The 16 PAHs targeted by the United State Environmental Protection Agency since 1976 are considered in this study. A standard mix of the 16 PAHs at 10 µg/mL in methanol was purchased from Sigma-Aldrich (USA) to prepare standard and spiked water samples. It contains naphthalene (NAP), acenaphtylene (ACL), acenaphtene (ACE), fluorene (FLR), phenanthrene (PHE), anthracene (ANT), fluoranthene (FTN), pyrene (PYR), benz(a)anthracene (B(a)A), chrysene (CHR), benzo(b)fluoranthene (B(b)F), benzo[k]fluoranthene (B(k)F), benzo(a)pyrene (B(a)P), indeno(1,2,3-cd)pyrene (IND), dibenzo(a, h) anthracene (D(h)A) and benzo(g, h, i)perylene (B(ghi)P). Number of rings, water-octanol coefficients and molecular weight of the 16 PAHs are given in TableS1 in Supplementary Material.

Purified water was obtained from a MilliQ system (Millipore, USA). Syringe filters with 0.45 µm polyethersulfone membranes (PALL, Port Washington, NY, USA) were used for filtering the natural environmental water samples from Isère river.

Silicon/glass microfluidic chips (Figure 1) were manufactured from 200 mm silicon wafers using standard micro-technology techniques such as photolithography and Deep Reactive Ion Etching (DRIE) as described elsewhere [27]. The internal chamber contains rows of squared micro-pillars of 17 µm side, 230 µm height and a 10 µm gap between pillars (Figure 1). External dimension of the chip is 30 × 12 mm². A 150 nm porous SiOCH thin layer was used as extraction phase and deposited by PECVD before glass top anodic bonding as described in [22]. Before chip dicing, a heating resistor (8.5 Ohms) and temperature probes (100 Ohms) were deposited on the back side by sputtering a thin Pt layer. Finally fused-silica capillaries (internal diameter 250 µm, outer diameter 360 µm) from Photon Lines (France) were glued to the inlets for fluidic connection.

The system architecture is shown in Figure 2. The system contains a peristatic pump (model P625 from Instech), a pressure regulator (IQ + flow from Bronkhorst), two 3-way and one 2-way valves (the Lee Company), a power supply board (NV1-4G5TT), electronic modules (NI9263, NI9217, NI9403 and cDAQ9174 from National Instrument) and a homemade interface board. A fused silica “Y” connector from Supelco is used for the fluidic tee connection, whereas 30 cm long fused-silica capillary (internal diameter 250 µm, outer diameter 360 µm) is

used between the tee and the GC capillary column. The originality of the assembly lies in the fact that the gaseous sample of PAHs never passes through a valve and that the transfer line and the fluidic tee through which PAHs pass after desorption are both heated using powered carbon fibers wrapped around the connections to avoid cold spots. Thus, the valves do not have to be regulated in temperature which makes it possible to reduce the power consumption and the analysis times since no long cooling time between each analysis is required.

All the system components are assembled in a 35 cm (l) × 35 cm (w) × 27 cm (h) module as shown in Figure 3. The system is connected through USB with a laptop for device control with a homemade Labview application. In particular chip and transfer line temperatures are regulated using PID Labview function and the feedback of the temperature measurements of the chip and along the transfer line.

The system operation is divided in 4 modes (Figure 4): in the extraction mode, the pump is activated and valves are configured so that sample is pumped through the microchip at 1 mL/min. In the drying mode, pump is stopped, helium pressure is set to 1 bar and valves are switched so that the gas flow is blowing

through the chip towards waste for 10 min. To ease the drying, the chip temperature is set to 50˚C and the capillary/fluidic tee temperature is set to 200˚C. The drying of the chip under a carrier gas flow at mild temperature before the thermal desorption removes residual excess water and avoids problems with premature degradation of the capillary column stationary phase or FID signal disturbance. Because of the difference in the fluidic impedance between the capillary towards waste and the capillary towards the GC column, there is a difference in pressure drop between the path to the waste and the one to GC column so that no water contamination is observed in the latter path.

During the desorption stage, valve 3 is closed, helium pressure is set to 0.5 bar and chip temperature is set to 250˚C for 5 min. The temperature of 250˚C is reached within 60 s with a 24 V power supply. Molecules that are desorbed from the chip are thus focused at the GC column inlet which is maintained at the GC oven temperature, i.e. 50˚C. Finally during the analysis mode, chip and capillary heating is stopped while maintaining helium pressure at 0.5 bar and valves configuration as in stage 3. GC analysis is operated using the thermal gradient described in the next session.

A 2 m long 5 MS capillary column (Supelco) of 250 µm internal diameter and 0.2 µm of phase was used. During analysis, the flow rate was measured at 1 mL/min. The column temperature was initially set at 50˚C, then increased to 180˚C at 20˚C/min, to 240˚C at 10˚C/min and to 250˚C at 5˚C/min and hold 1 min at this temperature. The temperature of the FID detector was set to 250˚C.

In the case of direct injection of the commercial mix for calibration procedure, the column was connected to the split/splitless injector. Liquid samples of 2 µL were injected with a split ratio of 9.3 and a temperature of 250˚C.

Extraction yields were estimated by comparing peak areas obtained with the FID after the whole protocol (sample extraction/drying/desorption/analysis) and those measured from the direct injection of the commercial mix of 16 PAHs.

In our previous studies we have shown that our microfluidic chips could extract efficiently within 20 min PAHs contains in 10 mL water samples, either by using PDMS extraction phase or SiOCH extraction phase [21] [22]. Desorption was performed using a solvent (acetonitrile) and the detection was done by HPLC-FLD. In this study we go further in the integration and development of a transportable system, using thermal desorption which does not use a consumable (solvent) and is more suitable for integration, automation of the protocol and coupling with GC.

An extraction rate of 1 mL/min was used to extract the PAHs contained in 10 mL water sample (i.e. extraction time of 10 min) since no degradation of the recovery rate was observed, compared with the 20 min time of our previous studies. Extraction is thus rather fast compared to other usual solid phase extraction methods [11].

The relevance of on-chip thermal desorption was assessed by comparing the analysis of a 10 mL water sample spiked at 1 µg/L with 16 PAHs extracted on the chip and recovered by thermal desorption with the analysis of a commercial reference sample directly injected in the GC. In Figure 5 are shown the chromatogram obtained after the extraction of the water sample followed by drying and thermal desorption, as well as the reference chromatogram. Except for the most volatile PAH (Naphthalene), all PAHs peaks can be identified on the chromatogram corresponding to the extracted sample. Two pairs of PAHs (CHR and B(a)A as well as B(b)F and B(k)F) are not resolved on both chromatograms. The use of a short 2 meter-long GC column, consistent with the perspective of portable miniaturized GC system, explains that not all peaks are resolved. Other peaks present in the first chromatogram are due to some interfering compounds also observed in the case of blank (pure water) samples as illustrated in Figure 6.

Some preliminary experiments have shown that a connection capillary and fluidic T temperature of 200˚C is required but sufficient to keep all the PAHs in the gaseous phase until the separation column inlet. The desorption duration and chip temperature have respectively been set to 5 minutes and 250˚C so that all the PAHs are well desorbed from the chip at one time: in Figure 7, the almost flat chromatogram confirms that desorption is complete at the first chip heating.

Moreover, we checked that no cross-contamination between the samples occurs. The blank samples were always clear, which means that there is no undesired adsorption along the system path.

These results prove that it is possible to carry out the extraction and the thermal desorption of PAHs on silicon-based chips before a GC-FID analysis with reduced time and energy consumption. The whole protocol, from extraction to desorption including the drying step, lasts only 25 minutes.

By comparing the peaks areas obtained after thermal desorption with those measured in the case of the direct injection of the commercial mix, one can deduce the extraction yield of our system. If two peaks are not resolved, the sum of the 2 PAHs quantities is taken into account. The extraction yield varies from 7.8% ± 1.2% for FLR to 105.4% ± 9.1% for (CHR + B(a)A), as detailed in Table 1 and Figure8. Quite good repeatability of the protocol is observed, as reflected by the average Relative Standard Deviation (RSD) of 12% calculated on the basis of 4 analysis performed in the same conditions. These analyses have been carried out over 2 days. The chip-to-chip dispersion of the yields is also low (cf. FigureS1 in Supplementary Material). Besides we have observed that the same chip could be used for dozens of analyzes without loss of performance.

These yields are slightly lower than the ones deduced from our previous study using solvent desorption [22]. This could be explained by some losses of the PAHs molecules during our protocol, in particular during the chip drying, some of the PAHs going possibly to the waste path. Measurement protocol may be improved to increase the recovery yield of the most volatile PAHs (shorter extraction, drying or desorption times, use of more polar extraction phase). Yet, this would probably imply to lose some signal for the heavier PAHs. By running sequential analysis with different parameters, it should be possible to get responses for the whole range of PAHs with the same system.

It is known that dissolved organic matter present in environmental samples could affect the performance of the solid phase extraction [28] [29]. In order to check that our system could be used for the monitoring of such environmental samples, we performed the extraction of spiked natural water sample. As described in the experimental part, sample from river Isère was collected, filtered and spiked with the commercial PAHs solutions up to 5 µg/L. We checked that no PAHs could be detected in the filtered river sample before the spiking step, either because they were not present in the sample, or because the concentrations were lower than the limit of detection of our system.

As seen in Figure 9, no large difference is observed between the recoveries in the case of purified water or natural water: heavier PAHs seem to get even better recoveries in the case of the natural water. This experiment therefore validates the use of our prototype for the extraction and desorption of PAHs in natural water samples prior to GC analysis.