Semiconductor Photocatalysis for Chemoselective Radical Coupling

Article pubs.acs.org/accounts

Semiconductor Photocatalysis for Chemoselective Radical Coupling Reactions Horst Kisch* Institute of Inorganic Chemistry, University of Erlangen-Nürnberg, D-91058 Erlangen, Germany CONSPECTUS: Photocatalysis at semiconductor surfaces is a growing field of general photocatalysis because of its importance for the chemical utilization of solar energy. By analogy with photoelectrochemistry the basic mechanism of semiconductor photocatalysis can be broken down into three steps: photogenerated formation of surface redox centers (electron−hole pairs), interfacial electron transfer from and to substrates (often coupled with proton-transfer), and conversion of primary redox intermediates into the products. Sun driven water cleavage and carbon dioxide fixation are still in the state of basic research whereas aerial degradation reactions of pollutants have reached practical application for the cleaning of air. In addition, a great variety of organic transformations (not syntheses) have been reported. They include cis− trans isomerizations, valence isomerizations, cycloaddition reactions, intramolecular or intermolecular C−N and C−C couplings, partial oxidations, and reductions. In all cases, well-known products were formed but very rarely also isolated. As compared to conventional homogeneous organic synthesis, the photocatalytic reaction mode is of no advantage, although the opposite is quite often claimed in the literature. It is also noted that a high quantum yield does not implicate a high product yield, since it is measured at very low substrate conversion in order to minimize secondary photoreactions. That is especially important in semiconductor photocatalysis since photocorrosion of the photocatalyst often prevents long-time irradiation, as is the case for colloidal metal sulfide semiconductors, which in general are photochemically too unstable to be used in synthesis. In this Account, we first classify the numerous organic photoreactions catalyzed by semiconductor powders. The classification is based on easily obtainable experimental facts, namely the nature of the light absorbing reaction component and the reaction stoichiometry. Next we discuss the problem of quantitative comparisons of photocatalytic activities or apparent quantum yields and propose a basic three-step mechanistic model. Finally, we address the question whether or not the unique photoredox properties of simple inorganic semiconductor powders may lead to previously unknown visible light induced organic syntheses. For that, we summarize novel radical C−C− and C−N− couplings photocatalyzed by self-prepared cadmium sulfide powders. Electron acceptor and donor substrates like imines or 1,2-diazenes, and cyclic olefins or unsaturated ethers, respectively, undergo a linear addition reaction. The hitherto unknown products have all been isolated in good to moderate yields and may be of pharmaceutical interest. In the first reaction step photogenerated electron−hole pairs produce through proton-coupled electron transfer the corresponding radicals. Their subsequent chemoselective heterocoupling affords the products, correlating with an insertion of the imine or 1,2-diazene into an allylic C(sp3)-H bond of the donor substrate. In the absence of an imine or 1,2-diazene, cyclic allyl/enol ethers are dehydrodimerized under concomitant hydrogen evolution. Even a visible light photosulfoxidation of alkanes is catalyzed by titania. In these heterogeneous photoredox reactions the role of the semiconductor photocatalyst is multifunctional. It induces favorable substrate preorientations in the surface-solvent layer, it catalyzes protoncoupled interfacial electron transfer to and from substrates generating intermediate radicals, and it enables their subsequent chemoselective coupling in the surface-solvent interface. Different from molecular photosensitizers, which enable only one oneelectron transfer with one single substrate, photoexcited semiconductors induce two concerted one-electron transfer reactions with two substrates. This is because the light generated electron−hole pairs are trapped at distinct surface sites and undergo protoncoupled interfacial electron transfers with unsaturated donor and acceptor substrates. The radicals diffuse in a solid-solute− surface layer to undergo chemo- and stereoselective C−C and C−N bond formation. Thus, the semiconductor photocatalyst functions like an artif icial leaf. Since several minerals are known to have semiconductor properties, solar photocatalysis may be also relevant for prebiotic and environmental chemistry.

1. INTRODUCTION The multistep reaction scheme of green plant photosynthesis may be broken down into four processes. Photochemical charge generation, charge trapping at photosystems (PS) I and PS II, interfacial electron transfer (IFET) to NADP+ and from H2O at spatially separated PS I and PS II, © 2017 American Chemical Society

respectively, and transformation of the primary redox products into oxygen and carbohydrates via O−O and C−C bond formation. Received: January 10, 2017 Published: April 5, 2017 1002

DOI: 10.1021/acs.accounts.7b00023 Acc. Chem. Res. 2017, 50, 1002−1010

Article

Accounts of Chemical Research hv

Charge generation occurs also upon photoexcitation of molecular photocatalysts, but recombination is usually too fast to observe concerted electron transfer to acceptor and from donor substrates. In general a molecular photocatalyst like an organic dye or a transition metal complex acts either as reductant or oxidant and requires the presence of a reducing or an oxidizing agent for its reformation.1,2 This differs from semiconductor surfaces (SC) in contact with gaseous or dissolved substrates. Now, the photogenerated charges are trapped at distinct surface sites and concerted IFET reactions with substrates A and D become feasible if their reduction potentials are located within the bandgap (Scheme 1, excitation via hν1). The thermodynamic

A + D( +C) → A − D( +A − D − C) SC

(2)

In general, the semiconductor absorbs the light, and we proposed to name this case direct semiconductor photocatalysis (Scheme 1, excitation via hν1). When light is absorbed by a substrate, like in dye degradation, the reaction can be named indirect semiconductor photocatalysis (Scheme 1, excitation via hν2). In general a reductive IFET from D* is operating, whereas an oxidative counterpart via A* is very rare. Commonly, neither SC nor D or A absorbs the light but a surface charge-transfer (CT) complex (eq 3) exhibiting a CT absorption band. Now, an optical electron transfer according to eq 4 and subsequent IFET to A affords the primary products (eq 5). Obviously, in many cases, it is difficult to identify the light absorbing species. However, quite often it was even not attempted.

Scheme 1. Schematic Description of Direct (via hν1) and Indirect (via hν2) Semiconductor Photocatalysisa

SC + D ⇄ [SCδ −···Dδ +]

(3)

hv

[SCδ −··· Dδ +] → [SC(e−)··· D·+]

(4)

[SC(e‐) ···D·+] + A → SC + D·+ + A·‐

(5)

The generally employed semiconductor powders like TiO2, CdS, and ZnS consist of 10−100 nm small nanocrystals forming micrometer-sized aggregates. In addition to the standard mechanism, wherein IFET occurs at one unique crystal (Scheme 2, a

Scheme 2. Details of Direct Semiconductor Photocatalysisa

The sphere symbolizes a micrometer large aggregate of semiconductor nanocrystals. Eg, Ec, and Ev correspond to band gap, conduction, and valence band edges, respectively.

feasibility of the IFET can be estimated by comparing the reduction potentials of the two substrates with the semiconductor band edge positions (Figure 1). The primary redox products are

a

Spheres symbolize SC nanocrystals, empty circles solvent and substrate molecules.

path 1), the new reaction path 2 contains an intercrystallite electron transfer (ICET). ICET processes may improve the efficiency of formation of reactive electron−hole pairs. According to Scheme 3 the light generated excitons may undergo relaxation via exciton−phonon coupling (within femtoseconds), geminate recombination, and migration to the surface, where they are dissociatively trapped into unreactive (etr−, htr+) and reactive (er−, hr+) surface sites (Scheme 3, processes 1, 2, 3, respectively). Surface-trapped charges can be described as reduced metal and oxidized anion centers. In competition with charge recombination (processes 4, 5), subsequent IFET reactions generate the primary redox products A•− and D•+ (processes 6 and 7) which may undergo a back-electron transfer (BET) to A and D (process 8) or produce the final products (processes 9, 10). The product quantum yield can be factorized into a product of three eff iciencies (eq 6). The efficiencies of reactive electron− hole pair formation (ηr) of IFET (ηifet), and of product formation

Figure 1. Positions of valence and conduction band edge potentials at pH 7.3

converted to stable reduced and oxidized compounds Bred and Cox, respectively (Type A, eq 1). In some rare cases the primary IFET products may undergo bond formation leading to addition products (Type B, eq 2). Thus, the semiconductor acts a heterogeneous photocatalyst enabling concerted reductive and oxidative IFET, followed by bond formation, analogous to photosynthesis. The state of knowledge in general semiconductor photocatalysis is summarized in a collection of review articles4,5 and in a recent text book.6 hv

A + D → Bred + Cox SC

(1) 1003


Article

Accounts of Chemical Research Scheme 3. Primary Processes at the Solid−Liquid Interfacea

Scheme 4. Visible Light Photofixation of Dinitrogen in the Presence of Ethanol

a

Full and wavy lines correspond to radiative and nonradiative processes, dashed lines to charge trapping, photocorrosion processes are omitted.

This effect is well-known as current-amplif ication in photoelectrochemistry. We proposed the term photoredox amplification for the field of general photoredox chemistry.6 The importance of the effect is nicely demonstrated in the visible light photofixation of dinitrogen to ammonia and nitrate, catalyzed by titanate thin films (Scheme 4). Product formation is observed only when reducing agent and film have photoredox and current amplification properties, respectively.10−14 In addition to ethanol also the ubiquitous humic acids are efficient reducing agents. Since iron titanates may be formed by solar weathering of ilmenite minerals, a nonenzymatic visible-light nitrogen fixation appears possible under natural conditions. It is known that rutile containing desert sands enable N2 fixation by UV light.15 A classical case for a reductive IFET coupled to bond breaking is organohalides (eq 16).

from the primary redox products (ηp) (eqs 7−9) wherein ksb corresponds to the rate constant of secondary electron transfer.7 Φp = ηr ηifetηp

(6)

ηr = k 3/(k1 + k 2 + k 3)

(7)

ηifet = k6,7/(k5 + k6,7)

(8)

ηp = k 9,10/(k 8 + ksb + k 9,10)

(9)

1.1. ηr

This efficiency depends on exciton dissociation energy, crystal phase, aggregation, charge diffusion constant, and surface properties.

ArCH 2 X + er − → ArCH 2· + X− 1.3. ηp

1.2. ηifet

From Scheme 3, it follows that primary BET (Scheme 3, process 8) and secondary BET (eqs 17 and 18) compete with product formation (Scheme 3, processes 9 and 10). As discussed above, radical ions, if formed at all, are usually very short-lived, and therefore process 8 seems less important.

The thermodynamic feasibility of the IFET is generally estimated by comparing the substrates reduction potentials with band edge positions. Omitted in such estimations is the reorganization energy that may reach values of half an electronvolt or more.6 Furthermore, the elementary IFET reactions according to eqs 10 and 11 may be coupled to bond breaking and bond formation, leading to a significant change of the driving force. A + er − → A·−

(10)

D + h r + → D•+

(11)

(12)

CH3OH + h r + → ·CH 2 OH + H+

(13)

Quite often the produced radical may inject an electron into the semiconductor conduction band resulting in the formation of two electrons from only one absorbed photon (eqs 14 and 15; see also Scheme 4). h r + + R 2CHOH → R 2C·OH + H+

(14)

R 2C·OH → R 2C = O + H+ + ecb−

(15)

A·− + h tr + → A

(17)

D·+ + er − → D

(18)

More relevant is the reactivity of the radicals generated therefrom or directly from the reactive electron−hole pair (not shown in Scheme 3). In homogeneous solution radicals are known to undergo unspecific conversions to products via disproportionation, addition, and electron transfer reactions. Different from that, a high chemoselectivity can be observed in heterogeneous system of semiconductor photocatalysis (vide infra). From the above discussion, it is obvious that changes of apparent quantum yields, observed when modifying photocatalyst or substrates, in general cannot be straightforwardly correlated with one of the three efficiencies.7 However, this is often done in the literature although even basic experimental information like nature of the light absorbing species is not given. In general, it is also unknown whether emissive and reactive electron−hole pairs are identical.

Classical examples are the oxidation of alcohols and the reduction of organohalides. In the case of methanol, the two-step process requires a potential of about 3.3 eV8 corresponding to the rate determining formation of CH3OH·+ (eq 12). But only 1.1 eV is necessary for the one-step process affording the hydroxymethyl radical and a proton (eq 13).9 CH3OH + h r + → CH3OH·+ → ·CH 2 OH + H+

(16)

2. TYPE A REACTIONS In Type A reactions, two or more substrates are converted into reduced and oxidized products (eq 1). Observed reactions 1004


Article

Accounts of Chemical Research

although water is reduced in the primary reductive step, it is not consumed as experimentally evidenced after production of two liters of H2. Therefore, the formation of D2 from D2O at early reaction stages indicates only an intermediary water reduction. Subsequent reformation of water is evident from the sum of eqs 19−21 affording eq 22. Thus, the formation of D2 from D2O, usually taken as experimental proof for permanent water reduction, is a necessary but not suf ficient criterion. Reformation of water is probably occurring in many so-called “sacrificial” systems for photochemical H2 production.

include cis−trans isomerizations, valence isomerizations, cycloaddition reactions, intramolecular or intermolecular C−N and C−C couplings, partial or exhaustive oxidations, and reductions.16 The nature of products is explainable by Scheme 3. In general, the reductive part is reduction of oxygen or water while the oxidative part is formation of the organic substrate’s radical or radical cation. All these reactions produce well-known organic compounds, which were just identified by analytical methods and only very rarely also isolated. One reason is that photocorrosion of the semiconductor quite often prevents its application in organic synthesis. This applies for colloidal metal sulfide, which are photochemically too unstable for synthetic reactions.17−19 In the following we shall discuss C−C and C−N couplings in the presence of metal sulfide powders leading to previously unknown compounds on a preparative scale. They all are examples of direct semiconductor photocatalysis, i.e. light is absorbed by the photocatalyst. Different from homogeneous photoredox catalysis no sacrificial agents are required for photocatalyst reformation subsequent to the electron transfer activation of the substrate (vide supra). Furthermore, product isolation is much easier than in homogeneous photoredox systems since the photocatalyst powder can be easily removed by filtration.

2RH + 2h r + → 2R· + 2H+ −

(19) −

2D2 O + 2er → D2 + 2OD ·

(20)

2R → R−R

(21)

2RH = H 2 + R−R

(22)

An experimental differentiation between emitting and reactive electron−hole pairs was feasible by emission quenching under the same experimental conditions as in the photoreaction and by reaction inhibition studies (Figure 2). Excitation of an aqueous

2.1. Intermolecular C−C Coupling through Dehydrodimerization

When an aqueous 2,5-dihydrofuran (2,5-DHF) suspension of ZnS or platinized CdS (Pt/CdS) is irradiated with UV or visible light, liters of hydrogen and gram-amounts of hitherto unknown dehydrodimers are isolated in yields of 60% (Scheme 5).17,20−23 Scheme 5. Anaerobic Dehydrodimerization of 2,5Dihydrofuran in D2O

Figure 2. Emission spectrum of an aqueous ZnS suspension in absence and presence of zinc sulfate, λexc = 300 nm.

suspension of ZnS affords an emission spectrum exhibiting broad peaks from band-to-band (at 366 nm, hν1 in Scheme 5) and band-to-surface state transitions (at 430 nm, hν2 in Scheme 5). It corresponds to the so-called self-activated emission, which originates from surface zinc ions. In the presence of zinc or cadmium sulfate the intensity of both bands slightly increases or does not change. the two emission bands. Also the substrate 2,5-DHF has no significant influence. Contrary to emission, product formation is inhibited strongly when cadmium or zinc salts are added. This indicates that emitting and reacting electron−hole pairs are different. Only traces of elemental zinc can be found, excluding an electron transfer inhibition. Differently, cadmium ions exert a stronger effect under concomitant reduction to the metal. In both cases, a Stern−Volmer plot of the relative H2 evolution rate as a function of concentration of adsorbed scavenger ionssurprisingly affords a straight line, as expected for a diffusion-controlled inhibition process (Figure 3). This suggests that the scavenger ions move in a pseudohomogeneous solvent−solute−surface layer.17 From Figure 4, one can conclude that the reduction of water is strongly exergonic. Contrary, the oxidation of 2,5-DHF to the radical cation is endergonic by at least 0.6 eV, taking a reduction potential of 1.6−2.0 V for the reactive hole and 2.6 V for the oxidation potential of 2,5-DHF.24 Contrary to that, the concerted process of a dissociative electron transfer (eq 23) is exergonic by at

When H2O is replaced by D2O, at low conversion the gas phase contains 90% of D2. Colloidal ZnS or CdS and high purity single crystals do not catalyze the reaction. This anaerobic dehydrodimerization constitutes the very first example of preparing a hitherto unknown compound by semiconductor photocatalysis. The expected dehydrodimers were also isolated from 3,4-dihydropyran, 3-methyl-2,3-dihydropyran, and cyclohexene in yields of 30−60%. Scheme 5 displays the proposed mechanism. In the reductive IFET, water is reduced to hydrogen while in the oxidative step 2,5-DHF is transformed into to an allylic dihydrofuryl radical and a proton. Statistical dimerization of the radicals leads to the products. At high conversion, corresponding to a production of one liter of hydrogen, the D2 part of the gas phase is lowered from 90% to 40%, while the sum of HD and H2 increases from 10% to 60%. This change in the isotopic composition is in accord with the mechanism proposed in Scheme 5. Accordingly, the concentration of HOD, formed as the sum of the two IFET reactions, is expected to increase with increased conversion. It is noted, that 1005


Article


case for semiconductor photocatalysis, irrespective of many claims. However, in the following we show that the unique properties of an excited semiconductor surface enable novel C−N− and C−C− coupling reactions of preparative value. Simple, self-prepared CdS is the visible light active photocatalyst. With these powders optimal rates are obtained only in the presence of 2−5% of water.26 3.1. C−N Coupling

Addition of azobenzene to a running CdS catalyzed photodehydrodimerization experiment of 2,5-DHF, results in a complete inhibition of hydrogen evolution. Instead, the novel allylhydrazine 2c, a linear addition product of 2,5-DHF to azobenzene, and small amounts of hydrazobenzene (3) are formed (Scheme 6).18,19,27 No reaction occurs in dry n-hexane or THF. Scheme 6. Addition of Cyclic Allyl/Enol Ethers and Olefins to 1,2-Diazenes

Figure 3. Dependence of relative product formation rate as a function of adsorbed inhibitor concentration.

However, when H2O or MeOH is added, the reaction becomes as fast as in pure MeOH. Using colloidal CdS instead of the powder, only photocorrosion is observable. The novel allylhydrazines 2a−f were obtained in gram amounts. Due to the poor crystallization properties, isolated yields reach only 10−40% whereas HPLC yields are about twice larger. Astonishingly, only a few allylhydrazines have been reported in the previous literature. In the proposed mechanism the oxidative IFET is the same as for the dehydrodimerization, whereas the reductive IFET consists of a proton-coupled reduction of the diazene to a hydrazyl radical. Heterocoupling of the hydrazyl with the allyl radical leads to the allylhydrazine (Scheme 7, path B). The hydrazobenzene derivative 3 is formed by consecutive disproportionation or reduction of the hydrazyl radical (eqs 24 and 25).

Figure 4. Some band edge and substrate reduction potential positions. (a) Single crystal, (b) self-prepared powder, and (c) RH = 2,5-DHF.

least 0.8 eV as estimated from the difference between the free enthalpy of C−H bond dissociation energy (3.22 eV) and the potential of the hydrogen electrode (−2.40 V in H2O). RH → R· + H+ + e−

(23)

The primary events at the semiconductor surface are schematically depicted in Scheme 5. The light generated electron−hole pair has a lifetime of 0.1−20 ns and either recombines through radiative or nonradiative processes or is trapped at emitting (etr−, htr+) or reacting surface (er−, hr+) sites. Surprisingly, the allylic radicals generated in the proton-coupled IFET almost exclusively undergo dimerization as evidenced by a complete material balance. This unexpected high chemoselectivity suggests that C−C coupling does not occur in homogeneous solution but in a H2O−2,5-DHF−surface layer. Although the saturated ether THF reacts only 10 times slower than the unsaturated one, no THF dehydrodimers or crossproducts are detected when THF is present in 10-fold excess. However, both THF and 2,5-DHF dehydrodimers and crossproducts are formed when in homogeneous solution the corresponding radicals are generated through hydrogen peroxide photolysis. In the ZnS photocatalyzed reaction these products are observable only at a five hundred-fold excess of THF. Thus, the solid−liquid interface induces an unexpected chemoselectivity of radical C−C coupling.

2ArN·−N(H)Ar → ArN(H)−N(H)Ar + ArNNAr (24)

Scheme 7. Addition of Cyclic Allyl/Enol Ethers or Olefins (RH) to 1,2-Diazenes Photocatalyzed by CdS or ZnS

3. TYPE B REACTIONS Organic synthesis via visible light induced molecular photoredox catalysis has recently received high attention.25 This is not the 1006


Article

Accounts of Chemical Research ArN(H)−NAr + er − + H+ → ArN(H)−N(H)Ar

Scheme 8. Addition of Unsaturated Ethers and Olefins to Trisubstituted Imines

(25)

The postulated C−N heterocoupling requires diffusion of the two radicals either in the solvent−solute−surface layer or in the bulk solution. In both cases, one expects that the reaction rate should decrease upon increasing solvent viscosity through the application of high pressure. Both the formation rates of 2c and 3 (R1 = R2 = Ph) decrease with increasing pressure.27 From a plot of ln(rate) vs pressure activation, volumes ΔV⧧ are obtained as +17.4 ± 3.4 and +15.8 ± 2.3 cm3 mol−1 for 2c and 3, respectively, (Figure 5). These positive values disfavor a radical C−N bond

tungsten halogen lamp. When all of the imine is consumed, CdS is removed by filtration and methanol by evaporation in vacuo. The remaining white powder of 5d (X = Ar, Ph) is recrystallized from n-heptane. Yield: 1.64 g (72%). Disubstituted imines (6a−d) in addition to the homoallylamine (7a−d) produce also the hydrodimer (8a−d) of the imine, i.e., the dimer of an expected α-aminobenzyl radical (Schemes 9 and 10). Formation of the homoallylamine can be Scheme 9. Addition of Cyclopentene to Disubstituted Imines

Figure 5. Pressure dependence of the formation rates of 2c and 3.27

formation as rate-determining step, for which a negative activation volume is expected. The only exception is radical recombination in the termination step of polymerizations.28 Their ΔV⧧ values are in the range of +13 to +25 cm3 mol−1 originating from a large positive contribution of diffusion and a small and negative part of radical C−C coupling. Therefore, the activation volume obtained for the formation of 2c may primarily stem from diffusion of the radicals, whereas the C−N coupling step is of second importance. Accordingly, ΔV⧧ should be comparable with the activation volume of the viscous flow of methanol. Since the latter value of +8 cm3 mol−1 is significantly smaller,29 it reasonable assuming that the radicals do not diffuse in the bulk homogeneous solution but in the solvent−solute−surface layer. It further suggests that reactive charges are not located at the same crystal but more likely at distant ones, generated by an ICET within the aggregate (vide supra, Scheme 2, reaction path 2). This basic question of the distance of the reactive charges certainly deserves further research.

Scheme 10. CdS Photocatalyzed Addition of Cyclopentene to Imines

3.2. C−C Coupling

3.2.1. Addition of Imines to Cyclic Olefins and Unsaturated Ethers. From the mechanism proposed above, it is expected that other substrates suited for photoinduced oneelectron reduction may undergo similar C−C couplings. Indeed, changing from the 1,2-diazene to an aromatic imine, analogous addition reactions occur.26,27,30−36 The trisubstituted imines 4 lead to novel homoallylamines 5a−g in isolated yields of 30−75% (Scheme 8). In a typical preparation, 0.30 g of selfprepared CdS is suspended in 200 mL of MeOH containing 1.55 g of 4 (X = Ar, Ph) and 37.0 mL of α-pinene. After transferring to a Solidex immersion lamp apparatus and subsequent sonication under Ar bubbling, the suspension is irradiated with a

explained by C−C heterocoupling of the α-aminodiphenylmethyl radical produced in the reductive IFET with the allylic radical generated in the oxidative IFET. In no cases products could be found, which are expected from a C−N heterocoupling. 1007


Article

Accounts of Chemical Research Scheme 11. Addition of Cyclic Olefins to N-Adamantylimines

Thus, different from thermal synthesis routes, which usually involve the use of organometallic reagents, the reaction is regioselective and much easier to perform. To find out how steric pressure at the imine nitrogen atom may influences the C−C coupling step, the aryl substituent Ar2 was replaced by the bulky 1-adamantyl group. The novel homoallyladamantylamines 10a−c were obtained in isolated yields of 21−85% when using cyclopentene, cyclohexene, α-pinene, and various N-adamantylimines (Scheme 11).37 Unsaturated adamantylamines are of pharmaceutical interest since they may show antibacterial, antitumor, antipyretic, and anti-inflammatory effects. Some of them were discussed as promising candidates for the treatment of Alzheimer’s and Parkinson’s diseases. In the absence of olefins, RH hydrodimers of 12 are also produced, although the rate is lowered by 90%. This suggests that the solvent is involved in the oxidative step. As expected, the corresponding addition products 13 and hydrodimers 14 are obtained when using various alcohols (Scheme 12).

Scheme 13. Support-Controlled Chemoselectivity

in Scheme 10. In the reductive part, the proton-coupled IFET generates an α-aminobenzyl radical. It is noted that in homogeneous chemistry αamino radicals are preferentially formed by oxidation. When surface SH and OH groups of CdS are alkylated with 3-bromopropyltrimethoxysilane, the resulting powder is completely inactive. But it becomes active again, when the iminium salt instead of the imine is employed. That indicates that the surface groups protonate the imine to the easier reducible iminium salt.33 Whereas the diastereoselectivity of C−C heterocoupling is low, the homocoupling of two α-aminobenzyl radicals is diastereospecific as exemplified by the hydrodimerization of the p-chlorophenyl derivative 15 (X = Cl). According to HPLC and X-ray structural analysis only one diastereomer of 18 (X = Cl) is formed in the reactions with cyclopentene and α-pinene, whereas cyclohexene induces formation of another diastereomer. Therefore, the stereochemistry of homocoupling is controlled by the nature of the olefin, although it is not directly involved in the reaction. This surprising result becomes explainable when recalling that the intermediate radicals diffuse within a solvent− solute−surface layer consisting inter alia of olefins adsorbed to CdS via hydrogen bonding with surface SH and OH groups. Steric interaction with the olefin may occur during this diffusion process. Thus, the olefin plays a dual role being substrate for the addition and stereodirecting spectator for the hydrodimerization reaction.39,40 Preliminary results with chiral substrates and chirally modified CdS surfaces suggest promising future developments for chiral semiconductor photocatalysis.40 These photoredox induced radical addition reactions seem to be of more general applicability as also implied by the recently reported ZnS photocatalyzed addition of carbon dioxide to acetylacetonate or 2,3-dihydrofuran.41,42 3.2.2. Addition of SO2 and O2 to Alkanes (Photosulfoxidation). The activation of alkanes represents a great challenge

Scheme 12. Addition of Alcohols to Imines

It is mentioned that in the presence of an olefin no alcohol addition product was observable, even in a 500fold molar excess of the alcohol. This impressively demonstrates the high chemoselectivity of the semiconductor-liquid interface. To further unravel the generality of these linear addition reactions for the synthesis of valuable organic compounds, the N-aryl substituent in the imines 4 (X = CN) was functionalized by an N-benzoyl group (Bz), which may be easily converted to an amino group. The thus obtained unsaturated amino acids could be of pharmaceutical importance.38 Unexpectedly, with CdS grafted silica (CdS-O-SiO2) no addition reactions of 15 to cyclopentene and cyclohexene are observable. Instead, a new thermal transhydrocyanation of the imine component produces the hitherto unknown malononitriles 17 in isolated yields of 40−50% (Scheme 13). Surprisingly, CdS, ZnS, or CdS grafted zinc sulfide (CdS-S-ZnS) inhibit this dark reaction and photocatalyze the formation of the novel addition products 16 (65−85% isolated yield). Commercial CdS and ZnS are inactive, most likely due to traces of corresponding metal oxides, which inhibit the photocatalytic activity. From the experimental results summarized above, a mechanism analogous to the C−N coupling reaction is proposed 1008


Article


extending the two-substrate addition to a three-substrate addition scheme.

in general chemistry. The well-known UV sulfoxidation of liquid alkanes by sulfur dioxide and oxygen represents the very rare example of an industrially applied process (eq 26). In general, C16−20 alkanes are employed and the resulting linear sulfonic acids are excellent biodegradable surfactants. R−H + SO2 + 1/2O2 + h□ → RSO3H

4. SUMMARY Different from molecular photosensitizers, which enable only one one-electron transfer with one single substrate, photoexcited semiconductors induce two concerted one-electron transfer reactions with two substrates. This is because the light generated electron−hole pairs are trapped at distinct surface sites and undergo interfacial electron transfers (IFET) with unsaturated donor and acceptor substrates. In most cases, intermediate radicals are formed by proton-coupled IFET. The radicals diffuse in a solid-solute−surface layer to undergo chemo- and stereoselective C−C and C−N bond formation. Thus, the semiconductor photocatalyst functions like an artificial leaf. The high photoredox activity of simple oxidic and sulfidic semiconductor powders is also exemplified by the visible light sulfoxidation of alkanes to alkanesulfonic acids and by the fixation of dinitrogen. Since several minerals are known to have semiconductor properties, solar photocatalysis may be also relevant for prebiotic and environmental chemistry.

(26)

Scheme 14 summarizes the general accepted mechanism. Accordingly, the alkane activation step consists of hydrogen Scheme 14. UV Photosulfoxidation of Alkanes

■

abstraction by triplet-excited sulfur dioxide. Subsequent addition of SO2 and hydrogen abstraction produces another alkyl starter radical and the persulfonic acid. Fragmentation and hydrogen abstraction (eqs 27,28) lead to the alkanesulfonic acid.43 As expected, the overall reaction is a photoinduced radical chain reaction and therefore product formation continues even when the light is turned off. RSO2 −O−O−H → RSO2 −O· + OH· ·

·

RSO2 −O + R−H → RSO3H + R

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Horst Kisch: 0000-0003-1042-7967 Notes

(27)

The author declares no competing financial interest.

(28)

Biography

Surprisingly, in the presence of titania, the photosulfoxidation proceeds also with visible light (λ ≥ 400 nm).44,45 This is because a yellow surface CT complex is formed with SO2, exhibiting a broad absorption band at 410−420 nm. Vis excitation according eqs 3 and 4 generates a reactive conduction band electron TiO2(er−) and an adsorbed sulfur dioxide radical cation (Scheme 15). The latter oxidizes the alkane under deprotonation to an alkyl radical. In the case of n-hexane a reduction potential of about 1.8 V is estimated for this reaction step (Scheme 15, reaction path 1). Alkyl radicals can be also generated through hydrogen abstraction by hydroperoxyl radicals, formed via the reduction of O2 by TiO2(er−) and subsequent protonation by water or surface OH groups (Scheme 15, reaction path 2). In summary, this novel visible light induced C(sp3)−H activation can be classified as semiconductor photocatalysis Type B,

Horst Kisch, born in 1942, studied chemistry at the Universität Wien, Austria, where he received his Ph.D. in 1969. From 1968 to 1984, he worked at the Max-Planck-Institut für Strahlenchemie (now MaxPlanck-Institut für Chemische Energiekonversion) in Mülheim a.d. Ruhr, Germany. In 1977 he completed his “Habilitation” in Organic Chemistry at the University of Dortmund, Germany, and became Professor of Inorganic Chemistry at the University of ErlangenNürnberg, Germany, in 1984. He retired in 2008.

■

REFERENCES

(1) For examples in organic synthesis, see ref 2. (2) Prier, C. K.; Rankic, D. A.; MacMillan, D. W. C. Visible Light Photoredox Catalysis with Transition Metal Complexes: Applications in Organic Synthesis. Chem. Rev. 2013, 113 (7), 5322−5363. (3) Kisch, H. Semiconductor Photocatalysis-Mechanistic and Synthetic Aspects. Angew. Chem., Int. Ed. 2013, 52 (3), 812−847. (4) Dionysiou, D. D., Li Puma, G., Ye, J., Schneider, J., Bahnemann, D., Eds. Photocatalysis: Fundamentals and Perspectives.; Royal Society of Chemistry: Cambridge, UK, 2016. (5) Dionysiou, D. D., Li Puma, G., Ye, J., Schneider, J., Bahnemann, D., Eds. Photocatalysis: Applications; Royal Society of Chemistry: Cambridge, UK, 2016. (6) Kisch, H. Semiconductor Photocatalysis; Wiley-VCH: Weinheim, Germany, 2015. (7) Kisch, H.; Bahnemann, D. Best Practice in Photocatalysis: Comparing Rates or Apparent Quantum Yields? J. Phys. Chem. Lett. 2015, 6 (10), 1907−1910. (8) Bard, A. J.; H, L. Encyclopedia of Electrochemistry of the Elements, Organic Section; M. Dekker: New York, 1987; Vol. XI. (9) Lilie, J.; Beck, G.; Henglein, A. Pulse radiolysis and polarography. Halfwave potentials for the oxidation and reduction of short-lived

Scheme 15. Proposed Mechanism for Visible Light Generation of Alkyl Radicals, R = n-Heptyl, 1-Adamantyl

1009


Article

Accounts of Chemical Research organic radicals at the mercury electrode. Ber. Bunsenges. Phys. Chem. 1971, 75 (5), 458−65. (10) Rusina, O.; Eremenko, A.; Frank, G.; Strunk, H. P.; Kisch, H. Nitrogen photofixation at nanostructured iron titanate films. Angew. Chem., Int. Ed. 2001, 40 (21), 3993−3995. (11) Rusina, O.; Linnik, O.; Eremenko, A.; Kisch, H. Nitrogen photofixation on nanostructured iron titanate films. Chem. - Eur. J. 2003, 9 (2), 561−565. (12) Rusina, O.; Linnik, O.; Smirnova, N.; Eremenko, G.; Kisch, H. Photocatalytic activity of nanocrystalline Fe/TiO2 films irradiated by UV and visible light. Khim., Fiz. Tekhnol. Poverkhni 2004, 10, 85−89. (13) Rusina, O.; Macyk, W.; Kisch, H. Photoelectrochemical Properties of a Dinitrogen-Fixing Iron Titanate Thin Film. J. Phys. Chem. B 2005, 109 (21), 10858−10862. (14) Linnik, O.; Kisch, H. On the mechanism of nitrogen photofixation at nanostructured iron titanate films. Photochem. Photobiol. Sci. 2006, 5 (10), 938−942. (15) Schrauzer, G. N.; Strampach, N.; Hui, L. N.; Palmer, M. R.; Salehi, J. Nitrogen photoreduction on desert sands under sterile conditions. Proc. Natl. Acad. Sci. U. S. A. 1983, 80 (12), 3873−6. (16) For a summary, see ref 6, and for a less synthetically oriented review, e.g., Friedmann, D.; Hakki, A.; Kim, H.; Choi, W.; Bahnemann, D. Heterogeneous photocatalytic organic synthesis: state-of-the-art and future perspectives. Green Chem. 2016, 18, 5391−5411. (17) Hoerner, G.; Johne, P.; Kunneth, R.; Twardzik, G.; Roth, H.; Clark, T.; Kisch, H. Heterogeneous photocatalysis, part XIX: semiconductor type A photocatalysis: role of substrate adsorption and the nature of photoreactive surface sites in zinc sulfide catalyzed C-C coupling reactions. Chem. - Eur. J. 1999, 5 (1), 208−217. (18) Kuenneth, R.; Feldmer, C.; Kisch, H. Heterogeneous photocatalysis. 10. Semiconductor-catalyzed photoaddition of cyclic enol ethers to 1,2-diazenes. Angew. Chem. 1992, 104 (8), 1102−3. (19) Kuenneth, R.; Feldmer, C.; Knoch, F.; Kisch, H. Heterogeneous photocatalysis. XIII. Semiconductor-catalyzed photoaddition of olefins and enol ethers to 1,2-diazenes: a new route to allylhydrazines. Chem. Eur. J. 1995, 1 (7), 441−8. (20) Buecheler, J.; Zeug, N.; Kisch, H. Zinc sulfide as catalyst for heterogeneous photoreduction of water. Angew. Chem. 1982, 94 (10), 792−3. (21) Hetterich, W.; Kisch, H. Heterogenous photocatalysis, V. Cadmium-zinc sulfides as catalysts for the photodehydrodimerization of 2,5-dehydrofuran. Chem. Ber. 1988, 121 (1), 15−20. (22) Zeug, N.; Buecheler, J.; Kisch, H. Catalytic formation of hydrogen and carbon-carbon bonds on illuminated zinc sulfide generated from zinc dithiolenes. J. Am. Chem. Soc. 1985, 107 (6), 1459−65. (23) Kuenneth, R.; Twardzik, G.; Emig, G.; Kisch, H. Heterogeneous photocatalysis. XI. Zinc sulfide catalyzed dehydrodimerization of dihydropyrans and cyclohexene. J. Photochem. Photobiol., A 1993, 76 (3), 209−15. (24) For an analogous estimation see e.g., Henglein, A.; Gutierrez, M.; Fischer, C. H. Photochemistry of colloidal metal sulfides. 6. Kinetics of interfacial reactions at zinc sulfide particles. Ber. Bunsen-Ges. Phys. Chem. 1984, 88 (2), 170−5. (25) Photoredox Catalysis in Organic Chemistry. Acc. Chem. Res. 2016, 49 (10), 2059−2370. (26) Schindler, W.; Knoch, F.; Kisch, H. Heterogeneous photocatalysis. Part XIV. Semiconductor-catalyzed photoaddition. γ,δ-Unsaturated amines from cyclopentene and Schiff bases. Chem. Ber. 1996, 129 (8), 925−932. (27) Reinheimer, A.; Van Eldik, R.; Kisch, H. On the Mechanism of Radical C-N Coupling in Type B Semiconductor Photocatalysis: A High-Pressure Study. J. Phys. Chem. B 2000, 104 (5), 1014−1024. (28) Yokawa, M.; Ogo, Y. Photosensitized polymerization of vinyl acetate under high pressure. Makromol. Chem. 1976, 177 (2), 429−36. (29) Isaacs, N. S. Liquid Phase High Pressure Chemistry; John Wiley: Chichester, New York, Brisbane, Toronto, 1981. (30) Keck, H.; Schindler, W.; Knoch, F.; Kisch, H. Heterogeneous photocatalysis. XVI. Type B semiconductor photocatalysis: the synthesis of homoallyl amines by cadmium sulfide-catalyzed linear

photoaddition of olefins and enol/allyl ethers to N-phenylbenzophenoneimine. Chem. - Eur. J. 1997, 3 (10), 1638−1645. (31) Schindler, W.; Kisch, H. Heterogeneous photocatalysis XV. Mechanistic aspects of cadmium sulfide-catalyzed photoaddition of olefins to Schiff bases. J. Photochem. Photobiol., A 1997, 103 (3), 257− 264. (32) Reinheimer, A.; Fernandez, A.; Kisch, H. Type B semiconductor photocatalysis: on the mechanism of the CdS-catalyzed linear photoaddition of 2,5-dihydrofuran to azobenzene. Z. Phys. Chem. (Muenchen, Ger.) 1999, 213 (2), 129−133. (33) Hopfner, M.; Weiss, H.; Meissner, D.; Heinemann, F. W.; Kisch, H. Semiconductor photocatalysis type B: synthesis of unsaturated a-amino esters from imines and olefins photocatalyzed by silicasupported cadmium sulfide. Photochem. Photobiol. Sci. 2002, 1 (9), 696− 703. (34) Kisch, H.; Sakthivel, S.; Janczarek, M.; Mitoraj, D. A Low-Band Gap, Nitrogen-Modified Titania Visible-Light Photocatalyst. J. Phys. Chem. C 2007, 111 (30), 11445−11449. (35) Pehlivanugullari, H. C.; Sumer, E.; Kisch, H. Semiconductor photocatalysis type B: synthesis of unsaturated a-cyano-homoallylamines from imines and olefins photocatalysed by silica- and cellulosesupported cadmium sulphide. Res. Chem. Intermed. 2007, 33 (3−5), 297−309. (36) Gaertner, M.; Ballmann, J.; Damm, C.; Heinemann, F. W.; Kisch, H. Support-controlled chemoselective olefin-imine addition photocatalyzed by cadmium sulfide on a zinc sulfide carrier. Photochem. Photobiol. Sci. 2007, 6 (2), 159−164. (37) Aldemir, M.; Heinemann, F. W.; Kisch, H. Photochemical synthesis of N-adamantylhomoallylamines through addition of cyclic olefins to imines catalyzed by alumina grafted cadmium sulfide. Photochem. Photobiol. Sci. 2012, 11 (6), 908−913. (38) Kollonitsch, J.; Perkins, L. M.; Patchett, A. A.; Doldouras, G. A.; Marburg, S.; Duggan, D. E.; Maycock, A. L.; Aster, S. D. Selective inhibitors of biosynthesis of aminergic neurotransmitters. Nature 1978, 274 (5674), 906−908. (39) Lindner, W., Cadmiumsulfid-katalysierte Photoaddition von Olefinen an Iminie - Stereochemische Aspekte. Ph.D. Thesis, University of Erlangen-Nürnberg, 2002. (40) Kohl, S., Zum Mechansimus der Cadmiumsulfid-katalysierten linearen Photoaddition von Olefinen an Imine. Ph.D. Thesis, University Erlangen-Nürnberg, 2007. (41) Baran, T.; Aresta, M.; Kruczała, K.; Macyk, W.; Dibenedetto, A. Photocatalytic Carboxylation of Organic Substrates with Carbon Dioxide at Zinc Sulfide with Deposited Ruthenium Nanoparticles. ChemPlusChem 2014, 79, 708−715. (42) Aresta, M.; Dibenedetto, A.; Baran, T.; Wojtyla, Sz.; Macyk, W. Solar energy utilization in the direct photocarboxylation of 2,3dihydrofuran. Faraday Discuss. 2015, 183, 413−427. (43) Ramloch, H.; Taeuber, G. Modern processes of large-scale chemistry: sulfoxidation. Chem. Unserer Zeit 1979, 13, 157−62. (44) Parrino, F.; Ramakrishnan, A.; Kisch, H. Semiconductorphotocatalyzed sulfoxidation of alkanes. Angew. Chem., Int. Ed. 2008, 47 (37), 7107−7109. (45) Parrino, F.; Ramakrishnan, A.; Damm, C.; Kisch, H. Visible-LightInduced Sulfoxidation of Alkanes in the Presence of Titania. ChemPlusChem 2012, 77 (8), 713−720.

1010


Semiconductor Photocatalysis for Chemoselective Radical Coupling

Recommend Documents