From 2-to 3-Substituted Ferrocene Carboxamides or How to Apply Halogen ”Dance” to the Ferrocene Series

: Two methods were compared to convert ferrocene into N , N -diisopropylferrocenecarboxamide, N , N -diethylferrocene- carboxamide, N , N -dimethylferrocenecarboxamide and (4-morpholinocarbonyl)ferrocene, namely deprotometalation followed by trapping using dialkylcarbamoyl chlorides and amide formation from the intermediate carboxylic acid. The four ferrocenecarboxamides were functionalized at C 2 ; in the case of the less hindered and more sensitive amides, recourse to a mixed lithium-zinc 2,2,6,6-tetramethylpiperidino-based base allowed to achieve the reactions. Halogen migration using lithium amides was next optimized. Whereas it appeared impossible to isolate the less hindered 3-iodoferrocenecarboxamides, 3-iodo-N , N -diisopropylferrocenecarboxamide proved stable and was converted to new 1,3-disubstituted ferrocenes by Suzuki coupling or amide reduction. DFT calculations were used to rationalize the results obtained.


INTRODUCTION
Polysubstituted ferrocenes are much appreciated scaffolds for various applications including catalysis, fuel additives, material sciences and medicinal chemistry. 1ethods to access 1,2-disubstituted ferrocenes have been largely developed from monosubstituted ferrocenes. 2 Among the methods used, deprotonative lithiations directed at neighboring sites by coordinating or/and acidifying groups are of crucial importance.Ferrocenecarboxylic acid derivatives such as hindered N,N-dialkylcarboxamides 3 and oxazolines 2b,4 have been often employed to this purpose, with stereoselective reactions being possible by using chiral ligands or chiral directing groups, respectively.
In sharp contrast, if 1,3-disubstituted ferrocenes appear as promising for different applications, 5 their synthesis is far less developed.Obtaining such structures by building the ferrocene core 6 is possible but subjected to tedious preparation of the required substrates.Thus, access by functionalization of ferrocene is an attractive approach.From monosubstituted ferrocenes, direct electrophilic substitutions are hardly regioselective. 7Stoichiometric and catalytic CH-functionalizations can take place at positions remote from substituents, but these reactions are limited to specific groups 8 or bases, 9 and can hardly be made stereoselective.Alternative ways to access 1,3-disubstituted ferrocenes consist in using retractable directing groups.11e Base-catalyzed aromatic halogen 'dance' is an elegant way to convert 2-halogeno substituted benzenes (I>Br) into the corresponding 3-substituted derivatives. 13Relatively well-developed in the benzene series, the reaction has been subjected to very few studies from ferrocenes.In 2010, we reported the competitive formation of 1-bromo-3-iodoferrocene in the course of the deprotometalation-iodolysis of bromoferrocene using the base in situ prepared from ZnCl2•TMEDA (TMEDA = N,N,N',N'tetramethylethylenediamine) and LiTMP (TMP = 2,2,6,6tetramethylpiperidino) in a 1:3 ratio, 14 and supposed to be a 1:1 mixture of LiTMP and Zn(TMP)2. 15More recently, Wang, Weissensteiner and co-workers showed that the reactions performed on ferrocenyl 1,2-dihalides (1-chloro-2-iodo-, 1-bromo-2-iodo-, 1,2dibromo-and 1,2-diiodoferrocene) using LiTMP more look like a 'scrambling' than a 'dance', with complex mixtures obtained. 16e thought a way to reduce this complexity was to involve in such reactions 1,2-disubstituted ferrocenes bearing only one halogen and a fixed N,N-dialkylcarboxamide directing group.We thus chose 2-iodoferrocenecarboxamides to attempt base-catalyzed halogen migration.Herein, we describe our efforts to efficiently synthesize the required substrates and to convert them into the corresponding 3-iodoferrocenecarboxamides.The molecular structure of the above mentioned ferrocenes was studied by 1 H and 13 C NMR spectroscopy, and DFT calculations were used to rationalize the results.

RESULTS AND DISCUSSION
Compared methods to access N,N-dialkyl ferrocenecarboxamides.N,N-dialkyl ferrocenecarboxamides are important substrates for subsequent elaboration.Therefore, we planned the synthesis of four ferrocenecarboxamides, namely the N,Ndiisopropyl, N,N-diethyl, N,N-dimethyl and morpholino.Direct reaction of ferrocene with dialkylcarbamoyl chlorides was reported to give the corresponding carboxamides in moderate yields. 17In our hands, reacting ferrocene with diethylcarbamoyl chloride in the presence of AlCl3 (1.1 equiv) at dichloromethane reflux provided N,N-diethylferrocenecarboxamide in 33% yield (5% yield by using 2 equiv of AlCl3; no product and 66% recovered ferrocene by employing 1.1 equiv of SnCl4).This disappointing yield led us to consider alternative syntheses.
We considered and evaluated two routes toward these substrates: ferrocene deprotometalation followed by trapping using dialkylcarbamoyl chlorides (Route A) 18 and amide formation from ferrocenecarboxylic acid 19 (Route B). 20 As shown in Table 1, the two routes work in similar yields although Route A provides the ferrocenecarboxamides 1-4 in only one step from cheap ferrocene.All the products obtained were fully characterized, and their main spectroscopic and X-ray diffraction data are furnished in Supporting Information.
When compared to 1 and 2, N,N-dimethylferrocenecarboxamide (3) and morpholinoferrocenecarboxamide (4) are more sensitive to nucleophilic attacks.Consequently, ketones are concomitantly formed upon their treatment by organolithiums, and lower yields of 2-substituted derivatives are noticed after subsequent quenching.Recourse to LiTMP with in situ trapping (e.g.ClSiMe3) makes functionalization of such substrates possible. 22Although iodine cannot be used as in situ trap, the zinc species formed next to the directing group by deprotolithiation-'trans-metal trapping' 23 using the combination of LiTMP (0.5 equiv) and Zn(TMP)2 (0.5 equiv) can be converted to the corresponding iodides 7 and 8 which were isolated in 78 and 84% yield, respectively.The main spectroscopic and X-ray diffraction data are given in Supporting Information and/or Figure 1.
Halogen 'dance' to afford N,N-dialkyl 3-iodoferrocenecarboxamides.Deprotonation-triggered heavy halogen migration 13 appeared to be a suitable approach for the conversion of 1-substituted 2-iodoferrocenes into their 1-substituted 3-iodo isomers. 14,16A N,N-dialkylcarboxamide being capable of coordinating lithium when located at a neighboring position on a ring, it can contribute to the stabilization of a lithio compound and thus direct halogen migration.To the best of our knowledge, such a group has only been used to direct halogen migration in the case of 3-iodo-N,N-diisopropyl-2pyridinecarboxamide (Chart 1, left). 24According to the generally accepted mechanism, 13 the reaction promoted by a lithium amide (LiDA = lithium diisopropylamide) proceeds through deprotonation at the 4 position and repetitive halogen/metal exchanges.Beside different electronic and geometrical features (e.g.angles), the position next to the carboxamide is locked by the pyridine nitrogen in the reported example whereas it can be attacked by a base in the case of the ferrocenecarboxamides 5-8 (Chart 1).Although silyl protection is possible, we did not consider this strategy as studies showed that hindered ferrocene carboxamides 22 bearing such silanes at C 2 can hardly be deprotonated at C 5 , but rather on the unsubstituted Cp ring.Carboxamide orientation seems to impact the metalation efficiency, the reaction being favored when the C=O group is in the plane of the substituted Cp ring. 25Chart 1. Substrate on which a carboxamide has been used to direct halogen migration (left) and planned substrates to attempt ferrocene halogen 'dance' (right).
Different aspects of the molecular structure of ferrocenes and their derived properties were studied by quantum chemical calculations. 26It is established that ferrocenes can exist as eclipsed and staggered conformations with a low internal rotation barrier. 27In our case (see Supporting Information), the investigated ferrocenes are predominantly in an eclipsed form with hydrogens slightly bent inward.We considered the carboxamide conformation space next.
Due to the presence of the heavy halogen at C 2 in 5, as in the case of silyl-protected carboxamide, the favored conformation seems to have the carboxamide C=O out of the Cp plane (even nearly perpendicular, see Figure 1), and thus not suitable to induce metalation at C 5 .We thus tried to assess, for 1 and 5, the energy difference between the most stable conformation and the conformation with coplanar C=O and Cp ring.Because we could not get this information by using dynamic NMR studies (see Supporting Information), we calculated their energy profiles upon rotation around their C 1 -C=O bond (Figure 2).It was found earlier 28 that for ferrocene carboxamides the bulkier the substituent is, the greater the value of angle between the Cp and amide planes is.
Whereas the computed conformations of lowest energy are very close to the structures obtained by X-ray diffraction, a local maximum (+24 kJ/mol) is noticed for the conformation of 1 with the C=O group in the plane of the Cp (Figure 2, top), and two maxima (local at +30 and global at +50 kJ/mol) were recorded for the two 'in-plane' conformers of 5 (Figure 2, bottom; respectively 0 ° in the case of the C=O group on the iodine side (syn) and 180 ° for the C=O group on the opposite side (anti)). 28This large value led us to suppose that the C=O group can hardly stabilize a 2-iodo-5-lithio-ferrocenecarboxamide by lithium coordination, and we were confident that halogen migration will take place without protective group.
Consequently, we kept -50 °C, and checked different reaction times (5 min, 15 min, 30 min, 6 h, 14 h and 20 h; see Supporting Information).The results show that the 2lithioferrocenecarboxamide is rapidly formed (giving 1 by hydrolysis), but the conversion to the 3-iodo-2-lithio derivative (affording 9 by hydrolysis) takes more time.Once the starting 5 exhausted (~14 h), there is no benefit to use longer reaction times.

Scheme 1. Proposed mechanism for halogen 'dance' on 2iodoferrocenecarboxamides.
Under the optimized reaction conditions, LiTMP proved superior to LiDA, often employed as halogen 'dance' mediator. 13ith LiDA, extending the reaction time favored the deiodinated product 1 to the detriment of the expected 3-iodoferrocenecarboxamide 9 (see Supporting Information).These results, that could be due to the higher propensity of diisopropylamine to protonate the 2-lithioferrocenecarboxamide when compared with LiTMP, led us to abandon LiDA.
One main issue of the approach is to isolate the isomerized products 9-12.Unlike the N,N-diisopropylferrocenecarboxamide 9, separable from the deiodinated compound 1 and the other iodides by column chromatography over silica gel (36% yield at a 1.0 mmol scale; 53% yield at a 4.0 mmol scale), the less hindered N,N-dialkyl ferrocenecarboxamides 10-12 proved much less stable.The N,N-diethylcarboxamide 10 could be purified by column chromatography over silica gel (32% yield at a 1.0 mmol scale), but in a non-reproducible way.Even worse, the N,N-dimethylcarboxamide 11 and the morpholino-based carboxamide 12 were never isolated.As a consequence, crystals suitable for X-ray diffraction were only obtained for the 3-iodo derivative 9 (Figure 1).
To understand why deiodination competes, we treated the ferrocenecarboxamide 1 by LiTMP (1.1 equiv) in THF at -50 °C for 6 h.After subsequent iodolysis, we isolated the 2-iodo derivative 5 in a low 7% yield.Similarly, iodolysis of the halogen 'dance' reaction mixture after 6 h at -50 °C produces much more 1 (~ 60% yield) than 2-lithioferrocenecarboxamide (~ 10%).That the latter easily reacts with H-TMP could explain why 1 first and easily accumulates in the reaction mixture before disappearing.In order to reduce such a protonation, we tried to use 2 equiv of LiTMP.Unfortunately, the formation of 9 is not favored under these conditions.By decreasing the amount of lithium amide to 0.5 equiv, the halogen 'dance' is considerably prevented with 5 remaining present in ~30% yield and deiodination similarly taking place (see Supporting Information).
In most cases, the deiodination is accompanied by formation of unwanted unstable mono-and diiodides.Because we could not isolate any of the side products, we tried to get information from the NMR spectra of fractions containing them.First, we completely assigned the 1 H and 13 C NMR signals of 1, 5 and 9, and deduced the NMR increments of both the CONiPr2 and iodo substituents (see Supporting Information).
In several experiments (notably by using an excess of base to attempt the reaction), we observed the formation (<10% yield estimated by GC mass spectrometry) of an isomer of 5 and 9.This compound was identified as being 1'-iodo-N,Ndiisopropylferrocenecarboxamide by NMR and GC comparison with the product resulting from silyl deprotection of 1'iodo-N,N-diisopropyl-2-(trimethylsilyl)ferrocenecarboxamide.
Concerning the diiodides, we could get information from the 13 C NMR spectra of fractions containing them.In particular, the chemical shift of the C 1 carbon (connected to C=O) largely depends on the presence of iodine atoms at C 2 and C 5 .Indeed, at 81.3 ppm in the absence of neighboring iodine (compound 1), this signal moves to 92.6 ppm in the presence of iodine at C 2 (compound 5), but is not modified significantly with iodine at C 3 (82.8ppm, compound 9).Thus, with a C 1 at 97.2 ppm, we are inclined to think that the diiodide most often formed in our halogen 'dance'-hydrolysis sequences is the 2,5diiodocarboxamide.After halogen 'dance'-iodolysis, a new diiodide is formed (longer retention time on GC mass spectrometry); we supposed it is the 2,3-diiodocarboxamide shown in Scheme 1, formed by iodolysis of the 3-iodo-2-lithioferrocenecarboxamide.Nevertheless, this second diiodide is rarely observed in our halogen 'dance'-hydrolysis reactions; instead, in addition to the 2,5-diiodocarboxamide, a third diiodide is often noticed.We have no clue to identify it, but the 2,4-diiodo derivative could be a possible candidate.This NMR study at least seems to show that deprotometalation of 5 lacks of regioselectivity (next to iodine vs. carboxamide).
To get more information on these diiodides, we attempted the use of the base in situ prepared from ZnCl2•TMEDA (0.5 equiv) and LiTMP (1.5 equiv) 14 to prepare diiodides from the iodocarboxamide 5.Under the conditions used in Table 2, the preponderant formation of 2,5-diiodinated ferrocenecarboxamides was suspected on the basis of the 13 C NMR data of the mixture obtained (peaks at 97.2 and 98.1 ppm for the main polyiodides formed).This evidences a favored kinetic deprotometalation of 5 next to the carboxamide function.
All these observations led us to consider the results in the light of the pKa values of key ferrocene derivatives.In 1973, Denisovich and Gubin described ferrocene as being more acidic than benzene, with a pKa value of 39±3 in MSAD scale (polarography). 31To our knowledge, this represents the only ferrocene pKa determination.Thus, we calculated the CH acidity (pKa values) in THF solution of Fc-H, Fc-I, 1, 5 and 9 within the DFT framework by using the approach elaborated earlier 32 (Chart 2).Whereas iodine exerts its known short-and long-range acidifying effects, the carboxamide function as such acidifies more moderately.When coordinated to lithium (calculations performed by using LiF), the carboxamide becomes a stronger directing group, as shown in Chart 3.
Upon carboxamide coordination to lithium, the 1' position of ferrocene 5 is greatly acidified and amenable to deprotonation; this might explain why 1'-iodo-N,N-diisopropylferrocenecarboxamide is observed in experiments.Besides, when compared to the site next to iodine, the position adjacent to the carboxamide is somewhat more activated; this could explain why deprotonation competitively takes place at C 5 , as demonstrated by 2,5-diiodocarboxamide formation.
Once coordinated to the metal, the C 2 position of the 3iodoferrocenecarboxamide 9 is highly activated; this allows to explain why the equilibria between the different lithioferrocenes are shifted toward the expected 3-iodo-2-lithioferrocenecarboxamide.
The reason why reaction times exceeding 14 h are not suitable to reach high yields could be due to LiTMP destruction, 33 favoring 2-lithioferrocenecarboxamide reprotonation.Competitive iodine/lithium exchange by LiTMP 34 could also be advanced to rationalize the iodine loss observed all along the reaction.
In order to obtain new kinds of ferrocenes, we made derivatives from the 3-iodoferrocenecarboxamide 9 (Scheme 2).Using 4-methoxyphenyl-and 4-(trifluoromethyl)phenylboronic acid in the presence of cesium fluoride, 35 and catalytic amounts of Pd(dba)2 (dba = dibenzylideneacetone) and triphenylphosphine, at the reflux temperature of toluene, 36 respectively afforded the Suzuki coupling products 13 and 14 in moderate yields.The N,N-diisopropylaminomethylferrocene 15 was in turn prepared by reduction of the carboxamide function using BH3•THF, as documented previously.3a Chart 2. Selected calculated pKa values in THF solution for Fc-H, Fc-I, 1, 5 and 9 (top and middle: most stable conformation; bottom: Cp and C=O coplanar -9 2 and 9 5 with the C=O group respectively pointing toward C 2 and C 5 ).

Chart 3. Selected calculated pKa values in THF solution for 5•LiF (most stable conformation of 5) and 9•LiF (Cp and C=O coplanar).
We have thus shown that it is possible to access 1,3disubstituted ferrocenes, which are promising substrates for different applications, 5 with recourse to halogen 'dance'.

CONCLUSION
We studied the halogen 'dance' reaction from different N,Ndialkyl 2-iodoferrocenecarboxamides.In spite of the low stability of the less hindered 3-iodoferrocenecarboxamides, we could optimize the reaction giving 3-iodo-N,N-diisopropylferrocenecarboxamide (9) and identify possible reasons at the origin of the side products formation.
One limit encountered in this halogen 'dance' is the formation of undesirable diiodides, notably due to the relatively high acidity found at C 5 on 2-iodo-N,N-diisopropylferrocenecarboxamide (5).By using deuterium to protect this position toward deprotonation, 4b one could favor the formation of the expected 3-iodo-2-lithioferrocenecarboxamide.
Competitive deiodination and lithioferrocene reprotonation proved to be main issues of the reaction.We will devote efforts in order to identify ferrocene substituents capable of making the generated ferrocenyllithiums more stable toward H-TMP.

EXPERIMENTAL SECTION
General Details.All the reactions were performed under an argon atmosphere using standard Schlenk techniques.THF was distilled over sodium/benzophenone.Column chromatography separations were achieved on silica gel (40-63 μm).Melting points were measured on a Kofler apparatus.IR spectra were taken on a Perkin-Elmer Spectrum 100 spectrometer. 1H and 13 C Nuclear Magnetic Resonance (NMR) spectra were recorded either (i) on a Bruker Avance III spectrometer at 300 MHz and 75 MHz, respectively, or/and (ii) on a Bruker Avance III HD at 500 MHz and 126 MHz, respectively. 1H chemical shifts (δ) are given in ppm relative to the solvent residual peak and 13 C chemical shifts are relative to the central peak of the solvent signal. 37Ferrocenecarboxylic acid is commercially available, but can be easily prepared according to a previously reported procedure. 19or 2-4, 6-9 and 13, the X-ray diffraction data were collected using D8 VENTURE Bruker AXS diffractometer at the temperature given in the crystal data.The samples were studied with monochromatized Mo-Kα radiation (λ = 0.71073 Å).The structure was solved by dualspace algorithm using the SHELXT program, 38 and then refined with full-matrix least-square methods based on F 2 (SHELXL). 39All nonhydrogen atoms were refined with anisotropic atomic displacement parameters.H atoms were finally included in their calculated positions.The molecular diagrams were generated by ORTEP-3 (version 2.02). 40neral procedure 1 (Route A).The protocol was adapted from a previously reported procedure. 41To a stirred mixture of ferrocene (0.93 g, 5.0 mmol) and t BuOK (56 mg, 0.5 mmol) in THF (45 mL) at -80 °C, was added dropwise t BuLi (~1.9 M in pentane, 10 mmol).After 1.5 h at this temperature, the mixture was rapidly transferred dropwise through a cannula to a solution of the required dialkyl carbamoyl chloride (the amount is given in the product description) in THF (15 mL) at -80 °C, and the resulting mixture was allowed to warm to -40 °C before quenching by an aqueous saturated solution of NH4Cl (50 mL).Extraction with AcOEt (3 x 20 mL), washing with brine (20 mL), drying over MgSO4, removal of the solvents and purification by chromatography on silica gel (the eluent is given in the product description) led to the expected compound (see Supporting Information for experimental data on the compounds 2-4).
General procedure 2 (Route B). 20 To a stirred mixture of ferrocenecarboxylic acid (4.6 g, 20 mmol) in CH2Cl2 (50 mL) containing DMF (5 drops), was added dropwise oxalyl chloride (3.5 mL, 40 mmol).After 2 h at room temperature, the solvent and excess of oxalyl chloride were removed under vacuum.The crude ferrocenoyl chloride was next dissolved in CH2Cl2 (65 mL) before dropwise addition of the required amine (60 mmol) at 0 °C.After stirring for 3 h, 1 M aqueous HCl (30 mL) was added.Extraction with Et2O (20 mL) and AcOEt (2 x 20 mL), washing with brine (20 mL), drying over MgSO4, removal of the solvents and purification by chromatography on silica gel (the eluent is given in the product description) led to the expected compound (see Supporting Information for experimental data on the compound 1).
Density Functional Theory (DFT) Calculation Details.All the DFT calculations were performed by using the Gaussian 09 software package. 44The structures from the X-ray diffraction analysis were used as starting guess.All the molecular geometries were completely optimized with no constraints.We used the B3LYP hybrid functional 45 together with the LANL2DZ basis set for both Fe and I, and the 6-31G(d) basis set for the other atoms to calculate the optimized geometries and vibrational frequencies.Relaxed PES scans for 1 and 5 were obtained at the same level of theory.The solvent influence was treated by using the polarized continuum model (IEF PCM) 46 with the default parameters for THF.The pKa values were calculated from the Gibbs energy of the homodesmic reaction between the studied and probe aromatic substrates. 47The single point energies were obtained at the CAM-B3LYP/LANL2DZ + 6-31+G(d,p) level of theory. 48

ASSOCIATED CONTENT Supporting Information
The Supporting Information is available free of charge on the ACS Publications website.
Scheme 2. Conversion of 9 by Suzuki coupling and carboxamide reduction.