Mitchell
Juneau
,
Daphna
Yaffe
,
Renjie
Liu
,
Jane N.
Agwara
and
Marc D.
Porosoff
*
Department of Chemical Engineering, University of Rochester, Rochester, NY 14627, USA. E-mail: marc.porosoff@rochester.edu
First published on 19th October 2022
Molybdenum carbides are promising catalysts for the reverse water–gas shift (RWGS) reaction, and we aim to understand if similar performance can be observed across the library of transition metal carbides. Although tungsten and molybdenum carbides exhibit similar catalytic properties for hydrogenation reactions, tungsten carbide has not been thoroughly evaluated for CO2 hydrogenation. We hypothesize that the extreme synthesis conditions necessary for carburizing tungsten can cause sintering, agglomeration, and carbon deposition, leading to difficulty evaluating the intrinsic activity of tungsten carbides. In this work, tungsten is encapsulated in silica to preserve particle size and demonstrate correlations between the active phase composition and RWGS performance.
Similar studies have been extended to tungsten carbide, another group VI transition metal carbide, where CO selectivity enhancements are also observed upon alkali metal promotion.11 Alkali metal promoters increase tungsten dispersion, CO2 and CO adsorption, and facilitate carburization, suggesting that the alkali metals function as structural and electronic promoters. However, the intrinsic effects of alkali promoters on catalytic properties, and in turn, performance are challenging to isolate.11
In general, tungsten carbides demonstrate inferior catalytic activity for CO2 hydrogenation relative to Mo2C, even though the two carbides are reported to have similar CO2 dissociation barriers. This could be a result of both W and Mo carbides exhibiting comparable charge transfer from the transition metal atoms to neighboring carbon atoms.12,13 Therefore, we hypothesize the lower activity of tungsten carbides is due to the low accessibility of active sites caused by particle agglomeration at the high temperatures required for tungsten carburization.14–17 Several methods of controlling particle size have been reported using reverse microemulsion,18,19 and micelle-assisted synthesis methods to support tungsten within a mesoporous silica scaffold.20–22 The mesoporous silica is hypothesized to stabilize the tungsten particles, preventing agglomeration during high temperature carburization.20,23,24
During carburization, tungsten oxide is first reduced to tungsten metal, followed by carbon deposition on the catalyst surface, which diffuses into the bulk to form W2C, followed by the thermodynamically stable WC. Complete carburization is particularly important to ensure inactive tungsten oxides and tungsten metal are carburized to W2C and/or WC.6,25 The hypothesized active W2C phase has been recently isolated when supported on SiO2, suggesting that the carburization conditions can be manipulated to synthesize tungsten carbide catalysts with improved control over reactor performance.15,26,27
Controlling both the tungsten carbide particle size and phase is particularly challenging because the extent of carburization is a function of particle size, resulting in mixed WC, W2C and WOx phases at larger particle diameters.28,29 Recent computational work indicates that the tungsten carburization pathway is a function of both surface energy and particle size, such that γ-WC and ε-W2C phases dominate below 10 nm, and the δ-WC is the thermodynamically favored phase at particle sizes above 10 nm.30 These findings have been validated experimentally, demonstrating that isothermal carburization conditions can produce the isolated tungsten semicarbide (W2C) phase at temperatures below 800 °C at the critical domain size of 10 nm.31
In this paper, we aim to isolate the effects of tungsten carbide phase and particle size on CO2 hydrogenation performance. For studying the effect of particle size, we are contrasting a micelle-based synthesis technique (M) with control experiments over catalysts synthesized by incipient wetness impregnation (IWI). The effect of carburization is studied by varying the carburization conditions at 600 °C, 835 °C and 1000 °C.18,20,32 Our findings demonstrate that tungsten carbide particles encapsulated within mesoporous silica exhibit increased RWGS activity relative to the IWI controls. The catalysts are stable for 12 hours on stream, despite observed particle agglomeration during the reaction, which suggests the well-dispersed tungsten carbide particles enhance the degree of carburization, and in turn, RWGS performance.
WxC/SiO2 (IWI) was synthesized via incipient wetness impregnation of ATOH on SiO2 (Alfa Aesar, amorphous fumed, S.A. 130–170 m2 g−1, 325 mesh powder) with a target loading of 15 wt% W. Both the M and IWI catalysts were calcined at 550 °C with 2 °C min−1 ramp rates for six hours. Catalysts were carburized using one of the three temperature profiles shown in Fig. 1 in a 21% CH4/H2 mixture (33 mL min−1 of CH4, 122 mL min−1 of H2). After a set time at max temperature (20 hours, 4 hours, and 0.75 hours for Profile 1, 2 and 3 respectively), CH4 flow was shut off and catalysts were cooled to room temperature under H2. The catalysts were then passivated under 1% O2/N2 (30 mL min−1) for three hours.
Fig. 1 Carburization profiles for M and IWI with a maximum temperature and duration of: (a) 600 °C, 20 hours; (b) 835 °C, 5 hours; (c) 1000 °C, 1 hour. |
W 4f7/2 binding energies were assigned as follows: W 4f7/2: W0 (W metal) at 31.0 eV, W2+(W2C) at 31.5 eV, W4+ (WC or WO2) at 32.4 eV or 32.9 eV and W6+ (WO3) at 36.3 eV.33–35
Fig. 2 (a–c) TEM images with 50 nm scale bars of nano-WxC (M) carburized at (a) 1000 °C, (b) 835 °C, and (c) 600 °C, with corresponding particle size distributions (d–f). |
Silica encapsulation is also observed over M 835 in Fig. 2b, but with a wider size distribution than the M 1000 sample, as illustrated by the particle size distribution in Fig. 2e. In Fig. 2c, encapsulation of tungsten carbide particles is not observed for the catalyst carburized at 600 °C, and the particle size is the largest of the three samples with median particle size of 15.7 ± 8.4 nm. The observed trend of decreased particle size and enhanced encapsulation with increasing carburization temperature is supported by the lower BET surface area of M 600 (1.7 m2 g−1) relative to M 835 (98 m2 g−1) and M 1000 (68 m2 g−1). Surprisingly, the data indicates that the median particle size decreases with increasing carburization temperature, suggesting that lower carburization temperatures result in tungsten agglomeration and silica collapse. It is important to note that time of carburization decreases with increasing temperature, with 600 °C, 835 °C and 1000 °C held at their maximum temperatures for 20 hours, 5 hours and 0.75 hours, respectively, as shown in Fig. 1.
Details on the effect of carburization time are included in Table S1 and Fig. S1 in the ESI.† We observe a bimodal distribution of particle size, with most particles clustered at 4–6 nm and 10–20 nm. While no overall increase or decrease in median or average particle size is observed with increasing or decreasing carburization time, the data in Fig. 2 and 3 suggest that particle size is a complex function of sample preparation, carburization time and maximum temperature.
Fig. 3 (a–c) TEM images with 50 nm scale bars of IWI WxC carburized at (a) 1000 °C, (b) 835 °C, and (c) 600 °C, with corresponding particle size distributions (d–f). |
TEM characterization of the catalysts synthesized by incipient wetness impregnation (IWI) show formation of large tungsten particles supported on amorphous silica in the TEM images in Fig. 3a–c. This is further supported by the particle size distribution analysis in Fig. 3d–f, where IWI 1000 yields a median particle size of 13.0 ± 7.4 nm, larger than the corresponding M 1000 catalyst in Fig. 2a of 6.7 ± 7.3 nm. In fact, the particle size of the M 1000 and M 835 catalysts are smaller or within error of the IWI catalysts carburized under analogous conditions. In contrast to the M catalysts, BET surface areas of the IWI catalysts appear invariant of carburization conditions, with surface areas of 128 m2 g−1, 173 m2 g−1 and 112 m2 g−1 obtained for IWI 600, IWI 835 and IWI 1000, respectively.
The residual sodium present from synthesis of the M catalysts could be affecting the particle size, as sodium has been shown to increase dispersion of tungsten carbides.11 Quantification of sodium content by ICP-MS indicates all M samples contain 8.0 wt% Na, which may contribute to the decreased particle size of the M catalysts relative to the IWI samples in Fig. 2 and 3, respectively.
Fig. 4 X-ray diffraction patterns of WxC-based catalysts. (a) Incipient wetness impregnation (IWI) and (b) nano-WxC (M). |
Additionally, while a broad amorphous silica structure is observed over IWI 600 around 20°, carburization at higher temperatures (IWI 835 and IWI 1000) appears to result in restructuring of the amorphous silica support into thermodynamically stable quartz, denoted by the peak at 21.8° (abbreviated as Q). The silica reformation into quartz observed over M 835 is likely occurring during carburization, which is supported by the lower BET surface area observed for the M 835 catalyst, as compared to M 1000 °C. The observed quartz formation for the IWI catalysts carburized at 835 and 1000 °C leads us to hypothesize that high carburization temperatures (>835 °C) may cause collapse of amorphous silica into thermodynamically stable quartz. However, we do not observe quartz formation for the M 1000 °C catalyst, possibly because of the shorter (0.75 h) carburization time.
Similar high temperature thermal degradation has been reported for analogous SBA-15 and KIT-6 silica-based catalysts. Exposure to temperatures above 800 °C in air results in silica collapse and complete loss of micro and mesoporous surface area, as measured by BET.36 Additionally, surface area degradation is observed at temperatures as low as 500 °C during air treatment for 5 hours. Therefore, the silica degradation that is observed for the M 600 catalyst during the 20 h carburization in Fig. 1 could be contributing to the lower surface area observed in BET.
Examining the degree of carburization over the IWI catalysts indicates there is an increasing degree of carburization as a function of temperature, following the proposed carburization pathway of W → W2C → WC.37,38 Careful inspection of the XRD in Fig. 4 reveals that the tungsten metal peak at 58.4° in IWI 600 disappears for the catalysts carburized at higher temperatures, and is replaced with W2C signatures at 61.9°, 69.7° and 75.1°. At the highest temperature carburization profile (1000 °C) for the IWI catalyst, we observe appearance of peaks at 31.4°, 35.7°, 48.4° and 64.4°, corresponding to WC. Together, the XRD data suggests a greater degree of carburization with increasing temperature, as expected.
Similar trends are observed over the M catalysts, with increasing carbide formation at increasing carburization temperature. More specifically, W2C peaks are observed in M 835, but the W2C signatures are virtually eliminated over M 1000, suggesting complete carburization to WC. Interestingly, all catalysts exhibit a large peak at 40.3°, which is attributed to metallic tungsten. This suggests that even at the highest carburization temperature of 1000 °C, the M catalysts did not achieve complete carburization, potentially due to mass transfer limitations because the tungsten nanoparticles are encapsulated within silica. Alternatively, because the catalyst precursors are exposed to a CH4/H2 mixture during a temperature ramp, catalyst reduction to W metal may suppress carbide formation.17,27 Furthermore, while W2C has traditionally been viewed as a thermodynamically unstable carbide phase, recent studies suggest SiO2 promotes W2C formation, rationalizing our observation of W2C peaks in IWI 835, IWI 1000 and M 835.31
Fig. 5 DRIFTS spectra of reduced M 1000, IWI 835 and bare SiO2 at 25 °C after CO2 adsorption. The peaks at 2077 cm−1, 2061 cm−1 and 2057 cm−1 suggest CO2 is dissociating into CO. |
Packed bed reactor studies for RWGS indicate all catalysts are highly CO selective, with methane observed as the secondary product along with some traces of ethane (<1%). However, as shown in Fig. 6, the catalysts display varying degrees of activity, represented by CO yield, as a function of time-on-stream. No deactivation is observed over the 12 hours, illustrating all catalysts are stable under RWGS conditions.
From examining the RWGS reactor data in Fig. 6, M 1000 is the top performing catalyst, with a TOF of 2.9 s−1, significantly improved relative to the IWI 1000 catalyst with a TOF of 1.4 s−1. The differences in reactor performance could be either attributed to the smaller particle size of M 1000 (6.7 ± 7.3 nm) versus IWI 1000 (13.0 ± 7.4 nm) or the composition of the bulk carbide phase. The catalysts carburized at 1000 °C exhibit XRD reflections attributed to WC and metallic tungsten in Fig. 4, while the IWI 1000 catalyst also contains features attributed to W2C.
For the M catalysts, the CO yield and activity are functions of the carburization profile, following the trend of M 1000 > M 835 > M 600, which is the same activity trend for turnover frequency (TOF) when the activity is normalized by number of active sites, measured by pulse oxidation in Table 1.40
Catalyst | Conversion (%) | Carbon-based selectivity (%) | CO yield (%) | O2 uptake (μmol g−1) | TOF (s−1) | ||
---|---|---|---|---|---|---|---|
CO | CH4 | C2+ | |||||
M 600 | 2.5 | 96.5 | 3.5 | 0 | 2.4 | 2.5 | 0.6 |
M 835 | 4.9 | 99.2 | 0.8 | 0.0 | 4.9 | 2.6 | 1.1 |
M 1000 | 13.9 | 94.6 | 5.0 | 0.4 | 13.1 | 2.7 | 2.9 |
IWI 600 | 6.5 | 88.0 | 11.3 | 0.7 | 5.7 | 2.7 | 1.5 |
IWI 835 | 9.2 | 94.9 | 4.9 | 0.2 | 8.7 | 6.4 | 0.9 |
IWI 1000 | 5.9 | 86.3 | 12.9 | 0.8 | 5.1 | 2.5 | 1.4 |
Na-IWI 835 | 1.3 | 73.4 | 25.4 | 1.2 | 1.0 | — | — |
Na-IWI 1000 | 0.8 | 95.5 | 4.5 | 0 | 0.8 | — | — |
Over the IWI catalysts, IWI 835 exhibits the highest RWGS activity, followed by IWI 600 and IWI 1000. However, IWI 835 exhibits the lowest TOF of the IWI catalysts due to the increased O2 uptake of 6.4 μmol g−1. The increased O2 uptake over IWI 835 could be due to the carburization profile resulting in increased availability of WxC active sites.
Comparing the TOF values in Table 1 of the two classes of catalysts shows that the IWI catalysts outperform M 600 but are roughly half as active as M 1000. However, catalysts with similar median particle sizes, for example, M 1000 (6.7 ± 7.3 nm) and M 835 (8.3 ± 8.1 nm), exhibit substantially different activity with respective CO2 conversions of 13.9% and 4.9%. This suggests that decreased particle size is not the sole factor determining catalytic performance, and that particle accessibility and the surface tungsten carbide phase are likely playing important roles. For example, the low BET surface area of M 600 reported in Table S2† suggests silica collapse, which agrees with the TEM images in Fig. 2c, showing tungsten carbide agglomeration within the amorphous silica. Lower TOF values over the M 600 and M 835 catalysts could also be from the Na that is present during M synthesis. Details of the BET surface areas are included in Table S2 of the ESI along with their respective isotherms in Fig. S2 and S3.†
To further investigate the effect of Na, two controls of IWI 835 and IWI 1000 with 8.0 wt% Na are included in Table 1. While 1 to 3 wt% Na have been shown to increase CO selectivity at the slight expense of CO2 conversion over tungsten carbide catalysts, our data suggests that excessive Na appears to diminish activity.9,11 Interestingly, M 1000 exhibits the highest conversion and highest TOF of all catalysts, despite the 8.0 wt% sodium content.
Textural properties are further described through TEM images and particle size distributions after 12 hours time-on-stream (shown in Fig. S4 and S5 of the ESI†) with the summary of particle size distributions included in Table S3.† Based on the TEM data, particle restructuring occurs under RWGS conditions with median particle size increasing for all catalysts except for IWI 1000 and M 600. All catalysts exhibit stable RWGS activity and stability for 12 hours on stream, suggesting that the particle restructuring occurring under reaction conditions is not a key parameter for determining RWGS activity. Furthermore, as evidenced in the deviations in activity observed over the range of IWI carburization temperatures with little change in particle size, the degree of surface carburization determined by X-ray photoelectron spectroscopy (XPS) is likely playing a key role in the observed activity correlations.
Fig. 7 W 4f XPS spectra for WxC-based catalysts after 12 hours time-on-stream. (a) M 600, (b) M 835, (c) M 1000, (d) IWI 600, (e) IWI 835 and (f) IWI 1000. |
XPS composition (%) | W | W2C | WC | WxC | WO2 | WO3 |
---|---|---|---|---|---|---|
M 600 | 59.0 | 3.9 | 1.2 | 5.1 | 8.7 | 27.2 |
M 835 | 40.2 | 7.8 | 22.4 | 30.2 | 0 | 29.6 |
M 1000 | 1.3 | 20.7 | 37.0 | 57.7 | 18.2 | 22.8 |
IWI 600 | 44.8 | 1.9 | 9.2 | 11.1 | 24.0 | 20.1 |
IWI 835 | 22.8 | 4.9 | 35.2 | 40.1 | 12.2 | 24.9 |
IWI 1000 | 19.2 | 3.9 | 1.8 | 5.7 | 36.7 | 38.4 |
Careful analysis of the quantitative XPS contributions in Table 2 show that the tungsten metal contributions decrease with increasing carburization temperature for each class of catalyst. The highest degree of tungsten carbide formation (W2C plus WC) is observed over M 1000 (57.7%), followed by IWI 835 (40.1%), M 835 (30.2%), IWI 600 (11.1%), IWI 1000 (5.7%) and M 600 (5.1%). The RWGS activity follows a similar trend, with M 1000 and IWI 835 exhibiting the highest CO2 conversion, suggesting that the degree of carbide formation is correlated with activity, as shown in Fig. 8.11,14
Fig. 8 Correlation between RWGS activity and XPS contributions of tungsten carbides (WxC), represented as WC plus W2C. |
More specifically, M 1000 exhibits the highest degree of W2C, which may be a result of the intimate contact between the WxC nanoparticles and the silica support stabilizing the W2C phase.31 The ∼10 nm particle diameter of M 1000 may enable formation of a W2C surface phase, which is not thermodynamically favored at larger particle diameters. This is consistent with the higher degree of W2C formation over the M catalysts versus the corresponding IWI catalysts in Table 2. The increased amount of W2C could be contributing to the enhanced CO2 hydrogenation activity over M 1000, as past studies have reported a decreased activation barrier for hydrogenation reactions over W2C when compared to WC.26,41 Additionally, the increased interfacial area of the M catalysts could be responsible for the high selectivity to CO.42,43
To contextualize the performance of M 1000 and the other WxC-based catalysts, Table S7 of the ESI† contains reaction conditions and performance metrics of previously reported transition metal carbides. As shown in the table, M 1000 exhibits a lower CO space–time yield (9.8 μmol CO g−1 s−1) at 350 °C with respect to K-promoted Mo2C/γ-Al2O3 (61.4 μmol CO g−1 s−1) and FeCrOx (85.8 μmol CO g−1 s−1) at 450 °C.10 Although the performance of M 1000 does not exceed some of the top-performing RWGS catalysts, the CO space–time yield is three times greater than that over previously reported WxC catalysts (3.2 μmol CO g−1 s−1)11 and more than five times greater than the M 600 (1.8 μmol CO g−1 s−1) reported in this work. Therefore, the findings of this paper show that the method of synthesizing WxC-based catalysts is critical for determining their intrinsic activity for CO2 hydrogenation reactions.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d2nr03281c |
This journal is © The Royal Society of Chemistry 2022 |