Intermolecular Interactions of Energetic Materials

A variety of intermolecular interactions occurs when an energetic material responds to its surroundings. With a better grasp of these energetic material contacts, improved performance on plastic-bonded explosives, superior swab materials for explosives detection, and novel insensitive munitions are possible. In order to further understand these interactions, the following relationships were researched: adhesion between energetic materials and polymer substrates; quantitative collection and detection of energetic materials on electrostatically charged swabs; and noncovalent derivative investigation between energetic material pairs. A number of explosives detectors rely on introduction of the analyte to the instrument via swabs. However, most swab materials are burdened by either poor sorption (pickup) or poor desorption (release). Therefore, finding a swab that can both easily sorb and desorb an explosive is highly desirable. Atomic force microscopy (AFM), while normally employing a sharp (~5 nm) tip for topographic and force measurements, can also be used to measure adhesion between a material and substrate surface. AFM force curve experiments were performed on eleven polymers with nine energetic materials, organic explosives, and energetic salts. Teflon was the least adhesive polymer to all energetic materials, while no distinct trend could be elucidated among the other polymers or energetics. Rather than create a novel swab material for explosives detection, improving current commercial off the shelf (COTS) swabs would be a fast and cost-effective way to increase analyte detection on existing security instrumentation. For this reason, the viability of electrostatically charging COTS swabs was explored. COTS swabs were charged both triboelectrically and inductively, and voltage degradation both in time and through changes in relative humidity was determined. For collection efficiency, transfer efficiency, and uncharged swab comparison, quantification of energetic materials on a triple quadrupole liquid chromatograph/mass spectrometer was performed. Limits of quantification for trace amounts of energetic material were typically in the single nanogram level. In addition to adsorption of energetic material comparable to traditional uncharged swabs, electrostatically charged swabs can also adsorb material at standoff, introducing a new noncontact sampling method. The synthesis of the next generation of explosives is increasingly difficult because available novel reagents and synthetic techniques are limited. Energetic material solvates have been known for nearly 65 years, but cocrystallization of an energetic material and another solid has only been demonstrated in the last decade. Relying on noncovalent derivatives (NCDs), cocrystals can tailor explosive properties such as density and detonation velocity, potentially yielding a new energetic material without novel molecule synthesis. The most common synthons for pharmaceutical cocrystallization involve carboxylic acid and amide functionalities, but the majority of common energetic materials are devoid of these groups. With the wealth of knowledge from pharmaceutical cocrystals, utilizing these groups could yield more effective screening for energetic cocrystal pairs. Herein, we present the TNT-nicotinamide cocrystal, an energetic cocrystal with an amide synthon.

charged both triboelectrically and inductively, and voltage degradation both in time and through changes in relative humidity was determined. For collection efficiency, transfer efficiency, and uncharged swab comparison, quantification of energetic materials on a triple quadrupole liquid chromatograph/mass spectrometer was performed. Limits of quantification for trace amounts of energetic material were typically in the single nanogram level. In addition to adsorption of energetic material comparable to traditional uncharged swabs, electrostatically charged swabs can also adsorb material at standoff, introducing a new noncontact sampling method.
The synthesis of the next generation of explosives is increasingly difficult because available novel reagents and synthetic techniques are limited. Energetic material solvates have been known for nearly 65 years, but cocrystallization of an energetic material and another solid has only been demonstrated in the last decade.
Relying on noncovalent derivatives (NCDs), cocrystals can tailor explosive properties such as density and detonation velocity, potentially yielding a new energetic material without novel molecule synthesis. The most common synthons for pharmaceutical cocrystallization involve carboxylic acid and amide functionalities, but the majority of common energetic materials are devoid of these groups. With the wealth of knowledge from pharmaceutical cocrystals, utilizing these groups could yield more effective screening for energetic cocrystal pairs. Herein, we present the TNT-nicotinamide cocrystal, an energetic cocrystal with an amide synthon.
vii LIST OF TABLES

Introduction
For safety, long-term storage, and good performance, various tests must be performed when explosives are formulated with other materials. Polymers in contact with explosives are used in a number of ways, e.g. "plastic bonded" explosives, particulate encapsulation of explosives, or as swab materials and filters to collect explosive particulates. Depending on applications, it is necessary to find polymers that adhere or repulse explosives; the ability to achieve a balance between attraction and repulsion can also be desirable. In this work, atomic force microscopy (AFM) is used to assess explosive/polymer interactions.
Typically, AFM generates topographic images of surface features from atomic to µm scale [1]. However, AFM can also generate force curves between a cantilever tip and sample surface [2][3][4][5][6]. These force curves yield adhesive parameters for the two test materials (i.e. tip and surface). By using the AFM cantilever and sample stage, an explosive particle affixed to the cantilever can be pressed onto a sample material, or, in reverse, the sample material can be deposited onto the cantilever tip and pressed into a monolayer of explosive [7]. Previous work on energetic materials and AFM focused on adhesion to terminal group-functionalized self-assembled monolayers [8], and metal coupon finishes [9].

Results and Discussion
Quantitative force measurements were collected for a virgin tipped cantilever and polystyrene microsphere on Teflon, polyethylene (PE), polystyrene (PS), and polyvinyl alcohol (PVA) ( Table 1). Adhesion forces could not be collected for a  Tip  23  14  991  32  12  991  39  21  885  27  12  891   PS  Microsphere  95  189  883  610  169  250  335  164  30  578  33  246 tipless cantilever with only glue because the overall adhesion was too great for the AFM to accurately measure. The adhesion forces for the virgin tipped cantilever were low with standard deviations no more than 60% of the observed values. The repeatability of these adhesion forces was likely due to the extremely wellcharacterized geometry and relatively small contact area of a manufactured AFM cantilever tip. Conversely, the standard deviation for adhesion forces was large for polystyrene microspheres. Though the microsphere has well-characterized geometry (as confirmed via SEM), the lower elastic modulus and increased contact area when compared to the silicon tip likely caused increased overall measured adhesion [10].
The large microsphere adhesion forces suggest that if polymer substrate transferred to the energetic particle during a series of force curves, the ensuing adhesion force on the next polymer would be artificially high. In practice, this was occasionally observed, requiring a new energetic particle to be adhered to a new cantilever and experiments repeated. Interestingly, the obtained polystyrene-polystyrene adhesion force from the microsphere (335 nN) closely correlated with a previously reported force (314 nN) [11]. Ultimately, these results suggest that none of the energetic adhesions resulted from artifacts of cantilever, glue, or polymer-polymer adhesion.
Quantitative force measurements were collected for the nine energetic materials on the eleven polymer substrates. AFM data sets were run over a period of 18 months. Table 2 presents the data collected over the last three intervals in order to exhibit the degree of reproducibility using different energetic material tips and different polymer substrates. Table 2 shows both the number of scans and the standard deviations. As typical for AFM measurements, standard deviations were large (Table 3) [12][13][14]. In cases where the standard deviation was larger than the measured force, the data are shown, with shadowing, but not included in the averages.
Examining the trends across the eleven polymers (bottom average), Teflon, poly(4vinylphenol) (P4VP), and styrene-butadiene rubber (SBR) had the lowest adhesion forces on the energetic materials. The average force exhibited with Teflon and P4VP was almost as small as that observed with the bare cantilever (Table 1). For Teflon, low adhesion force values are not surprising because it is valued for its "nonstick" properties. Its higher relative surface roughness (RRMS 342 nm) may account for values with high standard deviations. In addition, the small values observed with P4VP could be attributed to high surface asperities throughout a rough substrate. We encountered great difficulty in creating a smooth surface for this material, acquired as a powder, and the resulting surface could have been so rough as to only create a miniscule contact area and subsequent low adhesion force. The other eight polymers had average adhesion forces ranging from 108 to 204 nN, which, considering the standard deviations, were essentially identical. The nine energetic materials studied represent the major classes of military explosives, as well as energetic peroxide explosives and energetic oxidizers: nitrate ester (PETN); nitroarene (TNT); nitramine (RDX and HMX); peroxides (HMTD and TATP); and salts (KNO3, KClO3, and KClO4). (As explained above, shadowed data were not included in the averages.) For each energetic, the data sets collected in Batch 1 were averaged separately from those collected in Batch 2 and Batch 3.
Our purpose in averaging the three data sets separately was to see the magnitude of the differences in measured adhesion another researcher might observe using the same chemical but different microcrystals on the tip and same polymer but different surface preparation. It is notable that the adhesion forces recorded for Batch 3 experiments were always higher than those collected in earlier batches. This may be a consequence of using a different tip (but Batch 1 and Batch 2 used different tips); therefore, it is more likely that the polymer surfaces were smoother due to the use of a doctor blade in their preparation. While the data in Table 2   The slurries were poured onto an aluminum foil sheet and flattened to 1 mm thickness using a doctor blade. After solvent evaporation, 1 cm 2 pieces were cut from the film and used for adhesion measurements. Acquired via AFM topography images, surface roughness measurements of all polymers except PVC are shown in Table 4.

Energetic Material Adhesion
TNT, RDX, HMX, and PETN were obtained from military sources and used as received. Potassium nitrate, potassium chlorate, and potassium perchlorate (Fisher Scientific) were ground in mortar and sieved to obtain desired particle size. TATP and HMTD were synthesized and recrystallized according to previously published procedures [15,16]. Explosive microcrystallites were adhered to tipless cantilevers (Mikromasch CSC37-Tipless-Al BS, Nano and More) using a micromanipulator (Micromanipulator M2525) and polarized light microscope (Nikon Eclipse E400 POL). The cantilever platform was attached via double-sided carbon tape to a homemade probe (flattened wire) and inserted into the micromanipulator. Energetic material was added to a clean glass microscope slide. The powder was milled using another slide until particles of desired size (between 30 to 50 µm) were obtained (HMTD and TATP were not milled due to friction sensitivity). Particle size was estimated using an ocular micrometer; more accurate estimates of particle size were obtained by scanning electron microscopy (SEM). A glass Pasteur pipet, pulled to extremely fine points, was used to transfer a micro-drop of UV-curing glue (Loctite 352, Henkel) to the microscope slide. The cantilever was lowered into the glue microdrop, touched to the energetic microcrystallite (~40 µm long), and cured with UV light for 10 minutes. Cantilevers were also made with glue adhered to a cantilever or with a polystyrene microsphere adhered to the cantilever. If the energetic tip appeared damaged at any point, it was replaced and all measurements were repeated. After affixing each explosive to a tipless cantilever, SEM images were collected to confirm microcrystallite adhesion and that no glue or other artifact would contact the sample. As seen in Figure 2, solid contact between that energetic material and the surface was achieved. Because of its high volatility, TATP could not be imaged in the high vacuum of the SEM, while HMTD could not be imaged because the incident electron beam caused initiation of the material.

Force Curve Measurements
Before force curves were taken and after each polymer set, the modified cantilever was calibrated using the Thermal K function available on an Agilent 5500 AFM.
Experiments were performed at <20% relative humidity via flowing dry compressed air or nitrogen through the instrument's environmental chamber, and an electrostatic ionizer (Staticmaster) was employed for electrostatic dissipation. Because the polymer surface was easily deformed, the vertical displacement of the force curve was adjusted after every few force curve measurements to prevent indentation of the polymer.
Force measurements were taken using native tipped cantilevers, tipless cantilevers with only glue, a tipless cantilever with a polystyrene microsphere, and tipless cantilevers with fully adhered energetic microcrystallites. The order of polymers examined against a given tip was altered to show that one data set had no effect on another; repeat measurements of an initial polymer were conducted after collecting measurements from a second polymer for the same reason. After collection of a number of force curves (usually 1000), unrepresentative curves were culled for two primary reasons. First, significant indentation of the polymer after the jump-tocontact point was occasionally unavoidable, causing plastic deformation to the polymer or energetic material particle or transfer of significant amounts of polymer onto the particle. After the deformation or transfer, each successive force curve would be obtained with a unique particle (or polymer-coated particle), hindering comparison to other polymer force curves and other force curves within the same polymer set.
Second, surface roughness of polymer substrates was potentially too high, causing unrepresentative adhesion or detector saturation. Representative force curves were then baseline-normalized and calibrated using the measured cantilever deflection sensitivity and force constant. A histogram was then created to determine the adhesion force with highest frequency. A representative histogram is shown in Figure 3.

Conclusion
Adhesion forces of nine energetic materials were measured on eleven different polymer substrates. It was hoped that this study would allow us to match a particle explosive with a particle polymer that it best adhered to. However, examining each explosive (horizontal rows in Table 2), no "best" match could be identified due to the normal variation in results. It was noted that Teflon was the least adhesive polymer for every tip tested. Generally, P4VP and SBR also exhibited low adhesion. Despite wide variations in the chemical affixed to AFM tip, little bias for one energetic over another was observed. The lack of superior adhesion to one polymer over another is attributed to the effect of bulk properties, such as particle size, roughness, and contact orientation/angle, during force curve collection.

Introduction
Effective energetic material detection is of critical importance for homeland security, but every detection technology begins with collecting samples to analyze. Due to the low vapor pressure of explosives, many commercial detection technologies collect sample by physical contact "swabbing." [1][2][3][4] There are many variables that affect trace sampling, and significant research has been done to understand them more fully. [5][6] The size distribution, shape, and morphology of realistic trace explosive particles has been investigated, and the standardized creation of representative trace particles has been subsequently studied. 7 Specifically, drop-on-demand inkjet printing has become a staple in a number of industry, government, and academic laboratories as a method to reliably create standardized trace amounts of explosives. [8][9][10][11][12][13] This technique can dispense highly accurate and precise volumes of a variety of analyte solutions on a scale much smaller than traditional deposition methods such as syringes. [14][15] The contribution of biological substances (e.g. fingerprint oils) to trace explosives detection has also been investigated. 16 While understanding the properties of trace explosive particles and attempting to recreate them is important, it is only half of the sampling interaction. The role of the swab in the sampling mechanism has also been studied. Adhesion between a trace particle and a swab influences the ability of the swab to pick up an analyte. 17 In addition, swab material has a large effect on the ability of a given swab to adsorb trace particles. [18][19][20][21][22] Combining these variables can lead to a measure of the collection efficiency on a sampling protocol. 22  Noncontact electrostatic sampling from clothing or fingerprints has been demonstrated for biological samples, but its viability for energetic materials is currently unknown. [31][32] In this work, the potential for a noncontact, electrostatically-

Experimental Methods
Analytes Sucrose and sodium chloride were purchased commercially and acted as surrogate analytes for an organic explosive (e.g., TNT, RDX, PETN) and inorganic threat material (e.g., ammonium nitrate or potassium chlorate), respectively. PETN, RDX, TNT, and C-4 were either synthesized in-house or received from military or industrial sources. The materials were sieved to approximately 800 micron but otherwise used as purchased.

COTS Swabs
Teflon and Nomex, and Teflon-coated fiberglass and cotton swabs were supplied by FLIR and DSA Detection, respectively, and used as received.

Substrates
Teflon and Nomex COTS swabs were also used as substrates. Bytac was purchased as a ream and cut to size but was otherwise used as received. Vinyl fabric (90% polyvinyl chloride) was purchased in a fabric store and used as received.

Scanning Electron Microscopy (SEM)
A JEOL 5900 SEM was operated in backscatter mode at 20 kV to collect micrographs of COTS swab materials.

Method of Charging Swabs
For proof of concept, triboelectric charging was accomplished by simply rubbing the swab on a polyamide chair seat ("hand charged"). To charge more reproducibly, a charging setup was constructed using an electric drill to rotate two polyurethane paint rollers ("double roller charged") through which the swabs were fed (Figure 2.2). For swabs requiring use of a wand, a single roller charging method ("single roller charged") was used and can be seen in Figure 2.2. A Simco-Ion Chargemaster VCM-60 was used to inductively charge swabs ("inductively charged") at -10 kV, -20 kV, and -30 kV for five and ten seconds. Voltages were recorded using a 3M 718 static sensor, and a static dissipater was used to eliminate excess voltage before experiments.

Humidity Chamber
A custom-built humidity chamber was constructed from a plastic storage box and fitted with a humidistat and medical nebulizer with water to control humidity.
Experiments were conducted at 0% relative humidity (RH) using dry nitrogen, 25% RH, 50% RH, 75% RH, and 90% RH (water condensed in the chamber above 90% RH) (Figure 2.3). Drop analyses were performed by the instrument to determine drop volume, velocity, and diameter. Pictures were taken with a strobe delay to freeze the drops, and instrument software calculated the values previously listed.
In cases termed "direct deposit" the substrates were used without further processing. In others termed "dry transfer," a second transfer was performed. The explosive solution was placed on a 1" x 3" strip of Bytac, and, after the solvent was evaporated, the Bytac was rubbed/scraped against the final substrate. A pictoral representation of dry transfer can be seen in Figure 2.4. Dry transfer prevented settling of analyte into the grooves of a rough substrate and ensured that the analyte remained in a powdered form rather than an amorphous "coffee ring;" a common result of direct deposition methods. Finally, dry transfer protected the substrate from excessive exposure to organic solvents. 33

Swabbing Procedure
For sampling after direct deposit or dry transfer, a COTS swab was used to collect the analyte from the substrate. For contact swabs, the swab was physically wiped across

Quantification of Explosives
Using a Thermo Electron TSQ Quantiva mass spectrometer, standard curves were created for explosives PETN, RDX, TNT, and TATP. As seen in Table 2.1, with an injection volume of 20 μL, limits of detection for these energetics was as low as 100 pg and dynamic range was large. Extraction optimization was performed with an acetonitrile:water mixture at three different ratios. Using the inkjet printer, 500 ng of PETN solution was deposited on three different substrates; then, the solution was allowed to dry. Each substrate was extracted, and the extract was analyzed via LC/MS for retention time, peak shape, and quantification of both the analyte and internal standard. As seen in Figure 2.5, the 10% organic extraction solution yielded representative retention times and excellent peak shape of both PETN and the internal standard, HMX. The 50% ACN mixture showed similar retention time, though the peak shape was slightly broader, especially for the internal standard. At 90% ACN, the retention time and peak shape of the internal standard were altered enough to make accurate quantification difficult. As seen in Table 2       Uncharged contact swabs were treated with a Zerostat anti-static gun and weighted with a 50 g weight for consistent contact friction. Charged swabs were held over the trace explosive residue at a fixed distance of 10 mm. After exposure to the explosive, the swab was immediately placed into the FIDO, and results were recorded.

Confusants
To Itemiser ion mobility spectrometer was tested using PETN standards. To a swab (Nomex and Teflon-coated fiberglass for the Fido and Itemiser, respectively) was deposited 10 μL of PETN solution in acetone, and the solvent was allowed to dry.
Then, the swab was inserted into the instrument, and the signal intensity was recorded.
Three replicates of each concentration were analyzed, and means and standard deviations were calculated.

Scanning Electron Micrographs
Scanning electron micrographs of COTS swabs were taken using a JEOL 5900 SEM in backscatter mode. As seen in Figure 2.6, the COTS swabs have varying levels of surface roughness and topography, which could affect their adhesion or adsorption of energetic materials. Teflon has the smoothest overall surface, while cotton has the roughest, most irregular surface. The Teflon-coated fiberglass, though patterned, has pockets between its grids that can allow additional material to become attached.

Charging Methods
To determine charge viability, COTS swabs were triboelectrically charged either by rubbing them against polyamide fabric or in a double roller system with polyurethane foam rollers. Five types of swabs (of four materials) were successfully charged, and voltage decay over time was monitored (   Teflon, which is also on the high negative part of the triboelectric series. The soft polyurethane roller imparted more voltage than the hard polyurethane roller, probably because it could better contact the swab than the hard roller ( Having determined that the roller method of charging had a number of drawbacks, COTS swabs were inductively charged using a Simco-Ion Chargemaster VCM-60. This method was tested at three voltages (-10 kV, -20 KV, and -30 kV), and the imparted voltages were measured ( Table 2.5). Significantly higher and more consistent voltages were obtained by this method than by the roller method, and potential contamination of the swab by the roller material was avoided. In addition, the voltage on an inductively charged swab lasted for a longer duration.

Humidity
Because contact electrification is affected by humidity, 34 COTS swabs were charged in the double roller system in five different humid environments: 0% RH, 25% RH, 50% RH, 75% RH, and 90% RH. As the humidity increased, the voltage imparted to a given swab decreased (Table 2.3).

Release of Voltage
To analyze the ability of COTS swabs to release voltage (and, therefore, their analyte payload), swabs were triboelectrically charged and subjected to a thermal desorption analysis cycle in a Morpho Itemiser DX. As seen in Table 2.6, the voltage on a swab was significantly mitigated (if not eliminated completely) after an analysis cycle on a commercial explosives trace detector (ETD), which suggested the analyte would be released.       where almost all the TNT remained on the first Bytac surface (blue column). As might be expected, the amount of material picked up by contact swabbing (green column) was proportional to the force applied during swabbing ( could be extracted from these substrates when they were prepared by dry transfer and not subjected to an organic solvent. These results suggest that, for substrates that are sensitive to organic solvents, either performing dry transfer or depositing much smaller amounts of organic solvent (through a process such as inkjet printing) would yield more accurate and reproducible amounts of material for analysis. Moreover, dry transfer ensured that the analyte remained in a powdered form rather than an amorphous "coffee ring." Table 2.11 compares the uncharged contact swabbing data from Table 2.9 to electrostatically enhanced swabbing using either a Teflon or Nomex swab. In two of the four combinations, the electrostatic method was slightly superior or directly comparable. The signal response for the Itemiser (Figure 2.11) and Fido X3 (Figure 2.12), when fitted to a curve, shows logarithmic correlation with increasing concentrations of PETN. In the IMS experiments, toward the higher end of the detection limit, the instrument required a more thorough clearing process between samples. This suggests that the instrument was reaching a saturation point. All the results from the IMS were obtained in the same day. When attempting to continue analysis on a different day, the intensity varied and was therefore not included in the set. In addition, as each experiment ran, the retention times shifted from 9.3 seconds to 9.2 seconds, suggesting that as the detector is exposed to real signals, the values shift.
In the FLIR experiments, one nanogram and five nanograms showed a signal response, but they were both under the limit of an alarm. Though there is no precedent for FLIR signal response correlating logarithmically, IMS systems have shown a logarithmic correlation of signal intensity to increasing amounts of TNT. 23   The FLIR, while correctly alarming on every electrostatic swab event, also alarmed in ungloved fingerprint contact and electrostatic analyses performed concurrently with these tests. The signal for electrostatic swabs rose in time after introduction, reached the alarm threshold, and steadily decreased as the analyte was desorbed through the instrument. The signal for contact swabs, however, rose in time and alarmed as before, but the signal rapidly decreased significantly below the intensity origin. Specifically, after alarming correctly for C-4, the signal for one sebum sample dropped so rapidly and severely that three channels were reduced to values never before seen on the instrument. The instrument continued to operate correctly after this sample, but these results could shed light on the suspicious weakness of the FLIR to confusants. If the confusant is thermally desorbed from the swab before the analyte of interest, the signal could be decreased so much that the subsequent increase in signal from the analyte would be unable to overcome it. Conversely, if the analyte of interest is desorbed first, the instrument appears to operate correctly even with the resulting rapid and severe drop in signal. Electrostatically charged swabs adsorbed as much or more energetic material as conventional uncharged contact swabs.
Electrostatically charged swabs could become a novel tool for trace explosives collection, especially on areas that are difficult to sample, e.g headdresses or the corners of bags. In the field, an inductive charging unit could enhance swabbing efficiency with little to no loss in sampling time, a relatively small working footprint, and a reasonable cost through its ability to voltage swabs simultaneously for multiple ETDs. Moreover, we expect non-contact swabbing to result in lower false alarm rates due to the lack of contamination from oils often picked up in contact swabbing.

Introduction
Cocrystallization can rectify undesirable physical and chemical properties, such as solubility, high moisture affinity, melting point and stability. The most widely used applications for cocrystallization are for pharmaceuticals, where it has been utilized for over a decade. More recently, the energetic materials community has attempted to tailor explosive properties such as density, detonation velocity, and sensitivity by the same method.
The definition of "cocrystal" has been nebulous at least as far back as the 1970's. 1-2 Even the mandatory presence of a hyphen in the word, i.e. "co-crystal," has been the subject of disagreement. [3][4][5][6] The term "cocrystal" has become widely accepted as "co-crystal" has diminished, but an agreed-upon definition is still lacking. objections to this stricter definition, a cocrystal in this manuscript will nevertheless be defined as a unique crystal lattice composed of two or more different neutral, solid molecules in some definite stoichiometric ratio. 9 The principal energetic material used will be referred to as such, and the second molecule with which it is to cocrystallize will be called the "coformer." Regardless of the definition, cocrystals are incredibly important materials in a number of fields, most notably pharmaceuticals. The reason for this is their ability to tailor the properties of the constituent materials. Pharmaceutical cocrystals are often designed to improve bio-solubility, though a host of other advantageous alterations exists (Figure 3.1). [10][11][12] In the energetics community, cocrystals can change density, detonation velocity, thermal stability, and sensitivity (impact, friction, electrostatic discharge). [13][14][15][16][17][18][19][20][21][22][23][24][25][26] With the synthesis of the next generation of explosives becoming increasingly difficult because available novel reagents and synthetic techniques are limited, cocrystallizing energetic materials unlocks a new avenue for optimizing explosive properties.
Insofar as the free energy of cocrystal formation is favorable, nearly any method of single crystallization can be employed to create cocrystals. The easiest and most widely used process is through solvent evaporation, wherein the analytes are dissolved in a solvent of choice, and the solvent is evaporated either at ambient or slightly elevated temperature. 27 Reproducibility is a concern, and it is limited to solvents with adequate vapor pressure and in which the analytes are reasonably soluble. Modifications of this method include using more than one solvent, either as a mixture or separated in a sealed container to allow vapor diffusion of the second solvent. 28  Thermal methods are popular for creating cocrystals. In a solution-based method similar to recrystallization, analytes are dissolved in a hot solvent and allowed to precipitate as the temperature of the solution is gradually lowered. 27 Without solvent, compounds can be melted together on a hot stage and allowed to cocrystallize as the temperature is lowered. This method, popularized by Kofler, can be used to screen for potential cocrystals. [29][30][31] Less common, but still effective, cocrystallization methods include physical grinding with and without solvent, 31-34 spray drying, [35][36] resonant acoustic mixing, [37][38] and supercritical fluid precipitation. [39][40][41][42] Before cocrystallization experiments can be run, pairs of molecules to be cocrystallized must be chosen. Previously published cocrystals can be helpful in selection of compatible materials with potentially desirable properties (   10 DADP 132* 253* Trichlorotrinitrobenzene (TCTNB) n/a n/a n/a n/a n/a 15 11 DADP 132* 253* Tribromotrinitrobenzene (TBTNB) n/a n/a n/a n/a n/  In this reported work, many types of synthons (and resulting cocrystals) were attempted. In general, the energetic materials attempted were analogs of compounds

Cocrystal Preparation Solvent Evaporation
Two analytes were dissolved in a solvent, syringe filtered into a glass vial, and the solvent was allowed to evaporate at a given temperature. Normally, solvents

Vapor Diffusion
Analytes were dissolved, syringe filtered at 0.2 µm, and put into a 1 mL glass vial.
The smaller vial was placed in a 15 mL screw-cap vial that was subsequently loaded with an anti-solvent and the large container was sealed. (An example of this type of setup can be found in Reference 28.) After two weeks, if no precipitate was observed, the 15 mL vial was opened to the environment, and evaporation of both solvents was allowed. Solvents were acetone and ethanol, and anti-solvents were chloroform and cyclohexane, respectively. For large vapor diffusion experiments, 4 mL and 40 mL vials were used as the small and large containers, respectively.

Grinding
Analytes were placed into a small mortar in a given ratio with total mass 5 mg or 50 mg. Then, the compounds were ground by hand both with and without solvent for 30 s. If solvated, the solvent was allowed to dry, and the solid was scraped into a vial.
Note: Before grinding an energetic material, it is critical to confirm the insensitivity of the material to friction, impact, and electrostatic insult!

Thermal methods
Analytes were dissolved into a solvent, filtered, and placed into a glass vial. The vial was placed in a 3 °C refrigerator for one week. If no precipitate was observed, the vial was placed in a -20 °C freezer for one week.
TNT and nicotinamide were placed in a ceramic crucible and situated on a hotplate at 150 °C. When both materials melted, the liquids were mixed together with a metal spatula. The mixture was allowed to crystallize both quickly and slowly in separate experiments.

Thermal microscopy
Using the Kofler method in a hot stage microscope, 29-31 the analyte with the higher melting point was melted and allowed to recrystallize. Then, the lower melting compound was melted into it and allowed to recrystallize. If a new solid layer developed between the melted layers, a cocrystal was possibly formed.

Supercritical Fluid Precipitation
A Waters supercritical anti-solvent (SAS) reactor was used to precipitate a solution of TNT and nicotinamide that was of 10 mg/mL concentration for both reagents. The

Resonant Acoustic Mixing (RAM)
A Resodyn LabRam was used to acoustically mix analytes at high acceleration.
Analytes (different ratios equaling 50 mg total) were added to 1 mL glass vials, 50-100 µL of solvent was added (analyte was not dissolved), and the vials were acoustically mixed. Typical accelerations and durations were 30 gr for 1 hr, 50 gr for 1 hr, and 80 gr for 45 min.

Polarized light microscopy (PLM)
A Nikon Eclipse E400 POL polarized light microscope was used to image crystals both in unpolarized and polarized light. Crystals were typically imaged at 100x magnification. For hot stage experiments, a Mettler Toledo FP900 Thermosystem with a FP82HT hot stage was used. Crystals were heated at 10 °C/min and observed for changes as they were heating.

Differential scanning calorimetry (DSC)
A TA Instruments Q100 DSC was operated with nitrogen purge gas flowing at 50 mL/min. Samples were weighed to ~1 mg in aluminum pans and hermetically sealed.
The DSC ramp program was from 30 °C to 400 °C at 20 °C/min. For thermally cycled materials, the mixture was heated to slightly above the melting point of the higher melting material (e.g. 80 C for 2,6-DNT), cooled, and heated again to 400 C to ensure all thermal events were recorded.

Raman spectroscopy
An Andor Shamrock spectrograph coupled with Ondax SureBlock ultra-narrow-band filters and iDus CCD detector was used to collect low-frequency Raman spectra with a 785 nm laser source. Integration time was 60 s, and resolution was <1 cm -1 .

X-ray diffraction
A Rigaku Optima IV X-ray diffractometer was used to analyze crystallography of materials. A Cu source generated X-rays at 40 kV and 44 mA, a 10 mm slit was used for the source, the sampling rate was 0.75 °/min, the sampling width was 0.25°, and the sampling range was 5° to 105°.

Cocrystals Attempted
The vast majority of combinations tested were screened initially via solvent evaporation. Solvents were chosen first upon ability to dissolve the analytes; this meant that acetone was often used. The second consideration was ease of evaporation; therefore, vapor pressure and boiling point were examined. The synthons typically available in energetic materials fall into one of six generalized categories: pi-pi, nitro, peroxide, nitro/wurtzitane cage, amine/nitro, and amide interactions.

Pi-Pi interactions (Table 3.3):
Energetic materials 2,4-DNT, 2,6-DNT, DNAN, and TNT contain aromatic rings, leading to an increased likelihood of pi-pi stacking, both in-plane and out-of-plane. TNT has been shown to exhibit face-face stacking in a cocrystal, 16 and we anticipated that dinitrotoluene and dinitroanisole would do the same. However, with a wide range of solvents and both aromatic and nonaromatic coformers, no cocrystals were observed, either visually or by PLM.   14,17,36 Our attempt to cocrystallize RDX with the same materials that cocrystallized with HMX did not result in cocrystals. It may be that the six-membered ring of RDX can compact in a way that the eight-membered ring of HMX cannot, possibly leading to inaccessibility of part of RDX necessary to interact with another molecule. Although one example of a cocrystal has been reported for NTO 20 and one for TNAZ, 23 we were unable to discover further cocrystals with these species.    49 While the same interactions of DADP would presumably be present in TATP, we were unable to make similar cocrystals. We rationalize that the torsion in TATP caused by the unfavorable ninemembered ring structure could prevent adequate coupling. Conversely, the planar nitrogens on HMTD should present highly accessible synthon locations. We were unable to form any cocrystals with HMTD, but the poor solubility of HMTD in most organic solvents prevented us from thoroughly investigating it. Similarly, the volatility of TATP ruled out use of solvent evaporation and spray drying techniques as modes of making cocrystals.

Nitro/isowurtzitane cage interactions (Table 3.6): CL-20 readily
cocrystallizes with numerous compounds even though its six nitro functionalities prevent internal access to the cage structure. We speculated that TEX, which presents the same basic cage structure as CL-20 without four nitro groups, should form cocrystals with some of the same coformers used with CL-20. TEX has slight solubility (< 3 mg/mL) in acetone, methanol, ethanol, isopropanol, acetonitrile, and diacetone alcohol (DAA), but it readily crystallizes from these solvents. However, we have been unable to observe cocrystallization of TEX though a variety energetics, including those which formed cocrystals with CL-20, were used (Table 3.6).
Furthermore, we were unable to promote cocrystallization of CL-20 with TEX, even though published density functional theory suggested that possibility. 50 We attribute this to the fact that the only solvent of common and sufficient solubility was nitromethane. Neither CL-20 nor TEX has good solubility (> 5 mg/mL) in common organic solvents; nevertheless, CL-20 readily forms solvates with a variety of solvents (e.g. dimethylformamide and N-methylpyrrolidone), while TEX formed none. 51     well. Additionally, NU, a relatively low molecular weight, linear compound, has little steric hindrance and high accessibility. For those reasons, NU was added with numerous energetic and non-energetic coformers in an attempt to test this theory. In all but one case, NU grew concomitantly with its coformers rather than as a cocrystal.

Vapor diffusion
Four cocrystals were attempted with vapor diffusion (Table 3.9). With acetone/chloroform and ethanol/cyclohexane, few to no cocrystals were observed either visually or with the PLM. Interestingly, the TNT:nicotinamide cocrystal was first observed with the ethanol/cyclohexane combination, but it was never reproduced in multiple attempts.

Resonant Acoustic Mixing (RAM)
Seven cocrystals were attempted via RAM (Table 3.9). TEX was the most commonlyused material for these experiments because of its limited solubility in many organic solvents, though TATB was studied by this method for the same reason. Although a variety of solvents, mixing accelerations, and mixing times were used, no cocrystals were observed visually or via PLM and hot stage microscopy.

Supercritical fluid precipitation
Three cocrystals were attempted with supercritical fluid precipitation, wherein a concentrated solution of the energetic material and conformer were sprayed into a vessel saturated with supercritical carbon dioxide. No cocrystals were obtained (Table 3.9).

Spray drying
Five cocrystals were attempted via spray drying, though no cocrystals were observed either visually or via PLM (Table 3.9). The relative volatility of TNT may have caused difficulties in the ability of the spray dryer to properly condense it after being sprayed.

Grinding
Four cocrystals were attempted by manual grinding, both with and without solvent (Table 3.9). As with the RAM, TEX and TATB were common choices for this analysis because of their relative insolubility in common organic solvents and insensitivity to friction insult.

Kofler melting method
For materials with known melting points, the Kofler method was attempted to screen for cocrystals (Table 3.9). The analyte with the higher melting point was melted and allowed to recrystallize. Then, the lower melting compound was melted into it and allowed to recrystallize. If a new solid layer developed between the melted layers, a cocrystal was possibly formed. In all cases, no distinct third layer was noticed at the interface of the recrystallized lower melting compound and higher melting compound.
In addition, some compounds with high volatility or low stability would begin to decompose, leaving little solid material to analyze.

TNT and Nicotinamide Cocrystal
Of all attempted cocrystallizations, the only quantifiable success was TNT:nicotinamide. A widely used pharmaceutical coformer with both aromatic and amide functionalities, nicotinamide was chosen as a coformer with TNT. Numerous methods for cocrystallization were carried out, including solvent evaporation, vapor diffusion, 84

Grinding
In a mortar and pestle, TNT and nicotinamide were ground with and without solvent at a number of different ratios, times, and different solvents. No cocrystals were observed either visually, by microscopy, or by DSC.

Thermal modification
On the hot stage microscope, nicotinamide was melted and recrystallized on a glass microscope slide. TNT was then added to the nicotinamide and melted into it (Figure 3.12a on right). However, after TNT recrystallized, no distinct layer between the melted layers was observed (Figure 3.12b on right). Upon reheating, the TNT and interface both melted at 80 C, leaving only pure nicotinamide (Figure 3.12c on left).
At 115 C, the remaining nicotinamide began to melt (Figure 3.12d on left).
In a separate experiment, TNT and nicotinamide were mixed in a 1:1 mole ratio and melted on the hot stage microscope. Upon cooling, the mixture recrystallized from one nucleation center and grew outward in a circular pattern (Figure 3

Vapor diffusion
A solvent/antisolvent combination of acetone/chloroform never grew cocrystals.
Ethanol/cyclohexane was the mixture in which cocrystal plates were first observed, though in only one vial in a series of replicates. When the experiment was repeated at two different scales, no cocrystals were seen. This method, while affording the first look at the cocrystal, was neither faster nor more reliable at creation of cocrystals than nitromethane solvent evaporation.

Solvent evaporation
TNT and nicotinamide are both readily soluble (i.e. >10 mg/mL) in a variety of organic solvents, so dissolving each to a desired concentration was rarely a concern.
Because the melting point of TNT is 80 C, solvent boiling point occasionally  Occasionally, without impetus for nucleation, TNT will not crystallize upon thermal cycling, leading to a thermogram with no endothermic melt (Figure 3.16). The TNT has only undergone a phase change and is still viable and able to cocrystallize.     One way to determine whether a material has a eutectic or cocrystal phase is through the construction of a phase diagram: a plot of temperatures of thermal transitions as a function of the amount of one material present. If this phase diagram shows a clear minimum, a eutectic is likely formed. If the phase diagram shows two minima with a raised area between, a cocrystal phase is likely possible. 46 The phase diagram of TNT:nicotinamide mixtures shows two minima at ~10 mol% and ~50 mol% TNT, with a maximum at ~3:1 nicotinamide:TNT mole ratio (Figure 3.21). This ratio agrees with other observations, such as an excess of native TNT seen on most cocrystals and many thermograms showing a native TNT melt even after thermal cycling.

Raman spectroscopy
The Raman spectrum of the cocrystal is similar to an overlay of the spectra of TNT and nicotinamide (Figure 3

X-ray powder diffraction
The diffraction pattern of the cocrystal shows different peaks than those shown in the TNT and nicotinamide patterns (Figure 3.24). Specifically, a cocrystal peak at 19 is unique to that pattern, confirming a unique crystal structure from that of either TNT or nicotinamide.      Cocrystals typically exhibit a melting point between those of its constituents.
That this material does have a melting point above that of 2,6-DNT is encouraging, but the additional endotherm in the 39-42 °C range is curiously atypical.
The phase diagram of NU/2,6-DNT shows no discernable minima or maxima; in fact, the line is flat when both the first and second endotherm onsets are plotted (Figure 3.30). These data suggest no defined ratio at which the NU/2,6-DNT mixture demonstrates the observed thermal peculiarities.   NU was combined with a number of compounds analogous to 2,6-DNT in order to see if the same relationship existed (Table 3.