Regioselective Rearrangement of Nitrogen- and Carbon-Centered Radical Intermediates in the Hofmann–Löffler–Freytag Reaction

The Hofmann–Löffler–Freytag (HLF) reaction serves as a late-stage functionalization technique for generating pyrrolidine heterocyclic ring systems. Contemporary HLF protocols utilize in situ halogenated sulfonamides as precursors in the radical-mediated rearrangement cycle. Despite its well-established reaction mechanism, experiments toward the detection of radical intermediates using EPR techniques have only recently been attempted. However, the obtained spectra lack the distinct features of the N-centered radicals expected for the employed reactants. This paper presents phenylbutylnitrone spin-trapped C-centered and N-centered radicals, generated via light irradiation from N-halogen-tosyl-sulfonamide derivatives and detected using EPR spectroscopy. NMR spectroscopy and DFT calculations are used to explain the observed regioselectivity of the HLF reaction.


■ INTRODUCTION
Modern C−H functionalization chemistries have introduced late-stage functionalization (LSF) strategies in medicinal chemistry, targeting drug lead C−H bonds for creating new analogues.This toolbox includes photoredox-mediated and radical reactions and among them, amination reactions for the direct formation of C−N bonds. 1,2Recently, the focus is shifting from metal to organocatalytic protocols, paving the way to sustainability and adhering to green chemistry principles to minimize waste and improve yield and atom economy. 3This approach aligns with the EU's sustainable development policy. 4,5Numerous research groups are exploring new C(sp 3 )-H functionalization reactions with high chemo-, regio-, and stereoselectivity.The Hofmann−Loffler− Freytag (HLF) reaction, used for building pyrrolidine (and in some cases, also piperidine) ring systems, is among photoactivated amination reactions without metal catalysis. 6,7The HLF reaction, first discovered in synthetic studies of Nhaloamines, 8−11 is a multistep process involving nitrogen atom activation through halogenation, N-centered radical generation via irradiation, intramolecular hydrogen atom transfer (HAT), and radical termination with cyclization to form the final C−N bond (Scheme 1).
Contemporary adaptations of the HLF reaction employ toluenesulfonyl (tosyl, Tos)-activated amines (1), which undergo in situ iodination at the nitrogen atom (2) via an iodine source and a co-oxidant.The formation of an Ncentered radical (3) was recently examined using EPR spectroscopy (Figure 1). 12wever, the obtained spectra (I, in green), while presenting a triplet indicative of the nitrogen hyperfine splitting, raised questions due to issues such as broad line width, a high g-factor (2.0064) value for the proposed Ncentered radical, and the absence of α-hydrogen splitting.Calculated EPR spectra for a model compound of 3 are shown in Figure 1 (II, orange line). 13The spectra of the ditosylated aminoxyl radical (III, in blue) fit with the EPR parameters of I, thus suggesting this species to be the correct assignment of the EPR spectra I. 14 Neither C-centered radical (4) nor C 5 -iodo functionalized (5) intermediates were observed in the EPR and NMR studies.The only confirmed product in this reaction was the pyrrolidine ring compound (6).Experimental attempts for in situ generation of N-centered radicals and detection via time-resolved EPR included N-isopropyl-4-methoxybenzenesulfonamide as a radical precursor under electrochemical conditions. 15It is, however, likely that the detected radical is not an amidyl radical but a nitroxide radical instead.A reaction of the observed radical with 5,5-dimethyl-1-pyrroline-N-oxide (DMPO) did not occur, which is consistent with stable nitroxide radicals.The experimentally observed hyperfine coupling constant (hfc) value for hydrogen at C 2 is significantly smaller than those in typical amidyl radicals but is similar to the calculated values for the nitroxide radicals.
Numerous synthetic studies utilizing HFL chemistry 16−20 report the regioselective functionalization and subsequent formation of 5-membered pyrrolidine rings.−24 The observed regioselectivity in HLF reactions occurs during the HAT phase and can result from two different pathways.The intramolecular mechanism, governed by the kinetic preference of 1,5-over 1,6-HAT steps, represents one pathway.Alternatively, an intermolecular route directs the product distribution based on the thermodynamic stability of the resulting C-centered radicals.Munĩz suggested the latter mechanism in the context of selective piperidine formation, 24 where a phenyl-stabilized radical at the C 6 position is formed from an N-centered radical in a bimolecular reaction and a pyrrolidine product was not observed.Between the three distinct pathways, it is unclear which one is dominant for a given set of reaction conditions, and detailed investigations are thus warranted.
Moisture-sensitive reactions were performed using flamedried glassware under a nitrogen atmosphere (N 2 ).Air-and moisture-sensitive liquids and solutions were transferred with a Scheme 1. Modern HLF Reaction Sequence Involving Tosyl-Sulfonamides and In Situ Iodination with Iodine and a Co-Oxidant [PhI(OAc) 2 )] Figure 1.Simulated EPR spectra for tosyl-N-centered and ditosyl-N-centered radicals with the corresponding structures assigned by the authors.The green line is from EPR experiments, 12 the orange line represents calculated EPR parameters, 13 and the blue line is from EPR experiments. 14e Journal of Physical Chemistry A plastic or glass syringe.Chromatographic purification of the products was carried out using column chromatography filled with silica gel (Macherey-Nagel) 0.063−0.2mm, and appropriate solvent mixtures were used as eluents: petroleum ether/ethyl acetate.Thin-layer chromatography (TLC) was performed on precoated TLC plates ALUGRAM SIL G/ UV254, 0.20 mm silica gel 60 with a fluorescent indicator UV254 (Macherey-Nagel) in the appropriate solvent system.TLC spots were observed after illumination with UV light at a wavelength of 254 nm and after immersion in an aqueous solution of KMnO 4 (3 g KMnO 4 , 20 g K 2 CO 3 , 5 mL aq.NaOH 5%, and 300 mL water) followed by heating.If TLC spots were not visible after illumination with UV light, they were detected utilizing an iodine chamber.
Synthetic photocatalyzed reactions were performed in a custom-built photoreactor with built-in temperature control as The Journal of Physical Chemistry A well as standardized luminous intensity for a certain set of high-power LED light sources.The InGaN-based H2A1-420 LED (420 ± 20 nm) used in this experimental setup was purchased from Roithner Lasertechnik GmbH and mounted on a standard hexagonal aluminum package.Irradiation was performed in situ (EPR) and off-site (NMR) with Kessil PR-160L 370 ± 10 nm gen-2 LED UV, with an average intensity of 137 mW/cm 2 when the sample is 6 cm from the lamp, according to the manufacturer. 25he reactant and products were identified using 1 H NMR spectra, which were recorded on Varian Inova 400 and 600 machines in CDCl 3 at 400 or 600 MHz at room temperature.All 13 C NMR spectra were recorded, respectively, at 101 and 151 MHz.The chemical shifts are reported in ppm (δ), relative to the resonance of CDCl 3 at δ = 7.27 ppm of 1 H and for 13 C, relative to the resonance of CDCl 3 at δ = 77.16ppm.NMR spectra of the reaction mixture were obtained on a Varian Inova 400 NMR spectrometer operating at 399.90 MHz for 1 H NMR and 100.6 MHz for 13 C NMR and are reported as chemical shifts (δ) in ppm.The spectra were imported and processed in the MestreNova 11.0.4 program. 26C measurements were obtained at a Shimadzu GC-2010 Plus gas chromatograph with an AOC-20i autosampler (with a temperature-controlled sample holder) and an Optima 1701− 0.25 μm (25 m × 0.25 mm) column.
HR-MS spectra were obtained using a Thermo Finnigan LTQ FT machine of the MAT 95 type with a direct exposure probe and electron impact ionization (70 eV).
EPR spectroscopy was done by using a Bruker E500 ELEXSYS EPR spectrometer with an ER4122SHQE cavity resonator.As this cavity resonator does not have an optical window for illumination, the light source was mounted underneath the cavity, with light coming through the bottom of the EPR 4 mm-inner-diameter tube.EPR deconvolution and simulation was done using an EasySpin module with the MATLAB program package. 27EPR visualization and spectroscopy were done using the VisualEPR Web page. 28he conformational space was sampled and investigated using the Conformer−Rotamer Ensemble Sampling Tool� CREST 29 coupled with the xtb-GFN2 program package and MD simulation using xtb-GFN1 30 and xtb-GFN2. 31−34 For each structure with a stable wave function, frequency calculation was performed to identify the minima and transition-state structures.From the transitionstate structure, an intrinsic reaction coordinate search was performed to characterize the corresponding reaction and product complexes/reactive conformers.Improved thermodynamics were obtained using RO-B2PLYP 35,36 with a G3MP2 large basis set 37 on geometries obtained at the B3LYP/6-31G(d) level of theory, with additional D3 dispersion correction. 38Additionally, for smaller and model systems, the G3B3 37 composite method was used, with results matching the RO-B2PLYP-D3/G3MP2 large results.
Calculations of EPR parameters were done using the B3LYP functional and the mixed basis set: EPR-III for C, H, and O atoms, def2-QZVP for the S atom, and 6-31G(d) for the N atom.A small basis set on the N atom is necessary for the correct calculations of the g-factor and hfcs. 39,40When using a larger basis set for the N atom, e.g., EPR-III or def2-QZVP, the obtained results systematically underestimate the hfc.Calculations were performed on the Gaussian version 16.C01 41 using the advanced computing service (clusters Isabella and Supek) provided by the University of Zagreb University Computing Centre�SRCE 42 and the computational resources of the PharmInova project (sw.pharma.hr) at the University of Zagreb Faculty of Pharmacy and Biochemistry. 43

■ RESULTS AND DISCUSSION
Insights into the HAT steps of the HLF reactions can be gained through the simulation of thermodynamic and kinetic parameters.−46 Reaction kinetics are linked to thermodynamics via the Bell−Evans−Polanyi principle, a relationship that has been demonstrated for various HAT reactions for both inter-and intramolecular pathways. 47hen the C 5 and C 6 positions are both unsubstituted as in substrate 7-H, the regioselectivity toward the C 5 product is experimentally evident even when the kinetic and thermodynamic parameters for 1,5-and 1,6-HAT align closely. 48There are no notable differences in energies vs the geometrical parameter (N−H−C angle) between 1,5-and 1,6-HAT energy profiles.Both of them resemble those of the intermolecular HAT (see Chart S1 in Supporting Information).There is room for unknown factors governing the regioselectivity of HLF reactions.To determine whether the transient N-and Ccentered radicals influence the observed regioselectivity, we attempted in situ generation and trapping of these radical intermediates using the phenylbutylnitrone (PBN) spin trap and investigation of the resulting adducts with EPR.Additionally, the reaction progress was monitored using NMR techniques, with off-site light irradiation.
The same starting conditions reported earlier 12,16 were employed for N-hexyl-4-methylbenzenesulfonamide (7-H) as the substrate.This reactant was chosen as the simplest model for unsubstituted C 5 and C 6 positions.The corresponding Niodo (7-I) derivative was prepared in situ from 7-H using hypervalent phenyliodine (III) diacetate (PIDA) as an oxidant with elemental I 2 as the iodine source. 12,16After 3 h of irradiation of a reaction mixture containing 3 equiv of PIDA, 1 equiv of I 2 , and 1 equiv of 7-H with a 420 nm light source, two distinct products could be observed (Scheme 2b).These products were identified through GC−MS and NMR techniques as a 55:1 mixture of the five-membered ring compound 8 pyrrolidine and six-membered ring analogue 9 piperidine in a combined yield of 72% (see Supporting Information).When the same reactant mixture was stored in darkness for 7 days, the crude 1 H NMR spectrum revealed the unexpected presence of imine 10 imine together with 4-methylbenzenesulfonamide (15) and hexanal (16).Subsequent processing with aqueous sodium thiosulfate resulted in near-quantitative recovery of compound 15.
With these results in hand, an attempt was made to identify transient intermediates by monitoring reaction progress with NMR techniques.The mixture of 7-H/PIDA/I 2 was therefore irradiated off-site for 5 min (370 nm), followed by 1 H NMR measurements during a full reaction sequence (see Supporting Information).In addition to signals of the starting material, two sets of new signals appeared, which is consistent with the formation of two products.These two cannot be clearly assigned, but the observed signal motifs and chemical shift values support the formation of two halogenated structures, one of them being the expected C 5 -halogenated product 8-I.

The Journal of Physical Chemistry A
The assignment of the other product was not possible without ambiguity but clearly did not correspond to the C 6halogenated product 9-I, or 10 imine , observed previously.To our knowledge, this is the first reaction monitoring HLF with NMR spectroscopy, with a large signal from excess PIDA covering the aromatic region of the spectra.
After NMR experiments, we monitored the complete reaction sequence with EPR spectrometry.Using continuous irradiation with the same UV lamp from the bottom of the cavity resonator, compound 7-I was EPR-silent.This is in stark contrast to the published results, 12 but as mentioned before, radicals observed in that experiment might stem from different oxidation pathways and rearrangement reactions that are not part of the HLF sequence.
Next, we tried to capture nascent radicals from the reaction with PBN, but oxidizing and halogenating species present in the mixture reacted with PBN and produced an oxo-PBN (acylaminoxyl) radical, with α N,exp = 8.0 G, as well as additional (oxo-PBN)-PBN adducts (see Figure 2). 49To summarize, we have not been able to detect short-lived radical intermediates of the HLF reaction using this procedure.
As in situ halogenation inhibits PBN's ability to spin-trap radicals, the preparation of 7-Cl and 7-Br was performed in a separate step.While 7-Cl was easily isolated and proved stable for a couple of days, 50 7-Br had to be synthesized, cleaned, isolated, and measured without any delay.Another major point is the sensitivity of the reaction to air.Line widths measured with EPR and reaction yields were greatly influenced by the effectiveness of air removal using freeze−pump−thaw cycles with backfill of argon or nitrogen gas.Experimental line widths of less than 0.4 G were deemed satisfactory for optimal resolution of radical adducts.Under these conditions with illumination with 370 nm light, we were able to observe a Cl-PBN adduct, proving homolytic cleavage of N−Cl bonds generating a chlorine radical that quickly combines with PBN (see Figure S20 in Supporting Information).From both 7-Cl and 7-Br, a PBN adduct 7-PBN (g exp = 2.0064, α N,exp = 14.14 G, α N′,exp = 1.58 G, and α H,exp = 3.95 G) formed from an Ncentered radical 7 was detected for the first time using EPR spectroscopy, proving that this is the correct method for investigating the HLF reaction and corresponding intermediates (see Figure 2).Calculated EPR parameters for 7-PBN (g 7-PBN,calc = 2.00616, α 7-PBN,N,calc = 13.81G, α 7-PBN,N′,calc = 1.52 G, and α 7-PBN,H,calc = 2.89 G) are in satisfactory agreement with experimental values.
In the EPR spectrum (see Figure 2), one signal corresponding to the two PBN adducts of C-centered radicals was expected, namely, C 5 radical (8) and C 6 radical (9).PBN adducts of those radicals, 8-PBN and 9-PBN, have similar calculated hfcs and g-factors (g 8-PBN,calc = 2.00595, α 8-PBN,N,calc = 15.09G, and α 8-PBN,H,calc = 2.36 G and g 9-PBN,calc = 2.00597, α 9-PBN,N,calc = 14.80 G, and α 9-PBN,H,calc = 2.27 G), and it is thus difficult to distinguish between them, due to both being secondary alkyl C-centered radicals with similar environments around the radical center.A signal with g exp = 2.0061, α N = 13.84G, and α H = 2.47 G was observed that can be assigned to both 8-PBN and 9-PBN.The unexpected result was the formation and detection of PBN adduct 10-PBN.It is characterized with g exp = 2.0064, α N = 13.80G, and α H = 7.34 G, which is different from the previously described radicals (more information about deconvolution can be found in the Supporting Information).The 10-PBN radical adduct can be tentatively assigned as a C 2 radical (10) due to having an hfc from N and H atoms in PBN and to different

The Journal of Physical Chemistry A
connectivities closer to the radical center, although it has an unusually high hfc value for a C-centered radical.To confirm this assignment, extensive computational analyses were performed.The calculated Boltzmann averaged values for 10-PBN (g 10-PBN,calc = 2.00610, α 10-PBN,N,calc = 14.08 G, and α 10-PBN,H,calc = 5.46 G) demonstrated a trend similar to the experimental parameters, with significantly larger hfc values than computed for radicals 8/9-PBN.A detailed analysis of the structural factors contributing to these values revealed interactions between the Tos-N(alkyl)-H and the O−N− PBN radical center, significantly impacting the α H value in the lowest lying minima.Conformers of 10-PBN, where this interaction is not present, are higher in energy by more than 10 kJ/mol from the global minima and have calculated α 10-PBN,H,calc 2.40 G, which is almost the same as that for radicals 8/9-PBN.For more details on deconvolution and structure analysis, please consult Supporting Information.
From theoretical predictions calculated at the RO-B2PLYP-D3/G3MP2-large//B3LYP/6-31G(d) level of theory, 47,51  Rearrangement to a C 2 radical can occur from less stable distant C-centered radicals (8 → 10 and 9 → 10).Both interand intramolecular reactions are feasible with the latter (7-H + 8 → 10 + 7-H and 7-H + 9 → 10 + 7-H) being characterized with ΔH 298 ‡ = 33.5 kJ/mol and a thermodynamic driving force of ΔH rx,298 = −23.4kJ/mol.This was calculated with the propane/propyl radical reference system as a reasonable model for the distant C-centered radicals in the hydrocarbon chain, due to negligible differences in reactivity and stability between the C 5 radical, C 6 radical, and propyl radical (RSEs and TS) toward the C 2 radical.As seen in Figure 3, the barrier for intramolecular 1,4-HAT CC rearrangement of the C 5 radical to the C 2 radical (8 → 10) is prohibitively high (ΔH 298 ‡ = 84.1 kJ/mol), while 1,5-HAT CC from the C 6 radical to the C 2 radical (9 → 10) proceeds through a 6-membered transition state with ΔH 298 ‡ = 56.1 kJ/mol and a reaction enthalpy of ΔH rx,298 = −21.5 kJ/mol.These results indicate that this second intramolecular rearrangement, with the lowest energy of activation, is the probable origin of the experimentally observed regioselectivity in HLF reactions.It significantly decreases the lifetime of the C 6 radical required for halogen atom abstraction and thus the yield of C 6 -functionalized products.
4][45][46][47]51 In our system, going from a tosylated methylamine radical (7) to a secondary C-centered radical (8 and 9) corresponds to an exothermic reaction (ΔH predict,298 = −19.8 kJ/mo, see Supporting Information).Additionally, rearrangement from a secondary C-centered radical to a tosylamidesubstituted C-centered radical fragment such as 10 is also predicted to be exothermic (see Supporting Information), which is in line with our experimental observation.
The same rationale can be used to explain the regioselectivity reported in a recent study 52 with different sulfonamide derivatives.While radical generation was achieved using visible light photoredox catalysis with Ru(bpy) 3 Cl 2 in The Journal of Physical Chemistry A conjunction with blue LEDs, high yields and regioselectivity toward 1,5-HAT NC rearrangement were observed when C 5 and C 6 positions were secondary and primary radicals.This stems from a notable difference in the radical stability (ΔRSE prim/sec = 10.8 kJ/mol).When additional substituents at the C 6 position are introduced to generate a (stabilized) tertiary C 6 radical, a mixture of a 56% C 5 product and a 40% C 6 product is obtained (ΔRSE sec/tert = 6.3 kJ/mol), making both 1,5-HAT NC and 1,6-HAT NC equivalent reactions.Proof of the existence of a C 2 radical in HLF-type reactions is available in the literature.It was observed indirectly as an imine side product in electrochemically driven N-centered radical generation. 53The proposed generation of imine was either via elimination of HBr from an N-brominated precursor or via 1,2-HAT from a transient N-centered radical.The possible sequence of 1,6-HAT NC followed by 1,5-HAT CC was not explored.

The Journal of Physical Chemistry A
In the 1 H NMR spectra of the reaction mixture obtained after off-site irradiation of the 7-Cl precursor, only two (main) products were observed and measured in the same proportion.The two pairs of triplets in the upfield and downfield regions are consistent with product structures 10-Cl and 12-Cl (see Supporting Information), which supports the formation of C 5and C 6 -radical intermediates, the latter undergoing a 1,5-H shift to the respective C 2 -centered radical.The same NMR results were obtained after irradiation of precursor 7-Br.
In the EPR spectra, when 11-Cl was illuminated, a chlorine PBN adduct was formed (Cl-PBN), alongside N-centered radical adduct 11-PBN.Both 7-PBN and 11-PBN have similar hfcs, namely, g 11-PBN,exp = 2.0062, α 11-PBN,N,exp = 14.1 G, α 11-PBN,N′,exp = 1.3 G, and α 11-PBN,H,exp = 4.21 G.Those results fit nicely with the calculated results (see Supporting Information).As expected, the C-centered radical signal corresponds well to the C 5 -and C 6 -radical adducts 12/13-PBN, with the same hfc values as observed for 8/9-PBN (g 12/13-PBN,exp = 2.0061, α 12/13-PBN,N,exp = 13.6 G, and α 12/13-PBN,H,exp = 2.6 G).The tentatively proposed C 2 -radical 14, similar to radical 10, made an adduct with PBN, 14-PBN with a characteristic signal at g 14-PBN,exp = 2.0061, with hfc α 14-PBN,N,exp = 13.7 G and α 14-PBN,H,exp = 6.4 G, confirming our hypothesis that this species stems from rapid rearrangement from 13. Again, a rather high hfc value for α 14-PBN,H can be attributed to the interaction between the sulfonamide hydrogen moiety and the oxygen-centered radical of the PBN in the thermodynamically stable conformers.Extensive calculations on radical adducts confirm the assignments (see Supporting Information).Similar results were obtained with 11-Br, although the most pronounced peaks were 12/13-PBN.11-PBN was weak and 14-PBN is not directly visible, possibly overshadowed by 11-PBN.The lack of full correspondence between 11-Cl and 11-Br is most likely due to the instability of 11-Br, which was already decomposing in the dark (see Supporting Information).We plan to continue the investigation on the weak components of the EPR spectrum deconvolution, currently assigned as 10/14-PBN.Additional compounds with different substitution patterns on the C 5 and C 6 positions, with added modifications on C 2 (blocking the route) and C 3 positions (steric hindrance), will further test the origin of regioselectivity in HLF reactions.

■ CONCLUSIONS
Using NMR and EPR spectroscopy in combination with DFT calculations, we successfully monitored the reaction profile and identified all significant intermediate radicals and products in the HLF reaction.The initial rearrangement step allows the Ncentered radical to transform into both C 5 and C 6 radicals.The observed regioselectivity favoring 1,5-HAT products can be attributed to an additional rearrangement reaction exclusive to the C 6 radical.Through another 1,5-HAT step, the C 6 radical is transformed into the most stable C 2 radical.The existence of C 2 radical is experimentally proved not only through EPR spectroscopy but also via synthetic reactions and side products, notably imine and aldehyde.Future research should extend to more complex systems with additional substituents on the radical chain and varying functional groups, applying the methodology outlined in this study.
Experimental details, synthesis procedure, reactant and product characterization, EPR simulation parameters, calculation procedures, geometries and energies of optimized structures, and recorded NMR and EPR spectra (PDF) ■ AUTHOR INFORMATION

Scheme 2 .
Scheme 2. (a) Preparation of Precursors and Possible Intermediates in HLF Reactions of 7-H and 11-H, (b) Observed Products in the Reaction of 7-H with PIDA/I 2 under Dark and Irradiation Conditions, and (c) Investigated Reaction Sequences and Possible Products in Spin-Trapping Experiments

Figure 2 .
Figure2.EPR spectra of spin-trapped radical intermediates generated with 370 nm irradiation of 7-Br.The experimental spectrum is in purple, while blue, orange, green, and red correspond to simulated spectra for 10-PBN, 8/9-PBN, 7-PBN, and total simulated spectra, respectively.Residuals from subtracted simulation from experimental spectra are at the bottom.More information on deconvolution and simulation is deposited in Supporting Information.

Figure 4 .
Figure 4. RSEs for selected N-and C-centered radicals present in systems 7−14.The gray bands denote the anchor points for N-centered and Ccentered radical RSE scales to a global BDE scale.