Reaction Interface for Heterogeneous Oxidation of Aluminum Powders

Jun 20, 2013 - Oxidation of differently prepared Al-Mg alloy powders in oxygen. Hongqi Nie , Mirko Schoenitz , Edward L. Dreizin. Journal of Alloys an...
1 downloads 0 Views 4MB Size
Download PDF
Article pubs.acs.org/JPCC

Reaction Interface for Heterogeneous Oxidation of Aluminum Powders Shasha Zhang and Edward L. Dreizin* Otto H. York Department of Chemical, Biological and Pharmaceutical Engineering, New Jersey Institute of Technology, University Heights, Newark, New Jersey 07102, United States S Supporting Information *

ABSTRACT: Heterogeneous oxidation of aluminum is rate limited by diffusion through a growing aluminum oxide layer. If inward diffusion of oxygen ions is faster than outward diffusion of aluminum, the reaction will occur at the inner interface of the oxide. Conversely, the reaction will occur at the outer oxide surface if outward diffusion of aluminum is faster. The location of the heterogeneous reaction is identified processing results of thermogravimetric measurements for two oxidizing spherical aluminum powders with different but overlapping particle size distribution. For each experiment, the measured weight gain is distributed among particles of different sizes assuming that the rate of oxidation is proportional to the reactive interface area. Different models are considered to determine the interface area. For a ductile oxide shell, when there is no void between oxide and aluminum, two cases with the reaction occurring at both inner and outer surfaces of the shell were evaluated. In addition, a case with the reaction at the outer surface of a rigid oxide shell is considered, for which a void inside the particle forms when the aluminum core is shrinking. Oxidation weight gains for the same size particles present in different aluminum powders are expected to be identical to each other when the calculated reactive interface area reflects the true oxidation mechanism. It is concluded that the reaction at the outer surface of a rigid oxide shell describes the experiments most accurately. Thus, the outward diffusion of aluminum ions controls the rate of heterogeneous oxidation of aluminum in a wide range of temperatures of approximately 400−1500 °C. The conclusion is further supported by the electron microscopy of particles quenched at different temperatures, showing oxide surface features consistent with the identified reaction mechanism.

■

molecular dynamics simulation efforts8,12 investigating nanoaluminum particles heated at a very high rate suggest that outward diffusion of aluminum controls the reaction rate at a broad range of temperatures, including the temperatures well above those expected for the amorphous to γ-alumina phase change. In this paper, oxidation is investigated for micrometersized, spherical aluminum particles coated with natural amorphous aluminum oxide and heated at low rates. The location of the reaction interface for aluminum powders is identified based on processing the previously published thermogravimetric (TG) experiments9 on oxidation of two such powders with different but overlapping particle size distributions. TG data analyses are supported by the scanning electron microscopy (SEM) images of samples that are quenched at varied temperatures during oxidation.

INTRODUCTION Aluminum oxidation in oxygen-containing environments leading to its ignition has been studied extensively.1−8 It is established that the rate of oxidation is controlled by the rate of diffusion of reacting species, oxygen and aluminum ions, through a growing oxide scale. The diffusion rate experiences stepwise changes corresponding to polymorphic phase transitions occurring in the oxide,4,5,7,9 and affecting its transport properties. It remains, however, unclear which of the diffusing species, oxygen or aluminum, diffuse faster and thus determine where the reaction is occurring, inside or outside of the growing oxide layer. For oxidation of bare crystalline aluminum surfaces resulting in the growth of thin amorphous oxide scales, it was shown that outward aluminum diffusion is the process controlling the reaction rate.10 Conversely, inward diffusion of oxygen was suggested to control the reaction rate when crystalline γ-alumina polymorphs are forming.10,11 In more recent studies of oxidation of nanosized aluminum particles, it was suggested that the slow oxidation at temperatures below the melting point is dominated by the inward diffusion of oxygen, whereas the formation of hollow nanoparticles above the melting point indicates the significance of outward diffusion of aluminum ions.2 The © 2013 American Chemical Society

■

TECHNICAL APPROACH An assumption exploited for data processing is that the reaction rate for the heterogeneous reactions of interest is proportional Received: March 26, 2013 Revised: June 20, 2013 Published: June 20, 2013 14025

dx.doi.org/10.1021/jp402990v | J. Phys. Chem. C 2013, 117, 14025−14031

The Journal of Physical Chemistry C

Article

A change in the aluminum density upon its melting was also accounted for. For cases I and II, the oxide shell was assumed to be ductile, so that its dimensions were continuously minimized by the surface tension and no voids formed inside the oxidizing particle despite shrinkage of the aluminum core. For case I, Figure 3a, the reaction was assumed to occur inside the growing oxide layer, at the aluminum/oxide interface. For case II, Figure 3b, the reaction occurred at the outer layer of the ductile oxide shell. For case III, Figure 3b, a void was forming between a rigid oxide shell and shrinking aluminum core. The oxide shell was treated as ductile when the effects of thermal expansion of aluminum core or its expansion due to melting were more significant that its shrinkage due to oxidation, a situation observed for larger particles at the very beginning of the experiment. The oxide shell was assumed to become rigid after it reached its maximum radius. It began to separate from the aluminum core when the core radius started decreasing. In case III, the reaction was assumed to occur at the outer surface of the oxide shell, as in case II. A spherical aluminum core is shown in Figure 3c; however, the core shape was not important after a hollow shell formed. The calculations processed individual TG traces. For each time step, the mass of each particle for each size bin and its respective reactive interface area were adjusted considering the current area of reactive interface of this size bin, and based on the total measured mass increase. The size bins and time steps are respectively represented by subscripts i and j. The evolution of the particle mass mi with radius of the reactive interface ri in the time step, j, is calculated as

to the available reactive surface. Normalized experimental TG traces, shown in Figure 1, give the total powder rate of

Figure 1. Oxidation weight gain for Al 3−4.5 and Al 10−14 from TG. The blue and red curves represent aluminum powders with nominal particle sizes of 3−4.5 and 10−14 μm, respectively.

oxidation as a function of temperature. The data are available for two spherical Al powders by Alfa Aesar with nominal particle sizes of 3−4.5 and 10−14 μm. The mass load for this measurement is 1−6 mg. The particle size distributions for both powders are measured (see Figure 2) and used to partition the

mi , j = mi , j − 1 + (m*j − m*j − 1)

ri , j − 12 ∑i (Nri i , j − 12)

(1)

where Ni is the number of particles in the size bin i and m*is the mass of the entire sample. The aluminum core radius, Ri,j, in the size bin i at the time j is calculated as 1/3 ⎡ 2MAl(mi , j − mi ,1) ⎞⎤ 3 ⎛⎜ ⎢ ⎥ ⎟ R i,j = m − ⎢⎣ 4πρAl ⎜⎝ i ,1 MAl 2O3 − 2MAl ⎟⎠⎥⎦

Figure 2. Particle size distribution for aluminum powders with nominal particle sizes of 3−4.5 and 10−14 μm.

(2)

where ρAl is the aluminum density, and MAl and MAl2O3 are molar masses of aluminum and aluminum oxide, respectively. The evolution of the oxide thickness hi for cases I and II, (ductile oxide) is calculated as

mass increase inferred from the TG traces among powder particles in different size bins. No data aside from the TG traces and initial size distribution are used in this processing. Three cases representing different oxidation configurations, as illustrated in Figure 3, were considered to interpret the experimental data. The initial oxide thickness was assumed to be 2.5 nm for all particles. The dimensions of aluminum core and oxide shell were assumed to change due to both oxidation and thermal expansion13−16 occurring during TG experiments.

⎡ ⎤1/3 2 M ( m − m ) Al O i , j i ,1 3 2 3 ⎥ −R hi , j = ⎢R i , j 3 + i,j ⎢⎣ 4πρAl O MAl 2O3 − 2MAl ⎥⎦ 2 3 (3)

The evolution of the oxide thickness hi for cases III, (rigid oxide) is calculated as 1/3 ⎡ 2MAl 2O3(mi , j − mi ,1) ⎤ 3 3 ⎥ hi , j = ⎢R i ,max + ⎢⎣ 4πρAl O MAl 2O3 − 2MAl ⎥⎦ 2 3

− R i ,max

(4)

where ρAl2O3 is the density of aluminum oxide and Ri,max is the maximum core radius for size bin i. Only one aluminum oxide phase is assumed to exist for each time step. According to reference,9 for the heating rate considered here (5 K/min), the

Figure 3. Schematic diagram of oxidation configurations considered. Case I: ductile shell, reaction at the inner Al/Al2O3 interface; case II: ductile shell, reaction at the outer Al2O3 interface; case III: rigid shell, reaction at the outer Al2O3 interface. 14026

dx.doi.org/10.1021/jp402990v | J. Phys. Chem. C 2013, 117, 14025−14031

The Journal of Physical Chemistry C

Article

phase transitions from amorphous to γ-Al2O3 and from γ- to αAl2O3 occur respectively at 660 and 1031 °C. As the calculations continue, the dimensions of the oxidizing particles in all size bins grow; however, the particles originally placed into an individual size bin remain there. In other words, both the number of bins and number of particles per bin remain fixed, whereas the bin sizes are adjusted reflecting the change in the particle size distribution occurring when the powder is oxidizing. To determine which of the models describes experiments better, oxidation of particles with sizes present in both powders (cf. Figures 1 and 2) is considered. It is assumed that particles with the same diameters present in different powders would oxidize similarly to each other. Interactions between oxidizing particles are neglected. Thus, comparing oxidation of the particles from different powders belonging to identical size bins can clarify which of the reaction models is more suitable for describing aluminum oxidation. From Figure 2, it is apparent that particles with sizes in the range of approximately 2−40 μm are present in both powders.

■

Figure 4. Oxidation weight change (a) and oxide thickness (b) for each size bin of Al 3−4.5 by distributing weight gain by the area of the metal−oxide interface.

■

Figure 5. Oxidation weight change (a) and oxide thickness (b) for each size bin of Al 3−4.5 by distributing weight gain by the outer surface area of oxide layer for case II.

EXPERIMENTS TG experiments, with both 3−4.5 μm and 10−14 μm powders, were reproduced in order to recover and examine surface morphology of the partially oxidized particles. The experiments were conducted using a TA Instruments model Q5000IR thermogravimetric analyzer. The balance and the furnace were purged with argon at 10 mL/min and oxygen (purity 99.8%) at 25 mL/min, respectively. The sample was heated at 5 K/min. In preliminary experiments, it was observed that the partially oxidized powder particles adhered to one another so that removing the powder from the TG pan and placing it onto an SEM sample holder was destructive and altered particle surface morphology. To avoid the sample handling prior to its SEM inspection, a monolayer of aluminum particles was deposited on an alumina plate by coating it with a hexane-powder slurry. The plate was dried and placed in the TG sample pan. The weight change in these experiments was too small to be detected directly because of a very small mass of powder in the TG sample pan. However, the heating program reproduced conditions used for experiments shown in Figure 2, and samples were quenched at temperatures of 750, 850, 950, and 1050 °C, corresponding to the powders oxidized after the first and second steps observed in the TG traces. After quenching, the sample plate was coated with carbon by a Baltec MED 020 Sputter Coater by Leica Microsystems. The coated plate was recovered and transferred to the SEM directly, without disturbing the oxidized powder. An SEM, Phenom Tabletop Microscope by FEI Technologies Inc., was used to examine the powder particle surface.

small rapid drop in the oxide thickness at 1031 °C is due to a phase change from γ- to a higher density α-alumina. For the same powder and same size bins, a calculation following case II, in which the reaction was assumed to occur at the outer surface of the ductile oxide shell is illustrated in Figure 5. The difference in the mass gain between fine and

RESULTS Comparison of Different Oxidation Scenarios. Results of a calculation for five selected particle size bins for Al 3−4.5 μm performed following case I, i.e., assuming that reaction occurs on the metal−oxide interface (or on the inner surface of the ductile oxide layer), are shown in Figure 4. The thickness of the oxide layer is nearly uniform for all particle sizes until the finest particles become fully oxidized. Once a fraction of the powder is fully oxidized, the measured mass gain, mj* − mj‑1 * , is redistributed among all remaining partially reacted particles. A

coarse particles is somewhat greater than in Figure 4. The difference in the oxide thickness grown on particles of different sizes is more significant. As fine particles become more and more consumed, their reaction rates increase because the outer surface of the oxide shell (reactive surface area) increases, unlike its shrinking inner surface, considered as the reactive interface in the previous calculation. 14027

dx.doi.org/10.1021/jp402990v | J. Phys. Chem. C 2013, 117, 14025−14031

The Journal of Physical Chemistry C

Article

particle sizes of 3−4.5 and 10−14 μm (see Figure 2 for respective particle size distributions). Because the heating programs in experiments with different powders were identical to each other, it was expected that the mass change as a function of temperature would be the same for the particles with the same initial dimensions. From Figures 7 and 8 it is apparent that the experimental data interpreted assuming that the reaction occurs at the outer surface of the oxide show a much better match between the obtained mass change traces for identical particles belonging to different powders. Apparently, the best match for a 4.7-μm particle is observed from calculation for case II, while the best match for a 14.3-μm particle is found for case III. It is remarkable that the match between the two traces (for both 4.7 and 14.3 μm particle sizes) remains good for the entire range of temperatures considered. A qualitative comparison between different cases illustrated in Figures 7 and 8 was supplemented by a systematic quantitative comparison of calculated oxidation traces for all overlapping particle size bins. Differences between the mass change traces for each size bin were assessed calculating a square root of the square error

Figure 6 shows the calculation results for the same size particles for case III. The radius of the reactive interface for

Figure 6. Oxidation weight change (a) and oxide thickness (b) for each size bin of Al 3−4.5 by distributing weight gain by the outer surface area for case III.

Er =

∑j (Δmj1 − Δmj2)2 J

(5)

where Δmj1 is the percent weight change for sample Al 3−4.5, Δmj2 is the percent weight change for sample Al 10−14, both taken for the jth time step, and J is the total number of time steps in the experiment. The calculated results for three cases are plotted in Figure 9. The model with the smallest value of Er is expected to better describe the oxidation event. It is apparent that both models, assuming that the reaction occurs at the external surface of oxide layer (cases II and III), are better than the model assuming reaction occurring at internal surface of aluminum (case I). Surface Morphology of Partially Oxidized Particles. Figure 10 shows images of undisturbed samples of 3−4.5 μm powder quenched at 750, 850, 950, and 1050 °C. For samples quenched at 750 °C, particle surfaces exhibit concave depressions, likely formed due to the shrinkage of the thermally expanded particle upon their quenching. For particles quenched at higher temperatures, depressions on particle surfaces no longer exist. Instead, particles retain roughly spherical shapes. For particles quenched at 850 °C, surface oxide grain boundaries become noticeable. As the temperature increases to 950 °C, crystallites are observed to grow at the grain

each particle size is greater compared to case II. The difference in mass gain between fine and coarse particles is slightly greater than in case II. Because the inner radius of the oxide shell never shrinks (unlike for the ductile oxide model), and is, in fact, increasing due to the thermal expansion, the thickness of the oxide layer is greatly reduced compared to cases I and II. The rate of thickness growth is the same for all particle sizes. Oxidation of the Same Size Particles from Different Powder Samples. The calculations presented in Figures 4−6 represent the experimental data processing for one of the two spherical aluminum powders with the nominal particle sizes of 3−4.5 μm; similar calculations were performed for the second powder (Al 10−14 μm), and the processed experimental results for both measurements were compared to each other. Specifically, oxidation dynamics for particles of identical sizes present in both powders were considered. For example, such comparisons for particles with diameters 4.7 and 14.3 μm are presented in Figures 7 and 8, respectively. The mass increase for particles in these two size bins is calculated three times, assuming reaction mechanisms I, II, and III, as illustrated in Figure 3. Each individual plot includes two lines, representing the two spherical Al powders used in experiments with nominal

Figure 7. Mass increase for a 4.7 μm diameter Al particle as a function of temperature calculated using the experimental particle size distributions and TG traces obtained for two Al powders. Heating rate is 5 K/min. Results are calculated for cases I (left), II (middle), and III (right). 14028

dx.doi.org/10.1021/jp402990v | J. Phys. Chem. C 2013, 117, 14025−14031

The Journal of Physical Chemistry C

Article

Figure 8. Mass increase for a 14.3 μm diameter Al particle as a function of temperature calculated using the experimental particle size distributions and TG traces obtained for two Al powders. Heating rate is 5 K/min. Results are calculated for cases I (left), II (middle), and III (right).

Figure 11. Images of aluminum particles (10−14 μm) quenched at 950 °C. The sample surface is disturbed while transferring powder from TG pan to SEM sample holder.

Figure 9. Squares of the difference between the weight gain curves calculated for two samples for different size bins (see eq 5).

surface, clearly indicating that the hollow shell has formed during the oxidation, when aluminum melt could flow inside.

■

DISCUSSION Based on comparison of different reaction scenarios, in particular, from Figures 7−9, it can be concluded that the reaction occurs at the outer surface of the growing oxide, and thus, the outward diffusion of aluminum ions is faster than the inward diffusion of oxygen ions during oxidation of aluminum powders. Case III describes oxidation better for all except the two smallest particle size bins considered, indicating that, overall, it offers the best approximation to the actual oxidation scenario. The conclusion regarding superiority of the case III compared to other models and statistical significance of this conclusion are further validated using a detailed statistical analysis presented in Supporting Information. Note that the results for the smallest particle sizes, where cases II and III are close to each other, are most sensitive to the selection of the instant when the oxide shell stops being ductile and becomes rigid. In the present calculations, for simplicity, it was assumed that this transition occurs when the shell reaches its maximum radius. In reality, however, this transition depends on the oxide shell thickness and, likely, its radius. In recent research,17 it was shown that the oxide shells around aluminum nanoparticles become sufficiently strong to cause detectable stress in the thermally expanded aluminum when their thickness reaches about 20 nm. It is likely that larger size oxide shells (forming around larger particles) explored in this work would grow to a greater thickness before they become mechanically rigid. Consider Figure 12, where the evolution of the aluminum core radius and thickness of the growing oxide film are shown for the calculation for case III. For a smaller particle, the oxidation is very fast so that the aluminum core

Figure 10. Images of aluminum particle (3−4.5 μm) quenched at 750 (a), 850 (b), 950 (c), and 1050 °C (d). Note: the sample surface is intact after oxidation and quenching.

boundaries. The crystallites become prominent, for the particles heated to 1050 °C. In addition to the undisturbed particles, particles broken into several pieces as a result of handling the heated sample were examined. Particles with internal voids were observed for both quenched 3−4.5 μm and 10−14 μm powders heated above 850 °C. An image of the broken aluminum particles (Al 10−14 μm) quenched from 950 °C is shown as an example in Figure 11. It appears that aluminum particles oxidize forming a hollow oxide shell partially filled with molten aluminum. The solidified aluminum melt is observed to spread over the inner shell 14029

dx.doi.org/10.1021/jp402990v | J. Phys. Chem. C 2013, 117, 14025−14031

The Journal of Physical Chemistry C

Article

of reacting species through a growing product layer. The method is applied to study oxidation of aluminum powders by processing respective thermo-gravimetric measurements for two spherical aluminum powders with different but overlapping particle size distributions. It is found that, for the aluminum− oxygen reaction, the oxidation occurs at the outer surface of the growing rigid oxide layer and thus is controlled by the outward diffusion of aluminum ions for a broad range of temperatures from 400 to 1500 °C. The growing oxide shell becomes rigid once its thickness reaches approximately 100 nm. Oxide shells reaching this thickness do not shrink together with the consumed aluminum core, so that hollow particles are formed as a result of oxidation. The interpretations of TG experiments are consistent with the surface morphologies of the partially oxidized aluminum particles recovered at different temperatures and examined using SEM.

Figure 12. Calculated radius of aluminum core as a function of temperature for two particle sizes.

■

ASSOCIATED CONTENT

S Supporting Information *

does not expand. Based on eq 4 for case III, the oxide shell was always treated as rigid. However, if thin oxide shells are ductile, case II should describe the finest particles better, consistently with slightly better descriptions implied by Figures 7 and 9. For a larger particle in Figure 12, the maximum size of the aluminum core is reached when it expands due to melting, after which it remains nearly constant, until the shrinkage due to consumption of aluminum exceeds the effect of thermal expansion at about 850 °C. If such a particle is quenched from 750 °C, it will solidify so that the Al core will shrink noticeably. A ductile oxide shell expanded together with the Al core may collapse, forming depressions, as seen in Figure 10. The oxide thickness expected to form at these temperatures is just under 100 nm, indicating that even at this thickness, the shell may remain ductile. When heated to greater temperatures, the shell thickness exceeds 100 nm, and it is expected to become rigid. Thus, when the aluminum core shrinks further during its consumption or even upon quenching and solidification, the 100-nm thick or thicker oxide shell may preserve its spherical shape. This is consistent with the shape of particles quenched at higher temperatures (cf. Figure 10). It is also consistent with the shapes of broken spherical oxide shells observed in Figure 11. The shells contain what appears to be molten and solidified pool of aluminum; however, the shapes of the shells remain spherical. The thickness of the observed broken shells is, indeed, close to 100 nm, further supporting the proposed reaction scenario. Formation of hollow shells, qualitatively similar to results with nanoaluminum,2 further indicates a significant outward diffusion of aluminum. Unlike results,2 however, no aluminum is found outside of the oxide shells, indicating that aluminum was not removed by flowing through the cracks in the oxide. The crystallites observed to grow at the aluminum oxide grain boundaries at high temperatures (Figure 10) are likely to indicate a well-known fact that, once the oxide layer becomes crystalline, the grain boundary diffusion is much faster than its intercrystalline counterpart. The preserved and developed structures of the externally growing crystallites at increasing temperatures support the conclusion that their growth is controlled by the outward diffusion of aluminum, rather than inward diffusion of oxygen.

Detailed statistical analysis comparing different oxidation models. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: (973) 596-5751. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by Defense Threat Reduction Agency.

■

REFERENCES

(1) Mench, M. M.; Kuo, K. K.; Yeh, C. L.; Lu, Y. C. Comparison of Thermal Behavior of Regular and Ultra-fine Aluminum Powders (Alex) Made from Plasma Explosion Process. Combust. Sci. Technol. 1998, 135, 269−292. (2) Rai, A.; Park, K.; Zhou, L.; Zachariah, M. R. Understanding the Mechanism of Aluminium Nanoparticle Oxidation. Combust. Theory Modell. 2006, 10, 843−859. (3) Trunov, M. A.; Schoenitz, M.; Dreizin, E. L. Ignition of Aluminum Powders under Different Experimental Conditions. Propellants, Explos., Pyrotech. 2005, 40, 36−43. (4) Trunov, M. A.; Schoenitz, M.; Zhu, X.; Dreizin, E. L. Effect of Polymorphic Phase Transformations In Al2O3 Film on Oxidation Kinetics of Aluminum Powders. Combust. Flame 2005, 140, 310−318. (5) Trunov, M. A.; Umbrajkar, S. M.; Schoenitz, M.; Mang, J. T.; Dreizin, E. L. Oxidation and Melting of Aluminum Nanopowders. J. Phys. Chem. B 2006, 110, 13094−13099. (6) Watson, K. W.; Pantoya, M. L.; Levitas, V. I. Fast Reactions with Nano- and Micrometer Aluminum: A Study on Oxidation Versus Fluorination. Combust. Flame 2008, 155, 619−634. (7) Schoenitz, M.; Patel, B.; Agboh, O.; Dreizin, E. L. Oxidation of Aluminum Powders at High Heating Rates. Thermochim. Acta 2010, 507−508, 115−122. (8) Henz, B. J.; Hawa, T.; Zachariah, M. R. On the Role of Built-in Electric Fields on the Ignition of Oxide Coated Nanoaluminum: Ion Mobility Versus Fickian Diffusion. J. Appl. Phys. 2010, 107, 024901. (9) Trunov, M. A.; Schoenitz, M.; Dreizin, E. L. Effect of Polymorphic Phase Transformations in Alumina Layer on Ignition of Aluminum Particles. Combust. Theory Modell. 2006, 10, 603−624. (10) Jeurgens, L. P. H.; Sloof, W. G.; Tichelaar, F. D.; Mittemeijer, E. J. Growth Kinetics and Mechanisms of Aluminum-oxide Films Formed

■

CONCLUSIONS A new data processing method is established to identify the location of the heterogeneous reaction controlled by diffusion 14030

dx.doi.org/10.1021/jp402990v | J. Phys. Chem. C 2013, 117, 14025−14031

The Journal of Physical Chemistry C

Article

by Thermal Oxidation of Aluminum. J. Appl. Phys. 2002, 92, 1649− 1656. (11) Jeurgens, L. P. H.; Sloof, W. G.; Tichelaar, F. D.; Mittemeijer, E. J. Structure and Morphology of Aluminium-oxide Films Formed by Thermal Oxidation of Aluminium. Thin Solid Films 2002, 418, 89− 101. (12) Henz, B. J.; Hawa, T.; Zachariah, M. R. Molecular Dynamics Simulation of the Kinetic Sintering of Ni and Al Nanoparticles. Mol. Simul. 2009, 35, 804−81. (13) Wilson, A. J. C. The Thermal Expansion of Aluminum from 0 to 650 °C. Proc. Phys. Soc. 1941, 9, 235−244. (14) Drotning, W. D., Thermal Expansion and Density Measurements of Molten and Solid Materials at High Temperatures by the Gamma Attenuation Technique. Master Dissertation, Sandia Laboratories, 1979. (15) Sarikaya, O. Effect of the Substrate Temperature on Properties of Plasma Sprayed Al2O3 Coatings. Mater. Des. 2005, 26, 53−57. (16) Munro, M. Evaluated Material Properties for a Sintered αAlumina. J. Am. Ceram. Soc. 1997, 80, 1919−1928. (17) Rufino, B.; Coulet, M. V.; Bouchet, R.; Isnard, O.; Denoyel, R. Structural Changes and Thermal Properties of Aluminium Micro- and Nano-powders. Acta Mater. 2010, 58, 4224−4232.

14031

dx.doi.org/10.1021/jp402990v | J. Phys. Chem. C 2013, 117, 14025−14031