Profiling the oxidative activation of DMSO-F₆ by pulse radiolysis and translational potential for radical C–H trifluoromethylation

Abstract

The oxidative activation of the perfluorinated analogue of dimethyl sulfoxide, DMSO-F₆, by hydroxyl radicals efficiently produces trifluoromethyl radicals based on pulse radiolysis, laboratory scale experiments, and comparison of rates of reaction for analogous radical systems. In comparison to commercially available precursors, DMSO-F₆ proved to be more stable, easier to handle and overall more convenient than leading F₃C-reagents and may therefore be an ideal surrogate to study F₃C radicals for time-resolved kinetics studies. In addition, we present an improved protocol for the preparation of this largely unexplored reagent.

---

Introduction

The generation of C-centered radical species R˙ under laboratory conditions requires suitable precursors, P(R), and a matched set of activating conditions, A, since most of the members of this class of molecules cannot be stored in bottles (Scheme 1). For example, methyl radicals (H₃C˙) may be accessed most conveniently by the oxidative activation of readily available laboratory solvents, e.g. dimethyl sulfoxide (DMSO). For example, the hydroxyl radical (HO˙) readily generated under traditional Haber–Weiss or Fenton conditions with a suitable Fe(II) additive and hydrogen peroxide liberates the methyl radical from the parent radical precursor. Even though this process is not efficient in terms of yield, its striking operational simplicity allowed studying the important methylation of nucleic acid bases under conditions inspired by Nature.¹ Remarkably, the introduction of a single methyl group into a molecular scaffold can drastically alter a compound's biological profile, typically noted as the magic methyl effect.² As a consequence, synthetic organic chemists have dedicated considerable efforts towards expanding and optimizing the arsenal of radical methylation strategies available. For example, Baran and co-workers reported the use of methyl group surrogate radicals (PhSO₂CH₂˙) accessed by oxidative activation,³ and Antonchick and co-workers successfully modified the Fenton chemistry approach to achieve introduction of the CD₃ motif on synthetically relevant scales.⁴ Similar to the methyl radical, the installation of the trifluoromethyl analogue has garnered significant attention, especially in the past decade. Thus, the sagacious introduction of a single trifluoromethyl group into a organic scaffold typically increases the lipophilicity of a drug candidate (π = 0.88)^5,6 and together with this moiety's chemical inertness, has spurred the design of a host of matched precursor/activator pairs to provide F₃C˙ radicals (Scheme 1).⁷ Two general paradigms exist for the unfettering of the reactive intermediate: (1) reductive activation (RA) of a CF₃-reagent is often employed with photochemical strategies and exploits readily available bulk chemicals such as trifluoroacetic acid derivatives and sulfonyl chlorides⁸ whereas (2) oxidative activation (OA) depends on sulfoxide or sulfinate derivatives, e.g. Langlois’ reagent, which are also inexpensive and readily available. The latter strategy is widely adaptable and by tuning of the sulfinate's substituents, can readily provide access to other alkyl radicals, as most recently demonstrated by Baran and co-workers.⁹ Activation of the sulfoxide motif is most often achieved using tBuOOH as stoichiometric reactant and apart from an isolated example by Antonchick and co-workers,¹⁰ we are not aware of a hydroxyl radical being used as a promoter in Fenton-like chemistry. Furthermore, the direct application of the largely unknown, fully fluorinated analogue of the common laboratory solvent, dimethyl sulfoxide, DMSO-F₆ (1), as radical precursor to F₃C˙, i.e. in analogy to DMSO serving as precursor to H₃C˙, is conspicuously absent from the literature. Specifically, in the context of radiolysis experiments, which are carried out in water, the HO˙ radical constitutes one of the primary products¹¹ and DMSO-F₆ could potentially provide a convenient entry point into studying the kinetics of F₃C˙ radicals.

Scheme 1 General mechanistic paradigm for radical production (I), specific case for H₃C˙ generation (II), common precursors to the trifluoromethyl radical F₃C˙ and subject of the present study (IV).

Moreover, DMSO-F₆ would not necessitate specialized handling or additives as required for previously employed gaseous CF₃I or Togni's reagents,¹² nor would we expect any undesired cross-reactivity with commonly employed metal salts. Furthermore, the abovementioned commercial zinc sulfinate radical precursors are of limited use due to the formation of a precipitate with K₄[Fe^II(CN)₆],¹³ one of the standard indicators for HO˙ radicals. Herein, we explored the chemical kinetics of DMSO-F₆ with primary radiolysis products, juxtapose these values with parameters available for DMSO in the literature, and perform preliminary activation test of DMSO-F₆ with Fenton's reagent as a foundational study for an alternative radical trifluoromethylation reagent.

Materials and methods

Synthesis of DMSO-F₆

The synthesis of perfluorodimethylsulfoxide, DMSO-F₆, is based on a sole literature report from 1972.^14,15 However, the reported procedure (A) (requiring heating substantially above the boiling point of the product, and a substantial but delayed exotherm) required significant improvements to negate an unsafe and questionable protocol, to remove ambiguity, and update key spectroscopic details of DMSO-F₆. Spectroscopic examination only available by extended ¹³C NMR experiments revealed that product A was only about 70% estimated purity (measured against an internal standard by ¹⁹F NMR with a pulse delay value, d₁, of 10 s) using the reported procedure as the by-products including TMS-F and Me₃SiOMe form an inseparable ternary azeotropic mixture. We were not able to purify the mixture properly by any number of fractional distillations or liquid–liquid extraction. Therefore, an alternative synthesis strategy (B) was developed which led to the desired product with a purity of >95% based on GC-MS analysis (Scheme 2).

Scheme 2 Improved synthesis of DMSO-F₆.

Thus, a 350 mL oven-dried Ace Glass pressure vessel was charged with CsF (0.978 g, 6.44 mmol, 0.05 equiv.) and a large pTFE stirbar and subsequently dried at 200 °C at 1 × 10⁻² mbar for 12 h and then backfilled with dry argon. The reaction vessel was cooled to RT and then to −20 °C in an EtOH bath by means of externally controlled circulator. 1,3,2-Dioxathiolane 2-oxide (14.09 g, 130 mmol, 1.0 equiv.) and TMSCF₃ (38.20 g, 269 mmol, 2.1 equiv.) were added under an argon atmosphere. The flask was sealed and the contents were stirred at −20 °C for 1 h, warmed to RT and stirred for an additional 22 h behind a blast shield. To prevent loss of the volatile DMSO-F₆ during post-reaction workup, the light-yellow crude reaction mixture was cooled to 0 °C, and the contents transferred to a distillation apparatus containing a 25 cm rectifying column. The product was fractionally distilled twice, collecting the fraction with a b.p. of approx. 34 °C at 760 mmHg to afford hexafluorodimethylsulfoxide (DMSO-F₆) as a colourless, volatile liquid (16.63 g, 69% yield) still containing trace trimethylsilyl impurities (ca. <5% by GC-MS). ¹H NMR = product contains no protons. ¹³C NMR (101 MHz, CDCl₃) δ 123.59 ppm (qm, ¹J_CF = 338 Hz). Residual TMSF is observed at −0.1 ppm (d, ²J_CF = 15.3 Hz). ¹⁹F{¹H} NMR (282.38 MHz, CDCl₃, 298 K, int. ref. to 2.5% w/w CFCl₃) −67.45 ppm (s, 6F) (previously reported at 64.5 ppm (ref. 14)). IR (ATR-diamond, cm⁻¹): 942, 955, 1100, 1119, 1182, 1244. GC-MS (EI, 70 eV) and IR were also previously reported by Shreeve and are consistent with our observations (see ESI† for complete analysis).¹⁴

Most experiments with DMSO-F₆ were carried out with product derived from Method A. If the contaminants were less reactive than DMSO-F₆, then kinetics results are expected to be essentially unaffected. Alternatively, if the trimethylsilyl fluoride contaminants were significantly more reactive than DMSO-F₆, then the derived reaction rates will be too high. Controls (section 3) with product from Method B (see below) yielded comparable kinetics results as observed with product A. In keeping with expectations, rates derived with product A are smaller than rates derived with product B.

Vapor pressure of DMSO-F₆

The unbuffered sample solutions were thoroughly degassed and then saturated with a specific gas or gas mixture before addition of DMSO-F₆. To this end, 50 μL of the reagent (product A or B) were added to the closed vessel containing 10 mL of sample solution (ρ = 1.42 g cm⁻³ (see ESI†)), thereby minimizing losses by evaporation. The solution was then taken up with a Sample-Lock Syringe (Hamilton) and introduced into the irradiation cell by a syringe pump through PEEK tubing. DMSO-F₆ is highly volatile. We assumed that we observe the vapor pressure over a binary mixture of liquids, i.e. DMSO-F₆ in water. In this case, the amount of DMSO-F₆ in solution is dependent on its vapor pressure and the volume of the gas-phase. The assumption was validated and the vapor pressure derived as follows:

(A) The concentration of DMSO-F₆ remaining in the solution is proportional to its scavenging power. This was derived by appropriate competition experiments (section 3) with solutions prepared in a Schlenk tube with known volume. Three experiments were carried out with an identical solution. For each experiment, 10 ml solution were put in the same Schlenk tube and saturated with N₂O. Then, 0 μL, 50 μL or 100 μL DMSO-F₆ were added. Therefore, 0 mM DMSO-F₆, x mM DMSO-F₆, and (x + 38) mM DMSO-F₆, respectively, are present in the solutions. The (38 – x) mM DMSO-F₆ end up in the gas-phase. With the known volume of the Schlenk tube we estimated a vapor pressure of approx. 250 mbar at 22 °C.

(B) A 10 ml solution of 3 mM Ferrocyanide (K₄[Fe(CN)₆]), saturated with N₂O, was spiked in a gas-tight sample-lock syringe (Hamilton) with 50 μl DMSO-F₆. This corresponds to a concentration of 38 mM, and the derived rate constants k₄ and k₇ agree with the rate constants derived above (see section 3).

(C) 10 ml water at 24 °C was saturated in a Schlenk tube with N₂O to a total gas-pressure of 930 mbar. Then subsequently two portions of 50 μl DMSO-F₆ each were added. The total pressures in the Schlenk-tube measured after the additions were 1230 mbar and 1234 mbar, again in agreement with our assumption of a liquid–vapor equilibrium.

Methods

Radicals were generated by pulse-irradiation of unbuffered aqueous solutions with ionizing radiation (2 MeV-electrons) and the products of this pulsed radiolysis are known (eqn (1)).¹¹

The applied radiation deposits energy mass-proportionally and therefore, in dilute solution, all energy is transferred to the solvent. The product distribution and the yield of water radiolysis are known and depend on the applied dose. Specifically, the yields (“G-values”) are G(e_aq⁻) = 2.65, G(HO˙) = 2.65, G(H⁺) = 2.65 and G(H˙) = 0.55 with G-values given in species per 100 eV deposited dose. If the sample solution is saturated with N₂O prior to radiolysis (at 298.15 K: χ₁ = 4.367 × 10⁻⁴, [N₂O]_sat = 24.2 mM),¹⁶ then the solvated electrons e_aq⁻ can also be converted to HO˙ according to eqn (2), thereby doubling the yield of this oxidizing species.

N₂O + e_aq˙⁻ + H⁺ → HO˙ + N₂ (2)

Note, that if other fast reactions with e_aq˙⁻ do occur, they may compete with reaction (2). In particular and in analogy with other halogenated substances and carbonyl-derivatives, DMSO-F₆ is also expected to react quickly with e_aq˙⁻. As a consequence, for our kinetics analyses we aim for a ratio (k(N₂O + e_aq˙⁻) × [N₂O]_sat)/(k(DMSO-F₆ + e_aq˙⁻) × [DMSO-F₆]) ≥ 10, which correspond to >90% of the e_aq˙⁻ being scavenged by N₂O.

In certain experiments tBuOH was used as a HO˙–scavenger to avoid interferences by this oxidizing species (eqn (3)).

(CH₃)₃COH + HO˙ → (CH₃)₂C(OH)CH₂˙ + H₂O (3)

The rate constant of the reaction of a CF₃˙-precursor with a radical was always determined by competition with an indicator reaction. As example see reaction (4) and competitor in reaction (5). The product yield is given by eqn (6a) and results are shown in Fig. 1B. The data were not linearized for analysis, because such treatment would amplify the influence of measurement errors. Instead, a least squares fit according to eqn (6b) was performed by variation of the parameter r. Thus, errors were dominated by the uncertainties in the rate constants used as references.

Fig. 1 (A) Temporal evolution of dose-normalized absorbance at 605 nm after irradiation (10 Gy) of an argon saturated solution of methyl viologen MV²⁺ (135 μM) and tBuOH (10%, 1.1 mM) in absence (black) and presence (maroon) of DMSO-F₆ (14.5 mM). (B) The ratio [MV˙⁺]/[MV˙⁺]₀ varies non-linearly with the ratio [DMSO-F₆]/[MV²⁺]. At [DMSO-F₆]/[MV²⁺] = 35.2, half the solvated electrons e_aq⁻ are intercepted by DMSO-F₆. In light grey, predicted ratios [MV˙⁺]/[MV˙⁺]₀ for DMSO based on kinetic data available in the literature. Error bars represent two standard deviations.

Results

Reaction of DMSO-F₆ with e_aq˙⁻

The rate constant k₄ was determined by competition with methyl viologen (MV²⁺), reaction (5), observing the intensely blue colored methyl viologen radical (MV˙⁺) at 605 nm (ε₆₀₅ = 1.31 × 10⁴ M⁻¹ cm⁻¹).¹⁷

DMSO-F₆ + e_aq˙⁻ → products (4)

MV²⁺ + e_aq˙⁻ → MV˙⁺ (5)

Addition of DMSO-F₆ to a solution of MV²⁺ suppresses the formation of MV˙⁺ (Fig. 1) and competition predicts

[MV˙⁺] = [MV˙⁺]₀ × (k₅[MV²⁺])/(k₅[MV²⁺] + k₄[DMSO-F₆]) (6a)

[MV˙⁺]/[MV˙⁺]₀ = (r[MV²⁺])/(r[MV²⁺] + [DMSO-F₆]) (6b)

r = k₅/k₄ (6c)

with [MV˙⁺]₀ the yield in absence of DMSO-F₆. The product concentration was measured as temporal average over 4–6 μs after the pulse. When 14.5 mM DMSO-F₆ was introduced to an argon saturated solution containing 135 μM MV²⁺ and 10% tBuOH, the absorbance at 605 nm was lowered by 74% from 23.8 to 6.3 mAbs Gy⁻¹ (Fig. 1A). Variation of [MV²⁺] (0.135, 0.24, 0.35, 1.05 mM) at constant [DMSO-F₆] leads to Fig. 1B and to k₅/k₄ = 35.2 with a least squares fit of eqn (6b). With k₅ = (5.4 – 9.0) × 10¹⁰ M⁻¹ s⁻¹,¹⁸ we derive k₄ = (1.5 – 2.6) × 10⁹ M⁻¹ s⁻¹.

Based on the rate constant k₂ = (8.0 – 9.6) × 10⁹ M⁻¹ s⁻¹ and the solubility of N₂O in water at 298.15 K, [N₂O]_sat = 24.2 mM, we calculate that 84–91% of the solvated electrons are scavenged by N₂O in presence of 14.5 mM DMSO-F₆,^19a,20 in close agreement to our aim. In consequence, G(HO˙) = 4.87–5.07 and G(reaction (4)) = 0.23–0.43. Note, however, that this does not affect our competition experiments because the branching ratio for electrons through reactions (2) and (4) was always constant if DMSO-F₆ was used.

Reaction of DMSO-F₆ with HO˙

The rate constant k₇ was determined by competition with hexacyanoferrate(4-) (“ferrocyanide”) and hexachloroiridate(3-) (Fig. 2, maroon and green, respectively). Each data point is an average of >4 single determinations:

DMSO-F₆ + HO˙ → [(CF₃)₂S(OH)O]˙ (7)

Fe^II(CN)₆⁴⁻ + HO˙ → Fe^III(CN)₆³⁻ + HO⁻ (8)

Ir^IIICl₆³⁻ + HO˙ → Ir^IVCl₆²⁻ + HO⁻ (9)

Fig. 2 Relative yield of reference oxidation reaction for different [Precursor]/[Reference] ratios. Precursors to F₃C˙ employed as competitors: DMSO-F₆ (magenta, green), Langlois’ reagent (black) and trifluoroacetate (light and dim grey). Error bars represent two standard deviations.

DMSO-F₆ (14.5 mM) was added to aqueous, unbuffered and N₂O saturated solutions of Fe^II(CN)₆⁴⁻ (0.11–1 mM). After pulse irradiation with doses of 10–20 Gy yields were compared to respective measurements with 0.11 mM Fe^II(CN)₆⁴⁻ in the absence of DMSO-F₆. The reaction was followed spectroscopically at 420 nm (ε₄₂₀(Fe^III(CN)₆³⁻) = (0.9 − 1.1) × 10³ M⁻¹ cm⁻¹) and the final absorption (“yield”) was determined as a temporal average over 4–8 μs after the pulse.²¹ Alternatively, 14.5 mM DMSO-F₆ was added to aqueous, unbuffered, N₂O saturated solutions of 97, 291 and 873 μM Ir^IIICl₆³⁻. Solutions were pulse-irradiated, the kinetics was followed at 435 nm and plotted with reference to a solution of 97 μM Ir^IIICl₆³⁻ in the absence of reagent. Both, reaction (8) and (9), are diffusion controlled with k₈ = (0.92 − 1.1) × 10¹⁰ M⁻¹,²² and k₉ = (0.47–1.3) × 10¹⁰ M⁻¹.²³ Given k₈/k₇ = 4.8 and k₉/k₇ = 6.4 (Table 1), we derive k₇ = (0.73–2.5) × 10⁹ M⁻¹ s⁻¹. The uncertainty in k₇ originates from the large uncertainty in k₉ (see Table 1) and, therefore, the upper limit of the given range has the higher probability of being correct.

Table 1 Compilation of literature reference rates, k(Ref. + HO˙), and thereof derived rate constants for oxidative activation of F₃C˙ precursors, k(P + HO˙)

Precursor (P)	Reference (Ref.)	k(Ref. + HO˙)/10^10 M^−1 s^−1	k(Ref. + HO˙)/k(P + HO˙)	k(P + HO˙)/M^−1 s^−1
DMSO-F6	Fe^II(CN)6^4−	0.92–1.1 (ref. 22)	4.8	(1.9–2.5) × 10^9
	Ir^IIICl6^3−	0.47–1.3 (ref. 23)	6.4	(0.73–2.0) × 10^9
F3CSO2^−	Fe^II(CN)6^4−	0.92–1.1 (ref. 22)	2.0	(4.6–5.5) × 10^9
F3CCO2^−	Fe^II(CN)6^4−	0.92–1.1 (ref. 22)	15 140	(6.1–7.3) × 10^5
	SCN^−	1.4 (ref. 24)	7460	1.8 × 10^6

---

At longer timescales we observe in both cases additional processes, which we cannot explain quantitatively (see ESI†). The processes are clearly dose-dependent, suggestive for involvement of recombination reactions, i.e. reactions of/with products. Reaction (7) will be followed most probably by a fragmentation and we assume that CF₃S(O)O⁻ and F₃C˙ are formed (see below, reaction (14)). The sulfinate is a reducing agent and F₃C˙ is a moderately potent oxidant. It is not surprising that such species would further react in a mixture with the partially oxidized competitors and, therefore, induce concentration change of oxidized indicator.

Controls with product B

Competition according to section 1 was reproduced and r = 17.9 was derived, i.e. k₄ = (3 − 5) × 10⁹ M⁻¹ s⁻¹. For the determination of the vapor pressure and the concentration of DMSO-F₆ in a given Schlenk-tube, we measured the yield of Fe^III(CN)₆³⁻ (see above). Experiments were carried out with 0, 50 μl and 100 μl DMSO-F₆. With N₂O saturated solution of 3 mM Fe^II(CN)₆⁴⁻ we derived k₄ using the branching of solvated electrons via reactions (2) and (4). We observe a two stage reaction, a fast initial reaction followed by a slower process of pseudo-first order with k_obs = 2 × 10⁴ s⁻¹. Under the assumption that the total yield G(Fe^III(CN)₆³⁻)_tot = G(OH˙) and that G(OH˙) + G(e_aq˙⁻) = 5.8²⁵ we estimate k₄ = (3.4 − 4.1) × 10⁹ M⁻¹ s⁻¹. This assumption implies also, that the initial, fast absorption increase is due to reaction (8) and the slower is a consequence of reaction (7). Therefore, G(Fe^III(CN)₆³⁻)_initial/G(Fe^III(CN)₆³⁻)_tot = k₈[Fe^II(CN)₆⁴⁻]/(k₈[Fe^II(CN)₆⁴⁻] + k₇[DMSO-F₆]). We derive a rate constant of k₇ = (0.83 − 1.0) × 10⁹ M⁻¹ s⁻¹. If 10 ml of a N₂O saturated solution of 3 mM Fe^II(CN)₆⁴⁻ was directly spiked in the sample-lock syringe with 50 μl DMSO-F₆, we derive k₄ = (1.7 − 4.7) × 10⁹ M⁻¹ s⁻¹ and k₇ = (0.63 − 2.0) × 10⁹ M⁻¹ s⁻¹.

Reaction of CF₃SO₂⁻ (Langlois’ reagent) with HO˙

The reaction rate was measured in competition with reaction (8):

CF₃SO₂⁻ + HO˙ → [CF₃SO₂OH] ˙⁻ (10)

Unbuffered, N₂O saturated, aqueous solutions of 10.4 mM CF₃SO₂Na and varying concentration of K₄Fe^II(CN)₆ were pulse-irradiated with doses of 10–20 Gy. Two independent experimental series were measured, (1) [K₄Fe^II(CN)₆] = 0.4–6.3 mM and (2) [K₄Fe^II(CN)₆] = 0.63–3 mM. As reference a solution of 1.1 mM K₄Fe^II(CN)₆ was used. Results are shown in Fig. 2 (black) and for the obtained value of r = 2 we arrive at k₁₀ = (4.6–5.5) × 10⁹ M⁻¹ s⁻¹ (Table 1).

Also with Langlois’ reagent we observed reactions at later times, however, the reproducibility of those measurements was unsatisfactory. We suspect a non-negligible influence of impurities on our measurements at times >20 μs after pulse. While the corresponding Zn-salt is also commercially available and is typically provided in much better quality, its use is prohibited by the formation of Zn²⁺ precipitates with ferrocyanide (vide infra).¹³

Reaction of CF₃CO₂⁻ with HO˙

The rate constant of this reaction was measured versus ferrocyanide and thiocyanate:

CF₃CO₂⁻ + HO˙ → F₃C˙ + CO₂ + HO⁻ (11)

SCN⁻ + HO˙ → SCN˙ + HO⁻ (12)

SCN⁻ + SCN˙ → (SCN)₂˙⁻ (13)

The ratio k₈/k₁₁ = 15 140 was determined in unbuffered, N₂O saturated aqueous solutions containing either 0.11 mM K₄Fe^II(CN)₆/0.5 M CF₃CO₂⁻ or 0.05 mM K₄Fe^II(CN)₆/1 M CF₃CO₂⁻ (Fig. 2, light grey). As reference, a solution of 0.11 mM K₄Fe^II(CN)₆ was chosen.

Thiocyanate is often used as dosimeter in pulse radiolysis, reaction (13) is an equilibrium reaction, and (SCN)₂˙⁻ has a known molar absorptivity of ε₄₇₅((SCN)₂˙⁻) = 7580 M⁻¹ cm⁻¹.²⁶ Unbuffered, N₂O saturated solutions of 510 mM CF₃CO₂⁻ and 0.05–5 mM KSCN were pulse-irradiated and k₁₂/k₁₁ = 7460 determined. It is noteworthy that (pseudo-)halogenide radicals tend to form complexes with anions, e.g. reaction (13), and for chlorine atoms even the diffusion-controlled reaction with hydroxide is described.²⁷ Possibly, SCN˙ will react with the 0.5 M carboxylate present and such an equilibrium would compete with equilibrium (13) resulting, in turn, in an overestimation of k₁₁.

2,2′-Azino-bis(3-ethylbenzothiazoline-6-sulphonic acid (ABTS²⁻) as indicator for the reaction mechanism

One-electron oxidation of ABTS²⁻ produces a strongly colored radical, ABTS˙⁻, with ε₄₁₄ = 3.6 × 10⁴ M⁻¹ cm⁻¹.²⁸ It is a convenient reagent to test for a moderately or strongly one-electron oxidizing species in a solution. We suspect [(CF₃)₂S(HO)O]˙ (reaction (7)) to be of low reactivity and wished to gain kinetics information on reaction (14). Reaction (16) is very fast, k₁₆ = (1.2–1.9) × 10⁹ M⁻¹ s⁻¹, and we suspect k₁₇ to be of similar magnitude, large enough to possibly monitor reactions (14) and/or (15).

[(CF₃)₂S(OH)O]˙ → CF₃SO₂⁻ + F₃C˙ + H⁺ (14)

F₃C˙ + O₂ → F₃COO˙ (15)

ABTS²⁻ + Cl₃COO˙ → ABTS˙⁻ + Cl₃COOH + HO⁻ (16)

ABTS²⁻ + F₃COO˙ → ABTS˙⁻ + F₃COOH + HO⁻ (17)

ABTS²⁻ + HO˙ → ABTS˙⁻ + HO⁻ (18)

H˙/e_aq˙⁻ + O₂ → O₂˙⁻/HO₂˙ (19)

Solutions were saturated with a gas mixture of N₂O : O₂ ≈ 5 : 1. This will change the yields of the different radicals compared to experiments with N₂O saturation, as we have an additional reaction (19). If gas-saturation were perfectly reproducible, no variation of starting conditions after the pulse is expected. We measured DMSO-F₆ (14.5 mM) in presence of ABTS²⁻ (59, 98, 137 μM) and, as a reference in absence of DMSO-F₆, we used 98 μM ABTS²⁻ (Fig. 3, black). For this latter case, we determined k₁₈ = 1.3 × 10¹⁰ M⁻¹ s⁻¹, in agreement with the reported value.²⁸ The initial absorption of the kinetics traces, directly after the pulse, increases with increasing [ABTS²⁻], consistent with the aforementioned rate constant k₁₈. Also, we observe that the yield of [ABTS˙⁻] in reaction (18) is only about 57%, in full agreement with the work of Willson.²⁸

Fig. 3 Normalized temporal evolution of absorbance at 404 nm for ABTS²⁻ solutions saturated in N₂O : O₂ = 5 : 1 after irradiation with 10 Gy. Addition of DMSOF₆ (14.5 mM) to the ABTS²⁻ solution (98 μM) leads to a change in overall kinetics from pseudo-first order (black) to consecutive reactions (salmon). Rate of oxidation depends on the initial ABTS²⁻ concentration and increases in the order 59 μM (maroon), 98 μM (salmon) and 137 μM (purple). Inset: magnified area as encompassed by black rectangle to highlight initial offset and sigmoidal curve shape. Colored line segments are model fits.

The gas was mixed manually for every stock solution. We observed slightly increased yields of ABTS˙⁻ with each new solution. The effect was not dependent on the ABTS²⁻ concentration, as the different concentrations were measured in random order. The total signal increase over the whole day was approximately 15% (see ESI†). We judged that we were observing an artifact and, therefore, normalized the curves for the same end-absorption.‡ Normalization does not influence the derived rate constants. Kinetics traces can be seen in Fig. 3. The kinetics change dramatically in the presence of DMSO-F₆, and there are three important qualitative observations: (i) the reactions are distinctly slower, (ii) instead of a 1^st order growth we observe a lag-phase followed by an absorption increase to the level observed before and (iii) the rate of increase is dependent on [ABTS²⁻]. The lag-phase (ii) is typical for multi-step reactions, in support of our assumptions. We infer to observe consecutive reactions of reaction (7), presumably reactions (14), (15) and (17). It even seems, that the total yield of ABTS˙⁻ is independent whether ABTS²⁻ is oxidized directly via reaction (18) or indirectly via a reaction cascade starting at reaction (7).

Based on these qualitative observations we set out to model the kinetics traces. Again, we suspect a reaction cascade involving reactions (7), (14), (15) and (17). For our simulation of the shape of the curve, reaction (7) is irrelevant for the kinetics because it has a half-live of only t_½ = ln(2)/(k₇ × [DMSO-F₆]) < 80 ns (for k₇ > 0.6 × 10⁹ M⁻¹ s⁻¹). We know that reaction (17) is relevant, because the rate of absorption build up in Fig. 3 (color) is clearly dependent on [ABTS²⁻].

Because we want to limit the number of parameters in our model, we tried to model reactions (14) and (15) together using one single first order rate constant k_calc. With [O₂] ≫ [F₃C˙], this would imply that either k_calc = k₁₄ ≫ k₁₅[O₂] or k_calc = k₁₅[O₂] ≫ k₁₄. In the case that k_calc = k₁₅[O₂], the process of gas saturation would govern the results. The oxygen concentration, [O₂], in the samples were not analytically quantified between samples of DMSO-F₆ and therefore different oxygen concentrations could additionally provide a rationale for slightly different values observed between experimental series (Table 2). It is foreseeable that measured values for reactions (14) and (17) will also be subject to the analytical quality of DMSO-F₆ used.

Table 2 Compilation of model parameters for modeling the system comprising of eqn (14), (15) and (17)

[ABTS^2−]/μM	k calc /10^5 s^−1	k 17 ^b/10^9 M^−1 s^−1	Offset/10^−3 Gy
a First order intermediate step. b Standard deviations based on fitting N ≥ 5 kinetics traces. c Experiments performed with product A. d Experiments performed with product B.
59^c	2.7 ± 0.7^b	1.5 ± 0.3^b	2.4
98^c	2.7 ± 0.2^b	1.2 ± 0.0^b	5.6
137^c	3.6 ± 0.4^b	1.2 ± 0.1^b	4.0
68^d	0.6 ± 0.02^b	2.9 ± 0.1^b	2.7
125^d	0.9 ± 0.08^b	2.3 ± 0.2^b	4.5
250^d	1.1 ± 0.02^b	2.6 ± 0.1^b	10

---

The corresponding least-square fitted curves can be seen in Fig. 3, the corresponding parameters are shown in Table 2.

Fenton reaction in presence of DMSO-F₆ and product analysis

We characterized the radical species generated by oxidative activation by its reaction with caffeine. This xanthine is an ideal substrate: it is water soluble, features only one reactive site for radical addition and is reminiscent of purine nucleobases. Thus, the well-established oxidative activation of DMSO by the Fenton protocol for methylating adenine and guanidine should provide an ideal blue print.²⁹ To mixtures of caffeine, ascorbic acid, Na₂EDTA and Fe(II)SO₄·7H₂O in phosphate buffer we added at 0 °C a DMSO derivative followed by dropwise addition of H₂O₂. Products were extracted with ethyl acetate and analyzed by GC-MS (Fig. 4). Besides unreacted starting material (t_R = 11.0 min; 194m/z, grey) only the functionalized products occur, i.e. CF₃-caffeine (t_R = 9.8 min; 262m/z, green) or CH₃-caffeine (t_R = 11.6 min; 208m/z, magenta), as well as minor amounts of a decomposition product likely arising from oxidative imidazole cleavage (t_R = 11.6 min, 112m/z). The product CF₃-caffeine was quantified by ¹⁹F NMR spectroscopy through addition of PhCF₃ as internal standard to the organic extract and furnished a value of approximately 16%.

Fig. 4 GC-MS chromatograms of organic extracts of oxidative caffeine functionalization in aqueous, buffered media with Fenton's reagent and DMSO (A) or DMSO-F₆ (B).

Discussion

Rates constants for the reaction of DMSO-F₆ with e_aq˙⁻ and HO˙

Perfluorination has two distinct effects on the reactivity of DMSO. The inductive effect exerted by fluorine will decrease electron density around the sulfur center. Additionally, the C–F bond is easier to reduce than the C–H bond because of the very low lying σ_CF* orbital,³⁰ and the resulting instability of perfluorinated materials towards strongly reducing conditions is well documented.³¹ The reaction of the solvated electron with DMSO-F₆, k₄ = (2 − 5) × 10⁹ M⁻¹ s⁻¹, is three orders of magnitude larger than with ordinary DMSO (calculated curve depicted in light grey in Fig. 1B). Conversely and as inferred, the oxidation of DMSO-F₆ with HO˙, k₇ = (0.6 − 2.5) × 10⁹ M⁻¹ s⁻¹, is an order of magnitude slower than that of DMSO.³²

Reagents for production of F₃C˙ radicals in pulse radiolysis

Experiments with Langlois’ reagent and trifluoroacetate yielded unsatisfactory results for different reasons but serve as valuable references due to the reagents’ commercial availability. The oxidation of trifluoroacetate with HO˙ is simply too slow for fast production of F₃C˙ needed in time-resolved mechanistic studies. We find k₁₁ = (6 − 7) × 10⁵ M⁻¹ s⁻¹ in agreement with published data from flash photolysis experiments (<10⁶ M⁻¹ s⁻¹).³³ Langlois’ reagent is kinetically ideal for our purposes, k₁₀ = 5 × 10⁹ M⁻¹ s⁻¹. The corresponding Zn-salt (Baran modification) is stable and available in high purity, however, Zn²⁺ has a rich complex chemistry that distinctly limits the use of this compound. In our case, the low solubility of Zn[Fe(CN)₆] precluded certain measurements.¹³ On the other hand, the sodium salt proved to be either unstable or not pure: experiments yielded distinctly differing kinetics traces for timescales >20 μs from day to day, which is not acceptable for future mechanistic investigations. In addition, earlier experiments with reductive activation of CF₃I showed several disadvantages associated with this gas, such as the necessity of signal deconvolution.³⁴

DMSO-F₆ has the drawback of not being commercially available. Its rate of reaction with HO˙ is lower than that of Langlois’ reagent by half an order of magnitude. On the other hand, liquid handling is comparatively easy, and the reagent proved to be robust in daily use. This makes DMSO-F₆ the currently best, though not optimal, choice for time-resolved mechanistic studies. We therefore investigated the mechanism of the oxidative activation more closely, because in principle, fragmentation of the oxidized intermediate [(CF₃)₂S(OH)O]˙⁻ may not only occur via reaction (14), but also via reactions (20) and (21), also producing a moderately oxidizing radical.

[(CF₃)₂S(OH)O]˙ → [(CF₃)₂S(O)O]˙⁻ + H⁺ (20)

[(CF₃)₂S(O)O]˙⁻ → CF₃SO₂˙ + F₃C⁻ (21)

Kinetics experiments as well as product analysis were carried out to confirm the hypothesis of fragmentation via reaction (14). Product analysis, demonstrating trifluoromethylation of caffeine after oxidative activation of DMSO-F₆ by the Fenton reagent, indeed supports the notion of F₃C˙ radical production, reaction (14), very clearly, albeit at low yield. In this regard, it is noteworthy that the addition of ascorbate proved critical to observing the functionalized product. However, ascorbate (“antioxidant”) may reduce the F₃C˙ radical, thereby preventing oxidative functionalization of caffeine.³⁵ Because perfluoroalkyl radicals exhibit a very low molar absorptivity, we decided to monitor kinetics with a reporter molecule, ABTS²⁻. In the presence of oxygen, we observe a double-exponential behavior (Fig. 3), i.e. two sequential (pseudo)-first order reactions can be resolved. In aqueous environment F₃C˙ and Cl₃C˙ radicals have a similar electronegativity.^34,36 Similarly, F₃COO˙ and Cl₃COO˙ show comparable kinetic behavior.³⁷ While the former is a stronger oxidant, the overall reactivity of both molecules is dominated by the electron-withdrawing effect exerted by the halogen substitution pattern. With the ascorbate anion (trihalogen)methylperoxyl radicals have a reactivity H₃COO˙ ≪ Cl₃COO˙ ≤ F₃COO˙ with corresponding relative rate constants of 1 : 100 : 100.³⁸ For the reaction with Trolox C the corresponding rate constants have a relative magnitude of 1 : 2400 : 4700. Oxygen will cause formation of F₃COO˙ radicals, reaction (15), which we expect to exhibit a comparable reactivity as Cl₃COO˙, k₁₆ = (1.2 − 1.9) × 10⁹ M⁻¹ s⁻¹.²⁸ This is indeed the case: k₁₇ = (1 − 3) × 10⁹ M⁻¹ s⁻¹. (Table 2) Is the remaining rate constant, with k_calc = (0.5 − 3) × 10⁵ s⁻¹, to be attributed to reaction (14) or reaction (15)? Based solely on our data we cannot decide unequivocally. Nevertheless, reaction (15) is the more probable candidate for two reasons: (A) given [O₂] ≈ 200 μM,³⁹ we would calculate k₁₅ = (0.3 − 1.5) × 10⁹ M⁻¹ s⁻¹, in agreement with the corresponding rate constants for the reactions of H₃C˙ and Cl₃C˙ with O₂ of (3.0 − 4.7) × 10⁹ M⁻¹ s⁻¹ and 3.3 × 10⁹ M⁻¹ s⁻¹, respectively.⁴⁰ (B) Compared to a reported value of k₂₁ = 1.5 × 10⁷ s⁻¹ we consider a value of k₁₄ = (0.5 − 3) × 10⁵ s⁻¹ too low.⁴¹

[(CH₃)₂S(OH)O]˙ → CH₃SO₂⁻ + H₃C˙ + H⁺ (21)

The [S → C] substitution from Langlois’ reagent to trifluoroacetate underpins the importance of a nucleophilic central atom to achieve significant scavenging rates of the HO˙ radical. A loss in reactivity of over three orders of magnitudes was observed and this innocuous substitution was likely accompanied by a fundamental change in mechanism. Furthermore, the comparison of trifluoroacetate with acetate is an illustrative example for the often-quoted increase in metabolic stability achieved by the [CH₃ → CF₃] modification. In acetate, α-hydrogen abstraction predominates over oxidation (k(H₃CCO₂⁻ + HO˙) = (7.9 − 10) × 10⁷ M⁻¹ s⁻¹).^22a,33,34

Similarly, the [H → F] substitution in DMSO causes reactivity changes. As abovementioned, perfluorination of DMSO decreases the rate constant of its reaction with HO˙ about half an order of magnitude. The CF₃-group is a relatively strongly σ-withdrawing moiety and may, in addition, allow for stabilizing n → σ_CF* interactions.³⁰ The lone pair at sulfur in DMSO-F₆ may thus be rendered less electron-rich and presumably overall less available than in DMSO. A lower reactivity appears only reasonable. If oxidative activation occurs via addition of the HO˙ radical to sulfur, reaction (7), an additional steric argument should be considered: the electron density at the fluorine atoms gives rise to shielding. This may hinder the trajectory of an incoming radical. The comparison of the behavior of DMSO-F₆ and of Langlois’ reagent suggests that the replacement of a single CF₃ group in DMSO-F₆ fully reconstitutes the reactivity lost due to the perfluorination of native DMSO. However, whether this effect is mostly electronic in nature (e.g. anionic vs. neutral) or whether it also features a (pronounced) steric component cannot be deduced from the data at hand.

Conclusions

In analogy to the oxidative activation of DMSO with HO˙ providing H₃C˙, we explored the feasibility of using DMSO-F₆ as a precursor to F₃C˙. Pulse-radiolysis studies in conjunction to laboratory experiments corroborated that DMSO-F₆ undergoes a rapid reaction with HO˙ with k₇ = (0.6 − 2.5) × 10⁹ M⁻¹ s⁻¹, followed by a fragmentation reaction to furnish F₃C˙. In comparison to other commercially available precursors (Langlois’ reagent, trifluoroacetate), DMSO-F₆ proved stable, easier to handle and overall more robust and therefore, is well suited for time-resolved kinetics studies of F₃C˙. The major caveat is the requirement for its laboratory synthesis and hitherto time-consuming purification.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was generously supported by the Swiss National Science Foundation (N. S.; P3P3P2_167744), the National Science and Engineering Research Council of Canada (B. J. J.) and ETH Zürich (Prof. D. Günther, Prof. A. Togni).

Notes and references

1. (a) J. O. Kang, K. S. Gallagher and G. Cohen, Arch. Biochem. Biophys., 1993, 306, 178–182; (b) K. Kawai, Y.-S. Li, M.-F. Song and H. Kasai, Bioorg. Med. Chem. Lett., 2010, 20, 260–265.
2. H. Schönherr and T. Cernak, Angew. Chem., Int. Ed., 2013, 52, 12256–12262.
3. J. Gui, Q. Zhou, C.-M. Pan, Y. Yabi, A. C. Burns, M. R. Collins, M. A. Ornelas, Y. Ishihara and P. S. Baran, J. Am. Chem. Soc., 2014, 136, 4853–4856.
4. R. Caporaso, S. Manna, S. Zinken, A. R. Kochnev, E. R. Lukyanenko, A. V. Kurkin and A. P. Antonchick, Chem. Commun., 2016, 52, 12486–12489.
5. C. Hansch, A. Leo and D. Hoekmann, QSAR: Volume 2: Hydrophobic, Electronic, and Steric Constants, American Chemical Society, Washington DC, 1995.
6. K. Müller, C. Faeh and F. Diederich, Science, 2017, 317, 1881–1886.
7. A. Studer, Angew. Chem., Int. Ed., 2012, 51, 8950–8958.
8. (a) J. W. Beatty, J. J. Douglas, K. P. Cole and C. R. J. Stephenson, Nat. Commun., 2015, 6, 7919; (b) N. Kamigata, T. Fukushima and M. Yoshida, Chem. Lett., 1990, 19, 649–650; (c) Y. Ouyang, X.-H. Xu and F.-L. Qing, Angew. Chem., Int. Ed., 2018, 57, 6926–6929.
9. Y. Fujiwara, J. A. Dixon, F. O'Hara, E. D. Funder, D. D. Dixon, R. A. Rodriguez, R. D. Baxter, B. Herlé, N. Sach, M. R. Collins, Y. Ishihara and P. S. Baran, Nature, 2012, 492, 95–99.
10. K. Matcha and A. P. Antonchick, Angew. Chem., Int. Ed., 2014, 53, 11960–11964.
11. A. Henglein, W. Schnabel and J. Wendenburg, Einführung in die Strahlenchemie, Verlag Chemie GmbH, Weinheim, 1969.
12. (a) G. Bullock and R. Cooper, Trans. Faraday Soc., 1970, 66, 2055–2064; (b) N. Santschi and T. Nauser, J. Fluorine Chem., 2017, 203, 218–222.
13. G. Jander and E. Blasius, Lehrbuch der analytischen und präparativen anorganischen Chemie, S. Hirzel Verlag, Stuttgart, 1969.
14. (a) From dimethyl sulfite: D. T. Sauer and J. M. Shreeve, J. Fluorine Chem., 1971/72, 1, 1–11; (b) R. Singh, G. Cao, R. Kirchmeier and J. Shreeve, J. Org. Chem., 1999, 64, 2873–2876.
15. A similar protocol from thionyl fluoride, see: N. Patel and R. Kirchmeier, Inorg. Chem., 1992, 31, 2537–2540.
16. W. M. Hanyes, CRC Handbook of Chemistry and Physics, CRC Press/Taylor&Francis, Boca Raton, Florida, 97th edn, 2016.
17. H. Shiraishi, G. V. Buxton and N. D. Wood, Int. J. Radiat. Appl. Instrum., Part C. Radiat. Phys. Chem., 1989, 33, 519–522.
18. (a) M. A. J. Rodgers, D. C. Foyt and Z. A. Zimek, Radiat. Res., 1978, 75, 296–304; (b) A. Horvath, O. Horvath and K. L. Stevenson, J. Photochem. Photobiol., A, 1992, 68, 155–163.
19. (a) A. M. Koulkes-Pujo, B. D. Michael and E. J. Hart, Int. J. Radiat. Phys. Chem., 1971, 3, 333–344; (b) D. Razem and W. H. Hamill, J. Phys. Chem., 1977, 81, 1625–1631.
20. A. J. Elliot, Radiat. Phys. Chem., 1989, 34, 753–758.
21. (a) W. N. Perera and G. Hefter, Inorg. Chem., 2003, 42, 5917–5923; (b) J. J. Alexander and H. B. Gray, J. Am. Chem. Soc., 1968, 90, 4260–4271.
22. (a) R. L. Willson, C. L. Greenstock, G. E. Adams, R. Wageman and L. M. Dorfman, Int. J. Radiat. Phys. Chem., 1971, 3, 211–220; (b) W. L. Waltz, S. S. Akhtar and R. L. Eager, Can. J. Chem., 1973, 51, 2525–2529.
23. (a) C. L. Crawford, M. R. Gholami, S. L. Roberts and R. J. Hanrahan, Radiat. Phys. Chem., 1992, 40, 205–212; (b) N. Selvarajan and N. V. Raghavan, J. Chem. Soc., Chem. Commun., 1980, 336–337.
24. B. H. Milosavljevic and J. A. LaVerne, J. Phys. Chem. A, 2005, 109, 165–168.
25. R. H. Schuler, A. L. Hartzell and B. Behar, J. Phys. Chem., 1981, 85, 192–199.
26. R. H. Schuler, L. K. Patterson and E. Janata, J. Phys. Chem., 1980, 84, 2088–2089.
27. U. K. Kläning and T. Wolff, Ber. Bunsenges. Phys. Chem., 1985, 89, 243–245.
28. (a) B. S. Wolfenden and R. L. Willson, J. Chem. Soc., Perkin Trans. 2, 1982, 805–812; (b) J. E. Packer and R. L. Willson, J. Chem. Soc., Perkin. Trans 2, 1984, 1415–1419.
29. (a) J. O. Kang, K. S. Gallagher and G. Cohen, Arch. Biochem. Biophys., 1993, 306, 178–183; (b) K. Kawai, Y.-S. Li, M.-F. Song and H. Kasai, Bioorg. Med. Chem. Lett., 2010, 20, 260–265.
30. D. O'Hagan, Chem. Soc. Rev., 2008, 37, 308–319.
31. K. Brace, C. Combellas, M. Delamar, A. Fritsch, E. Kanoufi, M. E. R. Shanahan and A. Thiébault, Chem. Commun., 1996, 403–404.
32. D. Veltwisch, E. Janata and K. D. Asmus, J. Chem. Soc., Perkin Trans. 2, 1980, 146–153.
33. P. Maruthamuthu, S. Padmaja and R. E. Huie, Int. J. Chem. Kinet., 1995, 27, 605–612.
34. N. Santschi and T. Nauser, ChemPhysChem, 2017, 18, 2973–2976.
35. T. Nauser and J. Gebicki, Arch. Biochem. Biophys., 2017, 633, 118–123.
36. F. De Vleeschouwer, V. Van Speybroeck, M. Waroquier, P. Geerlings and F. De Proft, Org. Lett., 2007, 9, 2721–2724.
37. R. E. Huie, D. Brault and P. Neta, Chem.-Biol. Interact., 1987, 62, 227–235.
38. (a) X. Shen, J. Lind, T. E. Eriksen and G. Merenyi, J. Phys. Chem., 1989, 93, 553–557; (b) P. Neta, R. E. Huie, P. Maruthamuthu and S. Steenken, J. Phys. Chem., 1989, 93, 7654–7659.
39. R. Sander, Atmos. Chem. Phys., 2015, 15, 4399–4981.
40. (a) A. Sauer, H. Cohen and D. Meyerstein, Inorg. Chem., 1989, 28, 2511–2512; (b) J. K. Thomas, J. Phys. Chem., 1967, 71, 1919–1925; (c) J. Moenig, D. Bahnemann and K. D. Asmus, Chem.-Biol. Interact., 1983, 47, 15–27.
41. R. Herscu-Kluska, A. Masarwa, M. Saphier, H. Cohen and D. Meyerstein, Chem. – Eur. J., 2008, 14, 5880–5889.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c9ob02119a

‡ In the controls, this effect was not observed anymore.

---