A quantitative reactivity scale for electrophilic fluorinating reagents† †Electronic supplementary information (ESI) available. CCDC 1857922–1857928. For ESI and crystallographic data in CIF or other electronic format see DOI: 10.1039/c8sc03596b

Through kinetic studies we provide a quantitative reactivity scale for ten electrophilic fluorination reagents.


Introduction
Organouorine compounds have critically enabling roles in medicinal, agrochemical and material sciences due to the unique properties of the uorine atom. 1 The presence of a uorine atom can impart benecial changes to chemical properties and biological activities of drug molecules, such as improved metabolic stability and enhanced binding interactions. 1 Consequently, pharmaceuticals bearing uoro-aliphatic, -aromatic and -heterocyclic units have become widespread, e.g. ciprooxacin, 5-uorouracil, Prozac™. However, organo-uorine compounds are very scarce in nature; 2 therefore, the selective introduction of a uorine atom is a key challenge in organic chemistry. While uoroaromatic derivatives are synthesised industrially using anhydrous hydrogen uoride and nucleophilic halogen exchange processes that were rst reported a century ago, electrophilic strategies are less wellgrounded. Electrophilic uorination represents one of the most direct methods for the selective introduction of uorine into organic compounds. 1 Early work centred on reagents bearing an O-F bond (e.g. CF 3 OF, HOF, CsSO 4 F) or an Xe-F bond (i.e. XeF 2 ); however, these reagents were oen too reactive, unselective, difficult to prepare and not available commercially-all of which limited their adoption. Molecular uorine (F 2 ) is readily accessible, however, in order to use it safely, specialist equipment and training are required, and these factors limit its general applicability. A breakthrough came in the 1980s, with the introduction of bench-stable electrophilic uorinating reagents containing an N-F bond. 3 These reagents have since emerged as effective, selective and easy-to-handle sources of electrophilic uorine, that are now commercially available and do not require specialized handling procedures.
Electrophilic N-F reagents such as Selectuor™, 4 N-uoropyridinium salts [5][6][7] and NFSI 8 have been widely utilised by the pharmaceutical industry in both discovery and manufacturing processes. However, the choice of reagents for the uorination of a new scaffold at the discovery stage has generally been based on a "trial and error" approach rather than an understanding of reactivities of the electrophilic uorinating reagent and its nucleophilic substrate. Other fundamental transformations such as nitration, alkylation, halogenation, sulfonation and Friedel-Cras processes have been studied extensively by kinetic approaches and predictive reactivity proles for many reagents are well established. [9][10][11][12] Given the importance of uorination reactions in the chemical, pharmaceutical and materials industries, the lack of predictive reactivity data is surprising. We now present a rm kinetic underpinning for these widely-exploited reagents. Umemoto 13 initiated comparative reactivity studies with his power-variable scale for N-uoropyridinium salts, which centred on the electron-donating or electron-withdrawing natures of substituents on the pyridinium rings; however, the approach reected reaction yields rather than kinetic parameters. In 1992, Lal et al. 14 reported reduction potentials, E p , as measures of the relative reactivities of N-F reagents; and others have reported similar studies. 15 Unfortunately, access to data relating to the uorinating strength is oen precluded by experimental problems. Early kinetics studies by Stavber et al. [16][17][18][19] on the uorination of phenols and alkenes with Selectuor™ and Accuuor™ focused on the mechanisms of F transfer rather than reactivity comparisons. Togni and co-workers 20 obtained the relative rate constants of seven N-F reagents for competitive halogenations of b-keto esters in the presence of a titanium catalyst. However, the k rel values captured the whole catalytic cycle rather than individual uorination rate constants. Most recently, a computational reactivity scale was proposed by Cheng et al. 21 based on calculated uorine plus detachment values, however, nucleophiles were not included in the models.
Our strategy focuses on utilising a common nucleophile scaffold for the correlation of the uorinating abilities of N-F reagents. We chose 1,3-diaryl-1,3-dicarbonyls as the nucleophile basis set for our uorination kinetics owing to the ability to enhance or subdue nucleophilicity based on the introduction of electron-donating or -withdrawing substituents. The extended conjugation within these systems offered sensitive spectrophotometric output, where keto and enol tautomers have markedly different absorption proles. We capitalised upon the dominant enol content of the 1,3-diaryl-1,3-dicarbonyl starting materials being consumed during uorination to afford uoroketonic products.
Previous work involving the a-uorination of carbonyl, a 0ketocarbonyl, b-dicarbonyl and related carbonyl derivatives using oxidizing uorinating agents such as uorine, 22,23 XeF 2 , 24,25 alkyl hypouorite, 26 perchloryl uoride 27 and uoroxysulfate 28 generally yielded mixtures of undesirable a,a-diuorinated products in addition to the a-monouorinated products. 24 However, N-F reagents such as N-uoropyridinium salts, NFSI and Selectuor™ have been successfully employed for the selective a-monouorination of carbonyl derivatives. 29,30 Banks et al. rst reported the selective monouorination of 1,3diketones using Selectuor™. 31 An important eld of study that has emerged is the asymmetric a-uorination of carbonyl substrates, which has been explored with both chiral electrophilic uorinating agents and chiral catalysts. [32][33][34][35] Since the synthetic applications of N-F reagents are too numerous to cover in this paper, we refer to the excellent reviews from the recent literature to give an indication of topical uorination reactions. [36][37][38] Furthermore, in general, the uorination of 1,3dicarbonyl derivatives offers a convenient vehicle for the delivery of building blocks for the preparation of uoroaliphatic and -heteroaromatic systems 39 (e.g. voriconazolea billion dollar drug marketed by Pzer 40 ).

Development of the 1,3-diaryl-1,3-dicarbonyl platform
In order to capture the breadth of reactivities of commonly-used N-F reagents, we adopted the 1,3-diaryl-1,3-dicarbonyl derivatives 1a-m. These systems offered the potential to tune nucleophilicity in a predictable manner through the introduction of substituents that could be amenable to Hammett correlation. The 1,3-diaryl-1,3-dicarbonyl derivatives 1a-m (Fig. 1a) were synthesised using previously reported methods, in good yields. 41 Compounds 1a-m exist as mixtures of keto and enol tautomers and the ratio for each system was determined by 1 H NMR spectroscopy in CD 3 CN (see ESI Section 2.6 †). Each tautomer is easily distinguishable, with peaks at $4.5 ppm and $7 ppm corresponding to the keto and enol forms, respectively, and the OH signal of the enol form at $16 ppm. Compounds 1a-m exist in $90% enol form in CH 3 CN, except 1h which exists as $60% enol. Mono-uorinated products 2a-f were synthesised using Selectuor™ (compound 3 in Fig. 1e) and the ratios of tautomers were determined by 1 H and 19 F NMR methods (see ESI Section 2.6 †).
During the recrystallization of the uorinated 1,3-dicarbonyls we found that the keto and enol forms of 2b (R 1 ¼ R 2 ¼ F) and 2c (R 1 ¼ R 2 ¼ Me) crystallized separately from the same solution. For both compounds, the keto and enol tautomers formed white and yellow crystals, respectively (Fig. 1b). On the basis of the colour differences, crystals of each tautomer were picked from the supernatant solution and analysed spectroscopically. We found that both tautomers were stable with respect to tautomerization in CDCl 3 over the course of several days. So-called "tautomeric polymorphs" where tautomers crystallise in different crystal structures 42  The propensity for systems 2b and 2c to produce crystals of both tautomers rests on many kinetic and thermodynamic factors. In order to gauge the inuence of the intrinsic stabilities of each tautomer, calculations were carried out on enol and keto monomers and dimers of 2b using the procedures described elsewhere. 25 The enol form is more stable as a monomer by 2.0 kJ mol À1 but the keto form is more stable as a dimer by 2.0 kJ mol À1 when the dielectric constant of 3 ¼ 3 is applied in the solvent model. The dielectric constant of 3 ¼ 3 is typical in neutral organic crystals. 25 The very small relative energies support the possibility that crystals of both forms may be observed experimentally. The keto forms become more favourable as the solvent polarity (dielectric constant) is increased (see ESI Section 3 †).
With knowledge of the differing keto-enol tautomeric equilibria of starting materials and uorinated products in hand, we anticipated that the 1,3-diaryl-1,3-dicarbonyls should give a convenient nucleophile scaffold on which to base kinetics experiments.

Kinetics studies
Kinetic studies were performed on Selectuor™, NFSI, Syn-uor™, 2,6-dichloro-N-uoropyridinium triate, 2,6-dichloro-N-uoropyridinium tetrauoroborate, 2,3,4,5,6-pentachloro-N-uoropyridinium triate, N-uoropyridinium triate, N-uoropyridinium tetrauoroborate, 2,4,6-trimethyl-N-uoropyridinium triate and 2,4,6-trimethyl-N-uoropyridinium tetrauoroborate (Fig. 1e). All reagents were commercially available, except for 2,3,4,5,6-pentachloro-N-uoropyridinium triate 9, which we synthesised from pentachloropyridine and elemental uorine following the literature procedure. 6 The rates of uorination of nucleophiles 1a-m with electrophilic uorinating reagents 3-9 in CH 3 CN were monitored by UV-vis spectrophotometry. A representative time-arrayed multiwavelength study of the uorination of 1d by Selectuor™ 3 (Fig. 2a) shows clean, isosbestic behaviour, suggesting that no intermediate species are built up. The nucleophiles 1a-h show absorption bands at l max ¼ 340-360 nm, corresponding to their enol forms and at l max ¼ 250-270 nm, associated with a p* ) p transition of the diketone forms, as well as additional transitions due to the enol tautomer. 45,46 As each uorination reaction progresses, the absorption band at $250 nm increases in intensity, corresponding to the formation of the diketone form of the monouoro-products 2a-h, and the starting enol nucleophile signals at l $ 350 nm decrease. Plots of absorbance changes at four l values over time are shown in Fig. 2b, and tting of these data affords identical rst-order rate constants (k obs ). Similar behaviours were seen across the range of 1,3dicarbonyl derivatives and uorinating agents. By monitoring the decays in absorbance of the enol tautomer at l $ 350 nm, the kinetics of uorination reactions were conveniently monitored by UV-vis spectrophotometry. All kinetics experiments were carried out with excess electrophile in order to achieve pseudo-rst order conditions. Clean exponential decays of absorbance of the UV-active nucleophile were observed in all runs (Fig. 2c), and the rst-order rate constants k obs were obtained from the tting of plots of absorbance versus time. When k obs values were plotted against Selectuor™ concentration, a simple linear (i.e. rst order) correlation was observed (Fig. 2d). The direct dependence upon F + concentration demonstrates rate-limiting uorination and thus the slopes of these graphs give second-order rate constants k 2 [M À1 s À1 ] that report on both nucleophilic and electrophilic partners, according to the second-order rate eqn (1). The rate constants for the reactions of 1a-m with each uorinating reagent are summarized in Table 1.
Compounds 1a-g and 1i-m exist in $90% enol form whereas 1h exists as $60% enol. We conrmed that keto-enol tautomerism was rapid under our reaction conditions by using discontinuous LCMS assays on a number of systems. We found constant keto : enol ratios throughout the reaction courses (see ESI Section 7 †), where the keto and enol forms interchanged under the initially highly aqueous, acidic conditions of the LC elution gradient. Using the same LCMS approach, the uorinated products showed only small amounts of enol form. Furthermore, we monitored a reaction mixture containing 2aketo and Selectuor™ by 19 F NMR, and found that 2a-keto did not react to form the diuoro product over the course of 5 days. Hence, this suggests that diuorination does not occur in the UV-vis experiments (for further detailed discussion see ESI Section 8 †). We attempted to monitor the kinetics of uorination reactions involving reagents 6a, 6b, 7a and 7b by UV-vis spectrophotometry; however, the reactions were very slow at the low concentrations required by the UV-vis method. These studies were then conducted at higher concentrations using a discontinuous NMR reaction monitoring method, where the uorination reactions proceeded faster and at more measurable rates. Only nucleophile 1d was used in these kinetics reactions. An initial rates method by UV-vis gave a corroborating rate constant for the reaction of 7a, hence the UV-vis and NMR methods are in agreement (for all methods, spectra and rate constant graphs see ESI †).

Product analyses: reaction monitoring by NMR and LCMS
In order to corroborate and validate our ndings from UV-vis methods, NMR and LCMS experiments were employed to conrm the rates of progress of the uorination reactions and the identities of the expected mono-uorination products. NMR reactions were conducted in NMR tubes under pseudo-rst order conditions using excess nucleophile, at 25 C. A representative example is given in Fig. 3a, where compound 1b (R 1 ¼ R 2 ¼ F) was reacted with Selectuor™. Relative peak integrals from time-arrayed 1 H NMR experiments gave exponential behaviours for the uorination reactions (Fig. 3b), where each curve corresponds to a 1 H signal present in Fig. 3a. The k obs Table 1 Second-order rate constants (k 2 ) for the reactions of fluorinating reagents 3-9 with nucleophiles 1a-m in CH 3 CN, at up to four different temperatures (20 C, 25 C, 30 C and 35 C) values for each curve are in the range of 1.2-1.3 Â 10 À3 s À1 , hence they correspond to the same process. The second-order rate constant obtained was k 2 ¼ 2.2 Â 10 À2 M À1 s À1 , which is in very good agreement with that obtained from UV-vis studies (3.3 Â 10 À2 M À1 s À1 ). The multiplets at 3.7-3.8 ppm correspond to ClCH 2 -DABCO, which is the deuorinated product of Selectuor™. Given that the uorination reaction was rapid, this species was already in evidence in the rst NMR spectrum that was acquired.
LCMS experiments showed that keto and enol forms of both starting materials and products are clearly resolved, with their identities being conrmed through diode array analyses and the use of standards 1a-m and 2a-f (see ESI † for chromatogram traces). Reaction proles for uorination reactions were constructed via integration of peak areas. An example is shown in Fig. 3c, where nucleophile 1d (R 1 ¼ R 2 ¼ OMe) was reacted with 8b under bimolecular conditions (at 15 C). The increase in concentration of the uorinated product 2d is shown, and tting the data gave k 2 ¼ 3.4 Â 10 À2 M À1 s À1 , compared to k 2 ¼ 9.3 Â 10 À2 M À1 s À1 obtained from UV-vis kinetics studies (at 20 C). The two values are in good agreement considering the temperature differences.

Structure-activity correlations
The effects of the para-substituents on the rates of uorination were studied by Hammett correlation analyses of the reactions. Hammett plots were constructed for the reactions of disubstituted enols 1a-h with uorinating reagents 3, 4, 8a, 8b and 9 using s p + constants (Fig. 4a). The use of s p + values led to slightly better correlations than with s p constants in all cases (see ESI Section 5.2 † where representative Hammett plots for Selectuor™ are shown). The r + values obtained for reactions involving each uorinating reagent are between À1.4 and À2 (Fig. 4), where these negative values indicate moderate reductions in electron density on the substrates during the rate determining uorination steps. This magnitude of electron decit at the transition state is consistent with the S N 2-like mechanistic behaviors that are commonly attributed to N-F reagents.
For the mono-substituted enols 1i-m, Hammett plots were constructed using both s p and s p + values, with the latter giving better correlations (see ESI Section 5.2 †). Hammett plots constructed for reagents 3, 4 and 8b are shown in Fig. 4b. The r + values obtained were À0.83, À0.80 and À0.72 for reactions of 3, 4 and 8b, respectively. The similarity in each set of r + values suggests that the uorination mechanisms are analogous across the range of 1,3-dicarbonyl derivatives, which is a critical requirement for the construction of a predictive reactivity scale.

Reactivity scale for N-F reagents
Using the absolute rate constants obtained from kinetics studies via UV-vis reaction monitoring, relative rate constants (k rel ) were calculated, using eqn (2), with Selectuor™ as the reference electrophile (Fig. 5). Across the range of 1,3-dicarbonyl compounds 1a-m, the k rel values for each uorinating reagent are in good agreement, showing the predictive potential of the scale towards nucleophiles of differing potencies. With the k rel values in hand, we constructed a reactivity scale for uorinating abilities of the N-F reagents (Fig. 5), in CH 3 CN. The most reactive uorinating reagent on the scale is 2,3,4,5,6pentachloro-N-uoropyridinium triate 9. Selectuor™ 3, 2,6dichloro-N-uoropyridinium triate 8a and 2,6-dichloro-N-uoropyridinium tetrauoroborate 8b have very similar reactivities, with the counter-ion having little effect on the reactivity of the N-uoropyridinium salts. Synuor™ 5 is around 10 times less reactive than Selectuor™, although Synuor™ is very moisture sensitive and problems arose with competing decomposition reactions when using this reagent in our studies. Therefore, rate constants with this reagent were only obtained with the most reactive nucleophiles (R 1 ¼ R 2 ¼ OMe and R 1 ¼ OMe, R 2 ¼ H), where competitive hydrolysis processes were least signicant.
At the other extreme, NFSI 4 and N-uoropyridinium systems 6a, 6b, 7a and 7b were 4-6 orders of magnitude less reactive than Selectuor™ 3. Despite the low reactivity of NFSI 4, kinetic proles with nucleophiles 1a-e, 1h, 1j and 1k could be obtained using UV-vis monitoring within one week, owing to its high level of solubility in CH 3 CN, which allowed large concentrations of NFSI 4 to be used with consequent enhancement of observed rates. Selectuor™ 3, on the other hand, shows relatively low solubility in CH 3 CN thus, although it is more reactive, reaction rates are limited because of its poorer solubility.
Although their reactivities are similar to Selectuor™ 3, Synuor™ 5 and the 2,6-dichloro-N-uoropyridinium salts 8a and 8b are very moisture sensitive. Therefore, Selectuor™ 3 remains the most bench-stable and easy-to-handle uorinating reagent, as water can even be used as a solvent for uorination reactions involving this reagent. 48 Reagents 6a, 6b, 7a and 7b are less moisture-sensitive than the dichloro-derivatives, and our NMR studies show that they remain stable in CH 3 CN solution for several weeks. Furthermore, owing to their higher levels of solubility in CH 3 CN, appreciable rates of uorination can be achieved with these less reactive reagents through the use of higher concentrations. 2,3,4,5,6-Pentachloro-N-uoropyridinium triate 9 is highly reactive, even showing reactivity towards glass (as determined by our NMR studiestetra-uoroborate peaks are present due to uorination of borosilicate glass). We therefore suggest the use of plastic containers  Table 1, with Selectfluor™ 3 as the reference electrophile.
for transportation of this material. Compound 9 decomposes when heated in CH 3 CN, thus limiting the use of this reagent for reactions in this solvent at temperatures above $40 C.

Further insight into uorination of dicarbonyl compounds 1a-m
Activation parameters (DG ‡ , DH ‡ and DS ‡ ) were obtained from kinetic data for the reactions of Selectuor™ with 1a-e. These experiments were performed by collecting rate constants at 4 temperatures and the resulting parameters are summarized in Fig. 6a. The Eyring plots show excellent linear correlations, with R 2 > 0.99. The moderately negative values of DS ‡ support a bimolecular, S N 2-type mechanism for the uorination reactions. The free energy of activation (DG ‡ ) for the uorination reactions increases from 74.1 kJ mol À1 to 82.9 kJ mol À1 as the paryl substituent of the 1,3-dicarbonyl nucleophile changes from OMe to Cl. Enthalpy of activation (DH ‡ ) increases from 54.8 kJ mol À1 to 61.3 kJ mol À1 as the substituents become more electron-withdrawing. All three activation parameters are dependent on the electronic nature of the substituents, and the effect is most marked for the more electron-donating substituent OMe. A correlation of k 2 versus the number of para-substituents present on each 1,3-dicarbonyl was constructed using the rate constants obtained from kinetics studies with Selectuor™ and compounds 1a-e, 1g and 1i-m (Fig. 6b). As expected, nucleophiles with two electron-donating substituents (e.g. R 1 ¼ R 2 ¼ OMe) show an increase in reactivity towards uorination compared with the mono-substituted derivatives. Conversely, two electron-withdrawing groups at the para positions cause a greater decrease in nucleophilicity at C-2 than only one EWG, and hence the rate of uorination is slower with the disubstituted compounds. The para-substituents are thus working in synergy, rather than showing "push-pull" effects.
Furthermore, nucleophiles displaying substituents that have mostly inductive electron-withdrawing or electron-donating effects show a linear trend in the graphs of k 2 versus number of para-substituents. On the other hand, the OMe substituents have a non-linear correlation of rate constants versus number of substituents, and cause a strong increase in reactivity compared to 1a due to the strong electron-donating nature of each OMe group. A similar non-linear correlation was obtained with paranitro groups (Fig. 6b). The non-additive effects between monoand di-substituted substrates are consistent with the asymmetric nature of enol systems preventing identical conjugation effects by the substituents in the di-substituted systems (Fig. 6c).

Conclusion
We have provided a quantitative reactivity scale that spans eight orders of magnitude, for ten commonly-exploited uorination reagents. The reactivity of each uorinating reagent was assessed by directly monitoring the kinetics of uorination reactions with a family of 1,3-diaryl-1,3-dicarbonyl nucleophiles that mirrors the application of the reagents in C-F bond formation. The reactivities of the homologous nucleophiles span 5 orders of magnitude and allowed reactivity determinations to be performed in a genuinely comparative manner using a convenient spectrophotometric readout. Similar Hammett parameters across the range of uorination reagents revealed the mechanisms of uorination to be similar.

Methods
The ESI † contains details of kinetic experiments, product analyses and spectra of all characterized compounds.

Conflicts of interest
The authors declare no competing nancial interest.