Influence of shell thickness on thermal stability of bimetallic Al–Pd nanoparticles

RESEARCH PAPER Influence of shell thickness on thermal stability of bimetallic Al–Pd nanoparticles John Z. Wen • Ngoc Ha Nguyen • John Rawlins • Catalin F. Petre • Sophie Ringuette Received: 22 March 2014 / Accepted: 26 May 2014  Springer Science+Business Media Dordrecht 2014 Abstract Aluminum-based bimetallic core–shell nanoparticles have shown promising applications in civil and defense industries. This study addresses the thermal stability of aluminum–palladium (Al–Pd) core/shell nanoparticles with a varying shell thickness of 5, 6, and 7 A˚, respectively. The classic molecular dynamics (MD) simulations are performed in order to investigate the effects of the shell thickness on the ignition mechanism and subsequent energetic pro- cesses of these nanoparticles. The histograms of temperature change and structural evolution clearly show the inhibition role of the Pd shell during ignition. While the nanoparticle with a thicker shell is more thermally stable and hence requires more excess energy, stored as the potential energy of the nanopar- ticle and provided through numerically heating, to initiate the thermite reaction, a higher adiabatic temperature can be produced from this nanoparticle, thanks to its greater content of Pd. The two-stage thermite reactions are discussed with their activation energy based on the energy balance processes during MD heating and production. Analyses of the simula- tion results reveal that the inner pressure of the core– shell nanoparticle increases with both temperature and the absorbed thermal energy during heating, which may result in a breakup of the Pd shell. Keywords Bimetallic  Ignition  Energetics  Aluminum  Core–shell  Nanoparticles Introduction Energetic nanomaterial has shown promising applica- tions in developing advanced technologies related to chemical propellants, materials processing, compo- nents joining, propulsion, and power production. There are typically two classes of such kind of material. This first class is the metastable intermolec- ular composites (MICs), which contains Al nanopar- ticles and a nanostructured oxidizer such as CuO, Fe2O3, SnO2, and WO3. The other is the binary energetic nanostructures (e.g., core–shell nanoparti- cles and layered nanolaminates) of Ni/Ti, Co/Al, Ni/ Al, and Pt/Al. The latter has recently drawn much attention in developing micro-joining and micro- welding technologies (Hemeryck et al. 2010; Picard et al. 2008; Olia et al. 2012; Yang et al. 2012; J. Z. Wen (&)  J. Rawlins Department of Mechanical and Mechatronics Engineering, University of Waterloo, 200 University Avenue West, Waterloo, ON N2L 3G1, Canada e-mail: john.wen@uwaterloo.ca N. H. Nguyen Department of Chemistry, Center for Computational Science, Hanoi National University of Education, Hanoi, Vietnam C. F. Petre  S. Ringuette Defence Research and Development Canada – Valcartier, 2459 Pie-XI Blvd North, Quebec, QC G3K1Y1, Canada 123 J Nanopart Res (2014) 16:2486 DOI 10.1007/s11051-014-2486-z Levchenko et al. 2010; Fiedler et al. 2012; Gillner et al. 2006; Howell et al. 2011). In comparison with the extensive studies in the literature on characterizing thermal and chemical properties of MIC, few exper- imental studies have been reported on investigating the ignition and energetic properties of the bimetallic nanothermites. One of these challenges in this field is the fabrication and characterization of the bimetallic core–shell and laminated nanostructures, although the synthesis of larger particles such as the Ni-coated Al microparticles was recently reported (Wang et al. 2008). On the other hand, numerical and theoretical studies have been performed to a greater extent to reveal the structural transformation and inner diffusion processes within the nanoparticles during ignition and combustion. Recently, researchers carried out such computational tasks on some bimetallic nanothermites (Yang et al. 2012; Levchenko et al. 2010; Evteev et al. 2009; Nguyen et al. 2011; Song and Wen 2010). As bimetallic nanoparticles, these core–shell nanoalloys are expected to exhibit the morphological instability which usually correlates with the particle size as well as the shape and placement of the inner core (Bochic- chio and Ferrando 2013; Laasonen et al. 2013). The determination of the nanoalloy geometries therefore requires performing global optimization searches, based on the reliable potential energy data of the interested bimetallic system. In the literature, the global optimization has been normally carried out for small clusters with \100 atoms. It becomes very challenging for nanoparticles of a few nanometers, duo to computing difficulties, although a very recent study reported on the morphological instability of Ag-based nanoparticles with diameters up to 3–4 nm (Bochic- chio and Ferrando 2013). It was suggested that in that study, there is a critical value of the shell number of core atoms, for the given shell number of the shell atoms, at which the morphological instability becomes significant. And this critical value is system dependent. Heat released from the core–shell bimetallic ther- mite nanoparticles is comparable to the energy pro- duced by MIC. Table 1 summarizes the energy release data from a variety of Al-based thermite and bimetallic systems (Fischer and Grubelich 1998), which do not necessarily exhibit designated nanostructures. Shown in this table, the Al–Pd system, as the most powerful bimetallic thermite, can produce 2,890 cal/cm3 from its thermite reaction, compared with 4,976 cal/cm3 from the Al–CuO system. More importantly, due to its faster alloying rate and more intensive energy produc- tion rate, less gas is generated from the Al–Pd thermite reaction, which produces the liquid metal under the adiabatic condition. These characteristics are extre- mely critical and well desired in developing new wire bonding or material joining technologies for the microelectronics and MEMS (microelectromechanical systems) devices, which requires a larger power density, a greater reaction rate, appropriate wetting, and a less vibrational disturbance. As a predictive tool of composite design and optimization, theoretical studies using the classical molecular dynamics (MD) approach have become popular to investigate the ignition mechanism and energetic properties of both core–shell nanoparticles and multilayer nanolaminates (Yang et al. 2012; Levchenko et al. 2010; Evteev et al. 2009). This approach is not suitable to perform the global optimi- zation of stable geometries of nanoparticles, which usually requires combined simulations by means of atomistic and density functional theory (DFT) models (Bochicchio and Ferrando 2013; Laasonen et al. 2013). The MD approach, however, can address larger nanoparticles. Most such studies on nanothermites have been focused on the Al–Ni system, while little research has been done on the Al–Pd system. Recently, a force field has been reported for studying the Al–Pd binary system (Nguyen et al. 2011). That force filed, constructed by employing the EAM (embedded atom method) potentials developed from the structural parameters of pure Al and Pd metals and the Al–Pd binary, was used to study the thermal stability and energetic reaction of a Pd-coated Al Table 1 Energetic properties of Al-based thermites and intermetallic systems (based on Fischer and Grubelich 1998) Reactants Adiabatic temperature (K) Heat of reaction (cal/cm3) Gas production (g/g) Al–CuO 5,718 4,976 0.3431 Al–NiO 3,968 4,288 0.0063 Al–Fe2O3 4,382 3,947 0.0784 Al–B2O3 2,621 1,971 0 Al–Ni 2,362 1,710 0 Al–Pt 3,379 2,510 0 Al–Pd 2,725 2,890 0 2486 Page 2 of 12 J Nanopart Res (2014) 16:2486 123 nanoparticle with a diameter of 5 nm and possessing a fixed ratio of Al to Pd atoms. In that study, a new two- stage reaction process was proposed when the thermite was ignited below 900 K. At the first stage, the solid- state diffusion of Al atoms occurs within the nano- particle. At the second stage, when the Al core is molten, the alloying reaction between the liquid Al core and the Pd shell happens. Due to the limitation of computing resources, no effect of the shell thickness on the ignition and combustion was reported. In the present study, three different Al–Pd core/ shell nanoparticles, with the varying thicknesses of 5, 6, and 7 A˚, respectively, were numerically investi- gated. The diameters of these nanoparticles were set to about 3 nm. The classic MD simulations were per- formed to produce the histograms of temperature change and structural evolution of these nanoparticles, which were analyzed to study the effects of the shell thickness on their ignition and energetic behaviors. The objective was to further investigate and charac- terize the thermal stability and thermite reaction properties of the Al–Pd metallic nanothermite with different shell thicknesses, which are extremely important for exploring their applications. Methodology The determination of the stable geometries of interested core–shell nanoparticles at given temperatures was the starting point. In the literature, the main geometrical motifs for bimetallic AgCu, AgCo, AgNi, and AuCo clusters were found to be fcc nanocrystals, icosahedra, decahedra, and polyicosahedra (Bochicchio and Ferr- ando 2013; Laasonen et al. 2013). And it was found that centered cores are possible in icosahedral nanoparticles with small core sizes. For pure palladium clusters, calculations by Jennison et al. (1997) found that fcc structures are more favored in the Pd system than in the Ag system. Experimentally, the thiol-passivated Pd nanoparticles with the diameters of 1–5 nm exhibited a variety of structures, ranging from fcc (simple and twinned), to icosahedral, to decahedral, to amorphous ones (Jose-Yacaman et al. 2001). The stable geometries of the Pd clusters are, however, needed to be further investigated. (Baletto and Ferrando 2005). Due to the lack of literature data, the initial structures of these Al– Pd core/shell nanoparticles were built from an Al core with a radius of 20 A˚ (with 2,243 Al atoms) and a Pd shell with different thicknesses: 5, 6, and 7 A˚ (with 2,174, 2,774, and 3,238 Pd atoms, respectively). Both the Al core and the Pd shell were empirically constructed from their fcc structures when spherical nanoparticles were configured. Note that the excess energy was allowed within the initial geometries of three core–shell nanoparticles and lattice mismatch between metals was possible. Those initial geometries were further relaxed in order to remove the excess energy, as stated later on. These Al–Pd systems were denoted as AlPd5, AlPd6, and AlPd7, respectively, in this work. Figure 1 shows the constructed AlPd5 nanoparticle with a fcc structure. Note that this computational structure does not necessarily represent the optimized structure of the corresponding nanopar- ticle at different initial temperatures. It would experi- ence a molecular restructuring process within a buffer period at the specified initial temperature. For all three nanoparticles, the classic MD simulations were per- formed with three initial temperatures, i.e., 500, 600, and 700 K, which are close to the reported melting point of Al nanoparticles (e.g., 620 K for 4.4 nm pure Al nanoparticles, Song and Wen 2010) and are lower than 900 K above which the one-stage reaction was observed (Nguyen et al. 2011). In detail, the following three-stage procedure was employed in each temperature (initial) simulation, from the initial geometries of the AlPd5, AlPd6, and AlPd7 nanoparticles. Molecular dynamics relaxation/optimization Three initial geometries with the different shell thicknesses (hence the different atomic Al–Pd ratios) were relaxed, by the applied force field, at 0 K in order to represent the equilibrated structures of these Al core/Pd shell nanoparticles at 0 K. The buffering period was set to 5 ps. The major purpose of this step is to relax the empirically constructed Al–Pd core/shell nanoparti- cles to these equilibrated structures which satisfy the applied force field. This task was conducted at 0 K and mainly resulted in changing of lengths of bonds and, subsequently, the remove of the excess energy within the nanoparticle. For example, the Pd–Pd length, which was originally set to 2.751 A˚, was changed to 2.803 A˚ in the core–shell AlPd7 nanoparticle. As shown in Fig. 2, the relaxed AlPd7 nanoparticle still shows the fcc structures of the core and the shell. J Nanopart Res (2014) 16:2486 Page 3 of 12 2486 123 These slightly adjusted bonding lengths between atoms do not modify the crystalline structure signif- icantly. Moreover, since the shell number of the Pd shell (2–3) is less than the shell number of the Al core (7–8), the morphological instability is less notable according to the literature (Bochicchio and Ferrando 2013). Thirdly, because the thickness of the Pd shell is quite small in comparison with the diameter of the Al core, the lattice mismatch between two components is not expected to result in a large strain in the relaxed geometry and hence will not affect significantly the thermal behavior of the nanoparticles. Molecular dynamics heating The produced nanoparticles from the first stage were heated up to the specified ignition temperature Tig (also called the initial temperature) of 500, 600, and 700 K, respectively, over a given period of time tig. Two different heating periods were tested, i.e., 0.5 and 1 ps, in order to investigate the effects of different heating powers (of 1,200–2,140 and 614–1,161 eV/ps, respectively, as shown later in Table 2). Note that in practical, ignition can occur at the room temperature at which some external energy source such as a laser and an electrical arc is utilized to initiate the nanothermite reaction. This work, however, focused on the thermal ignition through heating up the nanoparticle. Depen- dent on the amount of thermal energy added, the ignition may occur or not. The heating rates investi- gated here (about a few 100 K/ps) are greater than the heating rates of a few or hundreds degrees per minute reported in the regular thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) Fig. 1 The initial structure of the AlPd5 core/shell nanoparticle (before optimization using buffering) Fig. 2 The relaxed structure of the AlPd7 core/shell nanoparticle (after optimization at 0 K) 2486 Page 4 of 12 J Nanopart Res (2014) 16:2486 123 studies (Dreizin 2009). But they are comparable with the heating rates applied during pico- and femtosecond laser heating, which ranged from 0.95 to 1,290 K/ps, when the kinetic superheating and melting of alumi- num nanolayers were studied (Hwang and Levitas 2013). In addition, it was reported that, when the nanothermite is ignited by a detonation front produced by an ignition primer or an expanding fire ball generated by a powerful explosive, the heating rates could reach the order of 106–107 K/s (Dreizin 2009). The canonical ensemble MD (NVT) calculations were employed at this stage. For different simulation cases, both the heating period tig and the initial temperature Tig determine how much thermal energy was added into the nanoparticle. Excess thermal energy provided at a given temperature could result in an energized structure which subsequently initiates alloying between Al and Pd. As will be discussed later, at the same initial temperature, the total energy level of the nanoparticle is controlled by the duration of the heating process. Molecular dynamics production During this stage, depending on whether the ignition has occurred or not from the heating stage, the phase change of the molecular structure could be observed when the particle temperature was increased from Tig. This process was described by the microcanonical ensemble MD (NVE) simulation. The three nanopar- ticles were isolated from the surroundings, and the total energy levels associated with these nanoparticles were fixed until the end of the simulation. No periodic boundary condition was applied at this stage. For these aforementioned three stages, the MD time step was set to 1 fs. During the heating period of tig, when the initial system was heated up from 0 K to the specified temperature Tig, the thermal energy absorbed by the Al– Pd nanoparticle was calculated from the change in the total energy (the sum of kinetic and potential energy) of the corresponding nanoparticle. Generally speaking, this heating process would produce energized nanoparticles at the same temperature Tig, but with the different potential energy, dependent on the additional thermal energy absorbed through heating (e.g., the excess energy added to the nanoparticle over tig = 1 ps, compared with the energy adsorbed over tig = 0.5 ps). Results and discussion Thermal energy added during molecular dynamics heating As mentioned earlier, the optimized initial geometries of the three Al–Pd nanoparticles were heated from 0 K to the desired temperature Tig over the specified period of tig. During this process [with the canonical ensem- ble MD (NVT) simulation], a certain amount of thermal energy Q was transferred from the surround- ings to the system, while the nanoparticle structure can be energized by this energy. The value of Q, in eV, is calculated as Q ¼ Etot Tig   Etotð0Þ ð1Þ where Etot is the total energy of the nanoparticle at the specified temperature. This total energy includes both kinetic and potential parts produced from the heating process. These Q values and the corresponding heating powers P (in eV/ps) were calculated for the three nanoparticles with the different shell thickness and over two heating periods, as shown in Table 2. Table 2 Thermal energy Q (in eV) and heating power P (in eV/ps) transferred to the Al–Pd core/shell nanoparticle during the heating process Tig AlPd5 (ps) AlPd6 (ps) AlPd7 (ps) 0.5 1.0 0.5 1.0 0.5 1.0 500 K Q 601.3 613.8 661.2 690.5 722.1 752.6 P 1,202.6 613.8 1,322.4 690.5 1,444.2 752.6 600 K Q 750.4 778.9 823.5 865.7 890.4 938.4 P 1,500.8 778.9 1,647.0 865.7 1,780.8 938.4 700 K Q 889.9 948.4 993.8 1,069.6 1,070.4 1,161.0 P 1,779.8 948.4 1,987.6 1,069.6 2,140.8 1,161.0 J Nanopart Res (2014) 16:2486 Page 5 of 12 2486 123 Table 2 clearly shows the larger nanoparticle absorbed more thermal energy during the heating process, due to its greater specific heat value. And a higher Tig resulted in a greater thermal energy transfer. In addition, when the heating period tig was increased from 0.5 to 1 ps, these nanoparticles become much energized due to the excess amount of thermal energy. Note that for AlPd5, a significant amount of excess thermal energy was supplied to the nanoparticle at Tig = 700 K (about 58 eV more) than the energy supplied at Tig = 500 K (about 13 eV more) when the heating period changed from 0.5 to 1 ps. A similar observation was made for the two other nanoparticles. For AlPd7, the amount of excess thermal energy supplied to the nanoparticle is about 91 eV at Tig = 500 K when the heating period increased from 0.5 to 1 ps. Generally speaking, the MD heating process, through adding different amounts of thermal energy into these nanoparticles in varying specified heating periods, can be viewed as a numerical ignition process during which an external energy was supplied to nanoparticles in order to overcome the activation energy of the thermite reaction at Tig. If the initial temperature Tig is high enough and close to the ignition temperature of the corresponding nanoparti- cle, less external energy is required to initiate the thermite reaction (