Fast energy release from reactive materials under shock compression

It has proven difficult to increase the energy output of conventional chemical explosives due to an inverse relationship between energy release and stability.^1–3 An opportunity to achieve such an increase has emerged due to the notable recent development of reactive nanomaterials (RNMs).^4–7 RNMs address the energy vs stability problem by spatially separating nanometric fuel and oxidizer components. Materials such as Ni/Al, Al/MoO₃, and Al/Teflon, on an energy per unit volume basis, can produce two or three times as much energy as powerful chemical explosives such as TNT (2,4,6 trinitrotoluene) or HMX (cyclotetramethylene-tetranitramine).⁸ For this reason, it might seem that RNMs could improve or even replace organic explosives. The problem is that RNM fuel-oxidizer reactions are too slow because they are ordinarily rate-limited by diffusion-controlled mixing of reactants.^9–11 Additionally, RNMs are generally not efficient generators of the gas-phase products needed to drive detonations. As a point of reference, HMX, a high-performance explosive, has a detonation reaction zone width of less than 50 ns and within that time generates gas-phase products at a pressure of 40 GPa.^12,13

In the present work, we study shock initiation of two kinds of RNMs prepared by arrested reactive milling.⁶ The fuel was oxide-passivated Al and the oxidizers were either Bi₂O₃ or BiF₃. The stable reaction products are Bi and either Al₂O₃ or AlF₃. Since Bi has a boiling point of 1837 K and RNM reaction temperatures are typically in the 3000 K range,⁶ gas may be formed from Bi vapor for both composites. Al₂O₃ has a high boiling point of 3250 K, whereas AlF₃ sublimes at 1545 K, so we expect significantly greater gas formation from the fluoride.

Electron micrographs of the micro- and nanostructure of the RNM powders used here are shown in Fig. 1. They were in the form of micrometer-scale powders where each micrometer-scale grain was a composite of fuel and oxidizer mixed on the nanoscale. Fuel-rich powders with Al/oxidizer mole ratios of 4/1 were prepared in 5 g batches using a SPEX Certiprep 8000 series shaker mill. Starting materials Al, Bi₂O₃, and BiF₃ were obtained from Alfa Aesar (Ward Hill, MA) and were milled for 30 minutes in stainless steel vials with 9.25 mm stainless steel balls, and a few ml of hexane was used as a process control agent. The charge ratio (balls/starting material mass) was 10 for the fluoride and 5 for the oxide. Electron microscope images of the milled composites are shown in Fig. 1. In these backscattered electron images, Al appears darker than Bi. The particle sizes were typically 10–20 μm and the scale of mixing was on the order of 100 nm. The powder samples were loaded into 1 mm diameter, 40 μm deep wells with a small amount (5%) of poly-vinyl alcohol polymer serving as a binder (Fig. 1). The optical images in Fig. 1 show the sample that faces the optical detectors while flyer plates impact the opposite side.

The tabletop shock spectroscopy apparatus diagramed in Fig. 2 has been described previously.^14,15 The flyer plates are Al-1100 disks of 0.5 mm diameter and 25 μm thickness launched by a pulsed laser at 3 km/s, as measured by photon Doppler velocimetry.¹⁶ The visible thermal emission was split 90% to an intensified camera with a 5 ns shutter and 10% to a 32-channel optical pyrometer.¹⁷ The pyrometer, calibrated with a 2960 K source, measures the spectral radiance from 450 to 825 nm every 0.8 ns.¹⁸ A graybody model was used to determine the temperature associated with the spectral radiance.¹⁹ 

Figures 3(a) and 3(b) show the time-dependent spectral radiance from the shocked RNM, where the pressure produced in the RNM by the flyer plate roughly approximates a powerful chemical detonation. Each is the average of 10 or more shots. The spectral radiances show minimal contribution from AlO (∼480 nm)²⁰ and Al emission (∼400 nm) was outside the range of observation. Consequently, the spectral radiances were good fits to a graybody model, as illustrated in Fig. 3(c), where the temperature was 3200 (±30 K) and the error bars represent 95% confidence limits. Figure 3(d) shows the time-dependence radiance (the wavelength integral from 450 to 825 nm) for the RNM and also a control consisting of fine-grained sucrose with a small amount of polymer binder.²¹ Shocked sucrose produces a brief weak burst of triboluminescence. The fact that sucrose, an inert often used as a control in energetic material studies, has emission more than 200 times weaker than the RNM confirms that the high temperatures observed for the RNM represent exothermic fuel/oxidizer reactions.

Figure 4(a) shows time-dependent temperatures from the shocked RNM, which are in the 4500 K range during the ∼4 ns duration flyer plate impact. After the shock has unloaded (e.g., 10 ns), the 4Al/BiF₃ temperature is quite close to the measured flame temperature of 3150 K (Ref. 22) and consistent with calculations of the adiabatic flame temperature of 3500 K.²² 

The temperature histories of the two shocked RNMs are similar until about 30 ns, when a large-amplitude cooling process occurs in the fluoride but not in the oxide [Fig. 4(a)]. Such a high cooling rate is inconsistent with thermal conduction and is caused by fast adiabatic volume expansion.^23,24 Time-resolved images of thermal emission from the shocked RNM at 30 ns in Figs. 4(b) and 4(c) clearly show a significant volume expansion of the fluoride relative to the oxide, far greater than what can be explained by thermal expansion, but consistent with rapid generation of volatile reaction products by the fluoride. Since the volume expansion typically occurs near the speed of sound (μm/ns), it takes tens of nanoseconds to produce the expansion in Fig. 4(c), which is consistent with gas production on the 10 ns timescale.

The rapid energy release from the shocked RNM is striking when compared to RNM combustion in the diffusion-limited regime, for instance, recent results on 8Al-MoO₃.²⁵ Even when the RNM was flash-heated with a short (microsecond) duration plasma discharge, high-temperature combustion took microseconds,²⁵ orders of magnitude slower than the shock-induced reactions.

Since energy release does not occur unless the fuel and oxidizer mix, the 10 ns energy release is clearly the result of nondiffusive mixing at the atomic scale in the shocked microparticles with individual components mixed on the nano-scale. Although the input shock was generated by a planar flyer plate, the complex micro- and nanostructure of the RNM converts the shock into a mixture of compressive and shear waves.²⁶ Shear waves, in particular, are especially efficient in producing plastic deformation that allows fuel and oxidizer to mix. The observed characteristic reaction time thus represents the time necessary for the shock-induced plastic deformation to reach the scale of the constituent particle nano-structure. The deformation of materials might initially result in yielding atomic F acting as an aggressive gas phase oxidizer generating gaseous products that expand beyond the sample well.

These experiments show that the RNM, when shocked at pressures comparable to detonation pressures, can produce energy in the form of heat or work done by rapid volume expansion on nanosecond time scales comparable to shocked conventional explosives such as HMX. The use of higher-valence, fluorine-rich fluorides of volatile metals like BiF₅, SbF₅, SbF₃, InF_4, SnF₄, and gas-generating oxides as oxidizers would likely improve gas generation even further. The gas production may not necessarily be sufficient to produce a detonation from RNM powders alone, but these results do suggest that RNM powders used as an additive might significantly boost the energy of chemical explosives.