A DFT study on the mechanism of NO decomposition catalyzed by short-distance Cu(I) pairs in Cu-ZSM-5

Molecular Catalysis 434 (2017) 96–105

Contents lists available at ScienceDirect

Molecular Catalysis journal homepage: www.elsevier.com/locate/mcat

Research Paper

A DFT study on the mechanism of NO decomposition catalyzed by short-distance Cu(I) pairs in Cu-ZSM-5 Simone Morpurgo Dipartimento di Chimica, Università degli Studi di Roma “La Sapienza”, P.le Aldo Moro 5, 00185 Roma, Italy

a r t i c l e

i n f o

Article history: Received 6 September 2016 Received in revised form 18 January 2017 Accepted 19 January 2017 Keywords: Zeolites Cu-ZSM-5 DFT NO decomposition reaction mechanism

a b s t r a c t The complete NO decomposition catalyzed by short-distance Cu+ pairs in Cu-ZSM-5 was studied by means of DFT calculations. After adsorption of two NO molecules, an hyponitrite species is formed. Further decomposition of hyponitrite occurs with activation energies ranging from about 4 to 24 kcal mol−1 , depending on the initial geometry of the substrate-catalyst complex. An oxidized form of the catalyst, [Cu O Cu]2+ and a copper-coordinating N2 O molecule are obtained. Further N2 O decomposition may occur with oxygen transfer from N2 O to [Cu O Cu]2+ and formation of N2 and O2 , both adsorbed on the catalyst. Three different kinds of transition states were identified for the latter step, which appears to be rate-determining due to activation energies ranging from 39–40, to 44–45, and to 50–52 kcal mol−1 , respectively. After this, N2 desorption occurs easily, whereas O2 desorption is endothermic (from 28.8 to 36.5 kcal mol−1 ), the highest value being associated to reductive O2 desorption from a peroxide-like complex. It turned out that the best way for N2 O elimination is the direct, spin-forbidden decomposition on a reduced Cu+ · · ·Cu+ pair, with formation of [Cu O Cu]2+ and N2 , as already suggested in the literature. The problem of how the reduced catalyst may be regenerated is left open. © 2017 Elsevier B.V. All rights reserved.

1. Introduction The direct decomposition of NO (2NO  N2 + O2 ) is thermodynamically favoured up to about 1273 K but is rather slow in the gas phase and also in the presence of many catalysts [1–4]. However, since Iwamoto’s discovery in 1986 [5,6] it is known that Cu-ZSM-5 is much more active than many other catalysts even at relatively low temperature (673–773 K). Despite the large number of publications on this topic many points are not fully understood, mainly the nature of the active site and the reaction mechanism. It is commonly accepted that the superior activity of Cu-ZSM-5 is related to a small fraction of Cu2+ ions, introduced in the ZSM-5 zeolite only at the higher exchange levels [7,8] and easily reduced to Cu+ under vacuum at 723–773 K [9–11], but there is no agreement about the number of copper ions per active site. Some authors suggested that the reaction occurs on single Cu+ ions [2,12] but the possibility that active sites may be pairs of Cu+ ions was also suggested because (i) the turnover frequency (TOF) of the NO decomposition increases with the copper loading, following a peculiar sigmoidal curve with a plateau at about the maximum exchange capacity of the ZSM-5 catalysts [7,8] and (ii) the maximum value of the TOF can be lin-

E-mail address: [email protected] http://dx.doi.org/10.1016/j.mcat.2017.01.028 2468-8231/© 2017 Elsevier B.V. All rights reserved.

early correlated to the number of aluminium atoms per unit cell of the ZSM-5 zeolite [13]. Several computational studies were undertaken on Cu-ZSM-5 and NO decomposition, with the aim to elucidate both the nature of the active site and the reaction mechanism. Some of them were focused on the coordination of copper ions within the ZSM-5 framework and calculated, by means of suitable computational methods [14,15], the binding energy of Cu+ [16], Cu2+ [17], and Cu+ pairs [18] for different sites and coordination numbers. As far as the reaction mechanism is concerned, it is known from theoretical calculations that the single-step, symmetric, concerted decomposition of two NO molecules is forbidden by orbital symmetry both in the gas phase and on a single Cu+ site [19]. As a consequence, multi-step mechanisms were proposed for the above reaction on single Cu+ centres in Cu-ZSM-5 [20–26]. The generally accepted mechanism involves the initial formation of N2 O and of an oxidized ZCuO catalytic centre (Z = zeolite) from two NO molecules. The final products are obtained by decomposition of N2 O, which reacts with ZCuO, leaving N2 and an O2 molecule coordinated to ZCu (ZCu· · ·O2 ) [22–24]. It was also suggested that N2 O may react with a different ZCu site, giving ZCuO + N2 . Although the latter reaction is spin-forbidden, a catalytic effect of Cu-ZSM-5 was demonstrated [25,27]. Alternatively, it was suggested that N2 O reacts with a third NO molecule, giving N2 and NO2 . The latter species reacts with a

S. Morpurgo / Molecular Catalysis 434 (2017) 96–105

ZCuO unit, leaving NO again and ZCu· · ·O2 . In this way, NO behaves both as the substrate and as an oxygen-carrier, which, through the formation of NO2 , moves an O atom from a catalytic site to another [23,25]. Alternative pathways, involving the formation of the ZCu(NO2 )(NO) and/or ZCu(N2 O3 ) intermediates, were found impractical because of the high activation energies required for the decomposition of such species [26]. In a previous work [28] we investigated the mechanism of NO decomposition catalyzed by Cu+ pairs located at the opposite sides of the ten-membered rings of Cu-ZSM-5 but relatively high activation energies (about 50 kcal mol−1 ) were calculated for the rate-limiting step of the process, i.e. the reaction of N2 O with the almost linear [Cu O Cu]2+ fragment, which turned out to be very stable. Liu et al. investigated by DFT calculations the decomposition of N2 O catalyzed by a single Cu+ site [29] as well as by Cu+ pairs [30]. In the latter case, activation energies of 47.2 and 63.9 kcal mol−1 were respectively calculated when N2 O reacts with the bare Cu+ · · ·Cu+ pair and with the binuclear [Cu O Cu]2+ species [30]. However, these activation energies turned out to be higher than the corresponding ones calculated by the same authors on a single ZCu or ZCuO site (35.2 and 28.1 kcal mol−1 , respectively) [29]. Recently, it was suggested both by experimental and computational work that Cu+ pairs in Cu-ZSM-5 and other zeolites may be the active site for O2 activation and consequent CH4 oxidation to CH3 OH [31–34], as well as for N2 O decomposition [35,36]. The latter reaction, although spin-forbidden, was shown to have a low activation energy (2–15 kcal mol−1 ) because of the high stability of the binuclear [Cu O Cu]2+ species formed after release of N2 . According to absorption and Resonance Raman spectroscopy, the active sites should consist of Cu+ pairs located within the ten-membered rings of the zeolite channels, with the Cu+ ions coordinated to the lattice oxygens of two Al T-sites separated by two Si T-sites. It was also shown that the most active Cu+ pairs are those where the Cu Cu distance is sufficiently short (<4.2 Å) so that N2 O can bind with a bridged ␮-1,1-O coordination before reaction [31,35,36]. In the light of the above results, the present work investigates the whole process of NO decomposition catalyzed by Cu+ pairs located at a short distance, within the so-called M7 ring of the ZSM-5 structure [16,17]. The reason for this choice is that such a potential catalytic site probably represents the lowest limit for the Cu Cu distance of a Cu+ pair within the ZSM-5 framework, and therefore the study was aimed at checking whether this may have any special mechanistic implications. In particular, the present research is aimed at evaluating whether a short-distance Cu+ pair allows N2 O decomposition and N2 /O2 formation to occur with a lower activation energy with respect to a single-Cu+ site [20–26,29] and/or to Cu+ pairs located at a longer distance [28,30], where the above step requires a relatively high activation energy. In general, the limited size of the adopted clusters allows the comparison of a large number of structures optimized at an accurate computational level, starting from NO adsorption, to the formation of N2 O, up to the final release of N2 and O2 . The main geometrical features of the optimized structures may represent a reasonable starting point for corresponding studies on similar catalytic sites as, for instance, the M6 ring [16,17] of Cu-ZSM-5, or specific sites along the 10-membered rings of the catalyst [31,35,36], where the Cu Cu distances of the Cu+ pair are only slightly larger than in the present system and the coordination of Cu+ to framework oxygen atoms is similar.

2. Methods The catalytic site, represented by the so-called M7 ring [16,17] of Cu-ZSM-5, was simulated by means of the Si7 Al2 Cu2 O26 H16 cluster, shown in Fig. 1. In a recent work [37], we showed that the

97

above structure is large enough to account for all the interactions of Cu+ ions with the zeolite framework as well as for the interactions involved in the adsorption of NO. It was also shown that, at least for the investigated system, the inclusion of a larger portion of the zeolite framework by means of the ONIOM [38–42] approximation does not appreciably change the results with respect to a simple cluster approach [37]. Taking the crystallographic structure of orthorhombic H-ZSM-5 [43] as the starting point, the Si7 Al2 Cu2 O26 H16 cluster was obtained by Al/Si substitution at the T1 and T7 sites and consequent Cu+ addition. The free valence of each terminal O atom was saturated by a H atom initially placed 1.0 Å far from the corresponding O, along the bond with the next Si atom not included in the cluster. The total cluster charge was set to zero throughout the calculations, so that a formal Cu+ ion corresponds to each Al3+ ion of the zeolite framework. As in previous work [37,44], geometry optimizations were performed as follows: in a first step, a pre-optimization, all interatomic distances and the Cu+ coordinates of the Si7 Al2 Cu2 O26 H16 cluster were optimized keeping the zeolite bond angles and dihedrals frozen at the corresponding crystallographic values. This was done in order both to adapt the system to the Al/Si substitution and to take into account that DFT calculations slightly overestimate Si O distances with respect to experimental values [14,37,44]. In the second step, all terminal OH groups of the above cluster were frozen at the positions obtained after the first step in order to maintain the specific geometry constraints of ZSM-5, and the remaining geometrical parameters were fully optimized. The so-obtained cluster was further employed with the same geometrical constraints (frozen OH groups) to simulate NO adsorption and reactivity. After optimization, each structure was vibrationally characterized, checking for the absence of imaginary frequencies in the minima and for the presence of only one imaginary frequency in the transition states. The great majority of the structures optimized in the present work, unless differently indicated, have a triplet-state wavefunction. All calculations were performed by the Gaussian-09 code [45] The B3LYP functional [46,47] was employed, both for the sake of comparison with previous work [24–32] and also because less computationally expensive functionals such as, for instance, BLYP [48] or PBE [49,50] were shown to overestimate the adsorption energy of NO [37], thus leading to potential artefacts in the energetics of the investigated reaction paths. The so-called Ahlrichs’ “def2” basis sets [51,52] were employed for all calculations as follows: def2-TZV for saturating H atoms, def2-TZVP for N, O, Al, Si and def2-QZVP for Cu. Such a combination of basis sets turned out [44] to minimize the Basis Set Superposition Error (BSSE) both in the coordination of Cu+ to the zeolite framework and in the interaction of NO with coordinated Cu+ . In particular, BSSE effects in the adsorption energy of NO were shown to be less than 1.0 kcal mol−1 [44], which is a rather small value with respect to the energy changes involved in the investigated reaction paths. For this reason, no BSSE corrections are provided in the present work. The output of the calculations was inspected and re-elaborated by the molecular visualization program Molden [53] and figures were obtained by the Mercury [54] software.

3. Results and discussion In the crystal structure of ZSM-5 [43], the so-called M7-rings [16,17,37,44] are found in specular pairs on the walls of the linear channels, which run along the b crystal axis (Fig. 1). Each pair of specular rings is delimited by a pair of parallel sinusoidal channels (which run along the a crystal axis) so that the M7-rings are also at the intersection between the linear and the sinusoidal channels. T1 sites are exactly at the edge between the linear and the sinu-

98

S. Morpurgo / Molecular Catalysis 434 (2017) 96–105

Fig. 1. Location of the investigated cluster within the crystal structure of ZSM-5 viewed along (a) the linear channels and (b) the sinusoidal channels; (c) the Si7 Al2 Cu2 O26 H16 cluster employed in the calculations with atom numbering according to [43]. Note that Si and O atoms are independently numbered.

soidal channels, whereas T7 sites are part of the common wall of both channels and are also easily accessible. Fig. 1(c) shows the Si7 Al2 Cu2 O26 H16 cluster effectively employed for calculations, with the Si/Al substitution at the T1 and T7 positions and both Cu+ ions coordinated by two framework oxygen atoms. Copper ions coordinated at the T1 and T7 positions will hereafter be referred to as Cu1 and Cu7, respectively. 3.1. NO adsorption The most relevant structures obtained after adsorption of one or two NO molecules are shown in Fig. 2. The cluster indicated as ZCu2 represents the bare catalyst and its electronic energy plus zero-point energy (ZPE), added to that of two isolated NO molecules, was adopted as the zero of the energy scale for all structures optimized in the present work, unless differently specified. Adsorption of one NO molecule (the second one is still considered as non-interacting) resulted in four distinct structures, the most stable of which (−40.8 kcal mol−1 ) is indicated as ZCu2 –NO, where nitrogen oxide is N-down adsorbed, bridging both copper ions of the catalyst (-1 coordination). ZCu2 –ON, with a similar geometry as the former one but O-down adsorbed, is the least stable (−12.3 kcal mol−1 ). In the structures indicated as ZCu(1) NOCu(7) and ZCu(1) ONCu(7) (−32.9 and −33.3 kcal mol−1 , respectively) nitrogen oxide interacts with both N and O atoms and its axis is nearly parallel to the Cu· · ·Cu axis (cis--1,2 coordination). Using the previously employed 0.9631 scaling factor [44], the calculated N O stretching frequencies for the last four structures are respectively 1683, 1401, 1603 and 1609 cm−1 . The evidence that all of them are lower with respect to that of isolated NO (1904 cm−1 ) [55] suggests that ␲-back donation from copper d orbitals to NO antibonding ␲ orbitals is prevalent over ␴-donation from NO to the 4s orbital of Cu. However, considering that the lowest adsorption energy and N O stretching frequency were calculated for ZCu2 –ON, where ␴-donation does not occur for geometrical reasons (the second HOMO of NO is a ␴ orbital mostly located on the N atom) and that the adsorption energy increases almost linearly with the N O stretching frequency, the observed trend should be ascribed to a concomitant increase of ␴-donation. The adsorption of a second NO molecule by ZCu2 –NO leads to the formation of ZCu2 (NO)2 , ZCu(1) NOCu(7) ON and ZCu(1) ONCu(7) NO (Fig. 2). The latter two structures, with ZCu2 (ON)2 , can also be derived from NO adsorption by ZCu2 –ON. It is worth mentioning that, although ZCu2 –NO and ZCu2 –ON are bridged structures, the coordination

of a second NO molecule to any of the copper ions leads, during geometry optimization, to bridge breaking and to the formation of a pair of isolated Cu-mononitrosyls. In this respect, the relative energies of the above dinitrosyls are approximately the sum of about −25 and −10 kcal mol−1 , respectively calculated for the ZCu–NO interaction in CuZSM-5 [37,44] and for the ZCu–ON interaction in zeolite-like models [28]. In general, the above dinitrosyls could not be shown to be involved in any further reactivity and are expected to undergo only adsorption-desorption equilibria. On the other hand, the O-down bridged mononitrosyl ZCu2 –ON, upon adsorption of a second NO molecule, can also give rise to some additional structures. Those indicated as P1a–P6a, in Fig. 2 are characterized the presence of the ONNO fragment, already shown to be involved [19,22–24,28] in the mechanism of NO decomposition. In P1a–P4a, ONNO is present in its cis form, with one of the terminal oxygen atoms bridging the copper ions and the other one coordinating a single copper ion of the pair. P1a differs from P2a for the slightly different orientation of the ONNO fragment with respect to the Al–Cu–Cu–Al plane, and the same is observed for P3a vs. P4a. P5a and P6a display ONNO in its trans form, with one of the terminal oxygen atoms bridging the copper ions of the catalyst and the other one pointing towards the inner part of the zeolite channel. The interaction energy of P1a–P4a is about 40 kcal mol−1 , whereas that of P5a–P6a is about 23 kcal mol−1 (Fig. 2). In general, the ONNO fragment observed in the present adducts can be identified as the hyponitrite ion, N2 O2 2− [19,22,23] for the following reasons: (i) the N N distance is about 1.23-1.25 Å, which is far shorter than the value of 2.18–2.28 Å experimentally observed [56,57] or calculated [58–62] for the simple NO dimer; (ii) a N N distance of 1.27 and 1.29 Å was respectively calculated in this work for the trans and the cis isomers of the hyponitrite ion optimized in vacuum at the B3LYP/def2-TZVP computational level; (iii) as shown in Fig. 2, the calculated Mulliken charges of copper ions increase from about 0.7 a.u. in ZCu2 to 1.0–1.1 a.u. in P1a–P4a and to 0.9–1.0 a.u. in P5a–P6a, whereas the absolute value of the negative charges on oxygen atoms in the ONNO fragment is larger than in other adsorbed species such as mono- and dinitrosyls; (iv) at the same time, the Mulliken spin density on copper ions (Fig. 2) increases from 0.0 a.u. of the bare catalyst (ZCu2 ) to about 0.6 a.u. in P1a–P4a and to 0.3 a.u. in P5a–P6a, suggesting that copper is at least partially oxidized from formal Cu+ (d10 electronic configuration) to Cu2+ (d9 ). Taking into account that the latter effects are not observed when isolated mononitrosyls are formed on Cu+ ions, the above observations suggest that the ONNO-catalyst interaction in P1a–P6a results

S. Morpurgo / Molecular Catalysis 434 (2017) 96–105

99

Fig. 2. The bare catalyst (ZCu2 ) and its complexes resulting from adsorption of one or two NO molecules. For the sake of clarity, terminal OH groups (see Fig. 1) are omitted. The relative energy (electronic energy + ZPE) of each structure, shown in brackets, was calculated with respect to that of the free catalyst (ZCu2 ) plus two isolated NO molecules and expressed in kcal mol−1 . All structures, unless differently indicated, are triplet states (2S + 1 = 3). Selected interatomic distances, Mulliken atomic charges (in red) and spin densities (in blue/italics) are also reported.

100

S. Morpurgo / Molecular Catalysis 434 (2017) 96–105

in copper oxidation with concomitant charge transfer to ONNO, i.e. in the formation of formal Cu2+ and hyponitrite-like species. NO adsorption by ZCu(1) NOCu(7) and ZCu(1) ONCu(7) generated the structures indicated as P7a, P8a, ZCu(1) (NO)2 Cu(7) and ZCu(1) (ON)2 Cu(7) , also shown in Fig. 2. In P7a and P8a, the trans isomer of ONNO coordinates the catalyst “side-down”, i.e. with the axis of one of its NO subunits nearly parallel to the Cu· · ·Cu axis. The N N distance of 1.37–1.38 Å, although longer than in P1a–P6a, and the distribution of Mulliken charges and spin densities suggest the formation of hyponitrite-like species also in the present structures. In ZCu(1) (NO)2 Cu(7) and ZCu(1) (ON)2 Cu(7) both NO molecules coordinate the same copper ion by their N atom. As a whole, the formation of the latter four structures from the corresponding mononitrosyls is only slightly exothermic (−2.7, −2.0, −3.5 and −2.6 kcal mol−1 , respectively). The existence of transition states connecting ZCu(1) (NO)2 Cu(7) to P7a and ZCu(1) (ON)2 Cu(7) to P8a could not be demonstrated, although the geometries of each pair of structures appear to be potentially related. 3.2. NO decomposition The structures reported in Fig. 2 which undergo further reactivity are those labelled with the P letter, namely P1a, P2a,. . ., P8a. Each of them gives origin to the corresponding reaction path, i.e. P1a to Path-1, P2a to Path-2, and so on. Fig. 3 shows the main geometrical features of the stationary points along the reaction paths labelled as 1–4, originated by P1a–P4a, as well as their relative energy (electronic energy + ZPE) with respect to the bare catalyst (ZCu2 in Fig. 2) plus two isolated NO molecules. In Path-1, the ONNO dimer coordinated to the catalyst in P1a undergoes the breaking the N O bond corresponding to the bridging oxygen through the transition state P1b(TS), and gives rise to the P1c minimum. The activation energy of the latter elementary step turned out to be 3.7 kcal mol−1 , a rather low value with respect to 15–20 kcal mol−1 previously calculated on isolated Cu+ sites [24] or for the reaction catalyzed by Cu+ pairs [28]. The P1c structure is particularly stable (−59.0 kcal mol−1 ) because of the formation of the [Cu O Cu]2+ fragment rather than for the coordination of N2 O to Cu1 through its terminal oxygen atom. The latter interaction was calculated to account for about 2–3 kcal mol−1 , whereas the [Cu O Cu]2+ bridge was already reported [28,35,36] to be a rather stable intermediate in similar processes. Coordinated N2 O in P1c is ready to react again through the P1d(TS) transition state, producing adsorbed N2 and O2 in the P1e minimum. After formation of the reaction products (N2 + O2 ) N2 is the most weakly interacting species, and is readily desorbed, giving the P1f adduct (already indicated as ZCu(cis-␮-1,2-O2 )CuZ) [63]. After final O2 desorption, the catalyst turns back to its initial form (ZCu2 ) and both O2 and N2 are released in the gas phase. Path-2, also reported in Fig. 3, is similar to Path-1, with the only difference of the almost specular orientation of the stationary points with respect to the Al–Cu–Cu–Al plane. Paths 3 and 4 differ from Paths 1 and 2 for the coordination to Cu7, instead of Cu1, of the terminal oxygen atoms of ONNO and N2 O, and of adsorbed N2 after products formation. As already described for Paths 1 and 2, Paths 3 and 4 also differ from each other for the nearly specular orientation of the structures with respect to the Al–Cu–Cu–Al plane. It should be noted that the last adducts along Paths 3 and 4 (O2 coordinated to the catalyst) are still indicated as P1f and P2f because their structures are actually coincident to the corresponding structures along Paths 1 and 2. In general, the optimized structures along Paths 3 and 4 display a slightly lower relative stability than those along Paths 1 and 2, but the energy changes involved in the elementary steps of the mechanism are nearly the same. Fig. 4 shows all minima and transition states along the reaction paths indicated as 5-8, originating from the P5a–P8a adducts

respectively. The trans isomers of hyponitrite, with only one of their terminal oxygen atoms bridging Cu1 and Cu7 such as in P5a and P6a, undergo the breaking of the N O bond corresponding the bridging oxygen through the P5b(TS) and P6b(TS) transition states, respectively. This leads to the formation of the P5c and P6c minima, where the [Cu O Cu]2+ fragment is present and the newly formed N2 O is coordinated by its terminal N atom to Cu1 or to Cu7, respectively. The stability of P5c and P6c is mostly due to the presence of the [Cu O Cu]2+ bridge, with only a modest contribution (about 2–3 kcal mol−1 ) from the N-down coordination of N2 O to copper ions. P5c and P6c undergo oxygen transfer to [Cu O Cu]2+ , with concomitant O2 formation, through the cyclic P5d(TS) and P6d(TS) transition states. The activation energy of the last steps are about 50 kcal mol−1 , similar to those calculated in [28], larger by 5–6 kcal mol−1 than those calculated for Paths 1-4 of the present work but lower by more than 15 kcal mol−1 with respect to those reported in [30]. O2 and N2 , just after formation, are coordinated to the catalyst as in the P5e and P6e minima, which are very stable (–81 kcal mol−1 ) and imply the coordination of O2 with its axis perpendicular to the Cu· · ·Cu axis. Further N2 desorption leads to the P5f structure, which is common to all reaction paths reported in Fig. 4. P5f, after final O2 release, gives rise to gas-phase N2 + O2 plus the bare catalyst (ZCu2 ). P7a and P8a (Fig. 4) are the starting structures of Paths 7 and 8. The N O breaking within the trans isomer of the ONNO dimer occurs in the NO subunit coordinated to the catalyst, through the P7b(TS) and P8b(TS) transition states. After formation of N2 O, coordinated to the [Cu O Cu]2+ fragment of the catalyst (P7c and P8c structures in Fig. 4), Paths 7 and 8 are the same as Paths 5 and 6 up to products formation. The main energetic features of the most representative reaction paths described so far are compared in the diagrams shown in Fig. 5. The first activated step along the reaction path is the decomposition of the hyponitrite species, which has an activation energy ranging from less than 4 kcal mol−1 in Paths 1–4 (Fig. 3) to about 10 kcal mol−1 in Paths 5–6 (Fig. 4) and to about 24 kcal mol−1 in Paths 7–8 (Fig. 4). The evidence that the bond length of activated N O is 1.43 Å in the P1a–P4a minima, 1.36 Å in P5a–P6a and 1.30 Å in P7a–P8a suggests that the degree of bond activation caused by coordination of hyponitrite to the catalytic site is expressed by the length of the activated bond before reaction and is finally reflected in the corresponding activation energy of hyponitrite decomposition. Further, the rate-determinig step of the mechanism is related to the transfer of an oxygen atom from N2 O to the [Cu O Cu]2+ fragment of the oxidized catalyst, independently of the considered path. Slightly lower activation energies (about 45 kcal mol−1 ) were calculated for Paths 1–4 and higher (about 50 kcal mol−1 ) when cyclic transition states were involved, as in Paths 5–8. Notably, the energy change related to O2 desorption (on the average, +28.8 kcal mol−1 in Paths 1–4 and +36.5 kcal mol−1 in Paths 5–8) makes the latter step relatively slow. All reaction paths are characterized by the presence of a rather stable intermediate (structures labelled as P1c–P8c) where N2 O interacts with an oxidized form of the catalyst in which copper ions are bridged (fragment previously indicated as [Cu O Cu]2+ ) by an extra-lattice oxygen atom (ELO), resulting from hyponitrite decomposition. Two isomers could be identified for this species, with the bridging oxygen slightly bent on either side of the Al–Cu–Cu–Al plane, and connected by a transition state. In the absence of interacting N2 O, the structures indicated as 9, 10(TS) and 11 (Fig. 6) were optimized. Their relative energies, calculated with respect to ZCu2 + 2NO including the electronic energy + ZPE of an isolated N2 O molecule, are −56.8, −53.7 and −56.2 kcal mol−1 , showing that 9 and 11 are easily interconverted into each other through the 10(TS) transition state. In the above structures, Mulliken charges (Fig. 6) range from 0.95 to 1.00 a.u. on copper ions (to be compared to

S. Morpurgo / Molecular Catalysis 434 (2017) 96–105

101

Fig. 3. Minima and transition states (TS) optimized along the reaction paths indicated as 1–4 (see text, Section 3.2). For the sake of clarity, terminal OH groups (see Fig. 1) are omitted. The relative energy (electronic energy + ZPE/kcal mol−1 ) of each structure (in brackets) was calculated with respect to that of the free catalyst (ZCu2 ) plus two isolated NO molecules. All structures are triplet states (2S + 1 = 3). Selected interatomic distances, Mulliken atomic charges (in red) and spin densities (in blue/italics) are also reported.

102

S. Morpurgo / Molecular Catalysis 434 (2017) 96–105

Fig. 4. Minima and transition states (TS) optimized along the reaction paths indicated as 5–8 (see text, Section 3.2). For the sake of clarity, terminal OH groups (see Fig. 1) are omitted. The relative energy (electronic energy + ZPE/kcal mol−1 ) of each structure (in brackets) was calculated with respect to that of the free catalyst (ZCu2 ) plus two isolated NO molecules. All structures are triplet states (2S + 1 = 3). Selected interatomic distances, Mulliken atomic charges (in red) and spin densities (in blue/italics) are also reported.

S. Morpurgo / Molecular Catalysis 434 (2017) 96–105

103

Fig. 5. Relative energy diagrams for the mechanism of NO decomposition catalyzed by short-distance Cu+ pairs in Cu-ZSM-5: (a) Path 1; (b) Path 5; (c) Path 7. For the description of reaction paths, see Figs. 3 and 4 and text, Section 3.2.

0.70 a.u. in ZCu2 ) and are about −0.60 a.u. on the bridging oxygen. Mulliken atomic spin densities (Fig. 6) are 0.54–0.56 a.u. on copper ions and 0.61–0.63 a.u. on the bridging oxygen, the latter value to be compared to 2.0 a.u. of the isolated O atom and to 0.0 a.u. of the O2− ion. Both observations clearly show an oxidation of both copper ions with respect to the bare catalyst. With the above structures in mind, some considerations should also be added on the adducts between O2 and the catalyst, namely P1f, P2f (Fig. 3) and P5f (Figs. 4 and 6). The former two structures, already indicated by Goodman et al. as ZCu(cis--1,2-O2 )CuZ [63], differ from each other for the nearly specular orientation of adsorbed O2 with respect to the Al–Cu–Cu–Al plane and for the different Cu· · ·Cu distance, which is 2.95 Å in P1f and 2.69 Å in P2f. In spite of this, the relative energies of both structures (−71.5 and −71.7 kcal mol−1 , respectively) are very similar. The O O bond length in both structures, is moderately longer than in gas-phase O2 (1.30–1.31 Å vs. 1.20 Å, calculated at the B3LYP/def2-TZVP level). Mulliken charges ranging from 0.82 to 0.92 a.u. on copper ions and from −0.13 to −0.18 a.u. on adsorbed O2 , as well as Mulliken spin densities (Fig. 3) show a charge transfer from Cu+ to

adsorbed O2 , but to a lesser extent than in the 9 and 10 structures. P5f (Figs. 4 and 6) corresponds to the structure labelled as ZCu(␮-␩2 :␩2 -O2 )CuZ in [63]. With respect to P1f/P2f, a stronger interaction between copper ions and O2 is suggested by the larger elongation of the O O distance (1.42 vs. 1.30 Å). Mulliken charges (about 1.0 a.u. on copper ions and −0.33 a.u. on both atoms of adsorbed O2 ) and spin densities (0.46–0.47 a.u. on copper ions) show that in the present adducts copper is oxidized almost to the same extent as in 9 and 11, where it can very likely be identified as formal Cu2+ . Interestingly, the sum of the Mulliken charges on both ELO atoms in P5f is close to the charge on the bridging oxygen in 9 and 11. Therefore, if the latter species are identified as O2− ions, it may be reasonable to consider adsorbed O2 in P5f as a peroxidelike species. As a term of comparison, the trans H2 O2 molecule was optimized at the B3LYP/def2-TZVP level, giving an O O bond length of 1.46 Å and a charge fraction of −0.33 a.u. on each oxygen atom, in good agreement with the corresponding parameters in P5f. Having considered that for all reaction paths examined so far the rate-determining step is the oxygen transfer from N2 O to the [Cu O Cu]2+ fragment of the oxidized catalyst, with activation

104

S. Morpurgo / Molecular Catalysis 434 (2017) 96–105

Fig. 6. Additional minima and transition states (TS) optimized for the mechanism of NO decomposition catalyzed by short-distance Cu+ pairs in Cu-ZSM-5 (see text, Section 3.2). For the sake of clarity, terminal OH groups (see Fig. 1) are omitted. For structures 9–13, the relative energy (electronic energy + ZPE/kcal mol−1 , shown in brackets) refers to that of non-interacting ZCu2 + 2NO. For structures 14, 15(TS) and 16, the relative energy refers to that of non-interacting ZCu2 + N2 O. All structures, unless differently indicated, are triplet states (2S + 1 = 3). Selected interatomic distances, Mulliken atomic charges (in red) and spin densities (in blue/italics) are also reported.

energies ranging from 45 to 50 kcal mol−1 , the considerable stability of P5f lead us to search for a transition state that could closely be related to the latter structure, hopefully allowing for a lower activation energy. This was achieved by performing a stepwise constrained optimization in which N2 was let to approach P5f along a direction nearly collinear with the axis of adsorbed O2 , and then optimizing as a transition state the point of maximum energy along the path. As a matter of fact, the transition states labelled as 12(TS) and 13(TS) reported in Fig. 6 were located, with the related imaginary frequency correctly associated to the transfer of the intermediate oxygen atom of the N O· · ·O structure from nitrogen to the extra-lattice oxygen and vice-versa. Unfortunately, no minima directly connected by the above transition states could be optimized because, unless geometrical constraints are applied, there is no way neither for N2 to interact with P5f along a direction collinear with the O2 axis, nor for N2 O to interact with 9 or 11 along a direction approximately perpendicular to the Cu O Cu plane. For these reasons, the 12(TS) and 13(TS) structures may be regarded as transition states connecting reactants and products where both N2 O and N2 are in the gas phase. The latter transition states are isoenergetic (−17.3 kcal mol−1 with respect to ZCu2 + 2NO) which implies activation energies of 39.5 and 38.9 kcal mol−1 (Figs. 5 and 6) for the reactions 9 + N2 O → 12(TS) → P5f + N2 and 11 + N2 O → 13(TS) → P5f + N2 , respectively. Finally, the stationary points associated to the direct N2 O decomposition on the bare ZCu2 catalyst were searched for, in order to compare the behaviour of the active site investigated in the present work with those described in [36]. The structures indicated as 14, 15(TS) and 16 (Fig. 6) were located imposing the wavefunction to be a singlet state. Their relative energies with respect to isolated ZCu2 + N2 O are −9.4, 6.5 and −32.4 kcal mol−1 , respectively. The activation energy for the 14 → 15(TS) → 16 reaction is therefore 15.9 kcal mol−1 in the singlet state, very similar to that obtained in [36]. Although the detailed crossing point(s) between the curves pertaining to the singlet and to the triplet state of the system as a function of the N O distance of reacting N2 O were

not determined, significant differences with respect to [36] are not expected and the active site investigated in the present work is also very likely to catalyze the direct N2 O decomposition with a reasonably low activation energy. 4. Summary and conclusions In the present work, a consistent number of stationary points were optimized for the complete NO decomposition catalyzed by short-distance Cu+ pairs in Cu-ZSM-5. The main conclusions can be summarized as follows: (1) Three general kinds of reaction path could be defined, depending on the geometry of the complexes in which two NO molecules are initially adsorbed on the bare catalyst. (2) As for the reaction promoted by a single Cu+ site [20–26,29] or by pairs of Cu+ ions located at a longer distance [28,30], the first step of the mechanism consists in the formation of hyponitritelike species [19,22,23], which may coordinate the copper ions of the catalyst in different geometries. (3) Hyponitrite decomposition occurs with activation energies ranging from about 4 to 24 kcal mol−1 , depending on its initial coordination to the active site, and results in the formation of an oxidized form of the catalyst, [Cu O Cu]2+ plus a N2 O molecule, which is weakly bound to the active site. The relative stability of the latter complex (59–60 kcal mol−1 with respect to the reactants) is both due to the formation of the [Cu O Cu]2+ bridge and of the stable N2 O molecule. The weak coordination of N2 O to the catalyst only accounts for 2–3 kcal mol−1 . (4) N2 O decomposition occurs with oxygen transfer from N2 O to [Cu O Cu]2+ , and with the formation of N2 and O2 , both adsorbed on the catalyst. Three different kinds of transition states were optimized for the latter step, with activation energies ranging from 39–40, to 44–45, and to 50–52 kcal mol−1 , respectively. As already calculated for single-Cu+ [20–26,29] and for other Cu+ -pair catalysts [28,30], this is still the slow

S. Morpurgo / Molecular Catalysis 434 (2017) 96–105

step of the mechanism, although with a slightly lower activation energy. This represents the main conclusion of the present study. (5) Further N2 desorption occurs easily, whereas O2 desorption is endothermic (from 28.8 to 36.5 kcal mol−1 ) especially when O2 remains adsorbed with its axis perpendicular to the Cu· · ·Cu axis. (6) It is additionally shown that, also on the present active site, the best way for N2 O to decompose may be represented by the direct, spin-forbidden reaction with the reduced Cu+ · · ·Cu+ pair, which has an activation energy of 15–16 kcal mol−1 or even lower, as already evidenced in the literature for a different site [35,36]. (7) However, if N2 O decomposition occurs as just indicated, the stable [Cu O Cu]2+ structure is produced again and the main problem of the mechanism is still represented by the need to reobtain a reduced Cu+ · · ·Cu+ pair, in order to restart the catalytic cycle. With such a consideration in mind, the kinetic advantage obtained by the spin-forbidden N2 O decomposition on a reduced Cu+ · · ·Cu+ pair over the spin-allowed N2 O decomposition on [Cu O Cu]2+ may be only apparent. As a perspective for future studies, it is experimentally known that NO itself acts as a reaction promoter for N2 O decomposition [11] and that NO2 is a possible reaction intermediate [64–66]. Computational studies aimed at elucidating a possible role of the above two species within the overall mechanism are currently in progress. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25]

M. Iwamoto, H. Hamada, Catal. Today 10 (1991) 57–71. M. Shelef, Chem. Rev. 95 (1995) 209–225. V.I. Pârvulescu, P. Grange, B. Delmon, Catal. Today 46 (1998) 233–317. F. Garin, Appl. Catal. A 222 (2001) 183–219. M. Iwamoto, S. Yoko, K. Sakai, S. Kagawa, J. Chem. Soc. Faraday Trans. I 77 (1981) 1629–1638. M. Iwamoto, H. Furukawa, Y. Mine, F. Uemura, S. Mikuriya, S. Kagawa, J. Chem. Soc. Chem. Commun. (1986) 1272–1273. J. Valyon, W.K. Hall, Catal. Lett. 19 (1993) 109–119. M.C. Campa, V. Indovina, G. Minelli, G. Moretti, I. Pettiti, P. Porta, A. Riccio, Catal. Lett. 23 (1994) 141–149. Y. Li, W.K. Hall, J. Catal. 129 (1991) 202–215. M. Lo Jacono, G. Fierro, R. Dragone, X. Feng, J. d’Itri, W.K. Hall, J. Phys. Chem. B 101 (1997) 1979–1984. P.J. Smeets, M.H. Groothaert, R.M. van Teeffelen, H. Leeman, E.J.M. Hensen, R.A. Schoonheydt, J. Catal. 245 (2007) 358–368. G. Spoto, A. Zecchina, S. Bordiga, G. Ricchiardi, G. Martra, G. Leofanti, G. Petrini, Appl. Catal. B: Environ. 3 (1994) 151–172. G. Moretti, Catal. Lett. 28 (1994) 143–152. M. Sierka, J. Sauer, Faraday Discuss. 106 (1997) 41–62. J. Sauer, M. Sierka, J. Comput. Chem. 21 (2000) 1470–1493. D. Nachtigallova, P. Nachtigall, M. Sierka, J. Sauer, Phys. Chem. Chem. Phys. 1 (1999) 2019–2026. D. Nachtigallova, P. Nachtigall, J. Sauer, Phys. Chem. Chem. Phys. 3 (2001) 1552–1559. P. Spuhler, M.C. Holthausen, D. Nachtigallova, P. Nachtigall, J. Sauer, Chem. Eur. J. 8 (2002) 2099–2115. R. Ramprasad, K.C. Hass, W.F. Schneider, J.B. Adams, J. Phys. Chem. B 101 (1997) 6903–6913. B.L. Trout, A.K. Chakraborty, A.T. Bell, J. Phys. Chem. 100 (1996) 17582–17589. A.T. Bell, Catal. Today 38 (1997) 151–156. W.F. Schneider, K.C. Hass, R. Ramprasad, J.B. Adams, J. Phys. Chem. B 101 (1997) 4353–4357. W.F. Schneider, K.C. Hass, R. Ramprasad, J.B. Adams, J. Phys. Chem. B 102 (1998) 3692–3705. N. Tajima, M. Hashimoto, F. Toyama, A.M. El-Nahas, K. Irao, Phys. Chem. Chem. Phys. 1 (1999) 3823–3830. D. Sengupta, J.B. Adams, W.F. Schneider, K.C. Hass, Catal. Lett. 74 (2001) 193–199.

105

[26] X. Solans-Monfort, V. Branchadell, M. Sodupe, J. Phys. Chem. B 106 (2002) 1372–1379. [27] X. Solans-Monfort, M. Sodupe, V. Branchadell, Chem. Phys. Lett. 368 (2003) 242–246. [28] S. Morpurgo, G. Moretti, M. Bossa, J. Mol. Catal. A-Chem. 358 (2012) 134–144. [29] X. Liu, Z. Yang, R. Zhang, Q. Li, Y. Li, J. Phys. Chem. C 116 (2012) 20262–20268. [30] X. Liu, Z. Yang, Y. Li, F. Zhang, J. Mol. Catal. A-Chem. 396 (2015) 181–187. [31] J.S. Woertink, P.J. Smeets, M.H. Groothaert, M.A. Vance, B.F. Sels, R.A. Schoonheydt, E.I. Solomon, Proc. Natl. Acad. Sci. U. S. A. 106 (2009) 18908–18913. [32] P.J. Smeets, R.G. Hadt, J.S. Woertink, P. Vanelderen, R.A. Schoonheydt, B.F. Sels, E.I. Solomon, J. Am. Chem. Soc. 132 (2010) 14736. [33] P.J. Smeets, J.S. Woertink, B.F. Sels, E.I. Solomon, R.A. Schoonheydt, Inorg. Chem. 49 (2010) 3573–3583. [34] P. Vanelderen, B.E.R. Snyder, M.-L. Tsai, R.G. Hadt, J. Vancauwenbergh, O. Coussens, R.A. Schoonheydt, B.F. Sels, E.I. Solomon, J. Am. Chem. Soc. 137 (2015) 6383–6392. [35] M.H. Groothaert, J.A. van Bokhoven, A.A. Battiston, B.M. Weckhuysen, R.A. Schoonheydt, J. Am. Chem. Soc. 125 (2003) 7629–7640. [36] M.-L. Tsai, R.G. Hadt, P. Vanelderen, B.F. Sels, R.A. Schoonheydt, E.I. Solomon, J. Am. Chem. Soc. 136 (2014) 3522–3529. [37] S. Morpurgo, J. Comput. Chem. 36 (2015) 660–669. [38] M. Maseras, K. Morokuma, J. Comput. Chem. 16 (1995) 1170–1179. [39] M. Svensson, S. Humbel, K. Morokuma, J. Chem. Phys. 105 (1996) 3654–3661. [40] M. Svensson, S. Humbel, R.D.J. Froese, T. Matsubara, S. Sieber, K. Morokuma, J. Phys. Chem. 100 (1996) 19357–19363. [41] S. Dapprich, I. Komáromi, K.S. Byun, K. Morokuma, M.J. Frisch, J. Mol. Struct. (Theochem) 461–462 (1999) 1–21. [42] T. Vreven, K. Morokuma, J. Comput. Chem. 21 (2000) 1419–1432. [43] H. van Koningsveld, Acta Cryst. B 46 (1990) 731–735. [44] S. Morpurgo, G. Moretti, M. Bossa, Theor. Chem. Acc. 31 (2012) 1180–1191. [45] M.J. Frisch, G.W. Trucks, H.B. Schlegel, G.E. Scuseria, M.A. Robb, J.R. Cheeseman, G. Scalmani, V. Barone, B. Mennucci, G.A. Petersson, H. Nakatsuji, M. Caricato, X. Li, H.P. Hratchian, A.F. Izmaylov, J. Bloino, G. Zheng, J.L. Sonnenberg, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, T. Vreven, J.A. Montgomery Jr., J.E. Peralta, F. Ogliaro, M. Bearpark, J.J. Heyd, E. Brothers, K.N. Kudin, V.N. Staroverov, T. Keith, R. Kobayashi, J. Normand, K. Raghavachari, A. Rendell, J.C. Burant, S.S. Iyengar, J. Tomasi, M. Cossi, N. Rega, J.M. Millam, M. Klene, J.E. Knox, J.B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R.E. Stratmann, O. Yazyev, A.J. Austin, R. Cammi, C. Pomelli, J.W. Ochterski, R.L. Martin, K. Morokuma, V.G. Zakrzewski, G.A. Voth, P. Salvador, J.J. Dannenberg, S. Dapprich, A.D. Daniels, O. Farkas, J.B. Foresman, J.V. Ortiz, J. Cioslowski, D.J. Fox, Gaussian 09, Revision B.01, Gaussian, Inc Wallingford CT, 2010. [46] C. Lee, W. Yang, R.G. Parr, Phys. Rev. B 37 (1988) 785–789. [47] A.D. Becke, J. Chem. Phys. 98 (1993) 1372–1377. [48] A.D. Becke, Phys. Rev. A 38 (1988) 3098–3100. [49] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865–3868. [50] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 78 (1997) 1396. [51] F. Weigend, F. Furche, R. Ahlrichs, J. Chem. Phys. 119 (2003) 12753–12762. [52] F. Weigend, R. Ahlrichs, Phys. Chem. Chem. Phys. 7 (2005) 3297–3305. [53] G. Schaftenaar, H.J. Noordik, J. Comput.-Aided Mol. Design 14 (2000) 123–134. [54] C.F. Macrae, I.J. Bruno, J.A. Chisholm, P.R. Edgington, P. McCabe, E. Pidcock, L. Rodriguez-Monge, R. Taylor, J. van de Streek, P.A. Wood, J. Appl. Cryst. 41 (2008) 466–470. [55] K.J. Irikura, J. Phys. Chem. Ref. Data 36 (2) (2007) 389–397. [56] W.N. Lipscomb, F.E. Wang, W.R. May, E.L. Lippert Jnr, Acta Crystallogr. 14 (1961) 1100–1101. [57] S.G. Kukolich, J. Am. Chem. Soc. 104 (1982) 4715–4716. [58] R. González-Luque, M. Merchán, B.O. Roos, Theor. Chim. Acta 88 (1994) 425–435. [59] B.S. Jursic, Chem. Phys. Lett. 236 (1995) 206–210. [60] A.L.L. East, J. Chem. Phys. 109 (6) (1998) 2185–2193. [61] M. Tobita, S.A. Perera, M. Musial, R.J. Bartlett, M. Nooijen, J.S. Lee, J. Chem. Phys. 119 (2003) 10713–10723. [62] Z. Hui, Z. Gui-Li, L. Gang, G. Yi-Zhao, J. Qing, Chin. Phys. B 24 (9) (2015) 093101. [63] B.R. Goodman, K.C. Hass, W.F. Schneider, J.B. Adams, J. Phys. Chem. B 103 (1999) 10452–10460. [64] B. Modén, P. Da Costa, B. Fonfé, D.K. Lee, E. Iglesia, J. Catal. 209 (2002) 75–86. [65] B. Modén, P. Da Costa, D.K. Lee, E. Iglesia, J. Phys. Chem. B 106 (2002) 9633–9641. [66] P.J. Smeets, B.F. Sels, R.M. van Teeffelen, H. Leeman, E.J.M. Hensen, R.A. Schoonheydt, J. Catal. 256 (2008) 183–191.