The Need for Improved Explosive Safety Testing During Design and Development

In 1956, an explosives worker at Los Alamos was scooping a large mass of thallium azide out of a porcelain tray with a metal scoop. The material detonated, killing the worker and destroying the bay, as well as a parked vehicle outside. Although the material had undergone safety evaluation using the drop-weight impact test, its behavior differed significantly from typical explosives of the time. Upon drop-weight impact from low heights, partial reactions would often occur, burning part of the sample and producing less noise than complete reactions from drops at greater heights. Lacking quantitative diagnostics to measure sound, the operator mistakenly believed that the material was fully inert at low drop heights, when it had, in fact, partially reacted. After this accident, drop-weight impact tests were given standards for calibration and decibel meters to quantify sound levels. These improvements have endured until the present day. For over 70 years, our go-to experiment for determining explosives safety utilizes this technology from the 1950s. There is an impressive history of scientific progress in the field of the shock initiation of explosives, with continuously updated technology that includes embedded gauge gas gun experiments, optical velocimetry, and gap stick tests. However, relatively little is understood about sub-shock initiation in safety scenarios, which we really should be concerned about. This is in part because during the era of the Manhattan Project and the decades immediately following WWII, the set of available explosives was narrowed down to a handful that displayed acceptable characteristics. Over time their safety characteristics were established somewhat inadvertently through their continuous use, a diversity of testing methods, and occasional accidents. Though we have established physically meaningful quantities to describe explosives performance and shock initiation, such as the Pop [1] plot, the sub-shock regime is so complex that it cannot be described with a simple number, equation, or plot. In Non-Shock Initiation of Explosives, Asay notes, through Smith, that “the sensitivity of an explosive is not a well-defined property of the material, expressible by a single number, say, but is, instead a complex pattern of behavior [2, 3].” As mentioned above, the drop-weight impact test has been the workhorse for new explosive materials. It allows operators to rank the safety properties quickly and cheaply, in small quantities (∼1 g), relative to a well-characterized standard. While useful, the only diagnostic is often something like decibel levels. The derived drop height sensitivity, H50—the value (in cm) at which samples are expected to initiate in 50% of attempts at the same drop height—is not a physically meaningful number that could be useful in predictions or for developing a fundamental understanding of initiation behavior. The proposed mechanisms for initiation in the drop-weight impact test are complex. Wenograd in the 1950s [4] and Field in the 1970s pioneered the use of drop-weight tests with a transparent anvil that showed that initiation typically occurs in the rapid, outward flow when samples are compressed between the striker and anvil [5, 6]. Our team recently showed that under certain drop heights (up to 41 J of energy with a sapphire anvil), PETN only ignited when pinching, misalignment, and/or solid–liquid interactions were present [7]. Similar to the thallium azide in the 1956 accident, some energetics only undergo partial reactions at some drop heights, emitting low sound levels. For example, the explosive 3,3′-diamino-4,4′-azoxyfurazan (DAAF) exhibits a wide range of decibel levels in the test, due to the formation of ignition sites that do not always propagate to reaction [8]. Unsurprisingly, drop-weight impact values vary widely [9], even for the same explosive, at the same site, under the same testing conditions [10]. The drop-weight test is therefore reliable at identifying hazardous behavior in explosives that ignite readily and in which reactions propagate easily. It is much worse at identifying rarer explosives that ignite readily, but with the extent of propagation governed by highly stochastic behaviors. Although such explosives comprise a small fraction of the known compositions, nowadays they may be the ones that go on to result in accidents during scale-up. It is likely not news to the reader that the energetics community has been searching for a new, rapid, cost-effective test that provides more physically meaningful data to either replace or augment the drop-weight impact test. Furthermore, like many before us, we have had a long-standing interest in establishing a connection between the chemistry of explosive molecules and their initiation properties in the sub-shock safety regime. The prevailing assumption is that reactions initiate thermally via localized high temperatures generated in the sample, either directly or indirectly through frictional forces. Therefore, a potential replacement experiment would involve a known temperature input along with measured times to reaction, neither of which are obtained from the drop-weight impact test. Further, some explosive properties, such as the propensity for drop-skid reaction, only become measurable at intermediate to large charge masses, suggesting that the study of explosive formulation safety cannot be performed solely at the small scale [11]. The time-to-explosion can be represented accurately by single-step Arrhenius kinetics. More sensitive explosives (small H50) transition to a thermal explosion at lower temperature inputs than less sensitive ones (large H50). These subsequent tests include differential scanning calorimetry (DSC), Brill's T-jump experiment [14], and Lawrence Livermore National Laboratory's one-dimensional time-to-explosion (ODTX) test [15, 16]. Each of these experiments is useful and necessary for establishing initiation characteristics but involves slower heating (DSC), unconfined geometries less relevant to safety scenarios (T-jump), or larger samples with slower heating (ODTX) and is therefore less applicable for establishing safety properties at the earliest design stage. Nevertheless, time-to-explosion data from a variety of experimental sources have been compiled by Henson and Smilowitz at LANL, which showed that Arrhenius ignition behavior might even extend into the strong shock regime [17, 18]. To obtain useful information about explosive initiation in the safety regime, explosive samples must be heated to about 1000 K at microsecond-millisecond timescales [19]. Wenograd developed a high temperature time-to-explosion test in the early 1960s at the US Naval Ordnance Laboratory, which used a high voltage power supply to pulse explosive samples confined within hypodermic needles. Although he was able to determine time-to-explosion at temperatures exceeding 1000 K, the small needle diameters required that samples be melt-cast into the needles [20]. In other words, solid fills could not be studied. Following Wenograd's work, we developed an experiment called the High Explosive Initiation Time (HEIT) to link heating mechanisms in accident scenarios to models of thermal ignition and growth [21]. Electrical power is rapidly discharged into a steel needle (inner diameter of 0.6 mm) to create resistive heating on microsecond timescales. When the explosive sample inside reacts, it bursts the needle wall, and high-speed imaging is used to determine times to explosion. At a given temperature, say 1100 K, the times-to-explosion rank similarly with drop-weight impact test results [22]. The reaction rates plotted versus inverse temperature result in linear Arrhenius behavior, though the slopes and intercepts of the lines differ from the ODTX results. We believe this is because the HEIT test provides non-uniform heating, where explosive near the needle wall starts to react heterogeneously before the middle of the sample has equilibrated. These observations highlight the importance of knowing the limitations of our methods and observables. However, the rapid heating rate, confinement, and heterogeneous ignition in this test are important because they may mimic accident scenarios better than tests with slower heating. For example, in the slow, minutes-to-hours timescales of the ODTX experiment and DSC, CL-20 appears safer and more thermally stable than HMX and RDX. However, drop-weight impact results (∼250 µs timescales) and the HEIT experiment indicate it is more sensitive than RDX and HMX, which is consistent with historical accidents and violent outcomes in recent testing. Measuring the time-to-explosion kinetics of an energetic material using these empirical tests requires first synthesizing it on small scales. The capability to reliably and accurately predict the reaction rates of energetic molecules and materials prior to their synthesis would allow us to pre-screen molecules and invest time and effort in only the most promising candidates. The concept of the trigger linkage—the weakest bond in the reactant molecule—has been useful in computational studies because insights into reaction rates can be obtained from routine, gas-phase quantum chemistry methods. In addition to investigations of trigger linkage kinetics, our team has led the use of quantum molecular dynamics simulations to calculate the single-step Arrhenius kinetics of the time-to-explosion in the condensed phase. The results of these simulations accurately capture the order of the thermal stabilities of explosives obtained experimentally from Henkin through to the present day with the HEIT experiment [20, 23]. With the ongoing development of fast algorithms for quantum molecular dynamics and new accurate interatomic potentials from the machine learning of ab initio data, we expect such approaches to predicting the safety and performance of explosives to grow in importance over the next few years. The development of small-scale safety testing of new explosives and formulations since the 1950s has greatly reduced the frequency and severity of accidents. From that important point of view, the current suite of tests should not be abandoned prematurely because it has served us very well. However, efforts to understand what those tests are really telling us about the explosive response and developing newer tests that are equally discriminating but that generate physically meaningful results should be the greater goal. In this way, injuries and deaths from new explosive candidates can be further minimized. This work was supported by the Laboratory Directed Research and Development program of Los Alamos National Laboratory under project number 20220068DR. Los Alamos National Laboratory is operated by Triad National Security, LLC, for the National Nuclear Security Administration of US Department of Energy (Contract No. 89233218CNA000001). Approved for public release (LA-UR-25-20751). The data that support the findings of this study are available from the corresponding author upon reasonable request.