CHEMICAL REACTIONS OF EXPLOSIVE MOLECULES FOR DETECTION APPLICATIONS

Explosive analytes and their decomposition products are of great interest to the scientific community, in large part due to events of international terrorism and warfare, but also as environmental pollutants. Chapter 1 of this dissertation shows for the first time that the hydroxide adducts of trinitrobenzene (TNB) and trinitrotoluene (TNT), TNB-OH – and TNT-OH – , are emissive while TNT – is not. This has great impact on pre-existing sensors, which may be affected negatively by an increase in emission competing with the observation of a quench. Additionally, we described a competing reaction with the solvent, N,N-dimethylformamide (DMF), which is also capable of nucleophilic attack upon TNT and observable due to its overwhelming quantity. Chapter 2 of this dissertation shows the similarities of TNT products formed by OH – exposure and amine exposure, covering a wide variety of amines. 1 H NMR and rapid absorbance measurements showed formation of TNT – or TNT adducts, the relative rates of which are observed to be strongly dependent on solvation. Intrinsic rate constants in methanol implied that all amine reactions were forming the same adduct (TNT-OCH3 – ). The credibility of amine adduct formation was further explored in Chapter 3 through computational approaches. High level ab initio calculations were performed to obtain models of reactants and products of the investigated reactions, and their relative energies and thermodynamic quantities were computed. The data showed a striking disparity in the calculated thermodynamics, with OH – and OCH3 – adduct formation being much more favorable than amine adduct formation. Charge transfer (CT) complexes between TNT and the explored amines were converged and also found to be of higher energy than alkoxide adducts. In Chapter 4, a sensing array based on highly fluorescent reporter molecules in DMF solution is described. Eight xanthene-based fluorophores were chosen based on their high quantum yields, and their interactions with twelve relevant explosive analytes were interpreted through absorbance and emission data. The resulting array showed promise in observing an identifying "fingerprint" response to each class of analyte, with the largest responses coming from the formation of TNB and TNT products. In addition, these products were observed to be involved in an electron transfer (ET) mechanism where they donated energy to cationic fluorophores, enhancing the fluorophores' emission. The findings of this dissertation indicate the need for caution in TNT sensor development, as trace water provides OH – and common solvents are capable of competing for TNT reactivity. Trace water should be monitored and minimized for appropriate sensor testing and subsequent use, and solution based sensors require an understanding of solvent competition. Amines are not capable of direct deprotonation of TNT, and do not compete significantly with OH – or alkoxides in forming a sigma adduct with TNT. Rather, the formation of colorful complexes that has been attributed to a TNT-NR3 adduct in the literature is due to the deprotonation of ambient water or solvent molecules, which in turn react with TNT. Finally, xanthene-based fluorophores may selectively interact with explosive analytes, with trinitroaromatic products capable of electron transfer to fluorescent reporters leading to emission enhancement.

emission competing with the observation of a quench. Additionally, we described a competing reaction with the solvent, N,N-dimethylformamide (DMF), which is also capable of nucleophilic attack upon TNT and observable due to its overwhelming quantity.
Chapter 2 of this dissertation shows the similarities of TNT products formed by OHexposure and amine exposure, covering a wide variety of amines. 1 H NMR and rapid absorbance measurements showed formation of TNTor TNT adducts, the relative rates of which are observed to be strongly dependent on solvation. Intrinsic rate constants in methanol implied that all amine reactions were forming the same adduct (TNT-OCH 3 -). The credibility of amine adduct formation was further explored in Chapter 3 through computational approaches. High level ab initio calculations were performed to obtain models of reactants and products of the investigated reactions, and their relative energies and thermodynamic quantities were computed. The data showed a striking disparity in the calculated thermodynamics, with OHand OCH 3 adduct formation being much more favorable than amine adduct formation. Charge transfer (CT) complexes between TNT and the explored amines were converged and also found to be of higher energy than alkoxide adducts.
In Chapter 4, a sensing array based on highly fluorescent reporter molecules in DMF solution is described. Eight xanthene-based fluorophores were chosen based on their high quantum yields, and their interactions with twelve relevant explosive analytes were interpreted through absorbance and emission data. The resulting array showed promise in observing an identifying "fingerprint" response to each class of analyte, with the largest responses coming from the formation of TNB and TNT products. In addition, these products were observed to be involved in an electron transfer (ET) mechanism where they donated energy to cationic fluorophores, enhancing the fluorophores' emission.
The findings of this dissertation indicate the need for caution in TNT sensor development, as trace water provides OHand common solvents are capable of competing for TNT reactivity. Trace water should be monitored and minimized for appropriate sensor testing and subsequent use, and solution based sensors require an understanding of solvent competition. Amines are not capable of direct deprotonation of TNT, and do not compete significantly with OHor alkoxides in forming a sigma adduct with TNT. Rather, the formation of colorful complexes that has been attributed to a TNT-NR 3 adduct in the literature is due to the deprotonation of ambient water or solvent molecules, which in turn react with TNT. Finally, xanthene-based fluorophores may selectively interact with explosive analytes, with trinitroaromatic products capable of electron transfer to fluorescent reporters leading to emission enhancement. x

FIGURE PAGE
Chapter 1:   While TNT and TNB share this type of reactivity, TNT also has acidic hydrogens that can be abstracted to form the TNT anion (TNT -). [9][10][11][12] Since both TNT products have absorbances in similar locations in the visible range, 9,10 and basicity and nucleophilicity tend to go hand in hand, it is a precarious process to be able to differentiate TNT sigma adducts from TNT -.
It is commonly believed that nitroaromatics and their nucleophilic substitution products are not fluorescent, and therefore additional mechanisms need to be imposed in order to address their sensing using a fluorescence method. Many sensors rely on the fluorescence quenching of a reporter molecule, where the emission of a selected fluorophore decreases as a function of interacting with an explosive analyte. This process may result from either dynamic or static quenching, but is most often observed as a combination of both mechanisms. [13][14][15][16][17][18] Since quenching decreases the observed signal as a function of addition of analyte, these collective mechanisms are commonly referred to as "turn-off" mechanisms.
Another typical approach utilizes absorbance features of an analyte in a resonance energy transfer mechanism (RET). In this mechanism, the excited state of a donor molecule is deactivated by the transfer of energy across space to an acceptor molecule, promoting it to its excited state. Since the mechanism does not necessitate direct contact between donor and acceptor, a sensor developed using RET has the capacity to be more sensitive. If the acceptor is emissive, then the observation is ratiometric with a decrease of donor emission concurrent with an increase in acceptor emission; otherwise, the observation is just a quenching of the donor's emission.
Because many explosive analytes do not have visible absorbance features or react under the same conditions, this is also touted as a selectivity parameter to differentiate nitroaromatics. Many proposals in the literature 3,6-8 use the visible absorbance features of nitroaromatic products to justify their use as a RET acceptor, but since these products lack emissive features the sensing remains a "turn-off" mechanism.
We describe herein the observation of fluorescence emission from TNT and TNB products formed in N,N-dimethylformamide (DMF), without addition of any strong base or nucleophile. The competing equilibria are complicated. TNT is observed to react with OH -(present from the trace water in nominally dry DMF) forming the non-emissive TNTor an emissive sigma adduct with DMF or OH -. TNB also shows the ability to form emissive sigma adducts with either DMF or OH -. In addition, the emission of TNT adducts may give a long desired method to clarify their abundance from that of TNT -. This is the first description of these products as being emissive, with evidence of the formation of products through a reaction with DMF.
This work also demonstrates the need to understand the role of even trace amounts of water in working TNT sensors.  Figure 2a shows the absorbance increase from the addition of NaOH and the decrease from the addition of HCl.

Emission & Excitation Spectra of TNB Products
Emission spectra were collected for both systems before and after each addition of water in the titration. In the case of TNB, the samples were excited at 446 nm in an effort to observe possible emissions from each species. Figure 4 shows the emission spectra for the initial addition of TNB into DMF and also after the addition of 30 μL H 2 O. The initial spectrum has a maximum at 650 nm, which decreases as water is added (see Supplementary Information, Fig. S3), while a new maximum is observed at 586 nm, which increases with the addition of water. In correlation to the absorbance spectra, the initial emission spectrum represents major features of the TNB-DMF adduct, which then gives way to the [TNB-OH]adduct. The previously discussed NaOH and HCl additions support this assignment. Figure   In an effort to understand the exchange between the emissive species in the titration, a separate pair of samples were prepared with identical concentration of nitroaromatic and water as the final point in the titration. These samples were allowed to equilibrate for one hour before obtaining emission and excitation data. Both sets of emission spectra were fit to a modeled set of Gaussian peaks, based on initial parameters from selective excitation of individual products; the TNB and TNT samples show a unique excitation spectrum for emission at 550 nm and evidence of a separate species at 700 nm emission (see Figure 5).  The TNB-DMF emission spectrum was observed using an excitation of 600 nm, as seen in Figure 7a modeled with two Gaussian peaks. These sets of peaks were used as initial parameters in the establishment of the overall model, fitting both species' spectra at the same time. An intensity ratio was held between peaks in the same spectrum, normalizing one intensity to the highest energy feature.

*Normalized in fitting equation
In the assigned [TNB-OH]spectrum, the energy difference between the two modeled peaks is 990 cm -1 , while the difference in the TNB-DMF spectrum is 1000 cm -1 . These energies may represent out of plane ring bending modes, formed by the nucleophilic attack. To verify that these modeling parameters are valid, the TNB fitting model was applied to emission data with varying excitation wavelengths.    Figure S4). Using these modeled fits, the area under each species' spectrum was calculated at each titration point.

Excitation and Emission Spectra of TNT Products
For TNT, the emission spectra of the titration do not show a great deal of shape change (see Supplementary Information S5). Figure 4 shows the normalized emission of the initial TNT solution and after the addition of 30 μL water, excited at 446 nm.
These two spectra appear to be superimposable, however the excitation data in Figure   5 refutes the existence of only one species. In similar fashion to the TNB data, two distinctly different excitation spectra are acquired using 550 nm emission and 700 nm emission. As previously mentioned, the methyl group on TNT gives additional reactions that TNB cannot exhibit, so identification of emissive species requires caution. Since the emission intensity increases ( Supplementary Information Fig. S5) while the absorbance of TNTis decreasing, TNTcannot be the emissive species.
Also, the established TNTabsorbance spectrum 9,10 has absorbance at lower energy beyond 700 nm; this rules out TNTemitting at higher energy wavelengths. These emission spectra show similarity to the products that are formed from TNB, which does not have the ability to be deprotonated. In addition, Figure  To further understand the TNT reactions, the emission data were also modeled using the same approach as described above. Figure 6b shows the emission spectrum resultant from 370 nm excitation, where only [TNT-OH]is excited. The outcome was a three peak spectrum with slight differences from the [TNB-OH]model, which was not able to reasonably fit all spectra in the titration by itself even when allowing for relaxation of initial parameters. TNT-DMF emission peaks were initially fit using a 600 nm excitation (see Figure 7b), and subsequently iterated with the [TNT-OH]parameters to establish an overall model. It became obvious that there was an additional emitting feature for several reasons. First, the emission maximum appears to shift to higher energy as the excitation shifts to higher energy (see Figure 8b), which was not observed for TNB. Also, the initial fits for [TNT-OH]and TNT-DMF were unable to account for all of the observed emission data collected between samples excited at varying wavelengths, even when the parameters were allowed to change dramatically. The addition of one extra Gaussian peak at a higher energy maximum than [TNT-OH]was able to globally satisfy all data without drastically altering the initial [TNT-OH]and TNT-DMF fits. The complete equation for the TNT model can be found in the Supplementary Information.
The existence of a third species seems to go against the observation of tight isosbestic points, which typically indicate only two absorbing species in solution.
TNTand [TNT-OH]are well established and clearly evident here. TNT-DMF has been observed through fluorescence, and supported through analogous chemistry to the formation of TNB-DMF. As a third contributing species, we cannot ascertain the magnitude of TNT-DMF's molar absorptivity nor its quantum yield. The addition of another [TNT-OH]species appears to cloud this even further; however, there are systems that may exhibit isosbestic points with three contributing species. Such a situation can arise in a system with absorbance contributions from a Lewis acid with two binding sites and one Lewis base, where the free acid and each form of the adduct absorb in the same range with different molar extinction coefficients. 21 In our case, TNT acts as the Lewis acid, which has two binding sites (C 1 Fig. S6). Absorbance computations at the B3LYP/6-311+G**//HF/6-311+G** level suggest that the HOMO/LUMO gap for the C 1 sigma adduct is of lower energy than the C 3 form. If each hydroxide sigma adduct has a similar Stokes shift, this would assign the higher energy, single peak emission spectrum to [TNT-OH] -C 3 . Figure 11 shows the initial emission spectrum of the TNT/DMF solution, the addition of 15 μL of water, and addition of 30 μL of water. All spectra in this titration fit well using our established model (full titration fitting can be found in Supplementary Information Fig. S7). The converged parameters for the TNT product emission spectra model can be found in Table 1

Conclusions
In summary, we have shown that TNB has the capacity to form two sigma adducts in DMF, [TNB-OH]and TNB-DMF, as shown in Scheme 1. Absorbance data for TNB/DMF solutions supports the existence of two species that absorb in the region expected for sigma adducts. Further investigation into excitation and emission data supports the two species. The development of a multi-Gaussian model (with parameters given in Table 1) clarifies the relative quantities of emission from each species, showing the two to be linked by equilibrium. The methyl group of TNT causes different reactivity than TNB. TNT has the ability to form four products in DMF (Scheme 1), leading to spectral differences and similarities which require careful analysis to discern. The major absorbing species, TNT -, is shown to be non-emissive.
Absorbance features consistent with sigma adducts of TNT are observed to grow as a function of water addition. The three possible sigma adducts, [TNT-OH] -C 1 , [TNT-OH] -C 3 , and TNT-DMF, all exhibit excitation and emission features that have distinct similarity to their TNB counterparts. A multi-Gaussian model was also established to differentiate the TNT sigma adducts' relative emission ( Table 1), showing that all grow together as a function of added water. This is not observed for the TNB sigma adducts, and implies different mechanisms for DMF sigma adduct formation. This is the first report of fluorescent emission from Meisenheimer adducts of TNB and TNT, with both nitroaromatics capable of nucleophilic attack by hydroxide and the less sterically hindered TNB capable of attack by DMF. All reactions show significant changes within minutes, which allow these reactions to be further explored in the context of sensor development. The emissions may be able to be enhanced by using a resonance energy transfer from a donor fluorophore with a long excited state lifetime to these adducts as acceptors; this could further amplify this form of a "turnon" mechanism. Alternatively, a solid state sensor capable of capturing these nitroaromatics through stronger nucleophilic attack may be able to serve in a facile turn-on approach. Future work should compare dinitrotoluene (DNT) emission characteristics to those described here. DNTs are common impurities in TNT, but since DNTs have a much higher vapor pressure than TNT they will each deliver similar amounts in the vapor phase to a sensor. Depending on the relative equilibrium constants for each adduct formation and their relative quantum yield, it may be found that DNTs are more observable in this capacity. In addition, the different reactivity between TNT and TNB means that caution must be used when using TNB as a surrogate for TNT in the development and testing of TNT sensors. Finally, this work shows that TNT sensors are expected to perform differently under wet or humid conditions than under dry conditions. Water can react with nitroaromatics even at trace levels.

Supporting Information.
UV/Vis Absorbance spectra, excitation and emission spectra of trinitroaromatics in DMF with water titration. This material is available free of charge via the Internet at http://pubs.acs.org.  : Absorbance spectra evolution from TNB titrated with water in DMF. Solution titrated directly after addition of TNB to DMF. Figure S2: Absorbance spectra evolution from TNT titrated with water in DMF. Solution titrated directly after addition of TNT to DMF. Figure S3: Emission spectra evolution from TNB titrated with water in DMF (λ ex = 446nm). Solution titrated directly after addition of TNB to DMF.

TNB Model:
The overall equation governing our TNB model (intensity as a function of wavelength) is given by: where and are the intensities of the [TNB-OH]and TNB-DMF spectra respectively, is the fixed ratio of intensity for the peaks that comprise the [TNB-OH]spectrum, is the fixed ratio of intensity for the peaks that comprise the TNB-DMF spectrum, represents the peak maximum for the highest energy feature of [TNB-OH] -, represents the peak maximum for the highest energy feature of TNB-DMF, represents the bandwidth of each peak, and represents the vibrational energy gap in the ground electronic state. Figure S4: Emission spectra evolution from TNB titrated with water in DMF (λex = 446nm), fit using static model established through spectral peak fitting. Figure S5: Emission spectra evolution from TNT titrated with water in DMF (λex = 446nm). Solution titrated directly after addition of TNT to DMF.

TNT Model:
The overall equation governing our TNT model is given by: where , , and are the intensities of the [TNT-OH] -C 1 , TNT-DMF, and the [TNT-OH] -C 3 spectrum respectively. All other parameters are as described previously for the TNB model. Figure S6: UV/Vis Computations performed at the B3LYP/6-311+G**//HF/6-311+G** level for the two hydroxide adducts of TNT, in vacuum and at 273.15K. Note that while the computed maxima are generally poor for this approach, the relative position of the maxima are typically representative of experimental observation. Figure S7: Emission spectra evolution from TNT titrated with water in DMF (λex = 446nm), fit using static model established through spectral peak fitting.

INTRODUCTION
The detection of trinitrotoluene (TNT) has become an increasingly prevalent research topic over the past decade, due in part to the rise in international terrorism and the toxic nature of TNT in the environment. Many methods of detection have been proposed in literature, including methods that rely on reactivity from TNT's electron deficient aromatic ring. Since these reactions generate species that absorb in the visible range, simple colorimetric sensors have been developed in the solution phase. [1][2][3][4] More complex and sensitive fluorometric sensors have been developed, [5][6][7][8] with many using the colorful species in a resonance energy transfer (RET) sensing mechanism. [9][10][11][12] In such a sensor, a donor fluorophore is chosen based on significant overlap of its emission spectrum with the absorbance spectrum of the TNT product acceptor. In this fashion, a sensor can be tailored for a species that absorbs in a specific range.
Alkoxide reactions with TNT have been studied for a number of years, with two major competing reactions resulting. Scheme 1 shows the possible reactions between TNT and hydroxide or a generic Lewis base. The literature reports the ability of hydroxide or alkoxide to deprotonate the methyl group of TNT, forming TNT − , or to attack the ring forming the sigma adduct TNT-OR − . [13][14][15][16][17][18][19] While alkoxides are believed to attack TNT's C 3 carbon, hydroxide is much less sterically hindered and is able to form the TNT-OH − C 1 adduct, pushing TNT's methyl group out of plane.
While there are other proposals for TNT mechanisms, these are the possible initial elementary steps for any further reactions; therefore, these are the most crucial to understand in the context of a rapid TNT sensor. Due to this, further reactions such the C 1 or C 3 carbon forming TNT-OH − or TNT-B, or can have electron density donated to its electron deficient ring forming a CT complex, TNT:B. as the "Janovsky complex" 14 (TNT − attacking TNT) will not be discussed in this study.
Given the multitude of TNT sensor proposals in literature that depend on TNT's interaction or reaction with amines, it is perhaps surprising that a thorough mechanistic study has not been undertaken previously. These types of sensors have amines, but their reactions were performed over a 24 hour period. 3 The reactions are commonly done in solvents that may have competing reactions with TNT, as well as solvents which are not easily dried of trace water. Therefore it is possible that in this early stage of development, the reactivity may not be coming directly from the amines, but rather from alternative sources like OH − generated through acid/base equilibria.
Additionally, the observed absorbance spectra are commonly misattributed to the incorrect major absorbing species. There is definite confusion as to the actual product(s) being formed in solution, as numerous accounts of the same or similar reactions lead to a different proposed product; for example, the same absorbance spectrum has been explained by a charge transfer complex, 28 TNT − , 21 or a sigma adduct with an amine. 27 The overall goal of this study is to clarify the product formation of the reaction of TNT with amines, and address the relative rates of the TNT reactions with hydroxide against those of amines. In addition, we set out to clarify the timescale in which amine functionalized TNT sensors may or may not be applicable.

MATERIALS AND METHODS
TNT was obtained from Drs. Jimmie Oxley and James Smith and was used without further purification. NMR experiments were conducted in either d 8  for dilution was purchased from Acros. TNT solutions were quantitatively prepared and 1 H and 13 C spectra were acquired before and after the addition of base on a Bruker 300 MHz NMR. Time delay from the start of reaction to data acquisition was approximately five minutes per spectrum, due to solvent locking and shimming.
Solutions containing TNT for absorbance measurements were made quantitatively using either HPLC grade THF (99.9%, inhibitor-free) or HPLC grade MeOH (99.9%), both purchased from Sigma. TNT stock solutions were added to a cuvette and observed prior to addition of amine. While stirring, amine solutions were quantitatively added using a micropipette. To rapidly gather absorbance data, a setup was constructed in house with a Peltier temperature controlled cuvette holder held at 20.00°C and a white light tungsten lamp directed by OceanOptics© fiber optics upon the sample. Our setup was capable of acquiring one spectrum about every 50 msec, limiting our temporal resolution to that value.
In order to address the kinetic evaluation in a reasonable fashion, a pseudo-first order kinetic approach was used. Experiments were prepared so that one reactant concentration would be 100x excess of the other, so that the loss of the excess reactant was negligible, simplifying the calculations. Data were fit to exponential trends using commercially available graphing software (Microsoft Excel, SigmaPlot).

RESULTS AND DISCUSSION
To elucidate the reaction between TNT and aliphatic or aromatic amines, our group utilized a similar approach to Fyfe et. al. by exploring relationships between 1 H NMR and absorbance spectra. In order to be consistent with what is known about hydroxide's reaction with TNT, we were able to perform our experiments first with hydroxide and subsequently with a group of selected amines of various basicity, nucleophilicity, and sterics. The aliphatic amines explored in this study included diethylamine (DEA), triethylamine (TEA), ethylamine (EA), propylamine (PA), and n-butylamine (BA). The aromatic amines explored in this study included 2,6-lutidine, collidine, pyridine, and aniline. Color changes were not observed with the addition of reasonable amounts of TNT and aromatic amines. Visible range absorbances were detected only when tens of milligrams of TNT were added to these as solvents; because of this, the further study of TNT reactions with aromatic amines was discarded and only the aliphatic amines were used. In our attempt to investigate the formation of TNT products in differing environments, methanol and tetrahydrofuran (THF) were used as solvents to assess polar and slightly polar surroundings respectively.

Product identification through NMR:
1 H NMR was employed to determine the possible contributors to visible range absorbance. While we recognize the time discrepancy between our ability to acquire 1 H NMR data (on the order of 5 minutes) and our ability to monitor visible absorbance changes (on the order of tens of milliseconds), we were able to correlate our spectral data to faster data acquisitions from the literature. 13 Since we anticipate ambient water to be present in even the most nominally dry solvents, our first investigation was the reaction of TNT with OH − .

Hydroxide/methoxide as competing species:
In order to understand the relation of TNT/amine products in systems with any trace water, reactions of TNT and sodium deuteroxide in deuterium oxide solution (NaOD/D 2 O) were carried out in deuterated THF and deuterated methanol. Figure 1 represents the reaction with increasing concentration of NaOD/D 2 O in d 8 -THF, while    H/D exchange on TNT in basic deuterated media has been observed in the literature, 30 and is indicative of the formation of TNT − being a reversible process.
The other species which appears to be growing in at a slower rate and as a function of increased NaOD concentration has a singlet at 8.41 ppm and another at 5.30 ppm. Their relative integration ratio is 3:2 which is consistent with the C 1 deuteroxide adduct formation, where the lower field feature represents the two aromatic protons (on C 3 and C 5 ) and the higher field feature represents the intact but shifted methyl group protons. This differs from Fyfe's observation of a 1:1 ratio for the C 3 adduct of methoxide, where two singlets were detected at 8.45 ppm and 6.18 ppm in 87.5% DMSO/12.5% MeOH. It is also important to note that the products only make up a small percentage of the total integration at equimolar TNT and NaOD concentration; this suggests that the equilibria lie much further towards reactants in MeOH than in THF. While there are unfamiliar features developed from the addition of excess DCl at the end of this titration, the implication of this product does not appear to impact our further absorbance investigation as we will discuss later in this paper. Also, TNT was exposed to DCl as a control to this experiment, which showed no change to the original TNT features. After the addition of DCl to the system, reaching a rapid equilibrium state, a 13 C DEPT 135 experiment was undertaken. The data ( Supplementary Information, Fig. S4) also support the full deuteration of TNT − , creating a CD 3 group; in this case, we observe the features to be negative whereas the features were not observable in d 8 -THF. The CD 3 group is expected to show as a septet, which may be what we see as the negative features at 15 ppm with very low S/N. This may be additionally hindered by the removal of the pre-existing protons, which enhance that carbon's signal. No other features besides unreacted TNT are apparent from the carbon spectrum, implying that TNT − is the major early product of this reaction.

Reaction of TNT with amines:
With an understanding of hydroxide's role in TNT reactivity, we subsequently investigated the formation of products of TNT with various amines using 1 H NMR, as seen in Figure 3. In a qualitative assessment of reaction products, it is first apparent that TNT − features exist in the reaction of TNT with EA or PA while similar features are not apparent for TEA, DEA and BA. This observation does not correlate well with the expected basicity of each amine in water; however, it does correlate well to their expected water content. EA is a gas at ambient temperatures, and thus was only available as an aqueous solution. As a larger quantity of PA was necessary for product observation, a small water impurity may exist in this system as well. Since the addition of water directly generates hydroxide, it is difficult to clarify which species seen here are resultant from hydroxide or amine reactivity. Other features are seen in these spectra, which may relate to amine adducts, but are barely observable over the baseline noise and are a very small contributor even with excess amine -this suggests that the equilibrium constant for these reactions is low, with mainly unreacted TNT being observed at equilibrium.
It is interesting to note here the possible appearance of two sigma adducts in the reaction of TNT with DEA, as Figure 3c shows two quartets in the spectrum for this reaction. The quartets may come from the inequivalence of the alkyl groups on DEA after covalently bonding to TNT, but may also come from two different attack sites. As the size of the nucleophile increases and sterics begin to play a larger role, it is conceivable that DEA can attack either the C 1 carbon or the C 3 carbon with similar energies. Computational analysis supports this proposal, as the relative energies of formation for the two sigma adducts shows the C 3 adduct to be slightly more favorable (vide infra). (not shown) are the predominant features with approximately 500 times the intensity of the products. Clearly, the equilibria for these reactions lie much closer to reactants than products or the kinetics for these reactions are very slow.

Absorbance Characteristics/Kinetic Evaluation
Although absorbance spectroscopy is not a clear identification method for product structure, it can be a powerful tool when coupled with a method like NMR.
Here, we set out to observe similarities in absorbance measurements to propose that there is no significant visible absorbance from TNT-amine adducts. In addition, we put to question whether amines are basic enough to deprotonate the methyl group on Under pseudo-first order conditions, the concentration of each product can be monitored as a function of time using a single exponential rise function: where "a" and "k" are functions of intrinsic rate constants as well as initial concentrations of reactants. Substituting in absorbance for concentration using Beer's Law yields: Since water content clearly plays a role in the TNT − equilibrium if OH − is the reactant, its concentration must be accounted for in our experiments. We have previously described the equilibria linking water concentration to these species in DMF solution. 33  Having good evidence for the products formed by TNT and hydroxide, we conducted experiments in which we observed the visible absorbance spectrum for five minutes after the reactions were started. In Figure 5, we observe the reaction using excess NaOH in THF with the formation of two products: one which grows in immediately and the other which equilibrates much more slowly. However, the second product is clearly the major absorbing species even before equilibrium is established. The early literature confirms the species which give rise to these absorbance features; the early product here has the shape of an anionic adduct, TNT-OH − , while the second spectrum belongs to TNT − . 13,14 This is consistent with our 1 H NMR assessment; due to sample preparation time, instrumental setup and data acquisition, TNT-OH − had already disappeared and TNT − was the only observed species. Figure 6 shows the reaction of 100x excess TNT with NaOH in THF.
Similar spectral shape to the excess NaOH experiments is seen here, with possible TNT-OH − formation early and a slower formation of TNT − . A control experiment was conducted to determine the effect of a possible scattering component of NaOH in THF on our kinetic fits. We observed a reasonably significant baseline increase with the addition of NaOH to THF, which dissipated within a few seconds. This lead us to omit some initial data when the absorbance signal was on a similar order of magnitude as the scattering. shows absorbance data from 520 nm and 700 nm fit to an exponential rise function, where the observed rate constant is the exponential variable; in this case, the data at 520 nm required two exponential rise functions to fit properly.
The same experimental approach was also taken in MeOH. Figure 7 indicates a different observation as a function of time. TNT − is the predominant absorbing species early on, which then gives way to absorbances at lower wavelengths from TNT-OH − . This is consistent with the order of appearance of these species in our NMR. While we are able to fit our entire five minute data set to a phenomenological set of exponential rise and decay functions (see Supplementary Information Figs. S11-13), we are not able to identify the secondary products nor do they appear to absorb strongly in the visible range. In addition, there is no analytic solution for the sequential equilibria that we have phenomenologically modeled. Due to this and the stronger importance of the initial processes on TNT detection, we continued by modeling the early absorbance evolution in MeOH.
The inset on Figure 7 shows a relatively close relationship between the observed rate constants at 520 nm and 700 nm, indicating the observation of the same species. However, it is interesting to note the late appearance of pseudo-isosbestic points, as seen in Figure    The early absorbance growths fit well to a set of exponential growths, as anticipated in our derivation and indicated on the insets to the absorbance evolution figures. The observed rate constants can be found collectively in Table 1.
[  Table 2. Intrinsic rate constants for the formation of TNT − calculated using kinetic model and absorbance data at 520 nm and 700 nm.
Upon calculation of the intrinsic rate constant for the TNT − reaction (in Table   2), we observe a major discrepancy between the excess NaOH and excess TNT experiments. The intrinsic rate constants should be the same for each condition, but they differ by at least an order of magnitude. We have shown above the relationship

Amine reactions in MeOH:
Solutions of TNT ranging from 10 -2 to 10 -1 M were prepared, as well as diluted amine solutions in MeOH. Using the same techniques described above, the amine solutions were quantitatively added to the TNT solutions in a stirring cuvette while absorbance measurements were taken. Absorbance spectra evolved very rapidly and equilibrated within several seconds, as in the case of OH − addition. Figure 9 shows the absorbance evolution of the reaction of excess TNT with EA, DEA, and TEA in MeOH. Insets on Figure 9 indicate the observed rate constants for the reaction, which are equivalent to . The observed rate constants at 520 nm and 700 nm are similar for each individual experiment, as seen in the hydroxide reactions. They are not similar, however, for the reactions with different amines; the EA and DEA experiments agree within random error, but the TEA reaction shows a much larger observed rate constant. The shape of each set of spectra are strikingly similar, which is not expected for these three amines. TEA is too sterically hindered to form a sigma complex with TNT, while EA and DEA are expected to form sigma complexes to different degrees.
A normalized set of spectra from the initial equilibrium are found in Figure 10a. All spectra were normalized to their shared amine reaction absorbance maximum at 450 nm. The amine reaction spectra overlap well from about 420 to 460 nm, implying that only one species absorbs in this range, while the OH − reaction spectrum shows slightly different shape. It is likely that TNT-OH − is the major absorbing species formed in the OH − reaction, and that the amines are deprotonating MeOH leading to the formation of TNT-OCH 3 − . This would account for the similar spectral shape, since both sigma adducts are expected to have similar MO energies. Towards lower energy, the spectra diverge but retain similar spectral shape consistent with the established spectrum of TNT − . Difference spectra of the amine reaction spectra subtracted from the OH − reaction spectra are found in Figure 10b  Difference spectra calculated using normalized data from (a). Since the addition of hydroxide is observed to generate the most TNT − , the amine reaction spectra were subtracted from it. The residual spectra show the same shape, consistent with the only other significantly absorbing product, TNT − .  Table 3. Calculated rate parameters for reactions of excess TNT + amines in MeOH.

Kinetic Model Derivation:
For the competing equilibrium reactions between TNT and hydroxide (or methoxide): We cannot assume one equilibrium is much faster than the other, because we observe both on a similar timescale. Therefore, they must be solved simultaneously.
The mass balance and charge balance conditions require: The latter two equations can be combined to show: The differential equations are: Derivation can proceed under either pseudo-first order condition, as long as water concentration remains in excess over TNT − concentration.
Letting , , and , the differential equations become: From equation [5],       vapor. [9][10][11][12][13] The competing reactions of TNT forming TNT − or a sigma complex, TNT-B, serve to complicate the design of these sensors, especially those that rely on resonance energy transfer (RET). 11,12,14,15 RET-based TNT sensors typically employ a strong fluorophore, whose emission is quenched as it donates energy from its excited state to the excited state of the TNT product acceptor. The choice of donor fluorophore is inherently dependent on the absorbance features and subsequent molecular orbital energies of the TNT product, which is not always identified properly in contemporary literature. In some cases, these absorbance features have even been attributed to the formation of a charge-transfer (CT) complex. 10,[16][17][18] Further, the rate of formation of these products is also an important feature, as environmental analysis has a much lower demand for rapid detection than the detection of landmines or improvised explosive devices (IEDs). Our group and others have investigated the relative kinetics of the formation of each of these TNT products, [19][20][21] which has proven to be intimately tied to solvation effects. In many cases, one product will act only as a transient species, yielding the dominance of the visible absorbance range to the second product rapidly. With a change in solvation, these roles can be observed to reverse, causing the other product to appear as the transient species. This indicates the importance of considering the effect of solvation on TNT sensor development.
Finally, the relative thermodynamics of each competing reaction are also solvent dependent, and secondary reactions are observed to occur removing TNT − or TNT-B from detection.
All of the discussion up to this point is made more convoluted by the proposed reaction of aliphatic amines with TNT to form TNT −13,18,22,23 or zwitterionic sigma adducts. 6,11,12,14,15,24,25 With lower basicity, we question herein the ability of aliphatic amines to deprotonate TNT significantly, as well as their ability to form a thermodynamically stable sigma complex relative to that of alkoxide bases. For many of these reasons, experimental approaches have proven to be limited in their ability to assess the actual chemistry that is being observed. Several groups have computationally addressed similar systems/reactions previously. However, some restricted their structural optimization computations to molecular mechanics forcefields, 26 semi-empirical methods, 27 and DFT methods. 28,29 No computational studies have been performed that compare alkoxide and amine reactivity through TNT − formation, sigma adduct formation, and CT complex formation using the same method, let alone a high level ab initio method and solvation modeling. By computationally modeling the reactants and products of these proposed reactions, we are able to calculate the relative thermodynamics of each reaction and observe a significant trend in solvation. Our results may also shed some light on the role of water in functional solid/vapor sensors, as the amines used in these sensors may only be deprotonating water to form hydroxide, which then gives rise to its related TNT products.

COMPUTATIONAL METHODS
In order to gain further insight into the proposed reactions, computational modeling was employed using the commercially available Spartan '10 software. 30 Reactants and products of TNT reactions with OH − , CH 3  Converged equilibrium geometries of TNT and possible acid/base and sigma attack products. Note that we were able to converge a sigma adduct for all bases except for TEA, which is expected to be too sterically hindered to attack the ring covalently. CT complexes were omitted from this figure due to the number of converged structures. solvation dependency. In these solvation computations, the converged structures from the HF outputs were used to perform single point energy calculations at the B3LYP/6-31G* level; this approach was chosen since it was used as a fitting method for the development of the SM8 model. To cover a wide range of dielectric constants, the following solvent models were used: vacuum (ε = 0), hexane (ε = 2.02), tetrahydrofuran (ε = 7.52), 1-propanol (ε = 20.1), methanol (ε = 33), formic acid (ε = 58), water (ε = 80), and formamide (ε = 109). All computations were performed using an input temperature of 298.15 K and a pressure of 1 atm.

STRUCTURAL ASSESSMENT
The set of converged structures for the acid/base and sigma attack reactions can be found in Figure 1. Although we investigated the formation of a multitude of CT complexes, they are omitted here for simplicity. We have also omitted the converged structure of the isolated protonated amines, which were used in the deprotonation reaction calculations to form free ions. The ion pairs are shown, and converged so that the positive hydrogens on the alkylammonium cation interact directly with the oxygens from the nitro group, where the greatest electron density from the TNT anion is located. This distorts the methylene group much further out of the plane of the ring than calculated for the isolated TNT anion. As expected, we were able to converge sigma adducts for all of the amines with the exception of TEA.
These optimizations pushed TEA away from the ring, and stabilized in an electrostatic interaction or CT complex type orientation. Table 1 shows some related structural parameters from each species' optimized geometry. As expected, the nucleophiles show a general trend of increasing bond length as their nucleophilic strength decreases and their sterics increase. The largest bond angle distortions from that of TNT are seen for the TNT-OH − and TNT-OCH 3 −

Species Bond Lengths (Å) Bond Angles (°) Dihedral Angles (°)
Nu -C  Table 1. Structural parameters from TNT products optimized at the HF/6-311+G** level. Note that "Nu" refers to the bonding site of the nucleophile.
adducts, stretching the ring in plane. These adducts also distort the calculated dihedral angles to the greatest degree of all of the nucleophiles, bending the ring out of plane.
The distortions may play a large contributing factor to the destabilization of the frontier MOs, whose density surfaces are localized on the ring. Since the amine adducts do not distort the ring as much as OH − or OCH 3 − , their MOs have closer relative energies to those of TNT. The amine adducts show a definite trend of increasing sterics resulting in increased distortion. TNT − formation distorts the bond angles and dihedrals the least, which is expected due to the enhanced resonance stabilizing the ring. The C 1 adducts appear to distort the ring less than their C 3 counterparts, which may explain their slightly greater MO stability. In addition, the C 1 adducts push the electron donating methyl group out of plane of the ring, replacing a portion of its electron donation with that of the nucleophile. Although the C 3 adducts are not displacing an electron donating group, their nucleophiles are significantly more orthogonal to the ring, which negatively affects their ability to donate electron density. This may be another stabilizing advantage of the C 1 adduct over the C 3 adduct.  Since the TNT/hydroxide reactions show much greater thermodynamic favorability and more significant experimental signal, NMR computations were done using those products. Table 2     Also, note that the ΔE rxn data for the methoxide adducts overlap such that the data points cannot be discerned.     ). Note that while the computed maxima are generally poor for this approach, the relative position of the maxima are typically representative of experimental observation.

CONCLUSIONS
High level ab initio computations were performed to assess the relative thermodynamics of TNT product formation, from reaction with OH − , CH 3  The following is in preparation to be submitted to Analyst, and is presented here in manuscript format.
another order of magnitude using metal enhanced fluorescence. 20,21 Selectivity, although difficult to achieve using a single interactive sensor, can be induced by the use of a sensor array where multiple sensors can be exposed to the same analyte stream and their responses observed together. This "fingerprint" can be read as a function of a specific analyte; whereas one of the sensors may not be able to discern one analyte from another, it is unlikely that two analytes will interact with the multitude of sensors in the exact same fashion.
Perhaps surprisingly, the rhodamine family has not been well explored in the field of explosive detection. Meaney & coworkers investigated the fluorescence quenching of common fluorophores including Rhodamine 6G against nitrobenzene, 4nitrotoluene and 2,6-dinitrotoluene, but did not explore more contemporarily relevant analytes. 9 Rhodamines have notoriety as being very strong fluorophores, and were therefore selected to test as fluorescence reporters against explosive analytes.
Rhodamines also have the advantage of possessing visible range excitations and emissions that are not subject to interference by absorbance bands of explosive analytes. The rhodamines investigated in this study were Rhodamine 560 Chloride (Rh560), Rhodamine 6G (Rh6G), Sulforhodamine B (SRhB), Rhodamine 640 Perchlorate (Rh640), Sulforhodamine 640 (SRh640), Rhodamine 700 (Rh700) and Rhodamine 800 (Rh800). Fluorescein 548 (Fl548) was also included, as another representative xanthene derivative with a high quantum yield. The structures of the selected fluorophores can be found in Supplementary Information Fig. S1. The  shows the absorbances characterizing the initial fluorophore solutions, spanning the visible range. In similar fashion, Figure 2 shows the normalized emission spectra from each initial fluorophore solution, spanning from 530 nm to 800 nm. Pairing absorbance and emission measurements together allows for determination of possible interactions between fluorophore and analyte, as both ground state and excited state profiles are examined.   The DNT isomers that were explored show a range of emission quenching and enhancement, but with fingerprint patterns that differentiate them from the other analytes. DNTs have a similar capacity to TNT to be deprotonated to a varying degree, and the ability to form a sigma complex with hydroxide from ambient water in solution. In some of our additions, absorbance features arise to support this assessment. For example, Fig. S5 shows an increase in absorbance for the addition of 2,4-DNT to SRh640 centered around 650 nm; this absorbance spectrum has been shown previously to be resultant from the deprotonation of 2,4-DNT. 32 Likewise, Fig. S4 shows an increase in absorbance for the addition of 3,4-DNT to Fl548 between 400 and 500 nm; this would be consistent with the formation of a hydroxide adduct.

Chemicals and materials
While these absorbance features contribute to an inner filtering effect, they are minimal compared to those of the TNB or TNT products, which will be described later. In addition, there are observable emission changes for DNT additions that either do not show these absorbance growths or are not fully resultant from an inner filter effect. 3,4-DNT causes a larger decrease in the emission of Fl548 than its product's absorbance growth (Fig. S4). Another example can be found in Fig. S6,  Once the maximum emissions were corrected using the above equation, the % change was recalculated. We estimate the relative error in the corrected emission change to be between 5-10%. The IFE corrected changes stand in stark contrast to the observed changes, indicating a hidden component of increased emission in the process. For example, the addition of TNT to Rh640 shows an observed quench of 44%. However, given the amount of TNT products that absorb in the visible range, IFE corrections should be applied. If no true collisional or static quenching occur, these corrections should bring the change back to 0%. Instead, the corrected change is calculated to be 104%, indicating a competing interaction that enhances the emission of the fluorophore. A similar effect can be seen for other rhodamine/trinitroaromatic pairings in Table 1, in which the corrected % changes are all more positive than the observed. The exceptions to this are the additions of TNT to Fl548 as well as the additions to Rh700 and Rh800, which do not change significantly within the estimated error. Fl548 especially shows significant quenching with TNT and TNB, which does not appear to be competing with an enhancing phenomenon.
Although we have previously described the emission of TNB and TNT products in DMF, 29 there is little shape change in the emission spectra to justify their direct additive role in the calculated increase in emission. TNB reacts with hydroxide from trace water to form the sigma complex TNB-OH − , which shows a narrow absorbance feature around 450 nm and a broader absorbance feature around 520 nm.
Additional absorbance bands are expected from the TNB-DMF sigma adduct around 650 nm, which we see in varying amounts. TNT is capable of similar chemistry, by forming two possible TNT-OH − adduct isomers, as well as a DMF adduct; both have been shown to be emissive. TNT is also capable of being deprotonated by OH − , forming TNT − , but this species is not emissive.
One plausible conclusion is that the increase results from resonance energy transfer (RET). This would cause the excited states of a trinitroaromatic product to donate energy to the fluorophore acceptor, populating their excited states leading to greater emission. Additionally, it appears that the addition of TNT leads to a greater enhancement than the TNB counterparts. While we cannot address RET efficiencies here without knowing the relative amount of each species and lifetimes, it is interesting to note the consistently higher response to TNT. To further investigate the possibility of RET between trinitroaromatic products and the fluorophores, the absorbance spectra of the fluorophore acceptors is compared to the emission spectrum of TNB product donors in DMF solution in Figure 4. The emission spectrum is comprised of two components, TNB-OH − and TNB-DMF, which emit at a maximum around 580 nm and 650 nm respectively. 29 The emission spectrum overlaps best with the absorbance spectra of Rh700 and Rh800, which do not exhibit enhancement when exposed to TNB in DMF. Also, the similar absorbance spectra of Rh640 and SRh640 but different response to TNB do not make sense in the context of RET shown here. In a RET mechanism, it would be expected that both Rh640 and SRh640 would show enhancements, which are not observed. In addition, the concentrations of each species in solution are on the order of 10 -3 M at best, which would make the average distance between molecules greater than 100 Å. Given the strong distance dependence in a RET mechanism, this distance is higher than the typical Förster distance exhibited by similar molecules. 19 The poor correlation of the spectral bands of the fluorophores and TNB products and the large intramolecular distances suggest that RET is not the enhancement mechanism.  Photo-induced electron transfer (PET) is unlikely to play a role in our observations in the context of these calculations. Typically, PET causes a quench in emission if electron density is transferred to or away from a fluorophore, as this can lead to structural and lifetime changes. 19 Emission enhancements can be justified through disruption of PET mechanisms, but since our initial solution is comprised of only the fluorophore in DMF this also seems unlikely. The only way that this might be conceived is if DMF is initially quenching the fluorophores, and that the displacement of DMF by explosive molecules in solvation leads to an observed enhancement. Given the relatively low quantity of analyte added relative to the overwhelming amount of DMF solvent, the small addition cannot account for the magnitude of enhancement observed.
In the context of a ground state ET mechanism, TNB and TNT have the lowest calculated HOMO energy of all of the modeled species. This feature as well as their electron deficient character makes it unlikely that the free trinitroaromatics would donate electrons to the fluorophores. The anionic products have a much higher calculated HOMO energy however, making them more likely electron donors. Fl548 and SRh640 have LUMOs that are slightly higher in energy than the HOMOs of TNT − and the OH − adducts, making them poor candidates as electron acceptors. The calculated LUMO of SRhB is lower but still close in energy to the HOMO of the anionic products (within 0.3 eV); this is likely an insufficient energy gap to cause significant ET at ambient temperatures. Rh6G and Rh640 have LUMOs of very similar energy (-5.39 and -5.13 eV respectively), which are about 2.4 eV lower than the HOMO of TNB-OH − . Since these two fluorophores show the greatest enhancement with TNB and TNT addition in solution and their relative LUMO energies to TNB-OH − are strikingly similar, electron transfer appears to be the most probable mechanism. The difference in energy is substantial enough to occur without significant reversibility. Rh560, Rh700, and Rh800 have LUMOs of lower energy than Rh6G and Rh640. Rh560 shows a calculated enhancement that is within the estimated error of our approach, and with a LUMO that is 3 eV lower than the HOMO of TNB-OH − the magnitude of the enhancement might be expected to be higher. The correlation of the relative enhancements is beyond the scope of our calculations here, however Rh560 is likely to follow the trend of enhancement shown by Rh6G and Rh640. Rh700 and Rh800 have a difference of about 2.8 eV, but an electron transfer is hindered by the lack of OH − adduct formation in the acidified solution. In these cases, only slight changes in emission are observed after IFE calculations. It is also interesting to note that the three enhanced fluorophores (Rh560, Rh6G, Rh640) happen to all be cationic, whereas the remaining fluorophores (excluding Rh700 and Rh800 for aforementioned reasons) are either neutral in charge or anionic. It is easy to conceive the possible formation of an ion pair between the cationic fluorophores and the anionic trinitroaromatic products. This driving force would bring the species in close proximity to allow ET to occur, without the limitations of diffusion and collision in solution.

Conclusions
In summary, we have investigated a sensing array based on highly fluorescent molecules in DMF solution that shows promise for the detection of explosive analytes.
In addition to typical collisional or static quenching, inner filter effect corrections indicate an ET component, which increases the emission of at least two of the fluorescent reporters (Rh6G and Rh640). The findings of ET may have an impact on pre-existing RET based sensors, as the emission of the fluorophores in this study was observed to enhance. The capability of trinitroaromatic products to be involved in ET may influence existing sensors that rely on TNB or TNT as a RET disruptor, leading to a competitive ET mechanism that would decrease the effectiveness of the sensor by diminishing an observed quench in emission. Further studies will be undertaken to expand the set of fluorescent reporter molecules and quantify the interactions with the most significant emission change, as well as experiments to give additional clarification of the ET with trinitroaromatic products.