Journal of The Electrochemical Society, 152 共6兲 H84-H93 共2005兲 H84 0013-4651/2005/152共6兲/H84/10/$7.00 © The Electrochemical Society, Inc. Theoretical Investigation of Structure and Stability of Reidinger Defects in Barium Magnesium Aluminate K. C. Mishra*,z OSRAM-SYLVANIA Development Incorporated, Central Research, Beverly, Massachusetts 01915, USA An atomistic simulation method has been used to study the equilibrated geometry, defect structure, and phase mixing of barium magnesium aluminate 共BAM兲 and related compounds. The calculated lattice energies are used to study the stability of defect structures in the -alumina lattice, particularly in the presence of excess alumina in the starting materials. It is shown that excess alumina in the starting compounds could lead to the formation of Reidinger defects in BAM. The calculated structural modification of the lattice due to a Reidinger defect is in excellent agreement with the experimental results. These defects are responsible for generation of F+ and F-type color centers by 共vacuum兲 ultraviolet radiation and bombardment of energetic particles from the discharge. BAM could be made resistant to the formation of color centers by adhering to a strictly stoichiometric formulation of this phosphor. © 2005 The Electrochemical Society. 关DOI: 10.1149/1.1914759兴 All rights reserved. Manuscript submitted October 19, 2004; revised manuscript received December 17, 2004. Available electronically May 5, 2005. Barium magnesium aluminate 共BaMgAl10O17兲 activated by divalent europium ions is a blue-emitting phosphor. This phosphor is popularly known as BAM in lighting applications. It is used as a blue-emitting phosphor in triband phosphor coatings in low-pressure Hg-discharge lamps. Unlike the other components of triband phosphors, namely, the red-emitting yttrium oxide phosphor activated by trivalent europium, and the green-emitting lanthanum phosphate phosphor activated by terbium, BAM is relatively unstable during high-temperature processing 共⬃400°C兲 of certain fluorescent lamps in an oxidizing environment.1,2 Similar processing conditions are also encountered in plasma display panel applications with similar results. BAM also tends to degrade faster during the lamp life under high loading conditions,a leading to “unacceptable” color shifts during the lamp life. The last few years have seen intense activities in the luminescent community toward developing an understanding of the nature of the thermal degradation process. It is believed that the thermal degradation of BAM results from the oxidation of europium ions.3 However, similar efforts to understand the nature of the degradation process during the lamp life are still lacking. This paper presents the results of a theoretical investigation of the structure and origin of defects that could contribute to the generation of color centers during the lamp life. One of the reasons for phosphor degradation in the lamp environment is the generation and growth of color centers in phosphors due to lamp operation. Inside the fluorescent lamps, the phosphors exist in a very harsh environment of an arc discharge. During the lamp operation, the surfaces of microcrystalline phosphor particles are continuously subjected to bombardment of energetic electrons, ions, and neutral atoms and irradiation by vacuum ultraviolet 共VUV兲 photons. The color centers are induced by these discharge-related processes. The lamp phosphors are usually made from complex inorganic hosts. The color center literature is sparse on the investigation of such systems. The haloapatite phosphors are the only class of lamp phosphors that have been extensively studied for color centers. For a brief review, see Ref. 4. Similar studies for the triband phosphors are lacking because of their robust performance in conventional fluorescent lamps. The instability of BAM was noticed only when this phosphor was applied to phosphor coatings in the highly loaded, * Electrochemical Society Active Member. z a E-mail: [email protected] The phenomenon of degradation of BAM during lamp life is well known in the lamp community. Many lamp-related causes have been identified. For example, see C. R. Ronda, V. W. Weiler, A. Johnene, J. A. F. Peak, and W. M. P. Van Kemenade, U.S. Pat. 5, 811, 154 共1998兲 for a discussion on the role of mercury adsorption in the degradation process. M. Raukas has shown that the nature of degradation of BAM during lamp life can be simulated by irradiating the phosphor with 193 nm laser radiation. This observation calls for a proper understanding of the role of color centers in the degradation process. electrodeless, fluorescent lamps and in VUV lamps, e.g., Xe*2 discharge. An understanding of the color centers in a crystalline solid requires reliable information on the crystal structure and the distribution of occupied Wyckoff sites of the host lattice. BAM crystallizes in the -alumina structure.5 The distribution of various ions in this structure is discussed later in detail. In the following discussions, arabic numerals within parentheses indicate the site occupancy of an ion in conformity with the X-ray diffraction 共XRD兲 studies. Fortunately, the color centers in other hexaaluminates with similar -alumina structures have been extensively studied.6,7 Based on their optical absorption characteristics, one could identify three specific color centers of interest from the lamp maintenance perspective: two electron centers F and F+, and a hole center V. The first two correspond to electrons trapped at an oxygen vacancy, and the third, to a hole trapped at an oxygen site adjacent to an Mg2+ ion substituting for an Al3+ ion. Both the electron centers are UV-absorbing color centers. The absorption bands associated with the F and F+ centers peak at 4.20 and 4.76 eV, respectively. They could compete with the divalent europium ions for 254 nm photons from the Hg discharge. The V center absorbs near 2.60 eV and could contribute to the loss of visible radiation. The electron centers are known to originate from the Reidinger defects.6,7 The Reidinger defects are intrinsic to barium hexaaluminate in phase I 共Ba.75Al11O17.25, BAL兲,8 but in other hexaaluminates, they appear as isolated, extrinsic defects.b The primary constituent of a Reidinger defect is an oxygen ion, OR, trapped at a 6h site in the intermediate plane with ideal internal coordinates being 共5/6,2/3,1/4兲. This site is referred to as an mO site and is indicated by a shaded sphere in Fig. 1. Subsequent generation of vacancies at these sites is responsible for the creation of electron centers. The Reidinger defect leads to a complex rearrangement of the local structure surrounding the interstitial OR atom. In Fig. 1, the local distortions induced by a Reidinger defect in BAL near the intermediate plane in the adjacent spinel blocks have been shown schematically in contrast to a regular half-cell. In the defect halfcell, the barium ion is missing and an oxygen ion occupies an mO site. The presence of OR in the intermediate plane creates two Frenkel defects involving Al 共1兲 ions in the spinel block 共Fig. 1兲. These aluminum ions are located in the spinel blocks, vertically above and below the OR ions along a direction parallel to the hexagonal c axis. They relax toward the interstitial OR ion through the layer of O 共2兲 and O 共4兲 ions to almost the level of Al 共3兲 ions. The spin resonance b +1 The Reidinger defects are known to exist in M1+x Al11O17+x/2 in the -alumina structure, where M+1 represents monovalent ions like Na+,K+, and Li+. In these systems, there are no cationic vacancies in the intermediate plane as in phase I of Ba -aluminate. Downloaded on 2016-05-12 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract). Journal of The Electrochemical Society, 152 共6兲 H84-H93 共2005兲 Figure 1. Schematic representation of two kinds of half-unit cells in presence of Reidinger defects. The large empty sphere, the solid large sphere, and the shaded large sphere represent oxygen ions at the 2c site, barium ions at the 2d 共BR兲 site, and oxygen ion 共OR兲 associated with the Reidinger defect at the interstitial 6h 共mO兲 site. The small spheres represent aluminum ions. The solid ones correspond to aluminum ions in tetrahedral coordination and the empty ones to ions in octahedral coordination. The mechanism of formation of the Reidinger defect with an interstitial oxygen ion is presented in the half-cell on the right. The displacement of an aluminum ion toward the intermediate plane is indicated by an arrow. studies of the F+ centers suggest that this distortion locks the Al 共1兲 ions to their new positions, even after a vacancy is created at the OR site.7,9-11 The Reidinger defects could possibly lead to the generation of several color centers in the following manner. First, a lattice defect is created at the OR site, possibly displacing the oxygen ion to an interstitial, metastable site. Then, the vacancy at the OR site captures one or two electrons, generating either an F+ or an F center. The oxygen ion at the interstitial site could then move to an adjacent mO site, inducing the formation of a new Reidinger defect. This is plausible because the mO sites are present in triplicate around a Beevers-Ross 共BR site兲, a 2d site normally occupied by a large cation in the -alumina lattice.5 Additionally, triple Reidinger defects exist in phase II of Ba-rich barium hexaaluminate.12 The F+ centers have been observed in nonstoichiometric -alumina compounds having Reidinger defects. These defects are formed for overall charge compensation of the lattice due to excess cations. If the charge compensation can be achieved without incorporating oxygen ions into the lattice, then the material has no such defects. Stoichiometric Na+-alumina, which contains no Reidinger defects, is known to be resistant to the generation of color centers.11 Nonstoichiometric Na+⬙-alumina, charge compensated by divalent ions at the aluminum sites, are not known to develop the color centers. Thus, one would expect that BAM with stoichiometric input would not be susceptible to generation of color centers. However, recent measurements by Raukas13 demonstrate their existence in BAM, even with stoichiometric input. To improve the performance of BAM, it is crucial to properly understand the factors that lead to formation of the Reidinger defects. Unlike the materials crystallizing in the magnetoplumbite structure, the anion close packing 共including large cations兲 of the -alumina structure is not complete.14 For the close packing to be complete, four anion sites 共per half-unit cell兲 in the intermediate plane should be occupied. In the -alumina structure, one Ba ion occupies a 2d 共BR兲 site, and the oxygen ions a 2c site. Considering the barium ion as a large ion, ideal close packing would have required occupancy of three mO sites surrounding the BR site by large H85 ions. The incompleteness of the anion close packing in the -alumina structure allows accommodation of large ions like barium in the intermediate plane. At the same time, the relatively open structure in the intermediate plane in the -alumina structure is the source of all the structural defects in BAM. This paper attempts to understand two aspects of the Reidinger defects. The first relates to the relaxation of the crystal structure surrounding the OR site. The second aspect involves the energetics of formation of the Reidinger defects. Both aspects have been studied using an atomistic simulation approach. Using the appropriate thermodynamic potentials calculated by this approach, the generation and variation of concentration of the Reidinger defects with initial conditions of synthesis have been studied. BAM is usually synthesized by solid-state reaction at ⬃1500°C. The ingredients 共starting materials兲 are thoroughly mixed and then held at a constant temperature to attain equilibrium by diffusion. When the chemical equilibrium is finally obtained, one expects the synthesis of the desired compound to be complete in accordance with the phase diagram for the chemical components of the mixture. The quality of the mechanical mixture of the chemical components, the thermal history, ambient conditions, and other relevant factors dictate the quality of equilibrium obtained and, therefore, the quality of the final products. Fluctuations in the concentration of the reactants either due to deliberate manipulation of the starting mixtures or loss of material during synthesis could lead to an undesirable level of defect concentration. In this context, the atomistic simulation methods provide an excellent tool for predicting the nature of the final product given the contents of the input materials and the thermal history using the calculated thermodynamic potentials for the optimized structures of the reactants and products. In the next section, the simulation method, as implemented in the computer codes “Shell,” has been described. The details of the theory of Shell are given in Ref. 15. Results from the atomistic simulation are then discussed, including the validation of the approach from a study of simple binary systems. Two distinct scenarios of formation of the Redinger defects in BAM are discussed. In the final section, conclusions based on the present calculations are presented. Simulation Method Interaction potential.— The simulation method employed in this work is based on the Born model of ionic solids.16 Any ionic solid is assumed to be a collection of point ions moving in the combined field of pairwise interaction potentials. These potentials have an attractive part, which is a sum of the Madelung and van der Waals potentials, and a repulsive component arising from the overlap of electronic clouds surrounding each ion. The repulsive component reflects the effect of Pauli repulsion, and prevents the collapse of the lattice structure from the long-range attractive, electrostatic potential. It is usually modeled in simple analytical forms, such as the Morse potential, Born-Mayer potential, and inverse potential. We have used in this work the Buckingham function, which combines the short-range overlap potential in an exponential form and the van der Waals potential as an inverse potential Vij = Aije−共rij/ij兲 − Cij 1 rij6 关1兴 where rij represents the distance between ions i and j. The parameters A,, and C describing the pairwise potentials for various ion combinations in various chemical environments are available from the literature. Additionally, the polarization of the solid is included by using the shell model originally proposed by Dick and Overhauser.17 In this model, each ion is assumed to consist of a massive core and a “massless” shell, the total charge on the ion being partitioned between the shell and core. The shell and core of a given ion are connected by a “massless” spring with a spring constant, k. There are several conceptual issues concerning the calculation of Downloaded on 2016-05-12 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract). H86 Journal of The Electrochemical Society, 152 共6兲 H84-H93 共2005兲 cohesive energy of solids and other structural thermodynamical properties based on the classical Born model. They have been discussed extensively in the literature.18 First, the concept of additive two-body potentials is only approximate. Early quantum mechanical calculations for the cohesive energy of solids have shown that much of the cohesive energy cannot be described by the two-body potentials.19 Second, the validity of the concept of separation of the motion of the nuclei from that of the electrons rests on the validity of the Born-Oppenheimer approximation. For most ionic solids with large optical gaps, the Born-Oppenheimer approximation is a very good approximation at low temperatures. But, with increasing temperature, most solids tend to be conducting, and the forbidden energy gap decreases rapidly. Thus, at a high temperature, one expects the Born-Oppenheimer approximation to break down. Therefore, describing the high-temperature behavior of solids using potential parameters based on structural properties at room temperature may not be a valid approach. Third, there are many different sets of pairwise potential parameters available in the literature.20 The difficulties in choosing appropriate pairwise potentials are apparent when one chooses to use this classical approach for predicting properties of complex systems. In this work, we addressed this issue by comparing various material properties of constituent binary oxides of BAM, namely aluminum oxide, barium oxide, and magnesium oxide, using various potential parameters. It is shown that the structural properties of interest in this work do not vary significantly with the choice of parameters. This is discussed in detail in the Results and Discussion section. Despite these apparent drawbacks, atomistic simulation methods of solids have led to many interesting results. For a complex system such as BAM, this method provides a fast computational approach for studying equilibrated structural properties. When used carefully with a proper appreciation of the complexity of the system, this classical approach provides reasonable explanations of the observed structural properties and predictions of novel properties of the material. Lattice dynamics approach for atomistic simulation of solids.— In this work, atomistic simulations of BAM and its defect structure were performed using the computational codes Shell.15 The scope of Shell extends beyond the static energy calculation and optimization of crystal structures at absolute zero temperature. The code implements a lattice dynamics approach for calculating the temperature dependence of the thermodynamic properties of solids. It provides a powerful alternative to molecular dynamics and Monte Carlo simulation of solids. Some of the interesting features of the lattice dynamics simulations of solids are discussed in the following section. Shell implements the lattice dynamics formalism within a quasiharmonic approximation. The equation of state thus generated is correct to the first order in anharmonicity. The free energy of the crystal is assumed to depend on the vibrational motion of nuclei in the harmonic potential generated by the superposition of pairwise potentials. The effect of anharmonicity of the crystal potential derives from the change in mode frequencies as a function of coordinates and temperature. This approach provides a powerful, robust alternative to molecular dynamics and Monte Carlo simulation of solids.15,21-23 It is particularly relevant below the Debye temperature when the quantum mechanical effects cannot be excluded. Shell is designed for optimization of the periodic crystal structures. Because our main focus is on understanding the nature of the defect structure of BAM and related thermodynamic properties, we used Shell within the framework of supercell approximation of imperfect solids24 in preference to the Mott-Littleton embedded cluster approach.25 In this approximation, point defects are assumed to be distributed periodically throughout the bulk, but the corresponding unit cells are significantly larger than the unit cells of the ideal lattice. With a random distribution of point defects within a unit cell, the supercell approach provides a powerful tool for simulation of the defect structure of a material. Structural input and optimization scheme.— A periodic crystal structure is defined by a lattice and a basis of atoms. External strain can deform the shape of the lattice, which is described by the external coordinates, ext. The internal coordinates, int, define the positions of the basis of atoms within a unit cell. The optimization process involves minimization of a suitably chosen thermodynamic potential with respect to these coordinates. The positions of the particles including core and shell in a macroscopically strained coordinate system are given by ␣ rix = 兺 共␦ ␥ ␣␥ + e␣␥兲共x␥ + i␥兲 关2兴 where the Greek superscripts are Cartesian indices 1…3, x is a lattice vector of the unconstrained lattice, and ␦␣␥ represents the Kronecker delta. Components of the tensor, e␣␥, describe the state and orientation of the strain. ␥i describes the ␥ component of the internal coordinate for particle i within a unit cell. The Voigt macroscopic strain coordinates are the components of the six-element vector 冢 冣 关3兴 兺ua = 兺uA 关4兴 e11 e22 e33 = e23 + e32 e13 + e31 e12 + e21 The internal coordinates i␣ are given in dimensionless units from the XRD measurements i␣ = t ␣ i t t t t␣ i t where the matrix At␣ describes the metric obtained from the lattice vector at. It is helpful to use the internal symmetry to reduce the number of internal coordinates. When optimizing a crystal without any external stress, the lattice vectors describe the external coordinates. The variables uti are expressed in terms of symmetric internal coordinates wm uit = git + 兺w t mgm,i 关5兴 m where gti is a fixed vector for a given particle i, and the directions gtm,i are determined by symmetry. The geometry of a crystal is then determined by optimizing the appropriate free energy with respect to the lattice vectors and internal coordinates wm. Using wm reduces significantly the number of independent variables and preserves the symmetry of the crystal. The significance of wm is obvious from a consideration of input coordinates for atoms occupying 6h sites in the space group, P63 /mmc. For example, the coordinate of an atom occupying a site with coordinates 共x, 2x, z兲 can be written as 关x共1, 2,0兲 + z共0, 0, 1兲兴. The entire set of coordinates describing atoms at the 6h site can be described by a suitable choice of direction gtm,i which will be held fixed during the minimization procedure and two variables x and z. Here x and z represent the internal coordinates wm. Thus, using x and z as internal coordinate variables not only preserves the overall symmetry of the crystal structure but also reduces the number of independent variables in the present case from 18 to 2. At a finite temperature under an applied pressure P0. the stable structure of a periodic lattice can be obtained by minimizing the Gibbs free energy G G = U − TS + P0V = F + P0V = stat + Fvib + P0V 关6兴 where U is the internal energy, S is the entropy, and F is the Helmholtz free energy. F has two parts: stat is the static 共lattice兲 energy, and Fvib is the vibrational contribution to the Helmholtz energy. The stable configuration corresponds to that vector in the parameter space which satisfies the condition Downloaded on 2016-05-12 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract). Journal of The Electrochemical Society, 152 共6兲 H84-H93 共2005兲 H87 Table I. Interatomic potential parameters. A„eV…, C „eV · Å6… and „Å… are defined in the text. K„eV/Å2… is the spring constant between oxygen shell and core. The numbers in parentheses provide appropriate constant for tetrahedral coordination. a. Source: See text footnote c. Function Atom type Atom type A C k O O O O O Al O O 22,764 1460.3 - 27.879 - 74.92 0.149 0.29912 - Atom type Atom type A C k O O O O O Al O O 9547.46 1725.2 - 32 - 54.8 0.21916 0.28971 - Atom type Atom type A C k O O O O O O O Al Mg Ba O O 22,764 1474.40 共1334.31兲 821.60 共710.50兲 931.70 - 17.890 - 27.29 0.149 0.30059 0.32420 0.39490 - Short Short Inverse Spring b. Source: Grimes 共Ref. 28兲 Function Short Short Inverse Spring c. Source: Park and Cormack 共Ref. 26,27兲 Function Short Short Short Short Inverse Spring d. Distribution of charges on oxygen Source Cormack Catlow Grimes 冉 冊 G A =0 ⬘ 关7兴 where ⬘ corresponds to all other coordinates other than A. The algorithm that implements the optimization condition 共Eq. 7兲 in Shell is discussed in detail in Ref. 15. In addition to optimization with respect to all coordinates, Shell also determines the optimal geometry at two levels of approximations: the zero static internal stress approximation 共ZSISA兲 and constant internal strain parameter 共CISPA兲 conditions. The condition for full optimization is 冉 冊 冉 冊 G G = =0 lext ⬘ int k ⬘ The equilibrium condition in ZSISA is given by 关8兴 冉 冊 冉 冊 G Gstat = =0 关9兴 lext ⬘ ext k ⬘ Because the number of external coordinates never exceeds 6, this approximation for large unit cells is considerably faster. In CISPA, a set of internal coordinates is found by full minimization of the static energy 冉 冊 冉 冊 Gstat Gstat = =0 关10兴 lext ⬘ int k ⬘ These internal coordinates are then held constant during subsequent optimization with respect to the external coordinates. BAM being a very complex system, we have confined the entire defect-related calculations to minimization with respect to the static energy, also referred to as the lattice energy. This approach has been used in the past to study the lattice structure of barium and lanthanum hexaaluminates.26,27 Here we have extended this approach to more complex defect structures with very interesting results. Oxygen 共shell兲 Oxygen 共core兲 −2.207 −2.869 −2.8 0.207 0.869 0.8 Results and Discussion Validation of potential parameters.— Alumina was used as a test case for validating the approach described earlier. In Tables I, different sets of potential parameters for pairwise interaction of aluminum and oxygen ionsc 28 for simulating aluminum oxides are listed along with those used for studying hexaaluminates.d Although the variation in the values of the interatomic potential parameters is significant, these parameters yield a nearly identical functional dependence of the potential on interatomic distance. In Table I, the pairwise potentials used by Park and Cormack are listed.26,27 The suitability of these potentials has been established for studying the structural and thermodynamic properties of hexaaluminates.26,27 Note that the potential parameters are different for octahedral and tetrahedral coordinations of the aluminum and magnesium ions Table II lists the calculated values of the lattice energy, lattice parameters, and elastic constants of ␣-alumina. The lattice energy is the same as the static lattice energy, stat. The structural parameters have been calculated by minimizing the static energy with respect to the lattice constants and dimensionless parameters, u in Eq. 4. The full lattice dynamics formalism of Shell has not been utilized, i.e., contributions from the vibrational motion of the nuclei are not included. This is not a very serious approximation, as we found from calculations on BAM. It has also been shown by Gillan that this is often a good approximation at higher temperatures because the change in internal energy is, to the first order, equal to the difference between enthalpy 共zero pressure兲 and 0 K internal energy.29 The entropy effects on the structural stability are not serious. Finally, the c The potential parameters used by Catlow and his collaborators are taken from the web site: http://www.ri.ac.uk/potentials d The potential parameters used by Park and Cormack in Ref. 26 and 27. They are almost the same as those in Ref. 30. Downloaded on 2016-05-12 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract). Journal of The Electrochemical Society, 152 共6兲 H84-H93 共2005兲 H88 Table II. Comparison of structural data of Al2O3 using potential parameters from Catlow library (text footnote c) and Grimes (Ref. 28). Experimental values are taken from Table II of Ref. 30. Structural data Units Lattice energy a c u1 u2 c11 c12 c13 c14 c33 c44 c66 eV Å Å 共units of a兲 共units of a兲 1011 dyn/cm2 1011 dyn/cm2 1011 dyn/cm2 1011 dyn/cm2 1011 dyn/cm2 1011 dyn/cm2 1011 dyn/cm2 Experiment −160.4 4.7628 13.0032 0.352 0.306 49.49 16.36 11.09 −2.35 49.80 14.74 16.67 Grimes 共Shell兲 Grimes 4.812 12.734 0.357 0.291 69.58 32.22 22.56 −4.69 57.44 15.95 18.69 structural parameters calculated by Shell are compared with those by the earlier workers using a similar lattice energy minimization procedure. This comparison provides a check for the reliability of the Shell algorithm. Alumina is one of the binary oxide components of BAM. Corundum, ␣-alumina, crystallizes in the hexagonal space group, R3̄c 共167兲. The aluminum and oxygen atoms occupy the 12c and 18e Wyckoff positions, respectively. Two lattice constants, a and c, and two u parameters completely describe the lattice and the coordinates of the aluminum and oxygen ions. Two sets of potential parameters were utilized for studying the stable alumina structure. They lead to lattice properties that are in good agreement with each other and also with the experimental values. We have also indicated results from Shell using these parameters, which compare well with those calculated by Catlow et al.30 and Grimes.28 Similar results were also obtained for MgO and BaO. These calculations provide justification for using the parameters listed in Table I for investigating the structural properties of BAM. Optimization of BAM structure: Distribution of the Mg sites at Al(2).—The interatomic potential parameters listed in Table I are used here for further investigation of structural properties of BAM and point defects in this lattice. Occasionally, results from other variations of interatomic potentials are used to establish the adequacy of those used by Park and Cormack. 共See footnoted.兲 In this section, we report three different aspects of our structural investigation of BAM: the distribution of Mg ions at Al共2兲 sites, optimized structure of BAM with Mg atoms at the most stable site, and temperature dependence of thermodynamic potentials of BAM. The crystal structure of BAM has been investigated using single crystal5 and powder samples. 共Several recent papers are based on powder diffraction.兲31,32 BAM crystallizes in the -alumina structure in the hexagonal space group, P63 /mmc. The crystal structure of BAM is related to the -alumina structure of Na Al11O17. The hexaaluminates in -alumina, magnetoplumbite, and related phases exhibit layer structures, consisting of intermediate layers containing Ba ions and spinel blocks stacking alternately along the c direction. The structure of BAM can be better appreciated if we consider Na -alumina as the parent structure and BAM as a derived system in which the Na ions are replaced by Ba ions, accompanied by charge compensation with Mg ions at Al sites. Because each unit cell contains two formula units and, therefore, two barium ions, there are two magnesium ions distributed among the aluminum sites. The XRD pattern for BAM is often analyzed assuming a random distribution of the magnesium ions at the aluminum ion sites in the spinel block. The aluminum ions occupy three different Wyckoff sites in the -alumina structure: 12k 关Al共1兲兴, 4f 关Al共2兲兴, 4f 关Al共3兲兴, and 2a 关Al共4兲兴. The aluminum ions at the 12k and 2a sites are octahedrally coordinated, and those at the 4f sites are tetrahedrally coordinated. −161.10 4.811 12.732 0.359 0.297 69.60 32.23 22.58 −4.68 57.43 15.96 18.68 Catlow et al. −160.21 13.2 42.96 15.48 12.72 −2.99 50.23 16.66 13.70 Catlow 共Shell兲 −160.56 4.7792 12.5638 0.359 0.294 64.82 29.69 20.63 −4.53 50.49 15.36 15.36 Clearly, unless the Mg ions occupy the 2a sites, they must be randomly distributed over the available aluminum sites to retain the same space group symmetry. Based on modeling27 and qualitative arguments,31 the magnesium ions are believed to occupy the Al共2兲 sites. But there are four Al共2兲 sites: Site 1 1/3 1/3 z Site 2 2/3 1/3 z + 1/2 Site 3 2/3 1/3 −z Site 4 1/3 2/3 −z + 1/2 with z = .024. The question is then if there is a preference for any particular combination of sites on the basis of lattice energies. Lattice energies, stat, for Mg atom pairs for combinations of sites 共1,2兲, 共1,3兲, 共1,4兲, 共2,4兲, 共2,3兲, and 共3,4兲 are calculated using two sets of potentials differing only in aluminum-oxygen interatomic potentials 共Table III兲. One set of potentials is from Lewis and Catlow33 used earlier for calculations involving alumina. 共These parameters are listed in the web site of footnotec.兲 The potential parameters in this set differ slightly from those used by Park and Cormack,26,27 but provide a better agreement with the observed cohesive energy for alumina. The Gibbs energies for occupancy of these sites by Mg atoms were probed using two sets of potentials. Except for the occupancies of site combinations 共1,3兲 and 共2,4兲, the occupancies of all other site combinations are energetically favorable. At 300 K, ⌬G corresponding to the chemical reaction 5Al2O3 + MgO + BaO → BaMgAl10O17 关11兴 is ⬃−6 eV. This suggests first the stability of the BAM lattice structure, and second that the magnesium ions are distributed randomly in the lattice. However, the magnesium ions avoid occupancy of Table III. Comparison of lattice energies of BaMgAl10O17 with Mg atoms distributed over different Al(2) sites. The change in lattice energy, ⌬E, corresponds to E„BaMgAl10O17… − E„MgO… + E„BaO… + E„5Al2O3…. Compound BaO MgO 5Al2O3 共Catlow and Lewis兲 BaO + MgO + Al2O3 BAM: Mg共1,2兲/共3,4兲 BAM: Mg共1,3兲/共2,4兲 BAM: Mg共1,4兲/共2,3兲 5Al2O3 BAM: Mg共1,3兲/共2,4兲 BAM: Mg共1,2兲/Mg共3,4兲 Lattice energy 共kJ/mol兲 ⌬E 共kJ/mol兲 ⌬E 共eV兲 −3020.746 −3902.216 −77359.052 −84282.014 −84862.051 −84222.552 −84853.434 −76602.247 −83361.845 −83985.761 −580.037 59.462 −571.420 163.365 −460.552 −6.01 0.62 −5.92 1.69 −4.77 Downloaded on 2016-05-12 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract). Journal of The Electrochemical Society, 152 共6兲 H84-H93 共2005兲 H89 Table IV. Variation of thermodynamic potentials of BAM with temperature. a. Temp. 300 K, a = 5.701 Å, c = 22.475 Å stat 共kJ/mol兲 −3902.086 −3019.721 −76601.424 −83523.424 −83985.154 −461.73 共−4.78 eV兲 Compound MgO BaO 5Al2O3 MgO + BaO + 5Al2O3 BAM ⌬ U 共kJ/mol兲 −3883.146 −3003.278 −76336.908 −83223.332 −83682.693 −459.3605 共−4.75 eV兲 G 共kJ/mol兲 −3892.378 −3022.064 −76420.948 −83335.388 −83795.572 −460.183 共−4.77 eV兲 U 共kJ/mol兲 −3823.340 −2940.478 −75640.557 −82404.376 −82872.840 −468.463 共−4.85 eV兲 G 共kJ/mol兲 −3985.484 −3157.681 −77404.675 −84547.839 −85004.392 −456.552 共−4.73 eV兲 U 共kJ/mol兲 −3794.135 −2908.218 −75324.956 −82027.309 −82507.608 −480.299 共−4.98 eV兲 G 共kJ/mol兲 −4043.819 −3234.670 −78040.109 −85318.598 −85769.775 −451.177 共−4.67 eV兲 b. Temp. 1500 K, a = 5.728 Å, c = 22.582 Å stat 共kJ/mol兲 −3898.951 −3015.583 −76592.791 −83507.326 −83976.536 −469.211 共−4.86 eV兲 Compound MgO BaO 5Al2O3 MgO + BaO + 5Al2O3 BAM ⌬ c. Temp. 2000 K, a = 5.741 Å, c = 22.639 Å stat 共kJ/mol兲 −3894.458 −3008.179 −76584.636 −83487.274 −83968.192 −480.918 共−4.98 eV兲 Compound MgO BaO 5Al2O3 MgO + BaO + 5Al2O3 BAM ⌬ sites where they are close to each other within the same spinel block as for the occupancy of sites 共1,3兲/共2,4兲. This is expected if one considers BAM as a derivative of Na -alumina, in which Ba atoms at the Na sites are charge compensated by the Mg atoms occupying Al sites. The Mg atoms at the Al sites are negatively charged with respect to the rest of the lattice, and thus, clustering of the Mg atoms is prevented by mutual electrostatic repulsion. Additionally, calculations performed with the potential used by Park and Cormack26,27 show the same trend, although the magnitude of change in the lattice energy is slightly lower 共⬃−4.8 eV兲. In the subsequent calculations of BAM, we have used the minimum energy configuration with the Mg ions at the site 共1,2兲. We also investigated the stability of BAM and the variation of static energy stat internal energy U, and Gibbs energy G at 300, 1500, and 2000 K. The results of these calculations are listed in Table IV, and the optimized geometry at 300 K is given in Table V. Table IV also gives lattice parameters a and c at the corresponding temperatures. There are several reasons for studying the variation of thermodynamic potential with temperature. In the subsequent calculations, we focus our attention on the stability of defect structures with respect to composition of the starting mixtures and ideal BAM. Be- cause BAM is synthesized at fixed temperature and pressure, the correct potential is the Gibbs energy. A chemical reaction at constant temperature and pressure moves forward, if it minimizes the corresponding Gibbs energies, i.e., the change in Gibbs energies is negative. Second, the optimized structure also refers to a minimum Gibbs energy configuration. But the calculation of Gibbs energy for a complex system such as BAM can be very time consuming. The supercell calculations for defect structures could be both time consuming and unstable. Thus, it is essential to determine if optimization of the static energy could lead reliably to the observed equilibrated structures. It is clear from Table IV that the changes in static energy, internal energy, and Gibbs energy are nearly equal. The optimized geometries calculated by minimization of the Gibbs energy are almost the same as those obtained by minimization of the static energy. Table V gives the optimized structure by minimization of the static energies. It is assumed that the magnesium atoms occupy aluminum ion sites 共1,2兲. For comparison, the structural data for BAM from the XRD measurements are also given in Table V. The agreement between the observed5,31 and calculated structures is excellent. Table V. Comparison of structural parameters of BAM calculated by Shell with experimental ones (Ref. 5 and 31) Lattice constants: a = 5.628 Å (expt.), 5.701 Å (theory); c = 22.658 Å (expt.), 22.475 Å (theory). Atom Wyckoff site x 共expt.兲 y 共expt.兲 z 共expt.兲 x 共theory兲 y 共theory兲 z 共theory兲 Ba Al共1兲 Al共2兲 Al共3兲 Al共4兲 O共1兲 O共2兲 O共3兲 O共4兲 O共5兲 2d 12k 4f 4f 2a 12k 12k 4f 4e 2c 2/3 0.8343 1/3 1/3 0 0.1534 0.5042 2/3 0 1/3 1/3 0.6686 2/3 2/3 0 0.3068 0.0084 1/3 0 2/3 1/4 0.10544 0.0240 0.017416 0 0.05152 0.14799 0.05901 0.14437 1/4 2/3 0.8325 1/3 1/3 0 0.1541 0.5002 2/3 0 1/3 1/3 0.6650 2/3 2/3 0 0.3082 0.0047 1/3 0 2/3 1/4 0.10459 0.0248 0.17391 0 0.05128 0.1461 0.059263 0.13843 1/4 Downloaded on 2016-05-12 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract). Journal of The Electrochemical Society, 152 共6兲 H84-H93 共2005兲 H90 Table VI. Lattice energies for BaO, MgO, Al2O3, and the model compound Ba2Al22O35. Lattice energy 共kJ/mol兲 Model compound BaO MgO Al2O3 Ba2Al22O35 −3020.746 −3902.216 −15320.450 −175351.032 Local structure of oxygen interstitials in intermediate plane.— It was discussed earlier how the color centers play an important role in the performance of a phosphor in fluorescent lamps. They have high oscillator strengths, and thus, the presence of color centers in a ppm range could provide strong absorption centers. Two electron centers 共F and F+兲 and a hole center have been observed in Ba -alumina phase I and M+-aluminates upon irradiation by X-ray and ␥-ray irradiations, and to a lesser degree by VUV/UV radiation. Only two of these color centers, a hole center, OI⬘, and an electron center, F+, are identified to be paramagnetic in nature by electron spin resonance and electron-nuclear double resonance measurements.6,8-10,34 The 11 line hyperfine pattern with a sequence of peaks in the intensity ratio of 1:2:3:4:5:6:5:4:3:2:1 for the F+ center suggests that this center is located between two equivalent Al nuclei with I = 5/2.34 The hyperfine structure results from the interaction of an unpaired electron with the two symmetrical Al ions. This center is associated with an anion vacancy, namely an oxygen vacancy in -alumina. The question is then which oxygen atom is responsible for generation of this center. The oxygen atoms in the spinel block are coordinated either to four or three aluminum ions. There are only two oxygen atoms in phase I of Ba -alumina that are coordinated to two aluminum ions: the bridging oxygen atom in the intermediate plane O 共5兲, and the interstitial oxygen ion OR located at a mid-oxygen site. These oxygen ions generate the Reidinger defects. O共5兲 is the bridging atom between the spinel blocks in all -alumina structures. It is located in the intermediate plane and is coordinated to two symmetrical aluminum ions, Al共3兲. The Reidinger defect appears in barium hexaaluminates phases I and II in the -alumina structure. 共For more details, see Ref. 26 and 31 and references cited.兲 This defect is associated with a highly reconstructed site with an oxygen atom, OR, located in the intermediate plane and bonded to two tetrahedrally coordinated aluminum atoms. Thus, the F+ center is associated with either the vacancy of O 共5兲 or OR. The F+ centers are rare in stoichiometric Na+--alumina or nonstoichiometric Na+-⬙-alumina, where the charge compensation is due to divalent ions in the spinel block. However, both these systems contain bridging oxygen atoms, O 共5兲. Therefore, the F+ centers could be formed only at the Reidinger defects. The structural information of Reidinger defects has been obtained by modeling and diffraction measurements. There are two simple questions of fundamental significance to the synthesis of BAM: First, what are the sources of the the oxygen interstitials in BAM? Second, will such interstitials induce the Reidinger defects? Oxygen interstitials could be formed in two circumstances. The first case is when the two magnesium ions are replaced by two aluminum ions. These two substitutional point defects could be charge compensated by one oxygen ion. This oxygen ion could be trapped in the intermediate plane. Second, a solid solution of BAL and BAM is formed due to excess alumina in the starting mixture.35 The situation of excess alumina is realized by deliberate addition of alumina over stoichiometry or the loss of MgO 共and BaO兲 during synthesis. We have investigated both scenarios leading to formation of oxygen interstitials in the following sections. Formation of AlMg-AlMg-Ol defect structure.—This complex defect structure, which results from substituting aluminum ions for two magnesium ions in BAM, is simulated by calculating the lattice energy of a model compound, Ba2Al22O35. This compound can be considered as resulting from a chemical reaction 2BaO + 11Al2O3 → Ba2Al22O35 关12兴 If the change in the Gibbs energy 共⬃⌬G兲 is negative for this reaction, the reaction can go forward. The final product in Reaction 12 is equivalent to BAM with two magnesium ions being replaced by two aluminum ions. The lattice energies for BaO, MgO, Al2O3, and Ba2Al22O35 are summarized in Table VI. The change in the Gibbs energy is approximated by the change in lattice energy of the reactants and the products. The change in Gibbs energy, ⌬G, for Reaction 12 is −146.97029 kJ/mol 共−1.52 eV兲. Thus, if magnesium oxide is lost during solid-state synthesis, there would be a tendency to form BaAl22O35. In other words, Reaction 12 could lead to incorporation of oxygen defects in BAM. Next we consider the stability of oxygen defects with respect to BAM by studying the chemical reaction Ba2Mg2Al20O34 + Al2O3 → Ba2Al22O35 + 2MgO 关13兴 This reaction describes alumina being dissolved in BAM accompanied by precipitation of MgO. The change in Gibbs energy is positive for this reaction and is equal to 136.5070 kJ/mol 共1.414 eV兲. At a synthesis temperature of 2000 K for BAM, this could lead to ⬃270 ppm oxygen defects in the lattice. At this concentration, color centers generated at oxygen defect sites could affect the performance of BAM. Next, we examine the structural details of the model compounds to determine if these oxygen centers are structurally similar to the Reidinger defects. We do not have any structural data for such defects, so we compare the calculated structural parameters with those Table VII. Optimized structural parameters (Å) for Ba2Al22O35. Experimental data are from Ref. 31. Structural parameters x 0.3801 0.8211 0.8464 Bond length Al1RO4 Al1RO2共x2兲 Al1ROR 1.73 1.74 1.76 Lattice parameters a c Experiment 共structural data for Ba-hexaaluminate, Phase I兲 Theory Atoms Ba Al1R OR 5.6894 22.2497 y 0.1539 0.6710 0.6918 z 0.25 0.1711 0.25 x 2/3 0.84176 0.87303 y 1/3 0.68352 0.74606 z 0.25 0.17709 0.25 1.73 1.77 1.68 5.588412 22.72626 Downloaded on 2016-05-12 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract). Journal of The Electrochemical Society, 152 共6兲 H84-H93 共2005兲 Table VIII. Lattice vectors in Å. Material BAL BAM a 共theory兲 c 共theory兲 a 共expt.兲 c 共expt.兲 5.65262 5.69161 22.48881 22.44138 5.588412 5.62534 22.72626 22.65751 Table IX. Coordinates of Al and OR ions of the Reidinger defect. The subscript R is used to describe the interstitial oxygen atom in the conduction plane. The coordinates are given in terms of lattice vectors. Ion x 共theory兲 y 共theory兲 z 共theory兲 x 共expt.兲 y 共expt.兲 z 共expt.兲 O Al 0.89799 0.86683 0.73665 0.68862 0.25 0.17734 0.87303 0.84176 0.74606 0.68352 0.25 0.17709 Table X. AlAO bond lengths (in Å) of the Reidinger defect. Bond Al-OR AlO4 AlO2 共x2兲 H91 ton or an energetic ion from the discharge removes this oxygen atom from the lattice, the associated oxygen vacancy could capture one or two electrons, forming F+- or F- type color centers, respectively. The local geometry of a Reidinger defect has been determined using the XRD technique.31,37,38 The calculated structural parameters reported in Tables VIII-X are in excellent agreement with the experimental ones.31 In Table VIII, we have shown the lattice parameters for BAM and BAL. The lattice constant a decreases from 5.63 Å in BAM to 5.59 Å in BAL and c increases from 22.66 Å to 22.73 Å. The theoretical values of these parameters for BAM and BAL show the same trend. The calculated positions of OR and Al共1兲 ions associated with the Reidinger defect are in good agreement with the experimental ones. The experimental values of Al-O bond lengths for this defect are satisfactorily reproduced by theory. The good agreement of experimentally observed structural parameters with those calculated by theory justifies the goodness of the twobody potentials used in this work and the structural model of BAL and that of Reidinger defects. To study the stability of BAL, the lattice energy of BAL is compared with those of alumina and barium oxide in relation to the chemical reaction Length 共theory兲 Length 共expt.兲 3BaO + 22Al2O3 Ba3Al44O69 1.651 1.746 1.767 1.684 1.734 1.770 The change in the lattice energy is −1882.92 kJ/mol 共−19.51 eV兲, indicating that this reaction would lead to formation of BAL. The stability of BAL with respect to BAM is examined by investigating the lattice energies of materials involved in the chemical reaction for barium hexaaluminate in phase I 共Ba.75Al11O17.25兲 共see Table VII兲, in which the presence of the Reidinger defects has been established.31 The calculated parameters reveal some interesting details about the interstitial oxygen ions. 1. Oxygen interstitials are located at the mO sites near the large cation site. 2. In phase I of Ba-hexaaluminate, the large cation is actually missing near the oxygen interstitial. In the present case, the Ba ion moves to an mO site to allow the structural relaxation around the impurity ion. This is the first theoretical evidence of a cation occupying an mO site in the BAM lattice.3,36 3. Al共1兲 atoms above and below the interstitial oxygen ion move toward the intermediate plane to form a “Reidinger defect.” The coordination of the aluminum atoms changes from an octahedral to a tetrahedral configuration similar to those of the Reidinger defects. 4. The Al-O bond lengths for this reconstructed defect compare reasonably well with those for a Reidinger defect in ideal phase I 共Table VII兲. These results suggest the possibility of formation of the Reidinger defects in BAM at a ppm level when the starting materials have excess alumina or are deficient in MgO. Most importantly, it suggests that the energetically favorable configuration surrounding an interstitial oxygen ion is very similar to a Reidinger defect. We also examined several other plausible mechanisms to incorporate an interstitial oxygen ion, but they require substantially high energy for their formation and are unlikely to be formed. Stability of BAM-BAL solid solution.—The role of Reidinger defects, as the most likely precursors of color centers in BAM, was discussed earlier. The Reidinger defects are intrinsic to barium hexaaluminate in phase I, Ba.75Al11O17.25 共BAL兲. The chemical composition of BAL and the available sites in the -alumina structure in which BAL crystallizes imply that one of every four halfcells would have a vacancy at the site of a barium ion and an extra oxygen atom at the mid-oxygen site adjacent to this vacancy. These defect half-cells are randomly distributed throughout the crystal. The interstitial oxygen ion, near a barium ion vacancy and occupying a mid-oxygen site, generates the Reidinger defect. When a VUV pho- 关14兴 2Ba2Mg2Al20O34 + 2Al2O3 Ba3Al44O69 + BaO + 4MgO 关15兴 This hypothetical chemical reaction provides the theoretical basis to determine if excess alumina would induce generation of BAL. The change in lattice energy in this reaction proceeding from left to right is −40.713 kJ/mol 共−0.42 eV兲, and thus, it favors formation of BAL when the input mixture for BAM contains excess alumina. Reaction 15 also suggests that BaO and MgO would leave the BAM lattice to accommodate BAL. Based on lattice energies calculated for various oxides in this study, the error in lattice energies in Reaction 15 is estimated to be less than ±0.5%. The relatively small change in lattice energy should also be considered from the perspective of forming solid solution BAM and BAL. BAM has already been shown to be very stable with respect to BaO, Al2O3, and MgO, and So is BAL 共Reaction 14兲. Thus, we expect only a small change in lattice energy between the reactants and products. Additionally, the entropic contribution to the change in Gibbs energy at the high solid-state synthesis temperature is also expected to favor the forward reaction. The significance of this reaction is evident when one examines the phase diagram of BAM39 with recent modifications suggested by Diaz et al. 共Fig. 2兲.35 The chemical composition of the reactants in Reaction 15 corresponds to a point on BAM and the alumina join close to BAM. Using the phase diagram in Fig. 2, one would conclude that this composition in equilibrium would lead to two separate phases of BAM and alumina. However, the lattice energy calculation for Reaction 15 suggests that the chemical reaction favors phase separation of BAL, BaO, and MgO. Thus, our calculations and the original phase diagram are not compatible. Two simple modifications of the phase diagram would make it compatible with our calculations and, most importantly, with some recent experimental results regarding the solubility of alumina in BAM. There is probably another stable point in the phase field of BAM, alumina, and BAL which splits the phase field near BAM. This was suggested by Diaz et al.35 based on a study of color shift of BAM with excess alumina and XRD results indicating absence of alumina up to a certain amount of excess alumina in the starting materials for BAM. The revised phase field due to point P is indicated by the dashed lines. The other alternative is that the phase field characterized by BAM, BAL, alumina, and spinel-alumina solid solution can be di- Downloaded on 2016-05-12 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract). H92 Journal of The Electrochemical Society, 152 共6兲 H84-H93 共2005兲 Figure 2. Phase diagram of BaO, Al2O3, and MgO. The dashed lines are recommended changes in the phase field near BAM by Diaz et al.35 BAL-I is the same as BAL discussed in the text. BAL-II is the barium-rich phase of barium hexaaluminate, Ba2.33Al21.33O34.33. BAM-S is the solid solution join of BAL and BAM. BAM-II and BAM-IIS are Mg-rich phases of BAM. For more details, see Ref. 39. vided into two regions by a BAL and spinel join instead of a BAM and alumina join. In either case, the BAM and alumina mixture in Reaction 15 leads to two phases: a solid solution of BAM and BAL, and a solid solution of alumina and spinel. This requires that the following equation should allow solutions for x,y, and z 2共Ba2Mg2Al20O34 + Al2O3兲 xBa3Al44O69 + yBa2Mg2Al20O34 + z共MgAl2O4 + 2.07Al2O3兲 关16兴 The spinel and alumina contents on the right side of Reaction 16 correspond to the solid solution limit of spinel and alumina in the phase diagram. Reaction 16 leads to values of x,y, and z of 0.12, 1.81, and 0.37, respectively, corresponding to a mixture of BAL, BAM, and a solid solution of spinel alumina in the ratio of 5:79:16. Because BAM and BAL have a continuous solid solution range at this range, BAM and BAL would appear as a solid solution if the reaction is favored energetically. To establish if the chemical reaction hypothesized in Reaction 16 would lead to a solid solution of BAM and BAL, one needs the lattice energy for the spinel in addition to those of BAM, BAL, and alumina. Therefore, the structural and thermodynamic properties of spinel were calculated. The pair potentials for Mg, Al, and O ions are the same as those used in our earlier calculations involving BAM. The lattice energy for the spinel was calculated assuming a face-centered cubic structure. The magnesium and aluminum ions are assumed to occupy special Wyckoff positions 8a and 16d. The oxygen atoms occupy the Wyckoff site 32e. For a complete specification of all the coordinates, this site requires specification of a parameter u in addition to the lattice constant a. Thus, the lattice energy was optimized with respect to the lattice parameters a and u. The optimized values of the lattice constant a and the structural parameter u are 8.1002 Å and 0.38951, respectively. These values compare well with the experimentally observed values of a and u of 8.0625 Å and 0.38672 from XRD40 and 8.08 Å and 0.387 from neutron diffraction measurements.41 Using the calculated lattice energy for the optimized structure, the change in lattice energy for the chemical reaction Al2O3 + MgO MgAl2O4 is −142.47763 kJ/mol 共−1.47607兲. It is in reasonable agreement with the observed value of −0.38 eV.42 The over- estimation is partly due to the choice of separate potentials for aluminum and magnesium ions in octahedral and tetrahedral coordinations used in this work. As indicated by Park and Cormack,27 this is essential for a proper description of the lattice structures where the aluminum ions occupy both the octahedral and tetrahedral sites, as in BAM. Using the lattice energy calculated by spinel, BAM, BAL, and alumina, one could determine if the reaction of BAM and alumina 共Reaction 15兲 would result in BAL, BAM, and a solid solution of spinel and alumina. Using the corresponding lattice energies, the change in lattice energy is −114.89 kJ/mol. This suggests that the presence of alumina would drive the aforementioned chemical reaction forward. Because BAM and BAL form a continuous solid solution range, the resulting product would contain two phases: a solid solution of BAM and BAL, and a solid solution of spinel and alumina. Clearly, from the earlier discussion, a solid solution of BAM and BAL would be formed if alumina were present in the composition of the starting materials in excess of stoichiometry. It could happen in two different ways: deliberate addition of alumina to the starting compounds in excess of stoichiometry or loss of MgO and BaO during synthesis. In either case, excess alumina in the input mixture of materials would lead to a solid solution of BAM and BAL. BAL would introduce the Reidinger defects, which in turn would generate color centers during lamp life. Thus, the best way to control color centers is to search for synthesis procedures to generate stoichiometric BAM. The other alternative is to search for phosphor treatments that could prevent creation of oxygen vacancies by VUV radiation from the discharge. Conclusions The main objective of this study is to determine mechanisms of degradation of BAM during lamp life. Based on earlier works on color centers in -alumina, it was hypothesized that the Reidinger defects are precursors of the electron centers, F and F+. Based on this hypothesis, plausible different scenarios that could lead to formation of these centers were explored using atomistic simulation methods. Two scenarios are interesting from the perspective of phosphor synthesis. Both involve the presence of excess alumina. In one case, two aluminum ions substitute for two magnesium ions and incorporate an oxygen atom at an mO site. This would lead to a reconstruction of the site around the oxygen atom at the mO site, leading eventually to a Reidinger defect and the displacement of a Ba ion from the BR site to a nearby mO site. Two magnesium ions precipitate out of the lattice. This could lead to formation of a Reidinger defect near a Ba site. The other case involves a solid solution of BAL and BAM. It is also induced by excess alumina, but differs from the previous case in that a Reidinger defect is created near a Ba vacancy and Ba and Mg ions leave the lattice. This solid solution is thermodynamically more stable than the first scenario and is more likely to form if excess alumina is present in the starting mixture or if the starting mixture is deficient in both MgO and BaO. The loss of MgO and possibly BaO could occur during the synthesis of phosphor in hydrogen atmosphere. This loss could result in the enhancement of alumina in the starting material. Raukas observed color center absorption bands in a stoichiometric formulation of BAM 共see footnotea兲. One of the plausible explanations for this observation is the loss of Mg ions during the synthesis process. Bheemineni and Readey43 have observed that the rate of evaporation of MgO is inversely proportional to the water vapor content of the reducing hydrogen gas and directly proportional to the square root of the gas velocity. Probably one could use wet hydrogen during the synthesis process and optimize its flow rate to reduce the vaporization rate of MgO while reducing trivalent europium ions. Acknowledgments I am grateful to Professor N. Allan for a copy of Shell and associated codes for analysis of the results from Shell. Special ac- Downloaded on 2016-05-12 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract). Journal of The Electrochemical Society, 152 共6兲 H84-H93 共2005兲 knowledgments are due to Professor K. H. Johnson, Dr. J. Lister, E. Dale, and Dr. M. Raukas for many helpful discussions and suggestions. OSRAM-SYLVANIA Incorporated assisted in meeting the publication costs of this article. References 1. S. Oshio, T. Matsuoka, S. Tanaka, and H. Kobayashi, J. Electrochem. Soc., 145, 3903 共1998兲. 2. K. Yakota, S. Zhang, K. Kimura, and A. Sakamoto, J. Lumin., 92, 223 共2001兲. 3. P. Boolchand, K. C. Mishra, M. Raukas, A. Ellens, and P. C. Schmidt, Phys. Rev. B, 66, 134429 共2002兲 and references cited. 4. J. M. Flaherty, J. Electrochem. Soc., 128, 131 共1981兲. 5. N. Iyi, Z. Inoue, and S. Kimura, J. Solid State Chem., 61, 236 共1986兲. 6. D. Gourier, B. Viana, P. Bellenoue, J. Therry, and D. Vivien, Radiat. Eff. Defects Solids, 134, 431 共1995兲. 7. D. Gourier, P. Bellenoue, and J. Thery, Ann. Chim. (Paris), 16, 391 共1991兲. 8. W. L. Roth, F. Reidinger, and S. LaPlace, in Superionic Conductors, G. D. Mahan and E. L. Roth, Editors, p. 223, Plenum, New York 共1977兲. 9. R. C. Barklie, J. R. Niklas, and J. M. Spaeth, J. Phys. C, 13, 1745 共1980兲. 10. D. Gourier, D. Vivien, and J. Livage, Phys. Status Solidi A, 56, 247 共1979兲. 11. R. C. Barklie, J. R. Niklas, J. M. Spaeth, and R. H. Bartram, J. Phys. C, 16, 579 共1983兲. 12. N. Iyi, Z. Inoue, S. Takekawa, and S. Kimura, J. Solid State Chem., 60, 41 共1985兲. 13. M. Raukas, Private communication. 14. A. F. Wells, Structural Inorganic Chemistry, 5th ed., Oxford University Press, New York 共1984兲. 15. M. B. Taylor, G. D. Barrera, N. L. Allan, and T. H. K. Barron, Phys. Rev. B, 56, 14380 共1997兲. 16. M. Born and K. Huang, Dynamical Theory of Crystal Lattices, Clarendon, Oxford 共1988兲. 17. B. G. Dick and A. W. Overhauser, Phys. Rev., 112, 90 共1958兲. 18. See, for example, Computer Simulation of Solids, C. R. A. Catlow and W. C. Mackrodt, Editors, Springer-Verlag, Berlin 共1982兲. 19. P. Löwdin, A Theoretical Investigation into Some Properties of Ionic Crystals, Uppsala Press, Uppsala, Sweden 共1948兲. 20. Refer to the web site: http://www.ri.ac.uk/potentials 21. G. D. Barrera, M. B. Taylor, N. L. Allan, T. H. K. Barron, L. N. Kantorovich, and H93 W. C. Mackrodt, J. Chem. Phys., 107, 4337 共1997兲. 22. N. L. Allan, M. Braithwaite, D. L. Cooper, W. C. Mackrodt, and S. C. Wright, J. Chem. Phys., 95, 6792 共1991兲. 23. G. W. Watson, P. Tschaufeser, A. Wall, R. A. Jackson, and S. C. Parker, in Computer Modeling in Crystallography, C. R. A. Catlow, Editor, p. 1, Academic, San Diego 共1997兲. 24. See, for example, R. A. Evarestov, Phys. Status Solidi B, 72, 569 共1975兲; P. C. Schmidt, A. Weiss, S. Cabus, and J. Kübler, Z. Naturforsch., A: Phys. Sci., 42, 1321 共1987兲; K. C. Mishra, K. H. Johnson, P. C. Schmidt, B. G. DeBoer, J. Olsen, and E. A. Dale, Phys. Rev. B, 43, 14188 共1991兲. 25. N. F. Mott and M. J. Littleton, Trans. Faraday Soc., 34, 485 共1938兲; for details, see A. B. Lidiard and M. J. Norgett in Computational Solid State Physics, F. Herman, N. W. Dalton, and T. R. Koehler, Editors, Plenum, New York 共1972兲. 26. J. Park and A. N. Cormack, J. Solid State Chem., 121, 278 共1996兲. 27. J. Park and A. N. Cormack, J. Solid State Chem., 130, 199 共1997兲. 28. R. W. Grimes, J. Am. Ceram. Soc., 77, 378 共1994兲. 29. M. J. Gillan, Philos. Mag. A, 43, 301 共1981兲. 30. C. R. A. Catlow, R. James, W. C. Mackrodt, and R. F. Stewart, Phys. Rev. B, 25, 1006 共1982兲. 31. S. R. Jansen, Ph.D. Thesis, Technische Universiteit Eindhoven, Eindhoven, The Netherlands 共1998兲. 32. See, for example, Y. Kim, S. Kang, J. Lee, M. Jung, and K. H. Kim, J. Mater. Sci. Lett., 21, 219 共2002兲. 33. G. V. Lewis and C. R. A. Catlow, J. Phys. C, 18, 1149 共1985兲. 34. T. Gbehi, D. Gourier, J. Thery, and D. Vivien, J. Solid State Chem., 83, 340 共1989兲. 35. A. L. Diaz, C. F. Chenot, and B. G. DeBoer, Proceedings of the 19th International Display Research Conference, p. 65, SID 共1999兲. 36. K. C. Mishra, M. Raukas, A. Ellens, and K. H. Johnson, J. Lumin., 96, 95 共2002兲. 37. N. Iyi, Z. Inoue, S. Takekawa, and S. Kimura, J. Solid State Chem., 52, 66 共1984兲. 38. F. P. F. Berkel, H. W. Zandbergen, G. C. Verschoor, and D. J. W. Ijdo, Acta Crystallogr., Sect. C: Cryst. Struct. Commun., 40, 1124 共1984兲. 39. M. Göbbels, S. Kimura, and E. Woermann, J. Solid State Chem., 136, 253 共1998兲. 40. T. Yamanak, Y. Takeuchi, and M. Takonami, Acta Crystallogr., Sect. B: Struct. Sci., 40, 96 共1984兲. 41. G. E. Bacon, Acta Crystallogr. 5, 684 共1952兲. 42. A. Navrotsky and O. J. Klepa, Inorg. Chem., 5, 192 共1966兲. 43. V. Bheemineni and D. W. Readey, Vaporization of Magnesium Dioxide in Hydrogen, Preprint. Downloaded on 2016-05-12 to IP 130.203.136.75 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
© Copyright 2026 Paperzz