Iron ‑ chromium mixed metal oxides catalyse the oxidative dehydrogenation of propane using carbon dioxide

The oxidative dehydrogenation of propane to propene using carbon dioxide is an attractive novel synthesis route to produce a key platform chemical. We investigated iron-chromia catalysts prepared by sol-gel. The mixed oxides showed improved specific surface area and a high concentration of acid and base sites and were found to be highly active catalysts. Although coking was observed over the catalysts, oxidative regeneration restored the majority of the initial activity over 5 consecutive cycles. The initial space time yield was amongst the most active Cr-based catalysts in the literature.


Introduction
Propene is a major platform chemical and precursor to many chemicals and materials, most significantly polypropene.Worldwide annual propene production was 130 megatonnes in 2019 and it is predicted that demand will increase over the next decade [1].Current production methods mostly rely on steam and fluid-catalysed cracking technologies, which have several drawbacks: Cracking of naphtha produces propane as one of several products, meaning the atom economy is low.Additionally, the demand for propene has been expected to outstrip supply, hence the emergence of the 'propene gap'.In order to fill this gap, industry has responded by investing in 'on-purpose' propene production processes, namely the direct dehydrogenation of propane (DDH), shown in Eq. (1).
To date there are several commercial processes in operation, including Oleflex, Dow FCdh, STAR process, CATOFIN and K-PRO.In most cases, Pt-or Cr-based catalysts are employed.While selectivity above 90% can be achieved at conversions in excess of 40%, there are also disadvantages associated with this method of propene production.Coking remains a challenge in all cases, and frequent catalyst regeneration cycles are needed, which involves burning off coke.Secondly, the reaction is equilibrium-limited and so operating conditions, and ultimately C 3 H 6 yields, are restricted.
The oxidative dehydrogenation of propane to propene using carbon dioxide (ODH-CO 2 ) represents an interesting alternative to direct dehydrogenation: The addition of CO 2 offers a strategy to utilise carbon, which could have a significant effect on the carbon neutrality in the life cycle of propene production [2].CO 2 facilitates the reverse water-gas shift reaction (RWGS, Eq. ( 2)) to remove H 2 and form H 2 O, which means higher equilibrium yields of propene are possible.Furthermore, the reverse Boudouard reaction (Eq.( 3)) can take place, which in turn removes coke and extends the catalyst lifetime. (2) For the above reasons, ODH-CO 2 has attracted significant academic interest in the last 20 years.The most active catalysts reported to date are mixed metal oxide-based catalysts that use Cr or Ga oxides as the active component.Notably, Michorczyk and co-workers have made several contributions to the field and reported numerous supported CrO x catalysts with promising performance [3][4][5].In all cases, deactivation via coking was a challenge.In addition to Cr, Ga 2 O 3 /Al 2 O 3 [6,7], Fe 2 O 3 [8], In 2 O 3 [9] and various other supported transition and platinum group metals have been reported as active catalysts [10,11]. Generally

Catalyst characterisation
A series of FeCrO x catalysts were prepared using a typical sol-gel technique (experimental details provided in the Supplementary Information), with various Cr:Fe molar ratios.The catalysts are referred to as FeCrX, where X = mol Cr / (mol Cr + mol Fe ) ×100.For example, FeCr25 is a mixed metal oxide where there is 25 mol% Cr and 75 Fe mol%.For simplicity, the Fe-and Cr-only oxides are simply referred to as Fe or Cr.The series of catalysts was initially analysed by ICP-AES to measure the actual ratio of Fe:Cr.The results are presented in Table S1 and are in excellent agreement with the nominal values, showing that the sol-gel method is appropriate to prepare mixed FeCr oxide catalysts.
The catalysts were then characterised to understand how the addition of Fe to CrO x affects the physical and chemical properties.Table 1 shows that the specific surface area varies substantially with the molar ratio of Fe:Cr.The highest surface area, 29 m 2 g − 1 , was observed over the Fe 0.5 Cr 0.5 O x catalyst.FeCr100 and FeCr0 showed values of 8.5 and 23 m 2 g − 1 , respectively.XRD patterns of the catalysts are presented in Fig. S1.Typical diffraction patterns indicative of Fe 2 O 3 and Cr 2 O 3 were observed, with principal reflections at 24, 34 and 37 • [12,13].In the case of Cr, the reflections were observed at a higher angle (2θ), consistent with smaller lattice spacings in the crystal structure.The mixed metal oxides were clearly less crystalline than the Cr and Fe catalysts, which is also consistent with the higher specific surface areas observed in these samples.In the case of FeCr25 and FeCr50, there is a broad, but visible additional reflection at 31 and 42 • .The crystallite size was estimated using the Scherrer equation (Table 1) and is inversely proportional to the specific surface areas of the FeCr oxides.Cr 2 O 3 is the outlier, which shows a reasonable specific surface area but a larger crystallite size.
It is well known that acid and base sites play an important role in DDH and ODH-CO 2 catalysis.Acid sites are able to activate C 3 H 8 , but if too strong, can promote deep dehydrogenation and coke formation [2].Basic sites have also been implicated in coke formation over Zn/ZrO 2 catalysts [14], but are generally considered to be sites for CO 2 activation.Accordingly, the balance of these sites can influence the surface concentration of C 3 H 8 , CO 2 and the rate of dehydrogenation and RWGS.NH 3 -TPD and CO 2 -TPD analysis was therefore completed on each sample in order to establish the concentration and strength of the acid and base sites present (Fig. S2).In poorly acidic and basic samples, the interpretation of the data is challenging due to the low intensity of desorption peaks.However, there are clear differences in the profiles of each sample.In general, the mixed oxides had more acid and basic sites than the Fe and Cr.Regarding the acid sites, the mixed metal oxide catalyst exhibited two broad desorption peaks, centred approximately at 200 and 380 • C. Fe and Cr exhibited lower NH 3 desorption, although FeCr100 showed a high temperature desorption peak at 550 • C, indicative of a strong Lewis or Bronsted acid site.The CO 2 -TPD profiles also show an increase in the intensity of the desorption peaks, indicative of an increase in the number of basic sites on the catalyst surface.The FeCr50 exhibited the largest desorption peaks in the series of catalysts.As the sample mass was controlled in each experiment, the integrated area of the peaks gives allows the relative concentration of acid and base sites on the catalysts to be inferred.These estimations are presented in Table 2.This shows that FeCr50 has the highest concentration of acid   SEM analysis was carried out to probe the morphology of the prepared catalysts.Representative micrographs are presented in Fig. S3(ae).The morphology of the Fe, Cr and FeCr25 (a,b and e) show small, dense clusters of particles around 20 nm in diameter.Conversely, the FeCr50 and FeCr75 samples (c and d) are bimodal in their particle size distribution.There are very large particles in the order of several hundred nanometres in size, with small particles below 20 nm decorating the edges.
In order to probe the redox properties of the catalysts, H 2 -TPR was carried out (Fig. 1).The reduction profile of Fe is defined by two peaks, one at 350 • C and one around 550 • C. The first peak is the reduction of hematite, Fe 2 O 3 to magnetite, Fe 3 O 4 .The higher temperature peak is due to the complete reduction of magnetite to Fe 0 [15].The addition of Cr causes a significant shift to lower temperatures in these peaks, to 255 and 360 • C.This indicates that the introduction of Cr enhances the reducibility of the catalyst.The reducibility of bulk Cr 2 O 3 compared to Fe 2 O 3 is very poor and indicates that the addition of Fe to Cr introduces redox active sites.The dissociative activation of CO 2 can take place at oxygen vacancies, therefore the increased reducibility of the FeCr catalysts would be expected to facilitate more facile CO 2 activation than Fe or Cr.

Propane dehydrogenation
The series of catalysts were evaluated in the DDH and ODH-CO 2 reactions.It has been shown that for some catalysts the addition of CO 2 inhibits propane activation, hence the value in comparing DDH and ODH-CO 2 activity [16][17][18].The origin of this inhibition effect was determined by Xu et al. to be due to CO 2 blocking propane adsorption on basic sites [16].The effect is support-dependent and underlines the importance of balancing acid and base sites on the catalyst surface.Table 2 shows the initial conversion and selectivity of the catalysts in the DDH and ODH-CO 2 reactions.
There are several notable differences in the performance of the catalysis in the presence and absence of CO 2 .It is clear that the combination of Fe and Cr leads to higher activity in both reactions.When the conversion and C 3 H 6 yield are plotted as a function of the Fe:Cr ratio, a clear volcano plot is observed with the maximum at the FeCr50 catalyst (Fig. 2a).When normalised for differences in specific surface area (Fig. 2b), a trend is still observed, whereby the mixed FeCr catalysts are more active than the Fe or Cr-only catalysts.Across the range of catalysts, the yield varied by more than a factor of 2, which indicates other differences beyond surface area in the properties of the catalyst contributed to the differences in activity.The relationship between acid/ base sites and C 3 H 6 yield was also investigated.Fig. S4a shows that there is a positive correlation between the C 3 H 6 yield per m 2 and the acid sites per m 2 .Fig. S4b shows that there is a volcano-type relationship between the surface area normalised C 3 H 6 yield and the number of basic sites per m 2 .As discussed above, acid sites are considered to be activation sites for C 3 H 8 molecules, and so it is expected that there would be such a correlation.CO 2 activation takes place on basic sites and so it would be expected that more basic sites would lead to higher yields of C 3 H 6 .The correlation in Fig. S4b suggests that the reaction is not limited by CO 2 activation, rather it is limited by C 3 H 8 activation.The catalysts with the highest concentration of basic sites have an intermediate concentration of acid sites, and an intermediate C 3 H 6 yield.Therefore the improved performance of the mixed FeCr catalysts is most likely due to the facilitation of more active sites for C 3 H 8 activation.
The addition of CO 2 to these catalysts is clearly beneficial.The C 3 H 6 yield increases across all catalysts, but to the largest extent over the mixed metal oxide catalysts, suggesting that the mixed metal oxides can efficiently activate CO 2 and C 3 H 8 .A further benefit of CO 2 is the improved converted carbon balance.In the DDH reaction, the converted carbon balance improves with the addition of Fe, but in the most active catalyst (FeCr50), it is only 78%.In the ODH-CO 2 reaction, the carbon balance also improves as Fe content increases, but above 50 mol% Fe, the converted carbon balance reaches 94% and above.The missing carbon may be ascribed to deep dehydrogenation forming coke on the catalyst.It has been shown that dry reforming can occur over supported nanoparticle catalysts and so this pathway cannot be disregarded [11].Additionally, it is not possible to distinguish between dry reforming, steam reforming and the reverse Boudouard reaction.Unfortunately, quantification of CO, CO 2 and H 2 was not possible using the current analysis, and therefore the rate of CO formation or deep dehydrogenation to coke cannot be determined.It should be noted that the selectivity calculations are based on the moles of converted carbon, therefore are not impacted by the inability to quantify CO or coke formation.
C 3 H 6 selectivity also increases as Fe content increases, but values are modestly improved in the presence of CO 2 .Aside from C 3 H 6 , CH 4 and some C 2 products (C 2 H 4 and C 2 H 6 ) were also formed.Interestingly, the CH 4 formation was much higher in the presence of CO 2 , which is likely due to methanation of CO 2 and CO (Eqs.( 4) and ( 5)).Cracking (Eq.( 6)) or hydrogenolysis (Eq.( 7)) reactions are generally suggested to be responsible for CH 4 and C 2 H 4 formation in DDH, but C 2 H 4 may react further over the catalyst via deep dehydrogenation to form coke, hence the deviation of CH 4 and C 2 H 4 from stoichiometric values.
The stability of a catalyst is of prime importance in evaluating its performance.Fig. S5(a-d) shows the time-on-stream activity of the catalysts over 2 h.In the DDH reaction, the conversion of all the catalysts reaches 0% after 2 h.This rapid deactivation is most likely caused by coking of the catalyst from deep dehydrogenation, as is widely reported for DDH [2].The propene selectivity increases on-stream as conversion drops, as is often observed in dehydrogenation reactions.In the ODH-CO 2 reaction, the catalysts are clearly more stable, but still deactivate quite rapidly.The most active catalyst, FeCr50, retained only 33% of its initial activity after 2 h on-stream, but still maintained the highest conversion of the series of catalysts after this time, on par with FeCr25.Similarly, the propene selectivity increases on-stream as the conversion decreases.
It is well known that even commercial DDH catalysts require frequent regeneration cycles to operate efficiently.Therefore, the ability to recover activity after regeneration is an additional indicator of longterm stability.Fig. 3 shows the performance of the most active catalyst, FeCr50, after 5 consecutive reaction cycles.The regeneration step involved switching the reaction feed to flowing air for 2 h, purging with N 2 and then switching back to the DDH reaction mixture, as described in the Supplementary Information.In the DDH cycles, the C 3 H 8 conversion quickly decreases after regeneration, falling from 8.5% in the first run, to 2.5% in the fifth run and the C 3 H 6 selectivity remains constant throughout each run.In the ODH-CO 2 cycles, the C 3 H 8 conversion decreases from 14% to 10.5% after 5 runs, a more modest deactivation.As was the case for the DDH cycles, the C 3 H 6 selectivity remains fairly constant and may increase slightly, from 72% to 74%.Clearly, the presence of CO 2 enables the catalyst to be regenerated more easily, although the deactivation is not completely reversible.In addition to coking, loss of active surface area through sintering is possible, and this could also occur during the exothermic regeneration step.
Finally, the space time yield of the FeCr catalysts was compared with similar Cr-based ODH-CO 2 catalysts.A comparison of these is provided in Table S2 and indicates that FeCr50 is amongst the most active Crbased catalysts reported for the ODH-CO 2 reaction.The most active catalysts in the table are typically supported on high surface area supports, whereas in the current work the catalyst was prepared as a bulk oxide.Supported FeCr catalysts would therefore be of interest to study, but the structural complexity would warrant a full investigation.Specifically, Cr speciation (monomeric, oligomeric, nanocrystalline Cr 2 O 3 ) is well known to affect activity and selectivity in C 3 H 8 dehydrogenation, and the interaction of Fe with such species may be complex.

Conclusions
The combination of iron and chromium to form mixed metal oxides has been shown to produce highly active and selective ODH-CO 2 catalysts.The binary oxides were shown to be more active than either Fe-or Cr-only, partly due to the increased specific surface area of the samples facilitating a higher active surface area.However, the improved selectivity of the catalysts and suppression of coke formation points to an intrinsic improvement in the efficiency of the catalyst.The improved reducibility of the catalysts may play an important role in facilitating CO 2 activation (and RWGS) during the reaction, which improved the C 3 H 6 yield.Additionally, the increase in acid site density may contribute to the improved performance.The FeCr50 catalyst exhibited improved stability compared to Cr-only, and is amongst the most active Cr-based catalysts reported to date.This work highlights the potential of multicomponent redox active metal oxides to catalyse the ODH-CO 2 reaction, and shows that RWGS catalysts may be considered for ODH-CO 2 catalysis, Further optimisation of the catalyst structure and formulation is important to realise the potential of FeCr catalysts for this reaction.

Fig. 2 .
Fig. 2. (A) The relationship between Fe:Cr ratio and the C 3 H 6 yield and (B) the relationship between the Fe:Cr ratio and the surface area normalised C 3 H 6 yields in the DDH (■) and ODH-CO 2 (•) reactions.

Fig. 3 .
Fig. 3. Catalytic cycles of FeCr50 under DDH (A) and ODH-CO 2 (B) reaction conditions.Blue bar: C 3 H 8 conversion, red bar: C 3 H 6 selectivity.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) , supports for Cr-based catalysts are inert and most often they are SiO 2 -based.The main limitation in the field of ODH-CO 2 is the low conversion and C 3 H 6 yields compared to optimised DDH reactions.Therefore, the discovery of novel, highly active catalysts is of interest.The main mechanistic difference between DDH and ODH-CO 2 is the requirement of the catalyst to activate CO 2 and facilitate the reverse water-gas shift reaction.While Cr 2 O 3 is highly active in the DDH reaction, it's activity as a reverse water-gas shift catalyst may not be sufficiently high.Mixed metal oxides of Cr and Fe are applied as hightemperature water-gas shift catalysts and consequently may be interesting candidates for ODH-CO 2 catalysts.In this work, we investigate this hypothesis by preparing a range of FeCr mixed oxides and testing them for DDH and ODH-CO 2 .

Table 1
Physical and chemical properties of FeCr catalysts.

Table 2
DDH and ODH-CO 2 results for the FeCrO x catalysts.
a Others refers to sum of C 2 H 4 and C 2 H 6 .b C balance = Converted Carbon balance.