In Situ Imaging of Particle Formation and Dynamics in Reactive Material Deflagrations

Kyle Sullivan (16-FS-028)


Reactive composites utilizing nanometer-scale particles have been the topic of extensive research for the past two decades. A motivating factor is that as the particle size is decreased, the mixing scale between constituents is greatly reduced, which has long thought to increase the rate of chemical reaction. While a general trend of increased reactivity with reduced size has been seen for mixtures of metal and metal oxide reactive materials (thermites), some results have demonstrated diminishing returns as the particle size is decreased. Recent results have shown that nanoparticles, which are typically aggregates of several primary particles, can undergo very rapid coalescence to form micrometer-scale particles once a critical temperature is reached. Experiments to date have been performed on very small sample masses, and sometimes under vacuum—conditions that are not representative of the environment during a deflagration. In this feasibility study, 100-mg powdered thermite samples were ignited and reacted in custom long acrylic tubes. We performed x-ray imaging at Sector 32 of the Advanced Photon Source at Argonne National Laboratory to image the particle field as a function of distance and time as the rarefied particle cloud expanded and flowed down the tube. Five different thermite formulations were investigated: aluminum and copper(II) oxide (Al/CuO), aluminum and iron(III) oxide (Al/Fe2O3), aluminum and tin oxide (Al/SnO2), aluminum and tungsten trioxide (Al/WO3), and aluminum and molybdenum trioxide (Al/MoO3), along with aluminum and copper oxide formulations with different sizes of aluminum particles ranging from 80 nm to approximate 10 mm. The results clearly show that the sample powder reacts and unloads into a distribution of larger micrometer-scale particles (approximately 5–500 mm), which continue to react and propagate as the particle-laden stream flows down the tube. This was the first direct imaging of the particle field during a thermite deflagration, and gives significant insight into the evolution of reactants to products. Analysis of phase is currently being pursued to determine whether this method can be used to deduce reaction kinetics.

Background and Research Objectives

Thermites are mixtures of a metal and a metal oxide, which exothermically react upon ignition to release chemical energy in the form of heat and/or pressure. The classical use of thermite dates back to 1904 and involved the reaction between metallic aluminum and iron oxide to produce molten iron metal to weld railroad ties together in Hungary. This method is commonly referred to as the Goldschmidt method, named after the German chemist Hans Goldschmidt. In 1995, Aumann et al. demonstrated that the reactivity in thermite formulations of aluminum and molybdenum trioxide was observed to be dramatically increased by over three orders of magnitude when nanoparticles were used in place of micrometer-sized constituents.1 Additionally, thermochemical calculations performed by Fischer and Grubelich showed that some formulations of thermite have a higher mass and/or volumetric energy density than even the nitroamine CL-20, one of the highest-power explosives used today.2 In energetic material applications, it is the rate of energy release, or power, of the chemical reaction that ultimately quantifies its performance and applications.

By using nanoparticles, it has long been thought that the power in thermites could be increased to a level in which these materials could be competitive with high-explosive applications. However, there have been several recent studies showing diminishing returns in reactivity with particle size. Researchers have suggested that a problem with nanoparticles, which are typically aggregated, is that upon melting the surface tension forces serve to drive particles to coalesce into larger particle sizes. This has been directly measured to occur in aluminum aggregates in as little as 25 ns using the dynamic transmission electron microscope at Lawrence Livermore.3 This morphology change results in a decreased surface area, and has been thought to limit reactivity. Dynamic measurements like this yield valuable information of the length scales and timescales associated with reaction. However, the measurements are difficult to make because of the fine spatial scales and timescales associated with these processes.

Our primary goal was to directly image the reaction of a thermite during a deflagration using the collimated x-ray beamline at the Advanced Photon Source. A 100-mg sample of thermite powder was ignited in a long, clear tube and allowed to flow in one direction down the tube as it reacted. The time-resolved particle field was then imaged perpendicular to the direction of flow, and at various distances down the tube corresponding to different times after initiation. The imaging technique allowed us to measure the unloaded particle size and, additionally, to observe phase and dynamics occurring within these particles as the reaction proceeded. The results are the very first of their kind, and significantly enhance our understanding of the dynamics occurring in thermites.

Scientific Approach and Accomplishments

We used a Livermore-developed method known as the “extended burn tube" for testing the reactivity of thermites, as shown in Figure 1.4 A 100-mg powdered charge of a thermite is loaded into the closed end of a long, acrylic tube. An ignition wire is inserted into the powder, and resistively heated with a fast current pulse. This ignites the material, which subsequently leads to unloading and flow of a rarefied powder cloud as the material reacts and expands down the tube. It is assumed that there is a direct linkage between spatial position down the tube and reaction time. We performed x-ray imaging normal to the flow direction, and time-resolved movies of the particle field were measured. The tube could then be translated, and the experiment repeated to examine different positions. Because of the high reproducibility of the burn tube experiment, the results could be stitched together from separate experiments to draw conclusions of the overall deflagration.

Figure 1. (a) schematic of the extended burn tube test. (b) still shots of the reaction igniting and propagating down the tube. (c) analysis of burn time by plotting the optical intensity versus time and measuring the full width at half maximum of the intensity. (d) analysis of the flow velocity and quenching distance.

Figure 1. (a) Schematic of the extended burn tube test. (b) Still shots of the reaction igniting and propagating down the tube. (c) Analysis of burn time by plotting the optical intensity versus time and measuring the full width at half maximum of the intensity. (d) Analysis of the flow velocity and quenching distance.

We employed the extended burn tube setup at the Advanced Photon Source twice, once in March 2016 when the beam was operating in 24-bunch mode, and once in July 2016 while the beam was in hybrid-singlet mode. In 24-bunch mode, we found that the beam intensity was insufficient to yield high-quality images from a single pulse. Instead, multiple exposures were required but, because of the high particle velocity in the tube, this led to blurring of the images. In the second visit, the beam was in hybrid-singlet mode and provided sufficient energy to image from a single pulse. A schematic of the experimental setup is shown in Figure 2. We performed the x-ray imaging perpendicularly to the flow, and a high-speed camera was used to record the luminous streak as it flowed down the tube. The tube was imaged at different positions after initiation.

Figure 2. schematic of the experimental setup at argonne's advanced photon source. the reaction is initiated at the end of the tube, and x-ray imaging is performed normal to the direction of flow.
Figure 2. Schematic of the experimental setup at Argonne's Advanced Photon Source. The reaction is initiated at the end of the tube, and x-ray imaging is performed normal to the direction of flow.
Several representative images of the collected data are shown in Figure 3. A time-resolved sequence of the aluminum and iron oxide reaction is shown, and one can see the formation of large, spherical particles as the powder unloads and rarefies. Within the particles, one can observe the evolution of dark regions which are, presumably, the formation and phase separation of the molten iron product from the moltenaluminum oxide. Four different thermite images collected at late times after ignition are shown as well. One common observation is that all particles are much larger than the initial, nanometer-scale particle sizes used in the formulations. Additionally, one can observe morphological differences in the particles for different formulations. In particular, the aluminum and tungsten trioxide system appears most irregular in shape, and we hypothesize this is because of the high melting point of the tungsten product, which hinders flow and phase separation. Finally, we observed some dynamic phenomena such as the formation and collapse of gas bubbles within these particles (also shown in Figure 3).
Figure 3. (a) time-resolved sequence of the aluminum iron oxide (al/fe<sub>2</sub>o<sub>3</sub>) thermite. (b) comparison of different thermites in the unloaded state. (c) sequence of the dynamic collapse of a bubble in aluminum iron oxide.
Figure 3. (a) Time-resolved sequence of the aluminum iron oxide (Al/Fe2O3) thermite. (b) Comparison of different thermites in the unloaded state. (c) Sequence of the dynamic collapse of a bubble in aluminum iron oxide.
Because of the very large number of samples and images collected at the Advanced Photon Source, it was necessary to automate the data analysis. The movies were processed by first background subtracting a blank movie from the image deck. An edge-finding routine was then performed using the ImageJ (an open-source image-processing program) to outline the particles, upon which the average particle diameter could then be extracted. The number, area, and volume-averaged particle diameter were all measured. However, the volume-averaged diameter was chosen for reporting because it represents the average particle size with respect to mass, which is most relevant. The sample analysis routine is shown in Figure 4. Also included is a method for looking at phase within the particles. The analysis of phase is currently under investigation to determine whether we can extract chemical kinetics from such analysis.
Figure 4. schematic of the image analysis including background subtraction, particle size analysis, and phase identification.
Figure 4. Schematic of the image analysis including background subtraction, particle size analysis, and phase identification.
The average diameter (mass basis) is plotted as a function of time, thermite system, and position of data acquisition in Figure 5. All positions are reported as a percentage of their expansion distance (xend) from Figure 1(d). It can be seen that all thermites produce particles at least tens of micrometers in size. This is despite the fact that the average primary particle sizes were less than 100 nm for all formulations. Between when the particle front arrives at 0 ms, and when the particle size plateaus, one can observe a rising slope in particle diameter with time. This rise is sharper at positions closer to the initiation point (1 cm and 25% data), and rises more slowly at larger expansion distances (75–100%). We believe this rise is because of flow lag—that is, larger particles take longer time to accelerate to a steady state velocity compared to small particles, and so the distance between the fine and coarse particles grows with time.
Figure 5: average particle size as a function of position, time, and oxidizer.
Figure 5: Average particle size as a function of position, time, and oxidizer.
We also observed that in all cases, interesting dynamics could be observed within the particles. This included the formation and growth of two-phase particles. The darker phase is assumed to be the molten metal product (tin, iron, tungsten, molybdenum, or copper) as the oxidizer liberates the oxygen and proceeds towards the equilibrium phase. The formation and dynamics of this metal product provides some indication of the reaction kinetics. However, further analysis is being performed to verify this. One complication is that one cannot easily distinguish between what is reacting and what has reacted but not cooled yet. Phase separation will occur if the materials are immiscible and if the temperature is hot enough. Therefore, decoupling that from chemical kinetics is the major challenge at this point. In any case, these movies have provided significant insight into the size and reaction pathways of thermite reactions.

Impact on Mission

Energetic materials are of interest to several programs at Lawrence Livermore, and our investigation of formulations for thermites support the Laboratory's core competency in advanced materials and manufacturing, as well as to the strategic focus area in stockpile stewardship science. Our feasibility study results can help inform the design of engineered architectures for custom energy release, and the work also applies to the development of advanced diagnostics for probing materials under extreme conditions. This work is relevant to the NNSA and Department of Defense Joint Munitions Technology Development Program, which investigates structural energetics to add functionality to munitions. Structural energetics aligns with the Livermore strength in structural materials, whose purpose is to develop structure–property relationships, including dynamic properties. Our work has the added benefit of increasing Livermore's exposure at the Dynamic Compression Sector, a new NNSA-sponsored research facility at Argonne, which will continue research on reaction of thermites to investigate the processing, ignition, and combustion characteristics of these formulations.


With this feasibility study we collected very-high-quality movies of deflagrating systems for the first time. This work has now transitioned to programs at Lawrence Livermore. While the exact details of the thermite reaction still remain challenging to extract from the movies, the data represents an unprecedented view of the energetic reaction and provides significant insight for future experiments. As part of future work, we plan to image thermites with different morphologies to see what ultimately controls the unloaded particle size. Phase analysis will also continue, and the goal is to determine whether the absorption coefficient within these particles can be used as a metric of reactivity. If so, this would represent a generic solution for extracting chemical kinetic parameters from composite systems, and would be a highly significant result. We have begun a collaboration with the New Jersey Institute of Technology to look at milling of custom thermites, and both the Joint Munitions Program and the Defense Threat Reduction Agency are supporting projects in FY17 for use of thermites and other reactive materials as structural energetics. Two postdoctoral researchers have been hired within the Laboratory's Materials Science Division who are, or will be, working with thermites. A third postdoctoral researcher and a graduate student employed in the materials engineering division have also recently begun examining three-dimensional printing of thermites.


  1. Aumann, C. E., et al., “Oxidation behavior of aluminum nanopowders,” J. Vac. Sci. Tech. B 13(3), 1178 (1995).
  2. Fischer, S. H., and M. C. Grubelich, A survey of combustible metals, thermites, and intermetallics for pyrotechnic applications, 32nd AIAA/ASME/SAE/ASEE Joint Propulsion Conf., Lake Buena Vista, FL, July 1–3, 1996.
  3. Egan, G. C., et al., “In situ imaging of ultra-fast loss of nanostructure in nanoparticle aggregates,” J. Appl. Phys. 115(8), 10.1063 (2014).
  4. Sullivan, K. T., et al., "Quantifying dynamic processes in reactive materials: An extended burn tube test." Propell. Explos. Pyrotech. 35(1),1 (2014).

Publications and Presentations

  • Reeves, R. V., et al., Direct x-ray imaging of Al thermite reactions at the particle scale. (2016). LLNL-POST-693797.