Homolytic N−S Bond Cleavage in Vinyl Triflimides Enabled by Triplet–Triplet Energy Transfer

Abstract

Vinyl triflimides are a new compound class with unknown reactivity. A computational analysis identified homolytic cleavage of the N−Tf bond induced by triplet–triplet energy transfer (EnT) as a highly interesting reaction type that might be accessible. A combination of experimental and mechanistic work verified this hypothesis and proved the generated radicals to be amenable to radical–radical coupling. Thereby, vinyl triflimides were transformed into a range of α‐quaternary, β‐trifluoromethylated amines in a 1,2‐difunctionalization reaction with no need for external CF₃ reagents.

No need to pay for extras: Vinyl triflimides are a new compound class with unknown reactivity. Computational analysis identified homolytic cleavage of the N−Tf bond induced by triplet–triplet energy transfer (EnT) as an accessible reaction manifold. This hypothesis was verified, and vinyl triflimides were transformed into a range of α‐quaternary, β‐trifluoromethylated amines in a 1,2‐difunctionalization reaction with no need for external CF₃ reagents.

Visible‐light‐mediated photocatalytic reactions have become one of the most popular ways to generate reactive radical species. Most of these reactions engage electron‐transfer reactions for the generation of open‐shell intermediates. An interesting yet underutilized alternative for the formation of radicals is homolytic bond cleavage. The two simultaneously generated radicals can be amenable to the design of new strategies for difunctionalization reactions or radical–radical coupling. ^[1] Unfortunately, such processes require substantial amounts of energy, as the substrate must be infused with the entire bond dissociation energy. Therefore, they were mostly restricted to the realm of UV‐light irradiation for a long time, and thus plagued by low functional‐group compatibility and the need for special equipment.[ ² , ³ ] As a solution to these drawbacks, it was recently demonstrated that EnT from visible‐light photocatalysts might provide the required energy for homolytic bond cleavage under milder and more selective conditions[ ⁴ , ⁵ ] (e. g., Scheme 1a). ^[6] For such a reaction to be successful two requirements are essential: A substrate must have a triplet excited‐state energy (E(T₁−S₀)) that is lower than in the chosen EnT catalyst so that its T₁ state is efficiently populated. ^[1b] As commonly used catalysts have an E(T₁−S₀) around 60 kcal mol⁻¹ the triplet energy of the substrate must not exceed this value. Furthermore, to allow for efficient bond cleavage in the T₁ state the BDE must be below the E(T₁−S₀) of the substrate. ^[7] As these criteria have not yet been identified in many compounds the range of EnT acceptors for this new and highly interesting type of transformation remains limited.

Scheme 1.

Reaction precedence and predictions. Calculations were performed at the (U)M06‐2X/311++G(d,p) level of theory with an ultrafine integration grid of 99 590 points and Grimme's D3 version (zero damping) for BDEs ^[14] and M06‐2X/6‐31+G(d,p) for all other energies, which are displayed in kcal mol⁻¹. ^[13]

Vinyl triflimides, the nitrogen containing analogues of vinyl triflates, are a new compound class. Synthetic access was reported only recently. ^[8] Hence, their reactivity remains completely unknown. As N−S bonds in sulfonamides are known to have rather low BDEs, they seem to be ideal candidates for a fragmentation into radicals. ^[9] Also, their photodegradation by homolytic bond cleavage is known as a major pathway in the environment ^[10] and homolytic cleavage was shown to be efficiently promoted under UV‐irradiation in aryl triflimides ^[11] (e. g., Scheme 1b). ^[12] Hence, we were inspired to look further into the possibility of using vinyl triflimides as a new entry in EnT acceptors. Experimental data are unavailable for these new compounds, hence we turned to DFT calculations for an estimate of E(T₁−S₀) ^[13] and BDE. ^[14] As shown in Scheme 1c), the E(T₁−S₀) of a representative vinyl triflimide is predicted to be easily surpassed by a range of common triplet sensitizers at ∼51 kcal mol⁻¹. Furthermore, the BDE of the N−S bond is predicted to be not only remarkably low, at ∼44 kcal mol⁻¹, a value typically encountered for the O−O bond in the widely used peroxides, ^[15] but also significantly lower than the E(T₁−S₀). From a thermodynamic standpoint, these numbers seem ideal for a highly efficient EnT‐induced homolytic cleavage. For comparison we also calculated the BDE of the N−S bond in Fagnoni's aryl triflimide (Scheme 1b)) with the same method and found its value within the same range. Since it was shown experimentally to cleave efficiently, the same may be expected for vinyl triflimides if a sensitizer can be identified. Finally, the envisioned fragmentation is predicted to have a barrier of only 5 kcal mol⁻¹ from T ₁.

Encouraged by this analysis we set out to identify a suitable photosensitizer for the envisioned homolytic cleavage of the N−S bond in vinyl triflimides (Table 1). When comparing the experimental values for the E(T₁−S₀) of the tested sensitizers, it is directly evident that the reaction starts to proceed as soon as the estimated value for the model vinyl triflimide 1 a (E(T₁−S₀): 51 kcal mol⁻¹) is exceeded. We were surprised to find that fragmentation of the model substrate 1 a not only proceeded smoothly, but also yielded the imine 6 a as the product of what may be a radical–radical coupling reaction (see below for further discussion). While no reaction occurs with Ru(bpy)₃Cl₂ (E(T₁−S₀): 45.4–49.0 kcal mol⁻¹) the conversion reaches 70 % after only 4 h in the presence of benzil (E(T₁−S₀): 53 kcal mol⁻¹). The evaluation of a set of photosensitizers with E(T₁−S₀)‘s above the required range revealed thioxanthone as ideal for an efficient reaction. Even though the tested iridium catalyst, benzil and diacetyl also led to high conversion, the reactions were plagued by side reactions. Xanthone and anthraquinone were less efficient. Further optimization (see the Supporting Information) allowed for a reduction of the thioxanthone loading to 10 %.

Table 1.

Triplet sensitizer screening.^[a]

	Sensitizer	E T	Yield imine 6 [%][c]
		[kcal mol−1]	1 h	4 h	6 h
1	methylene blue	32.0[ 1b , 7 ]	0	0	0
2	rose bengal	39.0 [16]	0	0	0
3	Ru(bpy)3Cl2	45.4–49.0[ 7 , 17 ]	0	0	0
4	(−)‐riboflavin	50.0[ 1b , 7 , 18 ]	0	0	0
5	benzil[b]	53.7–54.0[ 18 , 19 ]	24	70	73
6	diacetyl	55.0–57.0 [20]	48	68	79
7	[Ir(dF(CF3)ppy)2(dtbpy)] PF6 [b]	61.8[ 1b , 7 , 21 ]	77	48	25
8	anthraquinone	62.4 [19]	20	35	50
9	thioxanthone	63.4–65.4 [ 1b , 7 , 19 ]	0	67	77
10	xanthone	74.0[ 1b , 17 , 18 ]	0	11	39

[a] Triflimide 1 (0.05 mmol), triplet sensitizer (50 mol %), CH₂Cl₂ (0.05 M), rt, blue LED, t=1–6 h. [b] hydrolysis of 6 to α‐trifluoromethylated ketone. [c] Yield of 6 determined by ¹H NMR analysis.

To render the procedure even more synthetically useful, we then developed these initial results into a one‐step difunctionalization reaction (Table 2). ^[22] Therefore, trifluoromethylated imine 6 was trapped by a nucleophile to yield the tetrasubstituted amine 7.

Table 2.

Scope of the EnT‐enabled trifluoromethylating difunctionalization of triflimides. Conditions: triflimide 1 a–p (0.05 mmol, 1.0 equiv.), thioxanthone (0.005 mmol, 10 mol %), CH₂Cl₂ (1 mL), rt, blue LED, t=18 h; then: nucleophile (0.15 mmol, 3.0 equiv.), rt, t=24 h. Yield of corresponding imine 6 a–p given in parentheses. Yields determined by ¹H NMR analysis.

[a]  3 mol % [Ir(dF(CF₃)ppy)₂(dtbpy)]PF₆ instead of thioxanthone.

An evaluation of the scope is summarized in Table 2. In general, a variety of difunctionalized amines 7 a–p were obtained in good to excellent yields using trimethylsilyl cyanide as the nucleophile. Triflimides with different para‐substituents in the aryl moiety, such as alkyl 1 b, c, aryl 1 d, halogen 1 e–g, trifluoromethyl 1 j, ester 1 l, nitro 1 m and methoxy groups 1 n provide the desired amines irrespective of their electronic properties (80–93 % yield). Disubstituted aryl 1 k and aryls with substituents in different positions 1 f–h, 1 n, o are well tolerated (47–80 %), while the efficiency of imine formation raises with increased distance of the substituent from the vinyl function. To our delight, even heteroaromatic triflimide 1 p was converted in good yield (86 %). Conversion of ortho‐substituted triflimide 1 i was initially challenging, yet product formation occurred readily when switching to an iridium‐based sensitizer (74 %). It may be worth noting that EnT to the triflimides proved highly efficient in all cases, yielding the imines in very high yields. Nevertheless, for some products (7 d–f, i–n) slightly diminished yields are caused by a hydrolysis to the α‐trifluoromethylated ketone. Changing to Grignard reagents as nucleophiles, amines 7 q–x were synthesized in good to excellent yields (87–98 %), providing α‐tertiary amines with alkyl 7 r, s, x, vinyl 7 t, allyl 7 u, aryl 7 v or benzyl 7 w moieties. Heteroaromatic amine 7 q was obtained in 86 % yield with MeMgCl as the nucleophile, while only unreacted imine was recovered with trimethylsilyl cyanide.

Mechanistic investigations were directed towards two central questions: 1) Is radical formation occurring through an EnT mechanism and 2) Is radical–radical coupling ^[1] driving the C−C bond formation or is a chain reaction initiated? ^[23]

Exclusion of light or the photosensitizer failed to promote any conversion of the triflimide 1a (Table 3, entries 1, 2). The addition of the radical scavenger TEMPO (2.0 equiv., see the Supporting Information) completely inhibited the reaction. This indicates that irradiation‐induced radical formation plays a crucial role in the reaction. Steady‐state ultraviolet–visible absorption spectroscopy reveals thioxanthone as the only absorbing species at irradiation wavelength, validating the initiation by triplet energy transfer by the irradiated sensitizer. Already discussed above, a clear correlation of the ET₁−S₀) values of the tested sensitizers with the predicted E(T₁−S₀) of the substrate results in a sharp threshold of reaction productivity at the predicted value. A comparison of different established photoredox catalysts shows no correlation of the redox potential and reaction efficiency (see Table S3 in the Supporting Information). Control experiments in the presence of (triplet) excited‐state quenchers Q1 ^[24] and Q2 ^[25] (Table 3, entries 3, 4) not only confirm the importance of excited‐state intermediates by inhibiting the reaction but also indicate that excited triplet states are involved. As expected, the reaction ceased in an oxygen atmosphere (entry 5).[ ^24b , ²⁶ ] All of these experiments clearly vouch for a radical formation through an EnT mechanism.

Table 3.

Triplet quencher and control reactions.^[a]

	Deviation from standard conditions	Quencher	Yield 6 [%][b]
1	dark	–	0
2	no triplet sensitizer	–	0
3	–	2,5‐hexa‐2,4‐diene Q1	0
4	–	cyclooctatetraene Q2	0
5	–	O2	0

[a] Triflimide 1 (0.05 mmol), thioxanthone (10 mol %), quencher (50 mol %), CH₂Cl₂ (0.05 M), RTrt, blue LED, t=18 h. [b] Yield determined by ¹H NMR analysis.

Our investigation towards a better understanding of the C−C bond forming mechanism started with a light on/off experiment (see the Supporting Information). As the reaction was found to cease immediately in the dark, a chain reaction, if occurring, is very short lived. ^[27] A second experiment to differentiate between a radical–radical coupling versus a chain reaction driven C−C bond formation is the crossover experiment shown in Scheme 2. Even though the results show some crossover, the numbers deviate significantly from the equal distribution expected for an unmitigated radical chain. This effect is more pronounced when the viscosity of the solvent is increased (see the Supporting Information for more details). This indicates for a combined mechanism, that is dominated by a radical–radical coupling with some participation of a radical chain.

Scheme 2.

Crossover experiment. Product distribution determined by ¹⁹F NMR analysis. Conditions: triflimide 1 (0.05 m/ mol, 1.0 equiv.), nonafluorobutyl triflate 8 (0.05 mmol, 1.0 equiv.), thioxanthone (0.005 mmol, 10 mol %), CH₂Cl₂ (1 mL), rt, blue LED, t=18 h. Solvent viscosity was modulated by the addition of increasing amounts of poly(ethylene glycol) dimethyl ether (M _W=530.65 g mol⁻¹).

To get a better understanding of the origins of this bifurcation into both reaction types we turned to DFT based computational analysis (see the Supporting Information for detailed discussion). According to this analysis we assign a central role to the C−S bond cleavage in the trifluoromethylsulfonyl radical 3 (Scheme 3). Formed as one of the initial fragmentation products after the energy transfer, the rate of its fragmentation into SO₂ 5 and the CF₃ radical 4 may govern the reaction path bifurcation. If it is fast enough to occur while still in the solvent cage with the radical fragment of the triflimide 2 the reaction occurs as a radical–radical coupling step. ^[28] If it is slow enough to allow for an escape from the solvent cage a radical chain is initiated because the addition of the trifluoromethyl radical 4 at the olefin in the triflimide starting material 1 has a barrier of only ∼10.3 kcal mol⁻¹. Under the applied room temperature conditions, the predicted small yet substantial barrier of ∼9.7 kcal mol⁻¹ for the C−S bond cleavage in the trifluoromethylsulfonyl radical 3 might be just high enough to allow for its escape from the solvent cage in a limited number of cases and thereby account for the share of products generated by a chain reaction. This solvent cage effect‐based hypothesis is reinforced by the observation of a viscosity dependent increase in product selectivity for the radical–radical coupling product (see the Supporting Information for details).

Scheme 3.

Proposed reaction mechanism.

In summary, a new type of EnT acceptor substrate has been discovered and shown to be amenable to radical–radical coupling. The computationally backed hypothesis that energy transfer induces an efficient homolytic cleavage of the N−Tf bond in vinyl triflimides was thereby verified. Alongside, a new protocol was developed for the 1,2‐difunctionalization of the highly electron deficient olefin moiety yielding a range of α‐quaternary, β‐trifluoromethylated amines. As the fragmentation of the N−Tf bond not only provides the reactive species but also the trifluoromethylating reagent, external and often costly and impractical CF₃ sources are unnecessary. Hence, this first study of the reactivity of this new compound class has proven vinyl triflimides to be interesting building blocks.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

C. Strauch, S. Schroeder, G. Grelier, M. Niggemann, Chem. Eur. J. 2022, 28, e202201830.

Data Availability Statement

Research data are not shared.