1 Introduction

A key element for mitigating Diesel vehicle pollution is the Diesel oxidation catalyst (DOC), which is typically located directly after the engine. DOC not only oxidizes CO and hydrocarbons (HC) into inert H2O and CO2, but it also converts NO to NO2. High NO2 yield is essential for the efficient abatement of NOx in the selective catalytic reduction catalyst placed downstream as well as for the so-called passive soot oxidation in the particulate filter device, typically directly downstream the DOC [1]. The DOCs come in the form of a catalyst coated on ceramic or metallic monolith substrates [1]. The catalyst contains platinum group metals (PGMs) as the active component and most often consists of Pt-Pd nanoparticles supported on Al2O3. While Pt and Pd are both known to be active for CO and HC oxidation, NO oxidation takes place exclusively on metallic Pt [2,3,4,5].

According to current emission regulations, heavy duty Diesel vehicles need to be emission compliant for at least 700,000 km [6] and the mileage of compliance is expected to increase in future regulations. Therefore, robustness and stability of the DOC is a crucial requirement and it is of great interest to understand and prevent its deterioration. The DOC suffers from gradual deactivation upon mileage due to thermal and poisoning effects. The catalyst is exposed to high temperatures during fast driving or particulate filter regeneration events that lead to sintering of active PGMs as well as the support. Regarding poisoning, the DOC performance can be compromised by the accumulation of phosphorus (P) and sulphur (S) compounds, which stem from fuel impurities and oil additives such as zinc dialkyldithiophosphates [1, 7,8,9]. S poisoning results in the sulphation of the PGM particle surface as well as the alumina support [10,11,12]. With the introduction of ultra-low-sulfur Diesel however, the impact of S poisoning has significantly decreased and P is often the poison present in highest concentration in used catalysts. Deactivation by P is proposed to occur due to reaction with alumina forming phosphate species, as well as due to direct coverage of active surface and changes in PGM redox properties [1, 8, 13,14,15,16,17,18].

An obstacle to fully understand the deactivation phenomena is the fact that thermal aging and poisoning effects overlap [9, 17, 19,20,21,22,23,24] making it difficult to separate individual and combined aging effects. High temperatures and S exposure for example, are reported to promote PGM sintering [9, 20, 25,26,27,28,29], while the presence of P prevents particle growth [17, 21]. Recent publications reveal that the presence of P also influences the degree of sulphation and modifies the deactivation path [24, 30, 31]. Understanding the aging becomes even more challenging if we consider the evolution of the DOC technology in recent years. The Euro VI regulations approved in 2014 considerably limited the admissible NOx emissions in Diesel vehicles [6] and much effort has been devoted since to improve the NOx removal in Diesel aftertreatment systems. Hence, modern DOCs comprise increased Pt content to promote NO oxidation and minimize NOx emissions; in addition, zeolites are often included to function as “cold traps” to minimize HC emissions during the engine cold start [32,33,34]. Importantly, a common approach to optimize DOC performance is to deposit multiple washcoats on the monolith creating layers of different composition and functionality [35, 36]. In spite of their commercial relevance, little has been reported on the aging of multilayer catalysts and the rationalization behind the deactivation of different layers is still missing.

On this line, our recent work investigated the aging of a commercial DOC composed of two catalytic layers [24, 31]. Studies on real and artificial aging revealed a stronger deactivation of NO oxidation performance while CO and HC oxidation reactions were less compromised. A hypothesis behind this behavior is that the top layer - Pt rich and with NO oxidation functionality - may be more susceptible to P exposure and thermal aging. To shed some light into the deactivation of commercial DOCs, the present investigation offers an approach towards understanding and separation of the aging phenomena in bilayered catalysts. For this purpose, the individual and combined effects of P poisoning and thermal treatments were evaluated in two monoliths: (1) the commercial DOC formulation comprising two layers (where the top layer is richer in Pt and holds NO oxidation functionality) and (2) a model formulation comprising only the same top layer of the commercial DOC. The comparison of activity and characterisation results after the treatments delivers new insights into the aging phenomena of layered DOC formulations.

2 Experimental

2.1 Materials and Methods

2.1.1 Catalysts

Two full size DOC monoliths (washcoated 400 cpsi cordierite honeycombs) were studied:

  1. a)

    Commercial DOC (Mc) with a washcoat loading of ca. 100 g L-1 (PGM content of 22 g ft− 3) and composed of two layers. Both layers contained Pt and Pd as the active elements and Al2O3 as the support. The top layer was richer in Pt and NO oxidation to NO2 is believed to occur exclusively in this layer. The bottom layer also contained a zeolite component providing hydrocarbon storage functionality.

  2. b)

    Model DOC (Mm), equivalent to the commercial monolith but containing the top catalytic layer only. Hence, the total washcoat loading was half of that of Mc and the precious metal content was ca. 13 g ft− 3. Mc and Mm were prepared by the catalyst manufacturing company using the same cordierite, precursors and coating procedures for both samples.

2.1.2 Treatments

Four rectangular segments were extracted (5.0 × 1.2 × 1.7 cm) from the centre of each monolith. To ensure stable catalytic performance, all segments were degreened under a continuous feed of simulated Diesel exhaust (10 vol% O2, 7 vol% H2O, 1000 ppm NO, 150 ppm_C1 C3H6/C3H8 (2:1), 500 ppm CO, balance N2) at a gas hourly space velocity (GHSV) of 65,000 h− 1 at 525oC for 4 h. The degreening was performed in the same setup used for activity testing. After degreening, the four segments of each monolith type were subjected to the following treatments:

  1. a)

    Thermal (labelled as T600 or T750): heating under the simulated Diesel exhaust (GHSV = 65,000 h-1) at 600oC for 20 h or at 750oC for 16 h.

  2. b)

    Poisoning (labelled as P): by immersing a segment in an aqueous solution of H3PO4 (Fluka, 85%) for few seconds followed by drying in static oven at 110oC. The H3PO4 concentration in the solution was adjusted to aim for 9.6 wt% P in the coating. The difference in washcoat loading between Mc and Mm was taken into account for the poisoning procedure to ensure equivalent P content in the catalysts. Note that 9.6 wt% in the washcoat corresponds to ~ 1.2 wt% (in Mm) and 2.4 wt% P (in Mc) for the coated catalysts (i.e. cordierite + coating). Considering that previous publications on real aged DOC report 0.1–5.3 wt% P in coated catalysts, the used poison content in this study is within the expected range in real application [24, 37].

  3. c)

    Poisoning and thermal treatment (PT600 or PT750): P poisoning was performed following the procedure described in b); then thermal treatment was performed at 600 or 750oC following the procedure in a).

The resulting samples are denoted by indicating the type of monolith used (Mc or Mm) followed by the treatment performed in such segment (Deg, T600, T750, P, PT600 or PT750). For example, the commercial catalyst segment after P poisoning will be referred to as Mc_P.

2.1.3 Activity Testing

After each treatment, the segments were wrapped with ceramic fibre and placed in a horizontal quartz tube reactor operating at ambient pressure. The outlet gas concentration was analysed using a Fourier transform infrared spectrometer (Nicolet Antaris IGS) equipped with a 2 m gas cell heated to 180°C. The temperature in the reactor was recorded with a thermocouple placed in the sample inlet. All catalytic tests comprised three consecutive heating-cooling cycles under simulated Diesel exhaust at GHSV of 65,000 h− 1.

Fig. 1
figure 1

Schematic representation of the approach followed for treatments of DOC segments

Full size image

In the first cycle, the binders present in the ceramic fibre were burned off when heating the sample to 500oC at 15oC min− 1. After dwelling for 10 min at 500°C, the temperature was decreased to 100oC at the same rate. The pretreatment was followed by two consecutive activity cycles, each performed by heating from 100oC to 500oC and then cooling to 100oC at 5oC min− 1. All the results presented in this work correspond to the data collected during the heating ramp of the second activity cycle. The treatment and activity workflow described above is summarised in Fig. 1.

2.2 Characterisation

The elemental composition of the samples was determined by inductively coupled plasma optical emission spectrometry (ICP-OES) using a Varian Vista AX spectrometer. The samples were dissolved using hydrofluoric acid. All calibration solutions were prepared using 99.999% pure chemicals. Nitrogen physisorption measurements were performed at -196oC in a Quantachrome Autosorb-1 instrument. The sample (ca. 150 mg) was outgassed at 350oC overnight under high vacuum prior to the sorption measurements. The Brunauer–Emmett–Teller (BET) equation was applied within the p/po range of ca. 0.05–0.30 and used to calculate the specific surface area. The pore size distribution was determined by the Barrett-Joyner-Halenda (BJH) analysis of the isotherm desorption branch. For the scanning electron microscopy and energy dispersive spectroscopy (SEM-EDS) analysis samples were embedded at room temperature in EpoFix resin. The samples were subsequently polished and coated with 10 nm layer of Au in an ion sputter (Leica EM ACE200). The maps were acquired in a ZEISS Ultra-55 Scanning Electron Microscope using an EDAX APOLLO XV Silicon Drift Detector to provide EDS functionality at an accelerating voltage of 12 kV and a working distance of 8.5 mm. For transmission electron microscopy (TEM), the catalyst washcoat was carefully scraped out from the monoliths and finely ground in an agate mortar. The powder was then suspended in ethanol and a drop of the suspension was deposited on a copper grid coated with a holey carbon film. High-angle annular dark-field (HAADF) images were recorded on a FEI Technai F30 FEG scanning transmission electron microscope equipped with super twin lens and Schottky type emitter at an accelerating voltage of 300 kV or a probe corrected JEOL ARM200F (NeoARM) equipped with a cold FEG at an acceleration voltage of 200 kV. From the electron micrographs of each sample, the particle size of at least 200 particles were evaluated using the ImageJ software [38]. The particles observed had a close to circular shape, the diameter of such circles was taken as the particle size. Using OriginLab software, the particle diameters were then grouped into evenly spaced bins of 1.5 nm wide. This was defined using the Rice estimator (nw = 2n1/3) and following the work-flow proposed by Alxneit [39].

3 Results and Discussion

The catalytic activity of Mc (commercial bilayered monolith) and Mm (model monolith containing only the Pt-rich top layer) was evaluated by ramping up to 500oC under simulated Diesel exhaust. The activity was tested after degreening and after each phosphorus and/or thermal treatment (either 600oC or 750oC) as shown in Fig. 1. Figure 2 summarises the results by comparing the temperature values at 20% conversion for CO (T20_CO, Fig. 2a) and HC (T20_HC, Fig. 2b) as well as the values for the maximum NO conversion (Max_NO, Fig. 2c). Tabular data of the results can be found in Table 1. While T50 is typically reported to compare catalytic activity, here T20 values are used; the C3H6/C3H8 mixture 2:1 as HC feed makes T20 a more robust for comparison among samples as indicated in our previous studies [24, 31].

The T20_CO values for Mm_Deg (151°C) and Mc_Deg (154oC) were very similar. Mm_Deg and Mc_Deg also showed close values of T20_HC (170 and 171oC, respectively) and Max_NO (60% and 63%). The comparable performance of the two monoliths in the degreened state suggests that the contribution of the top layer dominates the catalytic activity of the bilayer sample. This is expected in views of the higher Pt content in the top layer, which gives this layer with most of the DOC functionality.

The thermal treatment at 600oC resulted in the deactivation of the DOC performance in both monoliths. This is evidenced by the rise of T20_CO and T20_HC values compared to both Deg references as well as by the decreased Max_NO values. Thus, Mm_T600 and Mc_T600 showed T20_CO of 161oC and 163oC respectively, T20_HC values of 178oC and 179oC, and Max_NO of 49% and 48%. The similar results obtained for Mc_T600 and Mm_T600 point to an equivalent degree of deactivation in both samples upon the thermal treatment.

Table 1 Summary of catalytic activity and characterisation results for Mc (commercial) and Mm (model) monolith segments after poisoning and/or thermal treatments
Full size table

On the contrary, the addition of 9.6 wt% P affected Mm and Mc to a different extent. The results in Table 1; Fig. 2 suggest that P poisoning was more detrimental for CO and HC performance in the one layer sample (i.e. higher T20_CO and T20_HC values for Mm compared to Mc). Nonetheless, very similar NO_max values were obtained in both monoliths indicating similar degree of deactivation in the NO oxidation performance.

Analogous outcome was observed in the subsequent PT600, T750 and PT750 treatments. In all cases, CO and HC oxidation suffered from stronger deactivation of Mm (higher T20_CO and T20_HC values than for Mc) while the impact of the treatments on NO oxidation performance was equivalent in Mc and Mm.

Fig. 2
figure 2

(a) T20_CO, (b) T20_HC and (c) NO_Max values of Mc and Mm. Sample codes are specified in the text

Full size image

3.1 Characterisation

To confirm that H3PO4 treatments led to the aimed poison content, Mm_PT600 and Mc_PT600 were measured by ICP-OES and the results show overall P amounts of 1.2 wt% and 2.4 wt%, respectively. As ICP-OES measurements were conducted with crushed sections, the samples included the cordierite substrate as well as the washcoat. Considering the washcoat loadings on both monoliths, the obtained values translate in P contents of 9.7 wt% (Mm_PT600) and 9.6 wt% (Mc_PT600) in the washcoat of and confirm equivalent levels of poisoning.

The latter calculation works under the assumption that all the P deposits on the catalyst and not on the cordierite. To verify this, the P distribution was evaluated by SEM-EDS. The images in Fig. 3 - corresponding to cross sections of monolith channel corners - illustrate that P was homogeneously distributed in the washcoat in both monoliths. Importantly, it accumulated exclusively in the coating and no poison was found in the cordierite substrate. We have shown previously that formation of aluminium phosphate (AlPO4) is responsible for the diffusion of P through the washcoat [24].

To gain insight into the physicochemical properties of the two monoliths after each treatment the various sections were characterised by N2 physisorption (surface area, SABET) and transmission electron microscopy (TEM, PGM particle morphology).

Fig. 3
figure 3

SEM images and the corresponding EDS element maps for P of cross sections of channel corners (normal to the monolith channels) of (a) Mm_PT600 and (b) Mc_PT600

Full size image

The SABET values calculated from the N2 physisorption isotherms are given in Fig. 4a; Table 1. The results reveal that Mc exhibited higher surface areas than Mm; for example, Mc_Deg showed a SABET value of 55.5 m2 g− 1 while SABET of Mm_Deg was 20.3 m2 g− 1. The higher surface area of Mc is ascribed to the presence of the microporous zeolite which is a component of the bottom layer. Note that the SABET was also measured for the sections after crushing in a mortar and thus the analysis comprises the cordierite, which has a low surface area of ca. 5 m2 g− 1. The possible variations of the amount of cordierite present after crushing (typically ca. 75–90 wt%) compromises the accuracy of the analysis. Nonetheless, the general trends in Fig. 4a illustrate that the performed treatments decreased the surface area in Mm and Mc. The drop in SABET upon thermal and P treatments is ascribed to the sintering of the support material and the formation AlPO4 phases [24, 37, 40]. No sintering is observed for Mc_T600 (SABET = 56.3 m2/g) compared to Mc_Deg (SABET = 55.5 m2/g), while same treatments lead to sintering in the one layer counterpart (20.3 m2/g in Mm_Deg and 13.1 m2/g in McmT600); this suggests that the zeolite - which dominates the surface area in the bilayer catalyst - shows higher thermal stability than the Al2O3. The pore blockage as a result of surface area decrease is believed to contribute to DOC deactivation [13], which occurs due to the trapping of active particles (encapsulation) thus preventing them from interaction with the reactants. In case of P poisoning, deactivation can also occur via additional pathways such as direct blockage of the active PGM surface or changes in the redox properties [24].

Fig. 4
figure 4

(a) Specific surface area and (b) average PGM particle size derived from analysis of HAADF-STEM images of Mc and Mm before and after various treatments. Sample codes are specified in the text

Full size image

Barrett-Joyner-Halenda (BJH) analysis of the N2 desorption branches (Fig. 5) revealed a different pore size distribution of Mc and Mm in the mesopore range (20–500 Å). Mm exhibited pores with average size of ca. 130 Å (Fig. 5a), while Mc presented a bimodal pore size distribution with additional pores with an average diameter of ca. 330 Å. These results enable us to assign the small pores to the top layer and the large pores to the bottom layer present in Mc.

Fig. 5
figure 5

Pore size distribution of (a) Mm and (b) Mc derived from N2 physisorption isotherms (BJH method)

Full size image

Consistent with SABET, the pore size distribution plots result in decreased intensity upon P or/and T treatments. Interestingly, the behaviour of the relative intensities of the two types of pores in Mc (Fig. 5b) suggests a more pronounced drop in the fraction of the small pores in response to the P treatment providing evidence for pore blockage by reaction of P with the catalyst and indicating this is more severe in the top layer. This effect can be in part explained by the formulation of the coated layers. Considering the fact that the top layer is composed mainly of Al2O3, this support reacts with P leading to AlPO4 formation; in contrast, such reaction should occur to a lower extent in the bottom layer due to the presence of the other components (i.e. zeolite).

The changes in PGM morphology are known to have a strong impact on catalytic activity [1]. The average particle size of Mm_Deg and Mc_Deg (Fig. 4b; Table 1) obtained from TEM images was 6 and 7 nm, respectively. Both samples exhibited equivalent particle sintering upon thermal treatment at 600oC resulting in particle growth to ca. 13 nm for both Mm_T600 and Mc_T600. This was also the case after the treatments at 750oC, which lead to 20 nm and 24 nm particles in Mm_T750 and Mc_T750, respectively. Note that the thermal sintering lead to wide particle size distribution as shown in Fig. 6.

Fig. 6
figure 6

Distribution of particle sizes obtained for Mm and Mc catalyst in degreened form as well as after each of the phosphorus or/and thermal treatments

Full size image

It has been widely observed in the literature that Pt-rich formulations are more prone to thermal sintering while the addition of Pd (to form mixed Pt-Pd particles and alloyed particles) is known to improve the thermal stability of PGM particles [41,42,43]. The increased stability has been attributed to the formation of a Pd-rich shell, which prevents particle sintering. Our transmission electron microscopy data reveal that such core shell structures are indeed present in Mc. As shown in an example particle in Fig. 7a, large particles exhibit a clear contrast between the particle core (composed of Pt and Pd) and the shell (Pd only) caused by the different distribution of the two elements in the particle. For smaller particles TEM-EDS analysis was performed to evaluate the Pt/Pd ratio across a particle. The results in Fig. 7b suggest that while Pt and Pd enriched domains are present there is no homogeneous core-shell like distribution among the particles. This is evidenced in examples given with some particles showing a core enrichment with Pt (Fig. 7b left) and others with Pd (right).

Considering the presence of core shell-like structures in the samples and the increased Pt content in the top layer of Mc, a higher degree of PGM sintering was expected in Mm (containing only the Pt-rich top layer) compared to Mc. The equivalent degree of sintering measured in both samples from the average particle size needs to be further assessed as our TEM analysis of Mc did not allow to distinguish particles present in the top layer from those in the bottom layer. Further evaluation of a catalyst containing only the bottom layer would be needed to unequivocally assess the degree of thermal sintering in each layer.

Regarding the effect of poisoning on particle morphology, the P treatment alone (i.e. Mc_P and Mm_P) showed no significant changes compared to the degreened states, and Mc_P and Mm_P exhibited particle sizes of 6 and 7 nm, respectively. The combination of P and thermal (600oC) treatments resulted in milder sintering than for samples subjected only to thermal treatment; thus, particle sizes of ca. 8 nm were obtained for Mm_PT600 and Mc_PT600. This agrees with previous observations [21, 24] demonstrating that the presence of P prevents the sintering of PGM particles. This can be explained by the particle encapsulation in line with the surface area decrease in Fig. 4a, which hinders PGM mobility and sintering. This appears to apply also to bimetallic Pt-Pd particles as in Mm and Mc catalysts.

Fig. 7
figure 7

(a) HAADF-STEM of a selected large PGM particle exhibiting difference in contrast between the core and the shell (top) and the point EDS spectra showing emission from Pd only in the shell, and emission from Pt and Pd in the core; and (b) TEM image of a selected smaller particles (top) and the resulting EDS analysis (bottom) performed along the particle diameter showing normalized Pd (red) and Pt (blue) signals. Si, Al, Cu, O and C signals in panel a correspond to the catalyst support and the copper grid

Full size image

Similar conclusions can be drawn from the results obtained from the thermal treatment at 750oC. As mentioned above, strong sintering of active particles occurred in the absence of P. In contrast, the same treatment but following P poisoning resulted in milder particle growth to only 8 and 9 nm for Mm_PT750 and Mc_PT750, respectively.

Note that the activity results (Fig. 2; Table 1) suggest that P poisoning was detrimental for catalytic performance in all cases, despite the resistance it provided against sintering of the PGMs. We propose that the smaller particle size obtained in P and thermally treated samples results from the encapsulation of particles in the support (via AlPO4 formation [24], pore blockage and subsequent SABET decrease), which inevitably isolates the active sites from the reactants and compromises the overall catalytic activity. Evidence of encapsulation/coverage is given in our previous studies [24] where we report decreased CO adsorption on PGM particles upon P poisoning. Other deactivation pathways - such as changes in PGM redox dynamics in the presence of P [24] as discussed above - also contribute to the decreased activity of poisoned samples. However, the observation of pore blockage upon P poisoning from the analysis of the pore distribution is a new experimental tile that aids our understanding of catalyst aging.

4 Conclusion

Our previous publications on aging of commercial bilayer DOC showed that this system was particularly susceptible to the deterioration of NO oxidation performance; nevertheless, the role of each catalytic layer and their individual response to aging was unclear. To investigate the behaviour of each coating, P poisoning and/or thermal treatments were performed on two DOC monoliths: (1) the commercial bilayer catalyst (with the top layer richer in Pt), and (2) a model catalyst comprising only the top layer.

The catalytic activity results suggest that all aging treatments are detrimental for the catalytic performance. While the deterioration of NO oxidation activity occurred to equivalent extent in both DOCs, milder deactivation of CO and HC oxidation activity was observed in the bilayer catalyst.

The deactivation upon thermal treatments was attributed to the substantial PGM particle growth observed by TEM. In agreement with previous reports the sintering was minimised when the thermal treatments were accompanied by the presence of P. In spite of the decreased sintering, the addition of P was detrimental for catalytic activity. Based on the selective decrease of a defined size range of pores characteristic of the top layer – richer in Pt and responsible for NO oxidation – we propose that the smaller particle size of consecutively poisoned and thermally treated samples results from PGM encapsulation by the support most likely via formation of AlPO4 and surface area decrease (observed by N2 physisorption). The deactivation by P poisoning possibly occurs as combination of this encapsulation and else phenomena such as changes in PGM redox properties that we have presented in our previous work.

NO oxidation is expected to occur exclusively in the top layer (Pt rich) and the preferential chemical poisoning of this layer explains the decreased NO oxidation performance upon aging. The higher catalytic activity towards CO and HC oxidation of the two layer DOC (compared to the one layer sample) can be explained by the milder poisoning of the bottom layer which provides CO and HC conversion activity when the top layer is partially deactivated. Hence, the results indicate that a more durable NO oxidation functionality could be achieved by improving the P resistance of the top layer. This can be approached by using a less phosphating support in the top layer, as well as by implementing a P trap upstream of the DOC to minimize the poisoning. We believe that the comparison of multi and single layer catalysts shown in this work helps the rationalisation of aging phenomena in modern DOCs of complex formulation and sets the basis for improvement of multilayer catalysts for durable NO oxidation performance.