- Email: [email protected]
Oxidation of nano-sized aluminum powders
Accepted Manuscript Title: Oxidation of nano-sized aluminum powders Author: A.B. Vorozhtsov M. Lerner N. Rodkevich H. Nie A. Abraham M. Schoenitz E.L. Dreizin PII: DOI: Reference:
S0040-6031(16)30112-5 http://dx.doi.org/doi:10.1016/j.tca.2016.05.003 TCA 77509
To appear in:
Thermochimica Acta
Received date: Revised date: Accepted date:
13-2-2016 5-5-2016 7-5-2016
Please cite this article as: A.B.Vorozhtsov, M.Lerner, N.Rodkevich, H.Nie, A.Abraham, M.Schoenitz, E.L.Dreizin, Oxidation of nano-sized aluminum powders, Thermochimica Acta http://dx.doi.org/10.1016/j.tca.2016.05.003 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Oxidation of nano‐sized aluminum powders A.B. Vorozhtsov1, M. Lerner2, N. Rodkevich2, H. Nie3, A. Abraham3, M. Schoenitz3, E.L. Dreizin1,3 1
Tomsk State University, Tomsk, Russia
2
Institute of Strength Physics and Material Science, Siberian Branch, Russian Academy of Sciences, Tomsk, Russia
3
New Jersey Institute of Technology, Newark, NJ, USA
Corresponding author: Edward L. Dreizin, [email protected], 973‐596‐5751 Graphical abstract
1
Highlights 1. 2. 3. 4.
Weight gain measured in TG oxidation experiments was split between particles of different sizes Reaction kinetics obtained by isoconversion explicitly accounting for the effect of size distribution Activation energy is obtained as a function of oxide thickness for growth of amorphous alumina Oxidation mechanism for nanopowders remains the same as for coarser aluminum powders
Abstract Oxidation of aluminum nanopowders obtained by electro‐exploded wires is studied. Particle size distributions are obtained from transmission electron microscopy (TEM) images. Thermo‐gravimetric (TG) experiments are complemented by TEM and XRD studies of partially oxidized particles. Qualitatively, oxidation follows the mechanism developed for coarser aluminum powder and resulting in formation of hollow oxide shells. Sintering of particles is also observed. The TG results are processed to account explicitly for the particle size distribution and spherical shapes, so that oxidation of particles of different sizes is characterized. The apparent activation energy is obtained as a function of the reaction progress using model‐free isoconversion processing of experimental data. A complete phenomenological oxidation model is then proposed assuming a spherically symmetric geometry. The oxidation kinetics of aluminum powder is shown to be unaffected by particle sizes reduced down to tens of nm. The apparent activation energy describing growth of amorphous alumina is increasing at the very early stages of oxidation. The higher activation energy is likely associated with an increasing homogeneity in the growing amorphous oxide layer, initially containing multiple defects and imperfections. The trends describing changes in both activation energy and pre‐exponent of the growing amorphous oxide are useful for predicting ignition delays of aluminum particles. The kinetic trends describing activation energies and pre‐exponents in a broader range of the oxide thicknesses are useful for prediction of aging behavior of aluminum powders. Keywords: heterogeneous reaction, aluminum ignition, nano‐aluminum, aging, thermal analysis Introduction Development of practical methods to prepare nano‐sized aluminum powders 1 catalyzed research in new nano‐energetic materials, including metal nanopowders, nano‐thermites, and intermetallic systems 2‐4 . Although many compositions have been prepared and characterized in the last two decades, nano‐ sized aluminum powder remains the main component for multiple energetic formulations of interest for propellants, explosives, and pyrotechnics 5‐7. For most such formulations, oxidation of nano‐sized aluminum is the process affecting critically both aging and ignition phenomena. This explains an active and sustaining interest in understanding and quantitative characterization of reactions governing oxidation of nano‐sized aluminum, e.g., see 8‐14. In particular, oxidation similarities and differences between nano‐ and micron‐sized powders need to be established. For example, widely discussed changes in the melting characteristics of nano‐aluminum 15‐16 become significant for particles smaller than 10 nm and may not play an important role for most practical nano‐powders with dimensions in the
2
100 nm range. However, it remains unclear whether mechanisms and kinetics developed for oxidation of coarse aluminum powders are directly transferrable for nano‐sized particles. Oxidation mechanism of micron‐sized aluminum powders has been established relatively well 4, 17‐18. It is generally agreed that the reaction is diffusion controlled. It is also agreed that the diffusion rates are heavily affected by the structure of the growing aluminum oxide film. The natural 2 – 4‐nm thick oxide is amorphous; it transitions to ‐Al2O3 at elevated temperatures and growing oxide layer thickness 19‐20. Further oxidation processes involve transformation of ‐Al2O3 to ‐ and ‐Al2O3 as temperatures increase. It has been shown recently that the reaction occurs at the interface separating the growing alumina layer and gaseous oxidizer, and thus it is rate controlled by outward diffusion of aluminum ions 21‐22 . Detailed thermo‐analytical measurements were reported and processed to establish kinetics of the respective oxidation processes 23‐24. Reactions at early oxidation stages were addressed in particular, as most critical for modeling both ignition and aging phenomena 21. A recent study 22 addressing oxidation of nano‐sized aluminum mostly agrees with the above oxidation sequence. However, quantitative comparisons of oxidation rates and respective reaction kinetics are not available. This work is aimed to characterize oxidation of aluminum nano‐powders quantitatively. It is further desired to determine the kinetics of respective reactions and compare it directly to that reported for micron‐sized aluminum powders. Experimental details Spherical aluminum nanopowders were prepared in argon using electro‐exploded wires 16, 25. The powder was held at a reduced oxygen pressure for 72 hours for passivation. The amount of active aluminum in the as‐prepared powders was determined by reacting the powder with an aqueous solution of sodium hydroxide and measuring the volume of the released hydrogen gas1. Oxidation of spherical aluminum nanopowders is characterized in thermo‐analytical experiments, including differential thermal analysis (DTA) and thermo‐gravimetry (TG). For the purpose of obtaining partially oxidized materials, samples were heated in air at 10 K/min to 823 K, 923 K, and 1183 K. At the target temperature, the samples were allowed to cool, and recovered for analysis. Residual oxidation after reaching the target temperature has not been measured, but is assumed to be negligible. These experiments used a Netzsch STA 449F3 instrument. Additional TG measurements for the purpose of examining oxidation kinetics were performed using a TA Instruments TA Q5000‐IR instrument. For these experiments, heating rates varied from 2 to 10 K/min, and the atmosphere was pure oxygen. As‐prepared as well as partially oxidized particles were examined using a JEOL JEM‐100 CXII TEM. Particle size distributions were determined by image analysis, and are based on a count of approximately 4,600 particles. In addition, XRD (XRD‐6000 by Shimadzu using Cu K radiation) was used to determine phases formed during oxidation. The oxide thickness on partially oxidized samples was determined from multiple TEM images using ImageJ software26. Five measurements on each particle were averaged. Experimental results
3
TEM images of as prepared nanoparticles are shown in Fig. 1. The particles are mostly spherical. Particle sizes range from about 20 to 400 nm; a detailed particle size distribution obtained by processing multiple TEM images is shown in Fig. 2. Results of TG measurements are illustrated in Fig. 3. A small initial mass loss is observed. The measurements are consistent with earlier reports on oxidation of micron‐ and nano‐sized aluminum powders 2, 19, 27. A rapid acceleration of the oxidation rate is observed around 800 K, prior to aluminum melting, when amorphous oxide transforms into cubic spinel type alumina, typically ‐Al2O3. A DTA trace, omitted here for brevity, shows an exothermic peak corresponding to the accelerated mass increase, followed by a small endothermic peak at the Al melting point. At higher temperatures, reaction rate accelerates further, again in agreement with earlier reports 14, 16, 19 assigning that acceleration to thermally activated growth of transition alumina. The TG traces shift to higher temperatures at greater heating rates, as is expected for a thermally activated reaction. A TG curve for the oxidation in air, representative for the recovery of partially oxidized samples is shown in Fig. 3 for comparison. XRD patterns are shown in Fig. 4. The as‐prepared passivated powder contains no detectable impurities. Samples partially oxidized in air were recovered from different temperatures. Consistent with earlier results for coarser aluminum14, 16, there are no well resolved peaks of alumina for the sample recovered from 823 K. Thus, only amorphous oxide is expected to be formed at lower temperatures. This amorphous oxide may be partially hydrated, explaining the small initial mass loss observed by TG (see Fig. 3). At 923 K, alumina is present as the (JCPDS 01‐083‐2080) and polymorphs (JCPDS 98‐006‐ 9213). Additional peaks, marked by asterisks in Fig. 4 are consistent with AlN (JCPDS 04‐007‐4280). However plausible this appears for aluminum heated in air, based on the diffraction pattern alone a partially reduced oxide with composition Al3O4 (JCPDS 04‐016‐0539) cannot be ruled out. The presence of cubic, spinel‐type alumina is consistent with previous observation of alumina at this stage of oxidation. At 1183 K, alumina remains present. Fewer peaks of the spinel type alumina are observed, consistent with alumina (JCPDS 00‐056‐0458). The XRD patterns generally confirm that oxidation of nano‐sized aluminum particles results in the formation of the same sequence of alumina polymorphs as reported to form upon oxidation of coarser aluminum powders19‐22. TEM images of partially oxidized particles are shown in Fig. 5. At 823 K, the particles largely retain their spherical shapes. Some particles show apparent hollow features at the surface, as confirmed by negligible amounts of Al or O measured by EDX. The structure of particles recovered from 923 K changes substantially. All particles appear to contain a relatively thick oxide layer and a cavity. The presence of cavities is supported by mapping aluminum concentration using energy‐dispersive x‐ray spectroscopy (line scans shown in Fig. 5 as well). The cavity formation occurring on a time scale of minutes in the current TG experiments is further consistent with cavities observed in previous 4
research13,16, whereas in experiments13 individual Al nanoparticles were partially oxidized for durations on the order of a second at higher temperatures. Although particles mostly retain their spherical shapes, most particles are now fused together. Both oxide layer thickness and cavity size grow further for particles recovered from 1183 K. Technical Approach In order to establish a kinetic description of the oxidation process, the TG measurements in oxygen shown in Fig. 3 were processed accounting explicitly for the particle size distribution shown in Fig. 2. This allowed to distribute the observed total sample mass over the particles in all size bins, and to recover the oxide layer thickness for each particle size. The geometry of the oxidation model is shown in Fig. 6. It is based on earlier work on micron‐sized aluminum powders28. The location of the reactive interface on the outer particle surface implies that the oxide growth rate is limited by the outward diffusion of oxygen. As the aluminum is consumed, a void grows in the aluminum particle core, and aluminum is assumed to adhere to the inner surface of the growing oxide shell. However, location and shape of the void are not essential to the current calculation. The void could be located off‐center; it need not be spherical, and indeed may be present as multiple smaller voids. The model only accounts for the net volume of the void. The validity of this approach diminishes as the oxidizing particles sinter and/or change shape. However, this analysis is useful in the early stages of oxidation, which are particularly important for both ignition and aging processes. To recover the oxide layer thickness for each particle size separately, the mass increment measured at a given time for the entire powder, dmpowder, is taken as the sum of the mass increments of the particles in all size bins:
dm powder ni dmi
(1)
i
where ni is the number fraction of particles in the i-th size bin, and dmi is the corresponding mass increment of a particle in that size bin. In previous work on micron‐sized aluminum24, 28, a particle’s mass increment was taken as proportional to the instantaneous particle surface area. This was justified for early oxidation stages of large particles since the layer thickness on particles of different sizes was not significantly different. However, this assumption does not hold for nanosized particles, and was therefore relaxed in the current research. Instead, we use the solution of the diffusion equation for a spherical particle with a growing oxide shell, written assuming that pure Al and pure O2 gas are on the inner and outer sides of the oxide layer, respectively, and dropping the subscript i for brevity: 1
1 1 rr dm Al 4 Al D dt 4 Al D 1 2 dt r2 r1 r1 r2
(2)
5
where dmAl is the mass change of aluminum; D is the diffusion coefficient (generally following Arrhenius kinetics), Al is the Al density (taken as a function of temperature), and r1 and r2, are outer and inner radii of the oxide shell, respectively (see Fig. 6). The mass change of the particle is then
dm dmAl dmox ,
(3)
and the mass increments of oxide (taken as Al2O3) and Al metal are related via the reaction
stoichiometry, dmox dmAl M Al2O3
2M Al where M are the respective molar masses. Thus, the
mass increment for a particle in the i‐th size bin is proportional to the factor ai r1r2 r2 r1 which i is equivalent to the surface area at the geometric mean radius divided by the oxide layer thickness. The mass increment of the powder dmpowder is now distributed over all particles according to the distribution of ai over the whole size distribution:
ni dmi
ni ai dm powder ni ai
(4)
i
The inner and outer radii, used to calculate the factors a, are calculated from the total oxide mass of a particle, and the particle geometry:
3 mox t r2 r2,0 ox ,0 , r1 r23 ox T 4 ox T 1/3
1/3
(5)
while accounting for thermal expansion via the temperature dependent oxide density ox. Equations (5) specify that new oxide is accumulated at the outer particle surface (at r1), while the Al metal – Al oxide interface (at r2) remains constant save for thermal expansion. Equations 2‐5 are solved for each time step. The powder mass increment for each time step is determined by the TG data acquisition rate and is taken as the finite observed mass changes mpowder:
mi ,t mi ,t t
ai ,t
n a
m powder ,t
(6)
i i ,t
i
The subscripts t and t-t indicate values at the beginning and end of the current time step in the TG data. The summation starts at the beginning of the TG measurement with a particle mass mi ,t 0 corresponding to an Al particle with a 2.5‐nm oxide layer. This method for distributing the observed powder mass in a TG measurement over the experimental particle size distribution was originally developed elsewhere23‐24, and expanded for small particles in the present work. Ultimately, TG traces representing oxidation of particles of each given size bin are obtained for different heating rates. This made it possible to perform kinetic analysis using a model‐free isoconversion technique23‐24 to obtain activation energy as a function of the reaction progress defined for individual particle sizes. The processing explicitly accounts for the effect of particle size distribution on the TG measurements. Once the activation energies are obtained, the reaction model represented by Fig. 6 is 6
considered to identify the pre‐exponent and thus to fully quantify the rate of oxidation. The obtained reaction kinetics for aluminum nanopowders is finally compared to that reported earlier for coarser aluminum particles. Results of data processing Distributing the mass change of the whole powder (see Fig. 3) over the particle sizes shown in the particle size distributions in Fig. 3, allows to recover how the oxide thickness increases with temperature for particles of different sizes. The oxide thickness as a function of temperature is shown in Fig. 7. Finest, 10‐nm particles are predicted to be fully oxidized at temperatures close to 730 K. The thickness of the grown oxide layer is close to one half of the initial particle diameter. Larger particles are fully oxidized at increasingly higher temperatures. Note that the oxide layer is predicted to grow at the outer surface of the existing shell; thus the size of the particle continuously increases, while a cavity is growing inside. In addition, the thermal expansion is taken into account for both metal core and the oxide shell. Because the particle diameter increases as a function of temperature, the thickness of the oxide layer for the fully oxidized particles from larger bin sizes become substantially smaller than half of the initial particle diameter. A sudden reduction in the oxide thickness observed at 770 K represents re‐adjustment of the calculated thickness as a result of polymorphic phase transition from amorphous to a higher density ‐Al2O3. Predicted oxide thickness as shown in Fig. 7 was compared to that measured from the TEM images acquired for partially oxidized particles recovered from different temperatures. Diameters of particles observed in TEM images were compared to the diameters expected to be produced at the respective temperatures for the partially oxidized particles belonging to different particle size bins. The bin, for which the diameters matched was then selected for each image, so that the thicknesses of the calculated and observed in TEM oxide layers could be compared to each other directly. Results of such comparisons are shown in Fig. 8. The match between the TEM and processed TG data is excellent, supporting the validity of the adopted oxidation model used to convert the measured mass increase into the oxide thickness. Oxidation kinetics for nano‐sized aluminum particles Thickness of the grown oxide layer was considered as a convenient indicator of the reaction progress. Unlike mass increase, oxide thickness can be used to gauge the degree of oxidation for particles of different sizes. Model‐free isoconversion processing 29 was applied to TG curves assigned to individual particle size bins. Initial oxide thickness was assumed to be 2.5 nm 30. The processing was performed separately for the portions of the curves describing growth of amorphous alumina, at temperatures below 770 K, and growth of transition alumina polymorphs, principally, ‐Al2O3, above that temperature. The temperature, at which this transition occurs was fixed for simplicity in calculations using the oxidation model shown in Fig. 1 and transforming the measured TG trace into individual traces for different size bins. A more realistic model should account for both temperature and thickness of the grown oxide layer 19, 31‐33 to describe this transition with a greater accuracy. Apparent activation energy as a function of the oxide thickness for growth of amorphous oxide, Eamorph, is shown in Fig. 9 for several individual particle size bins. The results for different size bins are consistent 7
among themselves and show a rapid increase in the activation energy followed by its stabilization at about 155 kJ/mol. The change in the activation energy, in kJ/mol is well described by the following curve‐fitting function:
154.2
. ∙ .
(4)
where h is oxide thickness in nm. The above expression is only valid for oxide thicknesses exceeding 2.5 nm, which is the thickness assumed to represent the initial natural oxide layer. Once the activation energy, E, as a function of reaction progress is identified, an explicit oxidation model is used to determine the respective value of the pre‐exponent. Considering oxidation of a spherical particle with an oxide shell, as given by Eq. (2), and assuming that
∙
, the values of
pre‐exponent characterizing growth of amorphous alumina, Camorph, were obtained as shown in Fig. 10. The pre‐exponent was separately obtained for different heating rates and for different particle size bins. For clarity, shown in Fig. 10 are only the pre‐exponents obtained for two different size bins. The results were very consistent among themselves showing a trend that closely reflects the changes in the activation energy. As expected, the pre‐exponent changes in a much broader range of values than activation energy. Its change (in ) can be described as a function of the oxide thickness using a curve‐ ∙
fitting expression analogous to that used for the activation energy:
5.01
.
∙ .
(5)
Similarly to Eq. (4), Eq. (5) can be used for the oxide thicknesses exceeding 2.5 nm. Results of isoconversion processing of the TG measurements temperatures above 770 K, describing growth of transition alumina polymorphs are illustrated in Fig. 11. Apparent activation energy is calculated for thicknesses above ca. 5.4 nm while the calculated values in Fig. 9, for amorphous oxide, end around 3.5 nm. The gap between 3.5 and 5.4 nm is not described because the polymorphic phase transition was assumed to occur at a fixed temperature of 770 K. At different heating rates, different thickness of the amorphous oxide is predicted to grow by 770 K, varied from about 3.5 to 5.4 nm in present experiments. Respectively, the isoconversion processing, which uses data for all heating rates simultaneously, is limited to the data range with the same range of variation of the oxide thickness, selected as the reaction progress indicator. The values of the apparent activation energies describing growth of ‐Al2O3 (Fig. 11) obtained from processing TG data corresponding to different particle size bins are consistent with one another. The apparent activation energy increases from about 180 to approximately 230 kJ/mol as the oxide thickness grows up to ca. 12 nm. The activation energy then decreases to approximately 150 kJ/mol when oxide thickness reaches ca. 20 nm. Strong variations in the activation energy are observed at thicknesses around 22 – 24 nm, which occur near the aluminum melting point for the heating rates used in the TG measurements. Following the melting point, the activation energy is observed to increase, reaching
8
very high values at the terminal thickness of about 60 nm. It should be emphasized that Al melting has no causal relation with the decrease in activation energy. It is used here only as a reference point. Using the apparent activation energy as shown in Fig. 11, and accounting for the specific oxidation model (Fig. 1), the pre‐exponent values characterizing growth of transition alumina can also be obtained, as shown in Fig. 12. The pre‐exponent was only calculated for oxidation up to about 20 nm, e.g., before the aluminum melting occurring in TG experiments. A rapid increase in the apparent activation energy following melting is probably associated with substantial changes in the morphology of the oxidizing sample as a result of particle agglomeration, e.g., loss in the surface area available for reaction. This increase is unlikely to characterize heterogeneous oxidation reactions of interest in this study. For the pre‐exponent, the results are somewhat different when different particle size bins and different heating rates are used. A possible reason for this discrepancy is that in the present analysis, the polymorphic phase transition from amorphous to ‐Al2O3 was assumed to occur at a fixed temperature for simplicity. In reality, this transition is affected by both temperature and oxide thickness; it occurs at different temperatures for different oxide thicknesses, depending on the heating rate 33‐35. Neglecting the effect of oxide thickness would shift the starting point for the present analysis compared to the actual formation of ‐Al2O3. The shift would be different for different particle size bins and for different heating rates. Discussion In this section, results obtained here for oxidation for nano‐sized aluminum powders are compared to those reported for micron‐sized aluminum earlier 34. Possible reasons for the observed differences in the reaction kinetics for powders with different particle sizes are also discussed. Apparent activation energy characterizing oxidation of aluminum is shown in Fig. 13 as a function of the oxide thickness, where results for amorphous and transition alumina polymorphs shown in Figs. 9 and 11 are combined with the activation energy obtained for a coarser aluminum powder 35. The inset in Fig. 13 shows a close‐up to the range of small oxide thicknesses. Thin solid lines represent piecewise straight line fits proposed to describe the activation energy as a function of thickness in Ref. 25. It is remarkable that for both nano‐sized and micron‐sized powders, apparent activation energy describing growth of amorphous alumina is shown to increase substantially for very small thicknesses and at low temperatures. The straight line segments proposed to describe the activation energy as a function of the oxide layer thickness for coarse aluminum for practical applications offer a reasonable approximation for the present data. Using Eq. (4) instead is advisable for greater accuracy of predicted oxidation rate; however, the difference between the values of activation energy is minor. A fairly sharp increase in the apparent activation energy observed for thin amorphous alumina layers may be attributed to healing imperfections and defects present in natural alumina and described in earlier research using high‐resolution TEM images 25, 32. 9
The trend for the apparent activation energy describing growth of ‐Al2O3 diverges slightly from the trend implied by the current data for growth of the amorphous oxide. This small difference is not apparent from the data obtained for coarser powder in ref. 36, where the entire range of oxide thicknesses was processed, without subdividing it into portions related to amorphous and crystalline alumina polymorphs. Generally, there is good agreement between trends describing the apparent activation energy as a function of thickness of ‐Al2O3 for both nano‐ and micron‐sized powders for the oxide thicknesses up to about 13 nm. The apparent activation energies diverge somewhat between different types of powders for greater alumina thicknesses, most likely as a result of changes in the powder morphologies, primarily, sintering between oxidizing particles. The initial oxidation is expected to be better described using Eq. (5) for the pre‐exponent, while the piecewise straight line fits proposed in Ref. 31 for coarser aluminum are also expected to predict the oxidation rate with an acceptable accuracy. Comparisons of pre‐exponent values for different powders are illustrated in Fig. 14. The inset shows a close‐up to the range of small oxide thicknesses. As in Fig. 13, thin solid lines represent piecewise straight line fits proposed to describe the pre‐exponent as a function of thickness in Ref. 25. There are small but appreciable differences between the straight line fragments proposed to describe the pre‐ exponent as a function of the oxide thickness for coarse aluminum and the trends describing the pre‐ exponent for both amorphous and ‐Al2O3 based on the present measurements. The initial oxidation is expected to be better described using Eq. (5) for the pre‐exponent, using the present measurements, where the initial oxidation is resolved better. Description of growth of ‐Al2O3 is a bit more difficult to quantify precisely because of an approximation used here to determine the instant of the polymorphic phase change leading to its formation. Initial reactions starting from thin, natural oxide film are particularly important for descriptions of both ignition and aging of aluminum and related reactive materials. Therefore, oxidation kinetics identified in Ref. 25 and validated here, with slightly amended expressions for activation energy and pre‐exponent (Eqs. 4, 5) should be used in the respective reaction models. Aluminum ignition is most likely occurring soon after the polymorphic phase change transforming the natural amorphous alumina into a higher density ‐Al2O3 phase. The diffusion resistance of the oxide layer diminishes, reaction rate accelerates, causing ignition 19, 25. Predicting when the polymorphic transition between amorphous and ‐Al2O3 occur can thus be used to predict the ignition delay for an aluminum particle. Because the phase change depends on both temperature and the oxide thickness, calculating an increasing oxide thickness during the initial pre‐ignition particle heating is important. Equations (4) and (5) obtained here as fits to the experimentally inferred apparent activation energy and respective pre‐exponent can be used to perform such calculations for any aluminum particle heating scenarios. For example, to account for transitions between alumina polymorphs, the following equation was proposed in Ref. 20:
10
1
(6)
where F, K, and E are parameters describing a specific oxide polymorph. It is important that the rate of growth is affected by both temperature, as in usual thermally activated oxidation, and by the oxide thickness. Thus, to accurately predict when the rate of growth of ‐Al2O3 becomes significant (i.e., greater than the growth of amorphous alumina), one needs to simultaneously track both temperature and thickness of the oxide layer. Eqs. (4) and (5) enable one to track the thickness of the growing amorphous alumina, and thus enable one to predict when ‐Al2O3 will start to grow, or when the particle is expected to ignite. A main shortcoming of the present approach is an approximate nature of Eq. (6), although it properly accounts for the effects of both temperature and oxide thickness on structure of the growing aluminum oxide. For aging analysis, the effect of polymorphic phase change on the thickness of growing aluminum oxide layer can be neglected; thus the straight line fits proposed in Ref. 25 are useful, which also describe the present results reasonably well. Conversely, fitting lines connecting trends for activation energies and pre‐exponents for both amorphous and ‐Al2O3 using the present data can be obtained and used to predict the rate of aluminum oxidation in slow aging processes. Conclusions Thickness of the oxide layer growing on nano‐sized aluminum powders exposed to elevated temperatures in oxygen‐containing environments was obtained from both TEM images and processing of respective TG traces. The TG data processing accounted for the specific particle size distribution. The measured weight gain was split among different size bins of the powder, accounting for particle sizes and spherical shapes, and taking into consideration that the reaction occurs at the outer interface of the growing oxide layer. The oxide layers observed in TEM images correlated well with those implied at different temperatures by the TG measurements. Oxidation kinetics was characterized using isoconversion processing of the TG data. Separate trends for kinetic activation energies and pre‐exponents as a function of the oxide thickness were obtained to describe growth of amorphous and ‐Al2O3 polymorphs. The present results are in good agreement with the earlier studies characterizing kinetics of oxidation of coarser aluminum powders. The present results confirm that the mechanism of aluminum oxidation does not change as a function of particle size down to tens of nm. An increase in the apparent activation energy describing growth of amorphous alumina implied by earlier work is confirmed here. The higher activation energy is likely associated with an increasing homogeneity in the growing amorphous oxide layer, initially containing multiple defects and imperfections. The trends obtained here and describing changes in both activation energy and pre‐ exponent of the growing amorphous oxide are useful for prediction of ignition delays of aluminum particles. The ignition is expected to occur when ‐Al2O3 forms disrupting continuity of the oxide layer covering aluminum surface. Formation of ‐Al2O3 is affected by both temperature and thickness of the oxide layer. The thickness of the oxide grown during pre‐ignition processes can be predicted using the kinetics of growth of amorphous alumina obtained here. The kinetic trends describing activation energies and pre‐exponents in a broader range of the oxide thicknesses are useful for prediction of aging behavior of aluminum powders. 11
Acknowledgements Research at NJIT was supported by Defense Threat Reduction Agency, Grant HDTRA1‐11‐1‐0060. The work at Tomsk State University was financially supported by the Ministry of Education and Science of the Russian Federation within the framework of the Federal Target Program; Agreement No. 14.587.21.0019 (Unique identifier RFMEFI58715X0019).
12
References 1. Tepper, F., Electro‐Explosion of Wire Produces Nanosize Metals. Metal Powder Report 1998, 53, 31‐33. 2. Tepper, F., Metallic Nanopowders Produced by the Electro‐Exploding Wire Process. International Journal of Powder Metallurgy (Princeton, New Jersey) 1999, 35, 39‐44. 3. Jiang, W.; Yatsui, K., Pulsed Wire Discharge for Nanosize Powder Synthesis. IEEE Trans. Plasma Sci. 1998, 26, 1498‐1501. 4. Yetter, R. A.; Risha, G. A.; Son, S. F., Metal Particle Combustion and Nanotechnology. Proceedings of the Combustion Institute 2009, 32 II, 1819‐1838. 5. Dreizin, E. L., Metal‐Based Reactive Nanomaterials. Progress in Energy and Combustion Science 2009, 35, 141‐167. 6. Gromov, A.; DeLuca, L. T.; Il'in, A. P.; Teipel, U.; Petrova, A.; Prokopiev, D., Nanometals in Energetic Systems: Achievements and Future. Int. J. Energ. Mater. Chem. Propul. 2014, 13, 399‐419. 7. Meda, L.; Marra, G.; Galfetti, L.; Severini, F.; De Luca, L., Nano‐Aluminum as Energetic Material for Rocket Propellants. Materials Science and Engineering C 2007, 27, 1393‐1396. 8. Higa, K. T., Energetic Nanocomposite Lead‐Free Electric Primers. Journal of Propulsion and Power 2007, 23, 722‐727. 9. Pivkina, A. N.; Frolov, Y. V.; Ivanov, D. A., Nanosized Components of Energetic Systems: Structure, Thermal Behavior, and Combustion. Combustion, Explosion and Shock Waves 2007, 43, 51‐55. 10. Henz, B. J.; Hawa, T.; Zachariah, M. R. In Atomistic Simulation of the Aluminum Nanoparticle Oxidation Mechanism, Orlando, FL, Orlando, FL, 2010. 11. Rai, A.; Lee, D.; Park, K.; Zachariah, M. R., Importance of Phase Change of Aluminum in Oxidation of Aluminum Nanoparticles. Journal of Physical Chemistry B 2004, 108, 14793‐14795. 12. Park, K.; Lee, D.; Rai, A.; Mukherjee, D.; Zachariah, M. R., Size‐Resolved Kinetic Measurements of Aluminum Nanoparticle Oxidation with Single Particle Mass Spectrometry. Journal of Physical Chemistry B 2005, 109, 7290‐7299. 13. Rai, A.; Park, K.; Zhou, L.; Zachariah, M. R., Understanding the Mechanism of Aluminium Nanoparticle Oxidation. Combustion Theory and Modelling 2006, 10, 843‐859. 14. Trunov, M. A.; Umbrajkar, S. M.; Schoenitz, M.; Mang, J. T.; Dreizin, E. L., Oxidation and Melting of Aluminum Nanopowders. Journal of Physical Chemistry B 2006, 110, 13094‐13099. 15. Rufino, B.; Boulc'h, F.; Coulet, M. V.; Lacroix, G.; Denoyel, R., Influence of Particles Size on Thermal Properties of Aluminium Powder. Acta Materialia 2007, 55, 2815‐2827. 16. Coulet, M. V.; Rufino, B.; Esposito, P. H.; Neisius, T.; Isnard, O.; Denoyel, R., Oxidation Mechanism of Aluminum Nanopowders. Journal of Physical Chemistry C 2015, 119, 25063‐25070. 17. Puri, P.; Yang, V., Effect of Particle Size on Melting of Aluminum at Nano Scales. Journal of Physical Chemistry C 2007, 111, 11776‐11783. 18. Eisenreich, N.; Fietzek, H.; Del Mar Juez‐Lorenzo, M.; Kolarik, V.; Koleczko, A.; Weiser, V., On the Mechanism of Low Temperature Oxidation for Aluminum Particles Down to the Nano‐Scale. Propellants, Explosives, Pyrotechnics 2004, 29, 137‐145. 19. Trunov, M. A.; Schoenitz, M.; Zhu, X.; Dreizin, E. L., Effect of Polymorphic Phase Transformations in Al2o3 Film on Oxidation Kinetics of Aluminum Powders. Combustion and Flame 2005, 140, 310‐318. 20. Trunov, M. A.; Schoenitz, M.; Dreizin, E. L., Effect of Polymorphic Phase Transformations in Alumina Layer on Ignition of Aluminium Particles. Combustion Theory and Modelling 2006, 10, 603‐623. 21. Zhu, X.; Schoenitz, M.; Dreizin, E. L., Oxidation of Aluminum Particles in Mixed Co2/H2o Atmospheres. Journal of Physical Chemistry C 2010, 114, 18925‐18930. 22. Schoenitz, M.; Patel, B.; Agboh, O.; Dreizin, E. L., Oxidation of Aluminum Powders at High Heating Rates. Thermochimica Acta 2010, 507‐508, 115‐122. 13
23. Nie, H.; Zhang, S.; Schoenitz, M.; Dreizin, E. L., Reaction Interface between Aluminum and Water. International Journal of Hydrogen Energy 2013, 38, 11222‐11232. 24. Zhang, S.; Dreizin, E. L., Reaction Interface for Heterogeneous Oxidation of Aluminum Powders. Journal of Physical Chemistry C 2013, 117, 14025‐14031. 25. Nie, H.; Schoenitz, M.; Dreizin, E. L., Initial Stages of Oxidation of Aluminum Powder in Oxygen Journal of Thermal Analysis and Calorimetry 2016, under review. 26. Rasband, W. S., Imagej, U. S. National Institutes of Health, Bethesda, Maryland, USA. 1997‐2015. 27. Chen, L.; Song, W.; Lv, J.; Chen, X.; Xie, C., Research on the Methods to Determine Metallic Aluminum Content in Aluminum Nanoparticles. Materials Chemistry and Physics 2010, 120, 670‐675. 28. Zhu, X.; Schoenitz, M.; Dreizin, E. L., Aluminum Powder Oxidation in Co2 and Mixed Co2/O 2 Environments. Journal of Physical Chemistry C 2009, 113, 6768‐6773. 29. Vyazovkin, S., Model‐Free Kinetics: Staying Free of Multiplying Entities without Necessity. Journal of Thermal Analysis and Calorimetry 2006, 83, 45‐51. 30. Vyazovkin, S., Modification of the Integral Isoconversional Method to Account for Variation in the Activation Energy. Journal of Computational Chemistry 2001, 22, 178‐183. 31. Ramaswamy, A. L.; Kaste, P., A "Nanovision" of the Physiochemical Phenomena Occurring in Nanoparticles of Aluminum. Journal of Energetic Materials 2005, 23, 1‐25. 32. Gertsman, V. Y.; Kwok, Q. S. M., Tem Investigation of Nanophase Aluminum Powder. Microscopy and Microanalysis 2005, 11, 410‐420. 33. Reichel, F.; Jeurgens, L. P. H.; Richter, G.; Mittemeijer, E. J., Amorphous Versus Crystalline State for Ultrathin Al2o3 Overgrowths on Al Substrates. J. Appl. Phys. 2008, 103, 093515/1‐093515/10. 34. Jeurgens, L. P. H.; Sloof, W. G.; Tichelaar, F. D.; Mittemeijer, E. J., Growth Kinetics and Mechanisms of Aluminum‐Oxide Films Formed by Thermal Oxidation of Aluminum. Journal of Applied Physics 2002, 92, 1649. 35. 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. 36. Rufino, B.; Coulet, M. V.; Bouchet, R.; Isnard, O.; Denoyel, R., Structural Changes and Thermal Properties of Aluminium Micro‐ and Nano‐Powders. Acta Materialia 2010, 58, 4224‐4232.
14
Fig. 1. TEM images of the aluminum nanoparticles used in experiments.
Number of particles, ,%
25%
20%
15%
10%
5%
0% 1 10
2
10
Particle size, nm
3
10
Fig. 2. Particle size distribution obtained from TEM images.
15
Relative mass change
1.7 2 K/min (O 2) 5 K/min (O 2) 10 K/min (O 2) 10 K/min (air)
1.6 1.5 1.4 1.3 1.2 1.1 1.0 400
600
800
1000
1200
Temperature, K
Fig. 3. Mass gain for an oxidizing aluminum nanopowder heated at different heating rates in oxygen. A measurement in air, on which recovery of partially oxidized samples is based, is shown for comparison.
Al
Square root of intensity, a.u.
*
Al
*
Al
Al
*
*
1183 K
*
923 K
* 823 K
as prepared
20
30
40
50
60
70
80
Diffraction angle, 2 (Cu K) Fig. 4. XRD patterns showing as‐prepared powder and aluminum nano‐powders partially oxidized in air
16
and recovered from the indicated temperatures. Greek letters indicate polymorphs of Al2O3. Asterisks indicate diffraction peaks consistent with AlN.
Fig. 5. TEM images of partially oxidized aluminum particles. Particles were recovered from the indicated temperatures. Low‐temperature hollow features are marked with asterisks. Example EDX scans along the white lines are shown below the respective images.
17
Reactive interface r2 r1 Void Aluminum Oxide
Fig. 6. Schematic diagram of the geometry assumed by the oxidation model. Particle bin size, nm 842 336
Oxide thickness, nm
60 50
255 40 196 30 152 20 77 10 0 600
14 10 700
800
900
1000
Temperature, K
1100
1200
1300
Fig. 7. Thickness of the oxide layer grown on particles of different sizes as a function of temperature when the powder is heated at 10 K/min obtained as a result of splitting the measured TG trace among different powder size bins. Particle bin size represents the initial particle diameter.
18
Oxide thickness, nm
60 50
Measured from TEM Calculated from TG processing
40 30 20 10 0 700
800
900
1000
1100
1200
1300
Temperature, K
Fig. 8. Thickness of the oxide layer for different partially oxidized particles measured from TEM images and respective thickness of the oxide layer calculated to form on particles of respective sizes using the oxidation model shown in Fig. 1 and splitting the measured TG trace among different powder size bins.
Activation energy, kJ/mol
160 150 140 Bin size, nm
130 6. -11.1 Ea=154.2-1.8 .10 h
120
14 35 152 850
110 100 2.4
2.6
2.8
3.0
3.2
3.4
Oxide thickness, nm
3.6
3.8
Fig. 9. Apparent activation energy characterizing growth of amorphous alumina obtained from model‐ free isoconversion processing of experimental data and considering different particle sizes.
19
Pre-exponent, g/m/s
10
10
10
-1
152 nm 842 nm
-2
2 K/min 5 K/min 10 K/min
-3 7. -15.4 Ln(C)=-5.01-1.69 .10 h
10
-4
2.6
2.8
3.0
3.2
3.4
Oxide thickness, nm
Fig. 10. Pre‐exponent characterizing growth of amorphous alumina calculated considering different particle sizes and different heating rates.
Activation energy, kJ/mol
600 500 400
Near aluminum melting point
300 200
153 nm 850 nm
100 0
0
10
20
30
40
50
60
Oxide thickness, nm Fig. 11. Apparent activation energy characterizing growth of transition alumina polymorphs obtained from model‐free isoconversion processing of experimental data and considering different particle sizes.
20
Pre-exponent, g/m/s
10 10 10 10 10 10 10 10 10 10 10
4 3 2 1 0
Bin size, Heating rate, nm K/min
-1
152 152 152 842 842 842
-2 -3 -4 -5
2 5 10 2 5 10
-6
5
10
15
20
Oxide thickness, nm
Fig. 12. Pre‐exponent characterizing growth of transition alumina calculated considering different particle sizes and different heating rates.
Activation energy, kJ/mol
250
200 220
150
200 180 160
100
140 µm-Al n-Al, amorph. Al2 O3 n-Al, -Al2O3
50
0
3
5
8
10
120 100 80
13
3
15
4
18
Oxide thickness, nm
5
6
20
7
23
8
25
Fig. 13. Activation energy as a function of thickness of the grown oxide layer. Results for both amorphous and transition alumina from the present measurements are combined with the activation energy obtained for micron‐sized aluminum powders 25. Thin solid lines show straight line fits proposed for the activation energy in Ref. 25. A dashed line in the inset is calculated using Eq. (4).
21
4
Pre-exponent, g/m/s
10 3 10 2
10 1 10 0 10 10
-1 -2
10 -3 10 10
-4 -5
10 -6 10 10 10
µm-Al n-Al, amorph. Al2 O3 n-Al, -Al2 O3
-7
10 2 10 1 10 0 10 -1 10 -2 10 -3 10 -4 10 -5 10 -6
3
4
5
6
7
8
-8
3
5
8
10
13
15
18
Oxide thickness, nm
20
23
25
Fig. 14. Pre‐exponent as a function of thickness of the grown oxide layer. Results for both amorphous and transition aluminas from the present measurements are combined with the pre‐exponent obtained for micron‐sized aluminum powders 25. Thin solid lines show straight line fits proposed for the pre‐ exponent in Ref. 25. A dashed line in the inset is calculated using Eq. (5).
22
S0040-6031(16)30112-5 http://dx.doi.org/doi:10.1016/j.tca.2016.05.003 TCA 77509
To appear in:
Thermochimica Acta
Received date: Revised date: Accepted date:
13-2-2016 5-5-2016 7-5-2016
Please cite this article as: A.B.Vorozhtsov, M.Lerner, N.Rodkevich, H.Nie, A.Abraham, M.Schoenitz, E.L.Dreizin, Oxidation of nano-sized aluminum powders, Thermochimica Acta http://dx.doi.org/10.1016/j.tca.2016.05.003 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Oxidation of nano‐sized aluminum powders A.B. Vorozhtsov1, M. Lerner2, N. Rodkevich2, H. Nie3, A. Abraham3, M. Schoenitz3, E.L. Dreizin1,3 1
Tomsk State University, Tomsk, Russia
2
Institute of Strength Physics and Material Science, Siberian Branch, Russian Academy of Sciences, Tomsk, Russia
3
New Jersey Institute of Technology, Newark, NJ, USA
Corresponding author: Edward L. Dreizin, [email protected], 973‐596‐5751 Graphical abstract
1
Highlights 1. 2. 3. 4.
Weight gain measured in TG oxidation experiments was split between particles of different sizes Reaction kinetics obtained by isoconversion explicitly accounting for the effect of size distribution Activation energy is obtained as a function of oxide thickness for growth of amorphous alumina Oxidation mechanism for nanopowders remains the same as for coarser aluminum powders
Abstract Oxidation of aluminum nanopowders obtained by electro‐exploded wires is studied. Particle size distributions are obtained from transmission electron microscopy (TEM) images. Thermo‐gravimetric (TG) experiments are complemented by TEM and XRD studies of partially oxidized particles. Qualitatively, oxidation follows the mechanism developed for coarser aluminum powder and resulting in formation of hollow oxide shells. Sintering of particles is also observed. The TG results are processed to account explicitly for the particle size distribution and spherical shapes, so that oxidation of particles of different sizes is characterized. The apparent activation energy is obtained as a function of the reaction progress using model‐free isoconversion processing of experimental data. A complete phenomenological oxidation model is then proposed assuming a spherically symmetric geometry. The oxidation kinetics of aluminum powder is shown to be unaffected by particle sizes reduced down to tens of nm. The apparent activation energy describing growth of amorphous alumina is increasing at the very early stages of oxidation. The higher activation energy is likely associated with an increasing homogeneity in the growing amorphous oxide layer, initially containing multiple defects and imperfections. The trends describing changes in both activation energy and pre‐exponent of the growing amorphous oxide are useful for predicting ignition delays of aluminum particles. The kinetic trends describing activation energies and pre‐exponents in a broader range of the oxide thicknesses are useful for prediction of aging behavior of aluminum powders. Keywords: heterogeneous reaction, aluminum ignition, nano‐aluminum, aging, thermal analysis Introduction Development of practical methods to prepare nano‐sized aluminum powders 1 catalyzed research in new nano‐energetic materials, including metal nanopowders, nano‐thermites, and intermetallic systems 2‐4 . Although many compositions have been prepared and characterized in the last two decades, nano‐ sized aluminum powder remains the main component for multiple energetic formulations of interest for propellants, explosives, and pyrotechnics 5‐7. For most such formulations, oxidation of nano‐sized aluminum is the process affecting critically both aging and ignition phenomena. This explains an active and sustaining interest in understanding and quantitative characterization of reactions governing oxidation of nano‐sized aluminum, e.g., see 8‐14. In particular, oxidation similarities and differences between nano‐ and micron‐sized powders need to be established. For example, widely discussed changes in the melting characteristics of nano‐aluminum 15‐16 become significant for particles smaller than 10 nm and may not play an important role for most practical nano‐powders with dimensions in the
2
100 nm range. However, it remains unclear whether mechanisms and kinetics developed for oxidation of coarse aluminum powders are directly transferrable for nano‐sized particles. Oxidation mechanism of micron‐sized aluminum powders has been established relatively well 4, 17‐18. It is generally agreed that the reaction is diffusion controlled. It is also agreed that the diffusion rates are heavily affected by the structure of the growing aluminum oxide film. The natural 2 – 4‐nm thick oxide is amorphous; it transitions to ‐Al2O3 at elevated temperatures and growing oxide layer thickness 19‐20. Further oxidation processes involve transformation of ‐Al2O3 to ‐ and ‐Al2O3 as temperatures increase. It has been shown recently that the reaction occurs at the interface separating the growing alumina layer and gaseous oxidizer, and thus it is rate controlled by outward diffusion of aluminum ions 21‐22 . Detailed thermo‐analytical measurements were reported and processed to establish kinetics of the respective oxidation processes 23‐24. Reactions at early oxidation stages were addressed in particular, as most critical for modeling both ignition and aging phenomena 21. A recent study 22 addressing oxidation of nano‐sized aluminum mostly agrees with the above oxidation sequence. However, quantitative comparisons of oxidation rates and respective reaction kinetics are not available. This work is aimed to characterize oxidation of aluminum nano‐powders quantitatively. It is further desired to determine the kinetics of respective reactions and compare it directly to that reported for micron‐sized aluminum powders. Experimental details Spherical aluminum nanopowders were prepared in argon using electro‐exploded wires 16, 25. The powder was held at a reduced oxygen pressure for 72 hours for passivation. The amount of active aluminum in the as‐prepared powders was determined by reacting the powder with an aqueous solution of sodium hydroxide and measuring the volume of the released hydrogen gas1. Oxidation of spherical aluminum nanopowders is characterized in thermo‐analytical experiments, including differential thermal analysis (DTA) and thermo‐gravimetry (TG). For the purpose of obtaining partially oxidized materials, samples were heated in air at 10 K/min to 823 K, 923 K, and 1183 K. At the target temperature, the samples were allowed to cool, and recovered for analysis. Residual oxidation after reaching the target temperature has not been measured, but is assumed to be negligible. These experiments used a Netzsch STA 449F3 instrument. Additional TG measurements for the purpose of examining oxidation kinetics were performed using a TA Instruments TA Q5000‐IR instrument. For these experiments, heating rates varied from 2 to 10 K/min, and the atmosphere was pure oxygen. As‐prepared as well as partially oxidized particles were examined using a JEOL JEM‐100 CXII TEM. Particle size distributions were determined by image analysis, and are based on a count of approximately 4,600 particles. In addition, XRD (XRD‐6000 by Shimadzu using Cu K radiation) was used to determine phases formed during oxidation. The oxide thickness on partially oxidized samples was determined from multiple TEM images using ImageJ software26. Five measurements on each particle were averaged. Experimental results
3
TEM images of as prepared nanoparticles are shown in Fig. 1. The particles are mostly spherical. Particle sizes range from about 20 to 400 nm; a detailed particle size distribution obtained by processing multiple TEM images is shown in Fig. 2. Results of TG measurements are illustrated in Fig. 3. A small initial mass loss is observed. The measurements are consistent with earlier reports on oxidation of micron‐ and nano‐sized aluminum powders 2, 19, 27. A rapid acceleration of the oxidation rate is observed around 800 K, prior to aluminum melting, when amorphous oxide transforms into cubic spinel type alumina, typically ‐Al2O3. A DTA trace, omitted here for brevity, shows an exothermic peak corresponding to the accelerated mass increase, followed by a small endothermic peak at the Al melting point. At higher temperatures, reaction rate accelerates further, again in agreement with earlier reports 14, 16, 19 assigning that acceleration to thermally activated growth of transition alumina. The TG traces shift to higher temperatures at greater heating rates, as is expected for a thermally activated reaction. A TG curve for the oxidation in air, representative for the recovery of partially oxidized samples is shown in Fig. 3 for comparison. XRD patterns are shown in Fig. 4. The as‐prepared passivated powder contains no detectable impurities. Samples partially oxidized in air were recovered from different temperatures. Consistent with earlier results for coarser aluminum14, 16, there are no well resolved peaks of alumina for the sample recovered from 823 K. Thus, only amorphous oxide is expected to be formed at lower temperatures. This amorphous oxide may be partially hydrated, explaining the small initial mass loss observed by TG (see Fig. 3). At 923 K, alumina is present as the (JCPDS 01‐083‐2080) and polymorphs (JCPDS 98‐006‐ 9213). Additional peaks, marked by asterisks in Fig. 4 are consistent with AlN (JCPDS 04‐007‐4280). However plausible this appears for aluminum heated in air, based on the diffraction pattern alone a partially reduced oxide with composition Al3O4 (JCPDS 04‐016‐0539) cannot be ruled out. The presence of cubic, spinel‐type alumina is consistent with previous observation of alumina at this stage of oxidation. At 1183 K, alumina remains present. Fewer peaks of the spinel type alumina are observed, consistent with alumina (JCPDS 00‐056‐0458). The XRD patterns generally confirm that oxidation of nano‐sized aluminum particles results in the formation of the same sequence of alumina polymorphs as reported to form upon oxidation of coarser aluminum powders19‐22. TEM images of partially oxidized particles are shown in Fig. 5. At 823 K, the particles largely retain their spherical shapes. Some particles show apparent hollow features at the surface, as confirmed by negligible amounts of Al or O measured by EDX. The structure of particles recovered from 923 K changes substantially. All particles appear to contain a relatively thick oxide layer and a cavity. The presence of cavities is supported by mapping aluminum concentration using energy‐dispersive x‐ray spectroscopy (line scans shown in Fig. 5 as well). The cavity formation occurring on a time scale of minutes in the current TG experiments is further consistent with cavities observed in previous 4
research13,16, whereas in experiments13 individual Al nanoparticles were partially oxidized for durations on the order of a second at higher temperatures. Although particles mostly retain their spherical shapes, most particles are now fused together. Both oxide layer thickness and cavity size grow further for particles recovered from 1183 K. Technical Approach In order to establish a kinetic description of the oxidation process, the TG measurements in oxygen shown in Fig. 3 were processed accounting explicitly for the particle size distribution shown in Fig. 2. This allowed to distribute the observed total sample mass over the particles in all size bins, and to recover the oxide layer thickness for each particle size. The geometry of the oxidation model is shown in Fig. 6. It is based on earlier work on micron‐sized aluminum powders28. The location of the reactive interface on the outer particle surface implies that the oxide growth rate is limited by the outward diffusion of oxygen. As the aluminum is consumed, a void grows in the aluminum particle core, and aluminum is assumed to adhere to the inner surface of the growing oxide shell. However, location and shape of the void are not essential to the current calculation. The void could be located off‐center; it need not be spherical, and indeed may be present as multiple smaller voids. The model only accounts for the net volume of the void. The validity of this approach diminishes as the oxidizing particles sinter and/or change shape. However, this analysis is useful in the early stages of oxidation, which are particularly important for both ignition and aging processes. To recover the oxide layer thickness for each particle size separately, the mass increment measured at a given time for the entire powder, dmpowder, is taken as the sum of the mass increments of the particles in all size bins:
dm powder ni dmi
(1)
i
where ni is the number fraction of particles in the i-th size bin, and dmi is the corresponding mass increment of a particle in that size bin. In previous work on micron‐sized aluminum24, 28, a particle’s mass increment was taken as proportional to the instantaneous particle surface area. This was justified for early oxidation stages of large particles since the layer thickness on particles of different sizes was not significantly different. However, this assumption does not hold for nanosized particles, and was therefore relaxed in the current research. Instead, we use the solution of the diffusion equation for a spherical particle with a growing oxide shell, written assuming that pure Al and pure O2 gas are on the inner and outer sides of the oxide layer, respectively, and dropping the subscript i for brevity: 1
1 1 rr dm Al 4 Al D dt 4 Al D 1 2 dt r2 r1 r1 r2
(2)
5
where dmAl is the mass change of aluminum; D is the diffusion coefficient (generally following Arrhenius kinetics), Al is the Al density (taken as a function of temperature), and r1 and r2, are outer and inner radii of the oxide shell, respectively (see Fig. 6). The mass change of the particle is then
dm dmAl dmox ,
(3)
and the mass increments of oxide (taken as Al2O3) and Al metal are related via the reaction
stoichiometry, dmox dmAl M Al2O3
2M Al where M are the respective molar masses. Thus, the
mass increment for a particle in the i‐th size bin is proportional to the factor ai r1r2 r2 r1 which i is equivalent to the surface area at the geometric mean radius divided by the oxide layer thickness. The mass increment of the powder dmpowder is now distributed over all particles according to the distribution of ai over the whole size distribution:
ni dmi
ni ai dm powder ni ai
(4)
i
The inner and outer radii, used to calculate the factors a, are calculated from the total oxide mass of a particle, and the particle geometry:
3 mox t r2 r2,0 ox ,0 , r1 r23 ox T 4 ox T 1/3
1/3
(5)
while accounting for thermal expansion via the temperature dependent oxide density ox. Equations (5) specify that new oxide is accumulated at the outer particle surface (at r1), while the Al metal – Al oxide interface (at r2) remains constant save for thermal expansion. Equations 2‐5 are solved for each time step. The powder mass increment for each time step is determined by the TG data acquisition rate and is taken as the finite observed mass changes mpowder:
mi ,t mi ,t t
ai ,t
n a
m powder ,t
(6)
i i ,t
i
The subscripts t and t-t indicate values at the beginning and end of the current time step in the TG data. The summation starts at the beginning of the TG measurement with a particle mass mi ,t 0 corresponding to an Al particle with a 2.5‐nm oxide layer. This method for distributing the observed powder mass in a TG measurement over the experimental particle size distribution was originally developed elsewhere23‐24, and expanded for small particles in the present work. Ultimately, TG traces representing oxidation of particles of each given size bin are obtained for different heating rates. This made it possible to perform kinetic analysis using a model‐free isoconversion technique23‐24 to obtain activation energy as a function of the reaction progress defined for individual particle sizes. The processing explicitly accounts for the effect of particle size distribution on the TG measurements. Once the activation energies are obtained, the reaction model represented by Fig. 6 is 6
considered to identify the pre‐exponent and thus to fully quantify the rate of oxidation. The obtained reaction kinetics for aluminum nanopowders is finally compared to that reported earlier for coarser aluminum particles. Results of data processing Distributing the mass change of the whole powder (see Fig. 3) over the particle sizes shown in the particle size distributions in Fig. 3, allows to recover how the oxide thickness increases with temperature for particles of different sizes. The oxide thickness as a function of temperature is shown in Fig. 7. Finest, 10‐nm particles are predicted to be fully oxidized at temperatures close to 730 K. The thickness of the grown oxide layer is close to one half of the initial particle diameter. Larger particles are fully oxidized at increasingly higher temperatures. Note that the oxide layer is predicted to grow at the outer surface of the existing shell; thus the size of the particle continuously increases, while a cavity is growing inside. In addition, the thermal expansion is taken into account for both metal core and the oxide shell. Because the particle diameter increases as a function of temperature, the thickness of the oxide layer for the fully oxidized particles from larger bin sizes become substantially smaller than half of the initial particle diameter. A sudden reduction in the oxide thickness observed at 770 K represents re‐adjustment of the calculated thickness as a result of polymorphic phase transition from amorphous to a higher density ‐Al2O3. Predicted oxide thickness as shown in Fig. 7 was compared to that measured from the TEM images acquired for partially oxidized particles recovered from different temperatures. Diameters of particles observed in TEM images were compared to the diameters expected to be produced at the respective temperatures for the partially oxidized particles belonging to different particle size bins. The bin, for which the diameters matched was then selected for each image, so that the thicknesses of the calculated and observed in TEM oxide layers could be compared to each other directly. Results of such comparisons are shown in Fig. 8. The match between the TEM and processed TG data is excellent, supporting the validity of the adopted oxidation model used to convert the measured mass increase into the oxide thickness. Oxidation kinetics for nano‐sized aluminum particles Thickness of the grown oxide layer was considered as a convenient indicator of the reaction progress. Unlike mass increase, oxide thickness can be used to gauge the degree of oxidation for particles of different sizes. Model‐free isoconversion processing 29 was applied to TG curves assigned to individual particle size bins. Initial oxide thickness was assumed to be 2.5 nm 30. The processing was performed separately for the portions of the curves describing growth of amorphous alumina, at temperatures below 770 K, and growth of transition alumina polymorphs, principally, ‐Al2O3, above that temperature. The temperature, at which this transition occurs was fixed for simplicity in calculations using the oxidation model shown in Fig. 1 and transforming the measured TG trace into individual traces for different size bins. A more realistic model should account for both temperature and thickness of the grown oxide layer 19, 31‐33 to describe this transition with a greater accuracy. Apparent activation energy as a function of the oxide thickness for growth of amorphous oxide, Eamorph, is shown in Fig. 9 for several individual particle size bins. The results for different size bins are consistent 7
among themselves and show a rapid increase in the activation energy followed by its stabilization at about 155 kJ/mol. The change in the activation energy, in kJ/mol is well described by the following curve‐fitting function:
154.2
. ∙ .
(4)
where h is oxide thickness in nm. The above expression is only valid for oxide thicknesses exceeding 2.5 nm, which is the thickness assumed to represent the initial natural oxide layer. Once the activation energy, E, as a function of reaction progress is identified, an explicit oxidation model is used to determine the respective value of the pre‐exponent. Considering oxidation of a spherical particle with an oxide shell, as given by Eq. (2), and assuming that
∙
, the values of
pre‐exponent characterizing growth of amorphous alumina, Camorph, were obtained as shown in Fig. 10. The pre‐exponent was separately obtained for different heating rates and for different particle size bins. For clarity, shown in Fig. 10 are only the pre‐exponents obtained for two different size bins. The results were very consistent among themselves showing a trend that closely reflects the changes in the activation energy. As expected, the pre‐exponent changes in a much broader range of values than activation energy. Its change (in ) can be described as a function of the oxide thickness using a curve‐ ∙
fitting expression analogous to that used for the activation energy:
5.01
.
∙ .
(5)
Similarly to Eq. (4), Eq. (5) can be used for the oxide thicknesses exceeding 2.5 nm. Results of isoconversion processing of the TG measurements temperatures above 770 K, describing growth of transition alumina polymorphs are illustrated in Fig. 11. Apparent activation energy is calculated for thicknesses above ca. 5.4 nm while the calculated values in Fig. 9, for amorphous oxide, end around 3.5 nm. The gap between 3.5 and 5.4 nm is not described because the polymorphic phase transition was assumed to occur at a fixed temperature of 770 K. At different heating rates, different thickness of the amorphous oxide is predicted to grow by 770 K, varied from about 3.5 to 5.4 nm in present experiments. Respectively, the isoconversion processing, which uses data for all heating rates simultaneously, is limited to the data range with the same range of variation of the oxide thickness, selected as the reaction progress indicator. The values of the apparent activation energies describing growth of ‐Al2O3 (Fig. 11) obtained from processing TG data corresponding to different particle size bins are consistent with one another. The apparent activation energy increases from about 180 to approximately 230 kJ/mol as the oxide thickness grows up to ca. 12 nm. The activation energy then decreases to approximately 150 kJ/mol when oxide thickness reaches ca. 20 nm. Strong variations in the activation energy are observed at thicknesses around 22 – 24 nm, which occur near the aluminum melting point for the heating rates used in the TG measurements. Following the melting point, the activation energy is observed to increase, reaching
8
very high values at the terminal thickness of about 60 nm. It should be emphasized that Al melting has no causal relation with the decrease in activation energy. It is used here only as a reference point. Using the apparent activation energy as shown in Fig. 11, and accounting for the specific oxidation model (Fig. 1), the pre‐exponent values characterizing growth of transition alumina can also be obtained, as shown in Fig. 12. The pre‐exponent was only calculated for oxidation up to about 20 nm, e.g., before the aluminum melting occurring in TG experiments. A rapid increase in the apparent activation energy following melting is probably associated with substantial changes in the morphology of the oxidizing sample as a result of particle agglomeration, e.g., loss in the surface area available for reaction. This increase is unlikely to characterize heterogeneous oxidation reactions of interest in this study. For the pre‐exponent, the results are somewhat different when different particle size bins and different heating rates are used. A possible reason for this discrepancy is that in the present analysis, the polymorphic phase transition from amorphous to ‐Al2O3 was assumed to occur at a fixed temperature for simplicity. In reality, this transition is affected by both temperature and oxide thickness; it occurs at different temperatures for different oxide thicknesses, depending on the heating rate 33‐35. Neglecting the effect of oxide thickness would shift the starting point for the present analysis compared to the actual formation of ‐Al2O3. The shift would be different for different particle size bins and for different heating rates. Discussion In this section, results obtained here for oxidation for nano‐sized aluminum powders are compared to those reported for micron‐sized aluminum earlier 34. Possible reasons for the observed differences in the reaction kinetics for powders with different particle sizes are also discussed. Apparent activation energy characterizing oxidation of aluminum is shown in Fig. 13 as a function of the oxide thickness, where results for amorphous and transition alumina polymorphs shown in Figs. 9 and 11 are combined with the activation energy obtained for a coarser aluminum powder 35. The inset in Fig. 13 shows a close‐up to the range of small oxide thicknesses. Thin solid lines represent piecewise straight line fits proposed to describe the activation energy as a function of thickness in Ref. 25. It is remarkable that for both nano‐sized and micron‐sized powders, apparent activation energy describing growth of amorphous alumina is shown to increase substantially for very small thicknesses and at low temperatures. The straight line segments proposed to describe the activation energy as a function of the oxide layer thickness for coarse aluminum for practical applications offer a reasonable approximation for the present data. Using Eq. (4) instead is advisable for greater accuracy of predicted oxidation rate; however, the difference between the values of activation energy is minor. A fairly sharp increase in the apparent activation energy observed for thin amorphous alumina layers may be attributed to healing imperfections and defects present in natural alumina and described in earlier research using high‐resolution TEM images 25, 32. 9
The trend for the apparent activation energy describing growth of ‐Al2O3 diverges slightly from the trend implied by the current data for growth of the amorphous oxide. This small difference is not apparent from the data obtained for coarser powder in ref. 36, where the entire range of oxide thicknesses was processed, without subdividing it into portions related to amorphous and crystalline alumina polymorphs. Generally, there is good agreement between trends describing the apparent activation energy as a function of thickness of ‐Al2O3 for both nano‐ and micron‐sized powders for the oxide thicknesses up to about 13 nm. The apparent activation energies diverge somewhat between different types of powders for greater alumina thicknesses, most likely as a result of changes in the powder morphologies, primarily, sintering between oxidizing particles. The initial oxidation is expected to be better described using Eq. (5) for the pre‐exponent, while the piecewise straight line fits proposed in Ref. 31 for coarser aluminum are also expected to predict the oxidation rate with an acceptable accuracy. Comparisons of pre‐exponent values for different powders are illustrated in Fig. 14. The inset shows a close‐up to the range of small oxide thicknesses. As in Fig. 13, thin solid lines represent piecewise straight line fits proposed to describe the pre‐exponent as a function of thickness in Ref. 25. There are small but appreciable differences between the straight line fragments proposed to describe the pre‐ exponent as a function of the oxide thickness for coarse aluminum and the trends describing the pre‐ exponent for both amorphous and ‐Al2O3 based on the present measurements. The initial oxidation is expected to be better described using Eq. (5) for the pre‐exponent, using the present measurements, where the initial oxidation is resolved better. Description of growth of ‐Al2O3 is a bit more difficult to quantify precisely because of an approximation used here to determine the instant of the polymorphic phase change leading to its formation. Initial reactions starting from thin, natural oxide film are particularly important for descriptions of both ignition and aging of aluminum and related reactive materials. Therefore, oxidation kinetics identified in Ref. 25 and validated here, with slightly amended expressions for activation energy and pre‐exponent (Eqs. 4, 5) should be used in the respective reaction models. Aluminum ignition is most likely occurring soon after the polymorphic phase change transforming the natural amorphous alumina into a higher density ‐Al2O3 phase. The diffusion resistance of the oxide layer diminishes, reaction rate accelerates, causing ignition 19, 25. Predicting when the polymorphic transition between amorphous and ‐Al2O3 occur can thus be used to predict the ignition delay for an aluminum particle. Because the phase change depends on both temperature and the oxide thickness, calculating an increasing oxide thickness during the initial pre‐ignition particle heating is important. Equations (4) and (5) obtained here as fits to the experimentally inferred apparent activation energy and respective pre‐exponent can be used to perform such calculations for any aluminum particle heating scenarios. For example, to account for transitions between alumina polymorphs, the following equation was proposed in Ref. 20:
10
1
(6)
where F, K, and E are parameters describing a specific oxide polymorph. It is important that the rate of growth is affected by both temperature, as in usual thermally activated oxidation, and by the oxide thickness. Thus, to accurately predict when the rate of growth of ‐Al2O3 becomes significant (i.e., greater than the growth of amorphous alumina), one needs to simultaneously track both temperature and thickness of the oxide layer. Eqs. (4) and (5) enable one to track the thickness of the growing amorphous alumina, and thus enable one to predict when ‐Al2O3 will start to grow, or when the particle is expected to ignite. A main shortcoming of the present approach is an approximate nature of Eq. (6), although it properly accounts for the effects of both temperature and oxide thickness on structure of the growing aluminum oxide. For aging analysis, the effect of polymorphic phase change on the thickness of growing aluminum oxide layer can be neglected; thus the straight line fits proposed in Ref. 25 are useful, which also describe the present results reasonably well. Conversely, fitting lines connecting trends for activation energies and pre‐exponents for both amorphous and ‐Al2O3 using the present data can be obtained and used to predict the rate of aluminum oxidation in slow aging processes. Conclusions Thickness of the oxide layer growing on nano‐sized aluminum powders exposed to elevated temperatures in oxygen‐containing environments was obtained from both TEM images and processing of respective TG traces. The TG data processing accounted for the specific particle size distribution. The measured weight gain was split among different size bins of the powder, accounting for particle sizes and spherical shapes, and taking into consideration that the reaction occurs at the outer interface of the growing oxide layer. The oxide layers observed in TEM images correlated well with those implied at different temperatures by the TG measurements. Oxidation kinetics was characterized using isoconversion processing of the TG data. Separate trends for kinetic activation energies and pre‐exponents as a function of the oxide thickness were obtained to describe growth of amorphous and ‐Al2O3 polymorphs. The present results are in good agreement with the earlier studies characterizing kinetics of oxidation of coarser aluminum powders. The present results confirm that the mechanism of aluminum oxidation does not change as a function of particle size down to tens of nm. An increase in the apparent activation energy describing growth of amorphous alumina implied by earlier work is confirmed here. The higher activation energy is likely associated with an increasing homogeneity in the growing amorphous oxide layer, initially containing multiple defects and imperfections. The trends obtained here and describing changes in both activation energy and pre‐ exponent of the growing amorphous oxide are useful for prediction of ignition delays of aluminum particles. The ignition is expected to occur when ‐Al2O3 forms disrupting continuity of the oxide layer covering aluminum surface. Formation of ‐Al2O3 is affected by both temperature and thickness of the oxide layer. The thickness of the oxide grown during pre‐ignition processes can be predicted using the kinetics of growth of amorphous alumina obtained here. The kinetic trends describing activation energies and pre‐exponents in a broader range of the oxide thicknesses are useful for prediction of aging behavior of aluminum powders. 11
Acknowledgements Research at NJIT was supported by Defense Threat Reduction Agency, Grant HDTRA1‐11‐1‐0060. The work at Tomsk State University was financially supported by the Ministry of Education and Science of the Russian Federation within the framework of the Federal Target Program; Agreement No. 14.587.21.0019 (Unique identifier RFMEFI58715X0019).
12
References 1. Tepper, F., Electro‐Explosion of Wire Produces Nanosize Metals. Metal Powder Report 1998, 53, 31‐33. 2. Tepper, F., Metallic Nanopowders Produced by the Electro‐Exploding Wire Process. International Journal of Powder Metallurgy (Princeton, New Jersey) 1999, 35, 39‐44. 3. Jiang, W.; Yatsui, K., Pulsed Wire Discharge for Nanosize Powder Synthesis. IEEE Trans. Plasma Sci. 1998, 26, 1498‐1501. 4. Yetter, R. A.; Risha, G. A.; Son, S. F., Metal Particle Combustion and Nanotechnology. Proceedings of the Combustion Institute 2009, 32 II, 1819‐1838. 5. Dreizin, E. L., Metal‐Based Reactive Nanomaterials. Progress in Energy and Combustion Science 2009, 35, 141‐167. 6. Gromov, A.; DeLuca, L. T.; Il'in, A. P.; Teipel, U.; Petrova, A.; Prokopiev, D., Nanometals in Energetic Systems: Achievements and Future. Int. J. Energ. Mater. Chem. Propul. 2014, 13, 399‐419. 7. Meda, L.; Marra, G.; Galfetti, L.; Severini, F.; De Luca, L., Nano‐Aluminum as Energetic Material for Rocket Propellants. Materials Science and Engineering C 2007, 27, 1393‐1396. 8. Higa, K. T., Energetic Nanocomposite Lead‐Free Electric Primers. Journal of Propulsion and Power 2007, 23, 722‐727. 9. Pivkina, A. N.; Frolov, Y. V.; Ivanov, D. A., Nanosized Components of Energetic Systems: Structure, Thermal Behavior, and Combustion. Combustion, Explosion and Shock Waves 2007, 43, 51‐55. 10. Henz, B. J.; Hawa, T.; Zachariah, M. R. In Atomistic Simulation of the Aluminum Nanoparticle Oxidation Mechanism, Orlando, FL, Orlando, FL, 2010. 11. Rai, A.; Lee, D.; Park, K.; Zachariah, M. R., Importance of Phase Change of Aluminum in Oxidation of Aluminum Nanoparticles. Journal of Physical Chemistry B 2004, 108, 14793‐14795. 12. Park, K.; Lee, D.; Rai, A.; Mukherjee, D.; Zachariah, M. R., Size‐Resolved Kinetic Measurements of Aluminum Nanoparticle Oxidation with Single Particle Mass Spectrometry. Journal of Physical Chemistry B 2005, 109, 7290‐7299. 13. Rai, A.; Park, K.; Zhou, L.; Zachariah, M. R., Understanding the Mechanism of Aluminium Nanoparticle Oxidation. Combustion Theory and Modelling 2006, 10, 843‐859. 14. Trunov, M. A.; Umbrajkar, S. M.; Schoenitz, M.; Mang, J. T.; Dreizin, E. L., Oxidation and Melting of Aluminum Nanopowders. Journal of Physical Chemistry B 2006, 110, 13094‐13099. 15. Rufino, B.; Boulc'h, F.; Coulet, M. V.; Lacroix, G.; Denoyel, R., Influence of Particles Size on Thermal Properties of Aluminium Powder. Acta Materialia 2007, 55, 2815‐2827. 16. Coulet, M. V.; Rufino, B.; Esposito, P. H.; Neisius, T.; Isnard, O.; Denoyel, R., Oxidation Mechanism of Aluminum Nanopowders. Journal of Physical Chemistry C 2015, 119, 25063‐25070. 17. Puri, P.; Yang, V., Effect of Particle Size on Melting of Aluminum at Nano Scales. Journal of Physical Chemistry C 2007, 111, 11776‐11783. 18. Eisenreich, N.; Fietzek, H.; Del Mar Juez‐Lorenzo, M.; Kolarik, V.; Koleczko, A.; Weiser, V., On the Mechanism of Low Temperature Oxidation for Aluminum Particles Down to the Nano‐Scale. Propellants, Explosives, Pyrotechnics 2004, 29, 137‐145. 19. Trunov, M. A.; Schoenitz, M.; Zhu, X.; Dreizin, E. L., Effect of Polymorphic Phase Transformations in Al2o3 Film on Oxidation Kinetics of Aluminum Powders. Combustion and Flame 2005, 140, 310‐318. 20. Trunov, M. A.; Schoenitz, M.; Dreizin, E. L., Effect of Polymorphic Phase Transformations in Alumina Layer on Ignition of Aluminium Particles. Combustion Theory and Modelling 2006, 10, 603‐623. 21. Zhu, X.; Schoenitz, M.; Dreizin, E. L., Oxidation of Aluminum Particles in Mixed Co2/H2o Atmospheres. Journal of Physical Chemistry C 2010, 114, 18925‐18930. 22. Schoenitz, M.; Patel, B.; Agboh, O.; Dreizin, E. L., Oxidation of Aluminum Powders at High Heating Rates. Thermochimica Acta 2010, 507‐508, 115‐122. 13
23. Nie, H.; Zhang, S.; Schoenitz, M.; Dreizin, E. L., Reaction Interface between Aluminum and Water. International Journal of Hydrogen Energy 2013, 38, 11222‐11232. 24. Zhang, S.; Dreizin, E. L., Reaction Interface for Heterogeneous Oxidation of Aluminum Powders. Journal of Physical Chemistry C 2013, 117, 14025‐14031. 25. Nie, H.; Schoenitz, M.; Dreizin, E. L., Initial Stages of Oxidation of Aluminum Powder in Oxygen Journal of Thermal Analysis and Calorimetry 2016, under review. 26. Rasband, W. S., Imagej, U. S. National Institutes of Health, Bethesda, Maryland, USA. 1997‐2015. 27. Chen, L.; Song, W.; Lv, J.; Chen, X.; Xie, C., Research on the Methods to Determine Metallic Aluminum Content in Aluminum Nanoparticles. Materials Chemistry and Physics 2010, 120, 670‐675. 28. Zhu, X.; Schoenitz, M.; Dreizin, E. L., Aluminum Powder Oxidation in Co2 and Mixed Co2/O 2 Environments. Journal of Physical Chemistry C 2009, 113, 6768‐6773. 29. Vyazovkin, S., Model‐Free Kinetics: Staying Free of Multiplying Entities without Necessity. Journal of Thermal Analysis and Calorimetry 2006, 83, 45‐51. 30. Vyazovkin, S., Modification of the Integral Isoconversional Method to Account for Variation in the Activation Energy. Journal of Computational Chemistry 2001, 22, 178‐183. 31. Ramaswamy, A. L.; Kaste, P., A "Nanovision" of the Physiochemical Phenomena Occurring in Nanoparticles of Aluminum. Journal of Energetic Materials 2005, 23, 1‐25. 32. Gertsman, V. Y.; Kwok, Q. S. M., Tem Investigation of Nanophase Aluminum Powder. Microscopy and Microanalysis 2005, 11, 410‐420. 33. Reichel, F.; Jeurgens, L. P. H.; Richter, G.; Mittemeijer, E. J., Amorphous Versus Crystalline State for Ultrathin Al2o3 Overgrowths on Al Substrates. J. Appl. Phys. 2008, 103, 093515/1‐093515/10. 34. Jeurgens, L. P. H.; Sloof, W. G.; Tichelaar, F. D.; Mittemeijer, E. J., Growth Kinetics and Mechanisms of Aluminum‐Oxide Films Formed by Thermal Oxidation of Aluminum. Journal of Applied Physics 2002, 92, 1649. 35. 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. 36. Rufino, B.; Coulet, M. V.; Bouchet, R.; Isnard, O.; Denoyel, R., Structural Changes and Thermal Properties of Aluminium Micro‐ and Nano‐Powders. Acta Materialia 2010, 58, 4224‐4232.
14
Fig. 1. TEM images of the aluminum nanoparticles used in experiments.
Number of particles, ,%
25%
20%
15%
10%
5%
0% 1 10
2
10
Particle size, nm
3
10
Fig. 2. Particle size distribution obtained from TEM images.
15
Relative mass change
1.7 2 K/min (O 2) 5 K/min (O 2) 10 K/min (O 2) 10 K/min (air)
1.6 1.5 1.4 1.3 1.2 1.1 1.0 400
600
800
1000
1200
Temperature, K
Fig. 3. Mass gain for an oxidizing aluminum nanopowder heated at different heating rates in oxygen. A measurement in air, on which recovery of partially oxidized samples is based, is shown for comparison.
Al
Square root of intensity, a.u.
*
Al
*
Al
Al
*
*
1183 K
*
923 K
* 823 K
as prepared
20
30
40
50
60
70
80
Diffraction angle, 2 (Cu K) Fig. 4. XRD patterns showing as‐prepared powder and aluminum nano‐powders partially oxidized in air
16
and recovered from the indicated temperatures. Greek letters indicate polymorphs of Al2O3. Asterisks indicate diffraction peaks consistent with AlN.
Fig. 5. TEM images of partially oxidized aluminum particles. Particles were recovered from the indicated temperatures. Low‐temperature hollow features are marked with asterisks. Example EDX scans along the white lines are shown below the respective images.
17
Reactive interface r2 r1 Void Aluminum Oxide
Fig. 6. Schematic diagram of the geometry assumed by the oxidation model. Particle bin size, nm 842 336
Oxide thickness, nm
60 50
255 40 196 30 152 20 77 10 0 600
14 10 700
800
900
1000
Temperature, K
1100
1200
1300
Fig. 7. Thickness of the oxide layer grown on particles of different sizes as a function of temperature when the powder is heated at 10 K/min obtained as a result of splitting the measured TG trace among different powder size bins. Particle bin size represents the initial particle diameter.
18
Oxide thickness, nm
60 50
Measured from TEM Calculated from TG processing
40 30 20 10 0 700
800
900
1000
1100
1200
1300
Temperature, K
Fig. 8. Thickness of the oxide layer for different partially oxidized particles measured from TEM images and respective thickness of the oxide layer calculated to form on particles of respective sizes using the oxidation model shown in Fig. 1 and splitting the measured TG trace among different powder size bins.
Activation energy, kJ/mol
160 150 140 Bin size, nm
130 6. -11.1 Ea=154.2-1.8 .10 h
120
14 35 152 850
110 100 2.4
2.6
2.8
3.0
3.2
3.4
Oxide thickness, nm
3.6
3.8
Fig. 9. Apparent activation energy characterizing growth of amorphous alumina obtained from model‐ free isoconversion processing of experimental data and considering different particle sizes.
19
Pre-exponent, g/m/s
10
10
10
-1
152 nm 842 nm
-2
2 K/min 5 K/min 10 K/min
-3 7. -15.4 Ln(C)=-5.01-1.69 .10 h
10
-4
2.6
2.8
3.0
3.2
3.4
Oxide thickness, nm
Fig. 10. Pre‐exponent characterizing growth of amorphous alumina calculated considering different particle sizes and different heating rates.
Activation energy, kJ/mol
600 500 400
Near aluminum melting point
300 200
153 nm 850 nm
100 0
0
10
20
30
40
50
60
Oxide thickness, nm Fig. 11. Apparent activation energy characterizing growth of transition alumina polymorphs obtained from model‐free isoconversion processing of experimental data and considering different particle sizes.
20
Pre-exponent, g/m/s
10 10 10 10 10 10 10 10 10 10 10
4 3 2 1 0
Bin size, Heating rate, nm K/min
-1
152 152 152 842 842 842
-2 -3 -4 -5
2 5 10 2 5 10
-6
5
10
15
20
Oxide thickness, nm
Fig. 12. Pre‐exponent characterizing growth of transition alumina calculated considering different particle sizes and different heating rates.
Activation energy, kJ/mol
250
200 220
150
200 180 160
100
140 µm-Al n-Al, amorph. Al2 O3 n-Al, -Al2O3
50
0
3
5
8
10
120 100 80
13
3
15
4
18
Oxide thickness, nm
5
6
20
7
23
8
25
Fig. 13. Activation energy as a function of thickness of the grown oxide layer. Results for both amorphous and transition alumina from the present measurements are combined with the activation energy obtained for micron‐sized aluminum powders 25. Thin solid lines show straight line fits proposed for the activation energy in Ref. 25. A dashed line in the inset is calculated using Eq. (4).
21
4
Pre-exponent, g/m/s
10 3 10 2
10 1 10 0 10 10
-1 -2
10 -3 10 10
-4 -5
10 -6 10 10 10
µm-Al n-Al, amorph. Al2 O3 n-Al, -Al2 O3
-7
10 2 10 1 10 0 10 -1 10 -2 10 -3 10 -4 10 -5 10 -6
3
4
5
6
7
8
-8
3
5
8
10
13
15
18
Oxide thickness, nm
20
23
25
Fig. 14. Pre‐exponent as a function of thickness of the grown oxide layer. Results for both amorphous and transition aluminas from the present measurements are combined with the pre‐exponent obtained for micron‐sized aluminum powders 25. Thin solid lines show straight line fits proposed for the pre‐ exponent in Ref. 25. A dashed line in the inset is calculated using Eq. (5).
22