Challenges of Small-Scale Safety and Thermal testing of improvised explosives: Results from the integrated data collection analysis (IDCA) program proficiency test

The IDCA Program has been conducting a Proficiency (Round Robin) Test on the application of Small-Scale Safety and Thermal (SSST) testing to Home Made or Improvised Explosives (HMEs). This Proficiency test has been designed to test the accuracy and relevancy of SSST testing among explosives testing laboratories (3 DOE and 2 DoD), where the testing is performed on the same batches of materials (17 HMEs and 2 Standards), prepared the same way. The results so far have indicated that standard testing methods are not adequate for HMEs, as many conflicting and inconclusive results have been documented: Impact sensitivity non-predictively affected by testing conditions; Detection of positive reaction (go/no-go) has too much variability; Thermal testing has sampling issues As the IDCA continues to compare and evaluate results from the Proficiency Test, many issues are beginning to coalesce about the application of traditional SSST Testing methods to HMEs. Many of the issues show that traditional methods used for military explosives MUST be modified before meaningful results can be obtained for HMEs. The IDCA is finding if traditional methods are not revised, testing can give misleading results that could lead to developing handling practices that are not adequate for working safely.

INTRODUCTION A critical aspect in developing forensics methods for explosives is understanding the chemical and physical pro-cesses that occur when the energetic material reacts, such as in a detonation. Military explosives are well character-ized and offer pre--and post--blast signatures for forensics and attribution to some extent. Improvised materials or homemade explosives (HMEs) are less well characterized and little is known of their behavior and even less is known about the required forensics if an "event" oc-curs. HMEs have few documented pre--blast signatures and essentially no documented post--blast signatures needed for forensics. The first step in the process of developing signatures is identifying how to handle HMEs properly so forensic methods can be developed. The IDCA Proficiency Test [1] addresses many of the issues regarding safety test-ing these materials (important information for first re-sponders, EOD techs, three letter agencies, facilities that test performance) and developing the accurate and correct information about the HME.
SSST tests are critical and usually a first step in deciding whether an energetic material is safe to handle [1,2]. These tests were designed for explosives to indicate sensitivity of the material to specific conditions-drop hammer for im-pact sensitivity; friction for shear force sensitivity; electro-static discharge for spark or static sensitivity; Differential Scanning Calorimetry (DSC) for thermal stability; many others for specific types of reactivity.
SSST testing is performed when: 1) the sensitivity of material is not known, 2) direct handling is desired, 3) the performance of an explosive (e.g., release energy and veloc-ity of detonation) is not known (usually very small quanti-ties of less than 1 gram are tested as a first step), 4) synthe-sis/formulation is changed, and 5) scale--up is required (showing the effects of preparation equipment). Results determine (depending upon interpretation) whether a ma--Sponsor: Department of Homeland Security terial can be directly handled, remotely mixed, or requires complete robotic handling.
The IDCA has been conducting testing on a series of HMEs, utilizing standard SSST testing practices as applied to military explosives [3]. The results so far have indicated that standard testing methods are not adequate for HMEs, as many conflicting and inconclusive results have been documented. In this report, several of these issues are de-scribed.
II. SMALL--SCALE SAFETY AND THERMAL TESTING SSST testing as applied to the IDCA Proficiency Test has been reviewed elsewhere [3--5]. Briefly, impact sensitivity is measured by drop hammer where the data are analyzed by the Bruceton [6] or Neyer [7] method. Friction sensitiv-ity is measured by BAM or ABL friction systems where the data are analyzed by the Bruceton method or threshold initiation method (TIL) [8]. Spark sensitivity is measured by ABL electrostatic discharge where the data are analyzed by the TIL method [8]. Thermal sensitivity is measured by differential scanning calorimetry (DSC) and the thermal response of the material is analyzed by heat flow into and out of the sample [9]. Note that for a specific material, each laboratory used the material distributed from the same batch, prepared, mixed and handled the same way.

III. IMPACT SENSITIVITY NON--PREDICTIVELY AFFECTED BY TESTING CONDITIONS
Impact sensitivity is an assessment of how sensitive the material is to being dropped or struck. A sample, 35 mg, is placed on an anvil in the drop hammer apparatus. Solid samples are held on sandpaper and a striker rod is placed on the sample. Drop weight (1 to 2.5 kg) is dropped on the striker rod from variable heights until a reaction is detect-ed. The reaction is a pop, flash or smoke (does not neces-sarily mean a detonation). The drop height is adjusted during a test to map out the reaction region near the 50% reaction level of the material, designated as DH50. The higher the DH50 value, the less sensitive the material is to impact.   Fig. 1 shows impact sensitivity testing of selected HMEs at six different experimental conditions. Material DH50 values are set relative to an RDX standard (the DH50 of standard is subtracted from the DH50 of the material set-ting the standard to 0). A positive DH50 value means the material is more stable than the standard; a negative DH50 value means the material is less stable than the standard. The standard is tested under the same conditions at which the sample is tested. 3 mixtures were tested in drop ham-mer at two different conditions. The experiments are: 1. KClO4/Dodecane (120--grit sandpaper) [10]; 2. KClO4/Dodecane (180--grit sandpaper) [10]; 3. KClO3/Dodecane (120--grit sandpaper) [11]; 4. KClO3/Dodecane (180--grit sandpaper) [11]; 5. KClO4/Al (120--grit sandpaper) [12]; 6. KClO4/Al (180--grit sandpaper) [12]. Figure  1 shows both mixtures 1 and 2 being less sensi-tive than the standard, but with 1 being much less sensi-tive than 2; mixture 3 being less sensitive to than the standard; 4 being more sensitive than the standard; mix-ture; 5 being much less sensitive than the standard; and 6 being slightly more sensitive than the standard. Because the only difference in these mixture pairs is the use of 120-vs. 180--grit sandpaper to hold the sample, and the RDX standard responds in a different way than the mixtures, no relative or absolute assessment of the sensitivity is possi-ble. This difference in response of a specific mixture to sandpaper as compared to the RDX standard is clear evi-dence that measurement of the impact sensitivity of the HME by standard methods needs scrutiny. Although it is not clear what property is causing this difference in re-sponse, it probably relates to the details of the sandpapers. Fig. 2 shows scanning electron micrographs (SEMs) and photographs of the two sandpapers, showing some of the obvious compositional differences. Some possible relevant properties are: a mismatch in the relative grain size of the sandpaper as compared to the particle size of the mixture (see below); 120--grit sandpaper is a wet/dry type sili-con/carbide (Si/C), while the 180--grit is a dry--only garnet; the grit of the 120--grit sandpaper is harder than the grit on the 180--grit sandpaper, (9 to 10 vs. 6.5 to 7.5 on the Mohs hardness scale [13], respectively); the glue on the wet/dry paper is an insoluble resin, while the glue on the 180--grit sandpaper is a hide glue; the wet/dry paper is approxi-mately twice as thick as the dry only paper (0.406 vs. 0.229 mm thick, respectively). All of these factors could contrib-ute to the differences seen in the impact data when using 120--grit vs. 180--grit sandpapers. In Fig. 1, experiments 5 and 6 strikingly contrast the ef-fect of the two different sandpapers. Fig.  3 illustrates this difference may be caused by the mismatch of the relative grain size of the sandpaper to the particle size of the mix-tures. The figure displays the particle size distribution (by laser light scattering) for both the KClO3 and KClO4 starting materials. The KClO4 distribution is significantly smaller size as compared to the KClO3 distribution. The mean di-ameters of the grit particles of the 120-- and 180--grit sand-papers based on the CAMI specification are also shown [14]. For the KClO3 mixtures, both the 120--and 180--grit average size fall in the size range of the oxidizer. For the KClO4, only the 180--grit average size falls in the particle size range of the oxidizer. In the mixture cases, the 120-grit and the KClO4/Al mixture are greatly mismatched and the fine powder may fall between the grains of the sand-paper, preventing much contact of the striker. In the 180-grit case, the grit of the sandpaper and the particle size of the KClO3/Al mixture are closer in size (by virtue of the KClO3 size) allowing for better contact. A similar grit size particle size distribution relationship is seen when com--paring particle size distributions as measured by Coulter Counter [12].

MUCH VARIABILITY
The method of detection for a positive/negative reac-tion in SSST testing is highly dependent upon the testing facility. The general method is by observation, typically done by the operator of the equipment. Use of sound me-ters, cameras and chemical reaction product analyzers are some of the more sophisticated but less used methods. Most positive reactions in impact, friction and spark are marked by a spark, flame, smoke, discoloration and/or sound. These are not trivial to distinguish from back-ground because the testing equipment can make substan-tial noise, even in blank testing.
BAM friction is a common method for determining fric-tion sensitivity. The sample (~ 5 mg) is held on a ceramic plate, and a rounded ceramic pin is dragged across the plate, through the sample. Variable force is applied by adding weight to the arm holding the pin, and this weight is varied to cause a reaction. The reaction is usually a pop, or smoke, or jetting from the sample. The sensitivity is reported either as TIL or F50. TIL is the load (kg) at which zero reactions out of twenty or fewer trials with at least one reaction out of twenty or fewer trials at the next high-er load level occurs. F50, in kg, is determined by a modified Bruceton method, load for 50% probability of reaction. 1. KC = KClO3, KP = KClO4; 2. Threshold Initiation Level (TIL) is the load (kg) at which zero reaction out of twenty or fewer trials with at least one reaction out of twenty or fewer trials at the next higher load level; 3. F50, in kg, is by a modified Bruceton method, load for 50% Reaction; 4. ND = Not determined; 5. KClO3 separated through a 100--mesh sieve; 6. KClO3 separated through a 40-mesh sieve. Table  I shows the BAM testing results (TIL and F50) for selected materials by LLNL, LANL, and IHD. In most cases, LLNL testing results indicate a more stable material to fric-tion than the results from the other laboratories, suggest-ing a systematic issue for at least one of the testing labora-tories. However, for the BAM friction testing, all three par-ticipants have various versions of the same testing equip-ment and use observation as the method for detection. Fig. 4 shows the configuration of the IHD and LLNL BAM Friction testing systems. The configuration of the IHD system shows a vent hose that removes gases formed during testing. The configuration of the LLNL system shows complete enclosure of the system also to vent gases formed during testing. The enclosure was designed to con--Size (microns) % PASS % CHAN 120--grit sandpaper CAMI specification particle diameter 180--grit sandpaper CAMI specification particle diameter trol the atmosphere in the testing room, but it also damp-ens noise from a positive reaction. This affects determina-tion of positives events that generate sound (pop or crack-le), and to a lesser extent, flashes. LANL has a system that is similar to IHD and the results reflect this effect. The core aspect of this issue is the detection method for positive reaction relies on the senses of the operator of the equipment. Because that person must make a decision based on hearing or seeing the event, the detection be-comes subjective depending upon how acute these senses are in the individual. There is certainly variation from op-erator--to--operator that adds to the variability of the de-termination of a positive reaction. There are also no standards for testing the ability of operators to hear or see positive reactions. Additionally, the secondary contain-ment is optional to the standardized BAM friction system, and is custom designed. This only adds more variability in the detection. It appears that the LLNL system inhibits the ability to determine positive events above background, and therefore the material seems less sensitive relative to de-terminations by IHD and LANL. As a result of issues such as these, efforts are on--going to make the decision more equipment based [20].

V. THERMAL TESTING HAS SAMPLING ISSUES
Thermal stability in SSST testing is commonly deter-mined by DSC. This technique has advantages as it uses a very small sample size (< 1 mg), can be a fast survey and can be automated. The sample is placed in a sample holder (open or sealed) and is heated, usually at a constant heat-ing rate (10°C/min). The heat flow into and out of the sample is measured. Heat flow into the sample indicates an endothermic response and the material is generally not considered energetically hazardous. Heat flow out of sam-ple indicates an exothermic response and suggests an en-ergetic material if the response is large.
For standard military type explosives, this technique is reliable. However, for HMEs, the application to mixtures, both solid--solid and solid--liquid, has shown issues in re-producibility. Fig. 5 shows the DSC of KClO3/sugar mixture heated at 10°C/min [16]. Three exothermic features are visible which have been assigned previously [16]-Ex1, with Tmax at ~ 180°C is the KClO3/sugar mixture reacting (sugar melts and then mixes); Ex2, with Tmax at ~ 220°C is the sugar carbonizing (sugar that did not react); and Ex3, with Tmax at 340°C is the KClO3 melting and reacting with residual carbon. In Fig. 5 [21] shows LANL photographs of a 0.15 mg sam-ple in the DSC sample holder before and after exposure to a 10°C/min thermal heating ramp from room temperature to 400°C. The close--up taken before testing shows that the sample does not fully cover the surface of the sample hold-er. (Note that very small sample sizes must be used for energetic materials to avoid bursting the sample holder and potentially damaging the DSC equipment.) Because this is a mixture of two different solids, these areas could be materials where the ratios of the two solids are not uni-form. This is substantiated in the close--up of the sample after thermal testing where there are regions of different color reflecting possible differing localized reactions. For example, some of the areas are still white while some show brown and even black. It is postulated that the white re-gions are oxidizer rich because the starting components are white, while the brown and black regions are fuel rich because the thermal reaction products show carbonization of the fuel. The DSC profile of this sample suggests that these correspond with exothermic responses Ex1, Ex2, and Ex3.
In this particular case, Ex1 is the exothermic feature of most importance because it is the shows the lowest tem-perature for instability of the mixture. In the experimental series referred to above, as the sample size increased, Ex1 became the dominant exothermic feature, although other experimental complications were pushing the reliability of the measurement. This suggests that if a perfectly mixed sample could be obtained, there would be only Ex1. How-ever, this is not the case because of the experimental con-siderations for this material. KClO3/dodecane [10] and KClO4/dodecane [11] mixtures also exhibit sampling issues that produce variable DSC results, indicating that the standard conditions for DSC screening applied to military materials may be misleading for HMEs.

VI. SUMMARY
SSST testing techniques have been used for military explosives to develop safe handling practices. These tech-niques have also been applied to HMEs, using the same protocols for testing. The results have shown that for HMEs, the testing techniques are not as reliable in estab-lishing safe handling conditions because the results are, in many cases, ambiguous or confusing. The physical and chemical properties of the HME are the sources of the is-sues. These issues can produce misleading results for both absolute and relative stability. The problem caused in thermal testing by DSC is an example of misleading results when looking for absolute stability. The problem caused by different sandpapers is an example of misleading re-sults when looking for relative stability.
The above issues are not well appreciated by the ener-getic materials community probably because of the lack of experience with HMEs. However, the results reported here are just some of the many examples that have been ob-served in the IDCA testing of HMEs. The community needs to be aware that the SSST testing methods applied to mili-tary explosives may yield misleading results when applied to HMEs. Standardization is the solution for the testing community-develop testing methods, materials stand-ards and equipment calibrations for HMEs.