A DFT Study of CO2 Hydrogenation on Faujasite‐Supported Ir4 Clusters: on the Role of Water for Selectivity Control

Abstract Reaction mechanisms for the catalytic hydrogenation of CO2 by faujasite‐supported Ir4 clusters were studied by periodic DFT calculations. The reaction can proceed through two alternative paths. The thermodynamically favoured path results in the reduction of CO2 to CO, whereas the other, kinetically preferred channel involves CO2 hydrogenation to formic acid under water‐free conditions. Both paths are promoted by catalytic amounts of water confined inside the zeolite micropores with a stronger promotion effect for the reduction path. Co‐adsorbed water facilitates the cooperation between the zeolite Brønsted acid sites and Ir4 cluster by opening low‐energy reaction channels for CO2 conversion.


Introduction
Development of new technologiest hat utilise carbon dioxide as aC 1 buildingblock forthe production of value-added chemical and fuels is considered as one of the key steps towards the green and sustainable chemical industry.R eductive transformationo fc arbon dioxide attractsp articulara ttention of the scientificc ommunity as it provides am ethod to store renewable energy in chemical bonds. [1] For example, catalytic hydrogenationo fC O 2 is one of the pivotale lements of the so-called methanole conomy advocated by the Nobel laureate George Olah. [2] Methanol is ak ey chemical that can potentially substitute fossil fuels and, at the same time, serve as av ersatile platform for the production of aw ide range of chemicals. [2,3] Catalytic hydrogenationo fC O 2 can proceed along three paths, namely (i)hydrogenation to formic acid (FA) or methanol, (ii)methanation to CH 4 and (iii)reduction to CO. The last two processesa re usually performed in the gas phase in the presence of heterogeneous catalysts, whereas the direct hydrogenationr eaction for the production of FA from CO 2 is performed in the liquid phase. [3] The possibility to convert CO 2 to methane (methanation process) in the presence of bulk metals was discovered by Sabatier at the beginning of the 20th century.T his discoveryw as followed by extensive research focusedo nt he optimisation of both catalysta nd process. [4] The key step in this reaction was found to be the dissociation of the CO 2 molecule on am etal surface. [5] Ar epresentative example of as upported metal catalyst for such ap rocess is zeolite Y-supported Ni/Ce nanoparticles. [6] The high reactivity of this catalysts was attributed to the combined effect of ceria to promote the initial CO 2 dissociation and the subsequent fast hydrogenation catalysed by Ni. [6] The catalysts formally having lower hydrogenation activity form the basis for the reductivec onversion of CO 2 to CO, which can then be utilised as ac onventional C 1 buildingb lock in well-establishedp rocesses such as methanols ynthesis, Fischer-Tropsch synthesis, or carbonylation. [7] With respect to the latter,L osch et al. have recently reported an easy and green carbonylation technology involving CO generatedi nsitu upon the decomposition of FA over Brønsted acidic H-ZSM-5 zeolite at mild conditions. [8] Another route involves the direct hydrogenation of CO 2 with H 2 without prior CO 2 dissociation. This reaction is central to at echnology for H 2 storage in the form of al iquid fuel. [9] Significant progress in the direct CO 2 hydrogenation has been made in recent years with the development of new highly efficient and robusth omogeneous transition-metal catalysts. [3] In particular, Ru [10,11] and Ir-based [12,13] homogeneous catalysts at-Reactionm echanisms for the catalytic hydrogenation of CO 2 by faujasite-supportedI r 4 clusters were studied by periodic DFT calculations. The reactionc an proceed throught wo alternative paths. The thermodynamically favoured path results in the reduction of CO 2 to CO, whereas the other,k inetically preferred channel involves CO 2 hydrogenation to formic acid under water-free conditions. Both paths are promotedb ycatalytic amounts of water confined inside the zeolite micropores with as tronger promotion effect for the reduction path. Coadsorbed water facilitates the cooperation between the zeolite Brønsted acid sites and Ir 4 cluster by opening low-energy reactionchannels for CO 2 conversion. tract much attention in view of their exceptional activity and stabilityi nt he catalytic CO 2 hydrogenation. Very recently,i t was shownt hat the Ir-based pincer catalysts do not only have an exceptional catalytic performance in CO 2 hydrogenation to formates, [12] but also are able to promote the reductive cleavage to CO, [14] which involves ac ooperative action between the transition metal centre and an eighbouring Brønsted acid functionality of the ligand. [11a] FA is not only a promising energy carrier,b ut also an important intermediate for the chemical industry.T he introduction of more sustainable and efficient technologiesf or its production from renewable sources such as CO 2 is desired. Although homogeneous catalysts show great potentialf or practical reversible H 2 storaget echnologies, [9,15] commercial FA synthesis in this way is hampered by challenges in catalyst regeneration and separation of reactants. [7] Therefore, an efficient and selective heterogeneousc atalystf or CO 2 hydrogenationt oF Ai s desired.
The first example of FA synthesis from CO 2 and H 2 was reported in 1914 by Brediga nd Carter, who demonstrated the possibility of catalytic hydrogenation of CO 2 and aqueous alkali carbonates to FA under mild conditions in the presenceo f bulk Pd. [16] After approximately 100 years from these initial observations, ac atalystf ormulation of the 21st centuryt hat also utilised Pd as the key component has been reported. Bi et al. have recently described ah ighly active system for reversible CO 2 hydrogenation based on ar educed graphene oxide supported Pd catalyst. [17] Metal oxide supported Au nanoparticles have also been studied for the liquid-phase CO 2 hydrogenation. [18a,b] In addition, grafting of Ir and Ru complexes onto mesoporouss upports has been considered as ap romisings trategy towardsheterogeneous CO 2 hydrogenation catalysts. [18c,d,e] Zeolitesr epresent an importantc lass of aluminosilicate materials suitable for the design of new efficient heterogeneous CO 2 hydrogenation catalysts. They combineh igh stability, aw ell defined porous system and relatively low cost with an outstanding degree of chemical tunability.T he isomorphous substitution of ap art of the silicon atoms in their framework for aluminiumg enerates al ocal negative charge on the lattice that is compensatedb ya ne xtra-framework cation.T he compensation of such negativec harges by protons resultsi nt he formationo fs trong Brønsted acid sites. Alternatively,t hese anionic centres together with the surrounding basic lattice oxygen centres can form as uitable ligand environment for the stabilisation of aw ide variety of cationic species with desirable catalytic properties. [19] Furthermore, zeolitesallow the construction of reactive environments featuring multiple reactive centres of different chemical nature inside their molecular-sized cages and channels. These reactive sites can act cooperatively to enablel ow-energy reactionp aths. [20] Such an active site cooperativity is analogous to the related reactivity concept in homogeneous catalysis. [11a] Zeolite-based materials have already been considered as potential catalysts for CO 2 transformations. [21] The hydrogenation of CO 2 over zeolite Brønsted acids ites (BAS) and extra-framework alkali cations has been computationally investigated by Chan and Radom [21a,b] who proposed ap ossibility for the effi-cient conversion of CO 2 by acidic zeolitest hrough ac oncerted mechanism. The reaction involved protonation of the adsorbed CO 2 by zeolite BAS and simultaneoush eterolyticd issociation of H 2 to produce formate and regenerate the BAS. The CO 2 activation and the formationo ft he formate species is ak ey step in the overall process. It can be facilitated by more reactive hydride anions formed upon the H 2 dissociation over at ransitionmetal site. At ransition-metal species stabilised by the zeolite lattice,w hich at the same time provides mobile protons necessary to close the catalytic cycleo ft he FA formation, [10] is ap romising candidate as ah eterogeneous system for CO 2 hydrogenation.
The stabilisation of transition metal clusters inside the zeolite pores has been extensively studied theoretically by the groups of Rçsch and Vayssilov [22] and experimentally by Gates and co-workers. [23] In particular,m olecular-sized Ir 4 clusters stabilized in faujasite (FAU) zeolites were found to be highly active in hydrogenation catalysis. [23c-e] The possibility of the hydrogen spilloverbetween the Ir 4 clusters and the zeolite framework has been proposed as an important feature of the Ir 4 / FAUs ystems relevant to their catalytic performance. [22d] Such molecular-sized clusters show distinctively differentc hemical properties and catalytic reactivity from the bulk nanoparticulate catalytic ensemblesc ommonly employed in conventional heterogeneous catalysis. [24] Because of their small size, the clusters do not possess ad eveloped band structure such as that observed in transition-metal nanoparticles and rather resemble the sites encountered in homogeneous catalysts.
The unique ability of Ir 4 /FAU to create ar eactive environment inside zeolite micropores combining relatively strong BAS [20] and multiple hydride specieso nt he transitionm etal cluster [22,23] renders this system ap romising catalyst for reductive transformationsofCO 2 .Inthis work, we carry out aperiodic density functional theory study on ar ealistic FAUz eolite modelt oe valuate the potentialo ft his system for CO 2 hydrogenation. We investigate the possibility to control the reaction path of CO 2 conversion and analyse the effect of co-adsorbed water molecules inevitably present insidet he zeolite micropores on the catalystp erformance and the reaction mechanism.

Computational Details Models
The reaction paths were analysed in the framework of density functional theory (DFT) using ap eriodic faujasite (FAU) zeolite model with aS it oA lr atio of 2.42 representing the lattice composition of zeolite Y. Similar to our previous studies, [25] the low-symmetry triclinic FAUu nit cell was used as am odel. The optimised cell parameters were a = b = c = 17.51 a nd a = b = g = 608.T he negative lattice charges owing to substitution of the framework Si + 4 by Al + 3 ions were compensated by the extra-framework iridium cluster with an overall charge of + 2a nd protons. The Ir 4 cluster compensated for the negative charge of two lattice Al centres, whereas the remaining 12 anionic lattice sites present in the framework were compensated by H + giving rise to Brønsted acid sites. The faujasite structure and the respective model with embedded Ir 4 cluster are shown in Figure 1. The initial structure and location ChemCatChem 2016ChemCatChem , 8,2500ChemCatChem -2507 www.chemcatchem.org of the Ir 4 unit was constructed on the basis of extensive experimental [22] and computational data [23] previously reported for this system. The current zeolite model represents ar ealistic system, with full periodic structure of the FAUf ramework that accounts for all the effects present in the confined space of the zeolite channels. The only major approximation used in this work is the assumption of the defect-free infinite periodic structure of the zeolite.
The reactive site was represented by ap artially hydrogenated Ir 4 cluster ( Figure 1b). Previous experimental [23] and theoretical [22a,b,c,d] studies showed that such species preferentially adopts atetrahedral configuration inside the zeolite. The formation of alternative planar structures for such species has never been reported before. Our computational results are in line with the previous findings. Chemical transformations catalysed by the zeolite-stabilised Ir 4 cluster result in geometrical perturbations that are limited to minor distortions caused by the interactions with the reactants and do not lead to any significant change in the overall Ir 4 shape. The initial cluster bears three hydride ligands and is stabilised at the SII sixmembered-ring cation site, where it compensates for the charge of two anionic [AlO 4 ] À units in the FAUs upercage. Such ac onfiguration allowed us to assess different mechanisms of CO 2 transformations initiated by the chemisorption of CO 2 to the under-coordinated Ir centre at the vertex of the cluster.T his activation route was identified as the preferred one among as eries of alternative mechanisms considered by the initial computational screening. Besides the mechanistic paths discussed in detail below,t he initial computational assessment included mechanisms such as the outer sphere CO 2 hydrogenation and its activation at the edge sites of Ir 4 .T he structural models for the respective intermediates either converged to the species involved in the main paths discussed below or showed prohibitive thermodynamics of elementary steps and therefore were not considered further.
The results obtained in the course of this study indicate that even the smallest amount of water can significantly affect the reactivity and selectivity of the investigated system. In view of the complexity of the reaction mechanisms considered and the size of the systems, the current mechanistic analysis was limited to the static relaxation simulations. The investigation of the effect of higher water content, although of great interest, was beyond the scope of the current study and would require ap rolonged sampling of the Free Energies by using Molecular Dynamics simulations, which would be prohibitive in terms of computational demand for the analysis of all relevant paths.

Methods
The DFT calculations were performed using the Vienna Ab Initio Simulation Package (VASP). [26] The generalised gradient approximated PBE exchange-correlation functional was used. [27] The electron-ion interactions were described by the projected augmented waves (PAW)m ethod. [28] The Brillouin zone sampling was restricted to the G point. [29] The energy cut-off was set to 500 eV.T he convergence was assumed to be reached if the forces on each atom were below 0.05 eV À1 .S uch computational parameters provide ag ood compromise between the accuracy and computational demands for studying chemical processes in zeolite micropores as we have demonstrated in our previous successful studies on related subjects. [30] Am odest Gaussian smearing was applied to band occupations around the Fermi level, and the total energies were extrapolated to s!0. The nudged-elastic band method (NEB) [31] was used to determine the minimum energy path and to locate the transition state structures along the CO 2 hydrogenation paths. The nature of stationary points was confirmed by determining the vibrational frequencies using the finite difference method. Small displacements of 0.015 w ere used to determine the numerical Hessian matrix. Transition states were identified by verifying the occurrence of an imaginary frequency along the expected reaction coordinate.

Results and Discussion
The reductivet ransformations of carbon dioxide by Ir 4 /FAU are initiated by its chemisorption to the under-coordinated Ir centre at the vertex of the Ir 4 cluster.T his step proceeds with an egligible barriera nd is exothermic by 41 kJ mol À1 .T he adsorptiono fC O 2 is accompanied by the change of the hybridisation from the linear sp to the sp 2 -type resulting in the formation of as trong h 2 -(C,O)-type adsorption complex 1 with the under-coordinated Ir centre (Figure 2). The distance from the non-coordinated oxygen of the adsorbed CO 2 to the nearest Brønsted acid site of the zeolite support is rather large (4.36 ).
Further hydrogenation of the adsorbed CO 2 molecule can take place by Ir-boundh ydrides. The optimised structures of key reactioni ntermediates and the relevant reactione nergy diagram are summarised in Figure 2. Prior to CO 2 hydrogenation, ar eactive hydride ligand migratesf rom the base of the cluster to the Ir atom bound to CO 2 .T his step 1!2 is almost thermoneutral( DE = 12 kJ mol À1 )a nd shows al ow activation barrier (E°)o fo nly 48 kJ mol À1 .T he subsequenta ttack on the ad-sorbedC O 2 by IrÀHt of orm IrÀCOOH (3)i sm uch more difficult (DE = 14 kJ mol À1 , E°= 143 kJ mol À1 ). The highly activated natureo ft his reactiono riginates from the need of repolarisation of the reactive species-thep rotonation of adsorbed CO 2 is done by af ormally anionic hydride.S ubsequentf acile (DE = À8kJmol À1 , E°= 29 kJ mol À1 )r otationo ft he hydroxyl group leads to 4,w hich is the Ir 4 ÀCOOH isomer that is ready for further transformationsa long the hydrogenation reduction paths.
Alternatively, the ÀCOOH intermediate may be formed by proton transfer from the zeolite lattice. This path is however www.chemcatchem.org very unfavourable (E act = 421 kJ mol À1 ), because of the large distance for the proton transfer.W ee xplored the possibility of facilitation of the H + transfer step by aw ater molecule. The corresponding energy diagram for the water-assisted reaction path B is shown in green in Figure 2. To model this mechanism we placed aw ater molecule between the acidic centre of the faujasite and the Ir 4 site (E ads = 115kJmol À1 , 1 + H 2 O!5, Figure 2). One proton of the water molecule in 5 is directed towards the oxygen atom of the CO 2 .T he subsequentt ransfer of the protont of orm an Ir-boundÀCOOH species 6 is an endothermic(102 kJ mol À1 )b arrierless reaction.
Water desorption at the next step destabilises the system by 85 kJ mol À1 .T his step may or may not be neededd epending on which further reactions teps are assumed. The desorption is describedh ere forcompletenessasitdoes not signify the thermodynamic limitations. Ther emoval of H 2 Og ives 7 with the carboxyl orientation similar to that in 4 formed in path A.T he main difference between 4 and 7 lies in the distribution of H atoms in the system, and the energy differenceb etween them effectively determines the natureo ft he confined extra-framework species. Whereas 4 contains af ormally + 2c harged Ir 4 (H) 2 COOH cluster and aB r ønsted acid site remaining at its originalp osition, the "water-assisted carboxyl" path C gives an Ir 4 (H) 3 COOH cluster in 7 with af ormal + 3c harge that is compensated by two vicinall attice anions and one distant [AlO 4 ] À site. Such indirect charge-compensation mechanism has been thoroughly discussed elsewhere. [19a, 32] The interconversion between 4 and 7 can proceed through the reverse hydrogen spillover process, in which the adsorbed CO 2 molecule acts as the hydrogen mediator.I nt he absence of CO 2 ,t he hydrogen spillover from the zeolite Brønsted acid sites to Ir 4 was studied theoretically by the group of Rçsch [22d] and it was found to be strongly exothermic by 168 kJ mol À1 . The lower exothermicity (DE = 54 kJ mol À1 )o ft he reaction in our case is causedb yt he lower reactivity of the hydrogenated cluster bearing additional ÀCOO and Hl igands. Furthermore, the fully periodic nature of our model allowed us to consider the Ht ransfer from ad istant Brønsted acid site to the reactive Ir 4 cluster resulting in as ubstantial chargea lternation that contributes to the effective destabilisation of the system.T he difference in the nature and formal charge of the extra-framework iridium species in 4 and 7 resultsi nt he slight difference in the interatomic distances. The binding to am ore cationic species in 7 results in as lightly longer Ir-C distance of 2.02 than the value of 1.98 i n4.F or comparison, ar elated molecular carboxylate Ir-PNP pincerc omplex [14] has an IrÀCb ond of 1.99 a nd the respective value in bulk iridium carbides ystems is 2.03 . [33] The hydrogenation of CO 2 can also proceed throughamechanism involving formate speciesa st he key intermediate commonly discussed in homogeneous systems. [3,[10][11][12][13] In the respective path C (Figure 2), an Ir 4 -bound formate 8 (Ir 4 ÀOOCH) is formed upon the hydridea ttack on the carbon atom in the ad-sorbedC O 2 insteado ft he carboxyl moiety in path A.T his reac- Figure 2. Reactionenergydiagramfor the first Ht ransfer to the CO 2 molecule.T he relative energiesfor each state(in italic) are given in kJ mol À1 with respect to the sum of the energies of free reactants. The water-free (A)and water-assisted (B)p ath to carboxyl are showni nred and green,respectively.Pathway (C) via af ormatei ntermediatei ss hownin purple. Transition-state structures are denoted by the square brackets.T he Ir cluster is represented by blue spheres and Hiss hown as white spheres. C, O, Si and Al are shown in grey,red, yellow and pink, respectively. ChemCatChem 2016ChemCatChem , 8,2500ChemCatChem -2507 www.chemcatchem.org tion (2!8)f aces am oderate barriero fo nly 79 kJ mol À1 ,w hich is nearly twice as low as the one computed for the water-free carboxyl path A.H owever,i ti su nlikely that the hydride attack can be facilitated by protic solvents. This is in contrast to the alternative proton-transfers tep in the carboxyl routes. This, togetherw ith the comparable energetics computed for the respectivef ormate-(C)a nd water-assisted carboxyl (B)p aths, allow us to propose that depending on the composition of the reactionm edium, both paths can take place affecting thus the selectivity of the CO 2 reduction process. The Ir 4 -OOCH formate complex 8 can be further stabilised by 49 kJ mol À1 by af acile (E°= 12 kJ mol À1 )r otation of the formatea nion resulting in species 9.
Further transformationsa long these paths can yield either the formic acid or CO as the final products. For the formate path C,o ur calculations reveal as ingle path to FA,w hich will be discussed below.F or the carboxyl paths, the selectivity of CO 2 conversion is determined at the next reaction stage. The associatedc omputed reactione nergy diagrams and local optimised geometries of intermediates andt ransition states involved are summarised in Figure 3.
Both paths to FA or CO andH 2 Or equirea ddition of an H 2 molecule to the [Ir 4 (H) 2 -COOH] 2 + complexi n4.A lternative hydrogenation paths involving Ht ransfer from the zeolite lattice or other Ir-bound hydrides are highly unfavourable. H 2 adsorption to the top atom of the cluster is exothermic( E ads = 110kJmol À1 )a nd yields as trongly activated s-H 2 complex [34] 10 with as ubstantially elongated HÀHd istance (0.93 v s. 0.75 i ng as phase H 2 ). SubsequentHtransfert ot he carboxyl ligand determines whethert he formic acid (10!11,p ath A, Figure 3) or carbon monoxidea nd water (10!12,p ath A', Figure 3) are formed. Hydrogenation at the Oa tom in COOH will lead to water formation and subsequent releaseo fC O( A': 10!12!CO + H 2 O + Ir 4 H 3 ), whereas the Ht ransfer to the C atom in COOH directly yields FA product (A: 5!6!FA + Ir 4 H 3 , Figure 3). The barrier to FA formation (E°= 116kJmol À1 )i s lower than that for the decarbonylation path A' (E°= 146 kJ mol À1 )s uggesting that FA is the kinetically preferred product in the current system.
As imilar conclusion can be drawn from the analysiso ft he reactionc hannel via the formate intermediate 8 (path C in Figure3). This catalytic path is continued by H 2 coordination to 8 that is much less exothermic (8 + H 2 !13, DE = À39 kJ mol À1 ) Figure 3. The energy diagram for the final steps of water-free CO 2 hydrogenation cycles on the Ir 4 cluster.T he relative energiesfor each state (in italic) are given in kJ mol À1 with respect to the sum of the energies of free reactants. The hydrogenation route of the "carboxyl path" to FA (A)iss howninred,a nd the respective reduction channelt oc arbon monoxide and water (A')i ss hown in black. The "formate" path C towards formicacid is showninp urple. Transitionstate structuresa re denoted by the square brackets. The Ir clusteri sr epresented by blue spheres and Hi ss hown as white spheres.C ,O ,S iand Al are shown in grey,r ed, yellow andpurple,r espectively. ChemCatChem 2016ChemCatChem , 8,2500ChemCatChem -2507 www.chemcatchem.org than the respective step in the carboxyl paths (4 + H 2 !10, DE = À110kJmol À1 ). Nevertheless, the resulting s-H 2 complex 13 is structurally very similar to 10.T he HÀHd istances in adsorbed H 2 moieties in these structure differ by only 0.01 . Hydrogenolysis of the IrÀOb ond in IrÀOOCH at the next step yields an FA adduct 11 common for both the carboxyl A and formate C paths.T he hydrogenolysis step is exothermic by 47 kJ mol À1 and has ab arrier of 90 kJ mol À1 .T he latter is very close to the overall barrierf or the initial CO 2 activation over this path (1!8, E app = 91 kJ mol À1 )s uggesting that depending on the conditions, either of the elementary processes may become rate-determining. [15b] In view of the strongp romoting effect of co-adsorbed water on the initial CO 2 activation by Ir 4 H 3 /FAU, we also considered aw ater-assisted path for the transformations of the resulting COOH adducts. The corresponding reactione nergy diagram and optimised structureso ft he intermediates and transition states are shown in Figure 4. The startingp oint for the watermediated paths is structure 14 that is a[ Ir 4 (H) 3 COOH] 3 + cation with ac o-adsorbed water molecule. H 2 adsorption to 14 (14! 15)r esembles the analogous process in the water-free system (4!10). The calculated adsorption energy is 63 kJ mol À1 .H ÀH bond length in the s-H 2 complex 15 is 0.90 w ith one of the Ha toms forming ah ydrogen bond with the co-adsorbed H 2 O.
The heterolytic dissociation of H 2 assisted by the co-adsorbed H 2 Om olecule regenerates the zeolite BAS, whichw as consumeda tt he initial step of CO 2 activation (15!16). This reaction is strongly exothermic (DE = À95 kJ mol À1 )a nd shows av ery low activation barrier(E°= 36 kJ mol À1 ).
The presenceo fH 2 Oa lters the selectivity trends discussed above for catalysis via 4.C ontrary to the water-free paths (Figure 3), the formation of FA (path B,g reen line in Figure 4) is now kinetically less favourable than the decarbonylation route A.T he activation barrier for the Ht ransfer to the carbon atom of the adsorbed COOH resultingi nF A( 16!17)i s 46 kJ mol À1 higher than the competing protonation step in path B' to form CO (16!18,F igure 4). However,o wing to av ery strong binding of CO to the Ir 4 cluster,t he regeneration of the initial active speciesi se xpected to be unlikely.O ne can expect that the strongly bound CO may undergo secondary hydrogenation transformations under the catalytic conditions, including the methanation reactiont hat is, however,b eyond the scope of the current manuscript. On the contrary,t he products of the directh ydrogenation path to FA do not form strong bonds with the iridiumc lustera nd can potentially be desorbed to close the catalytic cycle. Desorption of both FA and H 2 Of rom 17 to regenerate Ir 4 H 3 clusteri se ndothermic by only 117kJmol À1 . Figure 4. Water-assisted paths for the hydrogenationofI r-bound carboxyl (B and B')and formate (D)s pecies.T he relative energies for each state (in italic) are given in kJ mol À1 with respect to the sum of the energies of free reactants. The hydrogenation route of the "carboxyl path" to FA (B)isshowningreen,the respectiver eduction channeltoc arbon monoxide and water (B')iss howninblack. The water-assisted" formate" path D to FA is given in orange. Transitionstate structures are denotedbyt he square brackets.T he Ir cluster is represented by blue spheres and Hi ss hown as white spheres. C, O, Si and Al are shown in grey,red, yellow andpurple, respectively. ChemCatChem 2016ChemCatChem , 8,2500ChemCatChem -2507 www.chemcatchem.org The formate path can also benefit from the water assistance in the transfer of the protonf rom the BAS. H 2 Oa dsorption to Ir 4 -OOCHc omplex 9 gives species 20 (DE = À58 kJ mol À1 )t hat is the startingp oint fort he water-assisted reactionp ath (orangep ath D in Figure4). The subsequentp rotont ransfer from zeolite BAS to Ir-bound formate (20!21)y ields adsorbed FA.This step is endothermic by 58 kJ mol À1 and shows amoderate barriero f8 0kJmol À1 .N ext, the FA product is released by an endothermic ligand exchange reaction with molecular H 2 (21 + H 2 !22 + FA, DE = 77 kJ mol À1 ). With the assistance of coadsorbed H 2 O, H 2 undergoes ab arrierless heterolytic dissociation to regenerate Ir 4 H 3 site and BAS (22!23). The energetic cost of the removal of H 2 Of rom 23 (56 kJ mol À1 )i sa lmost identical to the value of H 2 Oa dsorption energy to 9,w hich evidences the catalytic role of co-adsorbed water in this reactionpath ( Figure 4).
These computational resultss uggestt he possibilityo fc ontrolling the reactionthermodynamically andkinetically.The formation of CO occurs with ab arrier of only 66 kJ mol À1 ,b ut leads to av ery stable system from which the products desorption (CO and 2H 2 Om olecules) faces ap rohibitive energy penalty of 366 kJ mol À1 .T he high endothermicity of this step is the result of as trong binding of CO with metals. For example, the experimentally determined CO adsorption energy to Ir (111) surfaceis1 75 kJ mol À1 . [35] In view of such as trong adsorption of CO, its further hydrogenationt oo btain methanol,m ethane or even to promote the CÀCb ond formation upon the coordination to the active Ir 4 site additional CO 2 moleculec annotb ee xcluded. Those anticipatedp roducts would bind weaker with the Ir cluster and as such, their release may become facilitated. The detailed computational analysis of such paths requires as eparate dedicated investigation that is beyondt he scope of this study.
Our findings suggest that by varying the concentration of water in the reactionm edium one can potentially selectively promote the FA synthesis by favouring the formate route or by lowering the kinetic barriers needed for the initial formation of the ÀCOOH speciesi nt he carboxyl path. However,a th igh H 2 O concentrations, the path towards CO formation would become strongly favoured resulting in ar apid deactivation of the catalyst or drastic selectivity change towards the CO hydrogenation products. Analysis of the theoretical results indicates the possibility of choosingthe preferred reaction path through variation of the experimental conditions and by performing the hydrogenation with simultaneous water removal or at very low water content.

Conclusions
CO 2 hydrogenation over zeolite Y-supported Ir 4 H 3 clusters was studied by periodicDFT calculations utilising arealistic faujasite model to simulatearealistic zeolite environmenta dequately. The computational resultsa llowed us to identify two possible reactionp aths for the CO 2 reduction,n amely the direct hydrogenationt owards formic acid and reduction to CO and H 2 O. The reactivity of the system and the reaction selectivity greatly dependso nt he presence of water molecules in the system, which act as proton mediators promoting the proton-transfer reactionsa long the catalytic cycles. Thep romotione ffect of water on hydride transfers to form CÀHb onds is low.
The competing paths are interrelated in an yin-yang manner because water molecules produced along the decarbonylation path can in situ modify the preferred reaction route. Under the water-free conditions, the initial CO 2 hydrogenation to Ir-bound COOH speciess hows ab arrier of 143 kJ mol À1 that is substantially highert han the alternative path to form the Ir-OOCH formate adducts, which are converted further into FA exclusively.H owever,t he initial formation of Ir-COOH carboxyl moiety can be greatly facilitated in the presence of water that mediates ap rotont ransfer from the zeolite lattice to thea dsorbed CO 2 .W ater reduces the barrier for the COOH formationb ya pproximately 40 kJ mol À1 and makes the respective reactionc hannel competitive with the formate path. The co-adsorbed water promotes the selective production of CO by an H 2 O-assisted decarbonylationo fC OOH species through reduction of the corresponding activation barrier by as imilar value of 46 kJ mol À1 with respect to the competing hydrogenation path to FA.
We conclude that it is virtually impossible to obtain the formic acid in an aqueouse nvironment by using the Ir/FAU catalyst, and an effective in situ water-removal has to be ensured if the selectivity towards FA is desired. By performing the reactionu nder H 2 O-free conditions, the decarbonylation path can be suppressed. This is necessary in view of the very strong binding of CO product to the active Ir 4 cluster (E ads = 246 kJ mol À1 ). On the other hand, the desorption of formic acid is much easier (E ads = 94 kJ mol À1 ). We propose that by optimizing process conditions, itss elective formation can potentially be achieved by the liquid-phaseh ydrogenation of CO 2 over an Ir/FAU catalyst.