Using Light to Modify the Selectivity of Transition Metal Catalysed Transformations

Abstract Light has a remarkable and often unique ability to promote chemical reactions. In combination with transition metal catalysis, it offers exciting opportunities to modify catalyst function in a non‐invasive manner, most frequently being reported to switch on or accelerate reactions that do not occur in the dark. However, the ability to completely change reactivity or selectivity between two different reaction outcomes is considerably less common. In this Minireview we bring together examples of this concept and highlight their mechanistically distinct approaches. Our overview demonstrates how these non‐natural, photo‐switchable systems provide key fundamental mechanistic insights, enhancing our understanding and stimulating development of new catalytic activity, and how this might lead to tangible applications, impacting fields such as polymer chemistry.


Introduction
Tr ansition metal catalysis has revolutionised synthetic chemistry,l eading to the development of entirely new reactivity,more-efficient transformations and enabling access to fresh chemical space.Atthe same time,photochemistry has seen ar esurgence of interest in the last decades,a llowing access to radical reaction pathways under mild and controlled conditions.U nsurprisingly,t he combination of these two fields is therefore currently an area of intense interest and anumber of recent reviews have dealt with various aspects of this area, from dual catalysis [1,2] to direct excitation of transition metal catalysts. [3,4] Them ajor focus when developing catalytic systems has been to tune their properties to maximise conversions, enhance functional group tolerance and improve selectivities. However,o nce an ew system has been developed, these characteristics are fixed, such that ar eaction will have ap reprogrammed outcome.O nly then, by physically changing the reaction conditions (for instance the reagent, the catalyst, the ligand) is it possible to modify the enantio-, diastereo-, chemo-or regioselectivity of ar eaction. Nevertheless,d ivergent strategies such as these are valuable because they allow chemists to obtain different products from one common precursor. [5] Furthermore,t he ability to selectively obtain different reaction outcomes on the same molecule can result in shorter and more efficient reaction sequences,a nd newly selective methods can give products that were previously inaccessible.
In contrast to these single-function catalytic platforms,an area of research that has received significantly less attention is the development of multifunction transition metal catalysts. Here,the activity of the system can be switched by an external stimulus,a llowing several different reaction outcomes to be obtained. Light is aparticularly attractive stimulus to use,asit is non-invasive and easily switchable,t hus offering excellent spatio-temporal control. However,m ost commonly,l ight is used only to switch on or accelerate ar eaction. [6] In this Minireview,w ewill focus on the examples in the literature where two different reaction outcomes can be obtained with the same transition metal catalyst, using light irradiation to dictate which product is formed, either by switching al ight source on, modifying its wavelength or changing the intensity.A lthough the current number of examples is somewhat limited, already ab road range of strategies has been reported, offering exciting opportunities to modify catalyst function in an on-invasive manner,a nd making the area ripe for further exploration.
Broadly,c urrent strategies can be split into two categories:light-induced modification of either the catalyst or of the substrate (Figure 1a,b). Where light is used to change the identity of the reactive catalytic species,t his can involve photoisomerisation of the ligand structure which affects either the geometry (and thus the shape of the catalyst binding site) or the electronic properties at the metal centre.A lternatively, excitation of the transition metal complex can lead to ligand photodissociation, affecting the ability of the substrate to bind to the metal centre,which can result in divergent mechanistic pathways.E xcitation of the catalyst may also result in alteration of the electronic configuration and its redox properties and, finally,particularly in heterogeneous catalysis, light irradiation can result in the alteration of the physical properties at the catalyst surface.
Alternatively,l ight can be used either directly or via asecond catalyst to modify the starting material or areactive intermediate,inducing either isomerisation or aredox event, such that the molecule undergoing the transformation by the transition metal is different, depending on the presence or absence of light irradiation.
Thee xamples highlighted in this Minireview range from the exquisitely designed to those which reveal unexpected divergences of mechanism. Notably,i tm ay not always be predictable which parts of the reaction that light will exert control over, or indeed whether the effect of light can necessarily be limited to only one aspect. Given the range of possibilities,i ti se ntirely possible that future research may involve multiple divergency points in the mechanisms between light and dark. We therefore expect the number of light-switchable transition metal catalysed processes to in-Light has aremarkable and often unique ability to promote chemical reactions.Inc ombination with transition metal catalysis,i toffers exciting opportunities to modify catalyst function in anon-invasive manner,m ost frequently being reported to switchono raccelerate reactions that do not occur in the dark. However,t he ability to completely change reactivity or selectivity between two different reaction outcomes is considerably less common. In this Minireview we bring together examples of this concept and highlight their mechanistically distinct approaches.O ur overview demonstrates howt hese non-natural, photo-switchable systems provide key fundamental mechanistic insights,enhancing our understanding and stimulating development of new catalytic activity,a nd how this might lead to tangible applications,i mpacting fields sucha spolymer chemistry.
crease significantly in the years to come,o pening up new avenues for on-demand selective synthesis.
Excitingly,due to the widespread availability of LED light sources,m ore and more synthetic research groups may find themselves working in this area. Unexpected photoreactivity which contrasts with thermal reactivity of catalysts will almost certainly lead to the discovery of new modes of catalytic activity,expanding even further the reach of transition metal catalysis and allowing previously unattainable selectivities in chemical synthesis.E xcellent mechanistic understanding, through spectroscopic and computational methods,w ill be essential to accompany developments in this area so that any discoveries can be fully capitalised on.

Light-Responsive Ligands
There has been significant interest in developing new molecular photoswitches that can be reversibly toggled between two thermally stable isomers. [7] These molecular motifs possess at least two distinct structural geometries and have found potential applications in molecular devices. However,a pplications of these switches for regulating chemical reactivity of catalysts had been limited until recently. [8] Several examples have now been reported which rely on carefully crafted systems,with two distinctive shapes, giving rise to active catalytic species with different steric or electronic properties at the reactive transition metal centre.
One of the first examples to demonstrate ad egree of photocontrolled selectivity in at ransition metal catalysed transformation was reported by Branda and co-workers in 2005. [9] Here,t he authors reported the careful design of ab is(oxazoline) ligand which incorporates a1 ,2-dithienylethene (DTE) moiety into the backbone.T he significant change in the metal-binding pocket of the ligand between ring-opened (1a)a nd ring-closed (1b)f orms was demonstrated in the context of acopper-catalysed cyclopropanation reaction (Scheme 1). Ther ing-opened form of the ligand yielded the products with moderate enantiomeric excess (ee); however, this was completely eroded upon UV irradiation, which ring-closes the DTE moiety of the bis(oxazoline) ligand.
Whilst this case demonstrated the loss of selectivity of at ransition metal catalyst, the first example of ac omplete switch of selectivity,such that asingle enantiomer of aligand could be used to obtain both enantiomers of ap roduct on demand, was reported by the Feringa group in 2015. [10] They designed the chiral phosphine ligand 2a-d,inspired by other C 2 -symmetric ligands,w hich could exist in two (pseudo-)enantiomeric forms owing to the incorporation of acrowded

Angewandte
Chemie alkene molecular motor core.N otably,t his four-stage switch also allows access to the racemic structures.
Thed evised catalyst system, using the chiral switchable bisphosphine ligands,w as applied for the enantioselective control of Pd-catalysed desymmetrisation reactions of biscarbamates.D ual stereocontrol is achieved by the use of au nidirectional light-driven molecular motor with an amide linker which undergoes a3 608 8 four-stage rotary cycle, involving photo-and thermal helix isomerisation steps,a nd leading to the change of helicity and geometry of the ligand (Scheme 2). Excellent enantioselectivity is attained by the stepwise control of the rotation cycle,w hich leads to the formation of three stable ligand isomers: P,P-trans 2a catalysing the formation of racemic mixture,a nd pseudoenantiomers M,M-cis 2b and P,P-cis 2c,both of which lead to the formation of as ingle enantiomer of oxazolidinone (Scheme 2). However,t hermally induced helix inversion between the pseudo-enantiomeric isomers is irreversible, and in order to recover the initial 2b,t hree consecutive isomerisations (light-heat-light) are necessary,instead of one isomerisation starting from the P,P-cis form.
More recently,t he Feringa group reported the development of as ystem with phosphoramidite molecular switches based on second-generation overcrowded alkene molecular motors. [11] Thes witch core contains as ix-membered ring in the upper half and af ive-membered ring in the lower half (Scheme 3, 3a and 3b). Compared to the previous motor, there are only two isomerisation stages,s ince the biphenyl motif suppresses the continuous unidirectional rotation about the central alkene.F urthermore,t he photogenerated meta-Scheme 1. Photoresponsive ligands that control enantioselectivity.

Angewandte
Chemie stable states possess long half-lives (up to 1.3 years at 20 8 8C), meaning both pseudo-enantiomers of the switch can be used without continuous irradiation over longer time periods.
Thed esign of the ligand is based on combining the photoswitchable biphenol motif with an amine at the phosphorus centre to make up the phosphoramidite.T he crowded alkene can be atropisomerised without disrupting the bridged biaryl unit or affecting its flexibility,r esulting in as ignificant change of geometry and steric environment around the phosphorus centre.I nt he phosphoramidite synthesis,amixture of diastereomers at the phosphorus centre are also obtained, depending on the identity of the amine. This turns out to be important for the effective switching of enantioselectivity,because the barrier for pyramidal inversion at the phosphine is high enough to prevent it at room temperature.
In total, the ligands have five stereochemical elements: the fixed phosphorus configuration (Scheme 3, pale blue box, R P or S P ); the fixed carbon stereogenic centre (Scheme 3, green box, S); and then three interconnected elements-the helicity of the overcrowded alkene,t he dynamic helical geometry and the axial chirality of the biaryl unit-which are given acombined stereodescriptor based on the helicity of the alkene chromophore (Scheme 3, grey box, M or P).
These newly developed ligand systems were studied in the context of copper(I)-catalysed asymmetric conjugate additions of diethylzinc to 2-cyclohexen-1-one.Reversible switching of chirality is achieved by light:i rradiation at 365 nm triggers isomerisation of the ligand and causes ac hange of catalytic activity and enantioselectivity.U nlike previous examples of ligands containing alkene photoswitches,w here photoisomerisation was possible only with the free ligand (due to quenching of the alkene excited state by energy transfer to the metal), this system allows dynamic switching even with the ligand bound to the copper centre.Interestingly, the overall catalytic performance of the ligand structure is dictated by the matched and mismatched relationship between the fixed chirality at the phosphorus and the dynamic helicity of the photoswitch. Therefore,t he most effective system for significant changes in enantioselectivity were those that contained two competing diastereomeric ligands in the two reversible states of the system. Fort he ligand system shown in Scheme 3, prior to irradiation, the mixture consists of the poorly active major diastereoisomer (S,S P ,M)w ith stereoselectivity for the (S)-product and the highly active minor diastereoisomer 3a (S,R P ,M)w ith stereoselectivity for the (R)-product. After photoswitching, the behaviour is inverted:t he weakly active major isomer becomes more active to give 3b (S,S P ,P), and the very active minor isomer is converted into its less active counterpart (S,R P ,P), therefore overall favouring formation of the (S)-product.
TheB ielawski group has investigated how the incorporation of aphotochromic dithienylethene (DTE) moiety into an N-heterocyclic carbene (NHC) ligand can modify its donicity. [12] In this context, they developed the first photoswitchable olefin metathesis catalyst 4a based on the Hoveyda-Grubbs second-generation catalyst (Scheme 4). [13] This novel Ru II catalyst has an NHC ligand which bears ap hotochromic DTE moiety.T he UV/Vis spectrum exhibits an intense absorption band at 298 nm that was assigned to ac ombination of the n!p*t ransition of the N-heterocycle and the p!p*transition of the aryl rings.Upon irradiation at 313 nm, ac olour change occurs concurrent with decreased intensity of the band at 298 nm and the appearance of new absorption bands centred at 453 and 639 nm. Together with other experimental data ( 1 HNMR shifts of the benzylidene resonance), this provided the evidence that electrocyclicring closure had occurred to form an extended p-conjugated system, 4b.E xposure to visible light (l > 500 nm) reverses this process,restoring the catalyst to the ring-opened state 4a.
Having demonstrated the reversibility of the photoisomerisation process,t he catalytic activity of the two forms of the catalyst were then investigated for both ring-closing metathesis (RCM) and ring-opening metathesis polymerisation (ROMP). It had been previously shown that the donating ability of this NHC ligand is significantly different in the two states.C alculations showed that this was important for RCM of 1,6-heptadiene,w here the rate-determining step (RDS) was the transition state of the retro-[2+ +2] cycloaddition (TS I,Scheme 4). In 4a,the NHC ligand is astronger donor, which stabilises the Ru IV metallacyclobutane preceding this transition state,r esulting in ah igher barrier for this step,and thus the process is faster for 4b.

Angewandte Chemie
ForROMP reactions,the ring-opened form of the catalyst is more effective,b ut the reasoning is substrate dependent. Forc yclooctadiene (COD), the increased steric bulk of the ring-closed form of the catalyst (arising from increased intrusion of the mesityl groups as ar esult of planarisation) results in an unfavourable interaction in the rate-determining retro-[2+ +2] step (TS II,Scheme 4).
Conversely,f or ROMP of norbornene,t he calculated RDS is the [2+ +2]-cycloaddition where the steric differences between the two forms of the catalyst are negligible.Here,the stronger donating ability of the ring-opened form of the catalyst promotes the cycloaddition due to the stabilisation of the ruthenacyclobutane intermediate (via TS III,S cheme 4). Overall, this means the ring-opened form of the catalyst 4a accelerates ROMP compared to the ring-closed form 4b, whereas,i nc ontrast, the ring-closed analogue shows the opposite activity by increasing the rate of RCM. Up to 1.7fold rate enhancement is observed between the two photoswitchable catalytically active states of the complex.
Light-responsive ligand motifs have also been applied for polymerisation reactions,w here the external control over selectivity between different reactions or substrates can have as ignificant impact for modifying the composition of polymers. [14,15] An umber of other light-responsive catalytic polymerisation systems have been reported [16][17][18][19] and these systems are likely to play as ignificant role in developing materials of the future. [20,21] In 2019, Chen and co-workers incorporated well-explored azobenzene photoswitches into the ligand structure of salicyaldimine Zn II complexes,w hich were investigated as polymerisation catalysts for ring-opening polymerisation (ROP). [22] Photoswitching was shown to be reversible in the case of the unsubstituted arene attached to the azobenzene (Scheme 5) in contrast to other substituted arenes.T he authors investigated the ability of the two forms of the catalyst to discriminate between different lactone-type monomers and thus alter the composition of the polymer based on an external stimulus.T he trans-Zn complex 5a can be isomerised to the cis form 5b with 365 nm light. Ther everse process can be achieved with 420 nm light. Both isomers are active in ROPb ut show different reactivities,w ith the trans-Zn complex being slightly more active for l-lactide (LA) in contrast to most other monomers such as trimethylenecarbonate (TMC) and caprolactone (CL) which are more efficiently polymerised by the catalyst in its cis form.
Interestingly,isomerisation results in some discrimination during copolymerisation of am ixture of TMC and CL monomers.T he initial rate of polymerisation for TMC for the trans-Zn complex is 1.4 times faster than for CL;however, upon UV-light-induced isomerisation to the cis-Zn complex, this changes to a6 .7-fold rate difference,d ecreasing the amount of CL incorporation into the copolymer. It was speculated that adifference in the electronic effects between the two forms of the catalyst is responsible for the differences in activity that is observed, because the steric environment of the metal is largely unaffected by cis-trans isomerisation.

Ligand Photodissociation
Modifying the coordination sphere of at ransition metal complex via ligand photodissociation has typically resulted in "on" and "off" states for catalytic transformations.However, it is also possible,though currently less predictable,touse this approach to give two different "on" states.B ys witching between open-and closed-shell intermediates,c ompletely different mechanisms can be favoured (e.g.asaturated coordination sphere might favour outer-sphere electron transfer whereas free-coordination sites would favour innersphere substrate coordination).
In this context, our group recently reported unique, photocontrolled chemoselective hydroboration reactions of a,b-unsaturated ketones using the well-defined transition metal complex CoH[PPh(OEt) 2 )] 4 6 in conjunction with pinacolborane. [23] Notably,c yclic a,b-unsaturated ketones were able to undergo selective 1,2-or 1,4-hydroboration reactions depending on the absence or presence of visible light, respectively (Scheme 6a). This approach also enabled the formation of cyclic boron enolates which had been al imitation of previous methodologies.T hese could be reacted directly with ar ange of aldehydes,i no ne-pot aldol reactions,l eading to the formation of syn-aldol products.I t was demonstrated that the divergent selectivity arises from distinctive mechanistic pathways in the dark and light. Irradiation with visible light triggers ligand dissociation (Scheme 6b), thereby generating the coordinatively unsaturated 16-electron cobalt(I) hydride complex 6a to which coordination of the substrate can occur. This results in ac oordination/insertion type mechanistic pathway,w ith insertion into the C = Cb ond favoured, leading to the enolborates,a nd ultimately formation of the 1,4-reduction product upon quenching.Inthe absence of light, however, the substrate is unable to coordinate at the saturated metal centre,r esulting in ad ifferent starting point to the mechanism. Instead, an initial reaction with the pinacolborane and electron transfer to the starting material occurs to generate the Co 0 intermediate 6b,w hich is an overall exergonic step. With the assistance of the starting material, the HBpin is activated by this Co 0 intermediate,r esulting in an HATt ype mechanism.

Angewandte
Chemie Using as imilar system, our group also reported the switchable reactivity of this system with terminal alkenes. [24] Using the same cobalt(I) hydride complex, it was possible to switch between anti-Markovnikov alkene hydroboration under visible light irradiation and alkene isomerisation in the dark via two distinct reaction pathways (Scheme 6c).
Recently,asimilar strategy has been reported by Chirik and co-workers involving photodissociation of aC Or ather than ap hosphonite ligand. [25] They focussed on the catalytic hydrogenation activity of ab ench-stable,c oordinatively saturated cobalt(I) precatalyst 7,( R,R)-(iPrDuPhos)Co-(CO) 2 H. Te rminal, di-and trisubstituted alkenes,a lkynes and carbonyls were investigated under both thermal and photochemical conditions.H ydrogenation occurs via divergent mechanistic pathways depending on the reaction conditions:upon heating to 100 8 8C, the reaction follows an HAT pathway facilitated by cleavage of the relatively weak CoÀH bond (56 kcal mol À1 ). In contrast, irradiation of 7 with blue light leads to carbonyl ligand dissociation and formation of a1 6-electron complex, which undergoes ac oordinationinsertion sequence with the olefin involving closed-shell intermediates.G enerally,t he photochemical method surpassed the thermal one in the case of hydrogenation of sterically hindered and electron-rich substrates,w ith one example of as electivity difference being described (Scheme 6d).
These three examples underscore the potential of coordination-sphere modification to impact the mechanistic pathway,r esulting in divergent selectivity.U nusually,t he radical mechanistic pathways occur under thermal/dark conditions rather than photochemical conditions,a smight be expected. Nonetheless,improved reactivity was observed under photochemical conditions,d emonstrating the ability of this approach to improve efficiency of catalysts.
Another example of photoswitchable behaviour was reported by Fumagalli et al.,who investigated the behaviour of [Cu(dap) 2 ]Cl (8)a st he catalyst for an alkene difunctionalisation reaction. [26] Here,t he hypervalent iodine Zhdankin reagent 9 was used as as ource of azide radical and at hird polar component served as the nucleophile.T he addition of azide to styrene-type double bonds is followed by addition of at hird component at the benzylic position. Light is able to control the degree of azidation in the following way:i nt he presence of visible light, oxidation of the benzylic radical occurs to yield the cation, which is subsequently quenched with the alcohol solvent resulting in methoxyazidation. In contrast, in the absence of light, double azidation occurs (Scheme 7). Thea uthors suggest that this results from ac hange of mechanism to ar adical chain process in this case,a lthough more recent work notes that there is some evidence that ligand dissociation from Cu(dap) 2 Cl can occur upon light irradiation, which may lead to as witch in the Scheme 6. Photocontrolled cobalt catalysis via ligand photodissociation.

Angewandte
Chemie mechanism between inner-a nd outer-sphere. [27] However, both alkene functionalisation reactions occur under mild reaction conditions at room temperature,a nd the reaction outcome is controlled solely by the presence or absence of visible light. It was shown that different substituent patterns are generally well-tolerated, although more electron-poor olefins do not yield the desired products under the reaction conditions.

Catalyst Excitation
Modifying the irradiation wavelength can be astraightforward way of influencing the reactivity of as ystem. [28] Alternatively,itisalso possible to do so by altering the intensity of the irradiation. [29] Thee ffectiveness of this underexplored approach has been recently demonstrated by Kerzig and Wenger. [30] They reported that the switching from the one-to two-photon activation of an Ir-based photocatalyst can lead to the formation of different products under very similar conditions.T he modulation of the light intensity per area is realised with inexpensive optics commonly used for beam collimation. Thew ater-soluble photocatalyst Irsppy (10), previously designed by the same group,can absorb aphoton under blue light irradiation to form the corresponding excited triplet state,w hich can be engaged in dehalogenation and isomerisation reactions (Scheme 8, left). If another photon is absorbed within the lifetime of the excited state (and this can be achieved by using al ight source with higher intensity per irradiation area), hydrated electrons (e aq C À )are generated. In this case,the system exhibits different behaviour that leads to the formation of different products compared to those formed by the one-photon pathway (Scheme 8, right).
Thea uthors focused their attention on three processes: 1) the selective reduction of C(sp 2 )-halogen bonds in an aromatic system;2 )the trans--cis isomerisation of olefins, rather than their hydrogenation;a nd 3) the reductive deha-logenation of abenzyl chloride,incontrast to its dimerisation. Fore ach process,t he different reactivity is ar esult of the different properties of 3 Irsppy and e aq C À .F or process 1), the selectivity is controlled by the redox potentials of the two species:w hile 3 Irsppy (E 1/2 = À1.6 V) is only able to reduce CÀBr bonds,e aq C À (E 1/2 = À2.9 V) is able to activate the more challenging CÀCl bond. In the case of process 2), the alkene isomerisation is induced by at riplet-triplet energy transfer (TTET) whereas the reduction occurs via ad irect electron transfer (ET). Lastly,i ns ystem 3), the starting material is consumed much faster in the two-photon conditions;t his means that the local concentration of benzyl radical is high enough to favour the dimerisation over the simple dehalogenation.

Aggregation State
Theo utcome of ar eaction can also be altered by influencing the rate of adsorption and desorption onto ah eterogeneous catalyst. Sarina and co-workers reported that the alkyne hydroamination reaction catalysed by Au 2 Co alloy nanoparticles leads to the expected cross-coupling product, an imine,w hen carried out under visible light irradiation but in the dark mainly homocoupling of the alkyne is observed (Scheme 9). [31] Scheme 7. Light-controlled switching between methoxyazidation and double azidation of styrenes.

Angewandte Chemie
Theswitch in the selectivity arises from the different rate of adsorption of the two substrates on the nanoparticle surface and it can be altered under low-flux visible light irradiation. Thel ight affects the system in two ways that contribute to the formation of the cross-coupling product: 1) light enhances the adsorption of aniline to the catalyst surface;2)light also weakens the adsorption of the alkyne.It had been previously reported that gold nanoparticles can display catalytic activity due to the surface plasmon resonance effect, whereby the oscillation of conduction electrons resonates with the electromagnetic field of the incoming light, thus enhancing local magnetic fields on the surface of the nanoparticle,a nd ultimately resulting in excited electrons. [32] Fort his example,aplasmon-generated optical plasmon force might explain why aniline adsorption is favoured in the light, but this force is expected to have an egligible effect on as ystem irradiated with al ow-flux visible light source such as the one used in this study. However,the interaction between the alloy and the substrate is also influenced by the polarizability of the substrate.Under these conditions,aniline can be converted to its electronically excited state,w hich presents ac onsiderably higher polarizability.T he optical plasmon force experienced by the excited aniline is therefore stronger and it is postulated to be sufficient to overcome the Brownian motion of the excited molecules,l eading to the adsorption onto the alloy surface. On the other hand, phenylacetylene does not absorb visible light and therefore it is not affected by the same force.
Zhao and co-workers studied the catalytic activity of gold nanorods (AuN Rs) supported on TiO 2 nanofibers. [33] After the synthesis and characterisation of this novel material, the authors turned their attention to ap otential synthetic application. They investigated the photochemical homocoupling of benzylamines under air and in solvent-free conditions.T he reaction leads to the formation of ab enzylidenebenzylamine as the major product and benzaldoxime,a s am inor byproduct when carried out under visible light or near-IR irradiation. Interestingly,t he selectivity of the reaction is partially reversed when it is carried out in the dark (Scheme 10). When benzaldoxime is used as reactant in the "light conditions" no homocoupling is observed, which indicates that this species is not an intermediate in the imine formation process.I ti stherefore likely that two different reaction pathways are responsible for the different selectivity.
EPR studies suggest that O 2 C À is present in the light pathway but not in the dark and, unsurprisingly,n or eactivity is observed in absence of oxygen. According to the proposed mechanism, when oxygen and the starting material are adsorbed onto the catalyst surface,t he light triggers the transfer of hot plasmonic electrons that leads to the formation of O 2 C À and an N-centred radical cation. This cation can be further oxidised to an aldimine that condenses with amolecule of benzylamine,g enerating the imine and ammonia. In the dark, benzylamine is directly oxidised to generate the aldoxime.

Substrate Modification
Another approach to alter the selectivity of transition metal catalysed reactions is to use light to modify the substrate.T his could involve altering the geometry of the starting material, or activating acertain functionality from the

Angewandte
Chemie ground state such that it becomes more reactive than another functional group which reacts in preference in the dark.
Forinstance,light has been used as atrigger to switch the enantioselectivity of ar eaction by acting only on the substrate,w here commonly ac hange in the configuration of the catalyst would have been required. Gilmour and coworkers reported the enantiodivergent synthesis of b-chiral phosphonates starting from stereodefined a,b-unsaturated phosphonates using as ingle enantiopure catalyst (Scheme 11). [34] Ther eduction, carried out under 10 atm of H 2 and catalysed by the rhodium complex 11,can be preceded by the photoisomerisation (at 365 nm) of the double bond. This is achieved by selective energy transfer using inexpensive anthracene as the photocatalyst. E-vinyl phosphonates are irreversibly isomerised to the Z isomer due to the 1,3-allylic strain that inhibits the re-excitation by weakening the conjugation between the p system of the arene ring and the double bond. Then ature of the substituents Ra nd, in particular,R ' are important to achieve ah igh Z/E ratio; indeed, when R = R' = H, the photostationary state consists of a1:1 mixture of the two geometric isomers.Moreover,the efficiency of the isomerisation is lost when Ri sa ne lectronrich group (such as am ethoxy). This can be explained considering ag reater contribution of resonance forms that can facilitate the rotation of the CÀCb ond, leading to the E isomer after relaxation. Thes ubsequent rhodium-catalysed reduction occurs with high stereospecificity,affording the two enantiomeric products with e.r. up to 99:1. When the Ephosphonate is reduced, the (R)-enantiomer is formed, whereas if the starting material is first converted to the Z isomer,t he (S)-enantiomer is obtained. Compared to more traditional approaches to enantiodivergent catalysis,t his method has the advantage of using as ingle enantiomer of the catalyst. Normally the switch in enantioselectivity is achieved by using the enantiomer of the catalyst, but sometimes accessing it might not be straightforward.
In ad ifferent context, Hashmi, Klein and co-workers demonstrated that aryldiazonium salts can show chemodivergent reactivity under the action of gold catalysis, depending on whether the reaction is carried out under blue LED irradiation (450 nm) or in the absence of alight source (Scheme 12). [35] In the first case,the reaction of the diazonium salt with o-alkynylphenols leads to arylated benzofurans;i n the latter,tosubstituted azobenzofurans.T he authors carried out extensive (experimental and computational) mechanistic studies to gain insight into the origin of the divergency. Based on their results,t he formation of akey vinyl Au I complex 12 was postulated. This species can form EDAc omplex 13 with the diazonium salt, and irradiation of this complex leads to N 2 extrusion, resulting in the formation of an ew C À Cb ond. In absence of alight source,the diazo moiety is retained and the diazonium salt behaves as an N-electrophile.Apart from the last step,t he mechanisms proposed for the two transformations are equivalent. Thei nitial p-coordination of the Au complex with the alkyne moiety activates the triple bond towards a5 -endo-dig cyclisation, in which the oxygen functions as nucleophile.T he resulting intermediate (14)i s deprotonated, generating vinyl Au I complex (12)t hat can then evolve into one of the two products,d epending on the conditions.
In one final example,Lemcoff and co-workers investigated ac hromatic orthogonal catalytic process in which two photochemical processes,t riggered by light of different Scheme 11. Enantiodivergent reductions using asingle enantiopure reduction catalyst via alkene isomerisation.

Angewandte
Chemie energy,a re combined:p hotocleavage of as upersilyl protecting group at 254 nm and the activation of the dormant Ru catalyst 15 a for RCM reaction at 350 nm. [36] Interestingly,the supersilyl group can influence the selectivity of ring-closing metathesis reactions,a ltering the preferred ring size that is formed compared to that with the unprotected alcohol starting material. This opened up the possibility of developing an oncommutative sequence,w hereby the selectivity of the reaction would be dictated by the order of the sequence of light irradiation of different wavelengths.
In the first sequence (Scheme 13, left), the supersilylprotected triene 16 was irradiated first at 254 nm, resulting in cleavage of the supersilyl group,w ith the metathesis catalyst remaining intact but inactive.Notably,the selection of solvent (1,1,2,2-tetracholorethane) was important to allow deprotection of the supersilyl group in the presence of the ruthenium catalyst, and aless-active precatalyst (where R = phenyl) was more effective in the one-pot sequence due to lessened decomposition. Subsequent irradiation of the reaction mixture (together with addition of CD 2 Cl 2 to enhance the cistrans isomerisation of the Ru catalyst) led to aRCM reaction which provided the six-membered dihydropyran structure 17 in 6:1p reference to five-membered dihydrofuran 18.I n contrast, for the second sequence (Scheme 13, right), the reaction mixture was first irradiated with 350 nm light, which led to preferential formation of af ive-membered ring via RCM due to the bulkiness of the protected alcohol. Irradiation at 254 nm led to the cleavage of the silyl protection group,giving product 18 as the major product (30:1 ratio with 17).
An important prerequisite for developing this orthogonal process was that the light-sensitive supersilyl protecting group remains intact upon irradiation with 350 nm light, and the RCM catalyst remains largely untouched under irradiation with 254 nm light. Thed evelopment of non-commutative sequences such as this holds exciting potential for the development of divergent synthetic routes controlled solely by light.

Summary and Outlook
To conclude,ithas been shown that light can influence the course of transition metal catalysed reactions in ar ange of different ways.Excitation of the catalyst, either directly or of the ligand, can result in alteration of the steric or electronic properties of the active catalyst. Alternatively,l ight-induced modification of the substrate,f or example via isomerisation or deprotection, can also result in divergent selectivity or reactivity.T hese approaches open up the possibility to access multiple different functions for as ingle catalytic system. Together with the advances in automation technology in synthetic chemistry,t his has potential to lead to more efficient, controlled sequences of catalytic transformations. Tr uly reversible,s witchable processes could be combined in production lines to allow efficient and swift diversification of common intermediates.
Despite the current limited range of examples on this topic, the reported approaches encompass ab road array of different strategies.Due to its non-invasive character and ease of set-up,light remains an easy and appealing way to modify the reactivity of transition metal catalysts.W ithout doubt, there are many opportunities going forward to design lightresponsive catalysts which will allow us precise control over selectivity of reactions in aw ay that was not previously possible.N ew work in this area is likely to arise both from exquisitely designed strategies,r elying on careful ligand design, but also from serendipitous discoveries.F uture research will undoubtedly lead to the discovery of many new divergent modes of catalysis which are poised to have applications across synthetic chemistry.