Insights into LiAlH4 Catalyzed Imine Hydrogenation

Abstract Commercial LiAlH4 can be used in catalytic quantities in the hydrogenation of imines to amines with H2. Combined experimental and theoretical investigations give deeper insight in the mechanism and identifies the most likely catalytic cycle. Activity is lost when Li in LiAlH4 is exchanged for Na or K. Exchanging Al for B or Ga also led to dramatically reduced activities. This indicates a heterobimetallic mechanism in which cooperation between Li and Al is crucial. Potential intermediates on the catalytic pathway have been isolated from reactions of MAlH4 (M=Li, Na, K) and different imines. Depending on the imine, double, triple or quadruple imine insertion has been observed. Prolonged reaction of LiAlH4 with PhC(H)=NtBu led to a side‐reaction and gave the double insertion product LiAlH2[N]2 ([N]=N(tBu)CH2Ph) which at higher temperature reacts further by ortho‐metallation of the Ph ring. A DFT study led to a number of conclusions. The most likely catalyst for hydrogenation of PhC(H)=NtBu with LiAlH4 is LiAlH2[N]2. Insertion of a third imine via a heterobimetallic transition state has a barrier of +23.2 kcal mol−1 (ΔH). The rate‐determining step is hydrogenolysis of LiAlH[N]3 with H2 with a barrier of +29.2 kcal mol−1. In agreement with experiment, replacing Li for Na (or K) and Al for B (or Ga) led to higher calculated barriers. Also, the AlH4 − anion showed very high barriers. Calculations support the experimentally observed effects of the imine substituents at C and N: the lowest barriers are calculated for imines with aryl‐substituents at C and alkyl‐substituents at N.


Introduction
Since its first synthesis LiAlH 4 has become one of the most commonlyu sed reducing agents. Saline lithium hydride (LiH) 1 is essentially unreactive towards double bonds of any kind due to its high lattice energy and low solubility in organic solvents. [1] Aluminium hydride (AlH 3 )i si nc ontrast highly reactive but even as it ether complex it decomposese asily in its elements. [2] Its combination LiAlH 4 ,h owever,i ss table and highly reactive and has since its discovery in 1947 been developed into av ery useful reducing agent. [3] This commerciallya vailable metal hydride source is well soluble in ethereal solvents and reacts readily with polar C=Ob ond in aldehydes, ketonesa nd carboxylic acids. [4] Nitriles react violently with LiAlH 4 and, under more forcing conditions, even reduction of the C=Nb ond in imines can be achieved. Despite the requirement of elevated temperatures, main group metal-mediated imine transformations are of prime industriali mportance. [5] Although these applications are based on stoichiometric use of LiAlH 4 ,t he last decades have seen some interesting examples of LiAlH 4 (or related compounds) in catalysis. [6][7][8][9][10][11][12][13] During the development of early main group metal catalyzed imine hydrogenation, [14] we found that commercially availableL iAlH 4 can be used under relatively mild conditions in catalytic instead of stoichiometric quantities (2.5 mol %c atalyst loading, 1bar H 2 and 85 8C). [13,15] Such an on-stoichiometric route prevents the generally hazardous aqueous work-up anda voids considerable amountso fL i/Al salts as side-products. Noteworthy is the fact that reactions could be carried out in neat imine,w ithouta dditional solvents. This eliminates the need for rigorously solvent dryinga nd makes the procedure highly atom economical and environmentally benign.
While LiAlH 4 performed well in imine hydrogenation catalysis, homologues such as NaAlH 4 and NaBH 4 were shown to be much less active. [13] This clearly shows that not only the nature of the hydride source (AlÀHo rB ÀH) but also the alkali metal (Li or Na) influence catalytic activity.M ostr ecently,w ei ntroduced group 2m etal alanates, Ae(AlH 4 ) 2 (Ae = alkaline earth), in imine hydrogenation catalysis. [16] Although Mg(AlH 4 ) 2 is less active than LiAlH 4 ,t he heaviera lanates Ca(AlH 4 ) 2 ,a nd especially Sr(AlH 4 ) 2 ,s howedh igh activities, considerably broadening the substrate scope. As the salt [nBu 4 N + ][AlH 4 À ]w as found to be essentially inactive, [16] the presence of the s-block metal is crucial.This strong indication for aheterobimetallic mechanism is supported by ac omprehensive study by the Mulvey group. [17] Comparison of the activities of neutrala nd anionic aluminium hydride compounds in hydroboration catalysis, clearlys uggestaheterobimetallic mechanism.
We proposed am echanism in which LiAlH 4 first reacts with two equivalents of imine to give am ixed hydride/amide complex LiAlH 2 [N] 2 which is the actual catalyst (Scheme 1, [N] = N(tBu)CH 2 Ph). This assumptioni sb ased on the fact that the second imine insertion is generally faster than the first while the third insertion is more difficult. [13] This does not only hold for LiAlH 4 /imine reactivity but also for LiAlH 4 /R 2 NH deprotonations and explainsn icely why there are many isolated examples of complexes like LiAlH 2 [N] 2 . [11,[18][19][20][21][22] Startingw ith LiAlH 2 [N] 2 as ac atalyst, the first step is insertion of at hird imine. We proposed this to be ah eterobimetallic process in which imine-Li coordination activated the C=Nb ond for nucleophilic attack by the AlÀHu nit. This is followed by amine formation in the reactiono fL iAlH[N] 3 with H 2 .S ince as light increasei nH 2 pressure has an accelerating effect, [13] this hydrogenation step was proposed to be the rate-determining most difficult step.
In this contribution, we report additional experimental proof for such ah eterobimetallic mechanism and support our observations with ac omprehensive computational study.

Metalm odifications
We recently presented that group 2m etal alanates, Ae(AlH 4 ) 2 , are considerably better catalysts with activitiesi ncreasing in the order Ae = Mg > Ca > Sr. [16] Herein, we explore further metal modifications of both, the s-a nd p-block metals.
While NaAlH 4 was found to be less activet han LiAlH 4 , [13] we becamei nterested in exploring KAlH 4 in imine hydrogenation. Analogue to Schlesinger's first synthesis of LiAlH 4 [3] and NaAlH 4 , [23] it was attempted to prepareK AlH 4 by reaction of four equivalents of KH with one equivalent of AlCl 3 .T his gave due to solubility problemsi ncomplete H/Cl exchange. More effectivew as the reactiono fK Hw ith AlH 3 ·(THF) 2 .T he product was only sparingly soluble in THF but dissolved sufficiently in imine PhC(H) = NtBu to obtain 1 Ha nd 27 Al NMR data (see Figure S1). These data confirm aT HF-free product containing the AlH 4 À ion;t he 27 Al NMR spectrum shows aq uintet with 1 J Al-H = 168 Hz. While neither pure KH nor AlH 3 (THF) 2 was able to catalyze imine reduction,t he catalystK AlH 4 led to product formation but only in trace quantities. Compared to the activities of LiAlH 4 and NaAlH 4 ( Table 1, entries 1-3), it is clear that catalyst performance decreases with the size of the alkali metal:L i > Na > K. Since KAlH 4 is notably insoluble, this may also be related to solubility.T he significantly better soluble complex KAlH 4 ·[(18-crown-6)/THF] was prepared accordingt ol iterature. [24] While its performance is farb etter than that of neat KAlH 4 (cf. entries3-4), it is less active than LiAlH 4 .T his implies that the s-block metal requires some Lewis acidity.  As we previously found that the borate NaBH 4 is essentially inactive in imine hydrogenation, [13] we becamei nterested in exchanging the p-block metal Al for Ga. Althought he synthesis of LiGaH 4 andN aGaH 4 has been described by Wiberg as early as 1951, [25] attempts to use these salts in catalysis gave already at room temperature decomposition in Ga and LiH (or NaH). The more stable TMEDAa dducts[ Li + (TMEDA) 2 ][GaH 4 À ] [Na + (TMEDA) 2 ][GaH 4 À ]c ould easily be prepared and isolated according to Bakum et al., [26] however,t hese were found to be insoluble in THF and gave after 24 hours at 85 8Co nly sub-stoichiometric imine-to-amine conversion (entries 5-6).

Isolation of intermediates
According to the proposedc atalytic cycle in Scheme 1, the first step is ad ouble addition of imine to LiAlH 4 to give LiAlH 2 [N] 2 ; [N] = N(tBu)CH 2 Ph. Indeed, reaction of LiAlH 4 with one equivalent of imine gave under catalytic conditions am ixture of un-reactedL iAlH 4 ,L iAlH 3 [N] and LiAlH 2 [N] 2 .T his implies that the second addition is faster than the first. While clean isolation of the single addition product is difficult to achieve, reaction of LiAlH 4 with two equivalents of imine cleanly gave LiAlH 2 [N] 2 , which we previously isolated in the form of the TMEDA and PMDTAc omplexes LiAlH 2 [N] 2 ·[TMEDA] 2 (I)a nd LiAlH 2 [N] 2 ·[PMDTA]( II). [13] This high preference for formation of the double addition product is generally observed. [11,[18][19][20][21][22] Addition of at hird imine needs much longerr eaction times and an excess of imine. We now introduce the isolation and characterization of LiAlH[N] 3 ·[THF] 4 (1)o btained after long term reaction of LiAlH 4 with an eightfold excesso fi mine at room temperature. Its crystal structure ( Figure 1, Table 2) shows as olventseparated-ion-pair.
We also isolatedi ntermediates along the catalytic pathway for alanatec atalysts with the heavier s-block metalsN aa nd K. Reactiono fN aAlH 4    size of Na + vs. Li + ,t here are additional contacts between Na + and the AlH 2 [N] 2 À ion. Reaction of KAlH 4 with two equivalents of PhC(H) = NtBu gave in the presence of 18-crown-6 the intermediate KAlH 2 [N] 2 ·[18-crown-6] (4)w hich despite extensive solvation of K + by the crown ether shows an intimate Al(m-H) 2 K contact( Figure 1, Ta ble 2);asimilar [(Me 3 Si) 2 N] 2 Al(m-H) 2 Li·(12crown-4) complex was isolated by the Mulvey and Hevia groups ( Figure 2). [11] Apart from intermediates with the standardi mine PhC(H)= NtBu, which hasb een the benchmark in catalyst screening, also intermediates with other imines have been isolated. As previously reported, catalytic reduction of tBuC(H)=NtBu is much more sluggish (Table 1, entry 7) [13] because the C=N bond in this bis-alkylated imineisn ot activated by conjugation with Ph. Under forced conditions, however,t he double addition product was formed and could be crystallized in the form of LiAlH 2 [N(tBu)CH 2 tBu] 2 ·(TMEDA) (5). Despite two equivalents of TMEDAh ave been used, ac ontact-ion-pair with only one TMEDA ligand and an intimateA l(m-H) 2 Li contact crystallized ( Figure 1, Table 2). This could be understood by the strongly electron-releasing character of the tBuCH 2 (tBu)N À ions that makes the AlH 2 unit much more hydridic. Sincet he C=Nb ond in PhC(H)=NPh is activated by conjugation over an extended p-system, addition to LiAlH 4 wasf ound to be extremelyf ast. In this case we have been able to isolate the fourfold addition product LiAl[N(Ph)CH 2 Ph] 4 ·(THF) 4 (6)w hich crystallized as as olvent-separated-ion-pair ( Figure 1, Table 2).

Possible pathways for catalystdeactivation
The thermald ecomposition of LiAlH 4 is well investigated, especially in the context of its potential in hydrogen storage. [27] Deactivation and thermald ecomposition of intermediates in the catalytic cycle for LiAlH 4 catalyzed imine hydrogenation,h owever,has so far not been investigated.
To this purpose, LiAlH 4 was reacted with an excess of PhC(H) = NtBu imine at 85 8Cf or multiple days. X-ray analysis of crystalso btained from this reaction mixture showed that reductiono ft he imine to amide has been followed by orthometallation. Complex 7 (Figure3)i sc omprised of an alanate anion with two C,N-chelating ligands and aL i + cation that bridges both Na toms and is additionally solvated by an imine. The latter neutral ligand is heavily disordered with reduced imine:P hCH 2 N(H)tBu.T his illustrates that deprotonation in the ortho-Ph positionp roceeds by an amide base. Alternatively, ortho-metallation takes place by the Al-H functionality and H 2 gas produced is responsible for amine formation by hydrogenolysis. The selective ortho-alumination with mixed Li/Al-amide  bases has been reported previouslyb yt he Mulvey and Hevia groups. [28,29] Recrystallization of 7 from aT HF/hexanem ixture gave complex 8,asolvent-separated-ion-pair ( Figure 3). Complexes 7 and 8 do not react with H 2 (1 bar,8 08C, 5d) to give AlH 4 À or other hydride species. This implies that the here observed ortho-metallation could be ac atalystd eactivation pathway. Complex 8 was isolated in crystalline form in yields up to 48 %. This combined imine insertion and subsequent ortho-metallation protocol may be exploitedi ns ynthesis.

DFT calculations
Given the complexityo ft he system with two different metals (Li and Al), two different anions (amide and hydride), the generally high dynamics of polar complexes and the importance of solvation, it is clear that anyc alculational study on LiAlH 4 catalyzed imine hydrogenation is challenging. Aim of this comprehensive study is to gain detailed information on possible reaction mechanismsa nd to answer following questions. What is the catalytically active species:L iAlH 4 ,L iAlH 3  Energy profilesf or the catalytic pathways have been calculated by DFT theory at the B3PW91/6-311+ ++ +G**//6-31+ +G** level. Since solvent effects in these polar reactions can be important,c orrections have been appliedu sing the polarizable continuumm odel (PCM) simulating THF (e = 7.4257). All energy profiles show DH values in kcal mol À1 .C omputational methods overestimate the entropic factors and therefore DG values for energy barriers of reactions with entropyl oss are calculated too high. [30] For completeness, energyp rofiles with DG values can be found in Supporting Information (Schemes S3-S7). Minima and transitions tates (marked by *) have been identified by frequency calculation.
In an explorative study on av ery simple model system, that is, the hydrogenation of MeC(H)=NMe with LiAlH 4 ,w efound that solvation of the Li + cation is highly important (see Scheme S1). This preliminary study demonstrated that modelling ap olar reaction medium only with PCM was not sufficient. Energy barriers dropped significantly when solvation effects were explicitly modelledw ith solvent molecules as well. Therefore, in as econd step we calculated the full energy profile for LiAlH 4 catalyzed hydrogenation of our benchmark substrate PhC(H)=NtBu, using this imine also fors olvation( Scheme 2). Starting with LiAlH 4 ,i nw hich Li + is bound by HAl (A2). Imine reduction proceedst hrough at ransition state (A3) in which the imine is activated by N-Li coordination and attacked by an Al-H unit. The amide [N],f ormeda fter reduction, is bound to Al in at erminal positiona nd Al(m-H 3 )Li bridging is restored. With + 17.2 kcal mol À1 ,t he activation energy needed for this heterobimetallic process is only moderate. Subsequent amide!amine transformation by reactionw ithH 2 has am uch higher barrier( A5!A6: + 30.7 kcal mol À1 )a nd is endothermic (A5!A7: + 7.6 kcal mol À1 ). Instead of following this high energy path, it is clearlye asier to insert as econd imine (B1! B2*: + 17.2 kcal mol À1 ). In product B3,t he Al and Li cationsa re bridged by an amide and hydride. Hydrogenolysis of the terminal amide by H 2 is again ah igh energy process (B4!B5*: + 36.1 kcal mol À1 ). The alternative reactiono ft he bridging amide with H 2 hasa ne ven higher barrier( 38.3 kcal mol À1 )a nd insertiono fathirdi mine is clearly preferred (B3!C1*: + 23.2 kcal mol À1 ). The product C2 is rather crowded and attempts to find at ransition state for insertiono fafourth imine failed due to space limitation. Attempts to optimize the geometry of LiAl[N] 4 led to dissociation into Li[N] and Al[N] 3 ,aprocess that also has been observed experimentally. [18] Twoo ft he three amide ligands in LiAlH[N] 3 bridge Li and Al and the third is bound only to Al in at erminal position. Interestingly,i nt his case the lowest barrierf or reactionw ith H 2 was found for hydrogenolysis of ab ridging amide (C2!C3*: + 28.6 kcal mol À1 ; that for hydrogenolysis of the terminal amide is slightly higher: + 29.2 kcal mol À1 ). This last step is clearly the bottle-neck in the catalytic reaction.
The high barrierf or this last rate-limiting step explains the high temperature of 85 8Cn eededf or catalysis. The high temperaturec ould also induce side-reactions, like the experimentally observed ortho-metallation (vide supra).S tartingf rom B3 we located two transition states for ortho-alumination. The lowest barrieri sf ound for deprotonation of the terminal amide by the bridging amide (B3!D1*: + 33.0 kcal mol À1 ). A much higher barrier was found for ortho-alumination by AlÀH (+ 48.4 kcal mol À1 ). Although ortho-deprotonation could be feasible at high temperatures, the insertion of an imine is still the preferred reaction (B3!C3*:23.2 kcal mol À1 ).
This comprehensive calculation study shows that LiAlH2[N]2 (B3) is the most likely catalyst. The role of Li + in this heterobimetallicc atalysti sc oordination anda ctivation of the imine substrate. The influence of alkali metal size was studied by calculating energy profiles for the series of catalysts MAlH 4 (M = Li, Na, K). Since we only aim to compare the different metal catalysts among each other,w es implified the model system to the "naked" catalysts and omitted solvation by additional imine ligandsa nd only included corrections for solvationb y the PCM model for THF.A ss olvation effects can be large for sblock metal complexes,a bsolute energy valuess hould be treated with caution. Lack of solvation leads to increased energy barriers,b ut the energy profiles clearly show trends and the effect of metal exchange. In all cases ac haracteristic Ph···M interaction was found in the transition states (Li2*, Na2* and K2*;S cheme3). The relative energies for these transition states differ only by 1-2 kcal mol À1 .S ince the preliminaryM Al-H 4 ·(imine) complexes Li1, Na1 and K1 becomem ore stable with alkali metal size, the energy barriers increase in the order: Li (+ 29.4 kcal mol À1 ) < Na (+ 33.0 kcal mol À1 ) < K( + 35.5 kcal mol À1 ). The barrierfor the subsequenthydrogenolysis step also increasew ith metal size:L i( + 26.4 kcal mol À1 ) < Na (+ 29.5 kcal mol À1 ) < K( + 31.1 kcal mol À1 ). The calculated order for these energy barriers (Li < Na < K) is in line with the experimental observation that LiAlH 4 is the most active catalyst.
These calculations clearly demonstrate the crucial influence of the alkali metal. It can, however, not be excluded that LiAlH 4 dissolved in neat imine or in ap olar solvent like THF forms a solvent-separated-ion-pair,[ Li + (solvent) n ][AlH 4 À ], in which AlH 4 À is the true catalyst. The electrostatic energy neededt o separateL iAlH 4 in Li + and AlH 4 À ions was calculated to be only + 22.7 kcal mol À1 ,avalue that could be partially compensated for by solvation.I ndeed, solvation of Li + with four equivalents of THF is exothermic by DH = À23.7 kcal mol À1 ,i ndicating that an equilibrium between contact-ion-pair and solvent-separated-ion-pair is feasible. However,t he energy profile for imine hydrogenation with only the AlH 4 À anion (Scheme 4) is clearly not competitive with that calculated for LiAlH 4 (Scheme 2). Especially the iminei nsertion barriers, which range from + 33.5 to + 36.3 kcal mol À1 ,a re affected by loss of the alkali metal cation and are much highert han barriers calculated for the contact-ion-pair LiAlH 4 ,r anging from + 17.2 kcal mol À1 to + 23.2 kcal mol À1 .T his is obviously related to the fact that both, Al and Li, play ar ole in the imine insertion step. The hydrogenolysis step, in which only the Al center plays ar ole, is less affected by ion separation. Considering the high barriers along the pathway,i ti su nlikely that the AlH 4 À anion alonec an be the catalyst. This is in agreement with the observation that [nBu 4 N + ][AlH 4 À ]i sn ot active in iminehydrogenation. [16] The influenceo ft he p-block metal in LiAlH 4 was evaluated by comparing energy profiles for the catalystsL iMH 4 (M = B, Al, Ga);S cheme5.I mine insertion with the borate LiBH 4 (Bo1! Bo2*)h as av ery high energy barriero f+ 44.5 kcal mol À1 , which is in agreement with the fact that imines cannot be reduced by borates. The barrier for LiGaH 4 (Ga1!Ga2*: + 28.7 kcal mol À1 )i ss lightly lower than that for LiAlH 4 (+ 29.4 kcal mol À1 ). The second hydrogenolysis step, however, has am uch higher energy barrier for LiGaH 4 (Ga4!Ga5*: + 40.2 kcal mol À1 )t han for LiAlH 4 (+ 26.4 kcal mol À1 ). This explains why for gallanate catalysts only sub-stoichiometric conversion was observed in catalytic imine hydrogenation ( Table 2, entries 5a nd 6). Catalyst activity therefore decreasesi no rder LiAlH 4 > LiGaH 4 > LiBH 4 ,t hat is, with decreasing bond polarity: AlÀH > GaÀH > BÀH.
The effect of different imine substitution wasi nvestigated by comparing energy profiles for PhC(H)=NtBu, tBuC(H)=NtBu and PhC(H)=NPh using aL iAlH 4 catalyst. To reduce computation time, we omitted solvation by additional imine ligands (but included PCM corrections in THF) and only calculated the first iminei nsertion and hydrogenolysis steps (Scheme 6). The nature of the imine substituents (phenyl or alkyl) have an enormous effect on the barriersf or imine insertion. The C=N double bond in PhC(H)=NPh is conjugated with two Ph rings and highly activated for insertion which is in agreementw ith a very low barrier (P1!P2*: + 20.4 kcal mol À1 ). The highest barrier is found for the non-conjugated C=Nd ouble bond in tBuC(H)=NtBu (T1!T2*: + 36.2 kcal mol À1 )w hile that for PhC(H)=NtBu is intermediate (+ 29.4 kcal mol À1 ). The very facile imine insertion of the highly activated C=Nb ond in PhC(H)= NPh is underscored by the experimental observation that LiAlH 4 can insert four of these imines (vide supra) while for PhC(H)=NtBu am aximum of three insertions can be achieved. For the least activated C=Nd ouble bond in tBuC(H)=NtBu only two insertions are feasible (this is supported by DFT calculation;s ee Scheme S2).A lso the stability of the products (T3, P3 and Li3)i ss trongly affected by the substituents. Charge delocalizationisp ossible for the amide anion PhCH 2 (Ph)N À in P3 but not for RCH 2 (tBu)N À in Li3 and T3.A lthough this form of stabilization is advantageous for the imine insertion step, it is ad isadvantage for the subsequenth ydrogenolysis reaction. The highestb arrier is found for reaction of resonance-stabilized PhCH 2 (Ph)N À with H 2 (P4!P5*: + 31.1 kcal mol À1 ). Amides with a tBu substituent at N( RCH 2 (tBu)N À )a re much more reactive which is underscored by lower barriersf or their reactionw ith H 2 (Li4!Li5*: + 26.4 kcal mol À1 ; T4!T5*: + 29.3 kcal mol À1 ). For this reason,t he best combination of substituents is aP h group at C, to give facile imine hydrogenation, and a tBu group at N, to give facile hydrogenolysis.

Conclusions
Exchanging the alkali metal in LiAlH 4 for heavier group 1 metals was found to be detrimental for its catalytic activity in imine hydrogenation.T he following ordero fc atalyst activity was observed:L iAlH 4 > NaAlH 4 > KAlH 4 .T his implies that the sblock metal should be relatively Lewis acidic. Indeed, the sblock metal-free catalyst [nBu 4 N + ][AlH 4 À ]i si nactive. Exchanging the p-blockm etal in LiAlH 4 for either Bo rG aa lso led to loss of activity.T he nature of both, the s-a nd p-block metal, is essential for catalysis. This is ac lear indicationf or ah eterobi-metallicm echanism in which synergyb etween both metals is key to success.
Reaction of LiAlH 4 with one equivalent of PhC(H)=NtBu gave am ixture of single and doublea ddition products which implies that the second additioni sf aster than the first. Addition of at hird imine is difficult but can be achieved with an excess of imine and longer reaction times. Addition of the fourth imine could only be observed for highly activated imines like PhC(H)=NPh. Severalo ft hese intermediates could be isolated and have been structurally characterized.
Reaction of LiAlH 4 with excess PhC(H)=NtBu led under forced conditions (85 8C, several days)t od ouble imine insertion and ortho-metallation in the Ph ring. The product does not react with H 2 and this reactionc ould therefore be ac atalyst deactivation pathway.H owever,D FT calculations demonstrate that the transition state for ortho-metallation is much higher than that for imine insertion which means that this side-reaction only plays ar ole at the end of the reactionw hen imine concentrationsa re low.S equential double imine insertion/double ortho-metallation may,h owever,b ea na ttractive reactivity that could be exploited in synthesis.
Calculationsa re complicated by many variables (two different metal cations, two differenta nions), the high dynamics of these polar molecules andt he importance of solvation. Treating solvation only with the PCM methodi si nsufficienta nd additional coordination of imine to Li is needed to reduce the barriers.
The Both metals, Li and Al, actively participate in the transition state for imine insertion:L ia ctivates the imine by coordination and Al delivers the hydride for reduction. In the hydrogenolysis step only the Al center is involved. Ac atalytic cycle based only on AlH 4 À shows very high barriers and is not feasible. Calculation of the energy profiles for the catalysts MAlH 4 (M = Li, Na, K) confirmt he experimentalo bservation that the activities decrease along the row Li > Na > K, that is, with de-creasingL ewis acidity.O nasimilar note andi na greement with experiment, calculations of the energy profiles for catalysts LiMH 4 (M = B, Al, Ga) show an activity order Al > Ga > B, that is, with decreasing polarity of the MÀHb ond.
Calculation of the energyprofiles for the LiAlH 4 catalyzed hydrogenation of PhC(H)=NtBu, tBuC(H)=NtBu and PhC(H)=NPh demonstrates the influences of the C-and N-substituents on the energy barriers for insertion and hydrogenolysis. While Phsubstituents activate the C=Nd ouble bond for insertion, the resultingr esonance-stabilized amide PhCH 2 (Ph)N À is much less reactivei nt he subsequenth ydrogenolysis step than PhCH 2 (tBu)N À .T he most favorable combination is therefore found in the imine substrate PhC(H)=NtBu which is often used as the benchmark imine.

Experimental Section
All experiments were carried out using standard Schlenk techniques or ag lovebox (MBraun, Labmaster SP) and freshly dried solvents. THF (THF AnalaR Normapur,V WR) was dried over molecular sieves (3 )a nd distilled from sodium. All other solvents were degassed with nitrogen, dried over activated aluminum oxide (Innovative Te chnology,P ure Solv 400-4-MD, Solvent Purification System) and stored over molecular sieves (3 )u nder inert atmosphere. LiAlH 4 (Sigma-Aldrich, 95 %) was purified by extraction with diethyl ether and dried under reduced pressure. NaAlH 4 (Sigma-Aldrich, 90 %) was purified by extraction with THF and dried under reduced pressure. 18-Crown-6 (TCI, 98 %) was dissolved in diethyl ether,s tirred over CaH 2 ,f iltered, and dried under reduced pressure. Starting materials were used as delivered unless noted otherwise. Imine PhC(H)=NtBu was purchased from Sigma-Aldrich, stirred over CaH 2 ,a nd distilled prior to use. Imine tBuC(H)=NtBu was prepared according to Momiyama et al. [31] and stirred over CaH 2 and distilled prior to use and imine PhC(H)=NPh was prepared according to Cattone tal. [32] and dissolved in pentane, stirred over CaH 2 , filtered, and dried under reduced pressure.
NMR spectra were measured on Bruker Avance III HD 400 MHz and Bruker AvanceI II HD 600 MHz spectrometers. Elemental analysis was performed with an Hekatech Eurovector EA3000 analyzer. Crystal structures have been measured on aS uperNova (Agilent) diffractometer with dual Cu and Mo microfocus sources and an Atlas S2 detector.