Ultraclean gas for downstream application
Campaign SXB20/07 was the first fully coupled run of the SXB gasification facility with final gas cleaning and synthesis. The SXB20/07 UC5 runtime was shorter than anticipated due to the long pre-heating requirement. To compensate for this, bed material changes were not performed for campaign SXB20/11, and a total runtime of 105 h with the same bed materials was achieved. The slipstream operation of the downstream processes was successful, and stable pressure levels in the atmospheric side were achieved. Detailed process measurements from select setpoints are presented in the Appendix Table 6. The gas purity was continuously monitored after the final gas cleaning process. The average cleaned gas composition is presented in Table 3 for select setpoints, along with the gas composition measured before final gas cleaning.
The major impurities that were monitored included the N-group compounds NH3 and HCN, the S-group compounds H2S and COS, and hydrocarbons including benzene and residual tars. HCl and SO2 were also analyzed during select setpoints, with none detected. The ammonia concentration in the gas varied significantly, between 350 ppm (wood) and 3300 ppm (sunflower), but this did not affect the removal performance, with ammonia not detected in any of the purified FTIR gas samples. HCN concentration in the gas after the reformer was in the single-digit parts per million range; again, sunflower husk feedstock showed the highest HCN concentration. HCN was also completely removed. Up to 1300 ppm benzene and some tars were detected in the sunflower-based syngas, which is an order of magnitude higher than in the previously reported BFB-UC5 campaigns. Benzene or tars were not detected after the UC5 process. The post-reformer syngas sulfur concentration was consistent with the analyzed feedstock sulfur content, with sunflower depicting the highest concentration, up to 220 ppm H2S and 11 ppm COS. The H2S:COS ratio remained above 10 in the analyzed setpoints. SXB20/07 COS after reformer was not analyzed, but the same feedstock was operated in SXB20/11D with a COS concentration of 6.7 ppm. The SXB20/07 setpoints showed a barely detectable breakthrough of COS, estimated at 0.1 ppm. This was not detected in the SXB20/11 campaign, which could indicate that the reaction temperature of SXB20/07 in WGB1 was insufficient.
Table 4 presents the post-run characterization results of the non-impregnated activated carbons VAC1 and VAC2 from AR.
The fresh carbons exhibited a basic pH of above 10, and the pH of the spent carbons dropped to 8.3–8.8 as a result of species deposition onto the AC surface. The BET surface area decreased, especially for the bed 1 sample. Micropore volume decreased, but mesopore volume remained unchanged for both spent samples. Sulfur analysis depicted significant increase for VAC1 in bed 1, while the bed 2 sample had only slightly increased from fresh base levels. Ultimate analysis of elements C, H, and N did not display significant deviations from fresh samples.
Thermogravimetric analysis was performed in a nitrogen atmosphere for the AR activated carbon samples and reference fresh equivalents. The results are presented in Fig. 2
There were three distinct mass change temperature ranges above baseline, which are grouped according to: 50–105 °C, 105–225 °C, and 200–450 °C. A significant mass change occurred at around 100 °C, which can be attributed to water vaporization. The low temperature range (25–105 °C) accounts for bed 1 around 25% of the excess mass loss, while for bed 2 it exhibits over 40%. A second high excess mass loss occurred at around a 150–175 °C peak, and a third wider range, only occurring for bed 1, at above 250 °C and up to 450 °C. The third range accounted for about 60% of the excess mass loss of the bed 1 sample and can be attributed to elemental sulfur desorption, according to previous studies [12, 13]. Since the equivalent bed 1 sample S-content was determined to be 4.4% and the excess mass loss in the range 200–450 °C ca. 4.1%, the majority of the 105–200 °C mass loss can be attributed to the desorption of compounds other than sulfur or water, most likely benzene and tars. The bed 2 excess mass loss in the 105–225 °C range was significantly higher than for bed 1, indicating higher affinity for benzene adsorption. This could also point toward competing adsorption between benzene and sulfur species, with bed 1 first capturing sulfur species to its micropores to the effect of breakthrough of benzene, which is subsequently captured by bed 2 at a higher capacity. Bed 2 was exposed to a lower on-stream sulfur concentration, which could explain the higher water content in the pores. Also the larger micropore volume of VAC2 likely improves the overall capture capacity over VAC1.
SEM and EDS analysis
The 4 mm diameter AR activated carbons pellet outer surfaces and cross-sections were imaged using SEM and elemental analysis by EDS. Relative to the reference fresh sample, only the sulfur concentration was higher in the spent samples. In the UC5 process, the AR bed 1 surface was exposed to gas with the highest levels of impurities, and was thus of interest for further characterization. Figure 3 displays the VAC1 sample from AR bed 1 after SXB20/11. For comparison, the figure also shows the spent CaAC sample from the previous BFB-UC5 campaign II AR bed 1 surface.
The spent VAC1 sample in Fig. 3b shows that the sulfur was fairly evenly deposited in a radial direction due to the high porosity of the non-impregnated activated carbon. By contrast, the caustic carbon from the previous campaign formed a sharp ring pattern with sulfur depositing on the outer surface, up to 1 mm in depth, due to the presence of caustic impregnate there. The WGB1 zinc oxide and deoxygenation catalyst were imaged and analyzed in a similar manner, shown in Fig. 4.
The ZnO1 zinc oxide adsorbent surface image with EDS sulfur and oxygen overlay (Fig. 4a) depicts two principal macroscale phases, one with significant sulfur content, about 5 atom %, and the other with less than 1 atom %. The sulfur-containing phase displays also more oxygen, about 50 vs. 40 atom %, with less zinc than the other phase. The cross-section of spent ZnO1 sample contained virtually no sulfur; thus, the captured sulfur was limited to the outer surface, depositing in the characteristic uneven pattern depicted in the figure. The surface EDS spectrogram in Fig. 4b of the CuZn1 outer surface in Fig. 4c shows no signs of sulfur or other impurities depositing onto the catalyst. Since the reduced copper surface, like other reduced forms of metal catalysts, is exceptionally receptive to impurities like sulfur or halogens, these results confirm that syngas impurities removal occurred in the prior steps. This result also increases the confidence in the gas analysis results shown earlier; there are no signs of undetected or under-detected inorganic impurities. Previous conclusions on the limited need for the final guard bed GCB2 are also confirmed, effectively allowing its elimination and resulting in process simplification.
Multipoint gas analysis
For the final campaign, SXB20/24, the packed bed masses were reduced to achieve more realistic operating conditions in space velocities that could be employed at upscaled operation.
For multipoint analysis, six sampling points, essentially after each major UC5 unit, were connected to the FPD-GC and FTIR analyzers. Manual switching of sampling points during campaign SXB20/24 setpoints was performed, with at least three samples taken from each sampling point. Sampling order was always in the upstream direction.
Figure 5 presents the multipoint FTIR analysis results from sampling during setpoints B and E2 for ammonia and benzene.
The FTIR analysis of benzene after passing through the acid wash condenser indicates similar concentrations to the offline analysis of sampled ammonia after the reformer due to the low solubility of benzene in water. Benzene in setpoint B was primarily removed by AR bed 1, but in setpoint E, with > 1000 ppm benzene concentration, a breakthrough occurred. The sample in Fig. 5b was taken at the end of setpoint E, and at this stage, > 750 ppm outlet concentration was measured after AR bed 1. The bulk of the benzene was thus removed by AR bed 2. A possible benzene desorption effect in the subsequent setpoints was observed, with higher concentrations measured after bed 1 than the feed benzene concentration. Once the bed was saturated, the weakly bound benzene, according to the analysis results, was seemingly partly released to the gas stream. The pore volume changes in SXB20/07 and 11 for the spent bed 1 and 2 carbons indicate that the benzene was adsorbed to the micropores. Analogous to the findings of Oliver et al. , it shows that benzene vapor adsorption occurred through physisorption and that breakthrough time was dependent on the adsorbent surface area and micropore volume. The equilibrium capacity of benzene and other hydrocarbon impurities, such as toluene, on microporous activated carbons is decent, which increases with higher impurity partial pressure. With commercially available carbons, equilibrium capacities up to between 200 and 350 mg g−1 have been measured [15, 16]. With a bed height of only 2.5 cm, a fast breakthrough is expected as fixed-bed operated benzene capture by physisorption is improved with increased contact time with the adsorbent. The small bed height is also suspected to lead to premature breakthrough due to the mass transfer zone (MTZ) exceeding the bed height . From a process design perspective, the possibility of frequent bed changes and rapid regeneration of the spent adsorbent would be required if high benzene removal efficiency is required. Fortunately, the purity requirement of benzene for catalytic synthesis applications is not expected to be as strict as with many other impurities.
Observation that a higher gas humidity negatively affects benzene removal has been reported . This is in contrast to H2S removal by activated carbons, which benefits from moisture in gas . Activated carbons are inherently hydrophobic, but if the pore size of the activated carbon is sufficiently small, then the adsorption energy increases, and water can be adsorbed. Water adsorbs to the narrow micropores and especially to hydrophilic sites, which blocks benzene physisorption [18, 19]. By tailoring activated carbons for low moisture affinity, benzene removal can be improved. On the other hand, fine-tuning of the porous texture negatively affects the regeneration potential of adsorbents, with the best regeneration results achieved with adsorbents that show a wider pore size distribution. Several methods for regeneration are available, including the most common, thermal regeneration, but pressure methods, solvent extraction, and other methods, such as microwave regeneration, exist [20,21,22]. Thermal regeneration involves an inert gas, steam, or air, with the previous results in  indicating high VOC capturing capacity after several regeneration cycles.
The 300 ppm ammonia in setpoint B was fully removed in the AWC, in which the circulating water pH was fixed at 3. Setpoint E2 contained 2200 ppm NH3, and this was also removed by acid washing. With these results, it appears that for similar ammonia removal performance, even a higher pH could suffice. This would decrease acid consumption which, here, was in the formic acid:NH3 mol ratio range of 5–6, relatively independent of ammonia concentrations. Increasing the pH would, however, increase the absorption of acidic gasses, which would potentially increase the costs of wastewater treatment. The treatment costs can be further decreased by minimizing the amount of wastewater by performing hot syngas water condensation at least partially separated from the acid injection water scrubbing. Choosing a strong acid, such as sulfuric acid, minimizes acid consumption by ensuring full dissociation in water. The product of ammonia removal by aqueous sulfuric acid is ammonium sulfate, which can be precipitated and used as a fertilizer .
For detected sulfur compounds, H2S and COS, the multipoint analysis is depicted in the bar charts in Fig. 6 for different sulfur concentration setpoints B (wood) and E2 (sunflower husk). Sulfur compound concentrations after AWC was considered to equal post-reformer concentration due to their low affinity for absorption in acidic water.
As presented in Fig. 6a, the H2S:COS ratio in the low-sulfur wood pellet syngas shifted to a significantly lower value in the post-reformer syngas. This is evidently due to changes in the catalyst loading of the second reformer bed, since the ratio far exceeds the predicted thermodynamic equilibrium. Before SXB20/24, the previously used combination of nickel and bimetallic precious metal catalyst was replaced by a thin layer of nickel followed by a larger layer of platinum-based methane-reforming catalysts. The COS concentration in setpoint B reached up to 6 ppm, with the majority removed in AR beds 1 and 2. These results establish in addition to the SXB20/07 and 11 post-run bed material characterization results that AC1 in bed 1 removed sulfur compounds, while the primary intention was to remove benzene and residual tars. To promote more selective and effective removal of each species, i.e., with bed 1 for removal of hydrocarbons and bed 2 for removal of sulfur compounds, air/oxygen feeding to the adsorbent reactor should be directed to bed 2. In our previous articles, it was established that H2S removal by the oxidative route in moist gas is limited to low space velocities in its application, while the removal rate significantly improves by the presence of ammonia in syngas [25, 26]. Although a high H2S removal rate was achieved even in the small bed 1, ammonia-enhanced desulfurization unlikely occurred. For H2S removal purposes, the bed size can be further decreased if partial ammonia breakthrough is allowed at the AWC. The most effective way to achieve this is to bypass a small slipstream of the raw syngas past the AWC.
The Al1 aluminum catalyst in WGB1 bed 1 did not achieve particularly high COS conversion. COS was not detected after the full final gas cleaning process and, thus, the remaining trace amount COS (< 0.5 ppm) was likely adsorbed by the carbons in CGB2. Setpoint E, with sunflower husk feedstock, displayed remarkably high COS concentrations, especially in E2, estimated at around 100 ppm. COS is less acidic and less polar than H2S, and therefore harder to remove with adsorbents [27, 28]. Surprisingly, COS was removed by AR bed 2 to levels around 7 ppm throughout the 15-h long setpoint E. Subsequently, an average of 70% removal of the remaining COS in WGB1 bed 1 was achieved, though no formation of H2S was detected. In setpoint E2, a 0.7 ppm COS breakthrough after the full process was detected. There are very few investigations of carbonyl sulfide adsorption on activated carbons since more abundant sulfur species, such as H2S and SO2, have been of primary interest. Fixed-bed breakthrough tests by Sattler et al.  illustrated that the COS adsorption rate is slow with a low uptake capacity, at < 5 mg g−1. In a gas mixture also containing H2S, the COS partially competed with H2S in adsorption, leading to significantly lower capacities (0.5 mg g−1 for COS, 4 mg g−1 for H2S). In addition, higher humidity negatively affected COS capture capacity, and ammonia did not affect performance. Reports by Qiu et al.  and Wang et al.  on modified activated carbons depicted higher capture capacity using, for example, metal or KOH impregnate. COS uptake was improved by higher O2 concentration and, thus, an oxidative reaction mechanism was proposed . Since the high COS concentration setpoint was operated in this study for only a short period, the feasibility of COS removal by adsorption cannot be determined. Nevertheless, COS adsorption remains an interesting topic for future work.
Finally, multipoint analysis was performed for setpoint G with the wood pellet gasification feedstock for HCN. Figure 7 displays the results.
The wood-based syngas contained a relatively small amount of HCN, about 1.4 ppm, as measured after the reformer sampling point with a condensing step. It was established that the acid-washing step did not effectively remove HCN. HCN is weakly acidic and, therefore, its absorption is low in water at pH 3. The first activated carbon bed in AR was not effective in HCN removal either, likely due to bed saturation; however, HCN was removed in bed 2 to concentrations below 0.05 ppm. Again, no HCN was detected after UC5. HCN adsorption on non-impregnated carbon occurs by weak physisorption. Methods to improve sorption include impregnation with Cu or Cr, which form stable complexes with cyanide [31, 32]. Seredych et al.  established that the surface chemistry of non-impregnated activated carbon greatly affects adsorption performance, and with basic surface pH induced by, for example, nitrogen groups, the surface reactions that result in the formation of stable compounds being deposited in micropores can be promoted.
Table 5 summarizes the impurity concentrations and their removal locations in the high impurity sunflower husk feedstock setpoint E2.
The gas purity goals listed in the table were used in the campaigns to decouple the downstream synthesis unit in case the breakthrough limit was achieved. Impurity concentrations from the literature were adapted for the detection limits of the analyzers employed in this paper. As there was COS breakthrough in high-impurity setpoint E, synthesis was decoupled during this period. Nevertheless, it stands noteworthy that such deep impurity cleaning was achieved with the smaller packed bed loadings in the dirty syngas.
The results, displaying the removal locations for impurities from campaign SXB20/24, provide valuable insights into the adsorption affinity of activated carbons in a complex impurity matrix, which would be difficult to replicate in laboratory conditions. These will be of benefit to continued development of the ultracleaning process and scale-up efforts. In the FlexCHX project, a preliminary 50-MW plant design was completed including an ultracleaning process based on a fully pressurized UC5 concept and single-stage compression to synthesis pressure.
The results of the campaigns presented in this study indicate that for activated carbons, competitive adsorption to microporous sites occurs not only between the removed impurities but also for the water vapor. Promoting more selective removal of impurities could improve the overall removal efficiency. For this, the selection of specific activated carbon types for each impurity removal step could be effective; on the other hand, optimizing the conditions, e.g., bed-specific moisture content and chemical injection adjustment, may suffice.