Improvised Explosives: X-Ray Detection & Euthectics of Erythritol Tetranitrate

This thesis focuses on the handling and detecting of improvised explosives, specifically erythritol tetranitrate (ETN). ETN is a nitrate ester with the unusual property of melting at a temperature significantly below its decomposition temperature. For this reason the first two manuscripts probe the potential usefulness of ETN as a military explosive and the long-term stability of this material. The third manuscript examines x-ray detection, a technique that readily reveals covert military explosives. Discussed are ways to evaluate detection capability without actually employing the explosive. With the recent availability of erythritol as an artificial sweetener, the synthesis of ETN has become economically feasible. This dissertation examines eutectics of ETN, to establish if it can be a useful component in melt-casted explosive formulations. To this end we report three novel ETN-energetic eutectics with trinitroazetidine (TNAZ), dinitroanisole (DNAN), and 2,3-hydroxymethyl-2,3-dinitro1,4-butanediol tetranitrate (SMX). We also report adjusted ratios and melttemperatures for two previously reported ETN eutectics with 2,4,6-trinitrotoluene (TNT) and with pentaerythritol tetranitrate (PETN), and examine a new eutectic composed of insensitive explosives, DNAN-TNAZ. Finally, we examined the eutectic of TNT-PETN as a reference to confirm the validity of our other results. Melting was examined on a differential scanning calorimeter (DSC) and eutectic properties were determined by constructing phase diagrams and Tamman plots. Theoretical eutectic properties were calculated using the van’t Hoff equation and a more recently published equation for predicting eutectics specifically of energetic compounds. Experimental results are consistent across analysis metrics but did not correspond with the theoretical results. With the increased ease of synthesizing ETN, and because we have been examining the usefulness of ETN, its complete characterization, especially its stability, is of great interest. This work examines the thermal decomposition of ETN and compares it to the well-studied nitrate ester PETN, both through experimental methods and computational theory. ETN was aged isothermally at temperatures of 60, 80, 110, 120, and 140 °C for multiple time points. At each point, the amount of ETN remaining was quantified using liquid chromatography mass spectrometry (LCMS). Kinetic parameters of decomposition were found by fitting this data to the integrated first order rate law equation, the obtained rate constants were then fit the Arrhenius equation to calculate the activation energy (Ea) and pre-exponential factor (A). It was found that ETN is less thermally stable than PETN, however ETN is more thermally stable than results from our previous DSC kinetics study would indicate. In addition to these kinetic parameters, decomposition products were examined to elucidate the decomposition pathway of ETN. These products were also studied using LCMS and consisted primarily of erythritol trinitrate, along with minor quantities of the dinitrate and mononitrate, presumably generated by the loss of NO2 from ETN. In addition to our work with ETN, a new approach to develop non-hazardous materials that simulate explosives, and other related threats, on x-ray detection equipment is presented in this work. Simulants for x-ray detection are needed to confirm proper working order of detection instruments in areas where actual threats cannot be examined (airports, vendors’ facilities, etc.). Rather than trying to make universal simulants, which often do not provide adequate matches across multiple instruments, this method focuses on making accurate simulants specific to the instruments on which they are designed. This is accomplished by matching the end signal output of these instruments rather than trying to match the properties inherent to x-ray interactions with materials. Our original methodology was developed on a liquid bottle screener on which we created simulants for five concentrations of hydrogen peroxide, four concentrations of nitric acid, and nitrobenzene. The methodology was applied to a computed tomography baggage scanner to make three hydrogen peroxide simulants and prove the applicability of the simulant development method. The method was tested for its applicability to solid threats and was found to require significant additional tweaks before moving on from liquid threats.


LIST OF TABLES
A binary eutectic mixture will always begin to melt at its eutectic temperature; however, only a specific ratio of components will result in the entire mixture melting at this lower temperature. This specific eutectic ratio will always have an isolated melt at the eutectic temperature. This is called the solidus melt since it is the temperature below which the mixture is entirely solid. 26 At all ratios other than the eutectic ratio, there will be a second melt at a temperature higher than the solidus melt. This second melt is the point above which the solution is entirely liquid and is called the liquidus melt. 27  The first time a mixture was melted it sometimes had multiple liquidus peaks due to insufficient mixing of the solid powder components. These samples were termed "first melts;" and if they had not been run to their decomposition temperature, they could be cooled to room temperature and heated again. The second run gave a much cleaner DSC trace than the first since the compounds were more thoroughly mixed after the first melt. However, some mixtures showed significant decomposition after being melted the first time, resulting in less heat absorbed in the solidus melt or no solidus melt at all. Re-melted sample data where decomposition was apparent and first melt data where the sample was clearly not mixed were discarded. The data judged as "good" was analyzed independently-first run data separate from pre-melted data--and eutectic ratio calculated.

Results and Discussion
Construction of Theoretical Phase Diagrams: Calculated eutectic melting points and eutectic ratios for each mixture are given in Table 1  Using this same approach for the TNT-PETN mixture allowed calculation of a TNT-PETN eutectic ratio which matched the literature value. This success supports the use of this method to provide reasonable estimations of eutectic ratios.
Generally the closer two compounds are in their pure melting points, the closer the ratio will be to 50-50 and the lower the eutectic melting point. If compounds have the same melting temperature and heat of fusion, then the theoretical eutectic ratio will be 50:50. If they have the same melting temperature, but compound A has a higher heat of fusion than compound B, the eutectic ratio will favor compound B. For example, DNAN and TNAZ have close melting points, 96°C and 101°C, respectively.
When mixed in ratios between 48/52 to 52/48 a melt at 66°C was observed; but unmelted material was observed when examined on a hot-stage microscope, and the DSC showed a long endothermic region trailing from the solidus melt returning to baseline in the 93-99°C range ( Fig. 1.9). The TNT-R salt eutectic exhibited a similar trailing region but to a lessor degree.  (15) Urbanski, T.

Introduction
In the U.S., erythritol tetranitrate (ETN) has become a material of concern because the precursors are readily obtained and nitration with mixed acid is relatively straightforward. 1 Because ETN has the same number of nitro groups and better oxygen balance than pentaerythritol tetranitrate (PETN), we have examined it for possible replacement of PETN in formulations. 2 However, its low melting point (60 °C) requires that its thermal stability be carefully examined. To investigate the thermal stability of ETN, isothermal aging was employed to provide an assessment of thermal kinetic parameters. Decomposition products were analyzed using mass spectrometry (MS) to determine if ETN decomposition is similar to PETN.

Experimental Section
Materials ETN was prepared by nitration with acetyl nitrate, a route which results in a high yield of product more completely nitrated than by the mixed acid method. 3 Glacial acetic acid (16.7 mL) and acetic anhydride (16.7 mL) were added to a round bottom flask and suspended in an ice bath. White fuming nitric acid (8.4 mL, 99%) was added dropwise maintaining a reaction temperature below 10 °C. The reaction mixture was allowed to stir for 30 minutes; then meso-erythritol (2 g) was added portionwise keeping the temperature below 10 °C. This mixture was allowed to stir for 2 hours at 0 °C before it was removed from the ice bath and stirred another 2 hours at room temperature. The reaction mixture was poured into a beaker of ice water, filtered, rinsed with copious amount of ice water, and allowed to dry. The crude product (4.54 g) was recrystallized from isopropanol and filtered to give the pure product (3.90 g). Purity was checked by melting point and liquid chromatography/mass spectrometry (LC-MS) detector.
Our experiments also required crude, partially nitrated ETN. Since mixed acid nitration generally produces less complete nitration, this method was used with slight modifications. 4 Sulfuric acid (3.0 mL, 96.5%) was added to a round bottom flask and cooled to 0 °C. White fuming nitric acid (4.6 mL, 99%) was added dropwise maintaining a reaction temperature below 10 °C. The mixture was allowed to stir for 30 minutes at 0 °C, and erythritol (1 g) was added portionwise keeping the temperature below 10 °C. This mixture was allowed to stir one hour at 0 °C rather than 30 °C. The nitration mixture was also not allowed to stand for an additional half hour; instead it was precipitated in a beaker of ice water and neutralized with sodium bicarbonate. The resultant crude product was filtered, rinsed with ice water, and allowed to dry. It was not recrystallized. Figure 2 ammonium acetate, and 0.1% formic acid (A); and acetonitrile (B). The solvent flow rate was 300 L/min with the following gradient program (A%-B%): 0 to 1 min isocratic 70%-30%, 1 to 4 min gradient from 70%-30% to 5%-95%, 4 to 5.5 min isocratic 5%-95%, 5.5 to 6 min gradient from 5%-95% to 70%-30%, 6 to 7 min Plotting the natural logarithm of the fraction of ETN remaining versus heating time in seconds (s) allowed determination of the rate constant, k (s -1 ), at each temperature from the slope ( gave a slope with activation energy, Ea, (kJ/mol) and y-intercept of the natural logarithm of the pre-exponential factor, A (s -1 ).

Product Analysis
To identify potential decomposition products, partially decomposed samples were reexamined by LC-MS. These partially decomposed samples, representing 10-25% decomposition, produced at various temperatures, were analyzed on a Quantiva LC-MS using a precursor ion scan. The mobile phase gradient, column, and ion source parameters were the same as for the quantification method described in the kinetics experimental section. To scan for decomposition products the MS was set for a precursor ion scan to look for precursor ions that fragmented into ions with m/z of 62 (NO3 -) and 46 (NO2). The first quadrupole was set to allow ions with m/z 80-800 through one m/z at a time. Each ion was then fragmented in the second quadrupole using argon collision gas at 1.5 mTorr with a collision energy ramp from 10 to 20 V.
The third quadrupole scanned for anything that created fragments of m/z 62 and 46.
The resulting chromatograms showed what parent ions in the decomposition products generated these fragments.
The partially decomposed samples were also run on a Thermo Exactive The resultant data was analyzed by generating extracted ion chromatograms for the exact mass (m/z +/-5 ppm) of potential decomposition products of ETN.
Decomposition product chromatograms were compared to crude, non-fully nitrated ETN chromatograms to verify some of the products.

Computation
The quantum chemical calculations were carried out using the Gaussian-09 package 5 . The B3LYP hybrid density functional 6

Kinetics
We have previously compared the rate of decomposition of ETN and PETN by differential scanning calorimetry (DSC); ETN appeared to decompose more rapidly.   that NO2 elimination is favored from the internal carbon (Figure 2.6). Computation also reveals that potential bimolecular H-atom migration between two PETN or two ETN molecules is unfavorable (Figure 2.7). The lack of bimolecular ETN decomposition is supported by the observation that the rate of ETN decomposition at 80 °C remained essentially identical whether performed neat or in benzene solution (   Table 2.4; no decomposition products other than the lesser nitrates were discovered. In well-sealed and highly decomposed samples, red gas, assumed to be NO2, was observed. This generation of NO2 from ETN would explain the autocatalytic decomposition observed in ETN kinetics experiments.
We have previous reported the IR and Raman spectra of ETN. 3 Theoretical computations now permit assignment of major features. Figure 2.9 shows three major IR peaks assigned as ONO scissoring, symmetric NO2 stretching, and asymmetric NO2 stretching. In Raman spectra three major peaks are assigned as OCH rocking, NO2 scissoring, and NO2 symmetric stretching (Figure 2.10) ppm corresponded to the two protons on the internal carbons. The acetone proton NMR was similar with four doublets observed at 5.01, 5.05, 5.25, and 5.29 as well as a multiplet at 6.04 (Figure 2.12).

Conclusions
Isothermal aging of ETN samples has provided insight into the thermal stability of ETN as well as details on the resulting decomposition products. Despite the ability to exist in a molten state, ETN has a lower thermal stability than its counterpart PETN. However, it was found that ETN was significantly more stable than estimated from DSC data. 3 Decomposition products identified were exclusively formed by loss of NO2 which is often the case for nitrate esters. DFT computations indicated that the first step in ETN or PETN decomposition could be NO2 or HONO ejection, though the former is more energetically favored. After the first step a number of decomposition routes become available. In experimental tests it was observed that ETN decomposition frequently accelerated rapidly beyond 25-30% to 100% decomposition; this can be explained by the reactive nature of the NO2 produced.
Wallingford CT, 2009.      Background X-ray scanner response (RES) to a compound is based on its density (ρ) and effective atomic number (Zeff) e as well as constants related to the energy of the x-ray.
Thus the development of simulants for x-rays has often focused on crafting materials that match both the density and Zeff of the threat. One problem with this approach is that it is challenging to match these values, specifically the Zeff; which is directly related to the energy of the x-rays being used. Thus, a "Zeff match" is only good for the x-ray energy level for which it was created. The only way to match the Zeff across all energies would be to match the exact elemental composition of the material being simulated.
The first step in creating simulants for x-ray systems is understanding how the x-ray instrument response is related to density and Zeff. Zeff is calculated by taking the sum of the Z's for each component of the system raised to some exponential power weighted by their fraction of protons in the entire system. The value of this exponent depends upon the x-ray energy and the type of x-ray interactions being measured (i.e. transmission or scattering). For transmission it is often assumed to be 2.94 (Equation Where Z is the number of protons in a given atom; f is the number of protons in that atom divided by the total number of protons in the composition (which means the proton in H and O in aqueous solutions must be accounted for); and n is the exponent constant related to x-ray energy.
If the Zeff exponent can be predicted based on the energy of the x-ray system, the simulant development method of matching the density and Zeff of a hazardous solution becomes an option. Originally, the goal of this project was to develop a theoretical model that could be used to create simulants for hazardous solutions based on density, the energies of the x-rays in a system, and Zeff at those energies. Our approach was to run sufficient samples of known composition and density so that by an iterative process the data could be fit to equation 2.
The value of n in equation 1 and the form of equation 2 were varied systematically as part of the iterative process. Based on the best fit, the exponent n in equation 1 could be established for that x-ray instrument. Unfortunately, this approach did not feasible, due to the polychromatic nature of x-ray sources. Nevertheless, we were able to identify a number of effective simulants for threat liquids and elucidate a method for simulant development applicable to certain instruments.
Explosive precursors were used as the threat materials in this study: nitrobenzene (NB), hydrogen peroxide (HP), and nitric acid (NA). Non-energetic materials used as potential simulant components are listed in List of aqueous solutions employed and their estimated density and Zeff are shown in Table 3.2.

Results and Discussion
Rather than trying to match the actual density and elemental composition of a threat material, we attempted to match the output signals that the instrument produced  (Fig. 3.4a), and sometimes it was a bit "off" (Fig. 3.4b) (Table   3.3). By this method simulants were prepared for five concentrations of HP (Fig. 3.5).
The simulant attempts and adjustments are included for 65-90% HP in supplemental information (SF 3.1-3.4), 60% HP simulant matched on first try.
For liquids with densities lower than water, the LE/HE trends lines approached the water LE/HE value from the lower left corner (Fig. 3.1). To reach these LE/HE values, it was necessary to prepare simulant mixtures using at least one low-density organic material (see Table 3.2, CHNO samples). Mathmatica could be used to generate two equations; one that predicted HE and one that predicted LE from the wt% BaCl2 and wt% KBr mixtures (Equations 2 and 3).
With equations 2 and 3, simulants for any liquid threat material bracketed by the original 25 simulants could be prepared. To demonstrate the robust nature of the calculation, equations 2 and 3 were used to create simulants for four concentrations of nitric acid (NA). The three simulants for nitric acid that were within our original simulant boundaries matched on the first attempts. The fourth, which was outside of our boundaries, was close but slightly "off," requiring further adjustments (Fig. 3.6).

Potential Extrapolation of Method for Dual-energy CT:
Having successfully demonstrated a method for preparing X-ray simulants for a rather unique X-ray device, this method was tested for applicability on a dual-energy computedtomographic (CT) X-ray. Using a subset of the innocuous liquids listed (Table 3.1), we found that as the sample concentration decreased the trendlines for instrument output reported as "Zeff" and "density" trended back toward water in a similar manner as observed on the original x-ray detector ( Fig. 3.8).
Using this data we attempted to create simulants for the dual energy CT using the trendline shifting approach described in Figure 3.3. However, the HE and LE of the innocuous liquids were replaced by the appropriate instrument output of density and Zeff. Simulants were created for three concentrations of HP, the results are shown in Figure 3.9.
During two sessions, threats were analyzed on the dual energy CT. The first session threat data is shown in green, and the second is shown in blue. Simulants, shown in orange, were created from the first session data and analyzed during the second. The initial results for the three HP simulants were promising for creating simulants in this fashion. The deviation from the first session threat data and the simulant data was less than 2% for the three HP simulants in both density and Zeff. The HCl data landed almost directly in the path of one of the aqueous solution trendlines ( Figure 3.10), collection of data on a potential simulant is in progress.

Methods for Simulant Development for Single-energy CT:
The one dimensionality of the single-energy CT response meant that our previous method could not be applied.
On a single-energy CT, the histogram of the instrument response to the threat material, i.e. HP (Fig. 3.11) can be matched. The ratio of the solutes to one another affects the histogram peak shape. Once the desired peak shape is achieved, increasing or decreasing the solutes in this ratio will shift the peak right or left, respectively, until their histogram overlays that of the threat material. A simulant made of aqueous glycerol and BaCl2 was adjusted in this manner for 30% hydrogen peroxide ( Fig.   3.11). Such an adjustment would not be possible on a dual energy X-ray; one of the approaches described above would be required.

Investigating Trendline Generation for Solids:
An attempt was made to investigate whether a series of solid solutions could be made that would also trend linearly to a single "zero" point, such as was observed with water. To simulate solid solutions, ground urea was chosen as the solvent and MgSO4 as the solute; thus, the "zero" point would be 100% urea. Urea and MgSO4 were mixed together in a variety of ratios.
These solutions were examined on single-energy CT.

Conclusions
Several methods, which generated successful simulants for threat liquids, have been outlined. These methods do not require knowledge of Zeff nor density. The simulants developed (Table 3.4) proved to be quite accurate for the instrument for which they were designed. Each of these methods require that a database be created for a particular model of X-ray instrument and the response to both hazardous and benign materials be recorded. Application of these methods to solid threat materials may be possible, but bulk packing density inconsistencies need to be overcome.      trendline shifted from water "zero" to 0.23 wt% BaCl2 "zero", c) BaCl2 trendline shifted from water "zero" to 0.23 wt% KBr "zero", intersection of shifted trendlines is at threat to be simulated.