Cu-BTC functional microdevices as smart tools for capture and preconcentration of nerve agents.

Cu-based Metal Organic Frameworks (MOF) microdevices are applied in sampling and preconcentration of nerve agents (NAs) diluted in gaseous streams. An in-situ electrochemical assisted synthesis of Cu-BTC thick film is carried out to functionalize a Cu modified glass substrate. This simple, rapid, reproducible and easy to integrate MOF synthesis approach, enables the microfabrication of functional microprenconcentrators with large BET surface area (above 2000 cm2) and active pore volume (above 90 nanoliters) for the efficient adsorption of nerve agent molecules along the microfluidic channel 2.5 cm in length. Equilibrium adsorption capacity of the bulk material has been characterized through thermogravimetric analysis after exposure to controlled atmospheres of a sarin gas surrogate, dimethyl methylphosphonate (DMMP), in both dry and humid conditions (30% RH at 293 K). Breakthrough tests at ppm level (162 mg/m3) reveal equilibrium adsorption capacities up to 691 mg/g. The preconcentration performance of such µ-devices when dealing with highly diluted surrogate atmosphere, i.e. 520 ppbV (2.6 mg/m3) at 298 K, leads to preconcentration coefficients up to 171 for sample volume up to 600 STP cm3. We demonstrate the potentialities of Cu-BTC micropreconcentrators as smart first responder tools for "on field" detection of nerve agents in gas phase at relevant conditions.


INTRODUCTION 38
Vulnerability of critical indoor infrastructures to intentional nerve agents release poses a 39 significant point of concern for authorities responsible for incident preparedness and prevention. 40 Such chemical compounds are odorless, colorless, highly persistent, volatile and lethal even at 41 low concentration. 1

Synthesis and characterization of Cu-BTC films on Borofloat substrates 144
The electrosynthesis of Cu-BTC was carried out as previously reported (see Section 2 of the 145 Supporting Information). 21 Using an Autolab potentiostat PGSTAT302N, cyclic pulses of 146 different current (ranging from 1 to 15 mA) were applied between the Cu modified glass 147 substrate used as working electrode and the counter-electrode. In this work, different 148 electrosynthesis parameters were explored (Table 1)  Resolution Diffractometer) and XPS (Kratos Axis X-ray spectrometer). 163 Table 1. Electrosynthesis conditions studied in this work. The textural properties of Cu-BTC films were assessed from Ar physisorption analysis, carried 169 out on commercial Cu-BTC in powder form, Basolite C300, purchased from Sigma Aldrich. The 170 as received bulk material was also used to quantify the adsorption properties towards DMMP 171 when exposed to 3748 mg/m 3 of DMMP at 293 K (in dry air and at 30 % RH at 293K). Similar 172 experiments were performed on previously activated samples (degassed overnight in an oven at 173 423 K and atmospheric pressure) to analyze the influence of water content on DMMP adsorption 174 properties. An aliquot of the exposed MOF as well as of the activated material were taken and 175 analyzed by thermogravimetry on a TA Instruments TGA-Q5000 (using N2 up to 873 K at 10 176 K/min as heating rate The preconcentration performance of the µ-device is evaluated from its preconcentration 197 coefficient, denoted as K 24 . This parameter mainly depends on the adsorbent type and loading, 198 the trapping efficiency, the number and nature of competing vapors in the sample and the sample 199 volume. According to IUPAC guidelines 25 , this value is defined as the ratio of the air sample 200 collected volume to the volume in which that same mass is released according to the 201 quantification at the point of detection, assuming no transfer mass losses take place. 202 Accordingly, the K estimation is carried out under experimental conditions that guarantee the 203 absence of target molecules in the outlet stream, i.e. no breakthrough conditions. 204 Preconcentration experiments were performed with atmospheres of 2.64 mg/m 3 (520 ppbV) of 205 DMMP in dry N2. As previously, atmospheres were generated by circulating 10 cm 3 /min of dry 206 N2 through a calibrated permeation tube of DMMP (VALCO, permeation rate of 148.41 ng/min 207 ± 3.07 at 90 ºC) and fed to the µ-preconcentrator. The atmosphere was sampled for 10, 30 and 60 208 min at 298 K, after which the µ-device was firstly swept with 3 cm 3 /min of the GCMS carrier At current densities above 1.5 mA/cm 2 , Cu-BTC crystallization begins to occur preferentially 237 at the borders of the metal electrode where the Gibbs free energy of the nucleation process is 238 decreased due to the presence of more grain boundary defects. By increasing the number of 239 cycles, the population of discrete Cu-BTC crystals grows rapidly all along the metal substrate, 240 but without forming a continuous and homogenous layer (Figure 3.a). At the highest current 241 density tested (Jsynthesis = 28.8 mA/cm 2 ), Cu-BTC layer formation is equally unsuccessful (Figure 3.c). This observation is attributed to the instability of the glass Cr/Au-Cu interface, more 243 pronounced at higher overpotentials (above 9 V) due to the electro-migration of Cr atoms into 244 the Au films (Figure 2.c). 27 In addition, the inter-diffusion of Cr into Au layers is temperature 245 dependent, i.e. solubility of Cr in Au at 323 K increases up to 5 atom %. 28  2p3/2 and Cu 2p1/2, respectively, which are presented on Cu-BTC films. 32 The peaks of Cu 282 2p3/2 and Cu 2p1/2 are also de-convoluted into two components (see Figure 4.d). Thus, the 283 fittings of Cu 2p3/2 lead to binding energies of 932.6 and 934.7 eV, which correspond to surface gives rise to the surface Cu2+/Cu+/0 M ratio of 1.017. Similarly, the Cu 2p1/2 peak is de-286 convoluted in two components at 952.7 and 954.5 eV. 287 In addition, there are three shake up satellite peaks, which are typical Cu2+ in cupric 288 compounds. These satellite peaks, on the high binding energy side of the core level Cu 2p XPS 289 data at 939.8, 944.1 and 963.1 eV, originate from multiple excitations in copper oxides and they 290 are known to be characteristics of CuO phase. 33 Therefore, the presence of the intense shake-up 291 satellite structures observed in the Cu 2p XPS spectra of D3 sample, accounting for circa 35% of 292 total copper content on the surface, was an indication of the simultaneous formation of  and CuO phases, aon the surface of the electrosynthesized films. 294 The starting Cu/Ti electrode was also examined. Peaks at 932.8 eV (Cu 2p3/2) and 952 eV (Cu 295 2p1/2) and the absence of satellites shake-up lines characteristic of Cu2+ give clear evidence that 296 Cu is present in the+1 oxidation state, in agreement with the XRD spectra show in Figure 4.c. 297 (red) and commercial Cu-BTC powder (black). Crystallographic patterns from. 14 d) Core level 303 Cu2p XPS spectra of the Cu-BTC film (blue) and pristine Cu-Ti substrate (red). 304  BET surface area is 1812 m 2 /g (correlation coefficient of 0.9999). Micropore size distribution 330 has been modeled using nonlocal density functional theory (NLDFT) and considering cylinder 331 shape pore. The pore network of Cu-BTC has a simple cubic symmetry. It consists on a 3 332 dimensional channel system with main pores of ca. 9 Å and tetrahedral side pockets of ca. 5 Å. 35 333 Such pore size distribution is illustrated in Figure 6.b with a total pore volume of 0.81 cm 3 /g. 334 Cu-BTC is extremely sensitive to water content due to the very strong interaction between 335 open Cu(II) sites and water molecules. 27 TGA of the as received Cu-BTC is shown in Figure 6.c. 336 The first weight-loss up to 373 K accounts for 1216 mg/gCuO and it is mainly associated with 337 water content due to the relatively hydrophilic large pores. On the contrary, the water content of 338 activated samples is notably reduced, i.e. 150-175 mg/gCuO (Table 2). This observation highlights 339 the importance of the thermal activation to empty the framework efficiently. 36  An aliquot of as received Cu-BTC sample was exposed to the fabrication process conditions to 351 assess the framework thermal stability during the anodic bonding, i.e. 523 K for 36 h (see Figure  352 6.d). A similar water weight loss is registered during the first heating ramp up to 523 K. During 353 the next 5 h at 523 K, no appreciable weight loss is recorded. Afterwards, the slope increases 354 gradually with time on stream. After 36 h exposure at 523 K, the weight loss due to the partial 355 degradation of the organic linker accounts for 11 %. Accordingly, the anodic bonding conditions, 356 523 K @ 1 kV, were kept for 5 h to preserve the Cu-BTC framework. 357 (DMMP@3748 mg/m 3 + water@7362 mg/m 3 ) in N2, respectively. The first weight-loss up to 363 373 K is attributed to water desorption, whereas those shown at temperatures above 373 K are 364 due to DMMP release. According to tabulated data, DMMP uptake is shown to be dependent on 365 the water content of the gas stream with a maximum DMMP sorption capacity of 771 mg/gCuO at 366 dry conditions. For single DMMP adsorption on activated samples, the differential thermogram 367 reveals the existence of three different sorption sites with different interaction energies (Figure 7.a). It is also found that most of the weight loss to DMMP d cc a 433 K, DMMP 369 = 473.0 mg/gCuO, vs. 112 mg/gCuO @ 500 K and 186 mg/gCuO @ 543 K. For the binary mixture 370 (Figure 7.b), the DMMP uptake of activated sample is notably reduced to less than half, i.e. 297 371 mg/gCuO; and the triple peak that was seen before, no longer appears. This value is only slightly 372 superior to the quantified for co-adsorbed water vapor, i.e. 238 mg/gCuO. simulations. In order to find minimal energy of the system, the temperature was modified 401 externally to simulate the annealing of the system from 100 K to 10 5 K for 20 cycles with 20000 402 of steps per cycle. The molecular interactions have been simulated by using UFF force field and 403 Ewald method as summation method. A similar study was performed on copper/copper oxides to 404 quantify the effect of the electrode and crystalline impurities. 405   Figure 10.b. Assuming the entire release of the collected DMMP mass by thermal 497 flushing at 473 K, the desorption volume is given by the carrier gas flow rate that sweeps the 498 microdevice, i.e. 3 cm 3 /min, and the full width at half maximum (FWHM) of the registered 499 desorption peak (see Section 4 of the Supporting Information for more details). There is a linear 500 correlation between K and sample volume, i.e., higher values of K are observed at higher sample 501 volume in agreement with published literature. 42 The maximum K is 171 and corresponds to 502 half-maximum injection peak width of ~1 min after collecting 1584 ng of DMMP (Vsample = 503 600 cm 3 ). Such preconcentration coefficient would require 60 min of sampling at 10 cm 3 /min. It 504 must be emphasized; however, that similar breakthrough volumes and K values are attained at 505 higher sampling rates, i.e. 100 cm 3 /min, for microfluidic cavities 40 m depth at the expense of 506 higher hydraulic losses. Thus, the latter time period may be reduced to 6 min, more suitable for 507 some applications requiring shorter analytical duty cycles. Analogous experiments carried out 508 with PC_Cu device, not shown here, did not succeed due to their dynamic sorption values were 509 almost null at these experimental conditions. 510   Section S1.

Fabrication of micropreconcentrators
Fabrication process is divided in five different steps: cavity etching, adsorbent incorporation, microdevice sealing, heater element integration and inlet/outlet ports mechanization ( Figure S1).
Cavity etching is the first step performed to fabricate the microdevice. It is divided in to two steps: 1) photoresist deposition, 2) etching and 3) photoresist removal. For the simple geometries as well as large features (>100 µm) and deep cavities (> 20 µm) studied in this work, wet etching with KOH was used.  For the fabrication of Cu-BTC based micropreconcentrators, the coating step for adsorbent deposition required the presence of a metallic thin film. This metallic thin film was not performed on the Si cavity but on the top Borofloat cover by standard lift-off process as squematically described in Figure S4.

Section 3. Breakthrough Testing of functional micropreconcentrators
Sorption dynamics of the functional µ-preconcentrators were evaluated by analysis of the monitored breakthrough curve. A typical breakthrough curve, as depicted on Figure S6, follows the evolution in time or volume (of circulated gas) of the eluted analyte concentration downstream of the sorption unit. Mostly, the eluted concentration Cx is normalized by the feeding concentration C0, so it ranges from 0 to 1.
For this work, breakthrough point is considered as the moment where Cx/C0 = 0.05, which implies that the adsorbent is becoming saturated and begins to be unable to trap all the analyte molecules carried by the feeding gas. The moment at which breakthrough takes place is denoted as breakthrough time (tb). Similarly, the volume that has been fed to the bed is defined as breakthrough volume (Vb = tb * Qfeed, being Qfeed: feed volumetric flow). The ratio of uptake sorbate  A conventional breakthrough test comprises the following steps (see Figure S7) depicts the schematics of the set up installed at the University of Zaragoza): 1. Adsorbent pre-treatment: before the adsorption experiment, the adsorbent is regenerated thermally by placing the microdevice on a hotplate at 200 ºC while inert N2 sweeps the cavity and degas the sorptive layer. Ideally this step is performed until no undesirable specie was detected.
2. Analyte baseline: analyte was fed directly into the detector to obtain the baseline signal that is used for calibration. To avoid contamination on the adsorbent material, dry N2 is circulated through the microdevice while this step takes place. The µ-device is connected to the GCMS 6-port valve through capillary tubing (320 µm inner diameter) though high temperature septum glued to the Borofloat side of the device. In order to avoid adsorption of the DMMP in the faces of the septa, an intermediate layer of Kapton has been placed between the Borofloat side and the septum. This approach relies on the rapid thermal desorption of DMMP and the adequate carrier flow rate to push out the bolus in a plug-flow fashion, without the need of a focusing stage. Thus, the experimental protocol is reduced to the sampling of DMMP at room temperature and 10 cm 3 /min for certain time, the flushing out of the DMMP molecules in gas phase with GCMS carrier for 5 min, after which, the adsorbent is rapidly heated at 523 K, releasing the DMMP and being pushed out to the detector by the GCMS carrier (see Figure S8).
The preconcentration performance of the µ-device as sampling unit is evaluated from its preconcentration coefficient, denoted as K. According to IUPAC guidelines, this value is defined as the ratio of the gas sample collected volume to the volume in which that same mass is released according to the quantification at the point of detection, assuming no transfer mass losses take place. Accordingly, the K estimation is carried out under experimental conditions that guarantee the absence of target molecules in the outlet stream, i.e. no breakthrough conditions. Thus, given a desorption peak (see Figure S9), the preconcentration coefficient K is calculated as follows: where Qsampling is the feeding flow rate, tsampling is the sampling time, Qdesorption is the flow rate at which the released sample is being flushing out and FWHM is the full width at half maximum of the desorption peak registered by the detector.