Improved explosive collection and detection with rationally assembled surface sampling materials

Abstract

Sampling and detection of trace explosives are critical steps in the analytical process necessary for modern transportation safety. In this work we have explored some of the fundamental aspects that influence collection and detection of trace levels of explosive residues from surfaces. We compared the analyte-release performance of standard muslin sampling swipes to that of rationally assembled fiberglass cloth, and used thermal-desorption ion mobility spectroscopy for detection. This collection–detection system is widely used for analyzing the trace chemical residues. The fiberglass cloth was chemically modified by covalently bonding phenyl-functional groups to the surface. The rationally assembled sampling materials provide significantly performance improvements over standard muslin sampling materials for detection of TNT, NG, RDX, TATP, and PETN. The phenyl-functionalized fiberglass swipes showed over 10 times greater TNT release, compared to muslin sampling swipes, as well as improved response and repeatability after multiple uses of the same swipe. The improved TNT release from the functionalized-fiberglass swipes resulted in significantly improved detection limits over muslin. To better understand the improvement offered by the phenyl-functionalized fiberglass, several commercially available fiberglass materials, each offering specific characteristics, were also compared, allowing several physical and chemical properties to be systematically explored to determine their influence on performance. These results are relevant to improving the detection of other explosive compounds, and potentially to a wider range of chemical sampling from surfaces.

---

1. Introduction

The collection and subsequent detection of trace amounts of explosives residue from surfaces is a common method for the discovery of hidden explosives.^1–6 Collection of explosive residue is typically conducted by manually sampling surfaces of interest with a fabric sampling material, generally referred to as a swipe or a trap. The performance of the sampling materials influences the initial analytical process upon which the entire collection, detection, and decision sequence depends for many transportation security needs. Herein we explore some of the fundamental factors involved with analyte release from swipe materials, comparing muslin, one of the common materials in use today, to highly-uniform sampling materials. We show how it is possible to significantly advance the state of art in explosives detection with the application of foundational material science and surface chemistry principles.

The presence of trace explosives on swipes can be analyzed by various instruments, with ion mobility spectrometry (IMS) being the principle method presently employed in the field. IMS enables rapid analysis, has low detection limits for many analytes of interest, has a low operating cost, and requires no sample preparation.^6–12 Consequently, IMS is one of the most widely-utilized, trace detection techniques for explosives detection throughout the world.^2,6,9–15 However, IMS sometimes produces inaccurate results, due to its low resolution, susceptibility to interferents and nonlinear analyte-release behaviors.^{2,4,7,16–19} Improving the front-end sample collection and analyte introduction into the IMS (and similar systems) should improve detection limits, signal stability, and selectivity, potentially resolving many of the principle problems presently plaguing field-deployed instruments tasked with trace organic detection.^17,20

Explosives detection depends upon the effectiveness of the measurement as well as the efficacy of collection.^2,4,10 For collection of trace residues from surfaces, the most widely used swipe materials are made from cotton fabric.^{1,3,17,18,21,22} Muslin is a woven cotton cloth that is the predominant surface-sampling material used for trace explosives collection and sampling, although polytetrafluoroethylene (PTFE), Nomex®, and paper are becoming just as common, and are growing in popularity (see below). Recovery of the analyte from the sampling material can be accomplished with solvent rinsing or thermal desorption.^1,8,22,23 In IMS systems, the sampling material is heated to thermally release the analyte for introduction into the instrument.

While effective and widely used, muslin cloth, which is made of cellulosic fibers, are not ideal sampling materials. Cellulose is a polysaccharide contributing hydroxyl and ether functional groups to a tertiary structure that results in a range of chemical environments. These contribute to heterogeneous interactions with an adsorbed analyte, which can result in a distributed or incomplete analyte release. Further exacerbating the analyte release problem, unprocessed muslin contains non-cellulosic compounds found in native cellulosic fibers (i.e., waxes, natural oils and starches), as well as sizing agents and lubricants added during and after textile processing.²⁴ Several industrial processes, such as harsh mechanical scouring, chemical cleaning and enzymatic methods, are used to remove these natural and added impurities. These processes weaken the cellulosic fibers, which degrade with repeated use, sometimes resulting in the introduction of contaminating fibers into detection instrumentation.²⁴ Additionally, muslin's limited thermal stability (decomposition at ∼150 °C,²⁵ although decomposition will also be related to time at temperature) adversely impacts analyte detection. Muslin's high specific heat and low thermal conductivity,²⁵ combined with surface heterogeneity, result in less than optimal release of analytes from the swipe surface.²⁴ Further, the thermal decomposition and degradation of the swipe material can interfere with the sample analysis by releasing decomposition products, thereby negatively impacting the detection process.^8,24

Recently, a range of materials has been evaluated and used for surface sampling including: PTFE (Teflon®), Kevlar®, Nomex®, and glass fibers, with Kevlar® and Nomex® being composed of aromatic polyamides.^{2,7,8,26–28} Compared to muslin, these materials have high thermal stabilities, contributing to the high desorption efficiencies observed for several explosive compounds. These materials are also resistant to mechanical and shredding stresses.^8,27,28 PTFE has been used as a sampling material, however, PTFE swipes showed lower collection efficiencies for explosives than did muslin.² In one instance, a silicon adhesive was applied to the surface of a PTFE-coated fiberglass swab to improve explosive particle collection.²⁹ These results suggest that for improving explosives collection and detection, a new swipe should have both excellent adsorption and desorption abilities for an analyte, as well as being physically and thermally stable.^8,24

Herein we explore and compare the performance of muslin sampling materials to fiberglass cloth (FGC) sampling materials, with and without phenylsilanes covalently attached to the fiberglass surface. FGC can be constructed with specific physical and chemical properties, and tailored for particular applications. The surface chemistry of FGC allows facile, and presumably, uniform surface modification. The impact of FGC structure, physical properties, and surface chemistries are explored as a means for improving the detection of trace explosives. TNT was chosen as a representative analyte for explosives in general, nitro-aromatics in particular, as well as a specific interest in its own properties as an explosive. The performance of the functionalized FGC swipe materials was also tested for several other explosive compounds.

2. Materials and experimental methods

2.1 Materials

Fiber E glass and S glass materials were obtained from BGF Industries, Inc. (Greensboro, NC) and JPS Composites Materials (Anderson, SC). Astroquartz® fiberglass (pure silica fiber) was provided by JPS Composites Materials. PTFE-coated FGC was purchased from McMaster Carr (Elmhurst, IL). For consistency, weave style 8HS (harness satin) was used, except when 4HS weave was used for E glass CS724, thin E glass 497A and PTFE, since weave 8HS was not available. Commercial muslin swipe materials (0.26–0.32 mm standard thickness) were purchased directly from Smiths Detection (Morristown, NJ).

Trimethoxyphenylsilane (97%) was purchased from Sigma-Aldrich Co. (St. Louis, MO) and used as received. Explosive chemical reference standards for 2,4,6-trinitrotoluene (TNT), nitroglycerin (NG), pentaerythritol tetranitrate (PETN), and triacetone triperoxide (TATP) were purchased from AccuStandard (New Haven, CT), and cyclotrimethylenetrinitramine (RDX) was purchased from Restek (Bellefonte, PA). Analytic grade toluene, methanol, HCl, and NaOH were purchased from Fisher Scientific (Pittsburgh, PA), and used as received.

2.2 Surface pretreatment and functionalization

As-received FGC was pretreated by sequentially washing the FGC in acetone and aqueous solutions of 1.0 M NaOH, and 0.1 M HCl, to prepare the glass surface for chemical modification (see detail in ESI†). A phenylsilane monolayer was applied by refluxing sections of the pretreated FGC overnight (18 h) in a 10% (v/v) solution of trimethoxyphenylsilane in toluene. The FGC was then rinsed twice with 100 mL of toluene, followed by two rinses using 100 mL of methanol. Secondary treatments using the same process, called “back-fill” (BF), were performed to enhance monolayer density and help passivate the surface. Finally, the monolayer-functionalized FGC was dried and conditioned at 180 °C under vacuum for 18 h to remove excess solvent and to continue driving incomplete silane condensation reactions.

2.3 Instrumentation and analysis

A Barringer Ionscan® 400A ion mobility spectrometer was used in this study. The instrument was operated in negative ion mode at a desorber temperature of 180 °C (unless otherwise noted) and an analysis time of 10 s. The drift and inlet temperatures were set at 114 °C and 240 °C, respectively. The sample gas was set to 239 mL min⁻¹ and the drift gas to 351 mL min⁻¹.

Sample analysis consisted of first obtaining a background (baseline) spectrum from a fresh swipe of the same material being tested. One μL of a 10 ng μL⁻¹ TNT-in-methanol solution was then deposited on the swipe material being tested. The methanol was allowed to evaporate for approximately 15–20 s at room temperature, leaving the explosive residue. The deposited sample swipe was analyzed one minute after the baseline run for consistency. Individual studies were collected at the same time interval to improve sampling consistency. Throughout most of this work, the data is presented as values normalized to the maximum muslin IMS response, unless otherwise noted. Normalized IMS responses were presented to provide an easily comparable response among sampling materials, and an instrument-independent value. All values were normalized to data obtained from 0.32 mm thick muslin unless otherwise noted. For other explosive compounds, 10 ng of each explosive was deposited on the swipe, and all normalized IMS responses were normalized to the maximum response obtained from the same amount of each explosive deposited on muslin.

2.4 Sampling cloth characterization

Swipe material thickness measurements were obtained using a Mitutoyo Absolute Digimatic Caliper Series 500 (Aurora, IL) by taking ten measurement points from five different pieces of each respective swipe (minimum 50 points). Surface area measurements were recorded using a Quantachrome Quadrasorb SI Four Station Automated Surface Area & Pore Size Analyzer (Boynton Beach, FL).

3. Results and discussion

FGC is a good candidate for an analytical collection material. The physical and chemical stability of FGC should support repeated mechanical stresses and provide high-temperature stability. The fiberglass weave can be engineered to accommodate a desired thickness, surface roughness, and air permeability—all of which are known to be relevant to collection and detection from swipes.^2,30 FGC materials are also commercially available in a range of surface chemistries, or “sizings,” as described by industry.³¹ The interaction of the analyte with the sampling material, during both collection and release, occurs on the surface; consequently, the surface chemistry of the sampling material is a critical analytical parameter. Furthermore, the silica surface is amenable to installation of a range of functional groups using liquid and gas phase methods.³⁰ These features, coupled with the fact that FGC is relatively inexpensive, make FGC a suitable material for development of improved sampling media.

Since this effort is focused upon improving the collection and detection of trace explosives from surfaces, a single relevant analytical method (IMS) and analyte (TNT) are utilized throughout this work to facilitate a comparative evaluation of material properties. TNT was selected since it is well-studied, and without the challenges of low vapor pressures seen in compounds such as RDX and PETN, although exploratory data for these and several other explosives are also presented.

3.1 Impact of fiberglass material substrate

As previously discussed, the materials used for surface sampling play a critical role in the performance of trace residue detection.^8,20 The analyte release of various, unmodified commercial materials that possess similar physical properties (i.e., thickness) was explored and the results are presented in Table 1. General descriptions of each material's surface chemistry are given as well as the industrial trade name. The exact surface chemistries of the industrial FGC materials are proprietary, and while not fully disclosed, the available information is sufficient to enable understanding of the general surface chemistries and processes.

Table 1 IMS response for TNT release from unmodified sampling materials^a

Material^b	Trade name designator	TNT normalized signal^c (untreated)	TNT normalized signal^d (thermally preconditioned)
a IMS signal normalized to the IMS signal for a 10 ng TNT deposition on muslin. Materials are all 0.20–0.29 mm thick. Additional information is provided in Table SI 2.b FGC is fiberglass cloth; E and S indicate glass types. 537A and Greige are finishes applied to E glass material.c Materials were used as received without thermal treatment.d Materials were preconditioned at 180 °C overnight under vacuum. Muslin was preconditioned at 100 °C; color change indicating that a chemical change was observed in muslin preconditioned above 120 °C.
Muslin	—	1.0 ± 0.1	0.4 ± 0.3
Organic coated FGC	Greige	0.9 ± 0.1	1.1 ± 0.2
Unfinished FGC	537A	1.9 ± 0.1	2.2 ± 0.2
Pure silica FGC silane finished (PSiSF)	Astroquartz®	1.7 ± 0.2	3.0 ± 0.2
Polymer primer (PP) E FGC	497A	2.0 ± 0.1	3.1 ± 0.3
Polymer primer (PP) S FGC	497A	2.7 ± 0.1	3.9 ± 0.2
Cl silane E FGC	CS724	2.9 ± 0.1	2.5 ± 0.2
Cl silane S FGC	CS724	2.8 ± 0.1	3.9 ± 0.1
PTFE FGC	Teflon	3.1 ± 0.2	3.9 ± 0.2
Activated carbon cloth	Zorflex®	0	0

---

Muslin sampling materials are a standard and utilized widely. The materials have poor thermal properties²⁵ and heterogeneous surface chemistry, resulting in a slower and more distributed release of analytes from the surface than desired.²⁴ The results in Table 1 show that almost all of the unmodified commercial FGC materials have a significantly higher normalized response than that of muslin. The improved response will subsequently be shown to be due in part to FGC having a much higher thermal conductivity and lower specific heat (values available in Table SI 1†) than muslin, promoting faster heating and better analyte release (if enabled by appropriate surface chemistry).^25,32,33

For the cases shown, thermal preconditioning of FGC at 180 °C under vacuum generally showed more intense TNT signals, compared to as-received FGC. The increased TNT-signal intensity is believed to result from improved release of TNT from the material surfaces. Further, lower background signals from contaminants associated with the as-received FGC, that appear to thermally desorb at temperatures less than 180 °C, also improve the TNT signal. Thermal and vacuum preconditioning contributes to improved analyte release in two ways. First, it removes volatile surface impurities associated with the FCG that remain from the manufacturing process, or exposure to incidental airborne contamination. These impurities can interfere with detection by interacting with the analyte via competitive ionization, thereby reducing the analyte signal or contributing to an elevated background noise.^8,24 Cleaner surfaces provide more uniform of release of TNT, as well as better mass and heat transfer. In the second case, preconditioning eliminates water associated with the swipe surface, contributing to a more uniform interaction between the surface and the analyte, as well as reducing the humidity created when the swipe is heated. It is generally accepted that capillary forces from humidity increase adhesion forces between explosive particles and the substrate surfaces.³⁴ As a consequence, lower than expected signals from TNT release are observed in more humid conditions. Therefore, it is likely that higher IMS signals from TNT were observed from the preconditioned swipes, because of cleaner surfaces and reduced surface-associated water, both of which may also be expected to increase over time when the materials are not properly stored, however, the effects of storage were not examined in this study.

FGC having an organic coating, known industrially as “Greige”, showed the lowest normalized signal of all FGC tested, with a response roughly equivalent to muslin. The Greige surface chemistry is composed of a variety of chemicals, such as starches, polyvinyl alcohol, lubricants, humectants, and other organic sizing/binders, each of which is likely to interact with TNT. Additionally, most FGC materials have organic surface additives to enhance flexibility, durability, product integration and manufacturing processes.^35,36 Such a heterogeneous surface may contribute to irreversible binding of a portion of the analyte, as well as contributing a diverse range of intermolecular interactions. Both of these require a thermally-broad range of analyte release energies, leading to a net reduction in analyte signal intensity. Compensating for the breath and range of interaction energies by using higher desorption temperatures could result in thermal decomposition of an explosive analyte with concomitant reduction of analyte signal.^19,37

The unfinished FGC (537A), cleaned of organics and with a reduced surface hydroxyl (–OH) population resulting from industrial thermal treatment, can be seen to provide better thermal release of TNT, than did the Greige or muslin. Here, the thermal treatment removed surface additives and served to reduce surface hydroxyls by promoting surface condensation reactions. Similarly, the chemical composition of the glass fibers can also be observed to have an impact upon the detection of TNT. The preconditioned S glass sample produces a greater TNT signal than E glass having the same applied surface chemistry. The principle difference between the materials is that E glass has calcium oxide (16–25%) and boron oxide (8–13%) added to the silica and alumina formulation which is not present in S glass. These E glass additives reduce the number of available sites for surface binding and may induce analyte-trapping surface defects. Additionally, S glass has slightly higher thermal conductivity and slightly lower specific heat (see Table SI 1†),^25,32,33 which provide faster thermal release of analytes than E glass, resulting in a higher amplitude IMS signal (Fig. SI 1†). Astroquartz (99.99% silica), a pure silica fiber, provides a higher density of surface hydroxyl sites for surface functionalization. As can be seen in Table 1, a silane finish on pure silica cloth provides a better signal (when thermally preconditioned) than unfinished FGC (537A). However, the brittle fibers of the pure silica material fracture easily, making the material physically much less durable (particularly for the thinner material), and are significantly more expensive than the other fiberglass materials.

Some silane-finished FGC materials (i.e., 497A and CS742) in Table 1 can be seen to provide almost four times the TNT signal intensity observed from muslin. Typically, coatings that result from commercial surface chemistries are formed from several components and are applied in several steps. For example CS742 has a coating formed by treating the FGC with an epoxy resin and an additional silane finish. The 497A has silane surface chemistry terminated in amines and vinyl groups for integration with polymers. In these cases, the silane-finished materials in Table 1 showed better TNT release than the unfinished FGC (537A). This indicates that the surface hydroxyl groups and the oxide surface on the unfinished FGC (537A) interact strongly with TNT, although higher desorption temperatures or longer desorption times can increase the amount of TNT released.^34,38 However, the quality of the surface layer can vary, and various silane functional groups interact differently with the analyte, impacting analyte recovery. For example, amine groups (present in 497A) have been observed to react with the nitro groups of TNT via hydrogen bonding, providing a different interaction than hydrogen bonding from hydroxyl groups.^39,40

A PTFE coating on FGC provides a relatively uniform and nonpolar surface chemistry, which results in a higher TNT release signal when compared with muslin swipes (Table 1). In addition, the low surface energy of fluorocarbons results in weak van der Waals forces, making PTFE-coated FGC a good material for thermal desorption of the analyte. However, the low attractive forces can also result in poor collection of organic residues via swipe sampling.²

Activated carbon is typically viewed as an ideal material for the collection of organic analytes. However, an IMS signal was not observed from the activated carbon cloth (Zorflex®) deposited with TNT. Thermogravimetric analysis (TGA) shows a gradual release of TNT from activated carbon at desorption temperatures in excess of 150 °C. This gradual thermal release is attributed to the heterogeneity of the activated carbon surface.^41,42 Partial thermal decomposition of TNT likely occurs before thermal release from the surface and it has been reported that activated carbon promotes rapid TNT oxidation through the TNT methyl groups,⁴³ forming nonextractable residues.^43,44 While the carbon material has a high affinity for TNT and other analytes, the difficulty in attaining thermal release of TNT demonstrates that a balance must be struck between analyte affinity and ease of release. Recovery of TNT (and other analytes of interest) from carbon sampling materials by solvent extraction¹ has been shown, but solvent extraction is not the preferred method for field use because of challenges presented by liquid processing and consumables.

3.2 Impact of swipe material thickness

Sampling material thickness directly impacts its heating characteristics, analyte release from the sampling material, airflow through the material and durability of the sampling material. Sampling material thickness has also been reported to impact analyte collection efficiency.^8,24 To quantitatively evaluate the impact of thickness on analyte release, sampling materials having the same surface chemistry and weave pattern, but of different thicknesses, were evaluated. Comparative results from these tests are provided in Table 2. IMS responses demonstrated that for all materials, including commercial muslin swipes, the thickness of the sampling material significantly impacted the TNT release. Thicker sampling materials resulted in lower TNT signal intensities when compared to thinner material. For example, a 0.06 mm increase in muslin thickness was observed to reduce the TNT signal by over 50%. Additionally, increased thickness for swipe materials increases the influence of physical characteristics, such as thermal conductivity and specific heat capacity. Representative reference data for several materials is shown in Table SI 1,† where fiberglass materials are shown to have higher thermal conductivity and lower specific heat capacities than cotton/cellulose. These trends suggest that fiberglass materials, for materials having the same thickness, will heat more rapidly than muslin, providing a higher IMS response.

Table 2 The impact of swipe thickness on IMS response to TNT release^a

Material	Thickness (mm)	Surface area (m^2 g^−1)	TNT normalized signal
a Material preconditioned at 180 °C overnight under vacuum. IMS signal normalized to the IMS signal for a 10 ng TNT deposition on muslin. Note: muslin was not preconditioned.
Muslin-thick	0.32 ± 0.01	1.07	1.0 ± 0.1
Muslin-thin	0.26 ± 0.01	1.04	2.1 ± 0.3
PP E FGC-thick	0.27 ± 0.03	0.40	6.0 ± 0.4
PP E FGC-med	0.20 ± 0.01	0.33	6.8 ± 0.3
PP E FGC-thin	0.10 ± 0.01	0.48	8.2 ± 0.2
PSiSF FGC-thick	0.29 ± 0.02	0.22	5.6 ± 0.4
PSiSF FGC-thin	0.14 ± 0.01	0.29	6.2 ± 0.4

---

Air permeability of the swipe materials, which is a function of weave and thickness, is also known to impact the release of an adsorbed analyte.^8,24 The results indicate that thinner swipe fabrics release more analyte, or release the analyte over a shorter time interval, resulting in a larger measured signal. Thinner materials having the same weave improve air flow and reduce the opportunity for desorbed analytes to encounter another fiber within the cloth, thereby improving performance. While there are clear advantages for thinner collection materials, the sampling material must have sufficient thickness to endure chemical production processes and the physical sampling forces encountered during field applications.

3.3 Impact of surface functionalization and fiberglass type

Organically-modified surfaces of collection materials have been shown to increase the capability of adsorption, selectivity, detection, sampling, and preconcentration of trace explosives.^{26,39,40,45–50} Silica-based materials are often used as the support material of trace collection materials due to the ease of installing a wide range of surface chemistries using organosilane ligands.^{39,40,45,46,48,50}

The nature of the fiberglass surface can provide organosilane ligands to be uniformly distributed across the surface, occupying nearly all of the available reactive surface hydroxyl sites. When a single surface chemistry is installed, the resulting surface is likely more uniform than muslin, or most commercially available fiberglass. In the case of muslin, cellulose presents a range of possible chemical interactions, while for commercial fiberglass, several chemicals are deposited on the surface during production. Phenylsilane monolayers were chosen as the surface chemistry to be installed on the FGC for evaluation in this study. A phenyl monolayer provides a lipophilic surface with general affinity for organic materials and some additional chemical selectivity for TNT and other nitroaromatics.^34,51 A combination of aromatic π–π interactions and polarizability of the aromatic ring are thought to make the phenylsilane monolayer more specific for TNT than the general hydrogen bonding that occurs between TNT (ring and nitro groups) and the fiberglass surface (hydroxyls or commercial finishes with amines and other groups).^34,51–53 This result relates to a previous work which reported that the phenyl end group showed a stronger adhesion force for TNT particles than those observed from other end groups including –OH, –NH₂, –COOH, CF₃, etc., which should result in better collection efficiency.³⁴ The phenyl end group was also observed to have strong adhesion to other explosive particles (i.e., RDX, HMX, and PETN) in ambient air and under water measurements.³⁴ Additionally the aromatic phenylsilane is also thermally robust, and was found to be stable to temperatures greater than 400 °C. TGA results show that phenyl-functionalized silica releases TNT between 145–230 °C, demonstrating that phenyl-functionalized materials are suitable for collection and thermal desorption of TNT (and many other organic chemicals).

Table 3 shows the impact of a sequential adjustment of the sampling material and surface chemistry on the TNT signals (for a selected material and normalized to muslin with a 10 ng TNT deposition). Surface chemistry, type of fiberglass material, functionalization method, and the thickness were systematically altered to increase the analytical signal. An additional test set with 5 ng TNT depositions was used to elucidate the maximum instrument response, since it is well documented that the IMS has a limited linear range and can easily become saturated. Saturation occurs when most of the reactant ions are depleted, such as when excess analytes are introduced into the system.²⁹ This appears to be the case for the final three entries, phenyl and BF-phenyl on S FGC, and BF-phenyl on thin S FGC, as is shown by these samples having similar response values for 10 ng TNT deposition, but differentiated response values for 5 ng TNT deposition.^7,54,55

Table 3 IMS response to TNT release from functionalized fiberglass sampling material

Material^a	TNT normalized signal^b (10 ng)	TNT normalized signal^b (5 ng)
a E and S indicate glass types.b Materials preconditioned at 180 °C overnight under vacuum. IMS response to 10 ng TNT depositions.
Muslin	1.0 ± 0.1	0.4 ± 0.3
PPE FGC	4.6 ± 0.2	2.3 ± 0.3
Phenyl E FGC	5.0 ± 0.2	2.5 ± 0.2
Phenyl S FGC	9.1 ± 0.2	5.8 ± 0.4
BF-phenyl S FGC	9.7 ± 0.5	7.1 ± 0.2
Thin BF-phenyl S FGC	10.2 ± 0.3	7.8 ± 0.2

---

As discussed and shown previously (Tables 1 and 2), different surface chemistries significantly affect the TNT release, with the polymer primer (PP) coating (497A) providing one of the more effective surfaces. Replacing the commercial PP coating on the FGC with a phenylsilane surface chemistry can be observed to increase the TNT signal-normalized to muslin (Table 3). Changing the fiberglass material from E glass to S glass also significantly improves the TNT release, which is due in part to the improved thermal properties of S glass (Fig. SI 1 and Table SI 1†). As mentioned above, improved thermal properties provide faster desorption and more intense analyte peaks. Additionally, with a higher fraction of silica in its composition, more surface hydroxyl sites are available on the S glass fibers for phenylsilane deposition, presumably resulting in a higher density monolayer, as compared to E glass. The importance of having a high surface coverage of the phenylsilane was recognized from observations that only a low TNT signal, or no TNT signal, was detected when using unmodified E or S glass FGC, pretreated to remove the as-received surface chemistry or chemicals. In this case, the pretreated FGC surface may irreversibly bind or decompose TNT molecules at the surface hydroxyl or metal oxide sites. To minimize the surface hydroxyl population, two consecutive phenylsilane treatment steps were performed, with the purpose of the second treatment to back-fill any hydroxyl sites that remained following the first treatment. A slight, but noticeable improvement in performance was observed following the second phenylsilane treatment, BF-phenyl, suggesting that further surface hydroxyls on the FGC were available for reaction with the phenylsilane. Comparing data sets for BF-phenyl materials loaded with 5 and 10 ng of TNT suggests that some IMS response saturation was occurring for the 10 ng TNT depositions, and this is shown to be the case in a comparative study presented in Fig. 3 and SI 2.† Finally, an additive improvement is shown by data in Table 3, demonstrating that reducing the thickness of the sampling material further increases the TNT release, as discussed previously (Table 2). Similar performance trends can be observed with other FGC substrates (see Table SI 3†).

Phenyl-functionalization of the FGC showed a significant improvement in TNT release and subsequent IMS detection. Efforts to improve surface functionalization and passivation of residual reactive sites are likely to yield further improvements. For example, using smaller silanes or alternative silane-deposition methods, such as supercritical fluids or plasmas, may passivate residual surface hydroxyls, further reducing hydroxyl–TNT interactions, and improving the TNT signal. However, installation of the phenyl surface chemistry depends upon the substrate material, substrate material preparation and deposition process.^56–58 Minor differences among these factors can produce diverse results, as observed in this work and reported in literature.^57,58

3.4 Material characteristics for TNT-signal enhancement

A comparison of TNT-signal profiles as a function of desorption time and temperature for standard muslin, FGC and phenyl-functionalized FGC is shown in Fig. 1. The as-received E FGC (CS724 silane finish) and phenyl E FGC were tested, with IMS signals normalized to the maximum TNT signal from muslin at a desorption temperature of 180 °C. At a desorption temperature of 150 °C, the as-received E FGC had a comparable signal to that of muslin, while the phenyl E FGC showed a normalized signal approximately 2.5 times greater than the normalized muslin signal. When the desorption temperature was increased, the as-received E FGC and the phenyl E FGC showed improved release of TNT, with the phenyl E FGC showing a normalized signal greater than 4.5 times and 5.7 times that of muslin, at 180 °C and 200 °C, respectively. At a desorption temperature of 290 °C, the absolute and relative performance of the as-received E FGC decreased by nearly a factor of two, but only slightly decreased for phenyl E FGC. This is likely a result of partial thermal decomposition of some of the TNT, a process that is known to occur near this temperature.⁵⁹ For these samples, a TNT decomposition product peak was also observed at a faster drift time than the TNT peak, as was also noted by Eiceman et al.³⁷ A lack of observed TNT signal at 290 °C for muslin was attributed to the concurrent partial thermal decomposition of the sample cloth, which showed significant discoloration after heating in the IMS. At 200 °C the phenyl E FGC demonstrated the greatest IMS response, in terms of absolute TNT signal, and normalized TNT signal.

Fig. 1 Impact of IMS desorption temperature on analyte (10 ng TNT depositions) release from commercial muslin, as-received FGC, and phenyl-functionalized E FGC. The TNT signals shown are normalized to the maximum TNT signal from muslin at desorption temperature of 180 °C. As desorption temperature is increased, normalized performance of FGC improves over muslin. No signal was observed for muslin at a 290 °C desorption temperature.

Independent of desorption temperature, Fig. 1 and 2 show there are three principle characteristics influencing TNT desorption: (1) surface chemistry, (2) inherent material properties, and (3) physical construction; which together enable functionalized FGC to provide a higher TNT release than muslin. The first characteristic, phenyl surface chemistry, installed on the FGC resulted in a larger integrated IMS signal than the muslin swipe at all desorption temperatures, as can be observed by the greater areas under the response curves in Fig. 1 and 2. This is significantly more pronounced for the BF-phenyl S FGC and thin BF-phenyl S FGC shown in Fig. 2. In contrast, FGC without surface functionalization, pretreated using a NaOH(aq) base rinse (which provides a large number of surface hydroxyl groups), had a much poorer TNT release than muslin. Fig. 2 shows the TNT response of the materials listed in Table 3. Taken together, these plots show TNT is released faster by functionalized FGC. An additional surface functionalization step (BF-phenyl) enhances the quantity of analyte released, as well as the rate of analyte release. Fig. 1 and 2 show that a more favorable surface chemistry significantly improves the fraction of analyte released and the rate of analyte release, contributing to larger TNT signals.

Fig. 2 Impact of surface modification and material thickness on 10 ng TNT detection using IMS at a desorption temperature of 180 °C. An increase in the uniformity of surface chemistry improves the normalized signal and narrows the desorption peak. Decreasing substrate thickness (0.26 mm to 0.10 mm) with better substrate properties further narrows the peak. Narrower release curves give rise to better detection limits. The peaks from left to right were obtained by using the following FGC substrates for surface modification: 0.1 mm S FGC, 0.26 S FGC, 0.27 E FGC, 0.23 E FGC (used as received), and muslin, respectively.

The beneficial effects FGC material properties and physical construction, the second and third characteristics, can also be observed in Fig. 1 and 2. The FGC materials, and particularly the functionalized FGC materials, provide a faster signal rise and more pronounced peak shape sharpening than muslin. These effects result from FGC having intrinsically better thermal properties than muslin, resulting in faster thermal desorption of adsorbed analytes. Compared to muslin, FGC has nearly five times the thermal conductivity, and a lower (approximately one-half) specific heat capacity (Table SI 1†), with both promoting faster heating and better analyte release (if enabled by appropriate surface chemistry).^25,32,33 The thin S FGC offers still faster desorption and a higher IMS signal because of better air permeability than thick S FGC, and better thermal properties than E FGC.

3.5 Impact on detection limits

The normalized IMS response curves for TNT are shown in Fig. 3 for as-received S FGC, thin BF-phenyl S FGC, and muslin. IMS responses were obtained by analyzing sampling swipes deposited with 0.1 to 50 ng of TNT, with all of the results normalized to the maximum IMS signal for 10 ng TNT deposited on muslin and thermally desorbed at 180 °C. Both FGC samples yielded better detection limits than muslin. The thin BF-phenyl S FGC was capable of producing a detectable signal with as low as 0.2 ng deposited TNT. Overall, the thin BF-phenyl S FGC showed an approximately 10 times signal improvement over muslin. At higher concentrations, the nominal increase in signal above 10 ng deposited TNT for thin BF-phenyl FGC suggests that the IMS system was becoming saturated, as discussed previously. These data suggest the phenyl group improves the release of TNT from phenyl-functionalized FGC (thin BF-phenyl FGC) over an as-received FGC surface and muslin, and in doing so, can lower the overall system detection limits.

Fig. 3 IMS response curves for muslin, thin as-received S FGC, and thin BF-phenyl S FGC. All values are normalized to muslin's average maximum signal intensity at 10 ng TNT deposition with a desorption temperature of 180 °C. Standard deviations from 5 measurements are shown for each data point.

3.6 Reusability

A brief reusability study was performed to assess thermal stability after multiple uses, comparing muslin and thin BF-phenyl S FGC. This study consisted of replicate deposition–desorption cycles on the same fabric swatch of sampling material to assess reusability of the sampling cloth. A cycle consisted of a 10 ng TNT deposition, followed by thermal desorption in the IMS. This was repeated a total of 10 cycles for each material. A plot of percent change, relative to the initial response for each material, can be seen in Fig. 4. Thin BF-phenyl S FGC demonstrated less than a 5% change in signal intensity over 10 replicate cycles with stable, and low IMS background noise (Fig. SI 3†). The muslin TNT response was observed to double for the second cycle and then generally increase for each additional cycle, suggesting that muslin improves after the first use, and continues to improve, but at a much lower rate, for subsequent cycles. The signal variation is not as pronounced for thermally preconditioned muslin, although some improvement is evident after the first cycle, roughly leveling off after the fourth cycle. Additionally, the background IMS noise for the thermally preconditioned muslin is more stable than muslin without preconditioning (Fig. SI 3†). While all of the tested materials improved with use, likely as a result of thermally desorbing contaminants, the stability and low noise associated with repeated uses of the thin BF-phenyl S FGC, suggest that these materials may provide a reliable platform for swipe materials.

Fig. 4 Absolute percent change of IMS signal intensity normalized to initial response for replicate detections using three sampling materials deposited with 10 ng of TNT for each cycle at a desorption temperature of 180 °C. The thin BF-phenyl S FGC, showed very repeatable responses compared to the first cycle. Standard muslin showed a two-times-signal increase and then a gradual signal increase. Thermally-treated muslin can be seen to have a much smaller signal variance. The small cycle-to-cycle variation for thin BF-phenyl S FGC suggests uniform performance for repeated uses.

3.7 Comparative collection and detection of explosive particles from surfaces

Comparative collection of TNT residues was explored to determine surface sampling efficacy. Each measurement was made by depositing a 30 ng TNT sample on a clean, nonporous PTFE surface, and allowing the volatiles to evaporate. This process resulted in small particles of TNT remaining on the surface. The surfaces were sampled with standard muslin material and thick phenyl-functionalized FGC (phenyl S FGC) and the TNT signals from surface-sampling materials (triplicate runs) were compared. As surface samplers, phenyl-functionalized S FGC showed approximately two times greater signal intensity over muslin for 260 °C desorption. A similar signal intensity difference was observed comparing results from 10 ng TNT deposited directly on the FGC and muslin. The similarity in signal differences derived from particle surface sampling and from deposited TNT suggests that the TNT particle collection efficiencies for both materials are similar, although a previous study suggests that the phenyl group installed on a substrate surface can increase the adhesion of TNT particles,³⁴ but this is not clearly demonstrated here. Other studies have investigated the influences of surface morphology and chemical forces on the collection of explosives particles and have attempted to quantify these effects.^60,61 However, considerably more work would be needed to make a definitive statement regarding relative collection efficiencies of FGC and muslin, or other materials.

3.8 Application to other explosives

While it was useful for this study to focus on TNT as a means of comparing the performance of muslin and several fiberglass sampling cloths, examining the general applicability of phenyl-functionalized FGC to the detection of other explosives shows a broader applicability to explosives detection. With that in mind, thin BF-phenyl S FGC, the material identified as offering the best performance for thermal release of TNT, was tested using a variety of explosives samples, as shown in Table 4. Similar to TNT, set amounts of each explosive were deposited on the BF-phenyl S FGC, thermally desorbed at 180 °C in the IMS, and the signal normalized to the maximum signal provided by the same amount explosive deposited on muslin. It is likely that much could be said about the differences in signal intensity observed for the explosive compounds; however, the data is presented here for comparative purposes and further study is necessary to understand the observed differences in signal intensities.

Table 4 IMS responses for 10 ng of various explosives release from BF-phenyl on thin S FGC

Compound	Normalized signal^a	Structure
a IMS signals normalized to the IMS signal for the same amount of each explosive deposited on muslin, at a desorption temperature of 180 °C.
TNT	10.2 ± 0.3
NG	3.4 ± 0.2
RDX	1.5 ± 0.1
TATP	1.4 ± 0.3
PETN	1.3 ± 0.1

---

4. Conclusions

In this work we have explored some of the fundamental attributes of fiberglass cloth used for surface sampling in explosives detection application. We show how to advance state of art in explosives detection with the application of foundational material science and surface chemistry principles. Using TNT as a relevant analyte of interest, the impact of basic properties inherent to the muslin and fiberglass sampling cloths were comparatively tested by systematically examining the impact of each attribute on the thermal release of deposited TNT into an IMS. We have shown that results from this study are applicable to the detection of other explosive compounds and similarly expect that the results are generally applicable to a wide range of other chemical sampling challenges.

We show phenyl-functionalized FGC, as a sampling material for trace collection and detection, can provide considerable improvements in detection limits. When compared to muslin, the phenyl-functionalized FGC showed improved analyte release, equivalent analyte collection, and repetitive use without degradation. Performance improvements were derived from three major factors. Physically, FGC has desirable thermal properties (i.e., higher thermal conductivity and lower specific heat capacity than muslin), enabling the sampling material to be heated more quickly and evenly than muslin, thereby providing better analyte release into the IMS. Second, thinner sampling material, with reduced thermal mass and better gas permeability, providing faster release and improved transport of the desorbed analyte to the detector. Finally, covalently bound, phenyl surface functionalization is considered to increase surface uniformity, reducing variations in analyte binding sites, and thereby promoting a more uniform analyte release, than the non-uniform surface chemistry of as-received FGC or muslin collection materials. When these factors are fully addressed in a single sampling material, such as for thin BF-phenyl S FGC, we achieved the best sampling performance of the materials tested.

Enhancing the collection and detection of trace organic residues from surfaces has immediate and significant applications to explosives detection but also has broader utility to forensic, biomedical, and environmental analytical challenges. Future efforts will report upon the synthetic challenges of installing high density low defect surface layers on FGC, examine additional surface chemistries to improve selectivity and applications of these collection materials to other analytical challenges.

Acknowledgements

This work was performed at Pacific Northwest National Laboratories (PNNL), which is operated for the DOE by Battelle Memorial Institute under contract DE AC06-76RLO 1830. We wish to acknowledge and thank Joe Davidson, Jackie Roberts, Melanie Waltman and Glen Fryxell for helpful conversations and input. This work was partially supported by the PNNL Laboratory Directed Research and Development (LDRD).

References

1. D. Perret, S. Marchese, A. Gentili, R. Curini, A. Terracciano, E. Bafile and F. Romolo, Chromatographia, 2008, 68, 517–524.
2. J. R. Verkouteren, J. L. Coleman, R. A. Fletcher, W. J. Smith, G. A. Klouda and G. Gillen, Meas. Sci. Technol., 2008, 19, 115101.
3. R. D. Voyksner and J. Yinon, J. Chromatogr. A, 1986, 354, 393–405.
4. D. A. Shea and D. Morgan, Detection of Explosives on Airline Passengers: Recommendation of the 911 Comission and Related Issues, Report Order Code RS21920, Congressional Research Sevice, The Library of Congress, 2007.
5. Guide for the Seelection of Commercial Explosives Detection System for Law Enforcement Applications Report NIJ Guide 100–99, U.S. Department of Justice, Office of Justice Programs, National Institute of Justice, 1999.
6. J. S. Caygill, F. Davis and S. P. J. Higson, Talanta, 2012, 88, 14–29.
7. G. W. Cook, P. T. LaPuma, G. L. Hook and B. A. Eckenrode, J. Forensic Sci., 2010, 55, 1582–1591.
8. L. L. Danylewch-May, US Pat., US2006/0192098 A1, 2006.
9. W. A. MacCrehan, Anal. Chem., 2009, 81, 7189–7196.
10. J. C. Oxley, J. L. Smith, L. J. Kirschenbaum, S. Marimganti and S. Vadlamannati, J. Forensic Sci., 2008, 53, 690–693.
11. G. A. Eiceman and J. A. Stone, Anal. Chem., 2004, 76, 390A–397A.
12. R. G. Ewing, D. A. Atkinson, G. A. Eiceman and G. J. Ewing, Talanta, 2001, 54, 515–529.
13. G. A. Eiceman, E. V. Krylov, N. S. Krylova, E. G. Nazarov and R. A. Miller, Anal. Chem., 2004, 76, 4937–4944.
14. L. Theisan, D. Hannum, D. W. Murray and J. E. Parmeter, Survey of commercially, available explosives detection technologies and equipment 2004, Report 208861, Sandia National Laboratories, 2005.
15. J. R. Verkouteren, J. Forensic Sci., 2007, 52, 335–340.
16. A. B. Kanu, C. Wu and H. H. Hill Jr, Anal. Chim. Acta, 2008, 610, 125–134.
17. Y. Takada, H. Nagano, M. Suga, Y. Hashimoto, M. Yamada, M. Sakairi, K. Kusumoto, T. Ota and J. Nakamura, Propellants, Explos., Pyrotech., 2002, 27, 224–228.
18. N. Talaty, C. C. Mulligan, D. R. Justes, A. U. Jackson, R. J. Noll and R. G. Cooks, Analyst, 2008, 133, 1532–1540.
19. D. S. Moore, Rev. Sci. Instrum., 2004, 75, 2499–2512.
20. S. Tao, Z. Shi, G. Li and P. Li, ChemPhysChem, 2006, 7, 1902–1905.
21. M. S. Beardah, S. P. Doyle and C. E. Hendey, Sci. Justice, 2007, 47, 120–124.
22. I. A. Popov, H. Chen, O. N. Kharybin, E. N. Nikolaev and G. R. Cooks, Chem. Commun., 2005, 42, 1953–1955.
23. N. Na, C. Zhang, M. Zhao, S. Zhang, C. Yang, X. Fang and X. Zhang, J. Mass Spectrom., 2007, 42, 1079–1085.
24. R. H. Bozenbury Jr, L. L. Danylewych-May, L. Fricano and L. Kim, US 2005/0288616 A1, 2005.
25. M. Harris, Harris's Handbook of Textile Fibers, Harris Research Laboratories, Inc., Washington, D. C., 1954.
26. D. S. Moore, Sens. Imag. Int. J., 2007, 8, 9–38.
27. R. Waddell, D. E. Dale, M. Monagle and S. A. Smith, J. Chromatogr. A, 2005, 1062, 125–131.
28. M. E. Sigman and C.-Y. Ma, Anal. Chem., 1999, 71, 4119–4124.
29. J. L. Staymates, J. Grandner and G. Gillen, Anal. Methods, 2011, 3, 2056–2060.
30. K. L. Loewenstein, The Manufacturing Technology of Continuous Glass Fibers, Elsevier Scientific, New York, 1973.
31. R. Charles and R. Farwaha, France Pat., EP1599427B1, 2009.
32. JPS Composite Materials Databook, http://jpsglass.com/.
33. G. Lubin, Handbook of Fiberglass and Advanced Plastic Composites, ed. E. Robert, Krieger Pub. Co., Huntington, N.Y, 1969.
34. Y. Zakon, N. G. Lemcoff, A. Marmur and Y. Zeiri, J. Phys. Chem. C, 2012, 116, 22815–22822.
35. J. Bjorksten, L. L. Yaeger and J. E. Henning, Ind. Eng. Chem., 1954, 46, 1632–1635.
36. E. P. Plueddemann, Silane Coupling Agents, Plenum, New York, 1991.
37. G. A. Eiceman, D. Preston, G. Tiano, J. Rodriguez and J. E. Parmeter, Talanta, 1997, 45, 57–74.
38. H.-X. Zhang, A.-M. Cao, J.-S. Hu, L.-J. Wan and S.-T. Lee, Anal. Chem., 2006, 78, 1967–1971.
39. F. An, B. Gao and X. Feng, J. Hazard. Mater., 2009, 166, 757–761.
40. F. An, X. Feng and B. Gao, J. Hazard. Mater., 2009, 168, 352–357.
41. C. O. Ania and T. J. Bandosz, Langmuir, 2005, 21, 7752–7759.
42. A. Dabrowski, P. Podkoscielny, Z. Hubicki and M. Barczak, Chemosphere, 2005, 58, 1049–1070.
43. G. K. Vasilyeva, V. D. Kreslavski and P. J. Shea, Chemosphere, 2002, 47, 311–317.
44. V. Marinovic, M. Ristic and M. Dostanic, J. Hazard. Mater., 2005, 117, 121–128.
45. F. An, B. Gao and X. Feng, Appl. Surf. Sci., 2009, 255, 5031–5035.
46. J. Feng, Y. Li and M. Yang, Sens. Actuators, B, 2010, 145, 438–443.
47. D. Gao, Z. Wang, B. Liu, L. Ni, M. Wu and Z. Zhang, Anal. Chem., 2008, 80, 8545–8553.
48. D. Gao, Z. Zhang, M. Wu, C. Xie, G. Guan and D. Wang, J. Am. Chem. Soc., 2007, 129, 7859–7866.
49. C. Xie, B. Liu, Z. Wang, D. Gao, G. Guan and Z. Zhang, Anal. Chem., 2008, 80, 437–443.
50. A. Yildirim, H. Budunoglu, H. Deniz, M. O. Guler and M. Bayindir, ACS Appl. Mater. Interfaces, 2010, 2, 2892–2897.
51. J. A. Greathouse, N. W. Ockwig, L. J. Criscenti, T. R. Guilinger, P. Pohl and M. D. Allendorf, Phys. Chem. Chem. Phys., 2010, 12, 12621–12629.
52. O. T. Ikkala, L. O. Pietilä, P. Passiniemi, T. Vikki, H. Österholm, L. Ahjopalo and J. E. Österholm, Synth. Met., 1997, 84, 55–58.
53. L. S. Reddy, P. M. Bhatt, R. Banerjee, A. Nangia and G. J. Kruger, Chem.–Asian J., 2007, 2, 505–513.
54. G. A. Eiceman and Z. Karpas, Ion Mobility Spectrometry, CRC Press, Boca Raton, FL, 1994.
55. M. Tabrizchi and V. Ilbeigi, J. Hazard. Mater., 2010, 176, 692–696.
56. C. Haensch, S. Hoeppener and U. S. Schubert, Chem. Soc. Rev., 2010, 39, 2323–2334.
57. S. Onclin, B. J. Ravoo and D. N. Reinhoudt, Angew. Chem., Int. Ed., 2005, 44, 6282–6304.
58. A. Ulman, Chem. Rev., 1996, 96, 1533–1554.
59. T. B. Brill and K. J. James, Chem. Rev., 1993, 93, 2667–2692.
60. M. N. Chaffee-Cipich, B. D. Sturtevant and S. P. Beaudoin, Anal. Chem., 2013, 85, 5358–5366.
61. M. N. Chaffee-Cipich, D. J. Hoss, M. L. Sweat and S. P. Beaudoin, Forensic Sci. Int., 2016, 260, 85–94.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra20157a

---