Three RTILs, namely, 1-butylpyridinum bis(trifluoromethylsul fonyl)imide ([BPY][NTf₂] (IL1), Methyltrioctylammonium bis (trifluoromethylsulfonyl)imide [N1888][NTf₂] (IL2) and 1-(2-Hydroxyethyl)-3-methylimidazolium bis(trifluoromethylsulfonyl) imide [HOEMIM][NTf₂] (IL3) were obtained from Ionic Liquids Technologies, Inc. (Tuscaloosa, AL, US). Silicon wafers (n-typed, 4 in diameter, 500 μm thickness, double sided polished) and Borofloat wafer (4 in diameter, 175 μm thickness) were purchased from University Wafer, Inc. (Boston, MA, US). Fused silica capillary tubes with 100 μm internal diameter and 200 μm outer diameter were purchased from Polymicro Technologies, LLC. (Lisle, IL, US). Fused capillary tube with 15 m length and 0.25 mm thickness, Capillary column with 15 m length and 0.25 mm thickness were purchased from Restek (Bellefonte, PA, US). JB Weld epoxy was obtained from the local store. Dimethyl ethyl silane, tris (pentafluorophenyl)borane, acetone, naphthalene, heptane, octane, nonane, benzene, toluene, ethylbenzene, p-Xylene, m-Xylene, o-Xylene, isobutylbenzene, styrene, butylbenzene, 1,2-Dichlorobenzene, 2,5-Dichlorotoluene, 1,2,4-Trichlorobenzene, benzyl chloride, 2-nitrotoluene, 3-nitrotoluene, 4-nitrotoluene, and standard of C7-C30 saturated alkanes (1000 μg/ml each compound in hexane) were purchased from Sigma Aldrich (St Louis, MO, US). Ultrapure helium, nitrogen, hydrogen, and air were obtained from Airgas, Inc. (Christiansburg, VA, US). Table 1 shows the order of retention times and temperature of ebullition for the 20 components. All solutions were kept in the refrigerator before using them. The retention time of individual chemicals was achieved separately.

All gas chromatography measurements used to characterize the behavior of the ionic liquids were performed on an Agilent 7890A GC (Agilent Technologies, Inc, Santa Clara, CA, US) system equipped with an automatic sampler (7693A), split/splitless inlet, and flame ionization detector (FID). Helium was used as a mobile phase. The flow rate of hydrogen and air for FID detector were kept at 30 ml/min and 300 ml/min, respectively. SEM images were obtained using Carl (Zeiss) EVO 40 Scanning Electron Microscope (SEM). The data were analyzed using the Agilent ChemStation Edition C.01.04.

The detail of the fabrication of the semi-packed columns was discussed in our previous reports [34, 35]. Briefly, the etching photoresist mask was pattern on surface of the silicon wafer using AZ9260 photoresist. Then silicon pillars were created using deep reactive ion etching (DRIE) process. Sulfur hexafluoride (SF6) and octafluorocyclobutane were used in the etchant and passivation steps. The etched pillar were the circular with around 20 μm in diameter with depth of 240 μm and column length of 1 m. After removing photoresist mask using acetone, piranha, and oxygen plasma, the etched silicon wafer was anodically bonded with a 175 μm thick Borofloat glass wafer at 1250 V and 400°C. After dicing the wafer into the individual microcolumns, the inlet and outlet were attached to 30 cm long silica capillary tubing. Stationary phase coating was performed using a freshly prepared solution with a concentration of 15 mg/ml in acetone for IL1, IL2, and 1.5 mg/ml in acetone for IL3 stationary phases.

Columns deactivation is the first step before starting the stationary phase coating process. Physical deactivation was achieved by depositing 13 nm of an aluminum oxide layer using atomic layer deposition (ALD) [29]. However, chemically deactivating silicon-based GC columns is typically used to enhance their efficiency. The most common chemical method used for this purpose involves the reaction between silanol groups (Si-OH) present on the surface of silicon and organosilicon compounds, such as dimethylaminosilanes, chlorosilanes, alkoxysilanes, and allylsilanes [36-39]. Despite the wide application of these approaches, undesirable polysiloxane networks are often created. These methods require long reaction times between 2-24 h. In this paper, the column efficiency was enhanced by using a fast chemical silanization reaction (Fig. S1). The chemical deactivation was performed with creating an organic monolayer by siloxanation of oxidized silicon pillars surfaces. The 1% dimethyl ethyl silane and tris (pentafluorophenyl)borane (B(C₆F₅)₃) as a catalyst were prepared in CH₂Cl₂ and injected inside the column with a flow rate of 20 μl/min. The supplementary video clip 1 shows the process. The bubbles seem in the video are related to the hydrogen generation during chemical deactivation. Following this step, the column was washed with CH₂Cl₂ for 30 minutes until no bubbles were detectable inside the column. This method is one of the fastest GC column deactivation process that the salinization reaction only takes a few minutes to accomplish [39, 40].

To achieve a mass produce of uniform stationary phase coating, designing a fast coating procedure free of bubbles is crucial. Fig. 1 indicates the schematic of the optimum coating steps. Several coating and drying strategies were investigated to achieve an efficient method. Fig. 2 shows the microcolumns filled with IL solution using the N₂ purging method (commonly used approach by our group and others) and the syringe pump. Using N₂ for solution purging leaves behind bubbles especially at the end parts of the column and the curved regions (Fig. 2A). Applying a constant flow rate with a syringe pump gives a chance to fill the ionic liquid solution uniformly (supplementary video clip 2). After the column packed with the solution, the flow rate was increased to cover up any remaining empty spots (supplementary video clip 3). Fig. 2B demonstrates smooth and bubble-free coated microcolumns using 20, 40, and 60 μl/min of the solution flow rates. In the next step of the coating, the solvent is required to evaporate and leave behind the thin layer of the stationary phase on the surfaces of the entire column. Different strategies were applied to dry the column. Fig. 2 provides the effect of them on the smoothness of the drying procedure. In one method, the withdraw mode of the syringe pump was used to create a local vacuum and push back the extra solution from the column. The process was examined in both sealed and open-ended situations. In both cases, nonuniform residues of the solution were observed inside the channels (Fig. 2C). In another technique, N₂ purging was used to accelerate the evaporation process. Still, the nonuniform residues were created (Fig. 2D). Finally, the N₂ purging was combined with a water bath (Fig. 2E). This process accomplished the drying procedure in less than 10 minutes. Fig. 2F shows after one minute N₂ purging, while Fig. 2G shows a uniformly dried column after four minutes. The column was kept for a while to make sure solvent residues escaped from the channels. Elevated uniform temperature using a water bath with a combination of N₂ flow provides fast and uniform removal of excess coating solution and avoids trapping microbubbles between silicon pillars. All these three procedures gave better drying quality than using an overnight vacuum inside a desiccator. So, the N₂ purging with a water bath of 40°C was selected through the entire work as the preferred drying step.

Fig. 3A shows the column after coating with a syringe pump and drying using N₂ purging in the water bath. The inset of the figure indicates SEM images of the four pillars silicon array. The SEM was used to monitor the smoothness and thickness of the IL stationary phase of the GC microcolumns. Fig. 3B shows the ionic liquid around the silicon pillars. It can be seen that with the selected coating method, the uniform layer of the ionic liquid was created around the silicon pillars. Inset demonstrated the magnified image of the stationary phase layer. The upper observed distortion happened during the peeling out of the upper layer to prepare for the SEM image. Figs. 3C and 3D indicate the optical microscope images of the column before and after the oven pretreatment. The column’s treatment in 240 °C removes all solvent residues on the glass slide and around the pillars and stabilizes the ionic liquid layer.

To characterize the quality of the column coating, column efficiency and column resolution were investigated. The efficiency of a column is reported as the height-equivalent-to-a-theoretical-plate (HETP) that is inversely related to the number of theoretical plates. The resolution denotes how the two adjacent peaks are separated. The geometry of the column and the thickness of the stationary phase play important roles in the column performance. Naphthalene was used as a probe to calculate plate number as it is well retained compound and was used in many studies to evaluate the column efficiency [33]. Naphthalene was injected into the columns under isothermal temperature (100 °C) to obtain the corresponding HETP values. The trend line between the HETP and the carrier gas velocity is shown in Fig. 4. It represents the effect of ionic liquid type, deactivation method, and coating thermal pretreatment temperature on the efficiency of the microcolumns. Fig. 4A indicates the effect of the pretreatment temperature on the efficiency of IL1. The three different temperatures of 200 °C, 220 °C, and 240 °C were applied. HETP for IL1 column treated at 240 °C is lower within a broader range of the carrier gas velocity. The 240 °C treated IL2 shows lower HETP at any given carrier gas velocity. Besides, IL2 demonstrates similar thermal treatment behavior to IL1. In addition to the thermal treatment, the effect of IL’s type was investigated (Fig. 4B). Different ionic liquids present different trajectories; IL3 treated at 200 °C shows lower HETP when compared to that of IL2 treated at 220 °C. Fig. 4C represents the effect of the deactivation methods of the silicon surface. Both chemical and physical deactivated columns show lower HETP and better efficiency compared to the non-deactivated column. While ALD surface treatment shows better efficiency, it may not be accessible in some nanofabrication facilities. The results show that the proposed chemical deactivation process can produce comparable results, especially in lower carrier gas velocities. The measurement was repeated for two columns, and an average variation of 400 was found between columns (R²=0.99) (Fig. 4D).

The efficiency of the semipacked column coated with ILs was compared with the commercial capillary column with 15 m length and 0.25 mm thickness coated with BP20 as the stationary phase and the same capillary column coated with IL1. Fig. S2 shows the HETP of the capillary columns. Semipacked column with a similar stationary phase compared to the capillary column demonstrates holding efficiency in higher carrier gas velocity making it possible to analyze the analytes in shorter times with the appropriate resolution values.

A plot of the theoretical plate height versus carrier gas velocity will never become completely linear and is different for various GC column types. In literature for the packed column, efficiency is represented by the van Deemter (equation 1).

Where A is the eddy diffusion results because in packed columns spaces between particles along the column are not uniform. Therefore, molecules take different pathways along the column, B is the coefficient for longitudinal diffusion, and C is the resistance to mass transport in the gas and liquid phases, and x is the carrier gas velocity.

For square spiral semipacked columns, Golay-Goucheon kinetic model was shown better regression fit in previous works (equation 2).

Where D was defined as the extracolumn band broadening [23, 41]. For the semipacked silicon pillar that was reported in this work, both these equations did not provide perfect fitness. We figured out the combination of them shows the best fit for the experimental Golay plot represented in Fig. 4. Fig. S3 shows the amount of fitness for different Golay equations for one of the cases of IL stationary phases. Based on data, equation 3 was used to calculate Golay coefficients.

This seems reasonable as silicon pillars can create different pathways similar to the particles in the packed column, so eddy diffusion should also be considered. Table 2 represents the calculated values for A, B, C, and D for different kinds of ionic liquid stationary phases. Data shows higher efficient columns indicate lower eddy and longitude diffusion coefficients and have lower mass transfer resistance.

The separation performance of the MEMS columns was evaluated using a sample mixture, including 20 compounds with the boiling point in the range of 80 °C to 238 °C. The separation of the saturated alkanes was examined for the IL1 stationary phase under different coating and different temperature ramp conditions (Figs. S4, S5). The performance of the columns’ coating was evaluated with temperature-programmed chromatograms. Fig. 5 shows chromatograms of the MEMS columns with the IL1 stationary phase treated at 200 °C (A), 220 °C (B), and 240 °C (C) oven temperatures. The starting temperature was 30 °C, and the programming rate was 40 °C/min. As the data demonstrates, the oven temperature treatment dramatically influences the efficiency and the peak symmetry of the IL stationary phases based columns. The plate number of the IL1 columns increased from 1600 at 20 psi (200 °C) to 8300 at 45 psi (240 °C) with increasing the treatment temperature. Besides, it affects the symmetry and the resolution of the peaks. Table 3 shows the tailing factor and the resolution of naphthalene (1), 2-nitrotoluene (2), 3-nitrotoluene, and 4-nitrotoluene (4) peaks. The resolution values refer to the resolution between the peak of interest and the preceding peak. Similar to the plate number, the resolution increases with the treatment temperature. Despite some increase in peak tailing, all values are below 2. Also, the Retention factor (k) of naphthalene was found 11.5 for IL1, 4 for IL2 at an inlet pressure of 45 psi, and 7.8 for IL3 at an inlet pressure of 35 psi.

The same pattern of the plate number and resolution was observed for the IL2 stationary phase columns (Fig. 6). Fig. 6A shows the mixture separation using IL3 stationary phase treated at 200 °C. The elution times for some chemicals have been changed compared to IL1, suggesting the different selectivity provided by ILs for these compounds. Among all prepared columns, IL3 can separate p-Xylene from m-Xylene (peak numbers 7 and 8). Figs. 6B and 6C show the chromatograms for IL2 thermally treated at 220 °C and 240 °C. Our data also indicate the crucial role of the thermal treatment in enhancing the plate numbers. The measured plate numbers were 3000 at 35 psi for the IL2 column before the treatment. After treatment at 240 °C, the plate number increased to 7800 at 35 psi. Table 4 shows the resolution and tailing analysis for IL2 and IL3 columns. With a similar trend for IL1, for these columns, the resolution also improved by increasing the treatment temperature. The increasing trend for efficiency and resolution can be related to the ionic liquids’ thermal dependence of viscosity. Previous researches indicated that decreasing viscosity of ionic liquids with increasing thermal treatment. Lower viscosity can lower mass transfer resistance and improve the interaction between the gas molecules with ionic liquid based stationary phases [42, 43].

Fig. 7 shows the chromatogram of the columns (A) compared with the chemically deactivated columns with dimethyl ethyl silane in the presence of the tris (pentafluorophenyl)borane as a catalyst (B). The deactivation improved the symmetry and the resolution of the peaks. This improvement comes from the deactivation of the active chemical sites on the surface of the silicon, such as hydroxyl and silica groups.

The presented preliminary data regarding this fast deactivation method is promising. However, more studies are required to investigate the effect of other types of silanes on the separation efficiency.

Besides, the performance of the semipacked column was compared with 15 m capillary columns. Figs. S6, S7 show the chromatograms of 20 components that were used to evaluate the semi-packed columns. The data was achieved in 45 psi to compare with semipacked columns. Moreover, as the capillary columns had the best efficiency at 10 psi carrier gas pressure, corresponding data were also reported. Tables S1 and S2 demonstrate the resolution and symmetry data for the three last peaks in the chromatograms. The resolution was reported between peak of interest and the preceding peak, and symmetry was reported based on tailing at 5% height of the peaks. At the same carrier gas pressure of 45 psi, the capillary column based on IL1 shows a more symmetric peak but lower resolution than the IL1 based semipacked column. In contrast, the capillary column based on BP20 demonstrates better resolution and symmetry in comparison to the semipacked columns.

The stability of the IL-based MEMS columns was also investigated for 100 repeated runs. The data was obtained for naphthalene under 30 psi carrier gas pressure with 0.3 μL injection of sample. For all ILs, there is a significant difference between the first and the second run, but it stabilizes after the second run. Results show height is the most affected variable. Besides, based on relative standard deviation (RSD) data, IL1 with 2.11 height RSD in comparison to 4.47 and 4.50 for IL2 and IL3, respectively, is a more stable column (Table 5). The effect of cleaning between each run also was investigated. More detailed data have been provided in the supplementary section (Figs. S8, S9).