Introduction

The presence of a fluorine atom considerably influences the physical, chemical and biological properties of organic compounds. It is often introduced to improve the bioavailability, lipophilicity or the metabolic stability of pharmaceuticals. Organofluorine compounds have therefore not only received increasing interest in pharma, but also in the agrochemical and material industry [1,2,3,4,5,6,7]. However, selective fluorination is a key challenge in organic chemistry [8].

Electrophilic fluorination is the most direct method to selectively introduce a fluorine atom into organic compounds. The most direct, atom economic, cost efficient and environmentally friendly electrophilic fluorinating agent for large scale manufacturing is by far fluorine (F2) gas [9,10,11]. Despite all these advantages, direct fluorination often suffers from poor selectivity, moderate overall yields, intolerance of many functional groups and high sensitivity to reaction conditions. In addition, due to the very high reactivity and toxicity, this reagent poses major hazards for the non-specialist and requires specific handling techniques and equipment [12]. To overcome these issues in particular on laboratory scale, a large number of stable and easy-to-handle, yet selective and efficient electrophilic fluorinating reagents of the N-F class have been developed [13, 14]. Out of this wide variety, Selectfluor [15, 16], N-fluorobenzensulfonimide (NFSI) [17, 18], and N-fluoropyridinium salts (NFPy) [19,20,21] are the most popular reagents, which is mainly due to their commercial availability, but also to their reactivity. Despite all these advantages, these reagents are comparatively expensive, and on closer inspection, most of them are synthesized from the corresponding cheap amine precursors with elemental F2 [13].

Considering large scale synthesis, it would therefore be beneficial to generate N-F reagents in-situ from the respective more economic precursors and F2 and directly use them in a telescoped downstream fluorination reaction. Handling F2 gas becomes significantly safer when continuous flow technology is involved, making such protocols particularly suitable for scale-up [12, 22,23,24]. In addition, efficiency and productivity can be improved in flow. Whereas Selectfluor and NFSI have been used in flow for performing fluorinations [25,26,27,28,29], to the best of our knowledge, the generation of N-F reagents using F2 has not been reported using a continuous flow format.

We were particularly interested in the synthesis of 2,6-dichloro-1-fluoro-pyridinium tetrafluoroborate (diClNFPy 2) starting from 2-6-dichloropyridine (DCP) for reasons of cost, atom economy and fluorination power. DiClNFPy 2 has a higher active fluorine content (3.94 mmol/g) compared to Selectfluor (2.82 mmol/g) and NFSI (3.17 mmol/g) and the fluorination power is similar to Selectfluor, which is one of the most reactive electrophilic fluorinating reagents. The reactivities of a series of common N-F fluorinating reagents (Fig. 1) have been determined by detailed kinetic fluorination studies [30, 31].

Fig. 1
figure 1

Fluorination power of N-F fluorinating reagents determined via kinetic studies

Full size image

Here, we report insights into the in-situ generation of 2,6-dichloro-1-fluoro-pyridinium tetrafluoroborate (2) from 2,6-dichloropyridine and F2 gas (10% in N2) and its telescoped downstream electrophilic fluorination reaction with enamine 3 as model compound (Scheme 1) in a continuous flow format. The only reported batch procedures for the synthesis of 2 involve treatment of DCP with BF3 (either as gas or MeCN complex), HF, and H2O in a Teflon-coated metal reactor in MeCN at − 15 to − 20 °C with 10% F2/N2 [32, 33]. DiClNFPy 2 was isolated in 76‒77% after evaporation and recrystallization.

Scheme 1
scheme 1

Reaction sequence for the in-situ generation of diClNFPy 2 and its downstream reaction with enamine 3

Full size image

Since we did not want to introduce an additional safety hazard by using HF, we decided on protocols developed by Umemoto and co-workers, which involved mixing a pyridine derivative with a Lewis acid, forming complex 1, and fluorination with 10% F2/N2 [19, 20]. After optimizations using a polymer tubing reactor set-up, the telescoped sequence was ultimately carried out in a commercially available modular lab-scale silicon carbide (SiC) flow reactor platform [34], which is not only perfectly suitable to carry out corrosive chemistry but also provides scale-up capabilities to a manufacturing scale.

Results and discussion

Continuous flow generation of diClNFPy 2

Investigations on the in-situ generation of 2 were performed closely following protocols reported by Umemoto and co-workers, although they have not described the synthesis of this specific 2,6-dichloro-substitued N-F reagent [19, 20]. First, a BF3 complex was formed in batch by mixing DCP and BF3·MeCN, then this mixture was reacted with 10% F2/N2 in continuous flow to generate diClNFPy 2. Optimizations for the fluorination step were performed with respect to temperature, residence time and equivalents of BF3 and F2 (Table 1). Initially, standard PFA coils were employed for optimization studies. This has the advantage that residence times can be easily changed while flow rates and thus the mixing regime of the liquid (DCP-BF3 complex 1) and F2 gas feed are kept the same. 10% F2/N2 was fed by a calibrated mass flow controller (MFC, Bronkhorst lowΔp) and mixed inside a SS T-mixer with DCP-BF3 complex 1, which was introduced using a syringe pump (see Figure S1 for the set-up). The original batch protocol by Umemoto uses a 1:1:3 ratio of pyridine:BF3:F2 with the fluorination step carried out at ‒40 °C for ca. 1.5 h [19, 20]. Preliminary studies for the in-situ generation of 2 in continuous flow using UV-Vis analysis suggested that an excess of BF3 is beneficial and that F2 equivalents could be reduced, as head space issues are eliminated in flow. Fluorinations are typically carried out at low concentration; therefore, we started the optimization studies using a 0.1 M solution of DCP-BF3 complex 1 using 3 equiv. of BF3 and 1 equiv. of F2 at − 10 °C and a total residence time of 10 min (entry 1, Table 1). Since a 10% F2/N2 mixture was employed, the gas flow rate was much higher than the liquid flow rate: 22.4 mLn/min vs. 1 mL/min. To reach a residence time of 10 min, a 234 mL coil reactor would have been required, which would be impractical. Therefore, excess N2 was removed by implementing a gas/liquid (g/l) separator. For our purposes, a simple gravity-based separator set-up consisting of a glass vessel and a bottle cap with three tubing connections was used (Figure S2). For safety reasons, the N2 outlet is directed to a NaOH quench solution. After a mixing unit (tres1: 1.5 s), the reaction stream entered the gas separator, and after a residence time of ca. 1 min inside the gas separator, the liquid mixture was further pumped through a second residence time coil with a flow rate of 1 mL/min (see Table 1). With this first set of conditions, 2 was generated in 46% yield, determined by 19F benchtop NMR spectroscopy with respect to the N-F signal at 31.4 ppm. Increasing the F2 equivalents to 1.5 (33.6 mLn/min) improved the yield to 69%, while a further increase to 2 equiv. had little impact (entry 2 vs. entry 3). Using 2 equiv. of BF3 (53%) or a temperature of − 20 °C (60%) did not improve the outcome either. Surprisingly, the best yield of 2 (73%) was obtained when the reaction was carried out at 20 °C (entry 6), since decomposition would actually be expected at higher temperatures. The residence time after the g/l separator could be reduced to 4.6 min (total residence time: 5.6 min), while omitting this residence time unit led to reduced yield of 2 (entries 7‒10).

Since an 19F NMR yield above 73% could not be achieved, high resolution NMR studies were conducted to gain further insight. Complex 1 containing 3 equiv. of BF3 proved to be stable for at least 1 day at room temperature: Apart from the signals at 7.85 ppm (t, 1 H) and 7.45 ppm (d, 2 H) in 1H NMR and at − 141.9 ppm in 19F NMR, no additional signals could be detected (see Figure S3). Full conversion of complex 1 was achieved according to the 1H NMR of the outlet feed solution of in-situ generated diClNFPy 2 from an experiment performed at − 10 °C (entry 8, Table 1). However, both 1H NMR and 19F NMR revealed several impurities particularly in the aromatic region (Figures S4‒S7). The signals could not be assigned unambiguously, but potentially some kind of decomposition of 2 could have occurred and/or 2 could have undergone hydrolysis, generating N-fluoro-6-chloro-2-pyridone [20]. An indication for the presence of traces of water is the signal of H2O-BF3 at 12.08 ppm in the 1H NMR. It also should be noted that a downfield shift from − 150.6 ppm to − 148.4 ppm of the BF4- signal is detected once BF3 is in excess (Figure S7).

Nevertheless, we decided to move forward to test the telescoped N-F generation/downstream fluorination involving 3 equiv. of BF3 for the formation of complex 1, 1.5 equiv. of F2, temperatures between − 10 and 20 °C and total residence times of 1‒5.6 min, as the yields of 2 from entries 7‒9 were in a similar range of 67‒72%.

Table 1 Generation of 2,6-dichloro-fluoropyridinium tetrafluoroborate (diClNFPy 2)a
Full size table

Downstream reaction with enamine 3

We selected deoxybenzoin-derived enamine 3 as model substrate for the downstream fluorination with in-situ generated diClNFPy 2. This reaction has been reported in batch by Mayr et al. within a comprehensive kinetic study of fluorinations of enamines employing various N-F reagents [31]. In their study, the fluorination of 3 was conducted at room temperature for 1 h using 1.05 equiv of 2. However, under those conditions, full conversion of the enamine was not achieved, and after hydrolysis, a 91:9 mixture of the mono- and difluorinated ketones 4 and 5 was obtained. In addition, other unidentified side products were reported.

Before exploring the telescoped N-F generation/enamine process, we tested the fluorination of 3 with commercial 2 in flow. Therefore, enamine 3 was reacted with diClNFPy 2 in a PFA coil reactor at room temperature (Scheme 2 and Table S2). As described in batch [31], 2 was fed at slight excess of 1.05 equiv. and concentrations similar to those that would be obtained from the in-situ generated diClNFPy 2 were chosen (Scheme 2). By translating those conditions to a flow protocol, the reaction time could be decreased from 1 h to only 2 min while obtaining similar results as for the batch process: 93% conversion of 3 was achieved with 97% selectivity for the monofluorinated product 4. The only detected side product was the difluorinated ketone 5 in 3%.

Scheme 2
scheme 2

Fluorination of enamine 3 in continuous flow with commercial 2

Full size image

With this information in hand, we integrated the fluorination of 3 to the N-F generation in a telescoped fashion by simply connecting a PFA coil as residence time unit to the outlet tubing of in-situ generated diClNFPy 2 (see Scheme S1 for the set-up). The downstream fluorination was performed at the same temperature as the formation of 2. Based on entry 7 in Table 1, 2 is generated in a concentration of 0.072 M at 20 °C and 4.6 min residence time after the g/l separator. Similar to batch, 2 was employed in slight excess (1.06 equiv) to enamine 3. Preliminary experiments revealed that although optimum results regarding formation of 2 were obtained at 20 °C and 5.6 min total residence time, the fluorination of enamine 3 furnished 20% higher conversion to the monofluorinated product 4 at − 10 °C. This can be most likely be ascribed to decomposition of 2 during the whole sequence at higher temperatures. It also turned out that the residence time unit after the g/l separator could be omitted, reducing the residence time for the generation of diClNFPy 2 to only ca. 1 min, which corresponds to the time inside the g/l separator. In agreement with the flow procedure using commercial 2, the fluorination of 3 proved to be very fast (0.5‒1 min), leading to a total reaction time of only 1.5‒2 min for the telescoped N-F generation/enamine fluorination. No difluorinated product 5 was obtained, but ca. 20% of another side product was detected by HPLC analysis.

This side product, which notably was not observed when using commercial diClNFPy 2, was identified as chlorinated ketone 6 (see SI and Figures S10‒S13). Such chlorinated side products have been reported before when performing fluorinations with Selectfluor or pentachloro-NFPy triflate [35, 36]. Whereas chloride oxidation leading to an electrophilic chloronium species was claimed in case of Selectfluor [35], chlorine derived from the hydrolysis of pentachloro-NFPy triflate was described responsible for the chlorination reaction [36]. In our case, chloride could be released either by hydrolysis generating species such as N-fluoro-6-chloro-2-pyridone [20], or by a fluoride/chloride replacement via a single electron transfer (SET) sequence [37] (Scheme 3). Another possible fluoride (F−) source that could undergo a similar fluoride/chloride replacement on 2 would be HF, which could be generated from the decomposition of F2 by water. Since N-F reagents are known to be strong oxidizing agents [14, 38], the released chloride could be oxidized by 2 either via a two electron transfer (TET) to generate Cl+ or a SET to generate Cl●. Both pathways have been suggested with Selectfluor as N-F reagent [39, 40]. Considering the 70% yield of diClNFPy 2 after the first step, hydrolysis followed by either SET or TET would be plausible, although extra dry MeCN was used as solvent and the entire set-up was purged with extra dry N2. Nevertheless, since the obtained side products in the N-F generation/enamine fluorination sequence could not be unambiguously assigned to any of the intermediates described in Scheme 3 and no further in-depth investigations on the mechanism of chlorination of enamine 3 were performed, we cannot exclude one of the proposed pathways.

Scheme 3
scheme 3

Possible reaction pathways toward the formation of chlorinated side product 6

Full size image

Based on the results obtained with the coil reactor set-up, we directly transferred those conditions to a modular plate-based SiC reactor (Protrix, Chemtrix) [34]. Since this reactor platform provides excellent reaction control, better cooling should be achieved compared to the coil reactor, which could be advantageous for preventing side product formation. In the Protrix we took advantage of the two independent reactor volumes/channels per module plate (see Figure S8). The N-F reagent was pre-formed on one side, the outlet stream was then directed into the g/l separator, and the liquid N-F stream entered the second channel within the same plate for the downstream fluorination reaction. Since diClNFPy 2 was generated in an average yield of 69% at − 10 °C when using DCP-BF3 complex 1 (3 equiv. of BF3) and 1.5 equiv. of F2, enamine 3 was pumped at a concentration of 0.068 M so that 2 is in slight excess (1.01 equiv.). We started optimizations with one reactor module with 0.96 mL reactor volume for the generation of 2 and 2.76 mL for the enamine fluorination (Table 2). By implementing the same flow rates as in the coil reactor set-up – 1 mL/min for complex 1 and 33.6 mLn/min for 10%F2/N2 – N-F reagent 2 was generated within ca. 1 min (tres inside the reactor channel: 1.7 s) and fluorination of 3 took place within 1.4 min (entry 1, Table 2). A similar outcome as in the coil reactor could be obtained: 86% conversion of 3, and product 4 was obtained in 58% yield and 72% selectivity (HPLC at 254 nm). Unfortunately, the chlorinated side product 6 was also detected in similar amounts. Increasing the equivalents of the N-F reagent by pumping enamine 3 at lower concentration did not improve the performance (entry 2). Since temperature control should be more reliable within this reactor system, the telescoped reaction was also carried out at 0 °C. Similar to experiments in the coil reactor, the yield of 4 dropped to 43% (entry 3). At this temperature, an increase in N-F equivalents led to only 23% of 4 (entry 4). Interestingly, when the g/l separator was removed and the outlet tubing of generated diClNFPy 2 directly attached as inlet to the second channel for the enamine fluorination (see Scheme S2), 4 was obtained in similar yield (51% vs. 58%) under otherwise identical conditions (entry 5 vs. 1). The reaction time for the fluorination step could thus be decreased to only 4.7 s, resulting in a total residence time for the entire sequence of 6.5 s compared to 2.5 min with the g/l separator. The best yield of 64% (entry 7) was achieved at a total residence time of 14.5 s, for which an additional residence time module (see Figure S8) was added. A 45 min scale-out experiment under these conditions furnished 75% (352 mg) of isolated product after extraction, containing 59% of fluorinated ketone 4, 25% of chlorinated ketone 6 and 16% of hydrolyzed enamine 3.

We also tested the reaction sequence with reduced BF3 equivalents in complex 1: 1‒2 equivalents led to 12‒39% lower yields, which is in agreement with our initial findings and proved the beneficial/activating effect of BF3 for the generation of diClNFPy 2. It should be noted that the selectivity for 4 did not change significantly, as chlorinated side product 6 was generated in essentially the same amount regardless of the reaction conditions. For comparison, we also performed the fluorination of 3 using F2 gas directly applying similar conditions (1.05 equiv F2, 7.6 s residence time). The reaction proceeded surprisingly clean, but converted enamine 3 in only 9% to the monofluorinated ketone 4 along with ca. 8% of difluorinated ketone 5. By employing 2 equiv. of F2, substantial amounts of difluorinated ketone were generated. This confirms the usefulness of highly selective N-F reagents in fluorination chemistry.

Table 2 Telescoped N-F generation/enamine fluorination using a SiC plate reactor (Protrix)a
Full size table

Conclusion

An explorative study on the continuous flow generation of the N-F reagent diClNFPy 2 from DCP and 10% F2/N2 and its telescoped downstream electrophilic fluorination reaction with enamine 3 revealed that this type of chemistry is not a trivial affair. Possibly because of the difficulties faced, the generation of N-F reagents in a continuous flow format has not been reported before. Compared to the reaction with commercial diClNFPy 2, difluorinated product 5 was not observed, but another side product was encountered, which was not obtained with commercial 2: the monochlorinated ketone 6. In general, diClNFPy 2 could be generated in up to 70% yield and monofluorinated ketone 4 in up to 64%, along with ca. 20‒30% of 6. The temperature for the N-F generation step could be increased to − 10 °C compared to − 40 °C in batch [20]. Both reaction sequences proved to be very fast when carried out in flow, the N-F generation step could be done within 7.9 s and only 6.6 s were necessary for the fluorination of 3.

Excess BF3, on the one hand, seems to be beneficial for the generation of diClNFPy 2, but, on the other hand, may hamper the downstream fluorination and/or promotes pathways leading to the chlorinated side product (see Scheme 3).

Further developments toward an improved process for the N-F generation and downstream reaction would need to involve in-depth investigations on the mechanism of chlorine side product formation and on potential other decomposition processes during the synthesis of N-F. If BF3 would have a negative impact on the whole sequence, excess could be removed via g/l separation. It is also known that fluorination reactions are substrate sensitive, if the pathway SET/F-Cl exchange would prevail (Scheme 3), potentially a change in the nucleophile would be beneficial. Another strategy to circumvent the formation of chlorinated side products could be the generation of non-chlorinated N-F reagents, such as trimethyl-NFPy. Although known to have a lower fluorination power [38], the performance potentially could be enhanced under flow conditions. In addition, to make the process more sustainable, the regeneration of the consumed N-F reagent would be the ultimate goal. Investigations on those topics are in progress.