The FTS reported in this work was carried out under both conventional gas phase and non-conventional supercritical and near supercritical phase utilizing the newly commissioned bench-scale novel high-pressure fixed bed reactor unit. The reactions were performed over a cobalt catalyst supported on alumina (15 wt% Co/Al₂O₃) that was prepared using the impregnation technique, and then calcined and reduced under hydrogen flow at elevated temperatures. For the gas phase FTS, the reaction was carried out at 240˚C, 20 bar syngas with 2:1 H₂:CO molar feed ratio. As for the SCF-FTS, the reaction was carried out at 240˚C, 65 bar with 20 bar syngas partial pressure and 2:1 H₂:CO molar feed ratio. These conditions were selected to provide the basis for comparing the two reaction media. Each experimental set was allowed to reach steady state conditions in terms of activity (CO conversion) and selectivity (hydrocarbon product distribution), as measured by the on-line GC setup. After reaching steady state, liquid samples were collected for off-line analysis.

The cobalt catalyst supported on alumina (15 wt% Co/Al₂O₃) used for the FTS reactions was prepared using the impregnation technique. Firstly, the alumina support (surface area 255 m²/g, median pore 70 micron and 5000 Å, total pore volume 1.14 cm³/g, packing density 0.395 kg/L) was crushed, sieved (75 - 150) μm, washed with deionized water and then calcined under airflow at 700˚C. Secondly, an aqueous 2 molar cobalt solution was prepared by dissolving cobalt nitrate precursor (Co(NO₃)₂∙6H₂O) in deionized water while stirring at room temperature. Then, the cobalt solution was gradually added to the treated support under constant stirring to avoid lumps formation. Afterwards, the formed catalyst paste was dried, crushed and sieved (150 - 250) µm. Thirdly, the obtained catalyst powder was calcined in a rotary oven under airflow and temperature ramp up to 350˚C. Then, the calcined catalyst was diluted with quartz sand by vigorously mixing the catalyst and quartz to assure homogenous catalyst distribution. Finally, different catalyst loadings were added into the reactor tube on top of a layer of glass wool and quartz.

The high-pressure bench-scale reactor unit used in this project consists primarily of five sections (Figure 1). The first section involves the feed delivery setup for gas stream (e.g. syngs, CO, H₂ and argon) and liquid stream (e.g. solvent hexane). The second section is the reaction zone where the FTS reaction takes place. The third section utilizes two separation columns (hot trap and cold trap) to separate permanent gases, liquid hydrocarbons and wax. The fourth section is composed of the reactants and products on-line GC analysis system. The fifth section, which is the utility section, is composed of three sub-sections: 1) pneumatic air, 2) chilled water system and 3) ventilation setup.

At the feed delivery section syngas (33.4% CO and 66.6 H₂), carbon monoxide (99.97% CO), hydrogen (99.999% H₂) and argon (99.999% Ar) were introduced into the system at controlled flow rates (0 - 500 nmL/min) using four mass flow controllers (MFCs). For the supercritical phase testing, liquid solvent, n-hexane (98.5% n-hexane) was introduced at controlled volumetric flow rates using a High Pressure Liquid Chromatography (HPLC) pump. A feed purification system composed of a cylindrical vessel filled with three layers of adsorbers/catalysts (BASF Selexsorb COS, BASF PuriStar R3-15 and BASF E-315) and an inline gas filter were employed at the solvent stream and at each gas stream for the removal of COS, H₂S, CS₂, CO₂, acetylene, O₂, arsine, sulfur compounds, moisture and particulates. The liquid solvent is then passed through a vaporizer before it is mixed with the feed gases in the static mixer that ensure proper mixing of the feed gases and the SCF.

The reaction mixture then enters the custom made tubular fixed bed reactor (Autoclave Engineers, 16" overall length, 12" heated length, 2/3" internal diameter, 1.0" outer diameter, 73 cm³ net volume in heated zone). The reactor tube was vertically embedded in a hollow ceramic insulated electric furnace with three heating zones that allow reactor temperature control within ±2.0˚C. The pressure control valve located between the reactor and the

hot trap controlled the reactor pressure.

Two flash separation columns were placed downstream of the high-pressure reactor. The first one is the hot trap surrounded by a heating tape and insulated by glass wool and aluminum foil. The hot trap is operated at 150˚C and 0.9 barg to facilitate the condensation of heavy hydrocarbons (wax) prior to the on-line GC system. The formed wax is collected in a special pressure vessel for future analysis. All the lines connecting the reactor with the hot trap, the wax collection vessel and the on-line GC were heated to 150˚C to prevent any wax condensation.

The remaining hot trap gas stream goes towards the cold trap used to separate liquid hydrocarbons and permanent gases. The cold trap is internally cooled using cooling coils with chilled water flow at 4.0˚C. Hydrocarbons and water condensed by the cold trap were collected in a tank and liquid samples were periodically collected for off-line GC analysis.

In a typical experimental run, the system was first flushed with n-hexane and then purged with nitrogen. Secondly, the catalyst was activated in-situ under hydrogen flow (100 - 350 nmL/min) at a temperature range between 180˚C to 350˚C. Thirdly, the system was purged and flushed again. Fourthly, the solvent was introduced (2.43 - 2.74 nmL/min) and the reactor pressure (20 - 80 bar) and temperature (230˚C - 240˚C) were slowly ramped to their desired set points. After the temperature and pressure had been stabilized syngas was allowed (50 - 150 nmL/min) at the desired flow rate. Each experimental set was allowed to reach steady state conditions in terms of conversion and product distribution, as indicated by the on-line GC. After reaching steady state, wax and liquid samples were collected for further off-line analysis.

A fraction of the gases leaving from the top of the hot trap is directed towards the on-line GC system through an air actuated 8 way selecting valve that is used to inject samples at specified intervals. Operation of the sampling valves was controlled by a custom-made Shimadzu GC Postrun software that acquires samples at preset times and record TCD and FID data. This system has been designed to ensure that the same sample to be injected simultaneously to two combined GCs with three detectors. This unique on-line analytical system consists of dual gas chromatographs with three detectors, TCD-1, FID and TCD-2 (Figure 2). TCD-1 detector analyzes permanent gases (O₂, N₂, CH₄, CO, CO₂, C₂ and H₂S) by utilizing two GC columns. Both columns use helium as a carrier

gas at 5.0 mL/min and 4.0 bar. The FID detector analyzes (C₃ - C₁₃) hydrocarbons by utilizing a single fused silica capillary column to separate the hydrocarbons. The capillary column also uses helium as a carrier gas at 5.0 mL/min and 4.0 bar, while air (350 mL/min, 4.0 bar) and H₂ (35 mL/min, 4.0 bar) were used to light the detector flame and N₂ as makeup gas at 25 mL/min and 4.0 bar. Hydrocarbon products heavier than C₁₅ were not detected by the on-line FID. One more detector TCD-2 is used with a single packed column to detect H₂, where in this column N₂ was used as a carrier gas at 25 mL/min and 4.0 bar.

Liquid products from the cold trap are periodically collected for off-line analysis using GC/MS system equipped with FID and Mass Spectroscopy (MS) that uses a capillary column to separate (C₅ - C₃₀) hydrocarbons. The column uses helium as carrier gas at 5.0 nmL/min and 4.0 bar, air at 350 nmL/min and 4.0 bar, and H₂ at 35 nmL/min and 4.0 bar to light the FID flame and N₂ as makeup gas at 25 nmL/min and 4.0 bar.

The CO inlet flow rate to the FTS reactor was measured and controlled by the MFC; however, CO participates in the FTS reaction, and thus its outlet flow rate cannot be known directly. In order to be able to calculate the outlet CO flow rate, it was necessary to use an inert (argon) as an internal standard. The method adopted herein is similar to that previously used by Nijs and Jacobs [23] . Argon does not participate in the FTS reaction, and thus its inlet and outlet flow rates are equal under steady-state conditions. Hence, it can be used to calibrate the outlet CO flow rate by relating the CO peak area/Ar peak area ratio (found using on-line GC TCD channel 2) to CO flow rate. The trend line (straight-line equation) of the peak area ratio (PA_CO/PA_Ar) against the CO flow rate can then be used to calculate the CO outlet flow rate as follows:

where Q_CO,out is the outlet flow rate of CO, “a” is the slope and “b” is the intercept. For example, syngas with H₂:CO of 2:1 (CO is 1/3 of syngas flow) was allowed into the system at different flow rates (10 - 200) nmL/min. Argon was allowed at constant flow rate (10 or 15) nmL/min. The resulting gas mixture was then directed toward the on-line GC/TCD, where CO and Ar peak areas were determined. CO and Ar peak ratio (PA_CO/PA_Ar) was then calculated and plotted against CO flow rate. Figure 3 shows the calibration line for argon flow rate of 15 nmL/min. The resulting straight line equation (y = 0.0548x − 0.0264), where y = PA_CO/PA_Ar and x = Q_CO,Out can be rearranged to find the outlet CO flow rate as follows:

The material balance calculation was based on carbon, where the amount of carbon (as CO) entering the reactor is equal to the amount of carbon (as unreacted CO and produced hydrocarbons) leaving it. Therefore, the CO consumption rate and conversion throughout the experimental run were calculated based on the difference between moles of CO entering the reactor and moles of unreacted CO leaving the reactor as shown in Equation (3)

and Equation (4). Note that syngas H₂:CO ratio of 2:1 was normally used unless CO, H₂ were introduced separately.

where x_CO is the consumption rate of CO and X_CO is the conversion percent of CO.

In order to calculate the CH₄ outlet flow rate, a calibration gas with 10.09 CO mol∙% and 4.00 CH₄ mol∙% (molar ratio = 0.397) was used to find the responding peak area for each component and relate CO flow rate to CH₄ flow rate. The peak areas for CO and CH₄ for the aforementioned concentrations were identified by the on-line GC-TCD and the CH₄/CO peak area ratio is measured as 0.331. After that, the molar ratio and peak area ratio were used to calculate CH₄ outlet flow (CH₄ formation rate) as follows:

CH₄ selectivity was then calculated as follows:

where is the selectivity towards methane.

The hydrocarbon product distribution was determined utilizing the on-line and off-line FID GC setups. The hydrocarbon formation rate (g/g_cat∙h) was calculated as the sum of formation rate of all species (isomers, normal paraffins, olefins, alcohols) with “n” carbon number. Then, the normalized weight percentages (W_n) of each carbon number was calculated from the ratio of the hydrocarbon formation rate of carbon number “n” to the total weight of hydrocarbons formed as detected by the FID.

It was difficult to determine the exact amount of C₆ hydrocarbons produced when running the FTS reaction under near and supercritical conditions since large quantities of n-hexane were present in the feed. This resulted in a very large n-hexane peak absorbing nearby peaks within the C₆ range. Thus, calculating the formation rate of C₆ hydrocarbons was impractical. Hence, the hydrocarbons within the C₆ product range were eliminated from the product distribution and were not included in the Anderson-Schulz-Flory (ASF) calculations. Additionally, for consistency purposes, C₆ hydrocarbons were also eliminated in the gas phase FTS when comparing the performance under the two reaction mediums.

To start the hydrocarbon product distribution calculations, a reference or starting point was needed. For the on-line GC, this was the CH₄ formation rate (g/g_cat∙h), previously determined using the GC TCD. The TCD CH₄ formation rate with the CH₄ peak area given by the FID was used as a reference ratio and held constant for each sample. The ratio was then used to calculate the formation rate of hydrocarbons with different carbon numbers. This method is similar to that previously used by Snavely and Subramaniam [21] . As for the off-line GC FID analysis, a similar approach was used, however the CH₄ formation rate was replaced with n-C₇ formation rate (g/g_cat∙h), determined using the on-line FID, and the on-line n-C₇ formation rate with the off-line n-C₇ peak area was used as a reference ratio in this case.

As discussed below in section 3.1, each hydrocarbon present in the FID spectrum was identified and assigned with its distinct carbon number and type (isomer, normal, olefin, alcohol). After that, the peak area for each carbon number with the same type was summed, and then the total peak area for each carbon number was summed. The weighted molecular weight for each carbon number (n) was then estimated as follows:

The total hydrocarbon formation rate for each carbon number (n), was then calculated as follows:

where F_HC is the hydrocarbon formation rate. is the methane formation rate estimated by the on- line TCD, is the peak are of methane given by the on-line FID.

The weight fraction of hydrocarbons containing “n” carbon atoms (W_n) was then calculated as follows:

where F_HC,n is the hydrocarbon formation rate for “n” carbon number.

The natural logarithm of the weight fraction of carbon number “n” over the carbon number [ln(W_n/n)] was calculated in order to plot the ASF distribution, where [ln(W_n/n)] was plotted against “n”. The slope of the trend line can then be used to find the chain growth probability as follows:

where ln(α) is the slope of the trend line and the chain growth probability is equal to the exponential of the slope (α = e^slope).

The exact and detailed identification of FTS products can be hindered by the large number of compounds involved [21] . One method for identifying FTS products is to use standard solutions that contain FTS products (n-paraffins, α-olefins, β-olefins, branched and oxygenates of C₁ - C₃₄) to obtain a satisfactory GC calibration. This is a far too complex and cumbersome process; thus a GC/MS analyzer was used for compound identification in the range of C₅ - C₃₄.

The FTS products under both conventional gas phase and supercritical phase were found to be qualitatively similar (but not quantitatively, as will be discussed later). The main difference between the GC product distributions under the two reaction mediums is the dilution of the FTS products in the SCF-FTS, as illustrated and compared in Figure 4 and Figure 5. This dilution complicates the identification process since hydrocarbons have a smaller relative peak area and are more difficult to detect when the solvent/syngas ratio is increased. Thus, gas phase FTS samples were primarily used for the product identification and then the product identification was validated for the SCF-FTS samples and found to be in good agreement.

For the identification of C₁ - C₄ hydrocarbons and permanent gases (CO, CO₂, H₂, Ar) a number of standard gas mixtures with known molar percentages were used. The standard gas mixtures included common FTS products within the (C₁ - C₄) range. These hydrocarbons include methane, ethylene, ethane, propylene, propane, 1-butene, n-butane, iso-butane, butylene, butane, as well as H₂ and CO₂. In addition, different solutions containing known concentrations of (C₁ - C₆) normal alcohols (methanol, ethanol, propanol, butanol, pentanol and hexanol) were injected into the on-line GC for identifying the alcohol peaks. After that, a calibration table was prepared by entering the retention time for each identified species into the GC data collection software. The identification of permanent gases is shown in Figure 6 and sample peak identification for C₄ hydrocarbon group with species fingerprints is shown in Figure 7. As illustrated in Figure 6, for the TCD 1 signal (channel 2), helium was used as the carrier gas. Thus, any gas slower than C₂ was back-flushed. The peak order obtained from the detector as shown in Figure 6 started with H₂ (small peak), Argon and O₂ (they do not separate), N₂, CH₄, CO, ethylene, ethane and CO₂.

The identification of C₅-C₃₄ hydrocarbon compounds were carried out off-line by injecting liquid FTS product samples into the GC/MS (Figure 8). The off-line GC/MS analysis of liquid products provided the full product distribution upon accurate identification of all hydrocarbons between C₅ and C₃₄. Most hydrocarbons (isomers, olefins and oxygenates) between C₅ and C₂₅ were easily identified with high accuracy by the GC/MS. However, higher molecular weight branched isomers and high molecular weight olefins were more problematic to identify. At that point the remaining peaks were identified via superposition since FTS products were observed to follow a similar pattern for each hydrocarbon group. Huff et al. [25] used a similar GC/MS and superposition technique for identifying FTS products. However, they limited their GC/MS product identification and verification to major hydrocarbons and oxygenates within C₁ - C₁₂ hydrocarbons. After that, they used superposition for identifying the remaining peaks.

For the product distribution and selectivity studies, FTS products were divided into five groups (alcohols, isomers, α-olefins, normal paraffins and alkenes). As previously mentioned, the FTS products followed a similar pattern for each hydrocarbon group (C_X), where it begins with (C_X) isomers (increasing in number as X increases). The peaks follow the aforementioned C_X of α-olefins (from 1 - 3), the normal (C_X) alkane (normally the peak with the highest area, which is common for cobalt-based catalyst). After the normal alkane peak comes, the cis and then the trans (C_X) alkenes peaks, as demonstrated in Figure 9 and Figure 10. On the other hand, the location of (C_X-3) alcohol peaks changed as the carbon number increased. For lighter alcohols (C₃ - C₇), the alcohol

peak was located before the C_X isomers, for (C₈ - C₁₃) alcohols, the alcohol peaks were located between C_X isomers and for (C₁₄ - C₂₀) alcohols, the alcohol peaks were found after the isomers and before the α-olefins peaks. It is also noticeable that, α-olefins were not detectable by the Agilent GC/MS beyond (C₁₇), alcohols beyond 1-eicosanol (C₂₀ normal alcohol) were not detectable, while cis and trans alkenes were not detectable beyond (C₂₂). All observations mentioned earlier agree well with the profile of FTS hydrocarbon product distribution obtained from cobalt-based catalysts tested under similar conditions [21] [24] . A sample of compounds (C₇ range) identified using the GC/MS is shown in Table 1. The sample retention time on both FID and MS were identified while the molecule identity was determined utilizing the MS (database), as shown in Figure 11. Subsequently, the retention time for each compound was inserted manually into the GC software (Enhanced Data Analysis), for automatic report generation with defined components.

During the off-line GC/MS FTS product identification minor aromatic, cyclic and acidic compounds were identified in the organic phase. This is expected since these compounds have higher affinity and would be more dissolved in the aqueous phase. Since these compounds were not prevalent in the FTS product and the error resulting by ignoring them would be minimal, they were not included in the FTS product distribution. For instance, the combined total weight percent of aromatics, acids and esters in the organic FTS phase was estimated by Hackett et al. [26] to be less than 4 wt%. While, acids and esters constituted about 8 wt% of the aqueous FTS samples as estimated by Anderson and White [27] .

In this section, a comparison study between the conventional gas phase FTS and the SCF-FTS was conducted to

validate the FTS product identification and analysis process. This was carried by investigating the enhancements in terms of activity, selectivity and chain growth probability (α-value). The reactor was operated at 20 bar and 240˚C for the gas phase FTS. On the other hand, for the SCF-FTS the supercritical solvent (n-hexane) was introduced at 240˚C and the total reactor pressure was maintained at 65 bar. In both cases, the syngas partial pressure was kept at 20 bar with syngas with H₂:CO of 2:1 feed ratio.

The on-line and off-line GC FTS product analysis confirmed a yield of a complex mixture of C₁ to C₃₄ hydrocarbons and C₁ to C₁₇ oxygenates, in addition to H₂O, CO₂ and unreacted syngas. The predominant FTS products were linear alkanes (n-paraffins), while branched (isomers), oxygenates (alcohols), α-olefins, and alkenes (cis and trans) hydrocarbons were produced to a lesser extent, as previously shown in Figures 8-10; this product distribution is typical for cobalt-based catalysts.

Starting with Figure 12, which compares the CO conversion percent under the two reaction media, it was noticed that CO conversion increased by approximately 14% upon the introduction of supercritical hexane. In literature [18] [28] [29] , the experimental data concerning CO conversion under the different reaction media is contradictory. On one hand, Fujimoto et al. [18] found that the conversion under supercritical condition was to some extent lower than that in the gas phase. They attributed this to the difference in syngas diffusion rate in the different reaction media where the diffusion rate in supercritical fluids is slower than that in the gas phase. On the other hand, Huang and Roberts [30] showed that conversion in supercritical hexane is higher (70%) than that in the gas phase (60%).

The overall hydrocarbon product distributions from C₁ - C₃₂ under the previously mentioned conditions were compared as shown in Figure 13. The overall product distribution figure elucidates the shift towards the production of heavier hydrocarbons under the SCF reaction medium. For instance, middle distillates (C₁₁ - C₂₂) constituted about 67% of the normalized product weight under the SCF-FTS compared to 44% in the gas phase FTS. In addition, a substantial drop in methane selectivity was observed when shifting the FTS reaction medium from gas phase to SCF. A number of studies [18] [28] [31] [32] also reported a suppression of methane formation when applying SCF as a reaction media for the FTS.

Higher α-value indicates higher selectivity towards heavier hydrocarbon production and vice versa. Figure 14 shows the ASF plots of the FTS under either reaction media (supercritical hexane and gas phase). From the ASF plots, it was first noticed that lower selectivities towards methane and light hydrocarbons were achieved under supercritical conditions, while the selectivity towards middle distillates and heavy hydrocarbons was increased (this is in agreement with the previous discussion). The improved product selectivities upon the introduction of

supercritical hexane resulted in a significant increase in chain growth probability from 0.76 to 0.85. Similar incensement were observed in previous literature [17] [24] [33] [34] under supercritical conditions.

The aforementioned enhancements in CO conversion, reduced selectivity towards light hydrocarbons and increased selectivity towards heavy hydrocarbons is explained in literature and is discussed briefly in this paper. The increased conversion is attributed to the increased reactants accessibility to the active sites after the in-situ wax extraction from the catalyst pores by the SCF [1] [16] [17] [28] [31] [35] -[37] . The reduction in methane and light end fuel cuts is attributed to the liquid-like heat transfer properties in the SCF reaction medium that results in better heat removal from the catalyst bed. Thus, it minimizes hot spots formation and reduces the rate of cracking reactions [16] [28] [31] [32] [38] . As for the observed enhancement in chain growth probability and the selectivity towards heavy hydrocarbons, this is explained by the liquid-like solubility of SCF that result in higher diffusivities and desorption rates of heavy hydrocarbons (wax extraction) from the catalyst pores. Thus, freeing more actives sites for primary products to incorporate in the chain growth process [1] [16] [17] [28] [31] [35] [37] . Moreover, SCF are capable of extracting olefins prior to undergoing hydrogenation reactions (termination of the chain growth), which increase the chance of their readsorption and incorporation in the chain growth, thus yielding heavier FTS products [1] [16] [17] [31] [33] -[35] .

During the off-line FTS product peak identification using the (GC/FID/MS) system, considerable amounts of several alcohols were identified in the organic FTS product samples. This is uncommon for cobalt-based low- temperature FTS; thus, further investigation was required.

A typical liquid sample of the gas phase FTS (240˚C, 20 bar, 65 nmL/min syngas flow with 2:1 H₂:CO molar feed ratio) was collected from the bottom of the cold trap. Then the organic phase from that sample was injected into the GC/MS for analysis. The alcohols identified within the sample’s organic phase are given in Table 2.

To further investigate these results, the aqueous phase from the same sample was injected into the GC/MS, where the resulting MS spectrum is shown in Figure 15, labelled with the identities of the peaks, as determined by the MS. The GC/MS analysis of the aqueous phase clearly shows that C₁ - C₆ alcohols were present in the FTS product.

As a double check and to confirm the MS results, a reference sample was prepared by mixing a small amount (~5 μL) of a mixture of C₁ - C₆ alcohols diluted with 1.5 mL of n-hexane (Sigma Aldrich, ≥95% purity, with the rest composed of small amounts of C₆ iso and cyclic hydrocarbons). The reference sample was then injected into the GC/MS, and the obtained MS spectrum is shown in Figure 16. By comparing Figure 15 and Figure 16, it can be concluded that the reference sample confirms that the GC/MS correctly identified the alcohol peaks.

The FTS produces a wide range of products including n-paraffins, olefins oxygenates and aromatics. Pei et al. [39] recently studied the promotional effect of silica over activated carbon supported cobalt catalyst for the production of higher chain alcohols. They reported that sufficient amounts of SiO₂ addition favored the formation of higher molecular mass alcohols (C₆ - C₁₈) by suppressing cobalt reduction and therefore producing Co(II) species and thus facilitating CO insertion. The same group previously also demonstrated that activated carbon supported cobalt catalysts are able to produce linear C₁ - C₁₈ alcohols from syngas via the FTS reaction due to the synergy between active Co₂C sites formed in-situ during the reaction and active cobalt sites [40] . Also, another experimental work [41] reported CO conversion of 30% and high alcohol selectivity of 20% when cobalt

molybdenum based catalyst is used under the operating conditions of 307˚C, 75 bar and syngas ratio H₂/CO = 2.

From the above discussion, it can be seen that while iron based catalyst are known to be selective towards alcohol production, cobalt based catalyst can also produce alcohols if the catalyst and reactions conditions were adjusted. Nevertheless, this finding requires further detailed investigation, which is currently being performed by the research team.