J. Phys. Chem. C 2010, 114, 13039–13046
13039
Idealizing γ-Al2O3: In Situ Determination of Nonstoichiometric Spinel Defect Structure Michelene E. Miller† and Scott T. Misture* Alfred UniVersity, 2 Pine Street, Alfred, New York 14802 ReceiVed: March 26, 2010; ReVised Manuscript ReceiVed: May 26, 2010
The ideal γ-Al2O3 structure is elucidated by solving the defect reaction for alumina-rich magnesium aluminate spinel. The selective reduction of nickel from NixMg1-xAl2O4 spinel through reaction with hydrogen enables the in situ analysis of the defect reaction as it progresses. The reaction is described by the following equation: NixMg1-xAl2O4 + xH2 f xNi0 + Mg1-xAl2O4-x + xVNi′′ + xVO•• + xH2O. Rietveld refinement of neutron and X-ray powder diffraction data is used to determine the following: (1) the weight percents of the products, (2) the nickel occupancy in the spinel phase, and (3) the Mg1-xAl2O4-x lattice parameter. The weight loss during the hydrogen reaction is determined using in situ thermogravimetric analysis. The hydrogen reaction is generalized to describe the defect reaction for alumina-rich spinel: Al2O3 f VMg′′ + VO•• + 2AlAlx + 3OOx. Application of this defect equation to γ-Al2O3 predicts a significantly lower number of ions per unit cell than is commonly accepted. This defect model predicts 16 Al3+ ions and 24 O2- ions per unit cell versus the commonly accepted 64/3 Al3+ ions and 32 O2- ions per unit cell. We relate the large concentration of vacancies to the porous microstructure of γ-Al2O3. The role of the anion and cation vacancies in oxidation and reduction reactions may be described as a Lewis acid and Bronsted-Lowry base, respectively. (eqs 1,2 2,12 and 313):
Introduction γ-Al2O3 is one of the most important materials in catalysis. Knowledge of its structure is essential to understanding its catalytic function. The structure of γ-alumina has long been debated. It is generally conceived and widely accepted to be a defect MgAl2O4 spinel for the limiting case in which there is no magnesium, i.e., Al8/3O4.1 In 1958, Jagodzinski and Saalfeld2 described this defect structure to contain aluminum vacancies (VAl′′′), which are charge-compensated by site-disordering AlMg• species. Site-disordering defects are intuitively a preferred charge-compensating defect because site disorder is an intrinsic defect in stoichiometric spinels of varying composition.3 As highlighted by Levin and Brandon,1 the Al distribution in γ-Al2O3 remains unclear despite many decades of experimental work and more recent atomistic simulation studies. The mechanisms of the polymorphic phase transformations in Al2O3 also remain unclear when applying the current defect model for γ-Al2O3. An added layer of complexity in elucidating the γ-Al2O3 structure is the debate over the role of hydrogen in the structure. One group provides evidence that hydrogen is necessary to stabilize the structure,4-6 while the other camp counters with equally justifiable evidence against this conclusion.7-10 γ-Al2O3 is metastable, as a transitional state between the stable spinel and corundum structures. Because γ-Al2O3 is not an equilibrium state, it is not surprising that the empirical data of researchers over the past 50+ years have had a fair degree of variation. In this study, we present an alternate approach to revealing the “ideal” transitional state of γ-Al2O3 based upon the defect reaction for the MgAl2O4sγ-Al2O3 solid solution.11 Three defect reactions for nonstoichiometric magnesium aluminate spinel have been proposed based on post situ data * To whom correspondence should be addressed. E-mail: misture@ alfred.edu. † Currently at Excelerant Ceramics, 200 North Main Street, Alfred, New York 14802.
4Al2O3 f 3AlMg• + VAl′′′+5AlAlx + 12OOx
(1)
4Al2O3 f VMg′′+2AlMg• + 6AlAlx + 12OOx
(2)
17Al2O3 f 5VMg′′+8AlMg• + 26AlAlx + 51OOx + VO•• (3) The majority of studies including nuclear magnetic resonance and a combination of neutron diffraction and X-ray diffraction (XRD) rely on complex relationships that are based on defect reaction (1) to describe the cation vacancy (VAl′′′) distribution over the tetrahedral and octahedral sites.1,14-17 Defect reaction (1) is also the model that predicts the γ-Al2O3 structure first proposed by Verwey.18 In this study, the defect reaction for alumina-rich spinel (MgAl2O4sγ-Al2O3) is determined by selectively reducing nickel from NixMg1-xAl2O4 spinel through reaction with hydrogen. This method enables the in situ analysis of the defect reaction as it progresses by observing the defect species as they evolve. With progressive formation of a more alumina-rich spinel as nickel is reduced from NixMg1-xAl2O4 spinel, the γ-Al2O3 structure is approached. A clear advantage of this approach is the study of microscale ceramic grains to largely eliminate the surface effects that complicate the study of nanoscale materials derived from alumina precursors. Finally, we apply our proposed defect equation for alumina-rich spinel to elucidate the ideal γ-Al2O3 structure. Experimental Section The compositional series NixMg1-xAl2O4, where 0 e x e 1, was synthesized from reagent-grade MgOH2, NiO, and Al2O3. Batches were ball-milled for 2 h using alumina media in
10.1021/jp102759y 2010 American Chemical Society Published on Web 07/14/2010
13040
J. Phys. Chem. C, Vol. 114, No. 30, 2010
Miller and Misture
TABLE 1: Reaction Temperatures and Times for NixMg1-xAl2O4 Spinel temperature (°C) 650 700 750 800 850 900 950 1000
time (h) 94 51 29 24 24 12 12 12
deionized water and subsequently dried at 90 °C. Pellets were sintered at 1450 °C in increments of 18 h until phase purity was achieved. Between sintering cycles, the pellets were crushed and ground using an alumina mortar and pestle for analysis using XRD. Once each single-phase spinel composition was formed, the samples were crushed, ground, and ball-milled for ∼24 h. These powders were used for all subsequent analyses. Heat treatments of the spinel powders were conducted in a closed system of ∼300 Torr of ultrapure H2 for various times. The heat treatment temperatures and times are recorded in Table 1. The reaction times exceed the time required to completely react the spinel samples with hydrogen for the given temperature. Completion of the hydrogen reaction was determined using in situ high-temperature X-ray diffraction (HTXRD) isothermal treatments in flowing 4H2-96N2. The end of the reaction was determined to be the time at which the formation of additional metallic nickel ceases. Thermogravimetric measurements were performed using a TA Instruments SDT 2960 DSC-DTA. The powdered samples having a weight of 15-30 mg were contained in alumina pans. Each measurement was repeated at least three times for each composition. Mass density was measured using He pycnometry. At least three density measurements for each composition were recorded. Room temperature powder XRD data was collected using Cu KR radiation in Bragg-Brentano geometry. Data was collected from 5°-160° 2θ using a 0.03° 2θ step size and 5 s count time. Topas V3 was used for Rietveld refinement of the XRD data.19 The diffraction-derived crystallite size was determined using the fundamental parameters peak shape approach in Topas. In situ XRD measurements were recorded using a custom high-temperature X-ray diffractometer capable of full atmosphere control.20 Powder samples were mounted onto polycrystalline alumina sample holders. The diffraction furnace temperature was calibrated using common phase transformation standards. Neutron powder diffraction was performed at room temperature at the High Pressure and Preferred Orientation diffractometer (HIPPO) at the Lujan Center at Los Alamos National Laboratory. HIPPO utilizes a high-flux, low-resolution moderator with an incident beam path of 8 m. The 90° and 144° detector banks were used in the refinement, covering a d-spacing range of 0.3 to 3.2 Å. The sample, a mixture of Ni + Mg0.75Al2O3.75 was contained in a vanadium sample holder and spun during measurement. Combined X-ray and neutron Rietveld refinements were performed using GSAS,21,22 using both unconstrained structure models and models constrained to represent defect models (1) and (4b/4c). (Defect model (4b/4c) is developed in this article and will be addressed in the Discussion.) For model (1), the nickel on the octahedral site was constrained to the aluminum such that aluminum was “created” as nickel vacancies were created to maintain charge balance. For model (4b/4c), the nickel
Figure 1. The change in lattice parameter as a function of x for NixMg1-xAl2O4. The error bars for this study are smaller than the data points.
was constrained to the oxygen such that as nickel vacancies were created an equal number of oxygen vacancies were created. A separate constraint was used to allow the cations to move between the tetrahedral and octahedral sites (site mixing). Two “free” refinements were also performed: (#1) the oxygen occupancy was fixed at 0.94 according to the TGA result and defect eq (4c), while the total magnesium and aluminum contents were constrained to 6 and 16 atoms per unit cell, respectively; and (#2) the oxygen, magnesium, and aluminum occupancies were free to refine. Results NixMg1-xAl2O4 Starting Spinel. Single-phase powders were produced for each composition, with average particle sizes of ∼1 µm as determined using SEM photos, and diffraction-derived crystallite size between 100 and 220 nm. The relationship between the spinel lattice parameter and the composition for NixMg1-xAl2O4 is shown in Figure 1. The lattice parameter decreases linearly with increasing amounts of nickel over the compositional range. The data for the spinels synthesized in this study are plotted with values reported in the literature. Overall, the lattice parameters are in good agreement for all compositions though there is some scatter within and between the different studies. The quench temperature or temperature at which the spinels were synthesized in the respective study is noted. The lattice parameter is larger for samples quenched from higher synthesis temperatures for a given composition. H2 Reaction with NixMg1-xAl2O4 Spinel. Rietveld analysis of the XRD data for Ni0.5Mg0.5Al2O4 spinel following reaction at 850 °C for 24 h is shown in Figure 2. The only products of the reaction of NixMg1-xAl2O4 spinel with H2 are metallic nickel and spinel for x e 0.50 for all reaction temperatures between 650 and 1000 °C. Below 650 °C, the reaction rate was slow; consequently, no additional studies were performed at lower temperatures. The corundum phase noted in Figure 2 is contamination that occurred during ball milling prior to reaction with hydrogen. The weight percent of corundum remains constant during the reaction with hydrogen at all reaction temperatures. The Mg-O-Al framework of the defect spinel phase is stable following the reaction with hydrogen up to 1000 °C for x e 0.50. For x g 0.75, the spinels become unstable during reduction, marked by the formation of transition alumina polymorphs for
γ-Al2O3: Determination of Spinel Defect Structure
Figure 2. An example refinement of XRD data for Ni0.5Mg0.5Al2O4 reacted at 850 °C for 24 h in H2. The markers identifying the diffraction line positions for the respective phases are as follows: black circles, top ) spinel; red triangles, center ) nickel; and green squares, bottom ) corundum.
reaction temperatures between 800 and 1000 °C. The products of NiAl2O4 reacted with H2 between 800 and 1000 °C are shown in Figure 3. The products of these reactions are dependent on the reaction temperature. The formation of γ-alumina occurs at 800 °C for x ) 0.75 (not shown) and x ) 1.0. Transformation of γ-alumina to θ-alumina is observed at 1000 °C with increasing amounts of θ-alumina forming at longer reaction times for x ) 1.0. No evidence of γ-alumina transitioning to θ-alumina is observed for x ) 0.75 though the appearance of weak unidentifiable diffraction lines in the XRD pattern is observed (not shown). Ni Reduction. The amount of nickel formed during the reactions, as well as the Ni site occupancy, was determined using Rietveld analysis. During refinement, the occupancies of all spinel sites were constrained to maintain charge neutrality by creating an equal number of nickel vacancies (VNi′′) to oxygen vacancies (VO••). The validity of this constraint is addressed in the following section. The results for Ni0.5Mg0.5Al2O4 are shown in Figure 4 for each reaction temperature, where the fraction of Ni metal is normalized, and the Ni site occupancy is the sum over tetrahedral and octahedral sites. These two independent variables in the refinement follow identical trends with temperature and demonstrate that all of the Ni initially present in the oxide is reduced to the metallic state by 850 °C. Formation of VO••. The Ni vacancies formed during reduction of the spinel must be charge-balanced by either the formation of hydroxyl species or oxygen vacancies. The formation of oxygen vacancies (eVolution of O2) will result in a measurable weight loss. To understand the mechanism, the weight loss of the spinel samples during reaction with hydrogen (flowing 4H2-96N2) is measured in situ using thermogravimetric analysis (TGA). An example set of data for Ni0.50Mg0.50Al2O4 is shown in Figure 5, demonstrating that the sample mass decreases with temperature until a minimum value is reached at 1000 °C. The weight loss for the samples is calculated relative to the value at 600 °C, or in other words after removal of physisorbed water. A plot of the experimentally determined weight loss is shown in Figure 6 for each composition, calculated as (initial - final)/ initial mass, again relative to the mass at 600 °C. Also included in the plot is the theoretical weight loss for each composition that is expected to occur if oxygen vacancies are created to charge-balance the formation of nickel vacancies during the reaction with hydrogen. In other words, we calculate the weight
J. Phys. Chem. C, Vol. 114, No. 30, 2010 13041 loss for the final-phase assemblage of NixMg1-xAl2O4-x + xNi metal, assuming full reduction of the Ni from the oxide. The experimental data is in excellent agreement with the theoretical data for all compositions. Density and Crystal Stucture. In situ diffraction data for Ni0.5Mg0.5Al2O4 spinel in flowing 4H2-96N2 reacted at 800 °C is shown in Figure 7. The formation of metallic nickel with increasing reaction time is observed, accompanied by a decrease in the spinel lattice parameter. The spinel lattice parameter decreases by ∼0.6% over the course of the reduction reaction, but the spinel framework structure remains intact. As the amount of nickel reduced from NixMg1-xAl2O4 increases, the spinel phase becomes increasingly defective or nonstoichiometric. The lattice parameters for the nonstoichiometric spinels for x e 0.50 formed by selectively reducing the nickel from NixMg1-xAl2O4 are plotted in Figure 8. These lattice parameters are compared with values reported in the literature for nonstoichiometric spinels synthesized through the solid-state reaction of oxides, demonstrating good agreement with previous work.11,13,23 Both the change in the site occupancies (VNi′′ and VO••) and the change in the spinel lattice parameter will result in a measurable change in mass density. Densities of the defect spinels are estimated using the pycnometer density for the Ni + spinel mixtures and the Rietveld-derived weight fraction of each phase. The densities of the nonstoichiometric spinels following reaction with H2 at 1000 °C for 12 h are shown in Figure 9. The density of the spinel decreases sharply with increasing alumina content. Combined refinements of X-ray and neutron powder diffraction data with constraints to simulate the various defect models produce the results shown in Table 2. The combination of X-ray and neutron diffraction provides sensitivity to both the cation and anion sublattices. All models tested yield similar quality of fit to the data, suggesting that the new defect model is applicable, although none of the models uniquely describe the diffraction data. All refined models have similar characteristics: (1) magnesium preferentially occupies the tetrahedral site; (2) aluminum preferentially occupies the octahedral site; and (3) the cation vacancies are distributed over both the tetrahedral and octahedral sites with a slight preference for octahedral coordination. Reoxidation/Reverse Reactions. All of the reduced specimens can be reoxidized to yield single-phase Ni spinels. An example is shown in Figure 10 for Ni0.5Mg0.5Al2O4. The reaction mechanism includes reoxidation of metallic Ni beginning at 300 °C to form NiO and going to completion by 500 °C after 1 h. The NiO does not react with the nonstoichiometric spinel until ∼800 °C, with the Ni spinel nearly fully recovered by 1000 °C. The change in the spinel lattice parameter during the reoxidation reaction shown in Figure 11 is compared to the lattice expansion in flowing 4H2-96N2. Reaction of NiO with the defect spinel over the range of 800 to 1000 °C results in a sharp increase in the spinel lattice parameter, thus recovering the notable spinel lattice contraction observed during the reduction reaction. Thermogravimetric analysis is used to measure the weight of nonstoichiometric spinel Mg0.5Al2O3.5 + Ni (formed by reacting Ni0.5Mg0.5Al2O4 with H2 at 1000 °C for 12 h) during reoxidation in air. The same temperature profile is used as is used in the in situ XRD experiment for the reoxidation of Mg0.5Al2O3.5 + Ni. Both the weight of the sample and the temperature profile are plotted as a function of reaction time in Figure 12. The sample gains a significant amount of weight
13042
J. Phys. Chem. C, Vol. 114, No. 30, 2010
Miller and Misture
Figure 3. XRD data illustrating the transformation of NiAl2O4 to metastable alumina polymorphs as the hydrogen reaction temperature increases from 800 to 1000 °C in 50 °C intervals. The last two scans are the XRD patterns for the reaction products following reaction with H2 at 1000 °C for 2 and 47 h.
Figure 6. Experimentally determined weight loss in the spinel compositions NixMg1-xAl2O4 during reaction with hydrogen at 1000 °C is compared with a theoretical weight loss due to the formation of oxygen vacancies, following defect reaction 4c.
Discussion Figure 4. Normalized fraction of nickel vacancies formed and the weight percent of metallic nickel (errors are smaller than the data points) formed during the hydrogen reaction with Ni0.5Mg0.5Al2O4 as a function of reaction temperature.
Figure 5. TGA data illustrating the weight loss for the Mg0.5Ni0.5Al2O4 spinel sample as a function of temperature. The data is recorded in situ in flowing 4H2-96N2 at 100 mL/min.
between 300 and 500 °C (between ∼100 and 400 min). A minimal amount of weight is gained between 600 and 1000 °C corresponding to an increase in the total weight gained from 4.56% to 4.74%, respectively.
NixMg1-xAl2O4 and Reduction Reactions. The spinel lattice contracts as nickel is substituted for magnesium in NixMg1-xAl2O4. This contraction is related to a change in the degree of inversion as the nickel content increases, where the nickel preferentially occupies the octahedral site, thereby displacing the smaller aluminum ion to the tetrahedral site.3,24 The smaller cation on the tetrahedral site results in a contraction of the oxygen sublattice as is evidenced by a decrease in both the lattice parameter and the oxygen positional parameter.25 The relationship between the degree of inversion and the oxygen positional parameter has been confirmed for the spinels synthesized in this study from the Rietveld refinement of XRD data though it is not reported. The effect of the quench temperature on the spinel lattice parameter is related to the degree of inversion and its temperature dependence.3,25 As the quench temperature increases the cation distribution becomes more random (the degree of inversion decreases). The lattice parameter is inversely related to the degree of inversion; hence, increasing quench temperature results in an increase in the spinel lattice parameter. The NixMg1-xAl2O4 spinels prepared in this work show the expected shift in lattice constant (Figure 1) and oxygen positional parameter with Ni content but all compositions show similar H2 reduction kinetics; the reaction is complete by 850 °C for 24 h for all compositions. We reiterate that the H2 reduction reaction was carried out using powders with average particle size of ∼1 µm, while the initial heat treatments were performed using powder compacts. The ability of the cation
γ-Al2O3: Determination of Spinel Defect Structure
J. Phys. Chem. C, Vol. 114, No. 30, 2010 13043
Figure 9. Experimental densities are compared with the predicted values which are based on various defect equations for alumina-rich spinel.
Figure 7. (a) In situ XRD of Ni0.5Mg0.5Al2O4 in flowing 4H2-96N2 at 800 °C. Successive XRD patterns were recorded in 1 h intervals. (b) Change in lattice parameter as Ni is reduced from Ni0.5Mg0.5Al2O4 during reaction with H2.
MgO · nAl2O3 spinelssynthesizedthroughsolid-statereactions.11,13,23 It is clear that both synthesis methods result in similar if not identical spinels. Reoxidation/Reverse Reactions. The combined results of the in situ XRD and TGA analyses reveal that nickel and oxygen ions are incorporated into the spinel structure simultaneously during the reoxidation of the Mg1-xAl2O4-x spinel + Ni. This behavior mirrors the hydrogen reaction for NixMg1-xAl2O4 spinel where an equal amount of nickel and oxygen vacancies form concurrently. Figures 11 and 12 demonstrate that the weight gain at temperatures