Introduction

Soot (black carbon) from diesel engines has become a highly hazardous pollutant which can cause serious environmental and health problems1. Catalytic soot combustion is an efficient after-treatment for soot clean-up and also a relatively complex gas-solid-solid reaction, which has always been thought to operate by an oxygen-transfer mechanism. However, truly conclusive examples are rare. In 1950 the so-called electron-transfer mechanism was put forward, unfortunately no firm proof was available2,3. Although this was mentioned in a review paper in 20013, no further information is available. Therefore, it is important to clarify the two mechanisms and their relationship. In this context catalysts containing alkali metals and potassium (K) in particular were chosen due to their extremely high catalytic soot removal activity4,5,6,7.

In most cases K is present to promote catalytic soot combustion and improve the contact of catalysts with soot and/or enhance the oxidation activity of the catalysts8. In addition K-compounds, for instance, K2O, KOH and K2CO3, can serve as independent catalysts or active components of catalysts for soot combustion and have high activity9. It is well known that as opposed to precious metals and transition metals with variable valence states10,11,12,13,14,15, K is present as K+ in the catalyst, which cannot be reduced or oxidized during the redox reactions. How then can K-containing compounds catalyze an oxidation reaction like diesel soot combustion? Okubo et al. proposed carbonates on the surface of thermally treated K2CO3/Na-nepheline as the active species16,17, possibly following the reaction of K2CO3 + C + O2 → K2O + 2CO218,19. However, this is not a catalytic cycle20. The role of K was suggested to effectively participate in a redox cycle between KxOy and KxOy+ 1, though these K species did not correlate with activity19,21,22,23. Apart from the nature of the active sites, the reaction intermediates play an important role in the elucidation of the reaction mechanism. Ketene groups, a carbon–oxygen complex containing the structure of C = C = O, were first reported as intermediates for soot combustion on K/MgAlO catalysts in our previous work24,25,26. The question is whether ketene species are common reaction intermediates when K+ is present?

In this work, we prepared different K-related species (K2CO3 and KOx) on three typical oxides (MgO, CeO2 and ZrO2) by a single method using K2CO3 as the precursor. Not only the K+ active sites but also a common intermediate were confirmed experimentally and theoretically. Thus an oxygen spillover mechanism including activation of gaseous oxygen, the formation and fate of ketene intermediates and electron transfer processes was proposed.

Results

X-ray powder diffraction (XRD) patterns of K-supported oxides after calcination at 850°C for 2 h and exposure in air (K/MgO, K/CeO2 and K/ZrO2) show typical diffraction peaks of the corresponding single oxide, suggesting that the K species was present as a highly dispersed phase (Supplementary Figure S1, Supplementary Table S1 and S2). However, the normalized absorption spectra of K-edge of K in Figure 1a in K/MgO, K/CeO2 and K/ZrO2 are almost the same as that of K2CO3, showing two prominent peaks at 3612 and 3619 eV, similar to the results reported by Gomilšek et al.27, which indicates that the K species was present as K2CO3. A small shift of the peak at 3619 eV for K/MgO may be associated with the formation of a small amount of K2Mg(CO3)2·4H2O, as observed for the same sample before calcinations (Supplementary Figure S2)24. No changes of oxides were observed after impregnation with K (Supplementary Figure S3). The presence of carbonate species can also be shown by IR spectra. As observed in Figure 1b, peaks at 1370 and 1460 cm−1 were observed, which are assigned to ν1 and ν4 of unidentate carbonate (–O–CO2)28. Furthermore, X-ray photoelectron spectroscopy (XPS) spectra show two peaks at 295.6 and 292.8 eV, identical with that of K2CO3 (Figure 1c). It can be concluded that the K2CO3 phase on K/MgO, K/CeO2 and K/ZrO2 has been confirmed using the described preparation conditions.

Figure 1
figure 1

(a) K K-edge normalized absorption, (b) IR spectra of fresh samples and the samples after O2 treatment at 850°C for 2 h, (c) XPS spectra of K 2p for K/MgO, K/CeO2, K/ZrO2 and K2CO3, (d) CO2–TPD patterns of K/MgO, K/CeO2 and K/ZrO2 heating from 150 to 850°C in He and holding at 850°C for 2 h.

Full size image

Temperature-programmed desorption of CO2 (CO2–TPD) experiments (Figure 1d and Supplementary Figure S4) show that the supported K2CO3 on MgO, CeO2 and ZrO2 has been completely decomposed after heat-treatment at 850°C for 2 h. As observed in Figure 1b, the IR peaks for the samples after O2 treatment at 850°C for 2 h at 1370 and 1460 cm−1, due to carbonates, disappeared. This would result in the formation of another K species, KOx, following the decomposition of supported K2CO3. Combining the above results, two kinds of K species, K2CO3 and KOx, have been successfully produced on the three typical oxides and these can be inter-converted by desorption/adsorption of CO2.

Discussion

Soot combustion in O2 was carried out by temperature-programmed oxidation (TPO) reactions to evaluate catalytic activity, which was expressed in terms of Tm (Figure 2, Supplementary Figure S5 and Supplementary Table S1). It was observed that both the fresh samples and those after O2 treatment exhibited almost the same Tm values. Furthermore, a weak CO2 desorption (insets in Supplementary Figure S5) was observed at higher temperatures (> 700°C) for the fresh samples, which is attributed to the decomposition of K2CO324. However, this is not the case for samples after O2 treatment. This shows that in the TPO experiments over catalysts after O2 treatment, the K-related phase is KOx rather than K2CO3. The TPO results confirmed that the KOx and K2CO3 species displayed almost the same catalytic activity. Specifically, in situ Laser Raman experiments (Supplementary Figure S6) also demonstrated that both K2CO3 and KOx can catalyze soot combustion reactions. In other words, no matter which K species is present, either K2CO3 or KOx in catalysts, the catalytically active sites are identified as K+. This is highly important because K2CO3 was always thought to be responsible for soot combustion16,17. After water-washing, no catalytic activity is observable due to dissolution of the K+ components (Supplementary Table S2)24, confirming that free (isolated) K+ is the active site. In order to exclude the effects of the support on soot combustion activity, SiO2 as an inert substrate was chosen to demonstrate the role of K+. It is showed that when K2CO3 was present (Supplementary Figure S7 and Figure S8), K/SiO2 showed relatively high catalytic activity (Supplementary Figure S9), similar to that of K/MgO, K/CeO2 and K/ZrO2 (Figure 2). Thus, it can be confirmed that K+ acts as an active site rather than a promoter.

Figure 2
figure 2

TPO patterns of COx for soot combustion on pure oxides, K-supported oxides, K-supported samples after treatment in O2 at 850°C for 2 h and K-supported samples after water-washing treatment for 24 h: (a) MgO, (b) CeO2 and (c) ZrO2.

Full size image

Now, it is important to determine the origin of the active oxygen species in catalytic soot combustion when K+ is present. For K-containing catalysts, the oxygen-transfer mechanism is the most important in which gaseous O2 is activated by the alkali metal and then transferred to the carbon surface24. Both Janiak et al.29 and Lamoen and Persson30 proposed that K can enhance the affinity and dissociation of gaseous O2, based on theoretical calculations. A similar view was presented by Jiménez et al. who found that the active oxygen on the catalyst for soot oxidation was increased by the presence of K31. However, the active oxygen species cannot be detected by temperature-programmed reduction (TPR) with H224. The existence of the active oxygen species needs to be confirmed by carefully designed experiments.

Soot–TPR results show that a certain amount of soot can be oxidized in the absence of O2 (Figure 3a), suggesting the existence of active lattice oxygen in K/MgO, K/CeO2 and K/ZrO2. The activity of active lattice oxygen, as seen in Figure 3a inset, follows the order of K/ZrO2>K/CeO2>K/MgO, which is in the same order as the Sanderson electro-negativity of the corresponding supports (Supplementary Table S1). Because strong electro-negativity means strong electron attraction, strengthened chemical bonds of pure oxides with K+ were formed, leading to weakening of the K–O bond in the catalysts. The Soot–TPR result is in good agreement with that of Tm from TPO (Supplementary Table S1, Figure 2 and Supplementary Figure S5). Additionally, it is noted that the strong CO2 signal below 700°C and CO signal above 600°C for the bulk K2CO3 are ascribed to the reactions of K2CO3 + C → K2O + CO2 and K2CO3 + 2C → 2K + 3CO or K2O + C → 2K + CO, respectively18,32.

Figure 3
figure 3

(a) Soot–TPR for K/MgO, K/CeO2, K/ZrO2 and K2CO3 after O2 treatment at 850°C for 2 h. The inset in (a) is the partially enlarged figure at low temperature range; (b) In situ IR spectra of NO adsorption (1000 ppm NO + He) on MgO and K/MgO at 100°C after O2 treatment in 5 vol.% O2 + He at 500°C for 30 min.

Full size image

In situ IR of NO adsorption was also performed, in which NO is used as a probing molecule, because NO can be efficiently stored on K+ cation sites only after NO has been oxidized to NO233,34. First, the catalysts were pre-oxidized and then exposed to NO (Figure 3b and Supplementary Figure S10). A strong and stable peak at 1248 cm−1 is observed, which can be attributed to nitrite species derived from the oxidation of NO by the active oxygen in the catalysts. However, over the corresponding potassium-free supports, only negligible peaks were present, possibly due to weak NO oxidation and adsorption. This is evidence of activation of gaseous O2 on the active K+ sites.

The role of activated O2 was further confirmed by isothermal anaerobic titrations35, in which the O2 flow was turned on at first and then turned off during the catalytic soot combustion process (Supplementary Figure S11, Supplementary Table S1). The results showed that once the O2 flow was stopped, the soot combustion activity gradually decreased, confirming the participation of active oxygen (O*) derived from gaseous O2 in real reaction conditions (the O* amounts are listed in Table S1). Similar results have been demonstrated on Li-doped MgO, on which the [Li+O−] active sites were formed from the interaction of Li+ with molecular oxygen, which were responsible for the activity of methyl radical formation36. Moulijn and Kapteijn proposed that oxygen-containing reactant molecules were incorporated in or dissociated by a K-oxide cluster to produce an O* species with a relatively high reactivity for carbon and this oxygen species could be exchanged extremely quickly by gaseous oxygen-containing reactants, which is an example of oxygen spillover37.

In situ IR experiments for soot combustion were carried out over K/MgO, K/CeO2 and K/ZrO2 (Figure 4a and Supplementary Figure S12). A characteristic IR band at 2162 cm−1 can be clearly observed, accompanied by the band at 2358 cm−1 and a series of bands in the range of 1000–1800 cm−1, which can be attributed to ketene species, physically adsorbed CO2 and carbonate species, respectively. The assignment of ketene species can also be supported by density functional theory (DFT) calculations. Given that the true structure of the ketene intermediates is not known and the ketene group is formed from free carbon atoms of soot which possess structures similar to polycyclic aromatic hydrocarbons. The geometry of the quinonoid-ketene molecular complex with K+ was optimized at the dispersion-corrected DFT level using the quantum program package Turbomole V6.438 and the IR vibration frequencies were thus calculated. The optimized structures are depicted in Figure 4b. The asymmetric stretch frequency of the ketene group for the complex quinonoid-ketene molecule has been calculated to be 2168 cm−1, which is in good agreement with the experimental value, confirming the assignment of the peak at 2162 cm−1 to the ketene group is reasonable.

Figure 4
figure 4

(a) In situ IR spectra for soot combustion in a flow of 5 vol.% O2 + He on K/MgO; (b) The optimized complex geometry of the quinonoid ketene molecule complex with K+; (c) Illustration of the unified oxygen spillover and the common electron transfer process for soot combustion on potassium-supported oxides.

Full size image

In view of the fact that the IR bands of the ketene group and CO2-related species appeared simultaneously, the ketene species were deduced to be reaction intermediates during the soot combustion process. Further reaction of the ketene species was demonstrated by transient reactions and the corresponding ex situ IR, in which K/MgO was taken as the representative catalyst and the mixture of soot and K/MgO (1/9 weight ratio) was heated in O2 to 350°C, then the O2 flow was stopped (Supplementary Figure S13). In the 1st stage, soot was oxidized to COx in O2 while the temperature was increased to 350°C (a and b). In the 2nd stage the flow of O2 was stopped at 350°C and the CO concentration sharply dropped to zero while the CO2 concentration declined slowly, implying that some surface active oxygen on the catalyst transferred to the ketene group. As the evolution of CO2 decreased to a negligible level, the ketene group disappeared, which can be shown from the vanishing of its characteristic peak at 2162 cm−1 in the inset in Figure S13 (c). These facts strongly support the transformation of the ketene species to CO2 by active oxygen24. Since the appearance of the ketene species is independent of the catalyst support, it should be a unified reaction intermediate for K-supported catalysts. As shown (Figure 4c), the K+ pulls O2 from the gas phase and active oxygen species were obtained by the formation of KOx ([K+O−]). The activated oxygen transfers to free carbon sites where the ketene intermediate is formed. This is further oxidized, by active oxygen, to form the end product CO2. This is a typical oxygen spillover mechanism proposed by us24 and others39, in which the catalyst, as an oxygen carrier, can promote the transfer of oxygen from the gas phase to the carbon surface, by means of the formation of an intermediate compound.

Most importantly in this work, the electron-transfer mechanism was proved for the first time both by ultraviolet photoelectron spectroscopy (UPS) experiments and by theoretical calculations. The UPS spectra gave direct evidence of the changes in the electronic structures of soot due to the presence of the K+ ions. As shown in Figure 5a, pure soot shows a broad UPS peak at approximately 9.0 eV, which is assigned to valence electrons of the p–σ bands on graphite soot40,41. This peak became weaker when soot was mixed with K/ZrO2 (Figure 5a) or K/CeO2 (Supplementary Figure S14a). However, the peak intensity of the mixtures of soot with supports (ZrO2 or CeO2) was nearly unchanged, indicating that the perturbations in the electronic properties of soot occur only in the presence of K+ ions42. The absence of photoemission signals of soot for the mixtures of soot + K/MgO and soot + MgO is possibly due to the relatively low electrical conductivity properties of MgO (Supplementary Figure S14b). The perturbations in the electronic properties of soot mixed with K–supported samples can be illustrated by semi-empirical theoretical calculations. The contour plots of net charges for the soot model (graphene) and the mixtures of soot + catalyst are given in Figure 5b and c, respectively, which clearly describe the changes in the electronic structures due to the presence of K+. The distribution of charges on pure soot is relatively homogeneous and the net charge is near to zero. When the K+ ions are present, the net charge of the edge carbons is substantially changed. As shown in Figure 5c, the edge charges on soot become negative while the inner charges are mainly positive, regardless of the locations of the K+ ions. This demonstrates an important role of the K+ ions that attract the electrons from the inner carbon atoms to the edge carbon atoms and this is in agreement with the calculations of Yang et al.43 The electron-rich carbon atoms favour donating electrons to the electrophilic species such as oxygen molecules to form active oxygen species such as O2− and O−44. In other words, the K+ ions facilitate the concentration of electrons on the soot surface with higher energy states, strengthening the driving force for efficient electron transfer from soot to O245,46.

Figure 5
figure 5

(a) UP spectra of soot, K/ZrO2 and soot + K/ZrO2; (b) Graphene as the model structure of soot. Three locations of the K+ ions near to the structures are labeled. Label p indicates the top location perpendicular to the plane of the soot model, while Labels z and a represent the locations on the same plane as the soot model, near to the zigzag edge and armchair edge of the soot, respectively; (c) Contour plots of net charges calculated by semi-empirical methods for soot and soot in the presence of the K+ ion at the locations of p, z and a as referred to in (b).

Full size image

The electron transfer and the oxygen spillover mechanisms can be effectively integrated by the interaction of soot with the K+ ions on K-supported catalysts47. On the one hand, the K+ ions act on π electrons of soot and covalent K–C bonds may be present, leading to electron transfer from soot to the electronegative oxygen, thus decreasing the aromatic character of soot and activating gaseous oxygen3,37; On the other hand, the activated oxygen spills over from K+ sites to soot and ketene species are formed, which weaken the neighbouring C–C bonds3,37 and the product COx is evolved. The transfer of electrons from soot to K+ was realized by way of oxygen species transferred from KOx ([K+O−]) to soot.

In summary, for soot combustion on the K-supported catalysts, the following three conclusions have been made: (1) the catalytically active site has been identified as free K+ rather than K2CO3; (2) the ketene intermediate has been found to be common to these processes; (3) the oxygen spillover mechanism has been interpreted as an intrinsic electron transfer process on an atomic scale through the active K+ sites.

Methods

Catalyst preparation

The catalysts were prepared by impregnating single oxides (MgO, CeO2 and ZrO2) with the aqueous solution of K2CO3. Prior to the preparation, the oxides were heat-treated at 850°C for 2 h. Their suspensions in the aqueous solution of carbonate salt were evaporated while being stirred at 90°C until achieving a paste, which was then dried at 120°C overnight and calcined at 850°C for 2 h. In this way, the obtained catalysts with different support were designated as K/MgO, K/CeO2 and K/ZrO2. According to our previous work24, the weight loading amount of K is determined as 8 wt.%. The as-prepared samples are also called as the fresh catalysts. Those after further O2 treatment at 850°C for 2 h are denoted as K/MgO–O2, K/CeO2–O2 and K/ZrO2–O2, respectively. While the samples after water-washing treatment were denoted as K/MgO–w, K/CeO2–w and K/ZrO2–w, respectively, which were obtained by stirring the suspension of the fresh catalysts in the deionized water, then filtering, drying at 120°C overnight and calcinations at 850°C for 2 h.

Characterizations

Powder XRD patterns were recorded on a Rigaku D/max-rc diffractometer. Surface area and pore size distribution were determined by N2 adsorption-desorption at 77 K with the BET method using a Micromeritics ASAP 2020 instrument after outgassing at 300°C for 5 h prior to analysis. XAFS measurements for the K K-edge were performed on the XAFS station of Beijing synchrotron radiation facility (BSRF, Beijing, China). K K-edge (3608 eV) data were collected at the 4B7A beam line of the Spectra in fluorescence mode with a Si (Li) detector. IR experiments were carried out using FT–IR spectrometer (Bruker Tensor 27) over 400–4000 cm−1 after 32 scans at a resolution of 4 cm−1. The samples were diluted with KBr in the ratio of 1:100. XPS data were obtained on an AXIS–Ultra instrument from Kratos Analytical using monochromatic Al Kα radiation (225 W, 15 mA and 15 kV) and low-energy electron flooding for charge compensation. To compensate for surface charge effects, the binding energies were calibrated using the C 1s hydrocarbon peak at 284.80 eV. X–ray Fluorescence (XRF) experiments were performed on a ZSX Primus II instrument from Rigaku. Inductively Coupled Plasma-Atomic Emission Spectrometer (ICP–AES) experiments were carried out on the IRIS Intrepid IIXSP instrument from Thermo elemental. CO2–TPD experiments were carried out in a fixed bed micro-reactor consisting of a quartz tube (6 mm i.d.). A 50 mg catalyst was pretreated in He (100 mL/min) at 850°C for 1 h and then cooled down to 250°C in He. When the temperature was stabilized at 250°C, 3976 ppm CO2 in He (100 mL/min) was introduced. After saturation, the flow was switched to He (100 mL/min) to flush the sample to remove the weakly adsorbed species at 250°C and then cooled down to 150°C. Desorption was then conducted by heating the catalyst from 150 to 850°C at a ramp of 10°C/min in He (100 mL/min). At 850°C, the sample was isothermally heated until the completion of CO2 desorption. The desorbed CO2 was detected by a quadruple mass spectrometer (MS, OminiStar 200, Balzers). The amount of CO2 adsorbed at 250°C was calculated by the integration of the CO2 desorption peaks. In situ Raman spectra were measured using a Raman spectroscope (HR800) with a CCD camera. The 632.8 nm line of a He–Ne laser was used to simulate the Raman spectra. The measurements were carried out with a microscope by using a ×50 objective lens (focus diameter larger than 1 micron) and the data were recorded in a backscattering geometry. Use of the cell allowed control of the sample temperature in static air at a heating rate of 5°C/min. UPS characterization was carried out using a HeI emission lamp (21.22 eV) as an excitation source and an analyzer resolution of 0.025 eV.

Activity tests

Temperature-programmed oxidation (TPO) reactions were conducted in the fixed bed micro-reactor. Printex–U from Degussa is used as the model soot. The soot was mixed with the catalyst in a weight ratio of 1:9 in an agate mortar for 30 min, which results in a tight contact between soot and catalyst. Isothermal reactions and isothermal anaerobic titrations were carried out to obtain the number of active redox sites (O* amount) and turnover frequency (TOF). Soot–TPR experiments were performed as the carbothermal reduction in the absence of gas phase oxygen in a fixed-bed flow reactor. The details for TPO experiments, isothermal reactions and isothermal anaerobic titrations, as well as Soot–TPR experiments are provided in supporting information.

In situ IR experiments

Soot combustion was further investigated using in situ IR spectroscopy. The IR spectra were recorded on the FT–IR spectrometer (Bruker Tensor 27) over 400–4000 cm−1 after 32 scans at a resolution of 4 cm−1. Additionally, in order to confirm the existence of active oxygen species, in situ IR experiments for NO adsorption were performed. The experimental details are provided in supporting information.

DFT calculations and Semi-empirical quantum chemistry calculations

The geometry of the complex of quinonoid ketene molecular and K+ was optimized at DFT levels using the well-known B3LYP hybrid exchange-correlation functional together with Ahlrichs split valence plus polarization (SVP) basis set for all atoms. A semi-empirical quantum chemistry program, MOPAC (Molecular Orbital Package) version 2012 was used to calculate the net charge of model soot based on NDDO (neglect of diatomic differential overlap) approximation. The program has been updated with a new and more accurate parameterization (PM7) for all the main group elements and transition metals. The details of calculation are provided in supporting information.