Investigation on the simultaneous removal of COS, CS2 and O2 from coke oven gas by hydrogenation on a Pd/Al2O3 catalyst

The present study deals with the processing of coke oven gas mainly composed of H2, CH4, N2 and CO to provide a feedstock for the synthesis of base chemicals. In this respect, the particular focus of this work is the simultaneous reduction of critical trace components like COS, CS2 and O2 by catalytic reaction with H2. The investigations were performed in synthetic coke oven exhaust using a Pd/Al2O3 catalyst. The results of the hydrogenation tests showed complete conversion of COS, CS2 and O2 at 200 °C and above with selective formation of H2S. However, below 200 °C the conversion of O2 was markedly reduced and CH3SH appeared as a by-product. Mechanistic studies were performed by in situ diffuse reflectance infrared Fourier transform spectroscopy coupled with mass spectrometry. These investigations demonstrated dissociative adsorption of COS on the catalyst at 150 °C resulting in the formation of bridged CO adsorbates and probably elemental sulfur. It is assumed that these species predominate the active Pd surface under reaction conditions. Consequently, the adsorption of O2 and the reaction to H2O is suppressed thus substantiating the decrease in performance at low temperatures. However, increasing the temperature to 200 °C and above leads to desorption of CO and sulfur compounds restoring the efficiency of the catalyst.


Introduction
Deep desulfurization is widely used for the removal of sulfur from natural gas as well as refined petroleum products, for instance gasoline and diesel fuels. The reduction of sulfur is primarily required to avoid poisoning of downstream processes such as fuel cells or depollution catalysts of combustion engines. Although well-established in petrochemical industry, deep desulfurization is still a matter of R&D, particularly to meet future legislative regulations addressing the sulfur fraction of fuels. [1][2][3][4][5][6] Additionally, the reduction of sulfur is also considered for coke oven gases aiming sulfur contents below 10 mol ppm, whereas up to now no cost-efficient process exists. Coke oven gas is a by-product of the steel production and mainly consists of 60 vol% H 2 , 20 vol% CH 4 and 6 vol% CO. Thus, the exhaust of coke ovens shows promising potential as a feedstock for basic chemicals, e.g. methanol. 7 However, the exhaust also contains critical sulfur components such as COS, CS 2 and mercaptans representing strong catalyst poisons, especially for Ni and Cu catalysts used in steam reforming and methanol synthesis, respectively. 7,8 Therefore, deep desulfurization is required, which is mostly realized by hydrodesulfurization (HDS). HDS implies hydrogenolysis of organosulfur compounds on CoMo or NiMo catalysts yielding sulfur-free hydrocarbons and H 2 S. The H 2 S produced can subsequently be removed by ZnO adsorbers or scrubbing processes. [9][10][11] As a result, HDS leads to drastic decrease in sulfur proportions down to 15-30 mass ppm as required for ultra-low sulfur fuels for instance. 12 Furthermore, for deep desulfurization of diesel fuels, noble metal catalysts based on platinum and palladium were investigated with promising results. [13][14][15][16][17] But, only little is known on the HDS performance of noble metals in gaseous streams containing COS or mercaptans. Additionally, noble metals provide advantageous features, which are relevant for their use in gas oven exhaust, such as (i) enhancement of the removal of oxygen by reaction with hydrogen to water, since it can be harmful in downstream processes (e.g. due to safety issues or corrosion effects), (ii) suppression of coke formation by hydrogenation of olefins and (iii) their superior HDS activity as referred to traditional CoMo catalysts. 14 However, there is a debate in the literature on the stability of noble metal catalysts towards sulfur poisoning. [13][14][15][16][17][18] In this respect, some papers investigated the interaction of sulfur compounds with Pt and Pd catalysts. Bartholomew et al. 19 reported on the dissociative adsorption of COS, CS 2 and methyl mercaptan (CH 3 SH) with strong sticking of the sulfur on the noble metals. Moreover, Reinhoudt et al. 14 studied the nature of the active sites on alumina-silica-supported Pt suggesting that the desulfurization takes place on vacancies present in small platinum particles. High activity was also attributed to the support acidity affecting the electron deficiency and weakening the noble metal-sulfur bond. 20 Additionally, Pt and Pd sulfides originated from HDS can effectively be regenerated by reaction with H 2 above 463 K. 17,21,22 With the above background the present study is the first that deals with the simultaneous removal of organosulfur compounds and O 2 from coke oven gas by hydrogenation on a commercially available Pd/Al 2 O 3 catalyst. The investigations were made in a synthetic exhaust using a laboratory test bench with fixed bed reactor. COS and CS 2 were taken as model components showing highest importance among the sulfur species in practice. 20 In addition to the evaluation of the catalytic performance, in situ diffuse reflectance infrared Fourier transform spectroscopy coupled with mass spectrometry was conducted to obtain some mechanistic insights into the complex reaction network, particularly at low temperatures.

Catalytic studies
The performance of the commercial Pd/Al 2 O 3 catalyst towards hydrogenation of COS and CS 2 combined with reduction of O 2 was studied in a coke oven model exhaust by employing a laboratory bench equipped with fixed bed reactor. The catalyst implied a Pd load of 0.3 wt% and was used in form of spheres with a diameter of 3 mm. According to Mears and Weisz-Prater criteria it was deduced that neither film nor pore diffusion limit the chemical kinetics. 23 The sample (68.6 g) was fixed in the stainless steel tube reactor (i.d. 30 mm) by quartz wool and a sieve resulting in a bed height of 15 cm. In front of the packed bed, glass beads (2.85-3.45 mm, Carl Roth) were placed to improve heating of the inlet gas. The temperature was recorded by two Pt100 shunts (type 27-7125, Bartec) directly located in front of and behind the catalyst bed. The catalyst was initially pre-treated at 350°C for 1 h with 50% H 2 in N 2 . Before the tests, the sample was heated up in N 2 flow (5.2 LN/min) to the specific test temperature and then the feed was adjusted until steady state conditions were reached. Each test point was held for 90 min. After this, temperature (120-400°C) or feed composition was varied. The model feed consisted of 60-120 mol ppm COS, 55-120 mol ppm CS 2 , 50 vol% H 2 , 20 vol% CH 4 , 1 vol% O 2 and N 2 as balance. The reactor pressure was regulated with a needle valve behind the reactor to 1.2 bara, whereas the total gas flow was kept at 5.2 LN/min corresponding to a space velocity of 3000 h −1 . All gases were purchased from Linde. Pure H 2 (5.0), CH 4 (2.5) and N 2 (5.0) were taken, while the sulfur compounds as well as O 2 were used as mixtures with N 2 . Each volume flow rate was checked by independent mass flow controllers (Brooks Instruments). All electronic devices were controlled by Freelance 2013 software from ABB.
The gas analysis was carried out by gas chromatography (Compact GC 4.0, Global Analyser Solutions) using a pulsed flame photometric detector (PFPD) for COS, CS 2 and H 2 S and a thermal conductivity detector (TCD) for O 2 . Simultaneous measurement of organic sulfur components, H 2 S and O 2 was not possible. Before entering the GC, the gas stream passed a condensate trap to reduce the amount of water in the columns.
Furthermore, a pressure regulator was implemented to ensure a constant pressure at the inlet of the GC. With this set-up, reproducible results were achieved.

In situ DRIFTS and MS studies
The reaction of COS, O 2 and H 2 on the Pd/Al 2 O 3 catalyst and the bare Al 2 O 3 support was investigated by diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS). The analyses were performed on a FTIR Spectrometer Tensor 27 (Bruker) equipped with MCT detector (Internal) and Praying Mantis DRIFTS optics (Harrick), while gas phase was simultaneously checked by mass spectrometry (Cirrus 2, MKS Instruments). In each experiment, fresh catalyst sample was placed into the cell and was subsequently pre-reduced with a mixture of N 2 and H 2 (50 vol% each) at 300°C for 15 min following the pre-treatment in the catalytic studies. DRIFTS data were recorded by accumulating 200 scans resulting in a time resolution of ca. 5 min. Background spectra were taken at 40°C and 150°C and under flowing N 2 (500 mL min −1 ). Subsequently, the sample was exposed to COS (200 mol ppm) at 40 and 150°C to investigate the resulting adsorbates. For temporal separation of the overall reaction, H 2 and O 2 were successively dosed after the pre-adsorption of COS at 150°C. Additionally, the simultaneous conversion of the reactive gases was checked at the same temperature by exposing the catalyst and bare support to a mixture of 200 mol ppm COS, 60 vol% H 2 , 1 vol% O 2 . N 2 was used as balance. After the adsorption and reaction steps, the sample was flushed with 500 mL min −1 N 2 for 25 min to differentiate gas-phase and surface species, while each exposure was made until equilibrium or steady state. The overall gas flow was always kept at 500 mL min −1 . The DRIFT spectra are presented in form of the Kubelka Munk function defined as FĲR) = (1 − R) 2 /(2R); R is the quotient of sample and background scans. At temperatures over 200°C the conversion of organosulfur compounds (60 mol ppm COS and 55 mol ppm CS 2 ) as well as 1 vol% O 2 was complete and stable for the established time on stream of 90 min (Fig. 1). Additionally, the formation of 140 mol ppm H 2 S was observed between 200 and 400°C, which is the thermodynamically favored product under these conditions. This proportion is in line with the stoichiometry of the conversion of COS and CS 2 according to eqn (1) and (2). Note that the discrepancy to the theoretical yield (170 mol ppm) is referred to influence of the condensate trap in front of the GC.

Results and discussion
Furthermore, when decreasing the temperature from 200 to 150°C a strongly unsteady behavior of the catalyst appeared implying continuous decrease of O 2 conversion, which remained almost constant after 200 min at 150°C (Fig. 2). The analysis of the sulfur components at 150°C at the end of the experiment showed still high conversion of COS and CS 2 .
However, in addition to H 2 S, CH 3 SH was also formed (approx. 30 ppm), whereas both compounds totally amounted to 120 mol ppm ( Fig. 3 and Table 1). For CoMo catalysts, Clark et al. 24 supposed that CH 3 SH is originated from the hydrogenation of the C-S bond in line with the thermodynamics (eqn (4)). Furthermore, it cannot be excluded that some adsorbed sulfur still remained on the catalyst surface 25 and that another sulfur compound was produced, which could not be detected by PFPD. For example, the production of SO 2 is basically feasible when the reaction of S with O 2 is faster than that with H 2 . Other organic sulfur components were not formed, they would have been detected by PFPD. Moreover, the catalyst was exposed to a model coke oven gas with higher sulfur mole fractions, i.e. each 125 mol ppm for COS and CS 2 , to evaluate the effect of sulfur on the catalytic performance. At 200°C the conversion of O 2 and organic sulfur is complete. Subsequently, the temperature was lowered to 120°C. This leads to a drop of oxygen conversion below 30% within less than 1 h (Fig. 4). Subsequently, it was heated again to 200°C resulting in an immediate rise of the conversion to over 95%. Entire removal of O 2 was achieved by further increasing the temperature to 250°C, whereas COS and CS 2 were also completely converted, i.e. the same performance was obtained than before ( Fig. 2). Measuring the concentration of the sulfur components at the outlet showed the initially measured full conversions and solely a H 2 S peak in the product stream; note that quantitative detection of H 2 S was not available. This shows that sulfur, which inhibits the O 2 conversion below 150°C, is released from the catalyst and therefore it can be regenerated by an increase of temperature.
Since the conversion of O 2 to H 2 O is inhibited below 150°C , while the sulfur compounds are still efficiently converted (<3 ppm), it can be concluded that COS, CS 2 and H 2 predominately chemisorb on the active sites of the catalyst. Furthermore, the higher fractions of the sulfur components obviously favor their adsorption as indicated by the stronger inhibition of the O 2 conversion (Fig. 4). This conclusion is in line with Bartholomew et al., 19 who also observed preferential adsorption of sulfur compounds compared to O 2 . Additionally, it is well known in the literature that sulfur forms strong bonds with metals of group 10 (Ni, Pd, Pt) due to unshared electrons. Therefore, it is assumed that in an   View Article Online initial step COS adsorbs on at least one Pd atom followed by dissociation into CO and S adsorbates as described by eqn (5) and (6). 17,19,[26][27][28][29] Obviously, the coverage of the Pd sites by S grows at lower temperatures (120°C) thus suppressing the H 2 -O 2 conversion as found in the experiments ( Fig. 2 and 4). However, when re-increasing the temperature (>150°C) the S adsorbates increasingly react with H 2 to yield Pd and H 2 S, 21 which was detected in the gas-phase (eqn (7)). Consequently, the active Pd sites are regenerated and are again available in sufficient abundance for the conversion of O 2 as evidenced in Fig. 4. 14,17,28 Since this reaction is highly exothermic (eqn (3)), it probably leads to a self-acceleration of the reaction rate. This explains the rapid rise of the conversion to over 95% when heating up from 120°C. Above that, the conversion rises slowly, but can be accelerated by increasing the temperature.
On the other hand, the complete inhibition of the hydrogenoxygen reaction upon cooling from 200 to 120°C is also fast within a narrow temperature window. Hence, it can be concluded that at temperatures below 200°C the formation of elemental sulfur from COS or CS 2 decomposition is faster than its reaction with hydrogen. The continuous chemisorption of sulfur on active Pd sites leads to decreasing levels of oxygen conversion which initiates further cooling of the catalyst internally and therefore increases chemisorption of sites with sulfur even more.
As shown by other groups, the supports' acidity has a positive effect on the regeneration of the catalyst: firstly, the number of active sites is increased because of stabilization of small noble metal particles. 17,28 Secondly, acid support material weakens the Pt-S bond and therefore promotes the formation vacant sites. 30,31 In this series of experiments no other side reactions were observed. However, it was previously described that metals of group 10 can catalyze the methanation reaction in this temperature range. 33

Mechanistic studies
In situ DRIFTS measurements were conducted with both Pd/ Al 2 O 3 as well as Al 2 O 3 to evaluate the role of sulfur in the inhibition of the O 2 conversion at low temperatures. In these studies, COS was taken, as it represents the most prominent organic sulfur compound present in real coke oven gas. 32 The DRIFTS studies were conducted by subsequently exposing the sample to O 2 , H 2 and COS as well as by simultaneously supplying the gases.
This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

View Article Online
HS species. 37,38 Since reference tests with pure Al 2 O 3 provided the same DRIFTS bands, all features are assigned to the chemisorption of COS on the Al 2 O 3 support. OH stretching bands of the support appear and increase over time in the range from 2700 to 3700 cm −1 . Subsequent flushing with nitrogen leads to disappearance of the band referred to molecularly adsorbed COS and to a slight decrease in the feature of HCO 3 − associated with desorption. During the entire time a small and decreasing amount of CO 2 is detected by MS in the outlet stream.
In the COS exposure at the reaction temperature of 150°C (Fig. 6) (Fig. 5). The decreasing peak at 150°C (Fig. 6) shows that HCO 3 − is decomposed at higher temperatures and CO 2 is released to the gas phase. Contrary, SH − remains on the catalyst surface, which is shown by the band located at 2586 cm −1 (ν(HS)), (not shown in Fig. 6). In contrast to 40°C, a double band with maxima at 1890 and 1810 cm −1 is obtained upon COS exposure at 150°C. According to the literature, 42,43 the band at 1890 cm −1 refers to bridged Pd x -CO (with x = 2-3) moieties, whereas the feature at 1810 cm −1 is likely due to threefold hollow binding CO (Pd 3 -CO). 42,[44][45][46][47] The presence of CO adsorbates is ascribed to the dissociation of COS on Pd sites in line with the literature. 39 However, the proof of resulting Pd-S species is difficult, as the corresponding stretching vibration is expected below 600 cm −1 , 48 which is not accessible by the DRIFTS cell used.
Compared to 40°C a higher partial pressure of CO 2 is detected in the outlet stream at 150°C decreasing over time (Fig. 7). This effect is referred to the above mentioned HCO 3 − decomposition. Additionally, it is likely that the adsorbed CO reacts with surface oxygen of the support to yield CO 2 which is released in the gas phase. When the excessive oxygen is consumed, CO remains on Pd as demonstrated by the growing DRIFTS bands between 1800 and 2000 cm −1 (Fig. 6). This leads to increased absorption of COS on the support (growing bands at 1578 and 1330 cm −1 in Fig. 6). Flushing the catalyst afterwards with nitrogen leads stable adsorbates. However, it has little influence on the HCO 3 − and HSCO 2 − deposits. The weakly adsorbed COS is removed immediately from the catalyst surface upon N 2 flushing. The COS adsorption on bare Al 2 O 3 performed at 150°C provides no CO bands indicating that Pd is necessary for COS dissociation. Besides that, all the other features appear which confirms that they are surface species formed on Al 2 O 3 and not on the active sites of the catalyst (Fig. 8).
When supplying H 2 to the catalyst at 150°C after COS adsorption and N 2 flushing all the DRIFTS bands decrease and disappear with the exception of the feature of the HCO 3 − species (Fig. 9). Considering the outlet stream, the MS trace showed strong release of water upon addition of H 2 referred to the reaction with some surface oxygen species of the support. 49 Simultaneously, the MS signal with m/z = 32 increased sharply followed by slow decay (inlay in Fig. 9). Since no oxygen has been added, it is presumably 32 S derived from small amounts of H 2 S which were removed from the catalyst by reduction of sulfur species. This is confirmed by small and noisy signals of mass 33 and 34 that were also detected. Since the catalyst reveals little load of Pd (0.3 wt%), Fig. 6 DRIFT spectra of the Pd/Al 2 O 3 catalyst exposed to COS at 150°C over time and upon subsequent flushing with N 2 (500 ml min −1 , 500 ppm COS, N 2 balance).

View Article Online
the amount of H 2 S is expected to be very small and hence hard to detect. During the addition of H 2 the bands in the region above 3000 cm −1 correlated to stretching vibrations of OH groups increase in intensity (Fig. 9), which is associated with the adsorption of H 2 O. Additionally, the formation of H 2 O enhances the hydrolysis of COS leading to the reaction of SH − and H 2 S, respectively. This is confirmed by detection of mass 32 that appears due to ionization and decomposition of COS into CO and S inside the mass spectrometer. Mass 60, that would indicate appearance of COS in the outlet stream was not detected. As shown in Fig. 9, the addition of H 2 to the catalyst led to further but not complete reduction of the CO peak(s). It is presumed that it is thermally desorbed due to the exothermic reaction of H 2 with surface oxygen species (eqn (3)). Flushing with N 2 had no effect on the CO bands.
Continuing with 1 vol% O 2 after exposure of the catalyst with COS and H 2 induced a slow but almost complete removal of CO (Fig. 10). A small band of threefold bound CO was still observed at 1807 cm −1 after 45 min. Additionally, MS indicates signals with m/z ratios of 33 and 34, which are clearly more pronounced as compared to the natural abundance of 16  ) and 48 (SO + ) associated with SO 2 (inlay in Fig. 10). Hence, residues of sulfur were still present on the catalyst surface after H 2 treatment. Since SO 2 is produced on Pd/ Al 2 O 3 and Al 2 O 3 , it can be assumed that remaining sulfur existed on the Al 2 O 3 support. 50,51 It was shown by Hoyos et al. 52 that H 2 S adsorbed on Al 2 O 3 can be completely     Fig. 10). This indicates that adsorbed hydrogen is still present on the Pd sites even after N 2 flushing which reacts with O 2 to form H 2 O. 54 3.2.2. Simultaneous exposure to COS, H 2 and O 2 to Pd/ Al 2 O 3 or Al 2 O 3 . In another DRIFTS investigation, COS, H 2 and O 2 were simultaneously fed to mimic practical reaction conditions, whereas the temperature of the catalyst was set to 150°C (Fig. 11). This approach led to an initial but weak formation of HCO 3 − within the first 5 min, continued by a decrease over the remaining 120 min until only a small peak at 1658 cm −1 remained. The previously identified CO peak on Pd (bridged Pd x -CO (with x = 2-3) moieties) appeared as a shoulder at around 1866 cm −1 . Reference experiments made with the bare Al 2 O 3 support basically shows very similar DRIFT spectra with exception of the CO shoulder, which again only appeared with Pd present (Fig. 12).
Considering that the catalyst is plain and active sites are free until the gas mixture is exposed to it, it can be assumed that initially all substances bind to it. Hence, besides the formation of CO-Pd and S-Pd by dissociation of COS, it is likely that oxygen and hydrogen also chemisorb to Pd. 54 Consequently, CO can be oxidized in a Langmuir-Hinshelwood mechanism to CO 2 and in addition H 2 O is formed. The initial increase of HCO 3 − bands indicate an adsorption of CO 2 on Al 2 O 3 . At the same time, CO 2 is detected in the MS with a decreasing trend that coincides with the removal of HCO 3 − deposits. Therefore, the adsorption is an intermediate phenomenon. In addition, the MS data show a peak of H 2 O in the beginning of the experiment, hence a release into the gas phase, that is more pronounced with Pd present. This is plausible, since Pd enhances the oxygen-hydrogen reaction.
At 150°C CO is favorably adsorbed over O 2 , so in the steady state Pd is preferably covered with CO and S, which inhibits O 2 adsorption and hence CO 2 formation. As investigated by Klauck et al. 55 and also shown in this work, for the same reason, the oxygen-hydrogen reaction is inhibited.
Since CO-Pd is formed by dissociation of COS, it can be once more assumed that in addition S-Pd is formed. While a high concentration of H 2 is present in the gas stream, it can be concluded that this bound sulfur reacts with dissociatively adsorbed H 2 on Pd to H 2 S and is removed from the surface. 17 It was previously described that this reduction of metal sulfides such as Pd-S is catalyzed by the adsorption of H 2 molecules on Pd and their subsequent dissociation into atoms. 17,21 The conversion of COS to H 2 S catalyzed by Pd is confirmed by the MS spectra. A comparison of the outlet streams of simultaneous gas exposure on the catalyst and the support shows that with Pd/Al 2 O 3 the partial pressure of mass 60 (COS) is lower (Fig. 13). In addition, comparing the spectra reveals that the masses 34, 33 and 32 showed steady signals in both cases, but with Pd/Al 2 O 3 the partial pressure is higher compared to the support only. This indicates the formation of H 2 S, HS − and 32 S, the latter derived from splitting of H 2 S in the MS. Fig. 11 DRIFT spectra of Pd/Al 2 O 3 support exposed to synthetic coke oven gas at 150°C over time and upon subsequent flushing with N 2 (500 ml min −1 , 500 ppm COS, 60 vol% H 2 , 1 vol% O 2 , N 2 balance). Fig. 12 DRIFT spectra of Al 2 O 3 support exposed to synthetic coke oven gas at 150°C over time and upon subsequent flushing with N 2 (500 ml min −1 , 500 ppm COS, 60 vol% H 2 , 1 vol% O 2 , N 2 balance).

View Article Online
Moreover, the DRIFT spectra show that CO is bound to Pd over the entire duration. Therefore, it can be concluded that COS adsorption is not inhibited by CO, sulfur or H 2 , hence all four can adsorb on the active sites under these conditions. Furthermore, HCSO 2 − appeared as a shoulder in the DRIFT spectra at 1578 cm −1 only in the beginning and exhibits a lower amount compared to initial COS dosing. This indicates that it is removed by hydrogen as shown before (Fig. 9). Moreover, a small feature at 1307 cm −1 was detected that increased with time and showed a smaller intensity with Pd present. A peak in that range was previously assigned to be physisorbed or chemisorbed SO 2 on Al 2 O 3 . 50,56 This accumulation of sulfates on Al 2 O 3 were smaller when Pd is present. It was explained by Luo et al. 21 that sulfates in proximity to noble metal particles are reduced stepwise to H 2 S by activated H 2 and are subsequently released. This reduction is temperature dependent, since higher temperatures increase the mobility of absorbed sulfates to the region where reduction is possible. Again, small and noisy signals of mass 64 and 48 were detected with both catalyst and support present. This confirms the formation of SO 2 . However, over time the signals decreased, which leads to the assumption that in the steady state H 2 S is the predominant sulfur species that is released into the gas phase. It this assumed that this again originates in the inhibited adsorption and therefore reactivity of oxygen molecules. In addition, COS molecularly adsorbed on Al 2 O 3 was present that was again instantly removed by flushing with N 2 .

Conclusions
The performance of a Pd/Al 2 O 3 catalyst was evaluated for the simultaneous hydrogenation of organosulfur compounds and conversion of O 2 , while some mechanistic investigations were made by DRIFT spectroscopy coupled with MS. The catalyst was tested between 120 and 250°C in a model coke oven gas consisting of H 2 , CH 4 , O 2 , COS and CS 2 . The results showed that at 200°C and above the organosulfur compounds are almost completely converted to H 2 S. Simultaneously, O 2 entirely reacted with H 2 to H 2 O. Upon lowering the reactor temperature below a critical point, the O 2 conversion is strongly inhibited, whereas COS and CS 2 are still completely converted. Also, the selectivity towards H 2 S is decreased and CH 3 SH appears as additional product. This altered performance is rapidly reversed by increasing the temperature just to 200 and 250°C, respectively.
The mechanistic examinations of the Pd/Al 2 O 3 catalyst as well as the bare Al 2 O 3 support were conducted by (I) successive gas dosage of COS, H 2 and O 2 and (II) simultaneous exposure to all the reactive gases at 150°C. It was shown that COS dissociates on the Pd sites to form CO and S adsorbates. These surface species located on the Pd sites prevent the adsorption of O 2 and they therefore strongly inhibit the oxygen-hydrogen reaction as observed in the catalytic performance tests. In addition, it was demonstrated that CO was removed from Pd due to displacement at 150°C and it is assumed that desorption is accelerated at higher temperatures. Furthermore, the presence of H 2 leads to a removal of bound sulfur to H 2 S.
Moreover, it was shown that COS also adsorbs on the Al 2 O 3 support resulting in the formation of HSCO 2 − . Subsequently, HSCO 2 − reacts with OH − surface groups to CO 2 and H 2 S, which can re-adsorb to yield HCO − 3 species. However, HSCO 2 − plays a rather minor role in the conversion of COS to H 2 S. To conclude, the strong blockage of the catalytically active Pd sites by CO and S inhibits side reactions. It is reversible by two effects: firstly, the increase of temperature that leads to the desorption of CO and secondly, by the presence of H 2 in the feed stream that removes bound sulfur to form H 2 S.
Finally, this study shows that commercially available Pd/ Al 2 O 3 catalysts can be effectively taken for the removal of organosulfur compounds and oxygen from coke oven model gas. Also, the high amount of H 2 simply facilitates the catalyst regeneration. Consequently, HDS by Pd/Al 2 O 3 catalysts is considered as a suitable method for the deep desulfurization of coke oven gas to supply synthesis gas for the chemical industry.

Conflicts of interest
There are no conflicts to declare.