N‐Heterocyclic Carbene Organocatalysis: With or Without Carbenes?

Abstract In this work the mechanism of the aldehyde umpolung reactions, catalyzed by azolium cations in the presence of bases, was studied through computational methods. Next to the mechanism established by Breslow in the 1950s that takes effect through the formation of a free carbene, we have suggested that these processes can follow a concerted asynchronous path, in which the azolium cation directly reacts with the substrate, avoiding the formation of the carbene intermediate. We hereby show that substituting the azolium cation, and varying the base or the substrate do not affect the preference for the concerted reaction mechanism. The concerted path was found to exhibit low barriers also for the reactions of thiamine with model substrates, showing that this path might have biological relevance. The dominance of the concerted mechanism can be explained through the specific structure of the key transition state, avoiding the liberation of the highly reactive, and thus unstable carbene lone pair, whereas activating the substrate through hydrogen‐bonding interactions. Polar and hydrogen‐bonding solvents, as well as the presence of the counterions of the azolium salts facilitate the reaction through carbenes, bringing the barriers of the two reaction mechanisms closer, in many cases making the concerted path less favorable. Thus, our data show that by choosing the exact components in a reaction, the mechanism can be switched to occur with or without carbenes.


Introduction
Since the discovery of thiamine (vitaminB1), [1,2] the mechanism of the biological processes it catalyzes have been in the focus of research. Next to the direct biochemical studies that revealed the role of this substance in life, numerous organic chemicalm odel reactions have been designed to understand these reactions in furtherd epth.T he synthetic value of the knowledge that was gathered through these organic chemical studies was recognized by Stettere tal., [3] defining the field that we call today N-heterocyclic carbene (NHC) organocatalysis. [4][5][6][7][8] Since then, these transformations offer an ever growing portfolioo fr eactions, including important CÀCc oupling reactions, which can be often performed in as electivea nd asymmetric manner.
One of the most prominente xamples for NHCo rganocatalysis is benzoin condensation, catalyzed by azolium salts in basic media. This reaction was discovered independently as an unexpecteds ide reaction of decarboxylases in in vitro experiments, [9] and in organic chemical research of thiazolium salts. [10] Subsequently,t he biological relevance of this process in the carbohydrate metabolism was discovered. In the underlying biochemical reactions an H 4 C 2 O 2 unit is transferred between sugar molecules,e nabling the transformation of various carbohydrates into each other. [11,12] Due to the importance and simplicity of this reaction, it becameaworkhorse for later studies, throughw hich the mechanism of NHC organocatalysis was investigated. Breslow observedt hat in deuterated solvents the position2protono ft hiazolium salts is spontaneously exchanged to ad euteron,i nferring the mobility of this proton. [13] Assumingt hat this mobility meantt hat the deprotonated species, an NHC, was actually present in the solution in as mall quantity as an intermediate, he adjusted the mechanism of cyanide-catalyzed benzoinc ondensation [14] to interpret the catalytic effect of thiazolium salts. [15] Through suggesting the formation of another key intermediate (called later Breslow intermediate), he defined am echanistic picture thati st oday still the main school of thought in NHC organocatalysis ( Figure 1).
The use of this mechanism as at emplate hasa llowed explaining and predicting selectivity in many relatedp rocesses and understanding the role of structural features in the catalysts, [16,17] which strongly supports the aforementioned reaction path. Computational studies predict low barriers for the reaction of free carbenes with variouss ubstrates. [17] Furthermore, Breslow intermediates [18][19][20] and related compounds [19,21,22] have been synthesized and structurally characterized. Similarly,s everal free NHCs have been shown to be stable, [23][24][25][26][27] and extraor-dinary persistence has been evidencedf or several of their derivatives. [23,27] In some specific cases, the formation of NHCs has also been observed by mass spectrometry in vaporized reaction mixtures, [28] vaporized catalytically active [29] solutions, [30] or confined within the active site of at hiamine-dependent enzyme. [31] However,i ti si mportant to stress that the direct evidence that free carbenes can catalyzet hese reactions does not mean they playa na ctual role in catalysis, and similarly,t he stability of NHCs prepared in an isolated environment does not prove their formation in the reaction mixture. Throughout the decades doubt has been raised regarding the first part of the catalytic cycle, [32][33][34][35][36][37][38][39][40] in which the bond between the azolium ring of the catalyst and the substrate is formed. In fact, already in this early step selectivity can become an issue, if multiple substrates are availablei nt he reaction mixture, such as in crossbenzoinc ondensations, in Stetter reactions, and in the biochemicalt ransketolase reaction. Thus, understanding this initial reaction between substrate and catalyst is of high practical relevance. [4][5][6][7][8] Washabaugh and Jencks measured the acidity of thiaminea ti ts active site in aqueous solutiont ob e pK a = 18.0. [33] They argued that the catalytic activity of this compound could be explained only with ap K a < 14. [33] To resolve this contradiction, they tentatively suggested that an alternative, concertedp athway might bypasst he formation of the energetically quite unstable free NHC, in which the deprotonation occurs simultaneously with the catalystÀsubstrate bond formation. [33] However,s peculating that the substrate would need to approach the ring in the very direction, in which the proton shouldl eave, [33] they dismissed this idea. Finally,t hey suggested that the reason for the activity of this compound within the enzyme must be due to the stabilizing effect of the protein on the carbene, which shifts the local acid-base equilibrium within the binding site. [33] There are indeed indications that the NHC might be more stable within the active site of the enzyme, [31] including the enhanced activity of thiaminei nt he presence of these proteins. [41][42][43] However,t he effect of the enzyme cannot explain how NHCo rganocatalytic reactions can be possible in organic syntheses in the absence of any biomolecule. Imidazolium (in water:p K a = 19-24, [44][45][46][47][48] in DMSO:p K a = 19-24, [45,49] in MeCN: pK a = 33.6 [48] ), triazolium (in water:p K a = 14.9-17.4 [50] )a nd thiazolium (in water:p K a = 17-19, [33,50] in DMSO:p K a = 14.5, [48] in MeCN:p K a = 25.6 [48] )s alts have all been applied as catalysts, [4][5][6][7][8]51] and despite their high pK a values, triethylamine (in water:p K a = 10.65, [52,53] in DMSO:p Ka = 9.0, [53] in THF:p K a = 12.5, [53] in MeCN:p K a = 12.5 [53] )h as been observed to be basic enough to deprotonate them in aq uantity,w hich is sufficient to exhibit reasonable to excellent catalytic activity. [51] Although the aforementioned contradictions regarding the most fundamental acid-base theorys hould already be enough to raise questions, the general wisdom regarding the stability of NHCs points to furtheri nteresting issues. The hitherto synthesized stable free NHCs generally possess bulky substituents on the ring to prevent decomposition reactions, for example, dimerization, with the exception of some imidazol-2-ylidene derivatives. [54] Carbene catalysts,onthe other hand, showedremarkable activity with the smallest substituents without any significant decomposition, and larger functionalg roups are introduced merely to increaset he stereoselectivity of the reaction. [4][5][6][7][8] This is especially surprising regarding thiazolium salts. Thiazol-2-ylideness howedahigh propensity to dimerize in the presence of acid traces, [26] even with the sizable1 ,3-diisopropylphenyl substituent on the nitrogen atoms. Strikingly,i nt he absence of acids, the same compounds did not dimerize. Arduengo andc o-workersa rgued that the dimerization takes place through the reaction of an NHC with at hiazolium cation ( Figure 2). These findings can be rationalizeda st he protonation increases the electrophilicity of the C2-carbon atom of the ring, facilitating thereby the nucleophilic attack of the NHC at the ring, which-after deprotonation-forms the NHCd imer. Should NHCs be present in the solution in the catalytic environment, the excesso ft he azolium cations shouldr esult Dr.H ollóczki obtained his PhD in Chemistry at the Budapest Universityo fT echnology and Economics. In 2012, he received the Humboldt Fellowshipf or PostdoctoralR esearchers, whichf inanced his stay at the Leipzig University and at the Universityo fB onn. Currently, he is doing his habilitation as aj unior group leader at the University of Bonn. In 2018, he was awardedb yt he ADUC Prize for his researcha ccomplishments. The researchi nterests of Dr.H ollóczki spanf rom catalysis through sustainable chemistry to the environmental effects of plastic wastes. Figure 1. Catalyticcycle of the benzoin condensation, as suggested by Breslow in 1958. [15] Figure 2. Dimerization reaction of thiazol-2-ylidenes in the presence of acid, reported by Arduengo and co-workers. [26] in the dimers of the catalyst in an analogous process. Such decomposition reactions of thiazolium catalysts have not been reported.I nfact, the role of these dimers in the catalytic mechanism has been discussed, [34][35][36][37][38][39] and it was subsequently dismissed. [55,56] For the carbene-like reactions of ionic liquidswith carbon dioxide, we discovered by static quantum chemical calculations an ovel reactionm echanism, which avoidsN HC formation, [57] and the imidazolium cation reacts directly with the substrate. This mechanism was later observed also through ab initio molecular dynamics simulations. [58] Similarly,w ef ound an alternative pathway for NHC organocatalytic reactions, [40] in which first ap re-aggregation of the components occurs. The proton transfer from the azolium cation to the base and the CÀC bond formation between the catalyst and the substrate take place in as ingle elementarys tep within that aggregate, avoiding thereby the formation of free NHC intermediates, analogously to the suggestion of Washabaugh and Jencks. [33] We found that the barrier of the classical (dissociative) pathway that assumes the formation of the free NHC is by 20-30 kcal mol À1 highert han that of the concerted (associative) path. [40] This mechanism can satisfyingly explain the contradictions above regardingN HC catalyst stabilitya nd acid-base equilibria, whereas suggesting that the base-being present at the catalystÀsubstrate bond formation step-might also be as ite, through which selectivity can be enhanced.
Rico del Cerro et al. showed in aj oint experimental-computationals tudy that the H/D exchange of azolium saltsf ollows a very similara ssociative mechanism, in which the formation of free carbenesi sa voided. [59] Nolan and co-workersr eported that the NHC-metal complexes from imidazolium salts and metal ions can form through an analogousp ath, [60] in which the preparation or even the in situ generation of an NHC is unnecessary, allowing very simple synthetic routes to thesep ractically highly important compounds. [60,61] Regardingt he carbene-like reactions of imidazolium acetate ionic liquidsw ith glucose, [62] we found indications that the concertedm echanism should be prevalent, as the aggregates that are required for the associativem echanisms pontaneously occurred in ab initio molecular dynamics simulations, [63] whereas the barriers of this mechanism were somewhat lower than that of the reaction through carbene intermediates. [64] However, we found that in such an ionic liquid environment, and with the acetate anion as base the differenceb etween these two mechanisms in barriers is lower, [64] which implies that there might be experimental setups, in which the dissociative mechanism is more facile. Thus, to fully uncover the mechanistic details of NHC organocatalysis, furtherresearch is necessary.
Althoughc onsidering this novel reaction mechanismh as already given ad eeper insighti nto the chemistry of azolium salts, [57, 59-61, 63, 64] and resulted in actual applications, [60,61] exploiting this process in its full potential-byf or example, introducing unprecedented ways to control selectivities-has not yet been achieved. To this end, hereby we aim to understand in detail the influential factors that facilitate or suppress this direct reaction mechanism, while assessing its relevance in catalytic systems. In this comprehensive workw ed iscussi nd etail the effect of possible azolium rings, substituents, solvents, bases,s ubstrates, and counterions on this reaction.

Computational Methods
Static quantum chemical calculations:A ll quantum chemical calculations were performed by the ORCA 4.0 program. [65,66] Geometry optimizations of the minima and transition states were undertaken by using the TPSSh functional [67] with the D3-BJ dispersion correction, [68,69] and the def2-TZVPP basis set. [70,71] For the SCF cycle and the geometry optimization the tight convergence criteria were applied. The nature of the obtained stationary points was verified by making sure that the Hessian had no negative eigenvalues for minima, and asingle one for transition states. Steepest descent optimizations were performed in both directions defined by the imaginary normal mode for each transition state to identify the minima the given transition state connects. On the structures obtained by the DFT calculations, DLPNO-CCSD(T) single-point energies [72][73][74] were calculated with the def2-TZVPP and def2-QZVPP basis sets [70,71] with tight settings for the localization. The obtained single point energies were extrapolated to the complete basis set limit. [75] These electronic energies were then corrected to enthalpies and Gibbs free energies by using the thermochemical data obtained from the corresponding DFT frequency calculations. Gibbs free energies of solvation were calculated through the COSMO-RS approach, [76,77] by using the BP-TZVPD-FINE method of the COSMOthrmX14 software [78,79] based on BP86/def2-TZVPD calculations [70,71,80] by the Turbomole program. [81] Non-covalent interaction analysis [82][83][84] was performed by the Multiwfn 3.6 program package. [85] Classical molecular dynamics simulations:F or performing classical molecular dynamics simulations, the LAMMPS program was used. [86] For modeling the 1,3-dimethylimidazolium cation, the chloride and triflate anions the force field parameters by Canongia Lopes et al. were applied. [87,88] The organic solvents were described by the OPLS-AA model, [89] whereas for water the SPC/E parameters were chosen. [90] In the simulations, periodic boundary conditions were applied. The starting geometry of the cubic simulation box was created by the Packmol program. [91] The initial cell vector of the simulation box was chosen according to the density of the pure solvent in question. After ag eometry optimization, 1nss imulations in the NpT ensemble have been performed, by using NosØ-Hoover chain thermostats and barostats at at emperature of 293 K and under 1bar pressure. The timestep was chosen to be 1f s. The volume of the periodic simulation box was averaged over the last 0.5 ns, and this value was used for the later steps of the simulations. An external harmonic potential was added to the system that kept the distance between centers of mass of the anion and the cation at av alue of 3 .A fter 100 ps of equilibration, the force that was required to maintain the cation-anion distance was measured, and averaged over ap roduction run of 250 ps. The anioncation distance was then increased by an increment of 0.5 until 20 ,r epeating the equilibration and production runs at each distance. Integrating the forces versus the distance gave the free energy profile of ion pair separation.

Results and Discussion
In this paper,w ec omparet he two reaction mechanisms for the initial steps of NHC-catalyzed benzoin condensation, in which the catalystÀsubstrate bond is formed. The classical, dissociativer eaction mechanism ( Figure 3, red curve)f ollows the originalp roposal by Breslow. [15] First af ree NHC intermediate II is formed in the reactionm ixture from the azolium cation by a proton transfer to ab ase (e.g.,a na mine), whichr eacts with the substrate, giving primary complex III.C oncludingt he formation of the catalystÀsubstrate bond, III is protonated again by the ammonium cation, yielding IV.T his mechanism will be referred to as dissociative mechanism. The key to this path is the availability of the free NHC intermediate, which is present in the solution in al ow concentration, dissociated from the rest of the components.
In contrast, in the associativep ath [40] (Figure 3, green curve) the free NHC is not an intermediate of the reaction. Instead, the aggregation of the azolium catalyst, substrate, and base occurs, producing complex 2.T he bond formation between the catalyst and the substrate is undertaken through transition state TS 2!IV by ar earrangement within aggregate 2 in as ingle elementary step. This step comprises the C-to-N proton transfer from the azolium cation to the base, the CÀCb ond formation between the catalysta nd the substrate, and the N-to-O proton transfer from the conjugate acid of the base (e.g.,a mmonium ion) to the oxygen atom of the substrate to give IV in ac oncerted asynchronous manner.A na nimation fort he reaction mechanism is available in the Supporting Information. In this mechanism, the role of the base is to shuttle the proton from the carbon atom of the catalyst to the oxygen atom of the substrate, instead of releasing af ree NHC intermediate. Formed by either of the two mechanisms, IV will transform to yield the Breslow intermediate, which can thereafter reactw ith another substrate molecule in the subsequent reaction steps.
The reaction mechanisms were calculated first for the three azolium catalysts that have been applied in NHC organocatalysis (imidazolium (A), triazolium (B), and thiazolium( C), see Figure 4) substituted with methyl groupso nt he relevantn itrogen atoms. The substrates were chosen to be four different aldehydes that represent aw ide range of electronic effects on their reactingc arbonyl group (formaldehyde, acetaldehyde, acrolein, benzaldehyde). The base was chosen to be trimethylamine.
Ak ey question regarding the reaction is the aggregationo f the reactants. The formationo fahydrogen bond between the base and the cation to give I wasf ound to be thermodynamically favorable in all cases, as shown by the DH and DG values in Table 1. The association of as ubstrate molecule to I to give 2 has an enthalpy benefito f8 -11kcal mol À1 ,w hich can be explained by the polarity of complex I and the aldehyde ( Table 1). Thissubstantial DH value is comparable in magnitude to the TDS entropic contribution, and therefore the relative Gibbs free energies of I and 2 compared to the three dissociated components are very similar. Consequently,a ggregation at least to I should occur in all investigated cases. The low relative Gibbs free energy of 2 suggest that this aggregate is easily accessible in the solution, and in some cases,i ts hould be even the (slightly) dominants tructure in the reaction mixture.
The enthalpy and Gibbs free energy barriers (DH°d issoc , DG°d issoc )f or the dissociative mechanism were measured in all cases as the enthalpya nd Gibbs free energy differenceb etween I and TS II!III .F or the associative mechanism, the enthalpy barriers (DH°a ssoc )w ere defined to be the enthalpy difference between TS 2!IV and 2.T he reference point for calculating the Gibbs free energy barrier of the associated path (DG°a ssoc ) was defined separately for each cases,t aking the more stable  The associative mechanism has significantly lower activation enthalpies in all cases than those of the dissociative path. As discussed before,t he two sets of barriers exhibit ac ommon trend, hence ah ighera ctivation enthalpy in the dissociative mechanism usually meansahigherb arrierf or the associative path as well. [40] The difference betweent he enthalpy barriers is substantial, ranging between 19 and 29 kcal mol À1 ( Table 2). [40] In fact, the activation enthalpies for the dissociative mechanism are so high, that these processes seem rather unlikely to occur at all at reasonable temperatures (37-49 kcal mol À1 ), whereas the associative mechanism exhibitsm ild 11-25 kcal mol À1 barriers. In the associativem echanism,t he imidazolium catalysts howed generally higher barriers (20-25 kcal mol À1 ) than those of the triazolium and thiazolium derivatives (11-17 kcal mol À1 ), whichi si nq ualitative agreement with the lower mobility of the active protonf or the imidazolium derivatives. [44][45][46][47][48]50] Catalyzed by certain imidazolium derivatives, the condensation of formaldehyde in the presence of triethylamine was too slow to observe at room temperature in earlier experiments, [51] which is in good accordance with our data. The activation enthalpies were found to be only slightly different for the four substrates, but in most cases the highest values were found for benzaldehyde, whereas the lowest were found for acrolein. Includinge ntropye ffects by calculating Gibbs free energies increases all barriers. The increase is generally higherf or the associative mechanism (2-10 kcal mol À1 )t han for the dissociative path (0-5 kcal mol À1 ,T able 2), in agreement with the highero rderi nTS 2!IV .N onetheless, the entropic penalty for reaching TS 2!IV is relativelyl ow,s ince the association occurred at the earlier steps of the reaction, namely at the formation of 2 (see discussion above).
Changing trimethylamine to other deprotonating agents, such as 1,8-diazabicyclo [5.4.0]undec-7-ene (DBU)a nd 1,4-diazabicyclo[2.2.2]octane (DABCO) changes both reaction barriers, but the dominance of the associativem echanism is retained (see the Supporting Information). Substituting the ring with larger groups to arrive at frequently applied catalysts does not affect the conclusions ( Figure 5). Some of the compounds considered here possess bulky functional groups that control enantioselectivity in the later stages of the reaction. [6] Nonetheless, the barriers of the reactions with formaldehyde and acetaldehyde indicatef or all catalysts ad istinct preferencef or the associativemechanism.
The most important catalyst is, of course, the biologically active thiamine (Figure 4). The calculations discussed above regardingt he synthetically relevant azoliumr ings, bases, and aldehydes, can be and have been [15] considered as model reactions for the biochemical reactions of thiamine. To test the feasibility of the associative mechanism on this biomolecule, we calculated the reactions of thiamine with pyruvic acid, and with glyceraldehyde, in which the bond between the catalyst and as ubstrate is formed. For these reactions it is not possible to define ad issociative mechanism, because the base andt he azolium ring are covalently attached to each other within thiamine. We have, however,s uccessfully located the transition Table 1. DLPNO-CCSD(T)/CBS//TPSSh-D3BJ/def2-TZVPP Gibbs free relative energies and relative enthalpies for I and 2 with respect to the separated azolium cation, trimethylamine, and substrate in the gas phase.  state of the associative mechanism, bringing an ovel insight into the correspondingb iologically relevant reactions ( Figure 6). The activation enthalpies were found to be reasonably low (24.4 and 15.1 kcal mol À1 for pyruvic acid and glyceraldehyde, respectively), which-in the presence of the enzyme-could be most likely decreased even furthert hrough stabilizing interactions within the transition state. Considering that the deprotonating agent and the catalysta re within the same molecule, entropyh as am ilder effect on this reaction, as shown by the activation Gibbs free energies (25.4 and 17.7 kcal mol À1 ,respectively).
Having seen the clear preference of the reaction to follow the associativep ath in the cases detailed above,t he question is apparent:w hat effects make this mechanism so much faster than the dissociative path defined by NHC intermediate formation?O ne of the effects that make the associative mechanism favorable is the aforementioned high association enthalpy of the components to form I and 2,w hich compensates for the entropyl oss in the association of three molecules into as ingle cluster.A nother reason mustb et hat the associative path avoids the complete liberation of ah ighly reactive carbene lone pair,a nd insteadt his nucleophilic and basic centeri s simply flippedf rom the proton to the electrophilicc arbonyl carbon atom of the aldehyde. Non-covalent interactiona nalysis [82][83][84] (see the Supporting Information) reveals that the interaction between the carbonyl carbon atom andt he carbene carbon atom is in general considerable, strongly supporting this hypothesis.
Next to these clear advantages,t here is another interaction in TS 2!IV ,w hichd eserves speciala ttention.I th as been shown that the C-to-N and N-to-O proton transfers and the CÀCb ond formation occur asynchronously within the single elementary step of the reaction. [40,64] The transition state is situated late along the reactionc oordinate, [64] where the C-to-N proton transfer has already been completed. The mobilized proton, while it is departing the azolium ring and approaching the nitrogen atom of the amine base shuttle, will form at ransient, but strong hydrogen bond donor species within this assembly, an ammonium cation. While the catalystÀsubstrate bond is being formed, the transient ammonium cation interacts strongly with the aldehyde oxygen atom through ah ydrogen bond, which is observable in all transition states TS 2!IV through various hydrogen bonding criteria (Table 3, and the Supporting Information).
Carbonyl compounds are known to be activated in their electrophilicity by hydrogen-bond donor molecules, which po-larize the C=Ob ond, and thereby decrease its strength. [92][93][94][95] Enzymatic reactions often take effect through such hydrogen bonding, which activate molecules in am anner to enhance their desired reactivity, [94,96] whereas the active site also offers a well-defined spatial arrangemento fh ydrogen-bonding sites, inducing thereby selectivity through template effects. Mimicking these biochemical reactions, hydrogen-bond-supported catalysis hasb een subject to intensive scientific attention through the last few decades, using forinstanceurea, thiourea, and guanidinium derivatives as catalysts. [92][93][94][95] Accordingly,i ti s reasonable to assumethat the presence of the aforementioned ammonium-substrate interplay in TS 2!IV reaches beyond a simple hydrogen bond, and it affects the reactivity of the aldehyde directly.C onsidering that the transition state is situated late within the concerted asynchronous elementary step that comprises the protont ransfers and the CÀCb ond formation, [64] where the protoni sa lready transferredt ot he amine, it can be expected that the barrier itself is largely determined by the CÀCb ond formation. [64] Therefore, the activation effect of the amine on the aldehyde may have at remendous effect on the barrieroft he reaction. Figure 6. Lewis structure of the associative transition state of thiamine with pyruvic acid as as ubstrate (left). Ball-and-stick representation of the same transition state with pyruvicacid (middle) and glyceraldehyde (right) as substrate. Table 3. Hydrogen bonding between the ammonium and aldehyde moieties in the key transition state of the associative reaction mechanism  To highlight the effect of this hydrogen bond on the reaction, the changes in the structure of the four selected aldehydes induced by presence of ammonium cations are shown in Ta ble 4. The activation should manifest in weaker C=O double bonds, which is ac lear sign of the polarization by movingt he electron density of the p bond toward the oxygen atom, rendering the molecule more electrophilic. All data indicates that the C=Ob ond of the substrate becomes weaker upon interacting witha na mmonium cation, implying-in agreementw ithl iterature [92][93][94][95] -thatt his interaction has an effect on the electrophilicity of the substrate, andt hus the reaction as well.
To assessh ow much this interaction facilitates the reaction, we compared the barriers for the reaction betweenafree carbene and af ree aldehyde in the absence and presence of at rimethylammonium cation in hydrogen bonding with the substrate (DDE°in Ta ble 3). The obtained data shows an approximately 12-19 kcal mol À1 decrease in the activation energiesb y the presence of the ammonium ion. This significant decrease in the barriers suggests that "umpolung" catalysis by NHCs, for instanceb enzoin condensation, also works partly as an inherent hydrogen bond-supported catalysis.
The nature of the interplay within the transition state can be perhaps best represented in the electrostatic potential maps of the carbene and the ammonium cation within the transition state TS 2!IV (Figure 7). The positively charged proton on the ammonium cation units and the negatively polarized carbene lone pair are situated in such am anner that they can together encompass the substrate in the possible mostf avorable arrangement, through creating as ort of template. Whereast he lone pair of the carbene interacts with the LUMO of the aldehyde at the partially positive carbonyl carbon atom, the ammonium cation can aid the nucleophilic attack through forming ah ydrogen bond with the negatively polarized oxygen atom of the substrate. This cooperative effect clearly shows that through modifying the template formed by the azoliumbase assembly,t he feasibility of the reactionm ay be influ-enced, creating further ways to improvet he selectivity of these reactions.
The effects that result in the lower barriers for the associative mechanism may also reveal the limitation of this path. Considering that hydrogen bondingw ith the aldehyde is of high importance,o ther speciesi nt he solutiont hat can affect the solvation and hydrogen bondings ituation of this moietymost importantly the solvent and the counterion of the azolium cation-might flip the balance between the two mechanisms toward the dissociative path.S olvents with hydrogen bondinga bility can take over the role of the ammonium cation in activating the substrate for the reaction. Furthermore, because intermediate III and transition state TS II!III are zwitterionic structures, the merep olarity of the solventm ay alter their stabilities, and thereby the corresponding barrier. Through ap rotont ransfer from the azoliumc ation to the amine, solvation energies might change significantly,b ecause such protic ammonium ions are generally stronger hydrogen bond donors than their azoliumc ounterparts, [97] shifting the relative free energies. Furthermore, the free NHC can be stabilized by hydrogen bonding, which also facilitates the deprotonation of the azolium cation, and hence stabilizes the dissociative path.
Thus, we investigateds olvent effects on the barriers in an array of organic solvents through the COSMO-RS approach. [76,77,79] This solventm odel can accurately account for the effects of the polarity of the solvent, and also hydrogen bondingi nteractions,m aking it ideal fort he present purposes. In agreement with the reasoning above, the polarity of the solvent makes ab ig difference ( Figure 8). Moving from the less polar solvents (toluene, hexane, THF) toward the highlyp olar ones (DMSO), the barriero ft he associative path slightly increases,w hereas that of the dissociative mechanism remarkably decreases.T his effect is even stronger for hydrogenb ond donors olvents (EtOH, MeOH, water). In aqueous solution the advantage of the associative mechanism is reduced to approximately 1kcal mol À1 .
Thus, the polar environmenta nd hydrogen bonding apparently decreases the barriero ft he dissociative path through stabilizing TS II!III .H owever,h ydrogen bonding can also have a Figure 7. Ball-and-stick representation of the transitions tate for the associative reaction mechanism for the reaction between 1,3-dimethylimidazolium cation, trimethylamine, and formaldehyde (left, C: orange;O:red;N:blue; H: white). Electrostaticp otential map of the samestructure, after the formaldehyde was removed (right). The negatively polarized (red) lone pair,and the positively polarized (blue) ammonium cation together define ab inding site for the aldehyde,w hich can reactw ith the catalyst, and canbea ctivated by the protonated base to lower the barrier. tremendous effect on the availability of the free NHC. NHCs are highly basic compounds, and hence they can form very strong hydrogen bonds, which can result in as tabilization up to even 10-20 kcal mol À1 . [30,[98][99][100] Although the stabilization by hydrogen bondinga ctivates the aldehyde substrate, it may also occupyt he NHC lone pair,w hich is the very site, to which the electrophilic substrate should bind in the reaction. Hydrogen bonding at this positionh as been, therefore, repeatedly invoked as af actor that diminishes the catalytic activity of NHCs. [102,103] For this reason,a tf irst glance the hydrogen bond between the solventm olecule and the NHC may seem rather counterproductive. Thus, next to the dissociative and associative mechanisms that were distinguisheda bove,i nw hicht he carbene participates as af ree or as ap rotonated species( i.e., azolium cation), hydrogen bondedc arbenes should represent another degree of freedom fort he carbene's lone pair,s omewhere in between these two extremes.
Similarly to the associative mechanism detailed above, the hydrogen bond donors olventm olecule can be replaced by the substrate withoutf orming an unstable, free NHC in the solution, through the transition state shown in Figure 9. The obtained Gibbs free energies of activation(for the reaction of 1,3dimethylimidazol-2-ylidene with formaldehyde: DG°s olv = 7.4 kcal mol À1 ;w ith benzaldehyde: DG°s olv = 15.7 kcal mol À1 ; both in methanol) are quite low.Comparing this mechanism to the one that proceeds through the free carbene is not possible due to technical reasons:T he COSMO-RS solventm odel in all cases considers the bestp ossible solute-solvent interactions, thus, it automatically forms ah ydrogen bond between the NHC and the implicit solvent whenever the lone pair is available, hindering the calculation of af ully free carbene molecule (i.e. with an availablel one pair,n ot blockedb yp rotonation or hydrogenb onding). However,t he values shown above for the mechanism in Figure 9a re very similart ot he dissociation Gibbs free energy of ac arbene-alcohol bond, thus, that of producingafree NHC. Ah ypothetical free NHC, with no deactivating hydrogen bonds from the solvent, would need to react through yet another barrier to form the CÀCb ond with the substrate. Accordingly,i ti sr easonable to assumet hat the mechanism in Figure 9i sm ore feasible. Thus, in protic (hydrogen bonding) solvents the dissociativem echanismc an be rationalized as as ubstitution of the ammonium cation by the solventa tt he hydrogen bond acceptor hypovalent carbon atom,f ollowed by the reaction depicted in Figure 9. Such exchange of hydrogen bond donors at NHCs hasb een discussed before, and has been found to depend on sterice ffects and the hydrogen bond acceptors trength of the NHC in its rate and mechanism. [99,100] The reactiono ft he solvent-NHC hydrogen-bonded complex with the substrate (Figure 9) can be considered as as pecialc ase of the associativem echanism depicted in Figure 3.
The presence of the anion in the solutionn ext to the reacting speciesm ight also have significant effectso nt he differences in the barriers of the associative and dissociative reactions. The first question in this regardi si ft he azolium cations and the anionsr emaina ssociated in solution in the form of an ion pair,a nd can thereby affect the reaction, or if they instead dissociate into individual ions in the solution, and the reaction mechanism remains mostly unaffected by the anion.I on pairing can be influenced by the solvent, and by the nature of the ions themselves.
To uncover in what cases ion pairing might occur in these reactionm ixtures, we performed umbrella sampling calculations in ac lassical molecular dynamics environment on a1 ,3dimethylimidazolium bromidei on pair,a nd a1 ,3-dimethylimidazolium triflate ionp air in different solvents (see Computational Methods, and the Supporting Information). Whereas the chloridea nion is considered highly coordinative, the triflate anion is considered non-coordinative in their imidazolium salts, for instancei ni onic liquids. [104] In these calculations, the free energy was obtained asafunctiono ft he distance between the anion and cation, providing an estimate fort he propensity of the given anion-cation pair to stay associated. The separation of the ions is in almost all cases endothermic,o nly in the most dissociating aqueous solution is the energy of dissociation close to zero. Based on these findings, imidazolium salts are apparently dissolved in all the investigated solvents as ion pairs. In the lack of such well-tested force field parameters for thiazolium and triazolium rings it is not possible to have such aq uantitative insight fort he corresponding salts. However, considering that those two cations are more acidic than imidazolium cations, it is reasonable to assume that they are stronger hydrogen bond donors, and therefore they would be even more strongly coordinated to the anions, and consequently Figure 8. Solventeffects on the Gibbs free energy barrierso ft he reaction between 1,3-dimethylimidazolium cation, trimethylamine, and acetaldehyde by using the COSMO-RS model and the DLPNO-CCSD(T)/CBS//TPSSh-D3BJ/ def2-TZVPP method. [67-74, 76, 77, 80] For numerical values,see the SupportingInformation. Figure 9. Associative reaction of an NHCdirectly from its hydrogen-bonded form without the formation of afree carbenes pecies. the formation of ion pairs can be expected in these solutions as well.
Thus, the presenceo ft he anions must be considered for the reactions, if the full picture is to be obtained regarding the two competing mechanisms. To this end, we chose as eries of anions (halides, tetrafluoroborate, and triflate) that are often appliedi ns ynthesis as counterions for the catalysts, and examined the Gibbs free energy barriers of the associative and dissociativep aths in as eries of solvents. Based on the obtained data, the difference between the barriers becamel ower, amounting to less than 11 kcal mol À1 in all cases (Tables 5a nd  6). In fact, the preference varies between the two mechanisms, depending on the aniona nd on the solvent. In general, as ob-served above,t he increasing polaritya nd hydrogen bonding ability of the solvent facilitates the dissociative mechanism, and hinders the associativeo ne. The increasing size of the halide anions shifts the preference toward the dissociative mechanism, although fort he reactions with the thiazolium cation the trend between bromide and iodide is reverse. The two other anions, tetrafluoroboratea nd triflate, apparently make the associativemechanism more dominant than bromide or iodide, exhibiting similard ifferences in the two barriers to those for chloride. Imidazolium-baseds alts appear to be the least favorable fort he associativep ath to occur,w hereas the very often appliedt riazolium catalyst shows the highest propensitytos upport this mechanism.
Underc loser scrutiny, the role of the anion in decreasing the difference between the barriers can be identified. The positive charge on the azolium cationsi sd elocalizedo ver the whole ring, whereas on the ammonium cation it is highly localized on the NÀHu nit. Therefore, although azolium cations have also been shownt ob eh ydrogen-bond donors, [63,104,105] ammonium cations offer an even stronger donor site. These differences are clearly observable in the electrostatic potential maps of the azolium and ammonium ions, as well as the charges of the hydrogen bond donor sites ( Figure 10). Thus, the strong hydrogen bond, formed between the ammonium ion and the anion, can compensate for the energy demand of moving the proton from the stronger base carbene to the weaker amine. This compensation effect can be estimated through the metathesis reaction[ azolium + X À ]+ +HNMe + !azolium + + +[HNMe + X À ], showingthe Gibbs free energy benefitofexchanging the azolium cation to an ammonium in interaction with anion X ( Table 7). All anionsi nteracta pparently stronger with the ammonium cation than with any of the azolium species. Generally,t he imidazolium cation A shows the lowest reactionG ibbs free energies, followed by the thiazolium cation C,whereas the least exergonic cation exchange was observed for the triazolium cation B.T his trend is in full accordance with the discus- Table 5. DLPNO-CCSD(T)/CBS//TPSSh-D3BJ/def2-TZVPP difference between the Gibbs free energy barriers of the catalyst + formaldehyde + triethylaminer eaction for the associative and dissociative mechanisms (DDG°= DG°a ssoc ÀDG°d issoc ,i nk cal mol À1 )i nn on-polar media, in the presence of various anions. Negative values mean that the associative mechanism is faster in that particularc ase.  Table 6. DLPNO-CCSD(T)/CBS//TPSSh-D3BJ/def2-TZVPP difference between the Gibbs free energy barriers of the catalyst + formaldehyde + triethylaminer eaction for the associative and dissociative mechanisms (DDG°= DG°a ssoc ÀDG°d issoc ,i nk cal mol À1 )i np olar solvents, in the presence of various anions. Negativev aluesm ean that the associative mechanism is fasteri nt hat particularc ase.  sion above,s howing that the preference for the dissociative mechanism decreases in exactly this order.T he Gibbs free energy is most negative in case of the halide anions, and movingt oward the larger speciesw ith ad elocalized charge the metathesis becomes less exergonic.I nterestingly,t he halides show an opposite trend, as the chloride-despite the significantly strongeri nteractions with the ammonium cation than with the azolium cation-decreasest he least the propensity of the catalysts to undergo the associative mechanism (Tables5and 6). The explanation for this discrepancy might lie in the small size of the chloridea nion,w hich allows some inter-action through hydrogen bondingb etween the anion and the ammoniumm oiety also in transition state TS 2!IV of the associativem echanism. Through this interplay, the system receives somes tabilizationa lso in this mechanism,a lbeit significantly less than in TS III!IV of the dissociative path,w here the HNMe + Cl À ion pair is separated from the reacting catalystsubstrate pair. The effects above result in the conclusion that controlling hydrogen bonding and polarity effects of the solvent and the anion lead to ac ontrol over the reaction mechanism of N-heterocyclic carbene organocatalysis. Thel ess polar solvents, and smaller halides or weakly hydrogen bonding anionsl ead to a preference in the associativem echanism, whereas the use of polar solvents and larger halidesf acilitates the dissociative mechanism. The numbers in Tables 5a nd 6s how that switching between the mechanismsi sp ossible. It is, however,v ery important to emphasize here the difficulties regarding the estimation of entropies in quantum chemical calculations, and it has been repeatedly discussed that entropye ffects are significantly overestimated, [106][107][108] also for related reactions, [64] which results in an overestimation of the barriers for the associative reactionm echanism. Thus, although the trendsf or the shift in preference between the two mechanismss hould be valid, it is possible that the actual numbers in Ta bles 5a nd 6s hould be somewhat more negative, and the associative mechanism slightly more preferred.The data presented here showthat the associativem echanism,w hich avoids the formation of actual carbenes in the solution, has mild barriers, and should be considered in all cases when investigating the mechanism of NHC organocatalysis, especially in case of higherc oncentrationso f the substrate, base and catalyst in the solution.

Summary and Conclusion
In the present computational study, two mechanismsw ere compared for the formation of the catalystÀsubstrate bond in NHC organocatalysis. These two mechanisms fundamentally differ in terms of the occurrence of free NHCs in the reaction mixture. In the widelya ccepted (dissociative) mechanism of this process, as described by Breslow, [15] the formation of free NHC intermediates via the deprotonation of the azolium salt catalysti sr equired, and the resulting low concentration of carbenes in the solutiono ffers the actual catalytic activity.I na n alternative, associative mechanism, [40] the deprotonation of the very weakly acidic azolium salts is bypassed in ac oncerted process, in which the proton transfer and the catalystÀsubstrate bond formationo ccurs in as inglee lementary step. Depending on the mechanism,t he selectivity of the reaction might be influenced, especially because in the transition state of the associative path the base is also present, and therefore its bulkiness and possible template effects can influence the outcomeo ft he reaction as well. In as eries of model reactions, imidazolium, triazolium, and thiazolium cationsw ere considered as catalysts with varying substituents on the relevant nitrogena toms, formaldehyde, acetaldehyde, benzaldehyde, and acrolein as substrates, and trimethylamine, DBU, and DABCO as bases.
In the absence of solvents and anions the reaction follows the associative mechanism, avoiding, thereby,t he formation of free carbenes in the solution. The reasons for this dominance were identified to be the high enthalpic benefit of the association of the components, and the interaction of the mobile protono no ne hand with the azoliumr ing, avoiding the complete liberation of the reactive carbene lone pair,w hereas on the other hand with the oxygen atom of the substrate, which activates this molecule for the nucleophilic attack of the catalyst. This complex network of stabilizing hydrogen-bonding interactions can be, however,d isrupted andp artly substituted in the presence of polar and protic solvents, which offer alternative, modes of stabilization in case of the dissociative mechanism.
Imidazolium salts in av ariety of organic solvents and water were found to stay associated within the same solvents hell in the form of ion pairs. The presence of anions vanishes most of the dominance of the associative mechanism, and the barriers become more similar.T he underlying reason for this effect was identified to be the stronger interaction of the anion with the protonated base than with the azolium cation,w hichs hifts the acid-base equilibrium toward the free carbene, and partially breaks up the azolium-base-substrate aggregates.I nc ase of stronger hydrogen-bond acceptora nions( e.g.,h alides), the dissociative process exhibits lower barriers, whereas in case of weak hydrogen-bond acceptora nions( e.g.,t etrafluoroborate and triflate), the associative path was found to be faster.Anexception here wast he chloride anion, which-due to its small size-can form stabilizing interactions with the hydrogen-bond donor speciesa lso within the transition state of the associative path, retaining therebyt he preference fort his mechanism in many cases.A ccordingly,t he mechanism can be controlled Table 7. DLPNO-CCSD(T)/CBS//TPSSh-D3BJ/def2-TZVPP Gibbs free energies (in kcalmol À1 )o ft he metathesis reaction[ azolium + X À ]+ +HNMe + 3 ! azolium + + +[HNMe + 3 X À ], showing the energyb enefit of exchangingthe azolium cation to an trimethylammonium cation in interactionw ith anion X.

Catalyst
Cl through varyingt he anion and the solvent, and therefore introducing novel kinds of selectivities into these mechanism should also focus on carefullyc hoosing thesee lements of the catalytic system.
The results above bring also new insight into the related biochemical reactions of vitamin B1, thiamine. Considering that in this reactiont he thiazolium ring and the base that should deprotonate it are covalentlyb ound, only the associative reaction mechanism is feasible. In the lack of significant entropic effects, the barriers of two model reactions of thiamine were found to be low,s uggesting the viability of the associative mechanism in biological systems.