An Assessment of Zeolite Framework Effect for Low-Temperature NO X Adsorbers

: Pd-promoted zeolites (Y, ZSM-5, FER, SSZ-13) were prepared and characterized to analyze their properties as low-temperature NO x adsorbers. The samples were investigated by BET and XRD and by in situ FT-IR spectroscopy of CO and NO adsorption to probe the Pd sites and the nature of the adsorbed NO x species. The NO x adsorption/desorption performances at low temperatures were examined by microreactor measurements upon NO/O 2 adsorption followed by TPD in the presence of water and carbon dioxide. It was enlightened that: (i) the zeolite framework inﬂuences the Pd dispersion: the smaller the zeolite cage, the higher the Pd dispersion, irrespective of the Si/Al ratio. Accordingly, the following Pd dispersion order has been observed, inversely to the zeolite cage size: Pd/SSZ-13 > Pd/ZSM-5 ~ Pd/FER >> Pd/Y; (ii) Pd is present as isolated Pd n+ species and in PdO x particles; (iii) the Pd dispersion governs the NO x storage capacity: the smaller the zeolite cage, the higher the Pd dispersion and the storage capacity; (iv) NO adsorbs mainly in the form of Pd nitrosyls and nitrates; (v) NO desorption occurs both at a temperature below 200 ◦ C and in a high-temperature range (near 350 ◦ C).


Introduction
The severe NO x emission regulations enforced in the automotive sector [1] call for highly efficient NO x reduction techniques. Currently adopted technologies for lean-burn engines are the NO x -storage reduction (NSR) technique, also referred to as Lean NO x traps (LNT) [2,3], and the urea selective catalytic reduction (SCR) [4]. In fact, both techniques are not efficient for the reduction in NO x emissions at low temperatures, as required by recent regulations, especially for hybrid vehicles where the endothermic engine is continuously subjected to start and stop cycles [5,6].
The use of passive NO x adsorber (PNA) systems installed upstream of the SCR or LNT systems has been recently proposed to mitigate cold-start NO x emissions. In this configuration, PNAs store NO x at low temperatures; the trapped NO x are then released at higher temperatures when the NO x reduction catalysts placed downstream effectively accomplish the NO x abatement [7][8][9][10].
In 2016, Johnson Matthey [11] reported a very efficient formulation for PNA systems based on acidic zeolites doped with Pd. In that work, the authors compared the behavior of different zeolite structures, including MFI (Mobil-type five), CHA (chabasite), and BEA (Beta). These systems showed remarkable NO x storage capacity at 100 • C, with release of the trapped NO x at a temperature suitable for practical application, i.e., near 200 • C. It was found that the framework structure of the zeolites affects the NO x storage capacity and the desorption temperature; in particular, the authors pointed out that at 100 • C the 2. Results and Discussion 2.1. Sample Characterization 2.1.1. Structural and Morphological Characterization Y zeolite exhibits the Faujasite (FAU) framework; it is a large-pore zeolite (pore openings of 7.4 Å), and its structure contains a 3D pore system limited by 12-member rings. Ferrierite (FER framework) and ZSM-5 (MFI framework) are medium-pore zeolites whose structure contains a 2D system of intersecting pores limited by 8-and 10-member rings for FER and 10-member rings for ZSM-5. SSZ-13 is a small-pore zeolite (ring opening 3.6 Å) with a chabazite framework; its structure contains a 3D pore system limited by 8-member rings. The main characteristics of these zeolitic frameworks are reported in Table 1, together with the compositional and textural characterization.
As expected, all the zeolites are characterized by a large surface area and very low pore volume; the addition of Pd does not substantially modify either the surface area or the pore volume.
X-ray Diffraction (XRD) patterns obtained on the bare zeolites and on the Pd-doped samples ( Figure 1) are in agreement with the diffraction patterns for the reference zeolite frameworks [22]. The zeolite structure is maintained in all the catalysts even after Pd addition; additionally, peaks corresponding to crystalline PdO, whose most intense peak should be observed at 33.8 • (JCPDS 75-0584), are not observed, suggesting that particles are too small to be identified by the XRD technique or absent.
SEM (Scanning Electron Microscopy) images reported in Figure 2 clearly show the maintenance of the zeolite morphologies also after Pd addition; moreover, the associated energy-dispersive X-ray (EDX) analysis confirms the Si/Al ratio given by the supplier and the Pd loading near the nominal value (1 wt%, see Table 1). Pd loading has also been confirmed by ICP-MS analysis in the case of the SSZ-13 sample.  X-ray Diffraction (XRD) patterns obtained on the bare zeolites and on the Pd-doped samples ( Figure 1) are in agreement with the diffraction patterns for the reference zeolite frameworks [22]. The zeolite structure is maintained in all the catalysts even after Pd addition; additionally, peaks corresponding to crystalline PdO, whose most intense peak should be observed at 33.8° (JCPDS 75-0584), are not observed, suggesting that particles are too small to be identified by the XRD technique or absent.  Figure 2 clearly show the maintenance of the zeolite morphologies also after Pd addition; moreover, the associated energy-dispersive X-ray (EDX) analysis confirms the Si/Al ratio given by the supplier and the Pd loading near the nominal value (1 wt%, see Table 1). Pd loading has also been confirmed by ICP-MS analysis in the case of the SSZ-13 sample.

FT-IR Characterization
CO adsorption-In Figure 3, the FT-IR spectra of CO (20 mbar) adsorbed at Room Temperature (RT) on the activated samples are displayed. The spectra are characterized by complex envelopes of bands, which are related to Pd carbonyl species in different oxidation states and coordination, as discussed in our previous paper [23], where the precise assignment of each component in the spectrum of Pd/SSZ-13 was made. As discussed in ref. [23], with reference to the Pd/SSZ-13 sample (trace a), it is possible to divide the bands into different spectral regions, namely 2220-2170 cm −1 and 2170-2110 cm −1 , associated with linear carbonyls of Pd 2+ and Pd + , respectively [13,[24][25][26]. The absorption near 2228 cm −1 is assigned to carbonyls of extra-framework Al 3+ [27].

FT-IR Characterization
CO adsorption-In Figure 3, the FT-IR spectra of CO (20 mbar) adsorbed at Room Temperature (RT) on the activated samples are displayed. The spectra are characterize by complex envelopes of bands, which are related to Pd carbonyl species in differen oxidation states and coordination, as discussed in our previous paper [23], where th precise assignment of each component in the spectrum of Pd/SSZ-13 was made. A discussed in ref. [23], with reference to the Pd/SSZ-13 sample (trace a), it is possible t divide the bands into different spectral regions, namely 2220-2170 cm −1 and 2170-211 cm −1 , associated with linear carbonyls of Pd 2+ and Pd + , respectively [13,[24][25][26]. Th absorption near 2228 cm −1 is assigned to carbonyls of extra-framework Al 3+ [27].   In the spectral region 2220-2170 cm −1 (where very strong bands are present in the case of Pd/SSZ-13 sample, Figure 3 trace a), it has to be considered that, due to the σ-donation contribution in the Pd 2+ -CO bond, the higher the band frequency, the more coordinatively unsaturated the site. With this in mind and on the base of the literature data, the bands at 2211 and 2192 cm −1 are assigned to di-carbonyls (νasym(CO) and νsym(CO) modes, In the spectral region 2220-2170 cm −1 (where very strong bands are present in the case of Pd/SSZ-13 sample, Figure 3 trace a), it has to be considered that, due to the σ-donation contribution in the Pd 2+ -CO bond, the higher the band frequency, the more coordinatively unsaturated the site. With this in mind and on the base of the literature data, the bands at 2211 and 2192 cm −1 are assigned to di-carbonyls (ν asym (CO) and ν sym (CO) modes, respectively) of isolated Pd 2+ ions formed by ion exchange with the Brønsted acid sites of the zeolite [24,28]; the shoulder at 2187 cm −1 is instead assigned to CO adsorbed on Pd 2+ sites of PdO x particles [23].
Along similar lines, for the spectral region 2170-2110 cm −1 , due to the π-backdonation contribution to the Pd + -CO bond, the lower the band frequency, the more coordinatively unsaturated the site. For this reason, in the spectrum of Pd/SSZ-13 ( Figure 3, trace a), the band at 2163 cm −1 is assigned to CO adsorbed on Pd + sites belonging to PdO x particles. The bands at 2150, 2142, 2127, and 2111 cm −1 are related to the presence of two different Pd + di-carbonyl species (two couple of bands, one component of the couple related to ν asym (CO) mode, the other component to ν sym (CO) mode), whose presence is well demonstrated in our previous paper [23]. This implies, for Pd/SSZ-13, the existence of two different isolated Pd + sites formed by ion exchange with the Brønsted acid sites of the zeolite: indeed, Pd + can be located into both 8-and 6-member rings of the zeolite framework [17,29].
For Pd/ZSM-5 and Pd/FER samples ( Figure 3, traces b and c, respectively), the bands observed in the same regions are less intense, but their assignments are similar to that examined for the Pd/SSZ-13 sample. In particular, for the Pd/ZSM-5 sample, the bands at 2214 and 2194 cm −1 are assigned to di-carbonyls of isolated Pd 2+ ions; the band at 2173 cm −1 to CO bonded on Pd 2+ belonging to PdO particles; the bands at 2142 and 2127 cm −1 to di-carbonyls of isolated Pd + sites. For the Pd/FER sample ( Figure 3, trace c), the broadness of the bands means higher surface heterogeneity. Notably, the two bands related to di-carbonyls of isolated Pd 2+ ions at about 2214 and 2194 cm −1 show shoulders, pointing out the presence of exchanged Pd 2+ ions located in different zeolite positions (e.g., 10-and 8-ring channels). Finally, it is worth noting that the spectrum of Pd/Y ( Figure 3, trace d) shows very weak bands, enlarged in the inset of Figure 3. Only bands related to di-carbonyls of isolated Pd + ions are present at 2142 and 2111 cm −1 , respectively. The band at 2228 cm −1 related to carbonyls of extra-framework Al 3+ is present as well.
The area of the carbonyl bands is proportional to the number of available Pd sites, both isolated Pd species inside the zeolite framework and Pd species belonging to external PdO x particles. Inspection of Figure 3 clearly indicates that the availability of the Pd sites (i.e., the dispersion) is highest for the Pd/SSZ-13 sample and follows the order: Pd/SSZ-13 > Pd/ZSM-5~Pd/FER >> Pd/Y. This suggests a relationship between pore size and the Pd dispersion: the smaller the pore size, the higher the Pd dispersion. In this respect, it has to be recalled that Zheng et al. [13] characterized Pd on zeolites with different pore sizes and found that Pd is largely present as isolated ions in the ZSM-5 framework, while the small pore opening of SSZ-13 inhibits the exchange with Brønsted acid sites. In our case, the availability of the Pd sites follows a different order, as reported above, apparently in contrast with the results of Zheng et al. This discrepancy might be related to the different pretreatment of the samples: in fact, while they evaluated the dispersion of isolated Pd ions on the freshly calcined samples, at variance, we evaluated the dispersion after an activation treatment in vacuum and oxygen at 500 • C.
In addition, our results seem to indicate that the Pd dispersion is not driven by the Si/Al ratio for zeolites having similar pore size: in fact, the Pd/ZSM-5 and Pd/FER samples, characterized by similar pore size but different Si/Al ratio, exhibit similar Pd dispersion. This suggests that the Pd dispersion, for the Pd loading conditions here adopted, is primarily affected by the zeolite pore size and only to a minor extent by the Si/Al ratio. At variance, Khivantsev et al. [28] analyzed the Pd dispersion over 1 wt% Pd-doped zeolites having different Si/Al ratios (and the same framework) but prepared in a different way and observed that the Si/Al ratio significantly affects the Pd dispersion. In particular, these authors, using a modified-ion exchange method to prepare 1 wt% Pd on NH 4 -SSZ-13 with Si/Al ratios ranging from 6 to 30, observed the increasing formation of PdO nanoparticles (i.e., low dispersion) upon increasing the Si/Al ratio. This is very likely related to the different preparation method used by these authors, which significantly differs with respect to our preparation route.
NO and NO/O 2 adsorption-Traces "a" in Figure 4 show FT-IR spectra of NO adsorbed at RT on activated Pd/Y (A), Pd/ZSM-5 (B), Pd/FER (C), and Pd/SSZ-13 (D). In all cases, bands in the region of 1900-1800 cm −1 are present with intensities that vary from one sample to another. These spectral features are well documented in the literature and are assigned to nitrosyl species of Pd 2+ and Pd + , even if there is no definitive consensus on the precise assignments of the single bands to either the Pd 2+ or Pd + species [28][29][30][31][32]. Accordingly, it is worth noting that, contrarily to CO, NO adsorption is not able to distinguish the different coordinations of Pd n+ ions (isolated ions or sites on particles). In fact, the complex envelopes of bands present in the spectra of adsorbed CO do not find correspondence in the spectral features shown upon NO adsorption. However, the intensities of the Pd nitrosyl bands in the various samples well match those of carbonyls observed upon CO adsorption. This confirms the Pd dispersion order already discussed in the case of CO adsorption, i.e., Pd/SSZ-13 > Pd/ZSM-5~Pd/FER >> Pd/Y.  Along with Pd nitrosyl bands, for Pd/FER and Pd/SSZ-13 samples (Figure 4, traces "a", sections C and D, respectively), peaks at 2250-2247, 2161-2155, 1972 (only for Pd/FER), and at about 1630 cm −1 are also present. The bands at 2250-2247 cm −1 and at 1972 cm −1 are related to adsorbed N2O and N2O3, respectively, present as impurities in the gas [33] or formed as detailed below. On the other hand, the bands at 2161-2155 and 1630 cm −1 are assigned to the stretching mode of NO + and to the bending mode of adsorbed molecular water, respectively. Hadjiivanov et al. [34] proposed the formation of these species upon the reaction of NO with Brønsted acid sites of an H-ZSM-5 in the presence of oxygen: Oxygen is not present in the gas phase, but the samples were oxidized before NO admission: accordingly, oxygen species are likely present on the Pd phase. This is supported by the fact that the bands related to NO + and adsorbed molecular water show lower intensities on the sample with lower Pd dispersion, i.e., Pd/FER with respect to Pd/SSZ-13. As a matter of fact, NO + species are observable with very low intensities also in the Pd/ZSM-5 sample (NO + band at 2128 cm −1 and water band at 1635 cm −1 ), and they are not  [33] or formed as detailed below. On the other hand, the bands at 2161-2155 and 1630 cm −1 are assigned to the stretching mode of NO + and to the bending mode of adsorbed molecular water, respectively. Hadjiivanov et al. [34] proposed the formation of these species upon the reaction of NO with Brønsted acid sites of an H-ZSM-5 in the presence of oxygen: Oxygen is not present in the gas phase, but the samples were oxidized before NO admission: accordingly, oxygen species are likely present on the Pd phase. This is supported by the fact that the bands related to NO + and adsorbed molecular water show lower intensities on the sample with lower Pd dispersion, i.e., Pd/FER with respect to Pd/SSZ-13. As a matter of fact, NO + species are observable with very low intensities also in the Pd/ZSM-5 sample (NO + band at 2128 cm −1 and water band at 1635 cm −1 ), and they are not present (except for a very small amount of water) for Pd/Y sample, for which Pd site availability is almost nil. It is important to point out that the presence of reactive oxygen species on the Pd phase can also induce the formation of NO 2 that, along with NO, can form the adduct N 2 O 3 (indeed observed for Pd/FER catalyst) and/or can form NO + and surface nitrates according to the following reactions [35][36][37]: Consequently, the observed NO + can be derived from both reactions (1) and (2), along with nitrates that should show characteristic vibrational modes in the region below 1700 cm −1 . As a matter of fact, nitrates could contribute to the band at about 1630 cm −1 , assigned to the bending mode of molecular water. All these observations were confirmed by oxygen admission, as described in the following section.
Starting from the condition obtained in the presence of 2 mbar of NO (spectra "a"), 10 mbar of oxygen were added: traces "b" and "c" in the different sections of Figure 4 show the evolution of bands after 5 and 60 min, respectively. First of all, the increase in bands below 1700 cm −1 , assigned to surface nitrates formed by reactions (2) and (3) mentioned above, where NO 2 is formed upon NO oxidation, is well evident for all the samples. In this region, the band at 1630 cm −1 is related to both nitrates and molecular water formed by reaction (1). Moreover, reactions (1) and (2) account for the formation of NO + , whose band suddenly increases after oxygen admission for Pd/SSZ-13 and Pd/FER samples. Even in the presence of oxygen, the band of NO + at 2128 cm −1 remains very weak for Pd/ZSM-5 and not detectable for Pd/Y sample. Indeed, reaction (1) needs Brønsted acid sites to occur: as a matter of fact, the intensity of the band and hence the amount of adsorbed NO + is directly related to the Brønsted acid site amount, i.e., to the Si/Al ratio of the zeolite. Moreover, it has to be taken into account the higher stability of adsorbed NO + in the smaller channels of SSZ-13 and FER [38].
Concerning nitrosyl species, oxygen admission causes an intensity increase in the bands for Pd/SSZ-13 and Pd/FER samples. Moreover, for Pd/SSZ-13 sample, nitrosyl bands blue-shift upon oxygen admission. The latter can be ascribed to the increase in the average positive charge of the Pd sites, especially of the Pd + sites, induced by oxygen adsorption, which could also cause changes in the absorption coefficients of nitrosyl species, justifying the increase in band intensities.
The areas (integrated intensities) of the bands related to nitrosyls and nitrates are indicative of the Pd dispersion, as also evidenced by CO adsorption. Indeed, along with nitrosyl formation, the presence of Pd is required to form nitrate species as well. In fact, nitrates are not formed in the absence of Pd, as evidenced by the comparison between the NO/O 2 adsorption on Pd/SSZ-13 and the bare SSZ-13 shown in Figure 5. This comparison also points out the markedly lower amount of NO + (band at 2161 cm −1 ) on Pd/SSZ-13 with respect to the bare zeolite. In fact, NO + is formed on the zeolite Brønsted acid sites, and their concentration has been reduced as a result of Pd exchange with the formation of Pd 2+ and Pd + sites.
The areas (integrated intensities) of the bands related to nitrosyls and nitrates are indicative of the Pd dispersion, as also evidenced by CO adsorption. Indeed, along with nitrosyl formation, the presence of Pd is required to form nitrate species as well. In fact, nitrates are not formed in the absence of Pd, as evidenced by the comparison between the NO/O2 adsorption on Pd/SSZ-13 and the bare SSZ-13 shown in Figure 5. This comparison also points out the markedly lower amount of NO + (band at 2161 cm −1 ) on Pd/SSZ-13 with respect to the bare zeolite. In fact, NO + is formed on the zeolite Brønsted acid sites, and their concentration has been reduced as a result of Pd exchange with the formation of Pd 2+ and Pd + sites.

NO x Uptake and Desorption Experiments 2.2.1. NO/O 2 Adsorption Phase
The NO x storage performances of Pd/zeolite were determined under realistic conditions, i.e., at 80 • C with a NO/O 2 mixture in the presence of CO 2 and H 2 O.
The concentration traces obtained upon NO x adsorption and subsequent Temperature Programmed Desorption (TPD) are shown in Figure 6A-D. For all samples, upon the addition of the NO/O 2 mixture, the NO concentration recorded at the reactor outlet rapidly increases without showing any dead time, eventually reaching the inlet concentration value with a dynamic response that depends on the catalyst sample. The response is very fast in the case of Pd/Y ( Figure 6A) and sluggish for the Pd/SSZ-13 sample ( Figure 6D), which, in fact, shows a large NO x uptake evaluated from the amounts of desorbed NO x during the TPD (Figure 7). Lower amounts of NO x are stored on the other Pd-doped samples, and the following order is observed: Pd/SSZ-13 > Pd/ZSM-5~Pd/FER >> Pd/Y (Figure 7). This ranking is in good agreement with the Pd dispersion evaluated from FT-IR measurements of CO and NO adsorption.
During NO adsorption, NO 2 evolution is also observed, although with different features. In fact, a small NO 2 peak is observed in the case of Pd/ZSM-5 and Pd/FER ( Figure 6B, C, respectively), with evolution of 8 and 5 µmol/g cat NO 2 , respectively, whereas a sluggish NO 2 evolution is seen in the case of Pd/SSZ-13. The overall amount of NO 2 evolved, in this case, is near 19 µmol/g cat . Almost no NO 2 formation is observed in the case of Pd/Y. NO 2 evolution observed during the adsorption phase is likely related to the reduction of Pd 2+ to Pd + operated by NO according to the following overall reaction [13,[39][40][41]:

3NO + 2[Pd(II)OH] + → NO 2 + 2[Pd(I)NO] + + H 2 O (4)
According to the stoichiometry of reaction (4), 3 moles of NO are consumed and lead to the formation of 1 mol of NO 2 evolved in the gas phase and to the storage of 2 moles of Pd nitrosyls. Reaction (4) also envisages the reduction of Pd 2+ with the formation of Pd + nitrosyls upon NO adsorption. 3NO + 2[Pd(II)OH] → NO2 + 2[Pd(I)NO] + H2O (4) According to the stoichiometry of reaction (4), 3 moles of NO are consumed and lead to the formation of 1 mol of NO2 evolved in the gas phase and to the storage of 2 moles of Pd nitrosyls. Reaction (4) also envisages the reduction of Pd 2+ with the formation of Pd + nitrosyls upon NO adsorption.  et al. [37], who pointed out that more stable nitrosyls are formed over Pd + species. Moreover, as reported by Castoldi et al. in the case of Pd/FER [42], and by Matarrese et al. [23] for Pd/SSZ-13, the low-temperature NOx desorption peak can also be associated with the decomposition of nitrate species. On this basis, it is speculated that the low-T desorption peaks can be associated with the decomposition of nitrates and/or Pd 2+ nitrosyls, whereas the high desorption peak is related to the decomposition of stable Pd + nitrosyls. The results reported above clearly indicate that the zeolite framework affects both the amounts and the stability of the stored NOx. The characterization data previously discussed point out the existence of different Pd species in terms of different structures and oxidation states, including isolated Pd cations and PdOx particles on the external zeolite surface, whose distribution varies with the zeolite framework [11,13]. Additionally, according to FT-IR characterization, the investigated samples show different Pd dispersion, and evidence has been provided on the fact that this parameter is influenced by the pore size of the zeolitic framework. The Pd dispersion is the key parameter affecting the NOx storage, having observed a clear correlation between NOx adsorption capacity and Pd dispersion. Notably, a very high NO adsorption capacity has been measured for the Pd/SSZ-13 sample, in spite of the fact that Pd species belonging to PdOx particles have been identified by CO adsorption. This suggests that also non-isolated Pd species participate in the NOx storage. Notably, as expected from the stoichiometry of reaction (4), the amounts of evolved NO 2 represent half of the amounts of formed nitrosyls. However, as will be shown in the following section, the amounts of stored NO x are higher than those expected from the stoichiometry of reaction (4) based on the NO 2 formation. This points out that other routes are active in the storage of NO x , possibly involving the direct NO uptake on Pd 2+ sites as well.

NO x Desorption Phase
During the thermal desorption (TPD) after the NO/O 2 storage at 80 • C, NO x release occurs mainly as NO, although NO 2 is also observed to desorb. The amounts of evolved NO x were evaluated from the area of the TPD peaks, and the results are summarized in Figure 7, along with the estimated desorbed NO x /Pd molar ratio.
For all samples, at the beginning of the temperature ramp a desorption of H 2 O is observed due to the heating ramp; notably, the amounts of water desorbed is inversely proportional to the pores size, i.e., the bigger the pores, the lower the amount of desorbed water. This is very likely related to water capillary condensation that is favored within the smaller pores.
In the case of the Pd/SSZ-13 sample ( Figure 6D), three main desorption regions can be identified: (i) from the closure of the NO/O 2 feed up to 120 • C a tail is apparent due to the physisorbed species, counting for 13 µmol/g cat NO; (ii) from 120 • C to 240 • C a well-defined peak is observed with a maximum at 180 • C, composed of NO and NO 2 (32 µmol/g cat and 6 µmol/g cat , respectively); and (iii) a high temperature peak (in the range of 240-500 • C) formed only by NO (41.4 µmol/g cat ) and centered near 360 • C. The overall amount of evolved NO x is near 93 µmol/g cat (see Figure 7). As previously recalled, the overall amounts of stored NO x are well above those expected from the amounts of evolved NO 2 , based on the stoichiometry of reaction (4). This indicates the occurrence of other adsorption reactions besides reaction (4), e.g., direct NO uptake on Pd 2+ with no reduction of Pd 2+ to Pd + (and no NO 2 formation).
The Pd/ZSM-5 and Pd/FER present two desorption regions ( Figure 6B,C, respectively), the former in the low-temperature region below 240 • C and the latter in the hightemperature region above 300 • C; negligible amounts of NO 2 are observed for these samples. The overall amounts of desorbed NO x are 41 and 50 µmol/g cat , respectively. Finally, in the case of Pd/Y ( Figure 6A), any NO x desorption is observed, as expected, due to the negligible storage capacity of this sample. These results are well in line with the CO and NO FTIR adsorption experiments previously discussed, showing a higher NO adsorption capacity for Pd/SSZ-13.
From the amounts of stored NO x , we have also estimated the adsorbed NO x /Pd molar ratio (Figure 7) under the hypothesis that all NO x are stored over Pd. In the case of the Pd/SSZ-13 sample, a value of desorbed NO x near 93 µmol/g cat was evaluated, corresponding to a NO x /Pd molar ratio of 0.86. This calls for a large utilization of Pd, thus indicating a good dispersion of the metal. The Pd/ZSM-5 and Pd/FER showed lower stored NO x /Pd values, whereas, as already stated the storage capacity for Pd/Y is negligible.
Concerning the stability of the adsorbed NO x species, low-and high-temperature contributions are observed in all cases, excluding the Pd/Y sample that shows no NO adsorption. NO x desorption peaks at low and high temperatures (near 150 • C and 360 • C) have also been observed, e.g., by Gupta et al. over Pd/SSZ-13 [39]; the authors attributed these two peaks to at least two different adsorption sites with distinct binding energies. In particular, they attributed the lower temperature NO peak to NO bound to Z − Pd 2+ Z − (where Zdenotes the zeolite anionic aluminum site (Al-O-Si) -) and the higher temperature NO peak to NO bound to Z − Pd + . Similar conclusions have been drawn by Villamaina et al. [37], who pointed out that more stable nitrosyls are formed over Pd + species. Moreover, as reported by Castoldi et al. in the case of Pd/FER [42], and by Matarrese et al. [23] for Pd/SSZ-13, the low-temperature NO x desorption peak can also be associated with the decomposition of nitrate species. On this basis, it is speculated that the low-T desorption peaks can be associated with the decomposition of nitrates and/or Pd 2+ nitrosyls, whereas the high desorption peak is related to the decomposition of stable Pd + nitrosyls.
The results reported above clearly indicate that the zeolite framework affects both the amounts and the stability of the stored NO x . The characterization data previously discussed point out the existence of different Pd species in terms of different structures and oxidation states, including isolated Pd cations and PdO x particles on the external zeolite surface, whose distribution varies with the zeolite framework [11,13]. Additionally, according to FT-IR characterization, the investigated samples show different Pd dispersion, and evidence has been provided on the fact that this parameter is influenced by the pore size of the zeolitic framework. The Pd dispersion is the key parameter affecting the NO x storage, having observed a clear correlation between NO x adsorption capacity and Pd dispersion.
Notably, a very high NO adsorption capacity has been measured for the Pd/SSZ-13 sample, in spite of the fact that Pd species belonging to PdO x particles have been identified by CO adsorption. This suggests that also non-isolated Pd species participate in the NO x storage.

Sample Preparation
The Pd-doped zeolites (Pd/Y, Pd/ZSM-5, Pd/FER, and Pd/SSZ-13; Pd loading 1% w/w) used in this study were prepared starting from commercial samples (Zeolyst International, Kansas City, KS, USA and ACS Material LLC, Pasadena, CA 91106) in their NH 4or H-form. The zeolites in the NH 4 -form were transformed into H-form by calcination at 700 • C in the air (heating rate: 1 • C/min).
Then, the Pd-doped zeolite samples were prepared by impregnation of the zeolite supports in excess of solvent using aqueous Pd(NO 3 ) 2 (solution at 4.5% Pd, Sigma-Aldrich, St. Louis, MO, USA), followed by drying in air at 80 • C overnight [42]. The dried precursors were calcined from room temperature up to 500 • C (rate: 1 • C/min, at 500 • C 3 h) and then from 500 • C up to 750 • C (rate 5 • C/min, at 750 • C 2 h).

Sample Characterization
BET analysis was performed by adsorption-desorption of nitrogen at −196 • C using a Micromeritics Tristar 300 instrument (Alfatest SRL, Rome, Italy). Before the measurement, samples were outgassed at 120 • C and 0.1 torr for 3 h.
Powder X-ray diffraction (XRD) patterns of bare and Pd/zeolite samples were obtained using a Bruker D8 diffractometer () with Cu Kα radiation. The XRD patterns were obtained on the as prepared samples in the range of 10-50 • (2θ) with a step size of 0.05 • at a speed of 12.5 • /min. SEM and elemental compositions analysis of bare zeolite and Pd-doped samples were performed using a ZEISS EVO 50 instrument (Oberkochen, Germany) equipped with energydispersive X-ray analyzer (EDX) model Quantax 200 6/30 (Bruker, Karlsruhe, Germany).
ICP-MS analysis has been performed on selected samples to evaluate the Pd loading. The Perkin Elmer NexION 2000 single quadrupole ICP-MS instrument (Shelton, CT, USA) was employed upon calibration with a limit of quantification (LOQ) of 0.040 ppb for Pd.
IR spectra were run in transmission mode at RT on a Perkin-Elmer FT-IR 2000 spectrophotometer (Shelton, CT, USA) equipped with a Hg-Cd-Te cryodetector, working in the range of wavenumbers 7200-580 cm −1 at a resolution of 2 cm −1 (number of scans~60). Self-supporting disks (10-15 mg cm −2 ) were obtained from the powder samples and placed in a quartz IR cell allowing thermal treatments in vacuum or in a controlled atmosphere. The samples were activated by treatment at 500 • C in vacuum, followed by oxidation in dry O 2 (40 mbar) at the same temperature. The samples were cooled down to room temperature (RT) in oxygen. CO adsorption was performed to characterize the nature and availability of Pd sites; NO and NO/O 2 adsorptions were performed to evidence the species formed during NO x storage. IR spectra were recorded at RT in static conditions before and after interaction with CO (20 mbar), NO (2 mbar), and NO/O 2 (1:5, p NO = 2 mbar) mixture. In the figures, spectra are reported as difference spectra, where the subtrahend spectrum is that of the sample after the activation, before the admission of the gas.

Gas phase Analysis during NO/O 2 Adsorption at Low
Temperature NO x adsorption experiments were conducted using 100 mg of powdered samples, sieved at 70-100 µm, and loaded in a quartz microreactor with an internal diameter of 8 mm. The outlet of the reactor is directly connected to a mass spectrometer (ThermoStar 200, Pfeiffer Vacuum GmbH, Aßlar, Germany)), a UV-analyzer specific for NO, NO 2 , NH 3 (Limas 11 HW, ABB, Zurig, Swizzerland), and a micro gas-chromatograph (3000 A, Agilent, Santa Clara, CA 95051 United States). In a typical experiment, the sample was initially pre-treated at 500 • C for 30 min under a flow of O 2 (3% v/v), CO 2 (2% v/v), and H 2 O (2.5% v/v) (balance He) at a flow rate of 100 Ncm 3 /min. The sample was then cooled down to the adsorption temperature (50 • C, 80 • C, or 120 • C) without changing the gas flow composition, and at this temperature 300 ppm of NO were stepwise added to the gas mixture for 75 min. Then, the NO supply was closed, and the sample was heated under temperature programming (TPD) up to 500 • C (heating rate 10 • C/min).

Conclusions
In this study, Pd-promoted zeolites have been prepared by impregnation of the zeolite supports in excess of solvent and investigated for potential use in low-temperature NO x adsorption. Four zeolitic supports with different structures have been considered (Y, ZSM-5, FER, and SSZ-13). These structures have different pore openings, i.e., large (Y), medium (ZSM-5 and FER), and small (SSZ-13); samples with different Si/Al ratios have also been considered. All the samples have been characterized and tested in the adsorption of NO under relevant operating conditions.
The FT-IR characterization using CO as a probe molecule evidenced the presence of a variety of Pd n+ sites with different coordination. In particular, isolated Pd cations formed by ion exchange with the Brønsted acid sites of the zeolite and on the surface of PdO x particles have been evidenced. The Pd dispersion estimated by CO and NO adsorption follows the order: Pd/SSZ-13 > Pd/ZSM-5~Pd/FER >> Pd/Y, i.e., increases upon decreasing the pore opening. These results suggest that the pore size plays a primary role in controlling the Pd dispersion. At variance, the relevance of the Si/Al ratio in affecting the Pd dispersion seems to be minor. In other words, the zeolite structure (and related pore openings) is apparently the key parameter to drive the Pd dispersion.
Reactivity experiments carried out under the relevant conditions pointed out that the amounts of adsorbed NO x follow the same order as the Pd dispersion. Notably, a very high NO x /Pd molar ratio has been estimated for Pd/SSZ-13, despite the presence of PdO x clusters and/or particles pointed out in this sample besides well-dispersed Pd sites. This suggests that non-isolated Pd species belonging to particles participate in the NO x storage along with isolated well-dispersed species.