## Introduction

Amines are ubiquitously present in nature and industry and play a pivotal role in organic chemistry1. In particular, tertiary amines are of great importance in the field of medicine; in fact, various tertiary amines exhibit biological activity and are utilised as drugs (Fig. 1a)2,3,4. For the development of novel compounds with medicinal properties, regioselective transformations such as late-stage C–H functionalisations of tertiary amines that preserve the structures of the parent amines are highly desired5,6.

To date, the α-C–H functionalisation of tertiary amines has been actively studied7,8,9 using various synthetic methods such as α-lithiation10, directed α-C–H bond activations11,12 and Rh-carbenoid insertions in α-C–H bonds5,13. Unfortunately, the application scope of these methods is intrinsically limited to certain amines and functionalisation. In contrast, oxidative α-C–H functionalisations can be applied, theoretically, to an almost unlimited range of tertiary amines and functionalisations if the regioselectivity of the amine oxidation can be controlled. In fact, numerous catalytic oxidative α-C–H functionalisations with various nucleophiles via iminium cations14,15,16,17 (or α-amino alkyl radicals)18,19,20 have been developed (Fig. 1b); however, these processes generally occur in an α-methyl-selective manner because the typical mechanism of amine oxidation involves a single electron transfer (SET)/deprotonation/SET sequence (Path A, Supplementary Fig. 1), where the selectivity-determining step is the deprotonation of aminium radicals generated via SET from amines, which requires the half-vacant nitrogen p-orbital and the vicinal carbon p-orbital to overlap21,22,23. In other words, the regioselectivity switch from an α-methyl position to another position is difficult. Similarly, the regioselectivity of the amine oxidation cannot be controlled in other reported mechanisms such as hydrogen atom transfer (HAT)/SET (Path B)24,25, SET/HAT (Path C)26,27 and the Polonovski–Potier reaction via amine oxide formation (Path D)28 (Supplementary Fig. 1), partly because α-C–H bonds of tertiary amines often possess nearly similar bond dissociation energies (e.g. 1-methylpiperidine: 92 kcal/mol for methyl vs 91 kcal/mol for methylene)29,30. In particular, the substrate scope for oxidative α-methylene functionalisations is generally limited to N-substituted benzylic amines like tetrahydroisoquinolines (THIQs), N-protected amines and symmetric amines (Fig. 1b)7,8,9,14,15,16,17,18,19,20. Considering the frequent appearance of α-methylene-substituted amines, especially cyclic methylene-substituted ones, in pharmaceutical fields2,3,4,31, the development of an oxidative regioselective α-methylene C–H functionalisation is highly desirable.

Exceptionally, there are rare examples of oxidative α-C–H functionalisations showing regioselectivity to an α-methylene C–H bond: cyanation and oxygenation (Fig. 1c). In the case of cyanation, α-methyl-cyanated products can be transformed into thermodynamically stable α-methylene-cyanated products via azomethine ylides under cyanation conditions32,33,34,35,36. The α-methylene selectivity is dependent on the formation of azomethine ylides from the cyanated amines, which renders this method inapplicable to the other oxidative α-C–H functionalisations. In the case of the α-oxygenation of tertiary amines to produce amides, the selectivity toward α-methylene-oxygenation has been realised using (super)stoichiometric amounts of oxidants37,38,39. In this context, in our previous work, we successfully developed α-methylene-selective oxygenation of tertiary amines catalysed by Al2O3-supported Au nanoparticles;40 however, the reason behind the observed regioselectivity could not be clarified. In addition, some examples of tertiary amine oxidations via β-hydride elimination exhibiting non-α-methyl selectivity (Path E, Supplementary Fig. 1) have been reported41,42,43, but these systems have been hardly utilised for the α-functionalisation of tertiary amines. As a rare example, Beller et al. reported a Ru-catalysed α-methylene alkynylation of tertiary amines via β-hydride elimination44; however, the substrate scope was very narrow, and hydrogenation of the products usually occurred as a side reaction. In addition, during the revision of this paper, Slowing et al. also reported α-methylene-oxygenation reactions and a few limited examples of α-methylene alkynylation of tertiary amines via β-hydride elimination in the presence of an Au nanoparticle catalyst supported on mesoporous silica with pyridyl groups, while the yields (8–16%) and turnover numbers (2–4) were quite low and insufficient for organic synthetic applications, and the detailed reaction mechanism for the amine oxidation was unrevealed45. Quite recently, Schoenebeck and Rovis et al. realised regioselective oxidative α-C–H alkylation of tertiary amines at the more-substituted positions by utilising the Curtin–Hammett principle via reversible and fast HAT catalysis;30 however, in principle, the unique regioselectivity is derived not from the amine oxidation step but from the reaction between amino alkyl radicals and electrophiles, which limits the types of functionalisation to reactions like Giese radical addition, and regioselectivity control between sterically hindered positions (e.g., linear methylene vs cyclic methylene) is quite difficult30. In addition to the aforementioned reports, although a few reports exist on oxidative functionalisation systems showing very limited examples of α-cyclic methylene functionalisations33,46,47,48,49,50,51, a wide range of tertiary amines, including unsymmetrical ones, cannot be utilised in the oxidative α-methylene selective functionalisation reactions developed to date (the previous main reports on oxidative α-methylene C–H functionalisations of tertiary amines except for benzylic amines, N-protected amines and symmetric amines are summarised in Supplementary Table 1). Considering this background, the development of novel general systems for the regioselective α-methylene functionalisation of tertiary amines containing a regioselective amine oxidation step would be an important breakthrough.

Herein, we develop an unusual oxidative regiospecific α-methylene C–H functionalisation of tertiary amines via iminium cation formation with alkynes that produces various propargylic amines, which are widely used in organic synthesis and the pharmaceutical fields52, by utilising a combination of Zn salts and hydroxyapatite-supported Au nanoparticles (Au/HAP) as a catalytic system (Fig. 1d). The present catalytic system is applicable to a variety of aerobic α-methylene alkynylations of tertiary amines and affords propargylic amines, including cyclic derivatives except for THIQs and N-protected amines, which are difficult to synthesise using traditional methods53,54,55,56,57,58,59. Surprisingly, even in the presence of α-methine and linear-α-methylene C–H bonds, cyclic-α-methylene C–H bonds are regiospecifically alkynylated in this catalytic system. Thorough experimental investigations reveal that the unusual α-methylene regiospecificity probably arises from a unique amine oxidation mechanism: a concerted one-proton/two-electron transfer from the amines to O2 on the Au nanoparticle catalyst, which differs from the conventional catalytic amine oxidation mechanisms.

## Results and discussion

### Effects of catalysts

We started out the investigation by screening various catalysts for the α-methylene-selective alkynylation of 1-methylpiperidine (1a) with phenylacetylene (2a) in trifluorotoluene (PhCF3) as a solvent (2 mL) at 95 °C under an O2 atmosphere (Table 1). In the presence of Au/HAP, the desired α-methylene-alkynylated product, i.e. 1-methyl-2-(phenylethynyl)piperidine (3aa), was obtained regiospecifically. Although the corresponding regioisomer 1-(3-phenylprop-2-yn-1-yl)piperidine (4aa, the α-methyl-alkynylated product) was not produced, the yield of 3aa was not satisfactory, presumably because of the low nucleophilicity of the alkynyl species (Table 1, entry 1). The addition of a catalytic amount of ZnBr2 to the reaction solution led to a drastic improvement in the yield of 3aa (Table 1, entry 3). The use of ZnBr2 as the only catalyst did not produce any amount of 3aa (Table 1, entry 2), indicating that ZnBr2 probably functioned as a co-catalyst to promote the nucleophilic addition of alkynes to iminium cations produced in the Au-catalysed amine oxidation56,60. Among the Zn salts examined, ZnBr2 yielded the best results (Table 1, entries 3–5 and Supplementary Table 2, entries 1–5). Other metal co-catalysts known to promote the nucleophilic addition of alkynes, such as CuBr, CuBr2, FeCl3 and Ag(OTf)61,62,63, proved inadequate for the present alkynylation (Table 1, entries 6–9). Redox-active species like copper salts were especially unsuitable, affording 4aa and 1,4-diphenylbuta-1,3-diyne (5a) via a SET process (Table 1, entries 6 and 7). We also investigated various supported Au nanoparticle catalysts (Table 1, entries 10–12 and Supplementary Table 2, entries 6–8). All the catalysts except for Au/layered double hydroxide (LDH) selectively afforded 3aa in moderate-to-high yields, and Au/HAP and Au/ZnO worked comparatively well. Differences in the Au nanoparticle size in the catalysts were revealed by transmission electron microscopy, e.g. the Au nanoparticles in Au/TiO2 were smaller than those in Au/HAP (Supplementary Fig. 2). Moreover, the X-ray photo-electron spectroscopy results reported in Supplementary Fig. 3 suggest that the differences of Au electronic states were not correlated with this α-alkynylation activity. Considering these results, the weak acid–base properties64 or the low surface areas of Au/HAP and Au/ZnO are probably responsible for the efficient alkynylation of amines without side reactions. In the presence of the other noble-metal catalysts supported on Al2O3, no 3aa was produced (Table 1, entries 13–15). Overall, the combined Au/HAP and ZnBr2 system proved the most efficient catalyst for the present unusual α-methylene-specific alkynylation of tertiary amines.

The choice of the reaction solvent was also a key factor; low-polarity solvents like PhCF3 and toluene were suitable (Supplementary Table 3). Under the optimised conditions, the desired product 3aa was obtained in 82% yield (Table 1, entry 16). A hot filtration test (Supplementary Fig. 4) and an inductively coupled plasma atomic emission spectrometry (ICP-AES) analysis of the solution revealed the heterogeneous nature of the Au/HAP catalysis (see details of the leaching test in the Supplementary Information). Au/HAP could be reused over at least two cycles of 24 h without lowering the 3aa yield, although the reaction rate decreased with the number of cycles (Supplementary Table 4 and Supplementary Fig. 5). A characterisation of the used Au/HAP (Supplementary Figs. 69) revealed an increase in the Au nanoparticle size and the attachment of Zn species onto Au/HAP (for details of the reuse test, see the Supplementary Information).

### Substrate scope

We examined the substrate scope of the present regiospecific α-alkynylation of tertiary amines. The combined catalytic system afforded a variety of propargylic amines (3) from tertiary amines (1) and alkynes (2) with α-methylene specificity. The propargylic amines thus obtained were easily isolated by simple column chromatography on silica gel, and the isolated yields are shown in Figs. 2 and 3. First, we conducted the α-alkynylation of 1-methylpiperidine (1a) with various alkynes (2) (Fig. 2), which proceeded with complete α-methylene specificity. As mentioned above, 3aa was efficiently obtained from 1a and 2a in a 78% isolated yield. When 2-, 3- and 4-chloro-substituted phenylacetylenes and the bromo-substituted analogue were used as substrates, the corresponding propargylic amines (3ab3ae) were produced without dehalogenation. The α-methylene-specific alkynylation also proceeded effectively to afford the 3af3ak products using 4-fluoro-, 4-methoxy-, 4-methyl-, 4-trifluoromethyl-, 4-cyano- and 4-nitro-substituted phenylacetylenes. 3-Ethynylthiophene produced the corresponding propargylic amine (3al), whereas 2-ethynylpyridine was inapplicable (3aq). Aliphatic alkynes also proved suitable, even including an enyne and a propargylic alcohol (3am3ao). Unfortunately, silyl alkynes and tosyl alkynes could not be used with this α-alkynylation system (3ap and 3ar).

Next, we evaluated the α-methylene alkynylation of structurally diverse tertiary amines with 2a (Fig. 3), finding that the corresponding single regioisomers were successfully obtained in almost all cases. The α-alkynylation of 1-methylpyrrolidine and 1-methylazepane effectively proceeded in a methylene-specific manner (3ba and 3ca). Surprisingly, when using 4-hydroxymethyl-1-methylpiperidine as the substrate, the trans-α-methylene-alkynylated product was stereoselectively obtained with the alcoholic hydroxy group intact (3da). Moreover, a substrate possessing a tetrahydroisoquinoline skeleton was applicable to the present regiospecific alkynylation at the α-benzylic site (3ea). The use of N-substituted piperidines as substrates produced the corresponding propargylic amines (3fa3ja) in moderate-to-high yields with selectivity toward the cyclic-α-methylene, even in the presence of a linear-α-alkyl group such as ethyl, benzylic or methine. Only in the synthesis of 3fa, the linear-α-methylene-alkynylated product was observed in ~2% gas chromatography (GC) yield (4fa). Notably, the alkynylation of a linear tertiary amine also proceeded in a methylene-specific manner (3ka), and a bulky amine like triethylamine was suitable (3la). On the other hand, α-methine alkynylation of N,N-dimethylcyclohexylamine (1m) was difficult because of the preferential hydrolysis of the corresponding iminium cation compared with nucleophilic addition of 2a (Supplementary Table 5, entry 1). When the amount of 2a or ZnBr2 was increased, the yield of the desired propargylic amine (3ma) was improved due to the promotion of alkyne nucleophilic addition (Supplementary Table 5, entries 2–4). The addition of molecular sieves 4 A (MS-4A) to remove H2O in the reaction drastically increased the 3ma yield (Supplementary Table 5, entries 5–8), and 3ma was obtained in 84% GC yield (64% isolated yield) under the optimised conditions using MS-4A (300 mg) (Fig. 3, Supplementary Table 5, entry 7). The conditions with MS-4A were also suitable for more difficult selective alkynylation reactions of the following several amine substrates. When using 3-chloro-N,N-diethylpropan-1-amine as the substrate, surprisingly, the α-methylene-alkynylated product at the ethyl group (3na) was regiospecifically obtained with the chloro group intact, suggesting that sterically non-hindered and/or electron-rich α-methylene C–H bonds can be selectively alkynylated among linear-α-methylene positions. Even in the presence of a cyclic-α-methine C–H bond with an ester group, cyclic-α-methylene selective alkynylation occurred to produce the corresponding trans-isomer stereoselectively (3oa). Likewise, the cyclic-α-methylene selective alkynylation of nicotine proceeded trans-stereoselectively (3pa), and as mentioned above, 3da was also obtained as the trans-isomer stereoselectively. The trans-stereoselectivity is possibly derived from the attack of the alkynyl species from the opposite side of the catalyst surface to the iminium cation adsorbed on the Au nanoparticles. In addition, albeit in low yields, cyclic-α-methylene selective alkynylation of buflomedil and cloperastine successfully proceeded to afford propargylic amines 3qa and 3ra, demonstrating late-stage functionalisation of medicines possessing tertiary amine structures. Unfortunately, several amine substrates were not applicable to this alkynylation, as summarised in Supplementary Fig. 10. For example, as expected from the high methylene selectivity, N,N-dimethylaniline was not converted to the desired product. The inability of Au nanoparticles to catalyse amide oxidations40 resulted in no reaction when using an amine protected by tert-butoxycarbonyl group. This reaction system was not applicable to the α-alkynylation of either quinuclidine or a secondary amine like piperidine. In the former case, the formation of the corresponding iminium cation would be forbidden by Bredt’s rule, which supports the expected reaction pathway involving the production of iminium cation intermediates.

### Mechanistic studies

Meanwhile, when the alkynylation reaction using Au/HAP and ZnBr2 was conducted under an Ar atmosphere, a small amount of 2a functioned as a hydrogen acceptor to produce an almost equal amount of 3aa and styrene (6a) (Fig. 4c). The same result was observed in the absence of ZnBr2, albeit with lower yields (Supplementary Fig. 18). Considering that Au–H species were not detected in the reaction without 2a under an Ar atmosphere (Fig. 4b), a concerted one-proton/two-electron transfer possibly occurs from adsorbed 1a to adsorbed 2a on the Au nanoparticle catalyst under the conditions described in Fig. 4c. Thus, under an O2 atmosphere, adsorbed O2 species on Au nanoparticles could effectively function as the sole oxidant of an irreversible concerted one-proton/two-electron transfer from adsorbed 1a to form iminium cations and an Au–OOH species, since the yield of 3aa under an O2 atmosphere was considerably higher than that under an Ar atmosphere (57 vs. 6% yield, respectively) and 6a was not detected at all during the α-alkynylation under an O2 atmosphere.

To gain more insight into this alkynylation, we performed a kinetic analysis of the reaction between 1a and 2a to produce 3aa in the presence of Au/HAP (100 mg) and ZnBr2 (0.05 mmol) (Supplementary Figs. 1921). Although the initial production rate of 3aa under air was independent of the initial concentration of 1a or 2a, it was clearly dependent on the partial pressure of O2 (~0.6th order) (Fig. 4d). These results indicate that the turnover-limiting step is a reaction involving O2 rather than amine adsorption or alkyne addition. Moreover, we examined the kinetic isotope effects (KIEs) using 1-methylpiperidine-2,2,2,2-d4 (1a-d4) or phenylacetylene-d (2a-d). The intermolecular KIE between 1a and 1a-d4 based on independently determined production rates was large (kH/kD = 3.6), revealing the α-methylene C–H bond cleavage of 1a as the turnover-limiting step (Fig. 4e and Supplementary Fig. 22)65. The kinetic results show that the turnover-limiting step is the amine oxidation with O2 on the Au nanoparticles, which is an evidence of the concerted one-proton/two-electron transfer from adsorbed amines to adsorbed O2. Although one-proton/two-electron (hydride) transfer from tertiary amines to stoichiometric oxidants has been reported to date58,66,67, to our knowledge, a concerted one-proton/two-electron transfer from tertiary amines to O2 via a catalyst exhibiting α-methylene selectivity is hitherto unknown. By contrast, when 2a-d was used as the substrate instead of 2a, the production rate of 3aa decreased a little (kH/kD = 1.3), although GC-MS patterns revealed the quick H/D exchange of 2a-d (estimated deuteration ratio as of 10 min: ~17%) (Supplementary Figs. 23, 24). Considering the aforementioned kinetic analysis and the quick H/D exchange, this decrease in the 3aa production rate was not probably derived from KIE of 2a C–H cleavage. On the other hand, surprisingly, 3aa was deuterated in ~62% as of 10 min after the reaction started using 2a-d without any deuteration of 1a (Supplementary Figs. 23, 25, 26). After the reaction for 24 h, the product was selectively isolated via column chromatography as 1-methyl-2-(phenylethynyl)piperidine-3,3-d2 (3aa-β-d2) (deuteration ratio: 13%). Thus, these results indicated that enamines formed from iminium cations accepted deuteron instead of proton to convert into deuterated iminium cations during the reaction (Supplementary Fig. 27). This is an evidence of iminium cation presence in the catalytic system, and irreversible concerted one-proton/two-electron transfer to O2 form iminium cations was strongly supported by the regiospecifically β-deuterated 3aa formation. Moreover, azomethine ylide formation34,35,68 and isomerization of iminium cations69 were excluded in this catalytic system. The deuteration ratio decrease of 3aa after the reaction for 24 h compared with the initially formed 3aa in the reaction is probably derived from the increase of proton source (e.g. H2O) as the oxidation reaction proceeds. In other words, the addition of D2O is assumed to give deuterated propargylic amines from non-deuterated amines and alkynes. In fact, in the presence of a large amount of D2O (3 mmol, three equivalents to 1a), selectively β-deuterated propargylic amine 3aa-β-d2 (deuteration ratio: 76%) was successfully synthesised in 44% yield (Supplementary Fig. 28), which will be beneficial in the synthesis of deuterated medicines70. To reveal the reason for the decrease of 3aa yield in the presence of D2O, either H2O or D2O was added to the present alkynylation, revealing the inhibition effect of water on 3aa production, possibly because of water adsorption on Au nanoparticles (Supplementary Fig. 29). Furthermore, the $${k}_{{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}}$$/$${k}_{{{{{{{\rm{D}}}}}}}_{2}{{{{{\rm{O}}}}}}}$$ = 1.5 was observed without much deuteration of 2a (~6–9%) and almost equivalent to that using 2a or 2a-d (kH/kD = 1.3). Thus, the regeneration step of Au–OOH species to Au species might be affected by the deuteration of 2a or the addition of D2O.

In addition, the intramolecular KIE using 1a-d2 was determined by 1H NMR (kH/kD = 3.3) to be almost equal to the intermolecular KIE (Supplementary Fig. 30), indicating that the present regioselectivity is determined by kinetic control at the concerted one-proton/two-electron transfer to O2 rather than by the thermodynamic stability of the iminium cation intermediates. At the transition state of the concerted one-proton/two-electron transfer, an iminium cation-like species is probably formed on the Au nanoparticles. Thus, in some cases like the α-alkynylation of 1a and 1-ethylpiperidine (1f), the regioselectivity is correlated to the stability differences of the corresponding iminium cations; however, the selectivity of other cases, such as the reaction of 1-cyclohexylpiperidine (1g) or 3-chloro-N,N-diethylpropan-1-amine (1n), is different from the stability order because of kinetic factors like steric effects and electronic effects (Supplementary Fig. 31), which probably leads to the unusual α-methylene selectivity.

According to the results, we propose the mechanism illustrated in Fig. 4f for this unusual α-methylene-specific alkynylation catalysed by the combination of Au/HAP and ZnBr2. First, a tertiary amine coordinates an Au nanoparticle and alkyne coordinates to a Zn species via deprotonation. Subsequently, O2 is adsorbed on the Au nanoparticle, followed by a concerted one-proton/two-electron transfer to produce the corresponding iminium cation and an Au–OOH species. This concerted amine oxidation step is both the selectivity-determining step and the turnover-limiting step. Then, the desired propargylic amine is produced via the nucleophilic addition of the alkyne to the iminium cation. The trans-stereoselectivity to alkynylation of substituted cyclic tertiary amines is thought to be derived from this step, possibly due to the adsorption of the iminium cations on the Au nanoparticle. In fact, DFT calculation results using an Au20 cluster model and the corresponding iminium cation of 1a indicated that the iminium cation is adsorbed and stabilised on the Au nanoparticle (Supplementary Fig. 32). In addition, when the oxidation of 1a in the absence of 2a was carried out using Au/HAP with/without ZnBr2, the conversion of 1a to the corresponding iminium cation, enamine and amide was quite low; however, the presence of 2a drastically improved the conversion of 1a, especially with ZnBr2 (Supplementary Table 6), suggesting that the nucleophilic addition of 2a removed the iminium cations (and enamines) adsorbed on Au nanoparticles to increase the turnover number of amine oxidation. Thus, the nucleophilic addition of the alkyne promoted by the Zn species to the iminium cation adsorbed on the Au nanoparticle presumably affords the desired propargylic amine in this step. Finally, the Au–OOH species accepts a proton to afford Au, H2O and O2, closing the catalytic cycle.

In summary, the combined catalyst system comprising Au/HAP and ZnBr2 promotes an unusual α-methylene-specific alkynylation of tertiary amines, which probably proceeds via an irreversible concerted one-proton/two-electron transfer from amines to O2 on the Au nanoparticle catalyst, differently from conventional amine oxidation mechanisms. In the presence of α-methine and linear-α-methylene C–H bonds, cyclic-α-methylene C–H bonds were regiospecifically alkynylated, and trans-stereoselective alkynylation was also observed in the case of substituted cyclic tertiary amines. Most of the as-produced propargylic amines exhibit unprecedented structures, including alkynylated buflomedil and cloperastine, demonstrating the great utility of this transformation to synthesise unexplored drug candidates and synthetic building blocks. This report provides a general example of α-methylene-specific oxidative C–H alkynylation of tertiary amines, which paves the way for other various α-methylene-specific functionalisations. We believe that this transformation will have a great influence on precise organic synthesis based on amine moieties and bridge the gap between heterogeneous catalysis and the development of novel reactions in organic chemistry.

## Methods

### Preparation of Au/HAP

Au nanoparticles supported on HAP (Au/HAP) were prepared as follows: HAP (2.0 g) was added to an aqueous solution of HAuCl4 (2 mM, 100 mL). After vigorously stirring the mixture for 2 min, aqueous NH3 (10%, 240 μL) was added, and the resulting mixture was stirred at room temperature for 14 h. The resulting slurry was filtered, washed with deionised water (1 L) and dried at room temperature in vacuo to give the HAP-supported Au precursor. The resulting species was dispersed in deionised water (100 mL) and treated with NaBH4 (80 mg) at room temperature for 1 h. The mixture was then filtered, and the residue was washed with deionised water (1 L) and dried to afford the Au/HAP catalyst as a reddish-purple powder (Au content as determined by ICP-AES: 1.5 wt%). Following the same procedure using ZnO, Al2O3, ZrO2, CeO2, TiO2 or LDH as the support instead of HAP, Au/ZnO (Au content: 1.5 wt%), Au/Al2O3 (Au content: 1.8 wt%), Au/ZrO2 (Au content: 1.3 wt%), Au/CeO2 (Au content: 1.6 wt%), Au/TiO2 (Au content: 1.6 wt%) and Au/LDH (Au content: 1.4 wt%) were prepared.

### Catalytic reaction

The catalytic reaction under the optimised conditions was typically performed according to the following procedure: ZnBr2 (10 mol%), Au/HAP (Au: 1.5 mol%), biphenyl (0.1 mmol, internal standard), 1-methylpiperidine (1a, 0.5 mmol), phenylacetylene (2a, 0.5 mmol), toluene (2 mL) and a Teflon-coated magnetic stir bar were successively placed into a Pyrex glass reactor (volume: ~20 mL). After purging with O2 for 1 min, the mixture was stirred at 95 °C under a closed O2 atmosphere (1 atm). The yields of the products were determined by GC analysis using biphenyl as an internal standard. With respect to the isolation of products, after the reaction, the catalyst was removed by simple filtration, and the filtrate was concentrated by evaporation of toluene. The crude product was subjected to column chromatography on silica gel (typically using hexane/ethyl acetate = 6/4 or 8/2 as eluent) to produce the pure propargylic amines (respective eluents were shown in spectral data of Supplementary Information). The products were identified by GC-MS, NMR (1H, 13C and 19F) and elemental analysis of C, H and N.