Introduction

Homopropargylic azides are important building blocks bearing two of the most versatile functional groups, allowing a rich panel of functionalization. They have been used as intermediates in numerous syntheses to access bioactive compounds or materials . In addition, azide reduction affords homopropargylic amines, which can be found in bioactive molecules and have been tested in structure–activity relationship studies (Scheme 1A) . Moreover, transformations have been developed to exploit the two functional groups simultaneously, for example through their intramolecular cyclization to form pyrroles in the presence of transition metal catalysts .

Scheme 1: Overview of homopropargylic azides importance and strategies for azido-alkynylation.

Currently, this motif is synthesized by sequential introduction of the two functional groups . Addition of a lithium acetylide to an epoxide affords the corresponding homopropargylic alcohol which can then undergo a sequence of mesylation and substitution with azide ions to produce the desired homopropargylic azide. However, this approach gives only access to products bearing the alkyne at the least substituted position. To the best of our knowledge, no general strategy has been employed to access the other regioisomer possessing a terminal azide, despite its implication in the synthesis of complex molecules . Therefore, the development of a straightforward reaction to synthesize homopropargylic azides would be of general interest.

The azido-alkynylation of alkenes would allow to generate the desired motif in a single step, greatly increasing the molecular complexity of the starting substrate. Using radical chemistry would lead to a regioselective addition of azide radicals to the alkene, forming selectively the most stabilized C-centered radical. A prominent method for the generation of azide radicals relies on hypervalent iodine reagents . Azidobenziodoxolone, also known as Zhdankin reagent, has often been used under thermal or photochemical conditions to generate the desired azide radical in a controlled fashion. However, recent safety issues arising from the shock and impact sensitivity of the compound led to the development of the azidobenziodazolone scaffold . This class of derivative showed an improved safety profile while retaining the redox properties of the original reagent.

A single example of azido-alkynylation has been reported by Ramasastry and co-workers during a mechanistic study for an azido-hydration reaction . The homopropargylic azide was obtained in only 28% yield using phenyl vinyl ketone. Based on reported aza-alkynylation reactions and modern azidation methods using radical chemistry three approaches could be envisaged. All of them would initially involve the addition of azide radicals to an alkene, generating a carbon-centered radical. Then, different trapping of this intermediate could be performed (Scheme 1B).

First, C-centered radicals are known to recombine with metal-acetylides, in particular copper . Reductive elimination of the organometallic intermediate would lead to the desired product (Scheme 1B, reaction 1). Unfortunately, this approach will not be compatible in the case of azidation since the copper, azides and alkynes present in the mixture are expected to undergo alkyne–azide cycloaddition reactions . Moreover, different azide sources are known to efficiently promote the diazidation of alkenes in the presence of a copper catalyst, often proceeding via a radical mechanism .

A second approach would involve SOMOphilic alkynes to trap the radical by a purely open-shell mechanism (Scheme 1B, reaction 2). Two classes of reagents are commonly used: ethynylbenziodoxolones (EBXs) and alkynylsulfones . A potential limitation of this method lies in the substitution of the transferred alkyne. The efficiency of the radical addition to those reagents is known to be highly dependent on the alkyne substituent. Arylalkynes are expected to perform well but in multiple cases alkyl substituents were reported to afford low yields or no reaction .

Finally, a radical-polar crossover (RPC) approach could be envisaged . Instead of attempting to trap the C-centered radical, further oxidation would generate the corresponding carbocation, which upon reaction with a nucleophilic alkyne would form the product (Scheme 1B, reaction 3). Based on precedence in the literature, this method should allow to transfer efficiently both aryl- and alkyl-substituted alkynes . On the other hand, the nature of the alkene might be limited as it would strongly influence the oxidation potential of the carbon radical and the stability of the resulting carbocation. Recently, we reported the first successful application of an RPC strategy for the azido-alkynylation of styrenes .

Herein, we describe our initial effort towards developing an azido-alkynylation of alkenes using the SOMOphilic alkyne approach instead. Then, the optimization of the RPC strategy will be discussed in detail, giving insights into the different steps of the optimization, which were available only as raw data in our previous work .

Results and Discussion SOMOphilic alkynes

We started to investigate the azido-alkynylation of styrene (1a) using EBX reagent 2 as SOMOphilic alkyne (Table 1). Tosyl-azidobenziodazolone (Ts-ABZ, 3), previously developed by our group , was selected as an azide source. Upon light irradiation, it can release an azide radical by homolysis of the I−N3 bond . We were pleased to see that irradiation of a mixture of styrene (1a), Ph-EBX (2) and Ts-ABZ (3) afforded 17% isolated yield of the desired homopropargylic azide 4a (Table 1, entry 1). Heating the reaction to 80 °C instead of using light to form the radical only afforded traces of the product (Table 1, entry 2). Changing the solvent to DCE slightly increased the yield (Table 1, entry 3). Blue light with an emission spectrum centered around 467 nm was initially selected since Ph-EBX is known to absorb light of lower wavelength, which is expected to cause degradation . Indeed, when the reaction was carried out using 440 nm blue light a lower yield of 10% was obtained and full conversion of the EBX reagent was observed (Table 1, entry 4). Next, we wanted to test different additives in the transformation in an attempt to increase the yield. Addition of acetoxybenziodoxolone (AcOBX) has previously been reported to help initiating the degradation of Ts-ABZ (3) to the azidyl radical . In our case, no difference in yield was observed (Table 1, entry 5). Currently, the reaction is expected to generate a large quantity of iodanyl radical from Ts-ABZ (3) homolysis and from the addition–elimination on Ph-EBX (2). Since no quencher is present in the mixture, we wondered if the accumulation of those radicals could be responsible for the low yields obtained. Addition of (TMS)3SiH, a H• donor, suppressed the reaction by reducing Ph­EBX (2) (Table 1, entry 6). Using diisopropyl ether as a milder donor had no effect on the reaction (Table 1, entry 7). Next, we tested reducing agents expected to react with the iodanyl radical. The addition of ʟ-ascorbic acid led to no improvement of yield and Hantzsch ester suppressed the formation of the desired product (Table 1, entries 8 and 9). Carrying out the reaction in the presence of sodium formate, which can play a dual role of H• donor and reductant , only led to a decreased of Ph-EBX (2) conversion, along with a diminished yield (Table 1, entry 10). The addition of DABCO or TBAI , two additives known to activate azidobenziodoxolone (ABX), afforded complex mixtures with no trace of 4a (Table 1, entry 11). Acids or fluorinated alcohols were tested to activate the different hypervalent iodine reagents. While AcOH, TFA and TFE had no impact on the reaction (Table 1, entry 12), the presence of 1.5 equivalents of HFIP slightly improved the yield (Table 1, entry 13). Increasing the amount of styrene in the reaction had no impact (Table 1, entry 14), highlighting that the issue might come from an inefficient trapping of the C-centered radical by Ph-EBX (2) and not from a sluggish addition of azide radicals to the double bond. We attempted to solve this issue by using Ts-ABZ (3) in excess, which should increase the overall quantity of carbon radicals formed, doing so slightly improved the yield of 4a to 29% with no styrene remaining (Table 1, entry 15).

Table 1: Optimization of the azido-alkynylation using Ph-EBX.

Entry Solvent Additive Equiv variation Yield 4a (%)a 1 CH3CN – – (17) 2b CH3CN – – <5 3 DCE – – 19 4c DCE – – 10 5 DCE AcOBX   18 6 DCE (TMS)3SiH – – 7 DCE (iPr)2O – 16 8 DCE ʟ-ascorbic acid – 16 9 DCE Hantzsch ester – – 10 CH3CN HCOONa – 8 11 DCE DABCO, TBAI additive (1.1 equiv) – 12 DCE AcOH, TFA, TFE – 20 13 DCE HFIP – 24 14 DCE HFIP styrene (2 equiv) 24 15 DCE HFIP Ts-ABZ (1.5 equiv) 29

aNMR yield determined using CH2Br2 as internal standard, yield in parenthesis correspond to the isolated yield. bReaction was heated to 80 °C without light irradiation. cReaction carried out using blue light (440 nm).

Styrene was initially selected as model substrate since the addition of azide radicals generated by ABX was well reported . We wanted to explore different classes of alkenes as the double bond substitution would greatly impact both the azide radical addition and the reactivity of the C-centered radical with Ph-EBX (2). Aliphatic alkenes, enamides, enol ethers and acrylates were tested in the reaction but did not lead to formation of the desired products (Scheme S1, Supporting Information File 1). In almost all cases >70% of the EBX reagent was left after 16 hours of reaction.

Radical-polar crossover

Due to the disappointing results obtained with EBX reagents as SOMOphilic alkynes, we turned our attention to the development of a radical-polar crossover approach using photoredox catalysis. The final results obtained were described in our previous publication , but only the raw data for optimization was given in the form of tables in the Supporting Information. We now provide full insights into the different steps of the optimization process, highlighting the decisions taken and the unexpected results obtained.

Ts-ABZ (3) was kept as the azide radical source since it is known to be reduced by photocatalysts such as Cu(dap)2Cl . This perfectly fits a catalytic cycle involving the reduction of Ts-ABZ (3) followed by oxidation of the carbon radical to form a carbocation and regenerate the ground state catalyst. Styrene (1a) was used as model substrate since the formation of a stabilized carbocation might be required for the reaction to occur. Xu and Molander previously reported the quenching of similar cationic species by alkynyl-BF3K salts. Boronate 5a was therefore selected as nucleophilic alkyne. Gratifyingly, using Cu(dap)2Cl in DCE under blue light irradiation afforded 4a in 17% NMR yield (Table 2, entry 1). The major byproduct formed during the transformation was identified as diazide 6. When a copper photocatalyst is involved, a lot of diazidation can be observed. We assumed it could be caused by the reaction of Ts-ABZ (3) with a non-complexed copper catalyst formed during the transformation . When iridium-based photocatalysts were tested, no product formation or only traces were observed (Table 2, entries 2 and 3). Using Ru(bpy)3Cl2·6H2O afforded 17% of 4a, a similar yield as with Cu(dap)2Cl with a reduced formation of 6 (Table 2, entry 4). In contrast, Ru(bpz)3(PF6)2 did not form the desired product probably due to its too high reduction potential compared to Ts-ABZ (3) (value reported for ABX: E1/2(ABX) = −0.43 V vs SCE) (Table 2, entry 5). In general, organic dyes could not catalyze the transformation except for 4ClDPAIPN which afforded 10% of yield of 4a (Table 2, entries 6–10).

Table 2: Photocatalyst screening.a

Entry Photocatalyst E1/2 [PC•+/PC*]
(V vs SCE) Yield 4a (%)b Yield 6 (%)b 1 Cu(dap)2Cl −1.43c 17 14 2 Ir(ppy)3 −1.88c – 7 3 [Ir(ppy)2(dtbbpy)]PF6 −0.96c <5 7 4 Ru(bpy)3Cl2·6H2O −0.81c 17 6 5 Ru(bpz)3(PF6)2 −0.26c – <5 6 4t-BuCzIPN −1.31c – <5 7 4DPAIPN −1.52d – 6 8 4ClDPAIPN −1.30d 10 8 9 DPZ −1.17c <5 6 10 rose bengal −0.68c – <5

aThe data were already published in the supporting information of ref. except for the yield of 6. bNMR yield determined using CH2Br2 as internal standard. cValue taken from reference . dValue taken from reference .

No correlation between the different redox potentials of the photocatalysts and the yield of the reaction could be established. Ru(bpy)3Cl2·6H2O was selected as the optimal catalyst since it afforded the highest yield and minimized diazide formation.

Next, a solvent screening was performed as it can vastly influence reaction proceeding via a carbocation intermediate. Alkynyltrifluoroborates have a low solubility in chlorinated solvents but are well soluble in acetonitrile. Although this solvent has been used in similar transformation before , in our case only 9% of the desired product was obtained (Table 3, entry 2). A large quantity of product resulting from a Ritter-type reaction between acetonitrile and the carbocation intermediate could be observed by NMR . Other highly polar solvents such as DMF and DMSO did not afford the homopropargylic azide (Table 3, entry 3). While alkynyl-BF3K salts are usually well soluble in acetone, carrying the reaction in this solvent only afforded traces of the product (Table 3, entry 4). Ethyl acetate or ether solvents such as dioxane and THF led to similar or slightly reduced yields (Table 3, entry 5). We were pleased to see that running the reaction in DME afforded 36% of 4a (Table 3, entry 6). A mixture of DME with HFIP, known to stabilize carbocationic intermediates , slightly increased the yield (Table 3, entry 7).

Table 3: Solvent screening.a

Entry Solvent Yield 4a (%)b 1 DCE 17 2 CH3CN 9 3 DMF, DMSO – 4 acetone <5 5 EtOAc, dioxane, THF 8–15 6 DME 36 7 DME/HFIP (9:1) 39

aThe data were already published in the supporting information of ref. . bNMR yield determined using CH2Br2 as internal standard.

DME was selected for further optimization as the increased yield with the addition of HFIP was not significant enough to compensate the downside of having an expensive co-solvent. Next, the stoichiometry of the different reaction components was examined. When styrene (1a) was used as limiting reagent instead of Ts-ABZ (3), a slightly higher yield was observed (Table 4, entries 1 and 2). Increasing the excess of Ts-ABZ (3) from 1.25 to 1.5 equivalents had no impact on the reaction (Table 4, entry 3). Reducing the equivalents of 5a to 1.25 only slightly diminished the yield while using 3 equivalents increased it to 44% (Table 4, entries 4 and 5). Surprisingly, 5a could be used as limiting reagent without impacting the reaction (Table 4, entry 6). Carrying out the azido-alkynylation at low or high photocatalyst loading had no impact (Table 4, entry 7).

Table 4: Equivalent screening.a

Entry Ts-ABZ/1a/5a (equiv) Yield 4a (%)b 1 1/1.5/2 36 2 1.25/1/2 42 3 1.5/1/2 41 4 1.25/1/1.25 39 5 1.25/1/3 44 6 1.5/1.5/1 43 7c 1.25/1/2 39–40

aThe data were already published in the supporting information of ref. . bNMR yield determined using CH2Br2 as internal standard. c1 or 5 mol % of photocatalyst were used.

Considering the robustness of the reaction to fluctuation in stoichiometry, conditions using styrene (1a) as limiting reagent, slight excess of Ts-ABZ (3) and 2 equivalents of 5a were selected, while keeping in mind that further fine-tuning can be done (Table 4, entry 2). They offered the best compromise of yield, waste of materials and reproducibility at this scale. So far, the reactions were carried out for 16 hours as it is typical for photocatalyzed reaction to be slow. We realized that full conversion of the alkene was reached very rapidly. In fact, the reaction could be carried out for 1.5 h with no difference in yield (Table 5, entries 1 and 2). It is only below this reaction time that a difference was observed, 21% of product was still formed after 10 min of reaction (Table 5, entries 3 and 4).

Table 5: Reaction time, light source and concentration screening.a

Entry Time (h) Light source c (M) Yield 4a (%)b 1 16 blue LEDs 467 nm (40 W) 0.05 42 2 1.5 blue LEDs 467 nm (40 W) 0.05 42 3 1 blue LEDs 467 nm (40 W) 0.05 39 4 0.17 blue LEDs 467 nm (40 W) 0.05 21 5 1.5 Kessil 467 nm (44 W) 0.05 42 6 1.5 Kessil 467 nm (22 W) 0.05 42 7 1.5 Kessil 440 nm (45 W) 0.05 41 8 1.5 green LEDs 525 nm (40 W) 0.05 35 9 4 CFL (8 W) 0.05 38 10 1.5 blue LEDs 467 nm (40 W) 0.025 36 11 1.5 blue LEDs 467 nm (40 W) 0.1-0.2 41

aBlue/green LEDs refers to LED strips attached to a crystallization flask. The data were already published in the supporting information of ref. . bNMR yield determined using CH2Br2 as internal standard.

We then turned our attention to the light source used to irradiate the reaction. Initially, a homemade set-up using blue LED strips was used. When it was replaced by a commercially available Kessil lamp of the same wavelength and intensity we observed a similar yield (Table 5, entry 5). Reducing the strength of the irradiation from 44 to 22 W had no impact, similarly using a more energetic wavelength afforded the same yield (Table 5, entries 6 and 7). When lower energy light source such as green LEDs or a compact fluorescent lamp (CFL) were used only a small decrease in yield could be observed, although 4 hours of irradiation were needed to reach full conversion using CFL (Table 5, entries 8 and 9). The concentration, a factor expected to play an important role in a three-component reaction, had surprisingly little influence on the transformation. At lower concentration only a slight decrease of yield was observed, whereas higher concentration led to a similar yield (Table 5, entries 10 and 11).

The source of nucleophilic alkyne was evaluated, changing the counter ion from potassium to tetrabutylammonium (7) reduced the yield to 14% (Scheme 2A). When TMS-alkyne 8 was used, no product formation occurred. In this case, we observed that the initial reaction mixture before light irradiation was colorless. This was surprising, as in all the previous experiments a yellow/orange mixture was obtained due to the presence of the photocatalyst. Further investigation revealed that Ru(bpy)3Cl2·6H2O is not soluble in DME (Scheme 2B). In contrast, when it is in the presence of alkynyl-BF3K it readily dissolves. The addition of a couple of water drops to a suspension of photocatalyst in DME also allowed solubilization. The solubilization due to residual water coming from the alkynyl-BF3K was ruled out by careful drying of 5a. We postulated that an ion exchange between Ru(bpy)3Cl2 and 5a can occur to afford a more soluble photocatalyst with the general formula 9 (Scheme 2C). Interestingly, when the solubilization experiment was carried out with the tetrabutylammonium salt 7 only moderate solubilization occurred (Scheme 2B), which could explain the lower yield previously observed (Scheme 2A). Neutral TMS-alkyne 8 cannot be involved in ion exchange and therefore this could be one of the reasons why no reaction occurred.

Scheme 2: Screening of nucleophilic alkynes and investigation of the photocatalyst solubility. n.o = not observed, PC = photocatalyst. aThe data were already published in the supporting information of ref. .

Next, we turned our attention to the temperature of the reaction, a factor rarely explored in photocatalyzed reaction due to the lack of available set-ups allowing an efficient light irradiation in addition to a proper temperature control. In our case, since the reaction previously showed to be very tolerant to different light intensity we could attempt to cool the reaction. Carrying out the transformation with the reaction vessel placed in an EtOAc bath cooled to 0 °C by an immersion cooler allowed to increase the yield to 50% (Table 6, entries 1 and 2). Decreasing the temperature to −20 °C slightly improved the yield and the mass balance of the reaction (Table 6, entry 3). At −41 °C the reaction time had to be increased to reach full conversion of styrene (1a), and the yield slightly decreased (Table 6, entry 4). As full conversion is still reached rapidly at −20 °C, we were interested to use a more convenient cooling bath made from ice and salt rather than the immersion cooler. We were pleased to see that running the reaction with this method of cooling afforded similar yield (Table 6, entry 5) even with a significantly more opaque bath mixture.

Table 6: Temperature screening.a