German Edition: DOI: 10.1002/ange.201712021FluorineInternational Edition: DOI: 10.1002/anie.201712021

Spectroscopic Characterization of a [C@F@C]+ Fluoronium Ion inSolutionCody Ross Pitts, Maxwell Gargiulo Holl, and Thomas Lectka*

Abstract: We report the first spectroscopic evidence for a [C@F@C]+ fluoronium ion in solution. Extensive NMR studies (19F,1H, 13C) characterize a symmetric cage-like species in whichfluorine exhibits substantial covalent bonding to each of thetwo carbon atoms involved in the three-center interaction.Experimental NMR data comport well with calculated valuesto lend credence to the structural assignment. As the culminat-ing experiment, a Saunders isotopic perturbation test con-firmed the symmetric structure. Congruent with the trend inother types of onium ions, the calculated charge at fluorinemoves in a more positive (less negative) direction from theneutral. It is this important trend that explains in part theextraordinary historical difficulty in making theoretical pre-dictions of fluoronium ions come true in solution, and why ittakes fluorine captured in a cage to produce, finally, a stableion and complete the historical arc of the organic halonium ionstory.

The chemical history of reactive intermediates generallyfollows a well-trodden path. Indirect evidence allows a postu-late; alternatively, a theoretical prediction is followed byindirect evidence. Direct evidence often appears some timelater in the form of a fast spectroscopic technique; finally,direct evidence in the form of a stable species completes thestoryQs arc and convinces most skeptics of the intermediateQsexistence. In the case of C@F@C fluoronium ions,[1] an initialpostulate by Olah[2] was followed by pioneering indirect gas-phase-based evidence reported by Morton and co-workers.[3]

Gabbai and co-workers more recently provided evidence fora novel cation that oscillates between two classical structureswherein a fluorine atom “jumps” between two carboncenters.[4] We contributed some time later with indirectevidence, heavily reliant on isotopic labeling studies, fora transiently formed symmetric [C@F@C] fluoronium ion (1)in solution derived from the hydrolysis of triflate 2(Scheme 1).[5] In the tradition of classical studies of reactiveintermediates, we now offer direct spectroscopic evidence forthe formation of a symmetric fluoronium ion in solution as

a surprisingly long-lived species generated at low temperaturein a non-nucleophilic medium.

Ionization of both triflate 2 and alcohol 3 was firstattempted in magic acid medium (SO2ClF solvent,@120 88C).[6]

In all cases, we observed decomposition, likely resulting fromthe protonation and putative dissociation of the fairly basicanhydride group. This result led us to conclude, erroneously,that the basic system represented by 2 and 3 was notappropriate for Lewis acid studies, though it had proved tobe extremely versatile for investigating other phenomena.[7,8]

Some time later, as a final approach, we attempted to formthe desired cation in solution through ionization of difluoride4,[9] which was synthesized from alcohol 3 by treatment withXtalfluor-E and DBU.[10] It is well known that SbF5, a powerfulLewis acid, is very fluorophilic.[11]

We reckoned that there was a reasonable chance that thedesired “outward”-pointing fluoride would be abstractedpreferentially as the resulting fluoronium would be substan-tially more stable (as predicted by various DFT methods)than other potential isomers (Scheme 2). On the other hand,we once again feared that coordination to the anhydride (thistime by SbF5) would lead to decomposition.

With difluoride 4 in hand, we attempted an ionization(SO2ClF, excess SbF5, @120 88C) and monitored it by 19F NMRspectroscopy. Starting difluoride peaks appear at d =

@179.6 ppm (in) and d =@198.6 ppm (out) in either SO2ClFor SO2 ; upon addition of SbF5, a new resonance appearscleanly in between the two substrate peaks (d =

Scheme 1. Formation of a fluoronium ion as a transient intermediate.Its symmetric nature was determined by isotopic labeling experiments.

Scheme 2. Possible ionization pathways of 4. Ion 1 is predicted to bemore stable than its counterparts 5 and 6.[5a]

[*] Dr. C. R. Pitts, M. G. Holl, Prof. T. LectkaDepartment of ChemistryJohns Hopkins University3400 North Charles St., Baltimore, MD 21218 (USA)E-mail: lectka@jhu.edu

Dr. C. R. PittsDepartment of Chemistry and Applied BiosciencesETH ZfrichVladimir-Prelog-Weg 2, 8093 Zfrich (Switzerland)

Supporting information and the ORCID identification number(s) forthe author(s) of this article can be found under:https://doi.org/10.1002/anie.201712021.

AngewandteChemieCommunications

1924 T 2018 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2018, 57, 1924 –1927

@188.2 ppm).[12] The new resonance is an apparent triplet ofpentets, explicable by strong coupling of the bridging fluorineto the two geminal protons (Scheme 3). The pentet feature is

a result of long-range W coupling[13] to the highlighted protonin green, of which there are four in total.

The 1H NMR spectrum shows a highly diagnostic down-field-shifted doublet (d = 8.11 ppm, 2 H) representing theexpected reciprocal geminal coupling to 19F. In contrast, the13C NMR spectrum at first glance appeared to be problematic.Instead of the expected C2v symmetry of the desired ion, theobserved species revealed extra peaks consistent with overallCs symmetry. A strong clue was provided by the chemicalshifts of the carbonyl groups, one of which rests considerablydownfield from the other (d = 159.5 vs. 186.7 ppm). Thisphenomenon can be interpreted to be a result of SbF5

coordination to one of the anhydride carbonyl groups in thecationic structure (Figure 1). Complexes of SbF5 and carbonyl

groups are well-known in the literature[14] and can becharacterized as metastable species. Still, we found it remark-able that coordination does not result in decomposition at lowtemperature. Calculation of the 13C spectrum of complex1·SbF5 (B3LYP/DGDZVP, Cs symmetry)[15] predicted chem-ical shifts consistent with experiment. Fortunately, coordina-tion still permits a fully symmetric C@F@C interaction;although the C2v symmetry is broken by SbF5, the remainingmirror plane bisects the fluorine atom (Cs symmetry) andleaves the bound carbon atoms as symmetry partners.

It is difficult to conclude anything about the precisecharge at fluorine from the 19F NMR spectrum as thecorrelation therebetween for this particular nucleus is some-what complex.[16] On the other hand, the various positions ofthe 13C and 1H chemical shifts in the cation suggest that most

of the positive charge thereupon is unsurprisingly shiftedaway from the divalent fluorine. Nevertheless, as is the trendwith other onium ions,[17] the DFT calculated partial charge atfluorine moves in a more positive (less negative) directionfrom neutral 4. It is this trend that represents the importantfactor, and explains the extraordinary historical difficulty inmaking theoretical predictions of C@F@C fluoronium ionscome true in solution.

The most important question to answer is whether thestructure is truly symmetric, with fluorine bound equivalentlyto both carbon atoms. The preliminary NMR data are stronglyindicative, but not quite definitive (all DFT levels indicatea symmetric structure as well). One notable approach to theclassical versus nonclassical question lies in the Schleyer–Lenoir–Prakash–Olah chemical shift additivity test.[18] Theauthors stated that “the total 13C chemical shift differencebetween a carbocation and the corresponding neutral hydro-carbon also provides a rough, but useful, structural index.Classical carbocations show large chemical shift differences,typically 350 ppm or more, whereas related nonclassicalcations display differences often hundreds of parts per millionless.” We measured the chemical shift additivity of system1·SbF5 versus 4 as 111.8 ppm, which is well within thenonclassical realm (Scheme 4). Factoring in the effect ofSbF5 coordination, this value presents an upper limit on thedifference that would be expected in the uncoordinatedcation.

To provide more convincing evidence of a symmetricstructure, we turned to the venerable Saunders isotopicperturbation test.[19] For an isotopically labeled symmetricstructure, the difference in the chemical shifts of the bridgingcarbon atoms should be on the order of 1 ppm or less. On theother hand, a labeled equilibrating structure (7·SbF5,Figure 2) should display a much larger chemical shift differ-ence (> 10 ppm).[20]

Synthesis of the labeled precursors was achieved bytreatment of alcohol 8[21] with Xtalfluor-E and DBU, as in 3(Scheme 5). The fact that the deuterium is scrambled isindicative of transient fluoronium ion formation in thefluorination reaction. In addition, while the fluorination of 3

Scheme 3. Formation of 1. The experimental 19F NMR signal of puta-tive 1 in SO2 at @50 88C; note that the triplet of pentets is well-resolved.

Figure 1. Formation of 1·SbF5 in SO2ClF or SO2.

Scheme 4. Chemical shift additivity of 1·SbF5 and its parent hydro-carbon.

Figure 2. Structure of 1 compared to hypothetical equilibrating classi-cal cations.

AngewandteChemieCommunications

1925Angew. Chem. Int. Ed. 2018, 57, 1924 –1927 T 2018 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org

to give 4 requires elevated temperature, labeled alcohol 8reacts at room temperature, as there is more strain to berelieved upon C@O bond cleavage. Upon ionization, bothlabeled precursors 9 and 10 lead to the identical cation 11(Scheme 5). The difference in chemical shifts between the twocarbon atoms is only 0.08 ppm, which supports the hypothesisof a nonclassical structure.

The integrity of cation 1 is maintained for hours even attemperatures as high as @40 88C. Quenching at this temper-ature with slightly wet trifluoroethanol (TFE) produces threemain products, namely TFE ether 12, alcohol 3, and startingfluoride 4 (Scheme 6). The reformation of the fluoride isunusual and implies either a very fast hydrolysis of thefluoroantimonic ions in solution or, also likely, abstraction offluoride from said ions by 1.

As we have generated an unusual type of chemicalenvironment for fluorine, many questions about its exactnature arise. For one, how does the fluoronium ion compareto onium ions of other second-row elements (O and N)? Aspreviously discussed, the inward fluorine atom in the startingmaterial becomes more shielded in the fluoronium product.There has been some success with the use of 17O NMRspectroscopy to characterize oxonium ions although theseexperiments are quite difficult to interpret owing to lowabundance and high signal broadening. Nonetheless, Olahand co-workers found that the 17O nuclei in oxonium ions aregenerally deshielded compared to those in their parentalcohols and ethers.[22] For nitrogen (observed by 15N NMRanalysis), the result seems to depend on the situation; thenitrogen atom in protonated trimethylamine is deshieldedrelative to the free base,[23] while the nitrogen atom intetramethylammonium bromide is shielded relative to trime-thylamine.[24]

In the case of fluoronium ions, there is another compli-cation. The fluorine atom is part of a molecule that is more

complex than the ammonium and oxonium examples, and itschemical shifts in both the precursor and cation arise froma multitude of different factors. The inward-pointing fluorineof the starting difluoride experiences “jousting” interactions[8]

with the in-hydrogen atom on the opposing bridge, leading toorbital compression[25] and strong nuclear deshielding. Thusthe fact that the 19F nucleus becomes more shielded as it istransformed from difluoride 4 into fluoronium 1 is notdominated by factors that would be present for a generalizedalkyl fluoride.

We turned to DFT calculations to shed more light on thisissue. Gauge-independent atomic orbital (GIAO) NMRcalculations (wB97XD/6-311 + G**)[26, 27] of 1 and 4 showthe fluoronium to be shielded relative to the difluoride by10 ppm (the experimental value is 7.5 ppm). For comparison,we calculated GIAO values for isopropyl fluoride and theunknown diisopropyl fluoronium as a much less strainedmodel system. In this case, the fluoronium is deshielded by68 ppm relative to isopropyl fluoride. Thus it appears that, aswith nitrogen, whether the fluorine nucleus is shielded ordeshielded and to what extent is highly dependent on thespecific species.

Another question that arises concerns the degree ofcovalency of the bonds between fluorine and the carbonatoms. An atoms in molecules (AIM) analysis[28] shows a bondcritical point (BCP) with a large electron density valuebetween the carbon atoms and fluorine, characteristic ofa strong bonding interaction. Furthermore, C@F couplingconstants also provide a rough guide to bond strength. Indifluoride 4, the one-bond C@F coupling (for the in-F) is213.7 Hz. In the fluoronium ion 1·SbF5, this value decreases to135.7 Hz, which is still quite large, indicating that significantspin transfer occurs between the fluorine and carbon nuclei;this is consistent with two fairly strong bonding interactions.While the difference between the coupling constants in 4 and1·SbF5 is large, the value for the latter is not so far outside therange observed for other covalent C@F bonds; for example,1JCF = 160 Hz in 1-fluoropentane and 1JCF = 162 Hz in tert-butyl fluoride, both of which are closer to the 1JCF value of1·SbF5 than to that of 4. In addition, the coupling constant of 4is more in line with that of strained alkyl fluorides (215 Hz forfluorocyclobutane and 221 Hz for fluorocyclopropane, com-pared to 174 Hz and 170 Hz for fluorocyclopentane andfluorocyclohexane, respectively).[29] The 1JCF coupling of1·SbF5 is expected to be dependent on the C-F-C angle aswell, with larger bond angles leading to larger couplingconstants. Notably, the experimental value of 137.5 Hz issome 30 Hz stronger than what is predicted for the 1JCF of theprototype model, [Me@F@Me]+, with the same bond angle atthe wB97XD/6-311 ++ G** level of theory. Significant cova-lent character is also corroborated by NBO calculations of themodel [Me@F@Me]+[5b] and of 1·SbF5, which provides a calcu-lated Wiberg bond order of 0.53 for each of the C@F bonds.

In conclusion, the spectroscopic evidence reported hereinis consistent with a [C@F@C]+ fluoronium ion in solution.NMR studies (19F, 1H, 13C) and an associated Saunders testreveal a symmetric cage-like species in which fluorine exhibitssubstantial and identical covalent bonding to each of the twocarbon atoms involved in the three-center bond, as well as

Scheme 5. Synthesis of a deuterium-labeled fluoronium ion.

Scheme 6. Products resulting from the quenching of the fluoroniumion.

AngewandteChemieCommunications

1926 www.angewandte.org T 2018 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2018, 57, 1924 –1927

ancillary coordination of the species to SbF5. The experimen-tal NMR data are consistent with calculated values and lendcredence to the structural assignment.

Acknowledgements

T.L. thanks the NSF (CHE 1465131) for support. T.L. andC.R.P. thank Professor Antonio Togni (ETH Zgrich) forgenerous use of ETH facilities, as well as Amanda Baxter(University of Southern California), Dr. Michael T. Scerba(NIH Baltimore), and Dr. Ren8 Verel (ETH Zgrich) for theirhelp.

Conflict of interest

The authors declare no conflict of interest.

Keywords: cage molecules · carbocations · fluorine · onium ions

How to cite: Angew. Chem. Int. Ed. 2018, 57, 1924–1927Angew. Chem. 2018, 130, 1942–1945

[1] A recent article (K. O. Christe, R. Haiges, M. Rahm, D. A.Dixon, M. Vasiliu, J. Fluorine Chem. 2017, 204, 6) takesexception to the use of the name “fluoronium” based oncalculated partial negative charges at fluorine in such species,which in fact we first published (see Ref. [5a,b]) and Christe andco-workers have reconfirmed. Instead, these authors advocatethe lengthy term “fluorine atoms with decreased negative chargecompared to the free F@ anion.” In contrast, we feel bound tofollow standard, long-standing IUPAC (Gold Book) nomencla-ture on halonium and other onium ions, in which calculatedcharges of the central atoms are irrelevant.

[2] G. A. Olah, Halonium Ions, Wiley, Chichester, 1975, p. 130.[3] N. Viet, X. Cheng, T. H. Morton, J. Am. Chem. Soc. 1992, 114,

7127.[4] H. Wang, C. E. Webster, L. M. Perez, M. B. Hall, F. P. Gabbai, J.

Am. Chem. Soc. 2004, 126, 8189.[5] a) M. D. Struble, M. T. Scerba, M. Siegler, T. Lectka, Science

2013, 340, 57; b) M. D. Struble, M. G. Holl, M. T. Scerba, M. A.Siegler, T. Lectka, J. Am. Chem. Soc. 2015, 137, 11476.

[6] a) G. A. Olah, Science 1970, 168, 1298; b) C. U. Pittman, Jr.,G. A. Olah, J. Am. Chem. Soc. 1965, 87, 5123.

[7] For a minireview, see: M. G. Holl, C. R. Pitts, T. Lectka, Angew.Chem. Int. Ed. 2017, DOI: https://doi.org/10.1002/anie.201710423; Angew. Chem. 2017, DOI: https://doi.org/10.1002/ange.201710423.

[8] L. Guan, M. G. Holl, C. R. Pitts, M. D. Struble, M. A. Siegler, T.Lectka, J. Am. Chem. Soc. 2017, 139, 14913.

[9] M. D. Struble, J. Strull, K. Patel, M. A. Siegler, T. Lectka, J. Org.Chem. 2014, 79, 1.

[10] A. LQHeureux, F. Beulieu, C. Bennett, D. R. Bill, S. Clayton, F.LaFlamme, M. Mirmehrabi, S. Tayadon, D. Tovell, M. Couturier,J. Org. Chem. 2010, 75, 3401.

[11] a) G. A. Olah, J. R. DeMember, R. H. Schlosberg, J. Am. Chem.Soc. 1969, 91, 2112; b) G. A. Olah, J. R. DeMember, R. H.Schlosberg, Y. Halpern, J. Am. Chem. Soc. 1972, 94, 156.

[12] NMR spectra for characterization of both the starting materialand the ion were collected at @50 88C out of convenience. This isthe lowest temperature that we could achieve using a compressor

instead of a liquid N2 vessel, which allowed for longerexperimental times for each sample. However, preliminarydata using liquid N2 at lower temperatures showed cleanconversion of the starting material to the ion between @120 88Cand @95 88C by 19F NMR analysis. The same signals wereobserved by 19F and 1H NMR spectroscopy upon raising thetemperature incrementally to @30 88C, and decomposition of theion was observed at temperatures above @30 88C by 1H NMRanalysis.

[13] T. Parella, F. Sanchez-Ferrando, A. Virgili, Magn. Reson. Chem.1995, 33, 196.

[14] J. Vancik, V. Gabelica, Z. Mihalic, S. Watanabe, D. E. Sunko, J.Chem. Soc. Perkin Trans. 2 1994, 1611.

[15] This basis set includes the Sb atom; see: a) N. Godbout, D. R.Salahub, J. Andzelm, E. Wimmer, Can. J. Chem. 1992, 70, 560;b) C. Sosa, J. Andzelm, B. C. Elkin, E. Wimmer, K. D. Dobbs,D. A. Dixon, J. Phys. Chem. 1992, 96, 6630.

[16] W. Adcock, A. B. Abeywickrema, J. Org. Chem. 1982, 47, 2945.[17] A. C. Hopkinson, I. G. Csizmadia, Theor. Chim. Acta. 1974, 34,

93.[18] P. v. R. Schleyer, D. Lenoir, P. Mison, G. Liang, G. K. S. Prakash,

G. A. Olah, J. Am. Chem. Soc. 1980, 102, 683.[19] a) M. Saunders, L. Telkowski, M. R. Kates, J. Am. Chem. Soc.

1977, 99, 8070; b) M. Saunders, M. R. Kates, J. Am. Chem. Soc.1977, 99, 8071; c) M. Saunders, M. R. Kates, K. B. Wiberg, W.Pratt, J. Am. Chem. Soc. 1977, 99, 8072.

[20] Recent work by Bogle and Singleton also highlights that “greatcare must be taken” when concluding asymmetry from anisotopic perturbation test as there are instances where isotopicsubstitution can have a substantial desymmetrizing effect onotherwise symmetric molecules. In short, a larger chemical shiftdifference observed for an isotopically labeled structure does notnecessarily indicate that the non-substituted structure is unsym-metric (i.e., rapidly equilibrating); on the other hand, a smallchemical shift difference is historically indicative of a symmetricstructure. See: X. S. Bogle, D. A. Singleton, J. Am. Chem. Soc.2011, 133, 17172.

[21] M. D. Struble, C. Kelly, M. A. Siegler, T. Lectka, Angew. Chem.Int. Ed. 2014, 53, 1521; Angew. Chem. 2014, 126, 1547.

[22] G. A. Olah, A. Burrichter, G. Rasul, R. Gnann, K. O. Christe,G. K. S. Prakash, J. Am. Chem. Soc. 1997, 119, 8035.

[23] A. Bagno, C. Comuzzi, G. Scorrano, J. Chem. Soc. Perkin Trans.2 1993, 283.

[24] O. Vogl, A. Rehman, P. Zarras, Monatsh. Chem. 2000, 131, 437.[25] a) M. G. Holl, M. D. Struble, P. Singal, M. A. Siegler, T. Lectka,

Angew. Chem. Int. Ed. 2016, 55, 8266; Angew. Chem. 2016, 128,8406; b) I. Nowak, J. Fluorine Chem. 1999, 99, 59; c) Y. Chang, L.Ho, T. Ho, W. Chung, J. Org. Chem. 2013, 78, 12790.

[26] J.-D. Chai, M. Head-Gordon, Phys. Chem. Chem. Phys. 2008, 10,6615.

[27] The wB97XD functional was employed previously to study thistype of fluoronium ion (see Ref. [5b]) as a reasonably cost-effective way to include long-range dispersion corrections.

[28] a) R. Bader, Atoms in Molecules: A Quantum Theory, OxfordUniversity Press, Oxford, 1994 ; b) R. Bader, Chem. Rev. 1991,91, 893 – 928.

[29] W. R. Dolbier, Jr., Guide to Fluorine NMR for Organic Chem-ists, Wiley, Hoboken, 2009, pp. 40 – 47.

Manuscript received: November 22, 2017Revised manuscript received: January 5, 2018Accepted manuscript online: January 9, 2018Version of record online: January 24, 2018

AngewandteChemieCommunications

1927Angew. Chem. Int. Ed. 2018, 57, 1924 –1927 T 2018 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org