1 Introduction

The emission of greenhouse gases (GHGs) is considered with great concern around the world, and accordingly a multitude of strategies are currently being developed to contain global climate change. Lean-burn gas engines can provide a relevant contribution to the sustainable supply of energy and transportation due to their high efficiency and GHG neutrality, particularly when biomethane is used as a fuel. However, one issue affecting lean gas engines is the emission of methane (CH4) and formaldehyde (CH2O) resulting from the incomplete combustion of the fuel [1,2,3]. The environmental relevance of these pollutants is associated with the strong GHG impact of CH4 (CO2 factor 25) and the toxicity and carcinogenicity of CH2O [4]. Therefore, emission limits have been introduced in the European Union requiring the after-treatment of gas engine exhaust emissions. For this purpose, catalysts based on platinum (Pt) and palladium (Pd) are considered due to their high oxidation efficiency (Eq. 1 and 2) [5, 6].

$${\text{CH}}_{{4}} + {\text{ 2O}}_{{2}} \to {\text{ CO}}_{{2}} + {\text{ 2H}}_{{2}} {\text{O}} \,\,\,\,\,\,\,\,\,\Delta {\text{H}}_{{\text{r}}} = \, - { 8}0{\text{3 kJ mol}}^{{ - {1}}}$$
(1)
$${\text{CH}}_{{2}} {\text{O }} + {\text{ O}}_{{2}} \to {\text{ CO}}_{{2}} + {\text{ H}}_{{2}} {\text{O}} \,\,\,\,\,\Delta {\text{H}}_{{\text{r}}} = \, - { 52}0{\text{ kJ mol}}^{{ - {1}}}$$
(2)

However, precious metal catalysts only exhibit limited long-term stability for CH4 oxidation [7]. Hence, alternative catalytic materials have been investigated such as Co [8,9,10], Cu [11, 12] and Mn [8, 10]. By contrast, only little is known about the CH4 oxidation activity of Fe-based catalysts, although iron is characterized by high availability, a low price and toxicological harmlessness. Recently, it was demonstrated that iron catalysts are basically capable of oxidizing CH4 with a conversion rate of 50% at 500 °C (4000 ppm CH4, 10 vol% O2, N2 balance; 50,000 h−1) [8]. Additionally, iron oxides have been reported as effective catalysts for the oxidation of CO, hydrocarbons and soot under O2-rich conditions [13,14,15,16].

The objective of the present study is the knowledge-based development of advanced iron oxide catalysts for the oxidation of CH4 and CH2O in lean gas engine exhaust. For this purpose, a systematic series of commercial iron oxide powder samples was characterized physico-chemically and tested in a model exhaust gas to evaluate structure–activity correlations. The most effective sample was transferred to the honeycomb level for practical laboratory tests followed by up-scaling to a real catalytic converter for tests on a stationary gas engine. Finally, mechanistic studies on CH2O oxidation were performed with the best catalyst using in-situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) and step-function experiments.

2 Experimental

2.1 Physico-Chemical Characterization of the Samples

In this study, five commercial iron oxide powder samples (Lanxess) with different physico-chemical properties were used. The samples were present in the form of a α-Fe2O3 modification, which is thermodynamically stable under the lean exhaust conditions of stationary gas engines, primarily involving temperatures above 450 °C [17]. Before physico-chemical characterization and catalytic testing, each material was pre-treated in air at 600 °C for 3 h (raising the heating at a rate of 4.5 K min−1). The samples were characterized by N2 physisorption, X-ray diffraction (XRD), Laser Raman spectroscopy, energy dispersive X-ray spectroscopy (EDX), the temperature-programmed reduction of CO (CO-TPR) and the temperature-programmed desorption of CO2 (CO2-TPD).

N2 physisorption was performed on a TriStar II (Micromeritics) at − 196 °C. The samples were pretreated at 350 °C for 16 h in a vacuum (10–4 bar). The BET surface area (SBET) was calculated from the adsorption data recorded at p/p0 ratios between 0.06 and 0.20.

XRD was carried out on a D8 Discover (Bruker) at room temperature. The diffractometer was equipped with a Bragg–Brentano configuration, an Fe-filtered Co-Kα radiation source (35 kV/45 mA) and a VANTED-1 detector. Each diffractogram was recorded from 20° to 80° (2Ө) with a step width of 0.01° and dwell time of 0.05 s step−1. TOPAS software (Bruker) was used to estimate the crystallite size (dcryst) according to the Scherrer equation.

Laser Raman spectra (LRS) were recorded with an inVia (Renishaw). The spectrometer was equipped with a Nd:YAG laser (532 nm, 127 kW m−2), grating with 1800 lines per mm and a CCD array detector. Spectra were taken between 56 and 1600 cm−1 with a resolution of 1.6 cm−1, an exposure time of 60 s and a laser power of 0.5%. Spectral data were processed using WIRE 4.2 software.

EDX was conducted on a Quanta 250 FEG from FEI equipped with an energy dispersive X-ray spectrometer (EDAX AMETEK—Octane Plus) providing a local resolution of about 1.7 nm (dwell time 200 µs). A tungsten incandescent cathode was taken as an electron source with an acceleration voltage of 20 kV. For analysis, the samples were prepared on standard carbon adhesive tabs. EDX measurements were evaluated with the EDAX TEAM software.

CO-TPR and CO2-TPD were made on a home-made laboratory test bench requiring a quartz glass tube reactor (i.d.: 8 mm). The gases (Air Liquide) were fed by independent mass flow controllers (Bronkhorst), whereas the total flow was always kept at 150 ml min−1 (STP). Online gas-phase analysis was performed by a hot measuring FTIR spectrometer (Multi-Gas 2030, MKS instruments). For CO-TPR, a sample mass of 100 mg (powder) was introduced into the reactor as a fixed bed, pre-treated for 15 min at 550 °C in synthetic air flow and then cooled to 50 °C. Subsequently, the TPR was started by continuously dosing a mixture of 3000 ppm CO and N2 (balance) and raising the temperature to 850 °C at 10 K min−1. Temperatures were controlled by using K-type thermocouples in front of and behind the packed bed (ΔT < 5 K). For CO2-TPD, the sample used was in the form of granules (500 mg, 125–250 µm). After the same pre-treatment as for CO-TPR, the sample was saturated with CO2 at 50 °C using a mixture of 3000 ppm CO2 and N2 (balance). After this, it was flushed with N2 until the proportion of CO2 was below 5 ppm, then TPD was started by heating to 850 °C at the rate of 20 K min−1.

2.2 Catalytic Studies

Catalytic investigations of powders and honeycombs were performed on a laboratory test bench. The powder catalysts (300 mg, particle size < 250 µm) were fixed in the tubular quartz glass reactor (i.d.: 10 mm) with quartz wool, while temperature was monitored by K-type thermocouples located directly in front of and behind the catalyst. Gases were dosed using independent mass flow controllers (Bronkhorst). Gas-phase analysis was carried out as for TPD and TPR, whereas O2 was measured by a lambda probe (LSU 4.9, Bosch). Reactor inlet concentrations were adjusted by bypassing the reactor. Before the tests, each sample was heated for 15 min at 550 °C in a flowing mixture of 20 vol% O2 and 80 vol% N2. Subsequently, the model exhaust gas, composed of 1000 ppm CH4, 100 ppm CH2O, 5 vol% O2, 10 vol% H2O and N2 (balance), was dosed with a total flow rate of 1000 ml min−1 (STP), corresponding to a space velocity (S.V.) of 150,000 h−1. Catalytic data were recorded under stationary conditions, while the temperature was decreased from 550 °C to 200 °C in steps of 50 K.

The best catalyst was coated on a cylindrical cordierite honeycomb (D: 25 mm, L: 40 mm, 400 cpsi) resulting in a load of 100 g l−1. For this purpose, the powder catalyst was dispersed in H2O using hydroxyethylcellulose (Sigma-Aldrich) as detergent (2 g l−1). Subsequently, the honeycomb was dipped into the dispersion and was then dried at 100 °C. This procedure was repeated several times. The honeycomb was finally annealed at 400 °C in air to remove the detergent. The catalytic test procedure was very similar to that for powders, whereas the quartz glass reactor had an i.d. of 26 mm and the total flow was 22.5 l min−1 (STP) corresponding to a S.V. of 65,000 h−1. The model exhaust gas consisted of 1500 ppm CH4, 100 ppm CH2O, 600 ppm CO, 250 ppm NO, 5 vol% O2, 12 vol% H2O, 12 vol% CO2 and N2 (balance).

The best catalyst was transferred to a real-size honeycomb (D: 390 mm, L: 450 mm, 200 cpsi) producing a loading of 85 g l−1. The coating was done as described in previous paragraph. The tests were conducted in the lean exhaust of a stationary gas engine (MWM TCG V12, 600 kW) fed with biomethane, while establishing constant engine operation points with loads of 50 and 75%, resulting in exhaust temperatures of 485 °C (S.V.: 25,000 h−1) and 465 °C (S.V.: 37,000 h−1). The temperatures were recorded by K-type thermocouples positioned directly in front of the catalytic converter.

The performance of the catalysts was evaluated based on the conversion of the respective gas component according to the equation X(i) = (y(i)in−y(i)out)/y(i)in.

2.3 Mechanistic Studies of the CH2O Oxidation

In-situ DRIFTS was performed using a FTIR spectrometer Tensor 27 (Bruker) equipped with a MCT detector and Praying Mantis optics (Harrick). For simultaneous surface and gas-phase analysis, the FTIR spectrometer mentioned in Sect. 2.1 was connected to the flow via a DRIFTS cell (Harrick). Gases were supplied as described in Sect. 2.2. Fresh catalyst powder (approx. 30 mg) was introduced into the DRIFTS cell and pre-treated at 400 °C in synthetic air for 15 min. Subsequently, the sample was cooled to 350 °C, then the temperature was decreased to 50 °C in steps of 50 K, a background spectrum being recorded at each temperature level. At 50 °C, the sample was exposed to a gas mixture of 500 ppm CH2O and N2 (balance) with a total flow rate of 200 ml min−1 (STP). After saturation, it was flushed with a blend of 5 vol% O2 and 95 vol% N2 and spectra were recorded until no more changes occurred. Then, the sample was heated up to 350 °C in steps of 50 K under a constant gas flow. The spectra were collected between 800 and 3800 cm−1 with a resolution of 2 cm−1, accumulating 400 scans per spectrum (5 min). The spectral data were processed using Opus Software (Bruker). Another series of experiments was performed following the same procedure but without dosing O2. Moreover, in further experiments, the adsorption of the CH2O (in N2) was also carried out at − 100 °C with subsequent flushing in 5 vol% O2 and 95 vol% N2.

Step function experiments were conducted on the laboratory test bench (Sect. 2.2) taking the best catalyst. The catalyst powder (200 mg) was pre-treated at 550 °C for 15 min in synthetic air and was subsequently cooled to 150 or 200 °C. Then, a gas mixture of 500 ppm CH2O and N2 (balance) was dosed into the bypass line, and additional runs were performed in the presence of 5 vol% O2. The step function was realized by rapidly switching the gas flow from the bypass to the reactor. The total flow was always kept at 500 ml min−1; gas phase analysis was carried out again using the FTIR spectrometer (Sect. 2.1).

3 Results and Discussion

3.1 Physico-Chemical Properties of the Fe2O3 Catalysts

The most important physico-chemical characteristics of the samples are demonstrated in Table 1, indicating BET surface areas from 5 to 25 m2 g−1. Based on these SBET values, the samples are denoted by Fe5, Fe6, Fe8, Fe12 and Fe25 throughout this article. The mean pore diameters (dpore) derived from the BJH model are between 32 and 52 nm (Table 1) indicating mesoporosity. The structural analyses by XRD (Fig. S1) and LRS (Fig. S2) confirm the exclusive presence of the hematite-type α-Fe2O3 phase in all the samples (Table 1). Additionally, it is obvious that the crystallite sizes (dcryst) derived from the Scherrer equation decrease with increasing SBET.

Table 1 Physico-chemical characteristics of the Fe2O3 samples
Full size table

The redox properties were investigated by CO-TPR as shown for Fe25 by way of example (Fig. 1, left). It is known from the literature that the conversion of Fe2O3 with CO into CO2 involves a three-step mechanism following the reduction sequence \({\text{Fe}}_{{2}} {\text{O}}_{{3}} \to {\text{Fe}}_{{3}} {\text{O}}_{{4}} \to {\text{FeO}} \to {\text{Fe}}\) [15]. The two initial steps coincide in the CO2 feature at 350 °C as inferred from the cumulative oxygen release (Fig. 1, left) amounting to 11% (Fe3O4, area I) and 33% (FeO, Area II). The subsequent production of elemental Fe is indicated by the CO2 evolution above 450 °C (Area III), resulting in the almost complete removal of oxygen from the Fe2O3 at 850 °C (96%). In total, 18.8 mmol g−1 of oxygen is removed. Additionally, the CO-TPR profiles reveal a minor pre-peak at low temperatures (TLT), which is observed at 230 °C for Fe25, whereas the remaining samples show higher TLT values shifting to 275 °C. Since this CO2 formation occurs at the beginning of the TPR, it is assumed that surface and subsurface oxygen species are predominately involved. For a rough estimation of the amount of oxygen required for a monolayer, the surface area of O2− (6.16·10–20 m2) [18], the mass and the BET surface area of each sample are used. With this value (14–67 µmol) and the amount of oxygen released during the low-temperature CO2 formation (nO,LT, 17–67 µmol), it is deduced that both quantities are in the same order of magnitude, indicating the participation of surface and/or subsurface oxygen. Thus, TLT is taken in Sect. 3.2 as a feature to express the reducibility of the sample surfaces.

Fig. 1
figure 1

CO-TPR of Fe25 with cumulative oxygen release in the form of CO2 (left) and CO2-TPD of Fe25 (right). CO-TPR conditions: m = 150 mg, F = 150 ml min−1 (STP), y(CO) = 3000 ppm, N2 balance, β = 10 K min−1. CO2-TPD conditions: m = 500 mg, F = 150 ml min−1 (STP); saturation: y(CO2) = 3000 ppm, N2 balance; TPD: N2 flow, β = 20 K min−1

Full size image

The number of active Fe sites is estimated from the uptake of CO2 determined by CO2-TPD (nCO2, Table 1). Note that before the TPD, the samples were pre-treated at 550 °C, resulting in some dehydroxylation of the surface that partially changed the coordination sphere of the surface Fe atoms, whereas the adsorption of CO2 led to the formation of carbonate surface complexes [19]. The CO2-TPD profile is shown by way of example for Fe25 (Fig. 1, right), demonstrating two prominent features at 100 and 400 °C that are associated with the decomposition of differently coordinated carbonato entities [19]. The nCO2 data range from 2.3 to 6.6 μmol g−1, roughly increasing with the BET surface area.

3.2 Oxidation Performance And Structure–Activity Correlations Of The Fe2O3 Catalysts

The performance of the powder catalysts during the simultaneous CH2O and CH4 oxidation was evaluated in a model lean gas engine exhaust gas. Figure 2 (left) shows that Fe25 exhibits the highest conversion of CH4, starting at 400 °C and reaching 22% at 550 °C. By contrast, the oxidation of CH2O occurs at far lower temperatures for all catalysts (Fig. 2, right), whereas Fe8 and Fe25 exhibit the highest activity, with total conversion at just 300 °C. The oxidation of CH2O and CH4 results in the selective production of CO2 at 250 °C and above, while a minor amount of CH3OH (< 10 ppm) is also formed at 200 °C, corresponding to a selectivity of 10%; the mechanism of the CH3OH production is examined in Sect. 3.3. The performance of the samples remains the same upon a following catalytic run and after exposure to the model exhaust for 1 h at 550 °C.

Fig. 2
figure 2

Conversion of CH4 (left) and CH2O (right) on the Fe2O3 catalysts (Fe6 (□), Fe5 (Χ), Fe8 (Δ), Fe12 (◊), Fe25 (○). Conditions: y(CH4) = 1000 ppm, y(CH2O) = 100 ppm, y(O2) = 5 vol%, y(H2O) = 10 vol%, N2 balance, S.V. = 150,000 h−1 (STP)

Full size image

When the activity (Fig. 2) is coupled with the physico-chemical properties of the catalysts (Table 1), it is clear that high BET surface areas and small crystallite sizes enhance the CH4 oxidation, as shown in Fig. S3 (left). This may be related to the number of active sites, which increases as the surface area rises. Indeed, the CH4 oxidation rates related to each surface area are similar; at 550 °C, for instance, they range from 21 to 45 µmol m−2 h−1, substantiating the importance of a large BET surface area to achieve high conversion levels. However, Fig. 3 (left) suggests that the surface reducibility of the catalysts overlaps with the effect of the surface area. The diagram shows that the CH4 oxidation activity increases if the TLT value during CO-TPR is decreased, indicating the relevance of the availability of active surface oxygen. It may be speculated that the oxygen species involved exist in the form of FeO6 entities, which are the thermodynamic coordination polyhedrons of α-Fe2O3.

Fig. 3
figure 3

Relation of the CH4 conversion at 550 (▲) and 500 °C (●) with the TLT value during CO-TPR (left) and relation of the CH2O conversion at 250 (Δ) and 200 °C (○) with the CO2 uptake inferred from CO2-TPD (right); the lines are a guide for the eyes

Full size image

With the amount of active surface oxygen, derived from CO-TPR (Sect. 3.1), and the conversion rate of CH4 (XCH4,550 °C), taken from Fig. 2, the turnover frequency (TOF) of the active FeOx sites is roughly estimated for the CH4 oxidation at 550 °C (TOF = \(\dot{\text{n}}\)CH4,0·XCH4,550 °C/nO,LT). As a result, the TOF values of the Fe2O3 catalysts range from 1·10–3 to 7·10–3 s−1. Compared to Pt sites, for which TOF data at 370 °C between 7·10–3 and 360·10–3 s−1 are reported [20], it is clear that the active Fe sites still exhibit inferior CH4 oxidation efficiency.

Moreover, similarly to the conversion of CH4, the CH2O oxidation is also enhanced by a rising BET surface area (Fig. S3, right). Figure 3 (right) also indicates a relation between rising CH2O conversion and an increasing uptake of CO2. The relevance of the CO2 adsorption for the CH2O oxidation performance of the Fe2O3 catalysts lies in the role of surface formates and carbonates in the reaction mechanism; this is discussed in detail in Sect. 3.4. CH2O and CO2 are expected to adsorb on the same surface sites due to their structural similarity.

3.3 Upscaling the Best Catalyst

Since Fe25 performs best for the simultaneous oxidation of CH2O and CH4 (Fig. 2), this powder sample was selected for the first level of upscaling, a 1″ cordierite monolith. For these catalytic tests, a more realistic model exhaust gas was used comprising CH2O, CH4, CO and NO, and higher proportions of H2O and CO2 (each 12 vol%). The reactant load of the catalyst was clearly higher for the monolith as compared to the powder, i.e., 143 vs. 723 mg(CH4) g−1 s−1, and 27 vs. 90 mg(CH2O) g−1 s−1. As a result (Fig. 4), the monolith showed markedly lower conversions of CH4 and CH2O. For instance, only 7% of CH4 is converted at 550 °C (powder: 22%), whereas the amount of CH2O converted at 250 °C is 13% (powder: 80%); almost complete CH2O oxidation occurs at 400 °C and above. The monolith also exhibits CO oxidation, with a maximum removal of 84% (550 °C), whereas only little oxidation of NO takes place, the maximum NO2 yield being 13% at 500 °C.

Fig. 4
figure 4

Conversion of CH2O (○), CO (Δ), NO (Χ) and CH4 (□) on the 200 cpsi monolith coated with Fe25 (100 g l−1). Conditions: y(CH4) = 1500 ppm, y(CH2O) = 100 ppm, y(CO) = 600 ppm, y(NO) = 250 ppm, y(O2) = 5 vol%, y(H2O) = 12 vol%, y(CO2) = 12 vol%, N2 balance, S.V. = 65,000 h−1 (STP)

Full size image

Due to the high CH2O oxidation efficiency, the catalyst was further upscaled to the real converter level using a 200 cpsi honeycomb substrate (D: 390 mm, L: 450 mm, 85 g l−1) and was evaluated on the stationary biomethane engine (Sect. 2.2). The two engine operation points that were established resulted in temperatures of 485 °C for a 50% load and 465 °C for a 75% load. The reactant load of the catalyst was 580 mg(CH4) g−1 s−1, 239 mg(CO) g−1 s−1 and 27 mg(CH2O) g−1 s−1 for a 50% engine load and 843 mg(CH4) g−1 s−1, 348 mg(CO) g−1 s−1 and 40 mg(CH2O) g−1 s−1 for a 75% engine load. Figure 5 shows that in the fresh state, the catalyst oxidized CH2O completely for both engine loads, whereas the CO conversion amounted to 80% (485 °C) and 60% (465 °C). This high efficiency in the case of CH2O and CO is in line with the results from the laboratory test bench (Fig. 4). Note that no significant CH4 oxidation appeared; this was related to the relatively low exhaust temperatures. Moreover, it should be mentioned that the conversion of CH2O and CO dropped within 21 h, but remained constant after that. The initial decrease in activity is typical for the deposition of ash components on the catalyst, originating from the motor oil [21]. No SOX was detected in the exhaust gas by the FTIR spectrometer (detection limit of approx. 1 ppm). However, even low concentrations in the ppb range might cause a significant accumulation and some deactivation of the catalytic converter. Indeed, SEM–EDX analyses revealed deposits, particularly in the upstream area of the converter, mainly consisting of Ca, Zn, Si, S and P.

Fig. 5
figure 5

Conversion of CH2O (○) and CO (Δ) for the 200 cpsi honeycomb coated with Fe25 (85 g l−1) in the exhaust system of the biomethane engine MWM TCG V12 (dashed lines: 50% load, solid lines: 75% load)

Full size image

In the literature, commercial Pt/Pd catalysts are reported to completely convert CH2O and CO in gas engine exhaust systems at a very similar temperature (475 °C), but with double the gas hourly space velocity, GHSV (80,000–100,000 h−1) [22]. CH4 is not converted.

3.4 Mechanism of CH2O Oxidation

The mechanism of CH2O oxidation, including CH3OH evolution during CH2O light-off, was investigated with Fe25, which is the best catalyst tested (Sect. 3.2). The in-situ DRIFTS studies were performed in a broad temperature range from 50 to 350 °C to provide a comprehensive understanding of the surface reactions. For clarity, selected spectral ranges from 2600 to 3800 cm−1 and 900 to 1800 cm−1 are depicted in Fig. 6, while all the DRIFT spectra are illustrated in Fig. S4.

Fig. 6
figure 6

DRIFT spectra of the Fe25 catalyst after CH2O exposure (50 °C) and subsequent flushing with N2 at 50 °C (grey), 100 °C (dashed grey), 150 °C (dark grey), 200 °C (dashed line) and 250 °C (solid line). Conditions: y(CH2O) = 500 ppm, y(O2) = 5 vol%, N2 balance, F = 200 ml min−1 (STP)

Full size image

Figure 6 shows the DRIFTS data recorded after the CH2O/O2 exposure at 50 and 100 °C, indicating the formation of [CH2O]n oligomers on the catalyst surface associated with the bands at 1406 (δCH2) and 2938 cm−1 (νCH2) [23, 24]. These species are clearly distinguished from the related polyoxymethylene (POM) entities, which are reported to form bands at 1470 (δCH2) and 2921 cm−1 (νCH2) [25]. Additionally, dioxymethylene moieties are not produced due to the absence of features at 1473 (δCH2) and 2882 cm−1 (νCH2) [26]. The oligomerization of CH2O occurs in the presence of bases with the consumption of OH species and the formation of H2O, which is partially adsorbed on the catalyst, as indicated by the DRIFTS band at 1644 cm−1 (δHOH) [19, 23, 27]. The OH groups required for the formation of the [CH2O]n oligomers are provided by the surface of the catalyst. The consumption of the OH sites is substantiated by bands with a negative formal charge between 3550 and 3700 cm−1 (νOH). Moreover, methoxy (OCH3) and bidentate formate species (HCO2) appear [23, 24, 26, 28]. The former are related to the bands at 2970 (νCH2), 2860 (νCH2), 1174 (νCO), 1091 (νa,CO) and 1069 cm−1 (νCO). Due to the same temporal progression, the bands at 2981 (νCH2) and 975 cm−1 (νCO) are also tentatively assigned to methoxy groups. By contrast, the formates are associated with the intense features at 1378 (δCH) and 1355 cm−1 (νs,COO). In addition, carbonate species appear initially at 100 °C, as deduced from the features at 1573 (νCO), 1311 (νas,OCO) and 1285 (νas,COO) cm−1 [19]. The assignments of all the DRIFTS bands are summarized in Table S1.

At 150 °C, the intensity of the bands identifying methoxy groups decreases and CH3OH is detected in the gas phase (Fig. S5). The bands of formate species increase at a temperature of 150 °C. Above 200 °C, methoxy groups are no longer observed, and the intensity of the formate bands decreases as the temperature rises. Simultaneously (at 200 °C and above), the intensity of the OH bands increases between 3550 and 3700 cm−1 (νOH). Surface OH groups are partially regenerated by the reaction of formate species and H2O, leading to the formation of formic acid (HCOOH). The H2O required comes from the CH2O oligomerization and is also present in trace amounts in the feed due to the CH2O supply. It should be noted that in real exhaust gas, a large excess of H2O is available. Traces of HCOOH and H2 are detected in the gas phase at 200 °C and above using FTIR and gas chromatography, respectively. The unstable HCOOH partly decomposes into H2 and CO2. In addition, above 200 °C strong carbonate bands appear at 1573 (νCO), 1311 (νas,OCO) and 1285 (νas,COO) cm−1 [19, 29, 30] and adsorbed H2O (1644 cm−1 (δOH)) [19, 23, 27] occurs, formed by the oxidation of CH3OH or formates, or HCOOH. Apart from the formation of H2O, surface OH groups are also reformed, and at 300 °C they are fully regenerated. Significant thermal decomposition of the carbonate species occurs at 250 °C and above, as indicated by the decrease in the corresponding DRIFTS features (Fig. S5) as well as the detection of gaseous CO2. At 350 °C, no bands are found any longer due to the enhanced desorption of the surface species, indicating that the catalytic cycle is closed.

For reference purposes, DRIFT spectra (Fig. S6) were recorded after CH2O exposure at -100 °C, providing bands at 2980 (νCH2), 2921 (νCH2), 2786 (νCH2), 1470 (δCH2), 1433 (ωCH2), 1382 (ωCH2), 1239 (νCO),1104 (νCO) and 947 cm−1 (νCO), which are all assigned to adsorbed POM [25, 26]. The iron-oxide-assisted polymerization of CH2O is known from the literature, while Brønsted acid OH groups on SiO2 and Brønsted basic sites on MgO are discussed as active sites. Meanwhile, there is a debate on the nature of the sites of Fe2O3 catalysts [23, 26, 31]. Since Fig. S6 reveals negative bands between 3550 and 3700 cm−1 (νOH) during the POM formation, it is suggested that at least OH sites of Fe25 participate in the CH2O polymerization. Additionally, the DRIFT spectra remain unchanged if the temperature is subsequently increased to 0 °C, indicating that the POM moieties are stable. However, when the temperature is further raised to 50 °C, the rapid and complete decomposition of POM takes place (Fig. S4) in line with the DRIFTS investigation reported above for CH2O exposure at 50 °C (Fig. 6). This clearly demonstrates that CH2O does not exist as a stable adsorbate on the Fe25 sample in the temperature range investigated (-100–350 °C).

From the DRIFTS findings, the mechanism for the lean CH2O oxidation on α-Fe2O3 (Fe25) is deduced (Fig. 7), primarily addressing conversion below approx. 350 °C, including the cold start phase of gas engines. The first reaction step involves the adsorption of CH2O, resulting in POM species below 50 °C and [CH2O]n oligomers at 50 °C subsequently forming dioxymethylene (II); the latter is not detected by DRIFTS, which may be related to its high reactivity and/or the superimposition of its vibration bands by those of the related [CH2O]n oligomers. Two adjacent dioxymethylene species are assumed to react to formate and methoxy species in the form of a Cannizzaro-type reaction (III–IV). In the subsequent hydrolysis of the methoxy groups, CH3OH is released at 150 °C and some OH surface groups (V) are reformed, as observed with DRIFTS. Overall, the Cannizzaro mechanism involves the reaction of two CH2O molecules with four OH groups of the catalyst surface, yielding gaseous CH3OH and a formate adsorbate [32]. Above 200 °C, the hydrolysis of formates occurs, leading to the production of HCOOH, detected in the gas phase, and the further reforming of OH groups (VI), as indicated by increase in the OH stretching vibration range and a simultaneous decrease in the formate bands. However, HCOOH decomposes into CO2 and H2, which is partly oxidized to yield H2O. Likewise, above 200 °C the oxidation of CH3OH into CO2 and H2O starts, which is verified by the disappearance of the methoxy groups with a simultaneous increase in the gas phase concentration of CO2. A separate experimental study (in a tubular quartz glass reactor, Sect. 2.2) examining CH3OH oxidation on Fe25 was performed under the same conditions to confirm this statement. An initial conversion of CH3OH to CO2 was evident at 200 °C. Complete conversion already occurred at 275 °C. Since no methoxy groups were present on the surface above 200 °C, the carbonates that are present are primarily formed by the decomposition of formates, which desorb up to 350 °C. Finally, based on the DRIFTS experiments, it cannot be completely excluded that some CH2O species are directly oxidized to create formate, although this route does not explain the appearance of methoxy and CH3OH species.

Fig. 7
figure 7

Mechanism of CH2O conversion on α-Fe2O3 catalyst according to a Cannizzaro-type

Full size image

Moreover, the O2 formally required for the CH2O oxidation according to Eq. 2 is likely involved in the follow-up oxidation of CH3OH and HCOOH (or its decomposition product H2) on the Fe2O3 catalyst.

An additional mechanistic investigation of the CH2O oxidation involved a step function experiment, in which Fe25 was immediately exposed at 150 °C to a gas flow composed of 500 ppm CH2O and 5 vol% O2 (N2 balance). The temporal progression of the relative fractions (molar) of the reactants related to the inlet proportion of CH2O is depicted in Fig. S7. In the first 200 s of the experiment, the incorporation of carbon species is evident (C balance significantly below 1). An initial formation of CH3OH is also observed, with a peak at an exposure time of 365 s, corresponding to a yield of 45%. After 1500 s, neither the incorporation of carbon nor the conversion of CH2O is observed any longer. During the total CH2O exposure lasting 1300 s, 48 µmol CH2O is converted with the formation of 23 µmol CH3OH and the accumulation of 26 µmol carbon. After this experiment, the catalyst was heated to 550 °C in a flow of synthetic air, resulting in the release of 27 µmol CO2. Similar results were obtained in the step function study performed at an exposure temperature at 200 °C (Figure S8). From the reaction mechanism suggested in Fig. 7, it is inferred that the accumulated carbon is related to formate species. This interpretation is in line with the stoichiometry of the Cannizzaro-type mechanism involving the conversion of two CH2O molecules (here: 48 µmol) into one CH3OH (here: 23 µmol) and one formate species (here: 26 µmol). It should be stated that according to DRIFTS studies (Fig. 6), formates only start decomposing above 150 °C, leading to this accumulation in the step function experiment. However, as a consequence of the lacking formate release, the surface of the catalyst is increasingly covered by the formates, while the OH groups are progressively consumed, diminishing the conversion of CH2O and the production of CH3OH as observed in the step function investigation.

4 Conclusions

Out of the systematic series of five α-Fe2O3 samples, the catalyst powder with the highest BET surface area (Fe25) showed the best CH4/CH2O oxidation performance under lean conditions. The investigations showed that α-Fe2O3 catalysts are suitable for the oxidation of lean CH2O (full conversion) and CH4 (up to 25%) above 300 and 550 °C, respectively. Structure–activity relations indicated that for CH4 conversion, the relevant properties driving the catalytic performance are a high BET surface area corresponding to a multitude of active sites, a small crystallite size and the high reducibility of the catalyst surface or availability of surface oxygen. By contrast, for the oxidation of CH2O, the total CO2 adsorption capacity according to the availability of OH adsorption sites is crucial. In accordance with the suggested Cannizzaro-type mechanism, these OH surface groups are involved in the conversion of CH2O into CH3OH and formate species between 100 and 200 °C, including a CH3OH selectivity up to 45%. Thus, CH3OH has to be considered as a significant secondary pollutant, particularly during the cold start phase of the gas engines. However, above 250 °C, CH3OH is no longer emitted, probably due to the inhibition of the Cannizzaro mechanism, associated with the enhanced oxidation of the methoxy species. Furthermore, above 200–250 °C, formates and carbonates become the predominating surface species, therefore their thermal removal is assumed to affect the overall CH2O oxidation rate.

The Fe25 sample was upscaled to the honeycomb level including a 1″ monolith for laboratory tests and a 390 × 450 mm substrate for tests in a real gas engine exhaust system. These investigations with the monolithic catalysts indicated a clearly lower CH2O/CH4 oxidation performance compared to the studies with the catalyst powder. This is related to the higher reactant load of the monoliths. The monoliths also exhibited considerable CO oxidation activity. In the gas engine exhaust system, the catalyst revealed substantial CH2O oxidation, with full conversion at 460 °C. By contrast, the real exhaust temperatures were too low to activate the CH4 oxidation on the monolith. The overall activity of the monolith is lower compared to precious metal catalysts [22, 33]. However, it can be postulated that by moderately increasing the residence time, similar conversions of CO and CH2O can be reached. Higher conversion levels might also be achieved with the Fe-based converter by improving the washcoat loading. Its use in stationary engines with higher exhaust temperatures between approx. 500 and 550 °C is also advantageous, particularly for the effective conversion of CH4.

The findings of the current paper can be used to enhance the performance of iron oxide catalysts. This could be achieved in further investigations by adding thermal promoters (e.g., tungsten and lanthanum) and electronic promoters (e.g., alkali (earth) metals).