The mechanism of N-heterocyclic carbene organocatalysis through a magnifying glass.

Abstract The term “N‐Heterocyclic carbene organocatalysis” is often invoked in organic synthesis for reactions that are catalyzed by different azolium salts in the presence of bases. Although the mechanism of these reactions is considered today evident, a closer look into the details that have been collected throughout the last century reveals that there are many open questions and even contradictions in the field. Emerging new theoretical and experimental results offer solutions to these problems, because they show that through considering alternative reaction mechanisms a more consistent picture on the catalytic process can be obtained. These novel perspectives will be able to extend the scope of the reactions that we call today N‐heterocyclic carbene organocatalysis.


Introduction
N-Heterocyclic carbenes (NHCs) define ah ighly versatile field of chemistry.Significant portion of this knowledge and the corresponding applications grew out of as eries of experiments at the end of the 19 th century, in which Eijkman observed that rice husk prevents beriberi-like symptomso fm alnutritioned hens. [1] In the following decades the compound responsible for this effect, thiamine( vitamin B1, Figure 1), was isolated [2] and its structure was determined, which led to an extensive research on the role of this compound in the human body,a nd on the mechanism, in which this role is fulfilled.T he underlying questionsc ould be answered only through the intensive and interdisciplinary collaboration of biologists and biochemists with organic chemists, who characterizedt he related enzymes and their function that gave ideas for synthetic applications, and designed model reactions that allowed explaining and even predicting reactions in living organisms.
The first biochemical reactiona ssociated with thiamine was the decarboxylation of pyruvate, [3] yieldinga cetaldehyde.H owever,G reen et al. found that carboxylase enzymes from pig hearts did not only decarboxylate pyruvate, but also coupled the product acetaldehyde in vitro to acetoin, which was the first reported benzoin condensation reaction withoutc yanides. [4] Independently,U kai dissolved thiazolium salts in ethanol, and reacted them with benzaldehyde, and he found that-ina greement with Green-benzoin was formed. [5] A decadel ater,H orecker [6,8] and Racker [7] simultaneouslyd iscovered ab iochemical reaction of the thiamine-dependent enzyme transketolase, in which thiamine catalyzes the transfer of at wo-carbon-atom carbohydrate unit between sugars in a reactiont hat is chemically analogous to the reactions of Green et al. [4] and Ukai. [5] Duet ot he mutual biochemical and synthetic importance of these CÀCc oupling reactions, benzoinc ondensation catalyzed by thiamine and its analogues became the workhorse for later mechanistic investigations.
Stetter recognized the synthetic value of the reactions. By extending the scope of theses yntheses, he and then others laid down the fundaments of the fieldc alled today "NHC organocatalysis", [9] which offersaremarkable portfolioofhighly efficient syntheticm ethods. As ar esult of these studies, throughout the last century, NHCs and their reactions played ap art in the developmento fb iochemistry and medicine, synthetic chemistry,g eneral chemistry and electronic-structure theory. Althought hese fieldsw ere from the beginning highly intertwined, and built on each other in as ynergistic manner,i ti s important to remember that the initial motivation to go down on this path in science wast ou nderstand the biochemical reactions of thiamine and the analogous organocatalytic reactions of NHCs.
In ac enturyofr esearch, the mechanistic picture on these reactions hasb een continuously refineda nd extended, and many detailso ft hese processes have been revealed. Nonetheless, there has been am ultitude of data in literature that does not fit into the general wisdom regarding these reactions, which suggests that our knowledge on thesep rocesses is far from complete. Collecting these contradictions is necessary,i f am ore complete view on these reactions is to be built. To this end, in this criticalr eview the findings that prove or challenge the widely accepted mechanism of NHC-related organocatalytic reactions are collected, aiming not at giving af ull account on these many times reviewed reactions and their applications, [10][11][12][13][14][15] but rather at focusingo nt he still open conceptual mechanistic questions.

Initial Mechanistic Investigations
Given the multiple functionalities in the thiamine molecule, over the decades several proposals had been published for the mechanism of the benzoin condensation ( Figure 1). It was suggested [16,17] that the amino group of thiamine is responsible for the decarboxylation of pyruvate through aS chiff base (imine) formation and as ubsequentd ecarboxylation andh ydrolysis. Althoughm ultiple model reactions of amines were presented as proof, it was shown that the amino group of thiamine itself was ineffective as ac atalystu nder the same conditions. [18] The reactionw as also surmised to involvet he open-chain isomer of the thiazolium ring, [19] but no direct evidenceh as been presentedf or the open form being active. In the light of the similaritiesb etween alkylpyridinium [20,21] anda lkylthiazolium cations, the methylene bridge of thiamine was also surmised to react with carbonyl compounds. [ In his early report, Ukai showed that the benzoin condensation can be catalyzed by thiazolium compounds with av ariety of substituents on the nitrogen atom, indicating that the activity of thiazolium salts-and thereby thiamine as well-should be related to the thiazolium ring. [5] Through using isotopically labeled substrates, further evidence was presented that the thiazolium ring is responsible for the catalytic activity of thiamine. [22] Breslow recognized that the protona t2 -position of the thiazoliumr ing can be exchanged to ad euteron in deuterated methanol. [23] Thus, he argued that the active species that in fact catalyzes the reactions of thiamine is an NHC, formed by the deprotonation of the thiazolium ring. [24] Considering that the benzoin condensationi sc atalyzed by thiazolium salts, he hypothesized that the process responsible for the reaction should be similar to the one cyanide-catalyzed reaction that had been discovered more than ah undred years earlier by Liebig and Wçhler. [25] Thus,h ea djustedt he mechanism established by Lapworth [26] for the cyanide catalyst, and createdt he mechanistic picturet hat is 60 years later still the dominant school of thought for azolium catalyzed benzoin condensations, and wasu sed as at emplate for designinga na rray of analogous reactions that comprise the majority of the socalled NHC organocatalysis, and to explain their action. [10][11][12][13][14][15][27][28][29][30][31] In this mechanism, [24] the initial step is the deprotonationo f azolium salt I into an NHC II ( Figure 2). This nucleophilic NHC reacts with the electrophilics ubstrate (e.g. an aldehyde), and forms an initial (or primary) adduct III.A dduct III can isomerize into V through ap rotonation/deprotonation mechanism.T his structure-nowadaysc alled Breslow intermediate-is another key intermediate of the mechanism, because the fulvenic structure makesi ts exocyclic double bond polarized in am anner that the electron density shifted away from the ring. This excess of electrons at the exocyclic carbon atom turns this originally electrophilic carbonyl carbon atom of the substrate into an ucleophilic site. Similarly to the "umpolung"i nc ase of the cyanide-catalyzed benzoin condensation, this polarity change allows an electrophile (e.g. another substrate) to bind to this carbon atom, which makes this reaction valuable for synthesis. Even more importantly,a lthought he benzoin condensation with cyanide only aromatic substrates can be applied, azolium cations can catalyze analogousr eactions with aliphatic substrates as well, increasing the scope of the corresponding applications. After the formation of this new bond and ap roton transfer,t he NHC II and the product VIII can dissociate, closing the catalytic cycle.
In support of this mechanism,s table Breslow intermediates [37][38][39] and analogous structures [38,40,41] have been detected or synthesized. Recently,athiazolium salt was tailoredf or a tandem MS study,w hich enabled the observation of the actual free NHC intermediate from the evaporated solution, interpreted as ap roof for the occurrence of this speciesi nt he solution. [42] Many intermediates of the biochemical processes have been observed as well, being consistentw ith the model reactions of Breslow. [43] In the last decades also several theoretical studies have been published, which showed that through this mechanism numerouse xperimentally observed features of these reactions can be reproduced and explained. [44][45][46][47][48][49] This mechanism assumes the in situ formation of NHCs in the reactionm ixture. Duringt he 1990s, when the "renaissance of carbenes" [50] was at its high point, Te les et al. showedt hat not only thiazolium, but also imidazolium and triazolium compounds catalyzet heser eactions, [51] presumably with the same reactionmechanism. Accordingly,the community startedtoexchange the term "thiazolium catalysis" (used by Breslow [24,52] ) to "N-heterocyclic carbene organocatalysis", whichs eemed to be am ore general term. However,i nm ost of the studies that followed Breslow in furthere xploring or exploiting the mechanism-including theoretical calculations-the formation of the NHC was considered granted, but it was barely investigated explicitly.I nf act, as will be shown below,s everals tudies have been reported that contradictt his hypothesis, particularly regardingt he involvement of NHCs therein.

Basicity of N-Heterocyclic Carbenes
The key to the mechanism above is the acidity of the azolium ring, whicha llows the formation of the actual NHC catalyst. The earliest estimates for the acid strength of thiamine at this site gave pK a = 12.7, [53] and pK a = 17-21, [54][55][56][57] until Washabaugh and Jencks gave exact measurements of pK a = 18.0 for free thiamine in water. [58] They argued that thiamine must have ap K a 14, for the formation of the carbene intermediate that would render the carbene formation feasible in the reactions. For the puzzlinglower acidity of the compound they gave two alternative explanations.F irstly,i ti sp ossible that the enzyme somehow stabilizes the NHC, shiftingt he acidity of thiamine below the given threshold. This is supported indirectly by earlier data, which showed an acceleration of the catalytic activity of this vitamin by af actor of 10 4 . [54,57,59] The NHC intermediate was recently also observed within the enzyme, [60] also in line with this hypothesis. Secondly, they tentatively suggested that this stepwise mechanism that involves the NHC intermediate could be bypassed by an alternative, concerted mechanism, in which the protont ransfer and the thiamine-substrate bond formation occurs simultaneously. [58] However,t hey rendered this explanation unlikely due to surmised steric considerations. [58] Although the hypothesis that enzymes change the acidbase equilibrium of thiamine might indeed explain how the reaction can occur through the NHC isomer even with the pK a valuesa bove,i td oes not explain the observed high catalytic activity of azolium salts in enzyme-free organic synthesis. So Figure 2. Catalyticcycle of the thiazolium-catalyzed benzoin condensationasd efined by Breslow [24] used frequently as aparagonfor NHC organocatalysis. far three groups of azolium derivatives have been found to be active in organocatalysis:t hiazolium, triazolium, and imidazolium salts. [51] Various derivatives of these catalysts have been found active in the condensation of formaldehyde into different carbohydrates in DMF,w ith triethylamine as deprotonating agent. [51] The basicity of these NHC derivatives( i.e.,t he acidity of their conjugate acids) was in the focus of research in the last decades. [58,[61][62][63][64][65][66][67][68] The strong basicity of imidazol-2-ylidenes-pK a = 19-24, depending on the substituent and slightly on the solvent-earned them the title "superbase". Although thiazol-2-ylidenes (pK a = 17-19) [58,61] and triazol-5-ylidenes (pK a = 14.9-17.4) [61] are somewhat less basic, they are stillo verwhelmingly more basic than the amine bases that they are deprotonated with (e.g. pK a = 10.65 for trimethylamine [69] ). Recently,b enzoate derivatives have been also found sufficiently basic to allow NHC organocatalysis. [70,71] In fact, the presence of the benzoic acid derivativew as evidenced in the later steps of these reactions, allowing ad ual NHC-Brønsted acid catalysis. [70,71] Given that the protont ransfer from the azolium cation to the benzoate should occur only in as mall proportion, it seems likelyt hat the benzoic acid stays associated with the catalystt hroughout the following reaction steps. Thus, in other words, NHC catalysis can be performedi nalocally acidic environment. [70,71] Although acid-base theory is one of the most fundamental principles of chemistry,t hese contradictions have never been thoroughly discussed after the aforementioned considerations of Washabaugh and Jencks. [58] The high basicity of NHCs makes them also strong hydrogen bond acceptors, af eature that has been suggested first by Wanzlick, [72] and evidenced later by theoretical calculations [73][74][75][76][77][78][79] and experiments. [74,[80][81][82][83] Thisi sa lso in accordance with the observations that solventr earrangement-that is,t he exchange of ah ydrogen bond donor at the basic site of the NHC (Figure 3)-is the rate limitings tep of H/D exchange reactions of azolium cations. Dependingo nt he NHC and the hydrogen bond donor,t he dissociation energy of the hydrogen bonds can be up to even 20 kcal mol À1 , [75,84] which is by far stronger than the approximately 5kcal mol À1 value for aw ater-water hydrogen bond. [85] This prominents trengths hould be an obstacle for NHC organocatalysis, because the availability of the lone pair acceptor site of the hydrogen bond is also the cornerstoneo ft he catalyst-substrate bond formation.T hus, if the lone pair is occupied by ah ydrogen bond, it should be stabilized against and therefore blocked from undergoing reactions. Considering that the protont ransfer from the azolium cation to the base should lead to the formation of av ery strong hydrogen bond between the NHC and the protonated base, it is puzzlingh ow carbenes, which are presumably generated in such as mall quantity due to their basicity,a nd then inactivated by the remarkably strong hydrogen bonding, can exhibit any kind of measurable catalytic activity.

Stable Carbenes
The involvement of NHCs in the organocatalytic reactions of azoliums alts was supported by the synthesis of free NHCs. Already in the early 1960s, Wanzlick reported that bis(1,3-diphenylimidazolidin-2-ylidene) dissociatesi nto monomers in an (NHC) 2 $2NHC equilibrium, [86,87] and exhibits the chemistry of free NHCs (Figure 4). He also provedt hat diaminocarbenes and thiazol-2-ylidenes are nucleophiles andh ence they can react   [86] and an alternativem echanism established by Lemal( below). [92] with carbonyl compounds, [72] seemingly confirming the mechanism established [24] by Breslow.T hese findings were strongly corroborated by the synthesis of the first stable NHC 1,3-diadamantyl-imidazol-2-ylidene 1, [88] and later the others 1,3,4-triphenyl-1,2,4-triazol-5-ylidene [89] 7 and 3-(2,6-diisopropylphenyl)-4,5-dimethylthiazol-2-ylidene [90] 6 ( Figure 5). NHC 1 exhibited extraordinary stability under inert atmosphere even at its meltingp oint 240-241 8C, [88] whereas the 4,5-dichloro derivative 5 was even identified as "airstable". [91] However,u nder closer scrutiny these arguments are somewhat less convincing. The successful synthesis of free N-heterocyclic carbenes in an isolated environment, which is very different from the catalytic mixture, is in fact no directp roof that during the synthesis these species are actually generated. To avoid undesired reactions even under inert atmosphere,s table free NHCs are, except for somei midazol-2-ylidene derivatives (e.g., 3), decorated with bulky substituents. It was shown that the dimerization of thiazol-ylidenes occurs through the reaction of at hiazolium cation and the corresponding NHC, [93] and even the bulky 2,6-diisopropylphenyl substituents of 6 are not enough to fully prevent these side reactions [90] in the presence of acid traces. These findings raise the question how the in situ generated thiazol-2-ylidenes can avoid reacting with the thiazolium catalyst in the reactionm ixture of an organocatalytic setup.F urthermore, in the presence of air,almostall hitherto synthesized NHCs reactw ith moisture or oxygen to give various decomposition products, [74,94] even if the hydrolysis of imidazol-2-ylidenes with traces of water appears is sluggish ( Figure 6). [74,94] Despite all this data, most reactions that are called NHC organocatalysis are performed under air, [10] and often with azolium cations possessing significantly smaller substituents than those mentioned above, [10] andt he introduction of larger substituents into NHCs is merely aw ay to introduce stereoselectivity (Figure 7). [10,12] It is, of course,apossible explanation that the concentration or the lifetime of the free NHC is just low enought oa void these reactions, but high enough to exhibit the desired reactions with reasonable rates. However, to fulfill these two criteria at the same time would mean al ack of robustness for the reactions, and there should be only an arrow basicity range for the reaction media that enables catalytic activity withoutt he decomposition of the catalyst. Given that many different kinds of NHCs are employedi nc atalysis, each of them with aw ide spectrum of bases ands olvents, this argument seems unlikely. Followingt he principle of Occam's razor,asimpler explanation may exist for these contradictions, namely the existence of an alternative mechanism that does not necessitate the presence of free NHCs in the solution.

Alternative Reaction Mechanisms
Shortly after Wanzlick presented [72,86,87] the dissociation equilibrium of (NHC)2$2NHC, Lemals uggested that the NHC-like reactions occur directly from the dimer,w ithout the involvement of free NHCs (Figure 4). [92] Based on these findings,L ópez-Calahorra hypothesized that in the thiazolium catalyzed benzoin condensation this NHC dimer plays the central role as the actual active species. This is supported by the aforementioned propensity of thiazol-2-ylidenes to form dimers in the presence acid traces. [90,93] They established two possible reaction mechanisms, one with the dimer dissociating after reacting with the substrate, resulting in the Breslow intermediate and af ree thiazol-2-ylidene, from whichp oint the reaction could follow the mechanism of Breslow. [95] In the other mechanism,t he connection between the two thiazolium rings is retained throughout the whole reaction. [95,96] López-Calahorra reported that the yields obtainedi nb enzoin condensations by catalysts, in which two thiazolium rings were linked by À(CH 2 ) n À (n = 2-8) groups through their nitrogen atoms, is highly dependento n the length of the link. They interpreted this dependency as a direct proof for the mechanism involving the NHC dimer. [97] However,i sotope-labeling experiments corroborated the original mechanism by Breslow,a nd thereby the involvement of the dimer in the reactionw as questioned, [93] and the related enzyme structures also showedn op ossibility for the formation of thiamine dimers in biochemical reactions. [93,98] Breslow showedt hat the reaction kinetics was first-order in thiazolium salts, and the transition state of the rate-limiting step contains two benzaldehyde molecules and as ingle thiazolium. [99] López-Calahorrap resented kinetic data that he rationalized as second order in thiazolium salts, [100] contradicting the earlier measurements.B ofill presented ac omputational study, [101,102] finding that the reaction occurs through ab iradical mechanismw ith the NHCd imer.I nt urn Breslow re-analyzed the data of López-Calahorra, showing that it is in fact first order in the thiazolium catalyst, andt hat their interpretation was erroneous. [52] Thereafter,t his mechanism was not discussed any further.M oreover, later theoretical calculations show that many NHCs do not form dimers, [103] which renders organocatalytic reactions through these structures as ag eneral mechanism somewhat dubious.
Through moleculard ynamics simulations we have observed that the exchange of as olvent molecule, which is in hydrogen bond with the NHC, does not necessitate the formation of the free NHC in the solution. [77,79] Instead, the singlel one pair can accommodate as econd hydrogen bond donor,a llowing for an associativee xchange, [77,79] which can facilitate the hydrogen bond dynamics-thus the solvente xchange-ofN HCs (Figure 3). We assumedt hat if the capacity of NHC lone pairs to serve as multiple interaction sites is ag eneral feature, it may also allow ac oncerted reactionm echanism for the related organocatalytic reactions, as was suggested (andi mmediately rejected) earlier. [58] We could identify two mechanisms for the reaction between azolium cationsa nd aldehydes in the presence of trimethylamine base. [104] The first, dissociative or stepwise reactionm ech-anism follows the mechanism as suggested by Breslow (Figure 2), including the explicit formation of af ree NHC in solution. In the second, associative or concerted mechanism (Figure 8), the associationo fa ll components occurs first, forming an initial cluster. Withint his cluster, the catalyst-substrate bond can form within as inglee lementary step througha proton transfer from the cation to the amine base and as imul-taneousC ÀCb ond formation between the ring carbon atom at the active site of the catalyst, and the substrate, yieldingd irectly the protonated adduct of the dissociativemechanism. [104] Through this path, the reactionc an occur despite the large basicity differenceb etween the NHCs and the bases withoutt he formation of the free carbene, which is therefore not present in the solution, and cannots how any decomposition reactions depictedi nF igure 6. The activation energies, enthalpies and Gibbs free energies indicate an overwhelming dominance of approximately2 0-30 kcal mol À1 for the associative mechanisms for all nine combinations of the three azoliumc ations and three aldehydes that were investigated. Using continuum solvent models did not change this general conclusion, although polar media decreased the differences in barriers. [104] Interestingly,c hanging the base from amines to an acetate anion decreasest he advantage of the associative mechanism, [105] which has implications for the chemistry of ionic liquids( for an excellent review see Ref. [106]).
The barrieri nt he dissociative mechanism originates largely from the differencei nb asicity between the NHC and the amine. In the associativem echanism, the transfer of the proton plays ar ole, therefore the basicityo ft he NHC could have an effect on the barrier. Indeed, the barriers of the two paths showed ac ommon trend, [104] which also explains why this possibility has been overlooked previously.I no ther words, al ess acidic azolium cation should have as lower reaction through both paths, in agreement with the generalq ualitative trends in the experiments. [51] The first qualitative measurementso fs uch processes, which aimed at observing theset wo reactionm echanisms, werep erformed by Rico del Cerro et al. [107] They calculated the rate constantsf or the two paths of H/D exchange reactions of imidazolium salts through DFT calculations and found that the experimental values compare better to the associativem echanism, [107] confirming our computational results. These findings,a nd the availability of the as- Figure 8. Alternative reaction mechanism for the initial step of the reaction between azolium cations and aldehydes, through as ingle elementary step, without the formation of free carbenes (seeFigure2). [104] sociativer eaction mechanism provide ac onsistent picture on the reactionm echanism of the organocatalytic reactions catalyzed by azolium derivatives.
The practical importance of this seemingly subtle difference can be recognized, if one considers the role of dissociative and associativem echanisms of nucleophilic substitution reactions, S N 1a nd S N 2p rocesses, in chemistry.I ti sw ell known that the presenceo ft he leaving group in the rate-limiting step of S N 2 reactions has as tructure-directing effect, which may be exploited to improve stereoselectivities. Similarly,t he presence of the (protonated)b ase in IX (Figure 8) maya llow influence in the formation of the initial substrate-catalystb ond, which might be important in the presence of multiple substrates. Recent studies on structure-directing effects by the presence of the protonated base at the later steps of NHC organocatalysis seem to underscoret his hypothesis. [70,71] Regarding the later steps of the catalytic cycle also further details have been revealed. Alreadyi nt he early work of Bofill it was recognized [101] that radical and biradical pathways may play ar ole in organocatalytic reactions by azolium salts, even if the actualm echanism he suggested has been disproven. [52] It was also shown that Breslow intermediates can be oxidized, to generatearadical species, [108] af eature that has been shown to play ar ole in biochemical reactions, [109] and has been exploited in the last decadet op erform redox catalysis with NHCs. [31,110,111] Rehbein observedE PR signals in ab enzoinc ondensation setup, with the exclusion of oxygen. The radicalw as observed at the onset of the reaction, and it was evidenced that the formation of the species in question requires both the Breslow intermediate and the aldehyde. Kinetic isotope effects did not only confirmt his finding, but were also found similar to those observed for the overall reactione arlier, [112] which could be explained through the rate-limiting step of the radical formation and the overall reactionb eing identical. [113] These findings were consistent with as ingle-electron transfer from the Breslow intermediate to the aldehyde substrate to give XI as an intermediateb efore the CÀCb ond formation. In as ubsequent study,i tw as shown that the quantum chemically calculated and experimentally measured kineticsc ompare well, [114] which provides further proof for this alternative mechanism. Accordingly,t he mechanism established by Breslow can be extended by ar adicalp athway ( Figure 9). However,i nafollowup study Regniere tal. showede videncet hat the one-electron oxidation of the Breslow intermediates leads to as ubsequent deprotonation of these species, yielding acylium radicals, which are therefore more likely to be present in the reaction than XI. [115] The authors also point out that even if the formation of radicals has been observed in the benzoin reactions, it has not been yet proven that these radicals are actual genuine intermediates of these reactions. [115] These considerations make it clear that further studies are necessary to assess the importance of radicalformation in these processes.

Summary and Outlook
Since the discovery of thiamine, al ot of hypotheses have been published regarding the biochemical and organocatalytic activity of this vitamina nd its analoguea zolium cations. Many of these possible ideas have been proven wrongs ince then, defining an evolution of the mechanistic picture for thesep rocesses. Although nowadayst he overall mechanism of Breslow with the formation of NHCi ntermediatesi ns olution is accepted, and on some examples it has been directly proven by a multitude of studies, there are still results that point to other possible paths for these processes. The two main current directions in this regard are am echanism, which bypasses the formationo ff ree NHCs in the solution,a nd an electron-transfer process between the Breslow intermediate and aldehydes ubstrates, resulting in the formation of radicals. Both of these mechanisms need more research in terms of NHC rings, substituents, bases, counterions, and solvents, to evaluate under which conditions are they dominant over the classical catalytic cycle as defined by Breslow.T he possibility of avoiding carbenes in the so-called "NHC organocatalysis" raises the question, if the community should change the name describing these reactions to "azolium catalysis", which fits better to the term "thiazolium catalysis" used by Breslow even in 1996, [52] while also describing more accurately the actual catalytic processes.

Conflict of interest
The authordeclares no conflict of interest.