transitions to higher excited states. Their energies, intensities and symmetry splittings yield an insight into the shapes of the potential and the wavefunctions for the vibrations of the light particles (Elsasser et al., 1998). 400 600 800 1000 1200 1400 1600 1800 2000 2 3 4 5 6 7 8 9 10 11 12 13 NiMg2H4 1st cycle CuLixMg2-x-H 1st cycle NiMg2H4 2nd cycle CuLixMg2-x-H 1st cycle I (a. u.) wavenumber (cm-1) Fig. 3.9 INS spectra for NiMg2H4 (1st and 2nd hydrogenation cycles) and for two samples containing CuLi0.08Mg1.92 (triangles correspond to a sample also contained Cu2Mg and the circles correspond to sample that contained Cu2Mg and CuMg2 as well). All samples show the formation of a similar monoclinic structure. As in NiMg2H4, in which Ni is bonded to four atoms of H forming the tetrahedral complex [NiH4]4-, Cu is also bonded to four atoms 97 of H forming the tetrahedral complex [CuH4]3-, which was previously referred on (Yvon & Renaudin, 2005). We have measured samples of the Cu-Li-Mg-H system by means of INS at FDS, LANSCE, Los Alamos National Laboratory. There is no doubt about the sequence of events; first there is the formation of CuLi0.08Mg1.92H5 (see Fig. 3.9) and then in subsequent cycles the formation of MgH2, either for disproportionation of CuLi0.08Mg1.92H5 or from hydrogenation of CuMg2. DSC/TGA experiments show that CuLi0.08Mg1.92H5 starts desorbing hydrogen at 313 K to 328 K. In this range of temperatures the sample can release up to 1.3 wt.% (results for a isothermal experiment with a sample initially containing approximately 78 wt.% of CuLi0.08Mg1.92 and 22 wt.% of Cu2Mg - which does not absorb hydrogen at the temperature and pressure that were used). In Fig. 3.10 it can be seen that 0.5 wt.% of a sample initially containing 61 wt.% of CuLi0.08Mg1.92, 23 wt.% of CuMg2 and 16 wt.% of Cu2Mg can be released at T < 350 K. In spite of the fact that there was some visible oxidation during this run, we think it was worth showing this initial desorption. This initial desorption seems to be due to CuLi0.08Mg1.92H5. At ~473 K, hydrogen starts to be desorbed at a different rate, probably due to the disproportionation of CuLi0.08Mg1.92H5, with the formation of MgH2, which will start releasing hydrogen at 553-573 K. Additionally, MgH2 can be formed upon hydrogenation of CuMg2. 300 350 400 450 500 550 600 650 700 97,4 97,6 97,8 98,0 98,2 98,4 98,6 98,8 99,0 99,2 99,4 ~473 K 1.6 wt.% 1.4 wt.% m (wt.%) T ( K ) 0.5 wt.% 315 K Fig. 3.10 TGA of two samples initially containing approximately 61 wt.% of CuLi0.08Mg1.92, 23 wt.% of CuMg2 and 16 wt.% of Cu2Mg. Samples were measured after hydrogenation but they are not expected to be saturated in hydrogen prior to the experiment (this figure refers to the same sample composition, where the lower curve seems to have oxidized). 98 The DSC/TGA results show that the system containing CuLi0.08Mg1.92 and Cu2Mg can destabilize MgH2 in a more efficient way than Cu2Mg by itself can. In fact, in a DSC experiment in which kinetics must be accounted for, MgH2 will release hydrogen at 553- 573 K, which can only be obtained when particles are reduced to nanopowders. 3.11 Conclusion The hydrogen storage world still offers a considerable amount of challenges since no universal solution has been found. Eventually, different solutions will be found to suite different applications. The Cu-Li-Mg system provides other possibilities for catalytic and destabilization effects yet not fully explored. There are several techniques that can be employed to study systems containing hydrogen. Nonetheless, Neutron Scattering is a very useful resource, in particular, Neutron Diffraction. In the latter, crystal structure of deuteride phases are directly studied since deuterium can be detected by ND and more accurate results can be obtained either in ex situ and in situ as shown previously. 99 3.12 References Aguey-Zinsou, K. & Ares-Fernández, J. (2010). Hydrogen in magnesium: new perspectives toward functional stores. Energy & Environmental Science, Vol. 3, No. 5, pp. 526-543, ISSN 1754-5706 Alapati, S., Johnson, J. & Sholl, D. (2007). Using first principles calculations to identify new destabilized metal hydride reactions for reversible hydrogen storage. Physical Chemistry Chemical Physics, Vol. 9, No. 12, pp. 1438-1452, ISSN 1463-9084 Andreasen, A., Sørensen, M., Burkarl, R., Møller, B., Molenbroek, A., Pedersen, A., Vegge, T. & Jensen, T. (2006). Dehydrogenation kinetics of air-exposed MgH2/Mg2Cu and MgH2/MgCu2 studied with in situ X-ray powder diffraction. Applied Physics A, Vol. 82, No. 3, pp. 515-521, ISSN 1432-0630 Ansara, I., Dinsdale, A. & Rand, M. (Ed(s)). (1998). COST 507, Thermochemical database for light metal alloys, Vol. 2, pp. 170-174, Office for Official Publications of the European Communities, ISBN 92-828-3902-8, Luxembourg Beattie, S., Setthanan, U. & McGrady, G. (2011). Thermal desorption of hydrogen from magnesium hydride (MgH2): An in situ microscopy study by environmental SEM and TEM. International Journal of Hydrogen Energy, Vol. 36, No. 10, pp. 6014-6021, ISSN 0360-3199 Bösenberg, U., Ravnsbæk, D., Hagemann, H., D’Anna, V., Minella, B., Pistidda, C., Beek, W., Jensen, T., Bormann, R. & Dornheim, M. (2010). Pressure and Temperature Influence on the Desorption Pathway of the LiBH4−MgH2 Composite System. The Journal of Physical Chemistry C, Vol. 114, No. 35, pp. 15212-15217, ISSN 1932-7455 Braga, M., Acatrinei, A., Hartl, M., Vogel, S., Proffen, T. & Daemen, L. (2010a). New Promising Hydride Based on the Cu-Li-Mg system. Journal of Physics: Conference Series, Vol. 251, pp. 012040 [4 pages], ISSN 1742-6596 Braga, M., Wolverton, M., Llobet, A. & Daemen, L. (2010b). Neutron Scattering to Characterize Cu/Mg(Li) Destabilized Hydrogen Storage Materials. Materials Research Society Symposium Proceedings, Vol. 1262, pp. 1262-W03-05, ISSN 0272-9172 Braga, M., Ferreira, J., Siewenie, J., Proffen, Th., Vogel, S. & Daemen, L. (2010c). Neutron powder diffraction and first-principles computational studies of CuLixMg2-x (x≅0,08), CuMg2, and Cu2Mg. Journal of Solid State Chemistry, Vol. 183, No. 1, pp. 10-19, ISSN 0022- 4596 100 Braga, M. & Malheiros, L. (2007a). CuMg2-YLiX alloy for hydrogen storage. International patent, WO 2007046017 (A1) Braga, M. & Malheiros, L. (2007b). CuMg2-YLiX alloy for hydrogen storage. National patent, PT 103368 (A) Braga, M., Ferreira, J. & Malheiros, L. (2007). A ternary phase in Cu–Li–Mg system. Journal of Alloys and Compounds, Vol. 436, No. 1-2, pp. 278-284, ISSN 0925-8388 Cermak, J. & David, B. (2011). Catalytic effect of Ni, Mg2Ni and Mg2NiH4 upon hydrogen desorption from MgH2. International Journal of Hydrogen Energy, Vol. 36, No. 21, pp. 13614-13620, ISSN 0360-3199 Chen, P. & Zhu, M. (2008). Recent progress in hydrogen storage. Materials Today, Vol. 11, No. 12, pp. 36-43, ISSN 1369-7021 Chen, P., Xiong, Z., Luo, J., Lin, J. & Tan, K. (2003). Interaction between lithium amide and lithium hydride. The Journal of Physical Chemistry B, Vol. 107, No. 37, pp. 10967-10970, ISSN 1520-5207 Choi, Y., Hu, J., Sohn, H. & Fang, Z. (2008). Hydrogen storage properties of the Mg–Ti–H system prepared by high-energy–high-pressure reactive milling. Journal of Power Sources, Vol. 180, No. 1, pp. 491-497, ISSN 0378-7753 Churchard, A., Banach, E., Borgschulte, A., Caputo, R., Chen, J., Clary, D., Fijalkowski, K., Geerlings, H., Genova, R., Grochala, W., Jaroń, T., Juanes-Marcos, J., Kasemo, B., Kroes, G., Ljubić, I., Naujoks, N., Nørskov, J., Olsen, R., Pendolino, F., Remhof, A., Románszki, L., Tekin, A., Vegge, T., Zäch, M., & Züttelc, A. (2011). A multifaceted approach to hydrogen storage. Physical Chemistry Chemical Physics, Vol. 13, No. 38, pp. 16955-16972, ISSN 1463-9084 Cui, S., Feng, W., Hu, H., Feng, Z. & Wang, Y. (2008). Structural phase transitions in MgH2 under high pressure. Solid State Communications, Vol. 148, No. 9-10, pp. 403-405, ISSN 0038-1098 David, W., Jones, M., Gregory, D., Jewell, C., Walton, A. & Edwards, P. (2007). A mechanism for non-stoichiometry in the lithium amide/lithium imide hydrogen storage reaction. Journal of the American Chemical Society, Vol. 129, No. 6, pp. 1594-1601, ISSN 0002-7863 Egami, T. & Billinge, S. (2003). Underneath the Bragg-Peaks: Structural Analysis of Complex Materials (First edition), Pergamon Press, Elsevier Ltd, ISBN 0-08-042698-0, Amsterdam 101 Elsasser, C., Krimmel, H., Fahnle, M., Louie, S. & Chan, C. (1998). Ab initio study of iron and iron hydride: III. Vibrational states of H isotopes in Fe, Cr and Ni. Journal of Physics: Condensed Matter, Vol. 10, No. 23, pp. 5131 [16 pages], ISSN 0953-8984 European Commission (2003). World energy, technology and climate policy outlook 2030 —WETO, in: Directorate-General for Research Energy, EUR 20366, Available from: http://ec.europa.eu/research/energy/pdf/weto_final_report.pdf Farrow, C., Juhas, P., Liu, J., Bryndin, D., Božin, E., Bloch, J., Proffen, T. & Billinge, S. (2007). PDFfit2 and PDFgui: computer programs for studying nanostructure in crystals. Journal of Physics: Condensed Matter, Vol. 19, No. 33, pp. 335219 [7 pages], ISSN 0953-8984 Felderhoff, M., Weidenthaler, C., von Helmoltb, R. & Eberleb, U. (2009). Hydrogen storage: the remaining scientific and technological challenges. Physical Chemistry Chemical Physics, Vol. 9, No. 21, pp. 2643-2653, ISSN 1463-9084 Fernández, A., Deprez, E. & Friedrichs, O. (2011). A comparative study of the role of additive in the MgH2 vs. the LiBH4–MgH2 hydrogen storage system. International Journal of Hydrogen Energy, Vol. 36, No. 6, pp. 3932-3940, ISSN 0360-3199 Fukai, Y. (1993). The Metal-Hydrogen System: Basic Bulk Properties (First edition), Springer-Verlag, ISBN 3540-556370, Berlin Grochala, W. & Edwards, P. (2004). Thermal Decomposition of the Non-Interstitial Hydrides for the Storage and Production of Hydrogen. Chemical Reviews, Vol. 104, No. 3, pp. 1283-1315, ISSN 0009-2665 Hlukhyy, V., Rodewald, U. & Pöttgen, R. (2005). Magnesium-Tin Substitution in NiMg2. Zeitschrift für anorganische und allgemeine Chemie, Vol. 631, No. 15, pp. 2997–3001, ISSN 1521-3749 Hohenberg, P. & Kohn, W. (1964). Inhomogeneous Electron Gas. Physical Review B, Vol. 136, No. 3B, pp. B864-B871, ISSN 1098-0121 Hong, S., Bae, J., Kwon, S. & Song, M. (2011). Hydrogen storage properties of Mg-23.5Ni- xCu prepared by rapid solidification process and crystallization heat treatment. International Journal of Hydrogen Energy, Vol. 36, No. 3, pp. 2170-2176, ISSN 0360-3199 Hsu, F., Hsu, C., Chang, J., Lin, C., Lee, S. & Jiang, C. (2010). Structure and hydrogen storage properties of Mg2Cu1−xNix (x = 0–1) alloys. International Journal of Hydrogen Energy, Vol. 35, No. 24, pp. 13247-13254, ISSN 0360-3199 102 Huot, J., Liang, G., Boily, S., Neste, A. & Schulz, R. (1999). Structural study and hydrogen sorption kinetics of ball-milled magnesium hydride. Journal of Alloys and Compounds, Vol. 293-295, pp. 495-500, ISSN 0925-8388 Jain, I., Lal, C. & Jain, A. (2010). Hydrogen storage in Mg: A most promising material. International Journal of Hydrogen Energy, Vol. 35, No. 10, pp. 5133-5144, ISSN 0360-3199 Jain, I. (2009). Hydrogen the fuel for 21st century. International Journal of Hydrogen Energy, Vol. 34, No. 17, pp. 7368-7378, ISSN 0360-3199 Jeon, K., Moon, H., Ruminski, A., Jiang, B., Kisielowski, C., Bardhan, R. & Urban, J. (2011). Air-stable magnesium nanocomposites provide rapid and high-capacity hydrogen storage without using heavy-metal catalysts. Nature Materials, Vol. 10, No. 4, pp. 286-290, ISSN 1476-4660 Jurczyk, M., Nowak, M., Szajek, A. & Jezierski, A. (2011). Hydrogen storage by Mg-based nanocomposites. International Journal of Hydrogen Energy, (In press - available online 27 April 2011), ISSN 0360-3199 Jurczyk, M., Okonska, I., Iwasieczko, W., Jankowska, E. & Drulis H. (2007). Thermodynamic and electrochemical properties of nanocrystalline Mg2Cu-type hydrogen storage materials. Journal of Alloys and Compounds, Vol. 429, No. 1-2, pp. 316-320, ISSN 0925-8388 Kalinichenka, S., Röntzsch, L., Riedl, T., Gemming, T., Weißgärber, T. & Kieback, B. (2011). Microstructure and hydrogen storage properties of melt-spun Mg–Cu–Ni–Y alloys. International Journal of Hydrogen Energy, Vol. 36, No. 2, pp. 1592-1600, ISSN 0360-3199 Larson, A., von Dreele, R. (2004). GSAS Generalized Structure Analysis System, LANSCE, Los Alamos. Massa, W. (2008). Crystal Structure determination. Springer-Verlag, Berlin Heidelberg. ISBN 978-3540206446 Milanese, J., Girella, A., Garroni, S., Bruni, G., Berbenni, V., Matteazzi, P. & Marini, A. (2010a). Effect of C (graphite) doping on the H2 sorption performance of the Mg–Ni storage system. International Journal of Hydrogen Energy, Vol. 35, No. 3, pp. 1285-1295, ISSN 0360-3199 Milanese, C., Girella, A., Bruni, G., Cofrancesco, P., Berbenni, V., Matteazzi, P. & Marini, A. (2010b). Mg-Ni-Cu mixtures for hydrogen storage: A kinetic study. Intermetallics, Vol. 18, No. 2, pp. 203-211, ISSN 0966-9795 103 Milanese, C., Girella, A., Bruni, G., Cofrancesco, P., Berbenni, V., Villa, M., Matteazzi, P. & Marini, A. (2008). Reactivity and hydrogen storage performances of magnesium–nickel– copper ternary mixtures prepared by reactive mechanical grinding. International Journal of Hydrogen Energy, Vol. 33, No. 17, pp. 4593-4606, ISSN 0360-3199 Noritake, T., Towata, S., Aoki, M., Seno, Y., Hirose, Y., Nishibori, E., Takata, M. & Sakata, M. (2003). Charge density measurement in MgH2 by synchrotron X-ray diffraction. Journal of Alloys and Compounds, Vol. 356-357, pp. 84-86, ISSN 0925-8388 Orimo, S. & Fujii, H. (2001). Materials science of Mg-Ni-based new hydrides. Applied Physics A, Vol. 72, No. 2, pp. 167-186, ISSN 1432-0630 Ramirez-Cuesta, A. (2004). aCLIMAX 4.0. 1, The new version of the software for analyzing and interpreting INS spectra. Computer Physics Communications. Vol. 157, No. 3, pp. 226- 238. Ravindran, P., Vajeeston, P., Fjellvåg, H. & Kjekshus, A. (2004). Chemical-bonding and highpressure studies on hydrogen-storage materials. Computational Materials Science, Vol. 30, No. 3-4, pp. 349-357, ISSN 0927-0256 Reilly, J. & Wiswall, R. (1967). The Reaction of Hydrogen with Alloys of Magnesium and Copper. Inorganic chemistry, Vol. 6, No. 12, pp. 2220-2223, ISSN 0020-1669 Reilly, J. & Wiswall, R. (1968). The Reaction of Hydrogen with Alloys of Magnesium and Nickel and the Formation of Mg2NiH4. Inorganic chemistry, Vol. 7, No. 11, pp. 2254-2256, ISSN 0020-1669 Reule, H., Hirscher, M., Weißhardt, A. & Kronmüller, H. (2000). Hydrogen desorption properties of mechanically alloyed MgH2 composite materials. Journal of Alloys and Compounds, Vol. 305, No. 1-2, pp. 246-252, ISSN 0925-8388 Ross, D. (2008). Neutron scattering studies for analysing solid state hydrogen storage, in: Solid State Hydrogen Storage Materials and Chemistry Walker, G. Ed. CRC Woodhead Publishing: Cambridge, England, pp. 135-172. ISBN 9781845692704 Rude, L., Nielsen, T., Ravnsbæk, D., Bösenberg, U., Ley, M., Richter, B., Arnbjerg, L., Dornheim, M., Filinchuk, Y., Besenbacher, F. & Jensen, T. (2011). Tailoring properties of borohydrides for hydrogen storage: A review. Physica Status Solidi A, Vol. 208, No. 8, pp. 1754-1773, ISSN 1862-6300 104 Sabitu, S., Gallo, G. & Goudy, A. (2010). Effect of TiH2 and Mg2Ni additives on the hydrogen storage properties of magnesium hydride. Journal of Alloys and Compounds, Vol. 499, No. 1, pp. 35-38, ISSN 0925-8388 Schimmel, H., Huot, J., Chapon, L., Tichelaar, F. & Mulder, F. (2005). Hydrogen Cycling of Niobium and Vanadium Catalyzed Nanostructured Magnesium. Journal of the American Chemical Society, Vol. 127, No. 41, pp. 1438-14354, ISSN 0002-7863 Schlapbach, L. & Züttel, A. (2001). Hydrogen-storage materials for mobile applications. Nature, Vol. 414, pp. 353-358, ISSN 0028-0836 Senegas, J., Mikou, A., Pezat, M. & Darriet, B. (1984). Localisation et diffusion de l'hydrogene dans le systeme Mg2Ni-H2: Etude par RMN de Mg2NiH0,3 et Mg2NiH4. Journal of Solid State Chemistry, Vol. 52, No. 1, pp. 1-11, ISSN 0022-4596 Setten, M., Wijs, G. & Brocks, G. (2007). Ab initio study of the effects of transition metal doping of Mg2NiH4. Physical Review B, Vol. 76, pp. 075125 [8 pages], ISSN 1098-0121 Siegel, D., Wolverton, C. & Ozoliņš, V. (2007). Thermodynamic guidelines for the prediction of hydrogen storage reactions and their application to destabilized hydride mixtures. Physical Review B, Vol. 76, No. 13, pp. 134102 [6 pages], ISSN 1098-0121 Simičić, M., Zdujić, M., Dimitrijević, R., Nikolić-Bujanović, L. & Popović N.(2006). Hydrogen absorption and electrochemical properties of Mg2Ni-type alloys synthesized by mechanical alloying. Journal of Power Sources, Vol. 158, No. 1, pp. 730-734, ISSN 0378-7753 Sohn, H. & Emami, S. (2011). Kinetics of dehydrogenation of the Mg–Ti–H hydrogen storage system. International Journal of Hydrogen Energy, Vol. 36, No. 14, (July 2011), pp. 8344-8350, ISSN 0360-3199 Squires, G. (1978). Introduction to the theory of Thermal Neutron Scattering. Dover Publications Inc., Mineola, New York, ISBN 978-0486694474 Suryanarayana, C. (2008). Recent developments in mechanical alloying. Reviews on Advanced Materials Science, Vol. 18, No. 3, pp. 203-211, ISSN 1605-8127 Tan, Z., Chiu, C., Heilweil, E. & Bendersky, L. (2011a). Thermodynamics, kinetics and microstructural evolution during hydrogenation of iron-doped magnesium thin films. International Journal of Hydrogen Energy, Vol. 36, No. 16, pp. 9702-9713, ISSN 0360-3199 105 M.H. Braga et al. / Journal of Alloys and Compounds 436 (2007) 278–284 279 Table 1 Chemical composition (in atomic fraction) of the studied samples of the Cu–Li–Mg system Samples x(Cu) x(Li) x(Mg) 1 0.313 0.032 0.655 2 0.304 0.056 0.640 3 0.312 0.066 0.622 4 0.355 0.067 0.578 5 0.290 0.096 0.614 Fig. 1. Isothermal section, at room temperature (293 K), of the Cu–Li–Mg system [8]. The symbols represent the compositions of the studied samples as well as of CuMg2Lix(x≈0.34) and CuMg2−xLix(x≈0.11). Samples were milled to powder of <200 # and studied by XRD on a Philips X’Pert Pro MPD using primary monochromated radiation CuK␣1 (λ= 0.154060 nm) by a symmetric Ge (1 1 1) crystal alpha 1R. Patterns were collected from 5◦to 120◦(2θ) under steps of 0.01◦and counting time of 10 s. Reflections of the compound under analysis were present on the XRD patterns of all the samples studied and indexed (after removal of the corresponding peaks of the other phases present in each sample). Crysfire [9] and McMaille [10] softwares were used for indexing and hence determining the phase cell and lattice parameters. In Crysfire, the hexagonal cell was found by Dicvol [11], Taup [12] and Treor [13]. It was used Checkcell [14] and Le Bail method [15], implemented on FullProf [16], for obtaining the “best solution” for the space group. For the refinement it was used Rietveld method, after Le Bail fitting, also implemented by FullProf software. Peaks profiles were modelled using pseudo-Voigt functions, whereas the background was firstly interpolated using WinPLOTR [17] and refined at the end. The presence of CuMg2−xLix, as well as of Cu2Mg, was detected in all samples (Figs. 2–6); Cu2Mg was even present in sample 1 although in a rather lower content (Fig. 3). Even if the presence of CuMg2in samples 1 (Fig. 6), 2 and 3 could not be detected from SEM/EDS analysis, the XRD patterns revealed its presence. On the other hand, no CuMg2phase could be detected on samples 4 and 5. As in the XRD pattern of sample 4 only two phases were observed: CuMg2−xLixand Cu2Mg (with a known cubic Laves-C15 structure and space group Fd-3m (2 2 7) – origin choice 2 with cell parameters a=b=c= 7.034 nm) [18], this sample was chosen for determining the structure of the ternary phase. 3. Results and discussion Table 2 presents the crystallographic data for sample 4 and, for comparison, the crystallographic data of NiMg2(H, D)x Fig. 2. Microstructure of sample 4 (magnification 500×, energy 15 keV)— Cu2Mg: light grey and CuMg2−xLix(x≈0.11):medium grey (sample’s micro porosity: dark grey). (with x≈0.3); Fig. 4 is a zoom of the obtained XRD pattern (2θ=10 ◦–90◦). Indexing results for CuMg2−xLixpointed out a hexagonal cell. The best fit was obtained with Dicvol (figures of merit M(20) = 50.8 and F(20) = 47.7) and with McMaille (figures of merit M(20) = 40.7 and F(20) = 40.5). It was used Checkcell for obtaining the “best solution” for the space group, based on the ratio between the observed and the calculated peaks for a particular cell/space group combination. Fig. 3. Microstructure of sample 1 (magnification 500×, energy 15 keV)— Cu2Mg: light grey and CuMg2−xLix(x≈0.11) + CuMg2: medium grey (sample’s micro porosity: dark grey). 112 280 M.H. Braga et al. / Journal of Alloys and Compounds 436 (2007) 278–284 Fig. 4. Zoom of the XRD pattern for sample 4. Reflections of CuMg2−xLix (x≈0.11) (not assigned) and Cu2Mg phases. The space groups obtained for the “best solution” were those for which there were no extinguished reflections due to the lattice type, although there were limiting reflections due to the space group symmetry corresponding to 0 0 l with l= 3n. P6222 space group was in the assigned group. Another method, Le Bail method implemented on FullProf software, was used for obtaining the space group: it was realized an extraction of intensities in a space group without extinctions (P6222). Extracted intensities were very carefully checked in order to determine systematic extinctions if any. The visual examination of a zoomed part of the pattern showing the refinement results was essential for concluding to the absence or presence of a reflection. For the obtained space groups, extracted structure factors were calculated. If the cell was false, this step of extracting structure factors would reveal it by a very bad correspondence between the observed and cell-constrained calculated patterns. The space group was found to be P6222 (1 8 0) (Tables 2 and 3) and the corresponding structure was refined, as previously mentioned, using the Rietveld method also imple- Fig. 5. Zoom of the XRD pattern for sample 5. Reflections of CuMg2−xLix (x≈0.11) (not assigned) and Cu2Mg phases. Table 2 Crystallographic data for CuMg2−xLix(x≈0.11) (on sample 4) and for NiMg2(H, D)x(x≈0.3), from Ref. [22], obtained by neutron diffraction (converted to Cu K␣1in (PDF-2 01-085-0910, 1997)) hkl d(nm) calc. for CuMg2−xLixa I/I0obs. by XRD I/I0obs. for NiMg2 (H, D)xfrom [5] converted in Ref. [22] 1 0 0 0.4556 (7) 18.7 20.3 0 0 3 0.4550 (7) 65.4 59.2 1 0 1 0.4321 (6) 32.9 31.2 1 0 2 0.3789 (5) 21.7 14.3 1 0 3 0.3219 (3) 9.4 4.6 1 0 4 0.2731 (2) 0.1 1.6 1 1 0 0.2630 (2) 0.3 0.1 1 1 1 0.2583 (2) 2.9 4.1 1 1 2 0.2454 (2) 42.7 38.3 1 0 5 0.2342 (2) 52.3 39.7 2 0 0 0.2278 (2) 48.2 51.1 1 1 3 0.2277 (2) 11.6 31.2 0 0 6 0.2275 (2) 22.5 19.0 2 0 1 0.2247 (2) 0.0 0.0 2 0 2 0.2161 (1) 0.0 0.1 1 1 4 0.2083 (1) 7.6 5.6 2 0 3 0.2037 (1) 100.0 100.0 1 0 6 0.2035 (1) 14.8 10.3 2 0 4 0.1895 (1) 0.0 0.0 1 1 5 0.1894 (1) 4.7 2.8 1 0 7 0.1793 (1) 7.5 3.2 2 0 5 0.1749 (1) 0.0 0.1 2 1 0 0.1722 (1) 2.5 1.2 1 1 6 0.1721 (1) 2.4 – 2 1 1 0.1708 (1) 4.3 3.4 2 1 2 0.1670 (1) 3.8 1.7 aThe d-spacings were determined from the best fitting of the raw data using the appropriate structural model. mented in Fullprof. A Rietveld refinement was carried out between 17◦and 115◦ (2θ). Although the refinement of the structure of the Cu2Mg phase was not the aim of this work, it was done together with CuMg2−xLix, only by control reasons, during the process. CuMg2−xLixand Cu2Mg were refined out of 133 and 21 reflections, respectively; five reflections (with I/I0< 1) were discarded due to the contribution (compare Fig. 7 with Fig. 4) of some impurities present in low contents. For Cu2Mg, with Cu (Wyck = 16d) for x=y=z= 1/2, and Mg (Wyck = 8a) for x=y=z= 1/8, it was obtained, after the above mentioned Rietveld refinement for control, a Bragg reliability factor RB= 7.34%. NiMg2and CuMg2metallic sublattices were found to be isostructural. Several alternatives were tried for possible Li positions on the ternary phase studied: (I) (x,y,z) Wyck = 12k, with x= 0.031 (1) nm, y= 0.018 (1) nm, z= 0.0515 (3) nm and Biso = 0.0213 (4) nm2; (II) (1/2, 0, z) Wyck = 6f, with z= 0.01081 (1) nm, Biso = 0.0106 (4) nm2(with Li occupying Mg1atomic positions); 113 M.H. Braga et al. / Journal of Alloys and Compounds 436 (2007) 278–284 281 Fig. 6. Zoom of the XRD pattern for sample 1. Reflections of CuMg2−xLix(x≈0.11) (not assigned), CuMg2and Cu2Mg phases (a) Zoom between 18◦and 48◦ (2θ). (III) (1/2, 0, z) Wyck = 6f (with Li occupying Mg1atomic positions) and (x, 0, 1/2) Wyck = 6h; (IV) (x,y,z) Wyck = 12k and (x,2x, 1/2) Wyck = 6j; (V) (1/2, 0, z) Wyck = 6f (with Li occupying Mg1atomic positions) and (x,2x, 1/2) Wyck = 6i (with Li occupying Mg2 atomic positions). Options I, III and IV are the same as those for hydrogen in NiMg2(H, D)x(x≈0.3) [4]. Fig. 7. XRD pattern for sample 4 after the refinement of CuMg2−xLix(x≈0.11). (1) observed—points, and calculated—continuous line (discontinuities correspond to the reflections not considered in the refinement), (2) Bragg positions, and (3) difference between observed and calculated patterns. The best reliability factors and the most stable parameters were obtained with alternatives I and II (Table 3). Final results of Rietveld refinement for CuMg2Lix(x≈0.34) – I and CuMg2−xLix(x≈0.11) – II were obtained at the same time as Le Bail fit was being performed on Cu2Mg. From SEM/EDS and XRD experiments, it can be concluded that samples 4 and 5 belong to the two-phase region: Fig. 8. Schematic representation of Cu6Mg11.367Li0.633 structure.White spheres denote Cu, grey Mg2and black Mg1/Li. 114 282 M.H. Braga et al. / Journal of Alloys and Compounds 436 (2007) 278–284 Cu2Mg + CuMg2Lix(Figs. 2, 4 and 5); on the other hand, the three remaining samples are located in the three-phase region Cu2Mg + CuMg2−xLix+ CuMg2(Figs. 3 and 6). These observations are more consistent with a ternary compound’s stoichiometry like CuMg2Lix(x≈0.34) – I than with one like CuMg2−xLix (x≈0.11) – II (Fig. 1). Results from SEM/EDS of the relation x(Mg)/x(Cu), for the present samples and for those that were studied in Ref. [19], lead to the conclusion that x(Mg)/x(Cu)av. = 1.87 which is more in agreement with the presence of a phase with stoichiometry CuMg2−xLix(x≈0.11) – II. The average content of Li in the phase in study, x(Li)av. = 0.04 (from SEM/EDS results), has to be considered carefully because Li was obtained by difference, as detailed in Ref. [8]. In the refinement of the structure of Cu–Li–Mg ternary’s phase, the best reliability and the most stable parameters concerning Li position were obtained for (x,y,z) Wyck = 12k corresponding to CuMg2Lix(x≈0.34) – I (Table 3). Neverthe- Table 3 Details of the refinement of the structure of CuMg2Lix(x≈0.34)—I and CuMg2−xLix(x≈0.11)—II 115 M.H. Braga et al. / Journal of Alloys and Compounds 436 (2007) 278–284 283 Table 3 (Continued ) Refinement was done together with Le Bail fit of Cu2Mg (both present in sample 4). 116 284 M.H. Braga et al. / Journal of Alloys and Compounds 436 (2007) 278–284 less, bond distances between Cu–Li seem to be rather small (Cu1–Li = 0.143 nm, Cu2–Li = 0.170 nm—Table 3). Hence, this compound was discarded at the last analysis, in spite of the fact that the literature values vary: for Li covalent radius from 0.068 nm [20] to 0.134 nm [21], and for Cu covalent radius from 0.138 nm [21] to 0.152 nm [20], making this last compound (II) as the chosen one (Fig. 8). 4. Conclusions 1. The results for the ternary phase of the Cu–Li–Mg system, when put in comparison with those for NiMg2(H, D)x (x≈0.3), seem to confirm the analogy between space groups, lattice parameters and crystal structures of the metallic sublattices (CuMg2and NiMg2, respectively). NiMg2presents, at room temperature, a hexagonal structure (space group P6222) and NiMg2(H, D)x(x≈0.3) is a solid solution of hydrogen in NiMg2that preserves the structure of the NiMg2. However, in the Cu–Li–Mg system, CuMg2has an orthorhombic structure, with space group Fddd, with a= 0.9044 nm, b= 0.5275 nm, c= 1.8328 nm and CuMg2−xLixis hexagonal, with space group P6222. It seems that Li stabilizes CuMg2into a hexagonal structure. 2. Mel’nik et al., [1] refers the existence of a ternary phase on the Cu–Li–Mg system: Cu8Li2Mg15, with a composition of x(Cu) = 0.34, x(Li) = 0.08 and x(Mg) = 0.60. Results obtained in this work pointed the existence of a phase CuMg2−xLix(x≈0.11) with composition: x(Cu) = 0.333, x(Li) = 0.035 and x(Mg) = 0.632 corresponding to a formula sum Cu6Mg11.367Li0.633. Acknowledgments The authors appreciate very much Dr. Armel Le Bail’s contribution on helpful discussions and Dr. M. H¨ am¨ al¨ ainen’s help for the supply of the samples. References [1] E.V. Mel’nik, M.F. Mitrofanova, P.I. Kripyakevich, M.Yu. Teslyuk (decd.), A.N. Malinkovich (L’vov), Russ. Metall. 3 (1976) 152– 156. [2] K. Schubert, K. Anderko, Z. Metallkd. 42 (1951) 321–325. [3] F. Gingl, P. Selvam, K. Yvon, Acta Cryst. 49B (1993) 201–203. [4] J. Schefer, P. Ficher, W. H¨ alg, F. Stucki, L. Schlapbach, J.J. Didisheim, K. Yvon, A.F. Andresen, J. Less-Common Met. 7 (1980) 65–73. [5] J. Senegas, A. Mikou, M. Pezat, B. Darriet, J. Solid State Chem. 52 (1984) 1–11. [6] B. Darriet, J.L. Soubeyroux, M. Pezat, D. Fruchart, J. Less-Common Met. 103 (1984) 153–162. [7] K. Zeng, T. Klassen, R. Oelerich, R. Bormann, J. Alloys Comp. 283 (1999) 213–224. [8] M.H. Braga, L.F. Malheiros, M. H¨ am¨ al¨ ainen, Thermochim. Acta 34 (2000) 47–54. [9] R. Shirley, The Crysfire 2002 System for Automatic Powder Indexing: User’s Manual, The Lattice Press, Guildford, Surrey, England, 2002. [10] A. Le Bail, http://www.cristal.org/McMaille/index.html (2002). [11] A. Boultif, D. Lou¨ er, J. Appl. Cryst. 24 (1991) 987–993. [12] D. Taupin, J. Appl. Cryst. 6 (1973) 380–385. [13] P.-E. Werner, L. Eriksson, M. Westdahl, J. Appl. Cryst. 18 (1985) 367– 370. [14] J. Laugier, B. Bernard, http://www.inpg.fr/LMGP, 2002. [15] A. Le Bail, H. Duroy, J.L. Fourquet, Mater. Res. Bull. 23 (1988) 447– 452. [16] J. Rodriguez-Carvajal, Program Fullprof (version 3.0, Nov 2004), D.B. Wiles, R. A. Young, A. Sakthivel, J. Appl. Cryst. 14 (1981) 149– 151. [17] T. Roisnel, J. Rodr´ ıguez-Carvajal, in: R. Delhez, E.J. Mittenmeijer (Eds.), Proceedings of the Seventh European Powder Diffraction Conference (EPDIC 7), 2000, pp. 118–123. [18] T. Ohba, Y. Kitano¨ y, Y. Komura¨ y, Acta Cryst. C40 (1984) 1–5. [19] M.H. Braga, PhD Thesis, Faculty of Engineering of the University of Porto, 1999. [20] Cambridge Crystallographic Data Centre, http://www.ccdc.cam.ac.uk/ products/csd/radii/, 2005. [21] M. Winter, The University of Sheffield and WebElements Ltd., UK, http://www.webelements.com/, 1993–2005. [22] J.L. Soubeyroux, D. Fruchart, A. Mikou, M. Pezat, B. Darriet, Mater. Res. Bull. 19 (7) (1984) 895–904, 01-085-0910 ICDD PDF-2 2003; from 030715 ICSD using POWD-12++, 1997. 117 118 Z. Kristallogr. Suppl. 26 (2007) 299-304 © by Oldenbourg Wissenschaftsverlag, München 299 HT-XRD in the study of Cu-Li-Mg M. H. Braga 1,* , J. Ferreira 2 , L. F. Malheiros 3 , M. Hämäläinen 4 1 GMM-IMAT, Dep. of Physics, FEUP, R. Dr. Roberto Frias s/n, 4200-465 Porto, Portugal 2 INETI Laboratory, R. da Amieira – P.O. Box 1089, 4466-956 S. Mamede de Infesta, Portugal 3 GMM-IMAT, Dep. of Metallurgical and Materials Engineering, FEUP, R. Dr. Roberto Frias s/n, 4200-465 Porto, Portugal 4 Laboratory of Materials Processing and Powder Metallurgy, HUT, Espoo, P.O. Box 6200, 02150 HUT, Finland * Contact author; e-mail: [email protected] Keywords: powder diffraction (RT and HT), Cu-Li-Mg, DSC/DTA, SEM/EDS Abstract. In a previous study on Cu-Li-Mg system, the authors of the present paper concluded that the ternary phase in that system corresponds to CuMg 2-x Li x (x ~ 0.11), with a hexagonal structure, space group P6 2 22 (180), and lattice parameters a = b = 0.5260 nm, c = 1.3649 nm [1]. The structure was refined by the Rietveld method [2]. In order to characterize the thermal behaviour of the ternary compound and to assess the Cu-Li-Mg phase diagram [3], HT-XRD measurements were performed on samples whose compositions were close to the one corresponding to the ternary compound. SEM/EDS measurements of the phases’ compositions in equilibrium, as well as DSC/DTA heating curves, contributed to the identification of the transition temperatures and the phases present in equilibrium. It was concluded that the ternary phase decomposes at ~ 702 ± 2K. Introduction The Cu-Li-Mg system has not been, till now, the object of many studies although it is one of the ternaries of the Al-Cu-Li-Mg system which has been deeply studied, at least near the quasicrystalline T2 (Al 6 Li 3 Cu) phase, and which has many applications in the aeronautic industry. Mel’nik et al. [4] referred to the existence of a ternary phase in the Cu-Li-Mg system: Cu 8 Li 2 Mg 15 with an orthorhombic structure (a = 0.524 nm, b = 0.899 nm, and c = 5.433 nm). Hämäläinen et al. [3] assessed the phase diagram of the system. Figure 1 presents two vertical sections for x(Li) = 0.04 and x(Li) = 0.07 from the Gibbs energy parameters obtained in [3]. In previous work, the present authors pointed out the existence of a phase with a stoichiometry close to that of Cu 8 Li 2 Mg 15 [5] and, in recent work, the latter phase was defined as being CuMg 2-x Li x (x ~ 0.11) [1]. Taking into account that there were no experimental data about the thermal behaviour of the ternary phase, nor the high temperature equilibria, 119 300 European Powder Diffraction Conference, EPDIC 10 and that quenching seemed ineffective for such a narrow temperature range, some HT-XRD studies were developed. On the other hand, an important feature for a successful assessment is the crystallographic data; thus the models used to describe the phases should be supported by the phase’s crystal structure. Experimental DSC/DTA measurements In the first part of this work, thirty samples were prepared in a resistance furnace and their compositions were analysed by AAS [5]. DSC/DTA experiments were performed on samples belonging to the vertical sections x Mg ~ 0.512, x Cu ~ 0.097, x Cu ~ 0.039 and x Li ~ 0.050 of the Cu-Li-Mg system under Ar atmosphere using quasi-hermetic stainless steel crucibles. Figure 1 presents some experimental data and figure 2 shows a DSC/DTA curve for a sample with x(Cu) = 0.088, x(Li) = 0.19, x(Mg) = 0.717. In the second part of this work, samples were prepared in order to have compositions close to Cu 8 Li 2 Mg 15 . Those samples were prepared and analysed in a similar way to the previous ones [1]. 200 400 600 800 1000 1200 1400 00.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1. 0 x ( M g) Liquid 1 (Cu) + Cu 2 Mg (Mg) + CuMg 2 Cu 2 Mg + CuLi x Mg 2-x + CuMg 2 Liq1 + Cu 2 Mg a) b) c) d) e) f) a) Liq1+(Cu) b) (Li)+Cu 2 Mg+CuLi x Mg 2-x c) Cu 2 Mg+CuLi x Mg 2-x d) CuLi x Mg 2-x +CuMg 2 +(Li) e) CuMg 2 +(Li) f) CuMg 2 +(Li)+(Mg) T(K) 200 400 600 800 1000 1200 1400 00.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1. 0 x ( M g) Liquid 1 (Cu) + Cu 2 Mg (Mg) + CuMg 2 Cu 2 Mg + CuLi x Mg 2-x + CuMg 2 Liq1 + Cu 2 Mg a) b) c) d) e) f) a) Liq1+(Cu) b) (Li)+Cu 2 Mg+CuLi x Mg 2-x c) Cu 2 Mg+CuLi x Mg 2-x d) CuLi x Mg 2-x +CuMg 2 +(Li) e) CuMg 2 +(Li) f) CuMg 2 +(Li)+(Mg) 200 400 600 800 1000 1200 1400 00.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1. 0 x ( M g) Liquid 1 (Cu) + Cu 2 Mg (Mg) + CuMg 2 Cu 2 Mg + CuLi x Mg 2-x + CuMg 2 Liq1 + Cu 2 Mg a) b) c) d) e) f) a) Liq1+(Cu) b) (Li)+Cu 2 Mg+CuLi x Mg 2-x c) Cu 2 Mg+CuLi x Mg 2-x d) CuLi x Mg 2-x +CuMg 2 +(Li) e) CuMg 2 +(Li) f) CuMg 2 +(Li)+(Mg) T(K) 00.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1. 0 x ( M g) g) h) i) j) l) g) (Li)+Cu 2 Mg+CuMg 2 h) Liq2+Cu 2 Mg+CuMg 2 i) Liq1+Liq2+Cu 2 Mg j) Liq1+(Mg) l) Liq2+Cu 2 Mg m) (Li)+Cu 2 Mg m) 00.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1. 0 x ( M g) g) h) i) j) l) g) (Li)+Cu 2 Mg+CuMg 2 h) Liq2+Cu 2 Mg+CuMg 2 i) Liq1+Liq2+Cu 2 Mg j) Liq1+(Mg) l) Liq2+Cu 2 Mg m) (Li)+Cu 2 Mg m) a) b) Figure 1. Assessed vertical sections of the ternary system Cu-Li-Mg [3]: a) x(Li) = 0.04 and b) x(Li) = 0.07. Some of the DSC/DTA experimental data are represented by squares. The samples were polished and cleaned in pure ethanol in order to remove the oxide layer and to ensure a flat surface for a better thermal contact between sample and crucible. Experiments took place in a DSC of continuous flux (indeed a DTA) Shimadzu 50 that operates between room temperature and 1003 K, using stainless steel crucibles especially conceived for this propose. Crucibles were quasi-hermetic and the volume of the samples and the inner 120 Z. Kristallogr. Suppl. 26 (2007) 301 volume of the crucibles were very close. For each composition, at least four cycles of heating /cooling were made, always with four different samples. The heating/cooling rates used were 5, 10 and 20 K/min. Transformation temperatures in figure 1 correspond to a "0 K/min." heating rate. SEM/EDS measurements SEM/EDS analyses were also performed on each sample, at room temperature and after being annealed for 60 min. at 773 K followed by quenching into liquid N 2 . The Li content was found by difference [5]; high quality standards of Cu and Mg were used as patterns. A JEOL, JSM - 35C or JFM 6301 F was used. Figure 2 shows one of the photomicrographs of the samples studied which reveals the presence of a eutectic constituent. In Table 1 it can be seen the obtained EDS results for three of the system's component phases. 520 560 600 640 680 720 760 800 840 880 920 960 1000 -80 -60 -40 -20 0 20 40 60 80 Amostra Nº 17 x Li = 0.195; x Mg = 0.717 exo endo T (K) Potência (mW) Taxa de aquecimento/arrefecimento: 5 K/min. Curvas de CDV (Shimadzu 50) Power (mW) heating and cooling rate 5 K/min. m μ 10 BS; 22 kV; x 480 520 560 600 640 680 720 760 800 840 880 920 960 1000 -80 -60 -40 -20 0 20 40 60 80 Amostra Nº 17 x Li = 0.195; x Mg = 0.717 exo endo T (K) Potência (mW) Taxa de aquecimento/arrefecimento: 5 K/min. Curvas de CDV (Shimadzu 50) Power (mW) heating and cooling rate 5 K/min. 520 560 600 640 680 720 760 800 840 880 920 960 1000 -80 -60 -40 -20 0 20 40 60 80 Amostra Nº 17 x Li = 0.195; x Mg = 0.717 exo endo T (K) Potência (mW) Taxa de aquecimento/arrefecimento: 5 K/min. Curvas de CDV (Shimadzu 50) Power (mW) heating and cooling rate 5 K/min. m μ 10 BS; 22 kV; x 480 Figure 2. A DSC/DTA heating/cooling curve for a sample with x(Cu) = 0.088, x(Li) = 0.19, x(Mg) = 0.717, for a heating/cooling rate of 5 K/min. SEM backscatter photomicrograph (x 480) of the same sample after being annealed at 773 K for 60 min. followed by quenching into liquid N 2 . Two phases can be distinguished: (Mg) and/or (Li) in black and CuMg 2 in white. Li content was obtained by difference [5]. Table 1. SEM/EDS results. Average contents were obtained from more than twenty measurements. The error associated with the Li content cannot be precisely determined because Li was found by difference. Phase x(Cu) +0.005 x(Li) x(Mg) +0.005 Phase x(Cu) +0.006 x(Mg) +0.006 Phase x(Cu) +0.005 x(Mg) +0.005 CuLi x Mg 2-x 0.334 0.040 0.626 CuMg 2 0.674 0.326 Cu 2 Mg 0.669 0.332 RT and HT - powder XRD At room temperature, some powder samples were measured on a Philips X´Pert Pro MPD using CuKα 1 radiation (λ = 0.15406 nm) primary monochromated by a symmetric Ge (111) crystal [1]. When those samples were studied at high temperatures, instead of the primary monochromated radiation CuKα 1 and CuKα 2 (λ = 0.15443nm) were used, because the inten- 121 ARTICLE IN PRESS with steps of 0.011and a counting time of 10 s at 303, 473, 523, 573 and 623 K [9]. Results from differential scanning calorimeter (DSC) have shown that CuLi x Mg 2x decomposes at approximately 702 K [9]. 2.3. Neutron powder diffraction measurements (below room temperature) Time-of-flight (TOF) neutron diffraction data were collected on the NPDF neutron diffractometer at the Manuel Lujan Neutron Scattering Center at Los Alamos National Laboratory. This instrument is a high-resolution powder diffractometer located 32 m from the neutron spallation target. The data were collected at 60 K (for 7.7 h), 150 K (for 6.4 h) and 300 K (for 6.2 h), at an average proton beam current of 100 m A, using the 46, 90, 119 and 1481banks, which cover a d-spacing range from 0.12 to 7.2 ˚ A. A sample powder with 38.0 at% of CuLi x Mg 2x , 44.8 at% of CuMg 2 and 17.2 at% of Cu 2 Mg was manually grinded to a diameter of less than 37 m m and placed in a vanadium sample holder in a glove box under He. The vanadium sample holder contribution, at each T, was subsequently removed from the diffraction pattern. The structure was refined using the general structure analysis system (GSAS), a Rietveld profile analysis program developed by Larson and von Dreele [12]. Background coefficients, scale factors, phase fractions, profile function coefficients (sigma-1), sample absorption, atomic positions, lattice parameters, temperature factors, and occupancies (in the case of the phase CuLi x Mg 2x and for Mg and Li) were refined for the three phases (whenever applicable) making a total of 81 variables. For NPDF data, pair distribution function (PDF), G(r), was obtained via the Fourier Transform of the total diffraction pattern as indicated below, GðrÞ¼4 p r½ r ðrÞ r 0 ¼ 2 rZ 1 0 Q½SðQÞ1sinðQrÞdQ ð1Þ where r ðrÞis the microscopic pair density, r 0 is the average atomic number density, and rthe radial distance. Qis the momentum transfer (Q¼4 p sinð y Þ= l ). S(Q) is the normalized structure function determined from the experimental diffraction intensity [13]. PDF yields the probability of finding pairs of atoms separated by a distance r. PDF fittings were performed using the software PDFgui [14]. 2.4. Neutron powder diffraction measurements (above room temperature) Time-of-flight (TOF) neutron diffraction data were also collected on the neutron powder diffractometer (HIPPO) at the Manuel Lujan Neutron Scattering Center at Los Alamos National Laboratory. This instrument achieves very high neutron count rates by virtue of a short (9 m) flight path and large detector solid angle. The data were collected at approx. 313 K (for 1.6 h), 373 K (for 6.3 h), 453 K (for 2 h) and 523 K (for 0.90 h), at an average proton beam current of 100 m A, using the 90, and 144.451banks, which cover a d-spacing range from 0.12 to 4.80 ˚ A. A sample powder with 70.4 at% of CuLi x Mg 2x , 21.0 at% of CuMg 2 and 8.6 at% of Cu 2 Mg was enclosed in an aluminum sample holder to which a resistive heater and a thermocouple was attached. A standard laboratory temperature controller maintained the desired temperature during the measurement. A capillary tube attached to the sample holder allowed for pressure measurements and gas loading during the experiments. The aluminum sample holder contributes to the diffraction pattern. An aluminum phase was refined in all our data sets. The measured sample (including Al) had: 42.1 wt% (14.6 at%) of CuLi x Mg 2x +CuMg 2 +Cu 2 Mg and 57.9 wt% (86.4 at%) of Al. The structure was refined using GSAS. Background coefficients, scale factors, phase fractions, profile function coefficients (sigma-1), Fig. 3. NPDF after Rietveld refinement diffraction pattern at 300 K for 461bank (the results at the graph are only due to the 461bank). PDF experimental and fitted. Table 2 PDF fittings’ results at 60, 150 and 300 K obtained from NPDF data. T=60 K (NPDF) R wp =18.6% CuLi x Mg 2x (x=0.0370.04) CuMg 2 Cu 2 Mg Hexagonal–P6 2 22 (180) Orthorhombic–Fddd (70) Cubic–Fd-3m (227) a=b=5.250 (2) ˚ A; c=13.666 (7) ˚ A; a=5.240 (5) ˚ A; b=9.003 (8) ˚ A; c=18.25 (2) ˚ A a=b=c=7.058 (2) ˚ A; T=150 K (NPDF) R wp =17.6% CuLi x Mg 2x (x=0.1070.04) CuMg 2 Cu 2 Mg Hexagonal–P6 2 22 (180) Orthorhombic–Fddd (70) Cubic–Fd-3m (227) a=b=5.258 (3) ˚ A; c=13.67 (1) ˚ A a=5.248 (8) ˚ A; b=9.01 (1) ˚ A; c=18.25 (2) ˚ A a=b=c=7.066 (2) ˚ A T=300 K (NPDF) R wp =17.6% CuLi x Mg 2x (x=0.0870.06) CuMg 2 Cu 2 Mg Hexagonal–P6 2 22 (180) Orthorhombic–Fddd (70) Cubic–Fd-3m (227) a=b=5.273 (4) ˚ A; c=13.70 (2) ˚ A a=5.30 (2) ˚ A; b=9.00 (2) ˚ A a=b=c= 7.082 (4) c=17.92 (5) ˚ A M.H. Braga et al. / Journal of Solid State Chemistry 183 (2010) 10–19 13 128 ARTICLE IN PRESS sample absorption, atomic positions, lattice parameters, temperature factors, and occupancies (in the case of the phase CuLi x Mg 2x and for Mg and Li) were refined for the four phases (whenever applicable) making a total of 63 variables. 3. First principles data In a solid, where there are upward of 10 24 interacting electrons and nuclei per cubic centimeter, the resolution of the many body Schr¨ odinger equation for the electronic wavefunctions and energy eigenvalues is a big challenge. However, based on the periodicity of the structure of pure elements and perfectly ordered compounds, Bloch’s theorem shows that it is only necessary to solve the many body Schr¨ odinger equation within one unit cell; in the case of disordered compounds that is not the case and it is necessary to solve it within a supercell. The basic information that one wishes to obtain from quantum mechanical calculations in solids is the total electronic energy for various arrangements of atoms on various lattices. One of the most sophisticated solutions to the quantum mechanical problem in solids lies within the framework of the density functional theory (DFT) [15] using the local density approximation (LDA) [16] or the Generalized Gradient Approximation (GGA) [17]. The basic notion of these theories is to replace the true interacting many-body-system with a system of one electron in an effective potential due to all of the other electrons and nuclei. From a fundamental point of view, the one-electron functions are a unique tool for calculating the total energy and the electronic density of states; these functions have no particular physical meaning. But this simplification of the problem needs a self-consistent calculation and it is one of the major technical problems in the ab initio approach [18]. Density functional theory calculations with projector augmented wave (PAW) pseudopotentials [19], as implemented in the Vienna Ab Initio Simulation Package (VASP) code [20], implemented in MedeA [21], were performed. A plane wave cutoff of 355.18 eV, and k-spacings of 0.230 0.230 0.230 ˚ A 1 were used. Calculations were done in real space and were performed with P1 space group supercells containing 144 atoms (48 atoms of Cu, 96-nof Mg, and n=0 to 12 of Li). The supercells contained as many atoms as possible to allow better approximations with the real Li concentrations (but such that the time spent on calculations were not completely impractical). Since CuLi x Mg 2x is a disordered structure, it had to be obtained by randomly substituting Mg by Li in several Wyckoff 6fpositions (1/2, 0, z) or Wyckoff 6ipositions (x,2x, 0) or in both positions within the supercells. The generalized gradient approximation (GGA), and the Perdew–Burke– Ernzerhof (PBE) functional [22] were used, and no magnetic moments were included in the model. The total energy was minimized with respect to the volume (volume relaxation), the shape of the unit cell (cell external relaxation), and the position of the atoms within the cell (cell internal relaxation). The ab initio calculations furnish the total energy (or the cohesive energy) at T=0K, E F . The energy of formation is easily Fig. 4. Lattice parameters of the phase CuLi x Mg 2x as a function of the absolute temperature. It can be observed a linear dependence from the lattice parameters towards the temperature between 60 and 300 K as well as a small difference between lattice parameters obtained by first principles and obtained after refinement of the experimental data. Notice that the slope of the fitting line between 300 and 523 K is very similar to that obtained by X-ray diffraction [9].It can also be observed that results obtained by PDF fitting differ 0.45% in the case of aand 0.55% in the case of c, at 300 K. Lattice parameters calculated using the slope obtained after plotting the interatomic distance d, as a function of temperature for the (101) peak were also plotted (note that the only peaks that were not overlapped with peaks from another phase, depend on aand c). As the phase is not isotropic, the use of this value is not completely correct. Nonetheless, it allows us to be more confident about the non-interference of CuLi x Mg 2x and CuMg 2 during calculations, since these two phases have several overlapping peaks. Fig. 5. Isotropic temperature factors for all the atoms of the phase CuLi x Mg 2x as a function of the absolute temperature. It can be seen that between 60 and 300 K all temperature factors behave linearly and increase similarly with temperature. U iso is related with the Debye–Waller factor, B iso , by: B iso =8 p 2 U iso . M.H. Braga et al. / Journal of Solid State Chemistry 183 (2010) 10–1914 129 ARTICLE IN PRESS calculated by the relation D f E¼EFXx i E O i ð2Þ where E O i is the total energy (or cohesive energy) of iin its stable state at T=0 K. In the following we will assume that the enthalpy of formation is equal to the energy of formation. In addition to the total energies, ab initio calculations allowed us to obtain the values of the lattice parameters at 0 K. We performed most of the calculations for CuLi x Mg 2x at least twice (from 1 to 5 atoms of Li calculations were performed three or more times) with different random Li positions. We also calculated the X-ray diffraction pattern from the ab initio results. The code is implemented in MedeA and is based on the LAZY-PULVERIX computer program [23] that calculates the position of the diffraction lines from Bragg’s law and their d-spacings. The diffraction intensity I hkl is calculated as, I hkl ¼MLPF 2 hkl ð3Þ where Mis the multiplicity factor of a powder line, Lis the Lorentz factor and Pis the polarization factor. The structure factor F hkl is defined by, F hkl ¼X unit-cell i f j O j exp½2 p iðhx j þkx j þlz j Þexp B j sin 2 y l 2 !ð4Þ where f j is the atomic scattering factor of atom j,O j is the occupation factor at site x j ,y j ,z j , for atom jand B j the Debye– Waller factor in ˚ A 2 for atom j. Table 3 Rietveld refinement’s results at 313, 373, 453 and 523 K obtained from NPDF data. 313 K (HIPPO) wR p =3.79%; R p =2.89% Wt. Frac. (CuLi 0.09 Mg 1.91 )=0.280 (8); Wt. Frac. (CuMg 2 )=0.090 (3); Wt. Frac. (Cu 2 Mg)=0.041 (1); Wt. Frac. (Al)=0.589 (7); CuLi 0.09 Mg 1.91 : hexagonal–P6 2 22 (180) CuMg 2 : orthorhombic–Fddd (70) a=b=5.254 (2) ˚ A; c=13.620 (6) ˚ A; r =3.384 g/cm 3 a=5.264 (4) ˚ A; b=9.026 (5) ˚ A ; c=18.33 (1) ˚ A; r =3.422 g/cm 3 Cu1: x=0; y=0; z¼ 1 2 ; occ.=1; U iso 100=1.3 (1) ˚ A 2 Cu2: x¼ 1 2 ;y=0;z¼ 1 2 ; occ.=1; U iso 100=1.4 (1) ˚ A 2 Mg1: x¼ 1 2 ; y=0; z=0.1092 (4); occ.=0.91 (2); U iso 100=1.4 (2) ˚ A 2 Mg2: x=0.164 (1); y=0.328 (2); z=0; occ.=1; U iso 100=2.1 (2) ˚ A 2 Li1: x¼ 1 2 ; y=0; z=0.1092 (4); occ.=0.09 (2); U iso 100=1.4 (2) ˚ A 2 373 K (HIPPO) wR p =3.52%; R p =2.36% Wt. Frac. (CuLi 0.08 Mg 1.92 )=0.297 (5); Wt. Frac. (CuMg 2 )=0.083 (3); Wt. Frac. (Cu 2 Mg)=0.051 (2); Wt. Frac. (Al)=0.569 (5); CuLi 0.08 Mg 1.92 : hexagonal–P6 2 22 (180) CuMg 2 : orthorhombic–Fddd (70) a=b=5.255 (1) ˚ A; c=13.626 (4) ˚ A a=5.262 (2) ˚ A; b=9.033 (3) ˚ A r =3.386 g/cm 3 c=18.327 (6) ˚ A; r =3.420 g/cm 3 Cu1: x=0; y=0; z¼ 1 2 ; occ.=1; U iso 100=1.64 (9) ˚ A 2 Cu1: x¼ 1 8 ;y¼ 1 8 ; z=0.4976 (3); occ.=1; U iso 100=1.3 (2) ˚ A 2 Cu2: x¼ 1 2 ;y=0;z¼ 1 2 ; occ.=1; U iso 100=1.04 (1) ˚ A 2 Mg1: x¼ 1 8 ;y¼ 1 8 ; z=0.0416 (6); occ.=1; U iso 100=1.1 (3) ˚ A 2 Mg1: x¼ 1 2 ; y=0; z=0.1111 (3); occ.=0.919 (8); U iso 100=1.5 (1) ˚ A 2 Mg2: x¼ 1 8 ; y=0.4708 (1); z¼ 1 8 ; occ.=1; U iso 100=1.37 (3) ˚ A 2 Mg2: x=0.1719 (7); y=0.344 (1); z=0; occ.=1; U iso 100=2.0 (1) ˚ A 2 Li1: x¼ 1 2 ; y=0; z=0.1111 (3); occ.=0.081 (8); U iso 100=1.5 (1) ˚ A 2 453 K (HIPPO) wR p =3.38%; R p =2.28% Wt. Frac. (CuLi 0.09 Mg 1.91 )=0.300 (6); Wt. Frac. (CuMg 2 )=0.088 (3); Wt. Frac. (Cu 2 Mg)=0.053 (2); Wt. Frac. (Al)=0.559 (6); CuLi 0.09 Mg 1.91 : hexagonal–P6 2 22 (180) CuMg 2 : orthorhombic–Fddd (70) a=b=5.279 (2) ˚ A; c=13.689 (4) ˚ A a=5.286 (2) ˚ A; b=9.072 (3) ˚ A r =3.337 g/cm 3 c=18.406 (7) ˚ A; r =3.376 g/cm 3 Cu1: x=0; y=0; z¼ 1 2 ; occ.=1; U iso 100=1.9 (1) ˚ A 2 Cu1: x¼ 1 8 ;y¼ 1 8 ; z=0.4984 (4); occ.=1; U iso 100=1.6 (3) ˚ A 2 Cu2: x¼ 1 2 ;y=0;z¼ 1 2 ; occ.=1; U iso 100=1.51 (9) ˚ A 2 Mg1: x¼ 1 8 ;y¼ 1 8 ; z=0.0411 (6); occ.=1; U iso 100=1.3 (3) ˚ A 2 Mg1: x¼ 1 2 ; y=0; z=0.1106 (3); occ.=0.91 (1); U iso 100=2.0 (1) ˚ A 2 Mg2: x¼ 1 8 ; y=0.4741 (2); z¼ 1 8 ; occ.=1; U iso 100=2.63 (3) ˚ A 2 Mg2: x=0.1712 (8); y=0.342 (2); z=0; occ.=1; U iso 100=2.4 (1) ˚ A 2 Li1: x¼ 1 2 ; y=0; z=0.1106 (3); occ.=0.09 (1); U iso 100=2.0 (1) ˚ A 2 523 K (HIPPO) wR p =2.71%; R p =1.91% Wt. Frac. (CuLi 0.09 Mg 1.91 )=0.282 (6); Wt. Frac. (CuMg 2 )=0.089 (4); Wt. Frac. (Cu 2 Mg)=0.050 (2); Wt. Frac. (Al)=0.579 (6); CuLi 0.09 Mg 1.91 : hexagonal–P6 2 22 (180) CuMg 2 : orthorhombic–Fddd (70) a=b=5.279 (2) ˚ A; c=13.687 (5) ˚ A a=5.285 (2) ˚ A; b=9.076 (4) ˚ A r =3.334 g/cm 3 c=18.406 (8) ˚ A; r =3.374 g/cm 3 Cu1: x=0; y=0; z¼ 1 2 ; occ.=1; U iso 100=1.9 (1) ˚ A 2 Cu1: x¼ 1 8 ;y¼ 1 8 ; z=0.4978 (4); occ.=1; U iso 100=1.5 (3) ˚ A 2 Cu2: x¼ 1 2 ;y=0;z¼ 1 2 ; occ.=1; U iso 100=1.55 (1) ˚ A 2 Mg1: x¼ 1 2 ; y=0; z=0.1093 (4); occ.=0.91 (1); U iso 100=2.0 (2) ˚ A 2 Mg1: x¼ 1 8 ;y¼ 1 8 ; z=0.0442 (6); occ.=1; U iso 100=1.3 (3) ˚ A 2 Mg2: x=0.1692 (9); y=0.338 (2); z=0; occ.=1; U iso 100=2.3 (2) ˚ A 2 Mg2: x¼ 1 8 ; y=0.4726 (2); z¼ 1 8 ; occ.=1; U iso 100=2.5 (4) ˚ A 2 Li1: x¼ 1 2 ; y=0; z=0.1093 (4); occ.=0.09 (1); U iso 100=2.0 (2) ˚ A 2 M.H. Braga et al. / Journal of Solid State Chemistry 183 (2010) 10–19 15 130 ARTICLE IN PRESS We used the MedeA empty space finder (ESF) [21] to analyze the empty spaces of the CuLi x Mg 2x supercell. The ESF algorithm divides the supercell into so-called Voronoi cells around each atom [24]. (A Voronoi cell is defined to be the volume enclosing all points that are closer to the center atom than to all other atoms). The ESF module positions non-overlapping spheres at the vertices of the resulting polyhedral grid and maximizes their radii. In doing so the physical size of different atomic species is taken into account through a set of covalent radii. 4. Results and discussion 4.1. Analysis of the diffraction data The results from the refined data from NPDF at 60, 150 and 300 K can be found in Table 1 and indicate very good agreement between the experimental and model powder patterns. Phase fractions were varied for each pattern; the difference between the results is less than the refinement error, indicating phase stability over the investigated range of temperatures. The same statement is valid for the Li occupation and atomic parameters for both the Cu x LiMg 2x and CuMg 2 phases. Figs. 1–3 show the diffraction patterns at 60, 150 and 300 K and the Rietveld fits. Results from PDF fittings at 60, 150 and 300 K can be found in Table 2 and Figs. 1–3. They indicate good agreement between the experimental and fitted curves. The error associated with each parameter obtained by PDF fitting is always higher than obtained with Rietveld, still, for 300 K where this difference is more valuable, the associated errors are D a/a=0.45% and D c/c=0.55% (Fig. 4). In Fig. 4 it can be observed the result of the calculation of the lattice parameters after obtaining the slop of the interplanar distance d(˚ A) of the peak corresponding to the reflection (101) for Cu x LiMg 2x as a function of temperature using different values for Fig. 6. HIPPO diffraction pattern at 313 K for bank 901(the results at the graph are only due to the 901bank). Fig. 7. HIPPO diffraction pattern at 373 K for bank 901(the results at the graph are only due to the 901bank). Fig. 8. HIPPO diffraction pattern at 453K for bank 901(the results at the graph are only due to the 901bank). Fig. 9. HIPPO diffraction pattern at 523K for bank 901(the results at the graph are only due to the 901bank). M.H. Braga et al. / Journal of Solid State Chemistry 183 (2010) 10–1916 131 ARTICLE IN PRESS aand cat T=0 K. Although this peak, as the other more intense non-overlapping ones, depends on both aand clattice parameters and thus the extrapolation for aand ctemperature dependence is not straight forward, we wanted to ensure that the calculations of the parameters of Cu x LiMg 2x were not interfering too much with those of CuMg 2 since these two phases have many overlapping peaks. The lattice parameters of the phase Cu x LiMg 2x and the isotropic thermal parameters seem to depend linearly on temperature (Figs. 4 and 5) between 60 and 300 K. A harmonic crystal does not undergo thermal expansion since its equilibrium size does not depend on temperature. The thermal expansion of solids is a consequence of the anharmonicity of the lattice vibrations. The earliest approach to lattice vibrations was Einstein’s [25]. He proposed that all the Natoms of the crystal vibrate with 3Nequal frequencies as harmonic oscillators. Beyond Einstein’s approach, the Debye–Gr¨ uneisen approximation [26] is a good starting point for describing the contribution of lattice vibrations to the thermal expansion of solid crystals [27]. The theory is only directly applicable to crystals containing a single kind of atom; however, it has been used with success with simple compounds and with elements which do not have cubic symmetry [28]. Krishnan et al. [29] proposed an extension of Gr¨ uneisen’s law for an anisotropic crystal that would depend on two different thermal expansion coefficients (with respect to a=b and to c). Thus, we are expecting to find the expression of Gr¨ uneisen’s law for a=band for c, independently (Fig. 4). For metals, on the other hand, the excitation of electrons is as important as that of phonons; when included, it gives different behavior for the thermal expansion coefficient at low temperature. We have obtained for Cu x LiMg 2x the expansion coefficient with respect to a, a a =(1/a 340 K )da/dT(at constant pressure), approx. equal to 2.4 10 5 K 1 . The expansion coefficient with respect to c, a c =(1/c 340 K )dc/dT(at constant pressure), is approx. equal to 0.8 10 5 K 1 . The expansion coefficient of Cu is 1.7 10 5 K 1 and of Mg is 0.8 10 5 K 1 at 273 K [30].Wehave chosen 340 K to calculate the thermal expansion coefficient because, at this temperature, we are already in the high temperature range of our measurements. The Debye–Waller factors, B iso , that can be obtained from U iso (Fig. 5), by making B iso =8 p 2 U iso were also found to vary linearly between 60 and 300 K; additionally, all atoms seem to behave in a similar way with increasing temperature since the slope of the straight lines in Fig. 5 is very similar. The results from the refined data from HIPPO at approximately, 313, 373, 453 and 523 K can be found in Table 3 and show good agreement between experimental and model powder patterns. On HIPPO the aluminum sample holder diffracts strongly, and the weight fraction of Cu 2 Mg becomes relatively low (4–5 wt%). Refinement results for this phase were not included in Table 3 because the isotropic temperature factors became unstable and we were forced to constrain them. Still, the results seem to be very consistent vide, for example, Li occupancy and the atomic parameters for both the Cu x LiMg 2x and CuMg 2 phases. Figs. 6–9 show the diffraction patterns at 313, 373, 453 and 523 K together with the Rietveld fits. Fig. 4 shows the difference between the XRD data and the neutron diffraction data obtained on HIPPO. Notice that the relative difference between results obtained with X-ray and obtained with neutrons never exceeds 0.4% in the case of aand 0.5% in the case of c. Additionally, the slope of the fitting line between 300 and 523 K seems to be similar in both cases. Small differences between neutron diffraction and X-ray diffraction are expected since these two techniques ‘‘see’’ different constituents of the atom. The NPDF 300 K results are very similar to the 313 K HIPPO results, which is one more indicator of the results’ coherence. Fig. 10. First principles (ab initio) calculated enthalpies of formation for a mole with 144 atoms of the phase CuLi x Mg 2x in which 48 atoms are of Cu, 96nare of Mg and nof Li (x=n/48). It can be observed that the most stable composition corresponds to nbetween 3 and 4 (xA[0.0625,0.0833]). Fig. 11. Supercell of the phase CuLi x Mg 2x (x=0) with calculated empty spaces (in light grey: empty spaces, in dark grey: Cu atoms, in red: Mg atoms). (For interpretation of the references to the color in this figure legend, the reader is referred to the web version of this article.) M.H. Braga et al. / Journal of Solid State Chemistry 183 (2010) 10–19 17 132 ARTICLE IN PRESS 4.2. Analysis of the first-principles data We studied the enthalpy of formation of the alloy CuLi x Mg 2x as a function of the number of Li atoms in a supercell with 144 atoms in which 48 atoms are of Cu, 96-nare of Mg and nof Li (x=n/48). The most stable composition (that has the lowest enthalpy of formation) corresponds to nbetween 3 and 4 (xA[0.0625,0.0833]) (Fig. 10). One of the purposes of this work was to determine which structure was more stable: CuLi 0.34 Mg 2 (where Li occupies some of the Wyckoff 12kposition of a P6 2 22 hexagonal structure) or CuLi x Mg 2x (with xffi0.1, in which Li occupies some of the Mg Wyckoff 6fposition of a P6 2 22 hexagonal structure). For that, we have used the minimized structure of a supercell (Fig. 11) with n=0 (x=0; CuMg 2 with P6 2 22 hexagonal structure) and with n=4 (x=0.0833; CuLi 0.0833 Mg 1.9167 with P6 2 22 hexagonal structure) and calculated the empty space using the ESF MedeA module (Fig. 11). Results show that in the first case the radii of the empty spaces found are 0.410 ˚ Arrr0.650 ˚ A and in the second case 0.397 ˚ Arrr0.666 ˚ A. As the covalent radius of Li is 0.68 ˚ A[31],we do not expect to have Li occupy interstitial sites, even if some atoms of Li substitute the Mg atoms. Still, we have tried to refine the NPDF data at 60 and 150 K for several possibilities that always contained Li atoms in interstitial sites (just in interstitial and both substituting Mg and in an interstitial site) and we have always obtained negative occupancies for these atoms. Upon using first-principles calculations, it was not possible to determine which Mg sites are occupied with Li (Fig. 10). In a 2007 article, Zhou et al. [32] calculated the enthalpy of formation of a mole of CuMg 2 : D H=13.20 KJ/mol within the Fddd space group. In the present work we have obtained for CuMg 2 with hexagonal structure belonging to the P6 2 22 space group D H=11.66 KJ/mol, which is consistent with the results of Zhou et al. [32] since the orthorhombic Fddd structure is the stable structure for pure CuMg 2 . What we expected is that the substitution of Mg by Li would make the Fddd structure of the solid solution of CuMg 2 less stable until CuLi x Mg 2x (x=0.08) in the hexagonal P6 2 22 structure became more stable (note that the difference between the enthalpies of formation is small, of 1.5 KJ/mol of atoms). In the same article [32] the enthalpy of formation of CuMg 2 at 298 K was also determined using the CALPHAD method [33]: D H=9.6 KJ/mol (CALPHAD uses experimental results and theoretical models for the structure of the phases and for its Gibbs energies). The experimental value from [34] is D H=9.55 KJ/mol. As the enthalpy of formation of CuMg 2 will vary from 13.20 KJ/mol (T=0 K) to 9.6 KJ/mol (T=298 K), it still seems possible that at room temperature CuLi x Mg 2x (x=0.08) in the hexagonal P6 2 22 structure will be more stable than the orthorhombic solid solution of CuMg 2 for the same composition of Li. In Fig. 12, the XRD calculated pattern of CuLi 0.08 Mg 1.92 is compared with an experimental one for a sample containing both CuLi 0.08 Mg 1.92 and Cu 2 Mg. 5. Summary Prepared samples were invariably contaminated with Cu 2 Mg, CuMg 2 , or both. Nonetheless, the final product contained approximately 81.0 wt% (75.6 at%) of Cu x LiMg 2x ,thephasewewantedto study. Thermodynamics is probably responsible for this limitation on sample composition. Indeed, preliminary tests indicate that the final amount of CuLi x Mg 2x material recovered depends (at the very least) on the reaction temperature and (to a lesser extent) on reaction time. We have taken advantage of this fact to extract refined parameters of CuMg 2 between 60 and 523 K, and of Cu 2 Mg between 60and300KaswellasofCuLi x Mg 2x between 60 and 523 K. Both neutron diffraction and first-principles calculations indicate that the composition of the ternary compound should be CuLi x Mg 2x (x=0.08). Neither with first-principles calculations, nor with neutron diffraction, was it possible to distinguish between structures where Li substituted Mg at each of its sites individually or at both Fig. 12. Comparison between the diffraction pattern obtained using first principles and the LAZY PULVERIX code implanted in MedeA and XRD obtained at room temperature with a sample powder containing 81.0 wt% (75.6 at%) of CuLi x Mg 2x and 19.0 wt% (24.4 at%) of Cu 2 Mg. The calculated pattern was obtained for a B iso =1.00 ˚ A 2 for all the atoms. M.H. Braga et al. / Journal of Solid State Chemistry 183 (2010) 10–1918 133 ARTICLE IN PRESS sites. The refinement of the diffraction data revealed that the reliability factors did not change much for the three different structures, and nothing unusual happened with the other parameters (such as, for example, the Li occupancy). The calculated enthalpy of formation curve also shows differences for different occupancies of Li for the same number of Li atoms that are within the error bar. Thus we cannot draw any conclusions regarding the substituted Mg sites based on Rietveld refinement and first-principles calculations. The neutron data agrees best with the results of first-principles calculations when Li occupies Mg1 sites (1/2, 0, z). In this case, the Li occupancy corresponds to x=0.08 (in CuLi x Mg 2x ) when this value is calculated by means of Rietveld refinement and using firstprinciples methods. This agreement does not happen for the other possibilities. With PDF fittings we were allowed to go further. PDF does not see the average but the local structure and with PDF all results but those in which Li would substitute Mg1 sites, gave negative occupancies for Li. For Li substituting Mg1 we have obtained an average composition for CuLi x Mg 2x (x=0.07) which is in agreement with the other obtained results. By calculating empty space in our structures, we found that it was unlikely that Li can occupy interstitial sites, but we note that it is possible for H to occupy these empty spaces. Acknowledgments M.H. Braga would like to acknowledge Portuguese Science Foundation, FCT, for the sabbatical grant (SFRH/BSAB/791/2008). This work has benefited from the use of NPDF and HIPPO at the Lujan Center at Los Alamos Neutron Science Center, funded by DOE Office of Basic Energy Sciences. Los Alamos National Laboratory is operated by Los Alamos National Security LLC under DOE Contract DE-AC52-06NA25396. The upgrade of NPDF has been funded by NSF through grant DMR 00-76488. Appendix A. Supplementary material Supplementary data associated with this article can be found in the online version at doi:10.1016/j.jssc.2009.09.010. References [1] M.H. Braga, J. Ferreira, L. Malheiros, J. Alloys Compds. 436 (2007) 278–284. [2] M.H. Braga, L.F. Malheiros, International patent, WO2007046017. [3] M.H. Braga, L.F. Malheiros, National patent, PT103368. [4] M.H. Braga, A. Acatrinei, M. Hartl, S. Vogel, Th. Proffen, L. Daemen, ICNS, Knoxville, 2009. [5] J.J. Reilly, R.H. Wiswall, Inorg. Chem. 6 (12) (1967) 2220–2223. [6] L. Schlapbach, A. Zuttel, Nature 414 (2001) 353–358. [7] J. Senegas, A. Mikou, M. Pezat, B. Darriet, J. Solid State Chem. 52 (1984) 1–11. [8] V. Hlukhyy, U.Ch. Rodewald, R. P¨ ottgen, Z. Anorg. Allg. Chem. 631 (2005) 2997–3001. [9] M.H. Braga, J. Ferreira, L.F. Malheiros, M. Hamal¨ ainen, Z. Kristallogr. 26 (Suppl.) (2007) 299–304. [10] Match, /http://www.crystalimpact.com/S, 2009. [11] P.M. de Wolff, J.W. Visser, Absolute Intensities, Report 641.109, Technisch Physische Dienst, Delft, The Netherlands. Reprinted Powder Diffract 3 (1988) 202–204. [12] A.C. Larson, R.B. von Dreele, GSAS Generalized Structure Analysis System, LANSCE, Los Alamos, 2004. [13] T. Egami, S.J.L. Billinge, Underneath the Bragg-Peaks: Structural Analysis of Complex Materials, Elsevier Science B.V., Amsterdam, 2003. [14] C.L. Farrow, P. Juhas, J.W. Liu, D. Bryndin, E.S. Bozin, J. Bloch, Th. Proffen, S.J.L. Billinge, J. Phys. Condens. Matter 19 (2007) 335219. [15] P. Hohenberg, W. Kohn, Phys. Rev. 136 (1964) B864. [16] W. Kohn, L.J. Sham, Phys. Rev. 140A (1965) 1133. [17] J.P. Perdew, Y. Wang, Phys. Rev. B 45 (1992) 13244. [18] C. Colinet, Intermetallics 11 (2003) 1095–1102. [19] P.E. Blochl, Phys. Rev. B 50 (1994) 17953. [20] G. Kresse, J. Furthm ¨ uller, Phys. Rev. B 54 (1996) 11169; G. Kresse, J. Furthm¨ uller, Comp. Mater. Sci. 6 (1996) 1; G. Kresse, D. Joubert, Phys. Rev. B 59 (1999) 1758. [21] MedeA, Materials Design, Inc., /http://www.MaterialsDesign.comS, 2009. [22] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865; J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 78 (E) (1997) 1396. [24] G. Voronoi, J. f ¨ ur die Reine und Angew. Math. 133 (1907) 97–178. [23] K. Yvon, W. Jeitschko, E. Parthe, J. Appl. Cryst. 10 (1977) 73–74. [25] A. Einstein, Ann. Phys. (Leipzig) 22 (1907) 180–800; A. Einstein, Ann. Phys. (Leipzig) 34 (1911) 170 590. [26] P. Debye, Ann. Phys. (Leipzig) 33 (1910) 442–489. [27] T.H.K. Barron, M.L. Klein (1962) reedited in Phys. Rev. 127 (1997) 77. [28] F. Sayetat, P. Fertey, M. Kessler, J. Appl. Cryst. 31 (1998) 121–127. [29] R. Krishnan, R. Srinivasan, S. Devanarayan, Thermal Expansion of Crystals, Pergamon Press, London, 1979. [30] Web Elements periodic table, /http://www.webelements.com/periodicity/ coeff_thermal_expansion/S, 2009. [31] Cambridge Crystallographic Data Centre, /http://www.ccdc.cam.ac.uk/pro ducts/csd/radii/S, 2008. [32] S. Zhou, Y. Wang, F.G. Shi, F. Sommer, L.-Q. Chen, Z.-K. Liu, R.E. Napolitano, JPEDAV 28 (2007) 158–166. [33] H.L. Lukas, S.G. Fries, B. Sundman, Computational Thermodynamics The Calphad Method, Cambridge, 2007. [34] R.C. King, O.J. Kleppa, Acta Metall. 12 (1964) 87–97. M.H. Braga et al. / Journal of Solid State Chemistry 183 (2010) 10–19 19 134 Study of the Cu–Li–Mg–H system by thermal analysis M. H. Braga •J. A. Ferreira •M. J. Wolverton MEDICTA2011 Conference Special Chapter ÓAkade ´miai Kiado ´, Budapest, Hungary 2011 Abstract Finite fossil-fuel supplies, nuclear waste and global warming linked to CO 2 emissions have made the development of alternative/‘green’ methods of energy production, conversion and storage popular topics in today’s energy-conscious society. These crucial environmental issues, together with the rapid advance and eagerness from the electric automotive industry have combined to make the development of radically improved energy storage systems a worldwide imperative. CuMg 2 has an orthorhombic crystal structure and does not form a hydride: it reacts reversibly with hydrogen to produce Cu 2 Mg and MgH 2 . However, CuLi x Mg 2-x (x=0.08) has a hexagonal crystal structure, just like NiMg 2 , a compound known for its hydrogen storage properties. NiMg 2 absorbsupto3.6wt%of H. Our studies showed that not only CuLi x Mg 2-x absorbs a considerable amount of hydrogen, but also starts releasing it at a temperature in the range of 40–130 °C. In order to determine the properties of the hydrogenated CuLi x Mg 2-x , absorption– desorption, Differential scanning calorimeter and thermogravimetric experiments were performed. Neutron spectra were collected to elucidate the behavior of hydrogen in the Li-doped CuMg 2 intermetallic. Using DFT calculations we were able to determine the best value for x in CuLi x Mg 2-x and compare different possible structures for the CuLi x Mg 2-x hydride. Keywords DSC/TG INS Mg based alloys  Energy storage Hydrides Introduction Efficient hydrogen storage remains a major technological obstacle toward the development of a hydrogen-based energy economy. Electric and hybrid vehicles use metal hydrides as the negative electrode (instead of cadmium) in their batteries [1,2]. In fact, metal hydrides are widely used in consumer electronics rechargeable batteries. In Ni–MH, M is an intermetallic compound. The most common one has the AB 5 type, where A is a rare-earth mixture (e.g., La, Ce, Nd, Pr) and B is Ni, Co, Mn, and/or Al. Note that LaNi 5 H 6 , which is commonly used in batteries, stores only a maximum of 1.4 wt% of H. Another disadvantage of these materials for hydrogen storage is that lanthanides tend to be expensive and heavy. We are currently investigating the Cu–Li–Mg–H system. The lighter and cheaper metals and our recent discovery that hydrogen can be reversibly stored in these compounds make them a very attractive alternative to lanthanide-based systems. CuMg 2 has an orthorhombic crystal structure and does not form a hydride: it reacts reversibly with hydrogen to produce Cu 2 Mg and MgH 2 [3]. However, CuLi x Mg 2-x (x=0.08) has a hexagonal crystal structure [4] (ICSD database [5]), just like NiMg 2 , a compound known for its hydrogen storage properties. NiMg 2 absorbs up to 3.6 wt% H (of the hydride’s weight) [6]. In spite of the fact that the percentage of hydrogen absorbed by NiMg 2 is enough to propitiate practical applications, the temperature at which the alloy desorbs hydrogen (282 °C (555 K) at 1 bar) is much too high for practical applications. M. H. Braga (&) CEMUC, Engineering Physics Department, Engineering Faculty, Porto University, R. Dr. Roberto Frias s/n, 4200-465 Porto, Portugal e-mail: [email protected] M. H. Braga M. J. Wolverton Los Alamos National Laboratory (LANSCE), Los Alamos, NM 87545, USA J. A. Ferreira Laboratorio Nacional de Energia e Geologia, Rua da Amieira, Apartado 1089, 466-901 S. Mamede de Infesta, Portugal 123 J Therm Anal Calorim (2012) 108:733–739 DOI 10.1007/s10973-011-2126-0 135 A comparison between the phase diagrams of the systems Cu–Mg and Ni–Mg shows that these binary systems form compounds with similar stoichiometry. NiMg 2 is formed by peritectic reaction of the elements at 759 °C (1,032 K) and CuMg 2 by congruent melting at 568 °C (841 K). The presence of Li lowers even further the melting point of CuMg 2 [7]. Since the enthalpy of formation of the hydride is related to that of the primary alloy [8], it was hypothesized that CuLi x Mg 2-x might also be a hydrogen storage material similar to NiMg 2 [9]. Presumably, its advantage would be that it would release hydrogen at a lower temperature (possibly close to room temperature) [10]. Preliminary studies at the Los Alamos Neutron Scattering Center (LANSCE) showed that hydrogen unsaturated samples could desorb up to 4.4–5.3 wt% of hydrogen. Experiments furthermore shown that samples containing CuLi x Mg 2-x will start desorbing hydrogen at a temperature from 40–130 °C where applications are easier to develop. Hence it should be possible to use this alloy with fuel cells or in batteries and hydrogen storage devices. Sample preparation The Cu–Li–Mg samples were prepared from the pure elements with a target composition of CuLi 0.1 Mg 1.9 . They were prepared by mixing stoichiometric amounts of Cu (electrolytic, 99.99% purity, 325 mesh), Mg (99.8% purity, 200 mesh, Alfa Aesar), and granules (approx. 2 929 3 mm) of Li (99% purity, Alfa Aesar). Because of the large vapour pressure of Mg, even below its melting point, the reagents were sealed in a stainless steel crucible in a dry box (He atmosphere). This eliminated possible reagent loss. The samples were heated in a tube furnace with a stirring device to ensure proper mixing of the heterogeneous starting mixture and complete dispersion of Li in the sample. Different reaction temperatures and times were used (from 450 °C for 24 h to 1,200 °C for 1–2 h). Regardless of reaction conditions, the samples invariably contained Cu 2 Mg, CuMg 2 , or both. Nonetheless, we obtained final products containing up to 82.5 at% (77.5 wt%) of CuLi x Mg 2-x . Since the structure of Cu 2 Mg and CuMg 2 is known as well as their hydrogen storage behaviour; this complication translated merely in the refinement of additional phases in the neutron/X-ray powder diffraction patterns. Samples were firstly characterized by means of X-ray diffraction (XRD) using a Rigaku Ultima III powder diffractometer, and their composition was roughly determined by means of the Match software, [11] which uses the ‘‘Reference Intensity Ratio method’’ RiR method) [12]to obtain phase fractions. Patterns were collected with CuK- alpha radiation with 2htypically from 15°to 70°with steps of 0.02°and a counting time of 10 s per bin. Hydrogen absorption experiments A sample with approx. 74 at% of CuLi 0.08 Mg 2-0.08 ,20at% of CuMg 2 and 6 at% of Cu 2 Mg was grinded to obtain a powder with particle size of the order of 37 lm. The sample was then studied in a HPVA high pressure absorption analyser [13]. The latter gas adsorption analysers are designed to obtain high pressure adsorption isotherms of gases, such as hydrogen, using the static volumetric method. Two additional samples with –79 at% of CuLi 0.08 Mg 1.92 and 21 at% of Cu 2 Mg and –57 at% of CuLi 0.08 Mg 1.92 ,34at%of CuMg 2 and 9 at% of Cu 2 Mg– were measured too, to determine H 2 equilibrium pressure at different temperatures during the absorption process (Fig. 1). The plot of the natural logarithm of the equilibrium pressure of hydrogen versus 1,000/T in which T is the absolute temperature in Kelvin allows us to understand that to equal equilibrium pressures will correspond very different temperatures when Cu–Li–Mg–H samples (with high initial content of CuLi 0.08 Mg 1.92 ) are compared with MgH 2 , NiMg 2 H 4 , or even with Cu 2 Mg ?3MgH 2 [13]. Additionally, when the equilibrium pressure, at 200 °C, for a sample containing CuLi 0.08 Mg 1.92 –H is compared with 0.0 0.5 1.0 1.5 2.0 0 5 10 15 20 25 30 (our work) NiMg 2 T = 298 ° C desorption P eq /bar H/(Cu + Li + Mg) T = 295 ° C desorption T = 200 °C absorption 100 at% CuLi0.083Mg1.917 CuMg 2 (actually Mg) (Reilly and Wiswall, 1967 and 1968) 2.00 2.04 2.08 2.12 2.16 2.20 2.24 2.5 3.0 3.5 4.0 Δ H = –70.3 kJ/mol (per mol H 2 ); Δ S = –0.176 kJ/mol; T dec = 127 ° C Δ H = –68.5 kJ/mol (per mol H 2 ); Δ S = –0.170 kJ/mol; T dec = 129 ° C 79 at% CuLi 0.08 Mg 1.92 + 21 at% Cu 2 Mg 57 at% CuLi 0.08 Mg 1.92 + 9 at% Cu 2 Mg + 34 at% CuMg 2 ln(P/bar) 1000/(T/K) Fig. 1 left Absorption curve for a sample containing CuLi 0.08 Mg 1.92 (considering that Cu 2 Mg does not absorb hydrogen at 200 °C and at the referred pressures). right Natural logarithm of the equilibrium pressure versus 1,000/T for two samples with initial compositions (prior to hydrogen absorption) referred in the plot’s legend 734 M. H. Braga et al. 123 136 the equilibrium pressure of a sample containing CuMg 2 ,at 295 °C, the difference between plateau pressures can be higher than 21 bar [13] (Fig. 1). The latter means that CuLi 0.08 Mg 1.92 –H will release hydrogen at a much lower temperature. Calculations of the decomposition temperature can be done, using: GMH GH2¼DG¼0,G0 MH G0 H2RTln Peq P0  ¼0 ð1Þ DG0¼RTln Peq P0  ¼DH0TDS0ð2Þ ln Peq P0  ¼DH0 R1 TDS0 Rð3Þ For Peq ¼P0)Tdec ¼DH0 DS0per mol of H2 ðÞð4Þ in which Gis the Gibbs energy, MH (metal-hydride), Rthe ideal gas constant, P eq and P 0 is the equilibrium and atmospheric pressure, respectively, H 0 the enthalpy and S 0 the entropy when P eq =P 0 . The previous calculations for the Cu–Li–Mg–H system lead to decomposition temperatures of 400 K (127 °C) for a sample containing – 79 at% of CuLi 0.08 Mg 1.92 and 21 at% of Cu 2 Mg– and of 402 K (129 °C) for a sample containing – 57 at% of CuLi 0.08 Mg 1.92 , 34 at% of CuMg 2 and 9 at% of Cu 2 Mg. Cu 2 Mg does not absorb hydrogen at the temperatures shown in Fig. 1. Hydrides stability depends mostly on the enthalpy term (DH 0 )asDS 0 is usually around -130 J (mol H 2 ) -1 K -1 , which roughly corresponds to the H 2 molecule losing its translational degrees of freedom upon transformation from the gas phase into the solid state of the hydride. Nevertheless, in this system, DS 0 is higher than usually but still into the boundaries of the entropy of formation of the hydrides studied in the literature (DS(FeTi)&-0.104 kJ/mol H 2 [14]; DS(LaNi 5 )&-0.105 kJ/mol H 2 [14]; DS(VH 2 )& 0.141 kJ/mol H 2 [14]; DS(PdH 0.7 )&-0.098 kJ/mol H 2 [14]; DS(CuLiMg–H)&-0.170 kJ/mol H 2 ;DS(CuMg 2 )& 0.138 kJ/mol H 2 [15]; DS(NiMg 2 )&-0.123 kJ/mol H 2 [14]; DS(MgH 2 )&-0.135 kJ/mol H 2 [16]; DS(NaH)& -0.164 kJ/mol H 2 [16]; DS(KH)&-0.169 kJ/mol H 2 [16]; DS(UH 3 )&-0.179 kJ/mol H 2 [14]; DS(TiH 2 )& -0.178 kJ/mol H 2 [14]; DS(LiH)&-0.126 kJ/mol H 2 [14]; DS(SrH 2 )&-0.141 kJ/mol H 2 [16]). DS 0 is mostly the responsible by the difference between the Cu 2 Mg ?MgH 2 and the CuLi 0.08 Mg 1.92 –H desorption temperatures. For all of the hydrides to be discussed, DH 0 and DS 0 are negative, i.e., the hydrogenation reaction is exothermic and the dehydrogenation reaction is endothermic. Inelastic neutron scattering Time-of-flight (TOF) inelastic incoherent neutron scattering was collected at low temperatures at LANSCE, in FDS. A sample with approximately –76 at% of CuLi 0.08 Mg 1.92 and 24 at% of Cu 2 Mg– was loaded with H 2 at 200 °C, at different pressures and loading times, before collecting a neutron vibrational spectrum (cycling treatments were performed over the same sample after each measurement. Between loadings, H 2 pressure was *1 bar and temperature dropped to room temperature). The sample was loaded: once for 1 h under 36 bar of H 2 ; twice for 6 h under 100 bar of H 2 and once for 21 h; once for 17 h and 25 min. and twice for 3 h (100 bar). All data were collected at 10 K. Figure 2shows the first Inelastic Neutron Scattering (INS) experiment in comparison with measurements of other samples of the same system – 33 at% of CuLi 0.08 Mg 1.92 , 35 at% of CuMg 2 and 32 at% of Cu 2 Mg- and – 42 at% of CuLi 0.08 Mg 1.92 ,32at%ofCuMg 2 and 26 at% of Cu 2 Mgafter being submitted to one hydrogenation cycle and of NiMg 2 H 4 . The latter hydride exhibits the same spectra whatever the hydrogenation treatment used or number of cycles. It seems that the lattice vibrations happen at similar wavenumbers possibly indicating a similar monoclinic structure comprising a [CuH 4 ] 3- anion complex. INS spectra show that cycling has a strong effect on structure and ion distribution. On the other hand, this effect on the structure can moreover be due to the loading time or to the waiting time (time the sample waited, at ambient pressure, before being measured). In other words, there 400 600 800 1000 1200 1400 1600 1800 2000 2 3 4 5 6 7 8 9 10 11 12 13 I/a.u. Wavenumber/cm –1 (33 at% CuLi xMg2-x + 35 at% CuMg 2 + 32 at% Cu2Mg)-H NiMg2H4 (76 at% CuLixMg2-x + 24 at% Cu2Mg)-H (42 at% CuLixMg2-x + 32 at% CuMg2 + 26 at% Cu2Mg)-H Fig. 2 INS results show similar spectra for different Cu–Li–Mg–H samples after one hydrogenation cycle. INS for NiMg 2 H 4 is similar to the others spectra shown and does not depend of the hydrogenation process Study of the Cu–Li–Mg–H 735 123 137 characterized by means of x-ray diffraction (XRD) using a Rigaku Ultima III powder diffractometer, and their composition was roughly determined by means of the Match software, [1] which uses the “Reference Intensity Ratio method” RiR) [2] to obtain phase fractions. Patterns were collected with CuKalpha typically from 2θ = 15 to 70º with steps of 0.02º and a counting time of 10 s per bin. Neutron scattering in deuteride and hydride samples. Time-of-flight (TOF) neutron diffraction data were collected at low and room temperatures on NPDF and HIPD neutron diffractometers at LANSCE. Neutron diffraction. Neutron diffraction was firstly performed in order to characterize the phase CuLixMg2-x. Rietveld refinements using GSAS [3] and Pair Distribution Function (PDF) refinements using PDFgui [4] were carried out to characterize the phases present in the samples. Results obtained show that Li will substitute Mg in (1/2,0,z) forming a hexagonal P6222 compound CuLi0.08Mg1.92 with lattice parameters, a = b = 5.250 (1) Å and c = 13.621(1) Å (at T = 300 K). A detailed study of the CuLi0.08Mg1.92, CuMg2 and Cu2Mg phases of the Cu-Li-Mg system can be found in [5]. Samples of the Cu-Li-Mg-D system were also studied in NPDF and HIPD Fig. 1. Samples were always deuterized at 200 ºC. The sample correspondent to Fig. 1 (left) was deuterized for 55h30min. under 69 bar (D2). Before being deuterized, it had around 38 at% of CuLi0.08Mg1.92, 45 at% of CuMg2 and 17 at% of Cu2Mg. The sample correspondent to Fig. 1 (right) suffered different hydrogen sorption/desorption cycles before being measured in HIPD; however before being cycled it had around 76 at% of CuLi0.08Mg1.92, and 24 at% of Cu2Mg. During cycling, the gas pressure was never higher than 100 bar H2/D2. When measured in HIPD, the sample correspondent to Fig. 1 (right) was not fully saturated (parent phases were still present). In the case of this last sample the presence of CuLi0.08Mg1.92D5 is clear although the compound contributes in a small percentage for the final weight of the sample. Inelastic Neutron Scattering. A sample with 76 at% of CuLi0.08Mg1.92 and 24 at% of Cu2Mg was loaded with H2 at 200 oC, at different pressures and with different loading times, before collecting a neutron vibrational spectrum (cycling treatments were performed over the same sample after each measurement. Between loadings, H2 pressure was ~ 1 bar and temperature dropped to room temperature). The sample was loaded: 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 I(a.u.) 2 θ θθ θ ( ° °° ° ) T = 60K; wRp = 0.05; RP= 0.04 composition of the sample (from Rietveld refinement): x(CuLi0.08Mg1.92) = 0.28 x(CuMg2) = 0.33 x(Cu2Mg) = 0.19 x(MgD2) = 0.20 NPDF 1.0 2.0 3.0 4.0 5.0 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 2 θ θθ θ ( ° °° ° ) I(a.u.) HIPD T = 5K; wRp = 0.04; RP = 0.02 Rietveld Refinement x(CuLi0.08Mg1.92) = 0.59 x(Cu2Mg) = 0.27 x(MgD2) = 0.11 x(CuLi0.08Mg1.92D5) = 0.03 Fig. 1(left) Sample containing around 28 at% of CuLi0.08Mg1.92, 33 at% of CuMg2, 19 at% of Cu2Mg and 20 at% of MgD2. This sample was not refined with CuLi0.08Mg1.92D5 since the presence of this phase was too small to make a difference in the refinement. (right) Sample containing around 59 at% of CuLi0.08Mg1.92, 27 at% of Cu2Mg, 11 at% of MgD2 and 3 at% of CuLiMg0.08Mg1.92D5. All phases had to be present to allow a reliable Rietveld refinement. 800 Advanced Materials Forum VI 144 - under 36 bar of H2 for 1 hour once; - twice under 100 bar of H2 for 6 hours and once for 21 h; - once for 17 h and 25 min. and twice for 3 hours (100 bar). All data were collected at 10 K. Fig. 2 shows the first and third experiment. Inelastic neutron scattering (INS) spectra show that cycling has a strong effect on structure and ion distribution. On the other hand, this effect on the structure can moreover be due to the loading time or to the waiting time (time the sample waited, at ambient pressure, before being measured). In other words there might be a phase transition during hydrogen loading / unloading. These results show that during the first loading the structure was monoclinic with the Cu-H, eventually forming [CuH4]3- and probably occupying C1 sites like [NiH4]4- in NiMg2H4 (monoclinic) [6], and that after 3 and 6 cycles the sample is likely to present a tetragonal structure similar to CoMg2H5 in which [CoH5]4- occupies square-based pyramidal C4v sites [7] or a tetragonal structure that can actually be MgH2 [8]. Differential scanning calorimetry (DSC) and Thermogravimetry (TG). Several samples were analyzed by DSC/TG using a Netzsch instrument. Samples were heated from room temperature to 450 °C, at 5 and 10 °C/min. Alumina crucibles and lids were used at all times under high-purity argon gas flowing at 27 ml/min. Samples in the form of powder were hydrogenised before measured. A literature search shows that MgH2 in form of powder releases hydrogen in DSC instruments at temperatures that vary from the minimum of 325 °C [9] to a maximum of 433 °C [10]. Although for MgH2 the equilibrium desorption temperature should be approximately 280 °C (553 K), in DSC the effect of kinetics counts as well, and therefore the lowest temperature found in the literature was 325 °C. Several DSC studies, that aimed to determine at which temperature MgH2 releases hydrogen, were reported in the literature [9-17]. Most of them focus on the effect of a catalyst on the desorption temperature of MgH2. One of the studies refers to nanoparticles of CuMg2 [17]. CuMg2 is hydrogenised to form MgH2 and Cu2Mg ( 2 2 2 2 332 MgHMgCuHCuMg +↔+ ). According to Shao et al., nanoparticles of MgH2 + Cu2Mg will start releasing H2 at 409°C when the sample is under 4MPa (40 bar) of H2 [17], while the equilibrium desorption temperature is 385 °C. The same study reports that Mg nanoparticles under the same previous conditions absorb hydrogen at 472°C, while the equilibrium desorption temperature is 442 °C [17]. 0 400 800 1200 1600 2000 0 2 4 6 I ntensity (arb. units) (1) 430 cm - 1 (2) 504 cm-1 (3) 616 cm-1 (4) 808 cm-1 75.6 at% CuLi0.08Mg1.92 (H) + 24.4 at% Cu 2 Mg (1) (2) (3) (4) 1 st loading 1h wavenumber (cm-1) 0 400 800 1200 1600 2000 0 2 4 6 I ntensity (arb. units) (1) 430 cm - 1 (2) 504 cm-1 (3) 616 cm-1 (4) 808 cm-1 75.6 at% CuLi0.08Mg1.92 (H) + 24.4 at% Cu 2 Mg (1) (2) (3) (4) 1 st loading 1h wavenumber (cm-1) Energy transfer (meV) (cm - 400 800 1200 1600 2000 0 1 2 3 871 cm-1 109 meV 6 cycles Intensity (arb. units) -1 621 cm-1 77.6 meV 2031 cm - 1 254 meV 437 cm - 1 54.6 meV 6 cycles 75.6 at% CuLi0.08Mg1.92 (H) + 24.4 at% Cu2Mg wavenumber (cm -1 ) Energy transfer (meV) (cm - 400 800 1200 1600 2000 0 1 2 3 871 cm-1 109 meV 6 cycles Intensity (arb. units) -1 621 cm-1 77.6 meV 2031 cm - 1 254 meV 437 cm - 1 54.6 meV 6 cycles 75.6 at% CuLi0.08Mg1.92 (H) + 24.4 at% Cu2Mg wavenumber (cm -1 ) Fig. 2 INS results for two different structures of the CuLi0.08Mg1.92Hx hydride. Results show different structures depending on the number of cycles / hydrogenation stage. In addition, the amount of desorbed hydrogen is not only related with the capacity of the hydride; it has moreover to do with kinetics, grain size, possible catalysts added, and also with the way powders are grinded [10]. Furthermore, if the samples are not saturated, the amounts of released hydrogen cannot be compared or controlled. Materials Science Forum Vols. 730-732 801 145 Results show that hydrogenised samples containing CuLi0.08Mg1.92, start releasing hydrogen at T < 55 °C / 328 K, and have a second release rate at around ~200 °C / 473 K and finally a third release rate at about ~280 °C/553 K. The first and second releases seem to be related with CuLi0.08Mg1.92-H and the third could be due to MgH2 desorption. If this assumption is correct, this means that CuLi0.08Mg1.92-H will disproportionate into other hydrides. Furthermore, if only CuMg2 had contributed to the formation of MgH2 (mass losses and peaks at 282 °C in Fig. 3), the maximum loss in weight percentage of the sample should be ~0.6 wt% since CuMg2 is not present with more than ~23 wt% (the theoretical amount of hydrogen that can be released by Cu2Mg + 3MgH2 is 2.6 wt% in a sample containing 100 % of the mixture). Likewise, if MgH2 desorbs hydrogen at 282 °C, then the catalytic effect of Cu2Mg is likely present and in addition a catalytic effect of CuLi0.08Mg1.92-H, since MgH2 desorption temperature is 43 °C lower than the minimum temperature found in the literature for DSC experiments [9-17] of samples containing 100% of MgH2. 0 100 200 300 400 500 0.0 0.7 1.4 98 99 100 101 416 °C 282 °C 1.3 wt% 1.2 wt% Wt (%) DSC (µ µ µ µV/mg) 10 °C /min. T ( ° °° ° C) 46 °C 208 °C endo 0 100 200 300 400 500 0.0 0.5 1.0 96 98 100 191 ° C 5 ° C /min. 2.5 wt% 426 ° C 53 ° C 287 ° C 1.4 wt% Wt (%) DSC ( µ µ µ µ V/mg) T ( ° °° ° C) endo 0 100 200 300 400 500 0.0 0.7 1.4 98 99 100 101 416 ° C 282 ° C 1.3 wt% 1.2 wt% Wt (%) DSC (µ µ µ µV/mg) 10 ° C /min. T (° °° °C) 46 ° C 208 ° C endo 0 100 200 300 400 500 0.0 0.5 1.0 96 98 100 191 ° C 5 ° C /min. 2.5 wt% 426 ° C 53 ° C 287 ° C 1.4 wt% Wt (%) DSC ( µ µ µ µ V/mg) T ( ° °° ° C) endo Fig. 3 DSC/TG curve of a sample containing 60.0 wt% of CuLi0.08Mg1.92, 27.6 wt% of CuMg2 and 12.4 wt% of Cu2Mg (62.4 at% of CuLi0.08Mg1.92, 28.2 at% of CuMg2 and 9.4 at% of Cu2Mg) and of a sample containing 68.6 wt% of CuLi0.08Mg1.92, 18.0 wt% of CuMg2 and 13.4 wt% of Cu2Mg (71.4 at% of CuLi0.08Mg1.92, 18.4 at% of CuMg2 and 10.2 at% of Cu2Mg) that were hydrogenated at 200 °C beforehand. Samples do not seem to be saturated with hydrogen since they still present a peak corresponding to the parent’s phase melting point (426 °C and 416 °C). Nonetheless, XRD patterns of the sample corresponding to the left curve showed saturation. On the other hand, the catalytic effects of CuLi0.08Mg1.92(-H) over CuMg2 were also present during hydrogen absorption. Hydrogen absorption experiments with nanostructured CuMg2 samples, without prior activation process as in the case of this study, reveal that this compound will start absorbing hydrogen at ~ 250 °C (5) [17]. All our samples absorbed hydrogen at 200 °C. Using the total amount of hydrogen released by the sample during heating in Fig. 3 (left), the stoichiometry of CuLi0.08Mg1.92-H can be determined. Nonetheless, the absorption of hydrogen by CuMg2 has to be taken into account; conversely, Cu2Mg will not absorb hydrogen at 200 °C. We have obtained a value of ~ 5.3 wt% that corresponds to CuLi0.08Mg1.92H6 (WtH% = 5.2 %). Ab Initio Calculations Parent phase, CuLi0.08Mg1.92. Density Functional Theory (DFT) calculations with Projector Augmented Wave (PAW) pseudopotentials [18], as implemented in the Vienna Ab Initio Simulation Package code (VASP) [19], were performed. A plane wave cutoff of 355.18 eV, and k-spacings of 0.230 x 0.230 x 0.230 Å-1 were used. Calculations were done in real space and were performed with P1 space group supercells containing 144 atoms (48 atoms of Cu, 96 - n of Mg, and n = 0 to 12 of 802 Advanced Materials Forum VI 146 Li). The supercells contained as many atoms as possible to allow better approximations with the real Li concentrations (but such that the time spent on calculations were not completely impractical). Since CuLixMg2-x is a disordered structure, it was obtained by randomly substituting Mg by Li in several 6f Wyckoff positions (1/2,0,z) of the initial hexagonal structure that conducted to the P1 supercells. The Generalized Gradient Approximation (GGA), and the Perdew–Burke– Ernzerhof (PBE) functional [20] were used, and no magnetic moments were included in the model. We have concluded that the stoichiometry that minimizes the energy of formation is CuLi0.08Mg1.92, which is in agreement with the experimental results [5]. Hydride phase, CuLi0.08Mg1.92Hx. Using DFT and the same conditions described for the parent phase (with one exception: we did not initially use supercells), several possible structures of CuMg2Hx were optimized. We did not replace Mg with Li because that would mean building a supercell and spend a large computational time. The goal was to obtain a starting point for further studies. 4.0 4.5 5.0 5.5 6.0 -40 -20 0 20 40 60 80 100 120 C 2/c, 5H P4/n m m Ef(kJ/mol of H2) n in CuMg2Hn P4/m m m C 2/c, 4H C 2/c, 6H Therefore, since most of the hydrides are cubic structures or structures that can be considered as distortions of a cubic structure [6], we have started calculations with tetragonal and monoclinic structures, similar to those of the hydrides formed by the nearest neighbours of Cu and Mg in the periodic table: NiMg2H4 and CoMg2H5 (e.g. CuMg2H4 was optimized as monoclinic C 2/c; CuMg2H5 tetragonal P 4/nmm, respectively). In Fig. 4 it can be observed that the most stable configuration corresponds to CuMg2H5 with a monoclinic C2/c structure. Using this structure, we have built a supercell and replaced Mg atoms by Li atoms in three different sites of the initial configuration. The new optimized structure is even more stable (the difference is: ∆Ef = - 4 kJ/mol of H2). Knowing what the most stable structure of the hydride is, is necessary, but not sufficient to make a statement about the reaction occurring upon hydrogen absorption. More information is needed since the compound may disproportionate into other compounds/hydrides like in the case of CuMg2. Nonetheless, when we compare these results with neutron diffraction and inelastic neutron spectroscopy, we conclude that it is highly possible that the first structure formed upon hydrogenation is a monoclinic C2/c. Summary After analysing the data obtained by neutron diffraction and neutron spectroscopy, DSC and first principles calculations we conclude that it is very likely that CuLi0.08Mg1.92 will react with hydrogen to form CuLi0.08Mg1.92H5. Nonetheless, according to DSC’s results, it is also possible that the stoichiometry of the product is CuLi0.08Mg1.92H6. Fig. 4 Energy of formation of the hydride CuMg2Hn as a function of n, for different types of crystal structures. Materials Science Forum Vols. 730-732 803 147 The monoclinic C2/c structure for the CuLi0.08Mg1.92 hydride is in agreement with the results obtained both using neutron scattering techniques and first principles calculations. Additionally, it can be determined the presence of MgH2, formed after hydrogen uptake, even for those samples that did not initially contain CuMg2. In this work it is also clear the roll of the Cu-Li- Mg-(H) as a catalyst for both CuMg2 and MgH2 during absorption and desorption. Acknowledgements The authors would like to acknowledge Portuguese Science Foundation, FCT, for the project (PTDC/CTM/099461/2008 and and FCOMP-01-0124-FEDER-009369). This work has benefited from the use of neutron scattering instruments NPDF, HIPD and FDS at the Lujan Center at Los Alamos Neutron Science Center, funded by DOE Office of Basic Energy Sciences, USA. This work has benefited from the use of 11-beamline at the Advanced Photon Source at Argonne National Laboratory, USA. References [1] Match, http://www.crystalimpact.com/, 2009. [2] P.M. de Wolff, J.W. Visser, Absolute Intensities. Report 641.109. Technisch Physische Dienst, Delft, Netherlands. Reprinted Powder Diffract 3 (1988) 202. [3] A.C. Larson, R.B. von Dreele, GSAS Generalized Structure Analysis System, LANSCE, Los Alamos, 2004. [4] C.L. Farrow, P. Juhas, J.W. Liu, D. Bryndin, E.S. Bozin, J. Bloch, Th. Proffen, S.J.L. Billinge J. Phys.: Cond.. Matter. 19 (2007) 335219. [5] M.H. Braga, J.J.A. Ferreira, J. Siewenie, Th. Proffen, S.C. Vogel, L.L. Daemen, J. of Sol. State Chem. 183(1) (2010) 10. [6] K. Yvon, G. Renaudin Hydrides: Solid State Transition Metal Complexes. Encyclopedia of Inorganic Chemistry, 2006. [7] S.F. Parker, U.A. Jayasooriya, J. C. Sprunt, M. Bortz, K. Yvon J. Chem. Soc. Faraday Trans. 94(17) (1998) 2595. [8] H.G. Schimmel, M.R. Johnson, G.J. Kearley, A.J. Ramirez-Cuesta, J. Huot, F.M. Mulder, Mat. Sci. Eng. B 108 (2004) 38. [9] A. Patah, A. Takasaki, J.S. Szmyd, Mater. Res. Soc. Symp. Proc. 1148-PP03-38, 2009. [10] N.W.B. Balasooriya, Ch. Poinsignon, IEEE Inter. Nanoelect. Conf. 2008, 894. [11] C. Z. Wu, P. Wang, X. Yao, C. Liu, D.M. Chen, G.Q. Lu, H.M. Cheng, J. Alloys Comp. 420 (2006) 278. [12] A. Montone, J. Grbovic, Lj. Stamenkovic, A.L. Fiorini, L. Pasquini, E. Bonetti, M.V. Antisari Mat. Sci. For. 518 (2006) 79. [13] M.A. Lillo-Rodenas, K.F. Aguey-Zinsou, D. Cazorla-Amoros, A. Linares-Solano, Z.X. Guo, J. Phys. Chem. C 112 (2008) 5984. [14] P. Tessier, E. Akiba, J. Alloys Comp. 302 (2000) 215. [15] S. Milovanovic, L. Matovic, M. Drvendzija, J.G. Novakovic, J. Micros., 232 (3) (2008) 522. [16] S. Deledda, A. Borissova, C. Poinsignon, W.J. Botta, M. Dornheim, T. Klassen, J. Alloys Comp. 404-406 (2005) 409. [17] H. Shao, Y. Wang, H. Xu, X. Li, J. Sol. State Chem. 178 (2005) 2211. [18] P.E. Blochl, Phys. Rev. B 50 (1994) 17953. [19] G. Kresse, J. Furthmüller, Phys. Rev. B 54 (1996) 11169; Comp. Mat. Sci. 6 (1996) 1; G. Kresse, D. Joubert, Phys. Rev. B 59 (1999) 1758. [20] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865; 78 (E) (1997) 1396. 804 Advanced Materials Forum VI 148                Chapter 4- Bi-Sn-Zn System 149 150 Chapter 4. Bi-Sn-Zn SYSTEM 4.1. Introduction Lead and lead-containing compounds are considered toxic substances due to their detrimental effect to the well-being of humans and the environment. Suitable development policy has been implemented by many countries around the world and, in order to protect the environment, the restriction of lead used in industry has been strongly promoted. In this context, since July 1st, 2006, both the European Union (RoHS — Restriction of Certain Hazardous Substances and WEEE — Waste Electrical and Electronic Equipment Legislations) and the U.S. Environmental Protection Agency have banned the leadcontaining electronic products (Islam et al., 2006; Yu et al., 2000). In order to replace the traditional Sn–Pb eutectic solder alloy by lead-free alternatives great efforts were developed. The Sn–Zn eutectic alloy is the nontoxic Pb-free solder alloy alternative having a melting temperature which is closest to that of the eutectic Sn–Pb alloy (471 K and 456 K, respectively). However, new solder alloys must fulfill a number of other requirements in both economic and physical/chemical points of view. In this context, the melting temperature should be in the same range as that of the traditional Sn– Pb eutectic alloy, strength and integrity should also be similar or superior, and manufacturing costs must be competitive (Garcia et al., 2010). Poor oxidation resistance and embrittlement behavior are the major problems with the Sn-Zn alloys. In order to overcome these drawbacks, and further enhance the properties of Sn–Zn leadfree solder alloys, a small amount of alloying elements (rare earths, Bi, Ag, Al, Ga, In, Cr, Cu, Sb, Ni, Ge) added into Sn–Zn alloys were selected by many researchers. For example, a small amount of Al, P, Bi, Ga can improve the high-temperature oxidation resistance of Sn– Zn solders remarkably as well as Cr (Zhang et al., 2010). We have studied the Bi-Sn-Zn as well as the Sn-Zn and Bi-Zn systems, on ambit of the COST 531 and COST MP0602, EU actions, on lead free solders. Our studies contemplated both 151 experimental (using DSC- Differential Scanning Calorimetry , HT-XRD - High Temperatures X-ray Diffraction, RT-XRD - Room Temperatures X-ray Diffraction, SEM – Scanning Electron Microscopy ), and assessments (using Thermo-Calc software). The need for new, improved solder alloys and a better understanding of reactions during the soldering process grows steadily as the need for smaller and more reliable electronic products increases. Information obtained from phase equilibria data and thermodynamic calculations has proven to be an important tool in the design and understanding of new lead-free solder alloys. A wide range of candidate alloys can be rapidly evaluated for proper freezing ranges, susceptibility to contamination effects, and reactions with substrate materials before the expensive process of preparing and testing candidate alloys is initiated (Kattner, 2002). Therefore, the latter was the objective of the above mentioned COST actions that joined researchers from many European countries around the same goal. 152 4.2. References Garcia,L., Osório, W., Peixoto, L., & Garcia, A., (2010). Mechanical properties of Sn–Zn leadfree solder alloys based on the microstructure array, Materials Characterization, Vol. 61, No. 2, pp. 212–220, ISSN 1044-5803 Kattner, U., (2002). Phase diagrams for lead-free solder alloys, JOM, Vol. 54, No. 12, pp. 45- 51, ISSN 1047-4838 Islam, R., Chan Y., Jillek W., & Islam S., (2006). Comparative study of wetting behavior and mechanical properties (microhardness) of Sn–Zn and Sn–Pb solders, Microelectronics Journal, Vol. 37, No. 8, pp. 705–713, ISSN 0026-2692 Yu, S., Lin, H., Hon, M., &Wang, M-C., (2000). Effects of process parameters on the soldering behavior of the eutectic Sn–Zn solder on Cu substrate. Journal of Materials Science: Materials in Electronics, Vol. 11, No. 6, pp. 461-471, ISSN 1573-482X Zhang, L., Xue, S., Gao, L., Sheng, Z., Ye, H., Xiao, Z., Zeng, G., Chen, Y., Yu, S., (2010). Development of Sn–Zn lead-free solders bearing alloying elements, Journal of Materials Science: Materials in Electronics, Vol. 21, No. 1, pp. 1-15, ISSN 1573-482X. 153