Photocatalytic α‐Tertiary Amine Synthesis via C−H Alkylation of Unmasked Primary Amines

Abstract A practical, catalytic entry to α,α,α‐trisubstituted (α‐tertiary) primary amines by C−H functionalisation has long been recognised as a critical gap in the synthetic toolbox. We report a simple and scalable solution to this problem that does not require any in situ protection of the amino group and proceeds with 100 % atom‐economy. Our strategy, which uses an organic photocatalyst in combination with azide ion as a hydrogen atom transfer (HAT) catalyst, provides a direct synthesis of α‐tertiary amines, or their corresponding γ‐lactams. We anticipate that this methodology will inspire new retrosynthetic disconnections for substituted amine derivatives in organic synthesis, and particularly for challenging α‐tertiary primary amines.


A. General Experimental S3
B. General Procedures S11

Light Sources:
The light sources employed in this work are shown in Table S1, with the relevant spectra given in Figs. S1 and S2.

Flow NMR Spectroscopy:
A peristaltic pump (Vapourtec SF-10) was used to circulate the mixture around the system to an InsightMR flow tube (Bruker) located within the spectrometer (Bruker Avance III + 500 MHz Ultrashield equipped with a Prodigy cryoprobe). In order to minimise the delay time between a change occurring in the reaction vessel and the arrival of the sample to the spectrometer for detection it is desirable to ensure that the volume of the tubing connecting the reaction vessel to the spectrometer is minimised, therefore narrow diameter polyetheretherketone (PEEK) tubing (0.762 mm i.d., Upchurch Scientific) was used. The PEEK tubing offers high chemical and mechanical stability (pH 0-14, -50-100 °C, >300 bar) along with good flexibility and low gas permeability. All other connections were made using standard HPLC-type PEEK connectors (Upchurch Scientific). All equipment was positioned inside a fume hood, located approximately 1 m from the shielded NMR spectrometer, without experiencing any adverse magnetic effects.
Data acquisition was performed without lock and with shimming performed using automated 1 H shimming routines, followed by manual fine adjustment. Data processing was performed using commercially available software. Solvent suppression of the MeCN resonance using a WET pulse sequence with a shaped pulse and low power 13 C broadband decoupling during acquisition was carried out with the Bruker pulse program "wetdc" (using a standard LC-NMR automated acquisition program, "au_lc1d" that first acquires a scan prior to starting the experiment to identify and subsequently suppress the desired number of solvent peaks areas throughout the experiment).

Supporting Information S7
A 20-mL scintillation vial equipped with a stirrer bar was flame-dried and transferred to a nitrogenfilled purge box whilst still hot, then was allowed to cool under the N2 atmosphere. The vial was then charged with stock solutions of 4CzIPN (2.28 mM in MeCN,6.58 mL,15.0 µmol,1 mol%) and tetrabutylammonium azide (70.3 mM in MeCN, 2.13 mL, 150 µmol, 10 mol%), and made up to a total volume of 10.0 mL by addition of MeCN (1.29 mL). Cyclohexylamine 5 (174 µL, 151 mg, 1.50 mmol, 1.0 equiv) was then transferred into the vial by microlitre syringe. Finally, methyl acrylate 14 (138 µL, 131 mg, 1.50 mmol, 1.0 equiv) was added, and the vial was sealed using an up-turned B24 rubber septa. It was then removed from the purge box, wrapped in tin foil before being transferred to the photoreactor.
The FlowNMR apparatus was purged with MeCN and then argon. The apparatus was connected to the reaction vessel by pushing the transfer lines through the septa, and the pump was started.
The flow tube was then inserted into the spectrometer and automated shimming and tuning routines were performed. Best results were obtained if automated shimming and tuning was performed on static samples; however, acceptable results were still obtained in flow. Frequency lock was switched off when using non-deuterated solvents, and shimming was performed on proton peaks.
Manual fine tuning of X and Y shims was often required to obtain a good peak line width. Spectra of the reagents were recorded without flow and again at the flow rate desired for the reaction.
Comparison of the integral areas of the peaks in each spectrum was used to calculate a correction factor for each reagent peak ( = peak integral, = correction factor). 1 corrected = (where = static / flow ) With the sample flowing, data acquisition was started using dedicated InsightMR reaction monitoring software, with solvent-suppressed 1 H spectra recorded at specified time intervals. To start the reaction, the light source was switched on. At the end of the reaction, or if intermediates of interest were observed, additional spectra were recorded with and without flow, and correction factors were calculated for the intermediate or product peaks, which were applied to each spectrum to give the final peak areas for calculation of species concentration and plotting of kinetic data.
Concentrations of species were determined by peak integrals referenced to tetrabutylammonium azide.
percolation through columns packed with neutral alumina under a positive pressure of nitrogen.

Preparation of tetrabutylammonium benzenesulfinate (13)
A 100-mL, Erlenmeyer flask equipped with a stirrer bar was charged with sodium benzenesulfinate (2.0 g, 12.2 mmol, 1.0 equiv) and H2O (20 mL), and the resultant mixture was stirred for ca. 5 min until a clear, homogeneous solution was obtained. Et2O (20 mL) and conc. aq. HCl (1.0 mL, 10.2 mmol, 0.84 equiv) were added sequentially, and the mixture was stirred for a further 5 min. The organic and aqueous layers were separated, and the aqueous phase was washed with portions of Et2O (28 mL). The combined organic layers were then concentrated in vacuo (in air) and the residue was dried under vacuum (ca. 0.05 mmHg) for 4 h to give benzenesulfinic acid as a white crystalline solid (1.20 g, 71%). This material was stored under N2 in a -20 °C freezer before use.
Next, aqueous solutions of Bu4NOH•30H2O and the freshly-prepared benzenesulfinic acid were made up in 25-mL volumetric flasks. These were each titrated against 1.00 M solutions of HCl and NaOH, respectively, using universal indicator solution to determine the end-points (in conjunction with pH test strips).  7, 124.3, 57.5, 23.05, 19.2, 13.5 Supporting Information S16

E.2.1. Optimisation Studies for α-C-H Alkylation
Under the standard conditions for α-C-H alkylation, the use of α-monosubstituted primary amine 1f results in a mixture of monoalkylated and dialkylated products: 17f and 20f, respectively. As generation of 20f consumes two equivalents of acrylate, the use of a 1:1 amine:acrylate stoichiometry (as per the standard protocol), leads to incomplete conversion of amine 1f (Entry 1).
A similar result is obtained upon changing the photocatalyst to Ir[(dF(CF3)ppy)2(dtbbpy)]PF6 (Entry 2). In order to increase the yield of α-C-H monoalkylated product 17f, it proved necessary to increase the equivalents of amine 1f to 2.0 equiv (Entry 3), and a further increase could be realised by switching the photocatalyst to Ir[(dF(CF3)ppy)2(dtbbpy)]PF6 (Entry 4). The origin of the better performance of the latter catalyst over 4CzIPN is unclear, but it may be due to enhanced stability against photobleaching (deactivation), and a consequent reduction in the extent of telomerisation occurring (determined by the lifetime of the α-carboxy radical). The significantly lower mass balance observed on increasing the amine loading seems to be a consequence of the formation of a complex mixture of unidentified, methyl ester-containing byproducts, as evidenced by a complex array of singlets in the -CO2Me region of the crude 1 H NMR spectra. The origin of these by-products is suspected to be telomerisation arising from additions of α-carboxy radicals to further molecules of acrylate, but we were unable to confirm this. They do Supporting Information S54 however complicate chromatographic purification. We also found that increasing the amine loading still further, to 3.0 equiv, affords no measurable increase in the yield of 17f.

E.2.2. Isolated Yields for α-C-H Monoalkylation of α-Monosubstituted Primary Amines
Prior to determining that 2.0 equiv of amine is necessary to mitigate α-C-H dialkylation (giving over-alkylated products 20), we had conducted a brief survey of α-monosubstituted amines (1yae) in the α-C-H alkylation process. All of these results (with the exception of amine 1ae) were performed with only 1.0 equiv of amine, and so the yields of monoalkylated lactams 17 are consequently low. Chromatographic purification was also challenging in these cases, which is why the more polar acrylate derivative 15 was employed. Nevertheless, we have summarised our findings in Fig. S6, followed by all the experimental procedures and characterisation data for these experiments. [a] 16% of the α,α-dialkylated product 20z was also isolated.
[b] Gave 16% of TsNH2 as a by-product.

Preparation of (RS)-4-methyl-1-azaspiro[4.5]decan-2-one (23d)
Following Purification via flash column chromatography on silica gel (12 g  The following amines were not successful in the α-C-H alkylation process: The following Michael acceptors were not successful in the α-C-H alkylation process:

F.1.1. Optimising the Lamp Power and Residence Time
The following solutions were prepared in 100-mL volumetric flasks under N2, and transferred to 250-mL, round-bottomed flasks capped with rubber septa under an N2 atmosphere:
A Vapourtec E-series flow reactor equipped with a UV-150 10-mL photoreactor and dry-ice cooling module was used. After priming the reagent lines for feeds A and B, and flushing the system with dry MeCN, 5.0 mL portions of feeds A and B were injected simultaneously into the photoreactor at various flow rates (residence time: 5-20 min 19 ), mixed in a T-mixer and passed through a 10 mL coil (0.8 mm inner diameter, fluoropolymer tube), irradiated with a 420 nm LED array (18 W or 54 W radiant output power) at 20 °C. The pressure was kept around 1 bar by using the third pump as a back-pressure regulator (BPR). After the entire 10-mL mixture had entered the reactor, it was followed with dry MeCN at the same flow rate. A 3.00-mL aliquot of the steadystate product mixture was collected and transferred to a microwave vial, then concentrated in vacuo on a spiral evaporator. Et3N (0.42 mL, 304 mg, 3.00 mmol) and MeOH (3.0 mL) were added, and the vial was crimp-sealed and heated at 70 °C in an aluminium block for 2 h. After cooling to rt, the vial was de-crimped and the mixture was concentrated in vacuo on a spiral evaporator, before subjection to 1 H NMR analysis. The NMR yield of 17b was calculated using the tetrabutylammonium ion as an internal standard, using the resonance at δH = 3.35 (8H, m) as a reference peak.

F.1.2. Attempts to Perform Lactamisation Step in Continuous Flow by Heating
We also attempted to perform lactamisation of the intermediate 8 in continuous flow by passing through a heated reactor after the UV-150 photoreactor, but even at 100 °C (20 min residence time) the lactamisation remained incomplete (66% conversion). Switching the reaction solvent to 90:10 MeCN/t-AmOH and subsequent heating at 100 °C also proved ineffective, with the lactamisation conversion only 47%.

F.2.1. Determining the Concentration of γ-Amino Ester Intermediate 8 Produced In Flow
The following solutions were prepared in 50-mL volumetric flasks under N2: nm LED array (54 W radiant output power) at 20 °C. The pressure was kept around 1 bar by using the third pump as a back-pressure regulator (BPR). After the entire 10-mL mixture had entered the reactor, it was followed with dry MeCN at the same flow rate. A 5-mL aliquot of the central part of the steady-state product mixture was collected over the time period 28-38 min, as a dark orange solution. From this solution, 3 × 1 mL aliquots were taken and transferred to microwave vials, then crimped and heated at 80 °C for 2 h. After allowing to cool to rt, the vials were de-crimped and the mixtures were each concentrated in vacuo on a spiral evaporator. Mesitylene (2.0 µL, 0.019 mmol) was added to each sample as an internal standard, followed by subjection to 1 H NMR analysis. The NMR yield for 17b (80% across all three aliquots) was calculated using the tetrabutylammonium ion as an internal standard, using the resonance at δH = 3.35 (8H, m) as a reference peak. Assuming the lactamisation of 17b is quantitative, the concentration of 8 in the steady-state product mixture was calculated as 0.08 M by reference to the mesitylene internal standard, using the resonance at δH = 6.77 (3H, s). The concentration of tetrabutylammonium azide in the steady-state product mixture was also calculated as 0.0088 M.

F.2.2. Determining the Quantity of Electrophile Needed to Trap Flow-Generated 8
For carrying out N-functionalisations of flow-generated 8 with electrophiles, we first needed to calculate the quantity of electrophile needed. An excess of electrophile is necessary due to the presence of unreacted cyclohexylamine 5 in the product flow stream, as well as the potential for reaction of the azide ion catalyst with the electrophilic reagent. Thus, in a 5-mL aliquot of the steady-state product mixture, the total quantity of reactive amine present (i.e., 5 + 8) is given by:

.3. N-Functionalisation of Flow-Generated γ-Amino Ester Intermediates 8 or 22g
The flow reaction was set up and run as per the procedure in section F.2.1. For generation of γ-amino ester 22g (R = t-Bu), methyl acrylate 14 was substituted for tert-butyl acrylate 21g, but all other parameters were kept unchanged (and the calculated concentration and NMR yield of 22g was found to be identical to 8). As before, a 5-mL aliquot of the central part of the steady-state product mixture was collected over the time period 28-38 min, as a dark orange solution. This was collected in a nitrogen-flushed, round-bottomed flask equipped with a stirrer bar, sealed with a rubber septum, and immersed in an ice-water bath. The requisite electrophile (and any other reagents) needed for N-functionalisation were then added dropwise to the reaction mixture (or, if stated, the inverse of this addition was performed).

propanoate (24a)
Following the general procedure described above, 5.0 mL of a steady-state aliquot of crude γ-amino ester 8 (0.40 mmol) was added dropwise to a stirred solution of 9-fluorenylmethoxycarbonyl chloride (170 mg, 0.65 mmol, 1.63 equiv) in MeCN (1.0 mL) at 0 °C under an N2 atmosphere. The reaction mixture was stirred at 0 °C for 3 h, then was allowed to
The reaction mixture was stirred at 0 °C for 3 h, then was allowed to warm to rt and was partitioned between CH2Cl2 (25 mL) and sat. NaHCO3 (25 mL

G.1. Gram-Scale Reaction on Vapourtec UV-150 Reactor (Back to Top)
The following solutions were prepared in 50-mL volumetric flasks under N2: nm LED array (54 W radiant output power) at 25 °C. The pressure was kept around 1 bar by using the third pump as a back-pressure regulator (BPR). It was necessary to replenish the dry ice in the cooling module at intervals of ca. 30 min to ensure that reactor overheating did not occur, and the set point was temporarily increased to 30 °C during each re-filling process (to account for the brief temperature rise on removing the cooling module lid, and prevent the run terminating due to 'high' temperature). After the entire 88-mL mixture had entered the reactor, it was followed with dry MeCN at the same flow rate. The steady-state product mixture (from 25-196 min) was collected in a 250-mL, round-bottomed flask as a dark orange solution (85.5 mL), and was then concentrated in vacuo. Et3N (7.36 mL,5.34 g,52.8 mmol) and MeOH (88 mL) were added, and the flask was Safety Note: It is known that inorganic azide reacts with CH2Cl2 to generate CH2(N3)2, diazidomethane, and that this material presents a serious explosion risk when concentrated. 21 In the above procedure, we performed chromatographic purification using CH2Cl2 as an eluent, so there is the possibility of forming some CH2(N3)2 during this step, by reaction of the tetrabutylammonium azide residues. The maximum theoretical yield would have been 74 mg of CH2(N3)2 and, if formed at all, this would likely have eluted very early in the run. We did not experience any issues in our own laboratory but, for future reference, it may be advisable when working on gram-scale to replace CH2Cl2 with an alternative chromatography solvent.

G.2. Optimisation on Uniqsis PhotoSyn Reactor (Back to Top)
The following solutions were prepared in 100-mL volumetric flasks under N2: The pressure was kept around 1 bar by using the third pump as a back-pressure regulator (BPR).

Reagent feed A:
After the entire 10-mL mixture had entered the reactor, it was followed with dry MeCN at the same flow rate. A 3.00-mL aliquot of the steady-state product mixture was collected and transferred to a round-bottomed flask, then concentrated in vacuo before subjection to 1 H NMR analysis. The combined NMR yield of 8 + 17b (with the latter as a minor component) was calculated using the tetrabutylammonium ion as an internal standard, using the resonance at δH = 3.35 (8H, m) as a reference peak.   Table S8. Summary of the above Optimisation experiments. Although the NMR yields of 8 + 17b are all appreciably lower than that for 17b under the optimised conditions on the Vapourtec UV-150 reactor (i.e., 90% NMR yield of 17b with 54 W LED radiant output power at 420 nm), we reasoned that increasing the residence time to 20 min with the Uniqsis reactor would be counterproductive in terms of material throughput (STY). It should be noted that the Uniqsis reactor wavelength is centred on 455 nm (with a half-height bandwidth of 25 nm), and we already know that our reaction is appreciably slower at 450 nm than 425 nm (see section D.3).

G.3.2. Performing a Decagram-Scale Reaction
The following solutions were prepared under N2:
A Vapourtec E-series flow reactor equipped with a Uniqsis cold coil tubing module and a PhotoSyn Blue HP LED photoreactor with a water-cooled 455 nm LED array (700 W radiant Supporting Information S109 output power) was used. After priming the reagent lines for feeds A and B, and flushing the system with dry MeCN, 320 mL portions of feeds A and B were injected simultaneously into the photoreactor at flow rates of 1.00 mL min -1 (residence time: 10 min), mixed in a T-mixer and passed through a 20 mL coil (1.0 mm inner diameter, fluoropolymer tube), irradiated with a 455 nm LED array (700 W radiant output power) at 10 °C. The pressure was kept around 1 bar by using the third pump as a back-pressure regulator (BPR). After the entire 640-mL mixture had entered the reactor, it was followed with dry MeCN at the same flow rate. The steady-state product mixture (collected over 5 h 17 min) was collected in a 1-L round-bottomed flask as a dark orange solution (634 mL), and was then concentrated in vacuo before subjection to 1 H NMR analysis. An NMR yield for 8 of 41% was calculated using the tetrabutylammonium ion as an internal standard, using the resonance at δH = 3.35 (8H, m) as a reference peak. Et3N (52.9 mL, 38.4 g, 380 mmol, 2.0 equiv) and MeOH (250 mL) were added, and the flask was fitted with a water-jacketed reflux condenser and heated at reflux in an oil bath for 3 h. After cooling to rt, the mixture was concentrated in vacuo, followed by dilution with EtOAc (500 mL). The unreacted cyclohexylamine 5 was removed by washing with 2 M aq. HCl (500 mL), and the aqueous layer was back-extracted with EtOAc (3 × 100 mL). The combined organic layers were then washed with sat. aq. NaHCO3 (500 mL), then the layers were separated and the organic phase was dried (MgSO4), filtered and concentrated in vacuo. Purification by recrystallisation was performed by dissolution in hot MeCN (ca. 150 mL) and gravity filtration through filter paper (whilst still hot) to remove small quantities of insoluble material. The filtrate was left to cool to rt over 1 h, and was then cooled in a refrigerator (5 °C) overnight. Vacuum filtration, followed by rinsing with cold (-5 °C) MeCN and drying under high vacuum gave a first crop of 17b as off-white crystals (2.43 g). The mother liquor was then concentrated in vacuo and a second recrystallisation was performed by dissolution in hot MeCN (ca. 100 mL), followed by cooling to rt and then in a refrigerator (5 °C) overnight. Vacuum filtration, as above, gave a second crop of 17b as off-white crystals (2.10 g). The mother liquor was again concentrated in vacuo and the residue was dissolved in CH2Cl2 (50 mL) and concentrated in vacuo onto Celite (ca. 6 g). Purification via flash column chromatography on silica gel (120 g) in CH2Cl2 (3 CV) then 100:0:0→90:9:1 CH2Cl2-MeOH-aq.
NH4OH (over 11 CV) then 90:9:1 CH2Cl2-MeOH-aq. NH4OH (10 CV) gave additional 17b as an off-white solid (5.61 g). When combined with the two crops of recrystallised material, this led to an overall isolated yield of 10.1 g of 17b.  Based on the typical detection limits of 1 H NMR (i.e., detection of peaks with ~0.5% intensity), we estimated a lower limit for the relative reactivity ratio of 5:18 of 20:1.
The anodic/cathodic peak potentials were measured using a glassy carbon working electrode, a platinum plated counter electrode, and a 3 M KCl Ag/AgCl reference electrode using a 100 mV s -1 sweep rate for the cyclohexylamine 5 and a 500 mV s -1 sweep rate for tetrabutylammonium azide. The measurements are reported in graphs vs SCE, by subtraction of 30 mV from all voltages relative to the Ag|AgCl electrode. 23 Ep were obtained from the graphs as the potential corresponding to the maximum current observed.
To better evaluate compounds' redox potential, Ep/2 were used, 48 and these were measured as the potential corresponding to the half value of the maximum current intensity observed. Cyclic voltammetry of both cyclohexylamine 5 and tetrabutylammonium azide both show completely irreversible oxidation waves with Ep A and Ep/2 (vs SCE) (Fig. S18).

H.4. Quantum Yield Measurement (Back to Top)
The quantum yield was measured for the reaction of cyclohexylamine 5 with methyl acrylate 14.

H.4.1. Determination of the Photon Flux
The photon flux of the LED setup was determined using standard ferrioxalate actinometry. 24 The number of moles of Fe 2+ formed was calculated using: where is the total volume of the solution after the addition of 1,10-phenanthroline (0.0015 L), ∆ is the difference in absorbance at λ = 510 nm between the irradiated and non-irradiated ferrioxalate solutions, is the optical path length of the irradiation cell (1.0 cm), and is the molar absorptivity of the Fe(phen)3 2+ complex at λ = 510 nm (11,100 L mol -1 cm -1 ).
The moles of Fe 2+ were plotted as a function of time (Fig. S22).  26.1 mg, 0.30 mmol, 1.0 equiv) were added, and the vial was sealed with a screw cap. It was then removed from the purge box and transferred to the same position in the same photoreactor used for the quantum yield measurement, and irradiated for 2 h at 425 nm (using the same lamp).
Following irradiation, the reaction mixture was concentrated in vacuo. The residue was dissolved in MeOH (3 mL) and Et3N (0.2 mL, 0.14 g, 1.35 mmol, 4.5 equiv) was added. The reaction was then heated at reflux for 2 h, followed by concentration in vacuo. The yield of 17b was determined by 1 H NMR using tetrabutylammonium azide as an internal standard to be 45%. The reaction was repeated a second time, giving a yield of 50%. Average yield = 47%.
The quantum yield (Φ) was then calculated using: Φ = mol product photon flux × × where is the time (7200 s) and is the fraction of light absorbed by the 4CzIPN catalyst at λ = 425 nm, where = 1 − 10 − (for a 1.5 × 10 -3 M solution in MeCN, this was determined by UV/Vis spectroscopy to be 1.000).