J. CHEM. SOC. FARADAY TRANS., 1994, 90(21), 3273-3280 EPR, ENDOR and TRIPLE Resonance Characterization of Three Paramagnetic Redox Stages of 5-Methylene-5H-dibenzo [a,d]cycloheptene M. Luisa T. M. B. Franco and M. Celina R. L. R. Lazana lnstituto Superior Tecnico, Laboratorio de Quimica Orgdnica Av. Rovisco Pais, P-1096 Lisboa Codex, Portugal Radical anions and radical trianions derived from 5-methylene-5H-dibenzo[a,d]cycloheptene 1, lO-deuterio-5-met h y Iene-5H-dibenzo[a,d]c yc loheptene 1 d and 5-d ideu te r iomet h y Iene-5H-dibenzo[a ,d]cyclo he ptene 1d, by reduction with lithium, sodium, potassium and caesium in ethereal solvents have been studied by EPR, ENDOR and TRIPLE resonance spectroscopy. The spin and charge distribution in 1'-and la3-are discussed both in terms of HMO-McLachlan and INDO calculations.The energies of the lowest unoccupied molecular orbital (LUMO) and of the next lowest unoccupied molecular orbital (NLUMO) are influenced differently by the strong interaction of the supercharged radical trianion with the counter-cations. A change in the orbital sequence of LUMO and NLUMO was predicted theoretically for the radical trianion relative to that found for the radical anion, in accordance with the experimental results. The radical cations lo+,Id'+ and Id," obtained via oxidation by AICI, in dichloromethane solution were also studied by EPR, ENDOR and TRIPLE resonance spectroscopy. The experimental results were also interpreted in terms of HMO-McLachlan and INDO methods.Although numerous reports on radical anions and dianions of polycyclic conjugated hydrocarbons have been published, only very limited data on higher negatively charged species have appeared in the literature. Since the earlier reports on the EPR of radical trianions were p~blished,'-~ the interest in these species has increased. In the review published by Gerson and Huber4 only 16 compounds were then referred to as giving radical trianions which had been characterized by ~1.~9~EPR. Since then, Hirayama et have examined the radical trianions of 2,2'-(4,9-dihydronaphtho[2,3-c][1,2,5] thiadiazole-4,9-diylidene)bis(propanedinitrile) and 2,2'-(4H, 8H-benzoC 1,2-c :4,5-c']bis[ 1,2,5]thiadiazole-4,8-diylidene)bis (propanedinitrile). Gerson and co-worker~~.~ described the radical trianions of 1,3,5,7-tetra(tert-butyl)dicyclopenta[u,e] pentalene7 and phenyl-substituted 1,2:9,10-dibenzo[2.2]-paracyclophane- 1,9-dienes'.The radical trianion of 4,7-phe- nanthroline was reported by Fujita and Ohya-Nishiguchig and the radical trianions of thio-and dithio-esters of benzenedi- and benzenetri-glyoxylic acid were studied by Sawluk and Vass." Radical anions, dianions and trianions are known to be powerful bases which are effective in abstracting a proton from very weakly acidic compounds. Nevertheless, they can also undergo electron-transfer reactions which are by far the most common process occurring even in systems favourable to proton abstraction. '' Significant differences in reactivity are expected depending not only on the electron affinity and charge or spin density distribution, but also on ion pairing.The interest in radical trianion research follows from the expected enhanced reactivity of these species compared with the corresponding less reduced radical anions and dianions. Here we report a detailed EPR, ENDOR and TRIPLE res- onance investigation of the radical anions and radical tri- anions which are formed in the course of the reduction of 5-methylene-5H-dibenzo[u,d]cycloheptene 1 with alkali metals in aprotic solvents. This study has been extended to the radical cation obtained by oxidation with AlCl, in CH,Cl, . For the assign- ment of hyperfine splitting (hfs) to protons in the 10,ll and 12 positions, isotopically labelled derivatives were also included, namely l0-deuterio-5-methylene-5H-dibenzo[u,d-J-cycloheptene (Id) and 5-dideuteriomethylene-5H-dibenzo [a,d]cycloheptene (ld2).For the remaining positions the assignment of hfs was based on theoretical spin density calcu- lations carried out by McLachlan and INDO methods. 1 R', R2, R3=H Id R' = D; R2, R3= H ldp R', R2 = H; R3=D Experimental 5-Methylene- 5H-di benzo [a,d] cyclo hep tene (1) was syn t he- sized in this laboratory from SH-dibenzo[a,d]cyclohepten-5-one (A) (Aldrich) according to the procedure reported in the literature.12 The same procedure was used in the synthesis of 5-dideuteriomethylene-5H-dibenzo[u,d]cycloheptene Id, using CD,MgI instead of CH,MgI. The final product was obtained in 75% yield as colourless crystals, mp 112.5-1 13 "C (from 95% ethanol).6, (300 MHz, solvent DCCl,) 6.812 (2H, s), 7.300 (SH, m); m/z 206, 179, 165, 152, 102. Mass spectral analysis and 'H NMR were both consistent with a 99% deu- teriation. The monodeuteriated derivative 1-d was synthesized from the same ketone A, in the following steps: bromination fol- lowed by dehydrobromination of A according to literature procedure^'^ led to l0-bromo-5H-dibenzo[u,d-Jcyclohepten-5-one (B). The Grignard reaction of methyl iodide with B fol-lowed by dehydration of the alcohol with iodine under vacuum, similar to the procedure described in the synthesis of 1,'' gave l0-bromo-5-methylene-5H-dibenzo[u,d]c~clo-heptene (C)in 52% yield.Recrystallization in hexane gave colourless crystals, mp 78-79 "C (lit.14 mp 80-82 "C). 6, [300 MHz, solvent (CH,),CO] 5.309 (2H, s), 7.400 (7H, m), 7.652 (lH, s), 7.843 (lH, d, J 7 Hz); m/z284, 282, 202, 101. 5 cm3 of butyllithium (7.5 mmol drn-,, 1.5 mol dm-j hexane) was added dropwise to a solution of C (2.72 g, 7.5 mmol drn-,) in 75 cm3 dry tetrahydrofuran cooled with a liquid-nitrogen- ethanol solution ( -100"C) under argon. The solution was stirred for 30 min, after which excess deuterium oxide (ca. 5 cm3) was added dropwise. The reaction mixture was further stirred for 30 min at -100°C and was slowly warmed to room temperature. After the solvents were distilled, the residue was extracted with hexane and the new residue after solvent evaporation was recrystallized from 95% ethanol, giving colourless crystals of Id in 72% yield, mp 114-1 15 "C.6, [300 MHz, solvent (CD,),CO] 5.210 (2H, s), 6.874 (lH, s), 7.360 (SH, m); m/z 205, 190, 179, 164, 152, 102. The isotopic purity achieved was 91%, as was verified by 'H NMR. 2-Methyltetrahydrofuran (MTHF), tetrahydrofuran (THF), 1,2-dimethoxyethane (DME) and hexamethylphosphoric tri- amide (HMPT) were dried by known techniques, stored under vacuum over Na-K alloy and distilled under vacuum directly into the sample tube. Acetonitrile (ACN), dimethyl- formamide (DMF) and dichloromethane (DCM), spectro- scopic grade, were further purified by being passed through a neutral alumina column (activity grade 1), under argon, directly into ampoules where each solvent was kept over molecular sieves 3A (Fluka).These solvents were finally dis- tilled under vacuum directly into the sample tubes. Tetra- butylammonium tetrafluoroborate (Bu,NBF,) and sodium tetraphenylborate (NaBPh,) were dried under vacuum at 100°C using P205 as the drying agent. Commercial alu- minium trichloride (Merck) was sublimed under high vacuum into the sample tube. Preparation of Radical Ions for Spectroscopic Studies The radical cations l*+,Id" and Id2*+were generated uia oxidation of the corresponding neutral compounds by AlCl, in CH,Cl,. The radical anions lo-,Id-and ld2*-were immediately obtained via reduction of the neutral compounds by the alkali metals Li, Na, K and Cs in MTHF, THF, DME as well as in mixtures of these solvents with HMPT.Metal films were used in the experiments involving Na, K and Cs, and metal sand was used for the reductions with lithium. The radical trianions lo,-, ldS3-, ld2'3-were produced by pro- longed contact with the alkali metal (after a few hours for reductions with potassium and caesium and after a day for reductions with lithium and sodium). EPR and ENDOR spectra were recorded on a Bruker ER 200D spectrometer equipped with an EN 810 ENDOR unit and an A-300 RF power amplifier. The temperature of the samples was monitored by means of a Bruker variable- temperature unit ER 400 VT. Voltammetric Measurements The cell used for low-temperature cyclic voltammetry was similar to that described by Terahara et all5 with a platinum wire working electrode and counter-electrode instead of gold wire.The preparation of the sample solutions was carried out in the vacuum line. The reduction studies were run in ACN or DMF containing Bu,NBF, (0.1 mol dm-3) as well as DME containing NaBPh, (0.1 mol drn-,). The oxidation of 1 was studied in DCM containing Bu,NBF, (0.1 mol dm-3) as supporting electrolyte. The solvent was distilled directly from an ampoule into the cell containing 1 (lo-, mol drn-,) and the supporting electrolyte. The solutions were degassed by the freeze-pumpthaw method under a dry and oxygen-free argon atmosphere. Ferrocene (Fc) (lo-, rnol dm-,) was added subsequently as an internal standard in order to cali- J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 brate the potential with the Fc+/Fc couple (Eo = 0.400 V us. NHE16). The cyclic voltammograms of 1 were taken in a PAR 273A Potentiostat/Galvanostat interfaced with a PC equipped with PARC 270 Electrochemical Software. Quenching Experiments These experiments were performed using a double Schlenk tube with the two arms separated by a break-seal. One arm contained excess dry methanol (protonation) or excess dry CH,I (methylation) under nitrogen, which were degassed several times before being sealed under high vacuum. The radical trianion was generated in the other arm, as usual. After exhaustive reduction of 1 and observation of the EPR spectrum from the radical trianion, the solution was trans- ferred through the break-seal to the other arm and mixed with the electrophilic reagent.After treatment with water, the organic layer was separated, washed with 5% HCl and water and dried over MgSO,. The residue after solvent evapo- ration was analysed by mass spectroscopy. Reoxidation Experiments These experiments were carried out in a double Schlenk tube with the two arms separated by a break-seal. The reduction of 1 by the alkali metal was carried out, as usual, in one arm. In the other one, two parts of the neutral compound were dissolved in the same solvent under nitrogen, degassed and then the tube was sealed under high vacuum. After 1 had been exhaustively reduced, the EPR spectrum revealed the presence of lo3-which was then mixed with the unreacted solution via the break-seal.The final spectrum showed that lo,-had been reconverted into lo-. Results Cyclic Voltammetry of 1 The cyclic voltammetric measurements were carried out at temperatures ranging from 210 K to room temperature. As the reduced species, which are formed through electron trans- fer, are very sensitive to low concentrations of protic and electrophilic impurities, the system, including cell, solvent and electrolyte, was carefully purified as described in the Experi- mental. During the cathodic sweep, in the voltage range from -1.2 to -3.0 V, three reduction waves were clearly distin- guishable. The first, corresponding to the formation of the radical anion, was clearly reversible with El,, estimated to be -2.26 V in ACN at 210 K and -2.24 V in DMF at 220 K us.NHE. The anodic peaks of the other two waves are very small compared with the cathodic peaks. The values of the observed cathodic peak potentials are -2.41 and -2.85 V in ACN at 210 K and -2.50 and -2.94 V in DMF at 220 K, us. NHE. In DME containing NaBPh, the voltammograms consisted of ill-defined consecutive electron transfers even at temperatures as low as 210 K. The cyclic voltammograms of 1 exhibited an oxidation wave in the voltage range 0.0 to +2.0 V with an oxidation peak potential of + 1.40 V in DCM at 210 K, us. NHE. Radical Anions Under the preparative conditions described in the Experi- mental, the radical anions la-,Id'-and ld2*-were persistent and could be studied by EPR spectroscopy in the tem-perature range 183-298 K.Analysis of the EPR hyperfine pattern was secured by ENDOR and TRIPLE resonance spectroscopy. The relative signs of the hfs were determined by J. CHEM. SOC. FARADAY TRANS., 1994, Fig. 1 EPR, ENDOR and TRIPLE resonance spectra of 1'-K+ in DME at 193 K: (a)EPR spectrum; (b)ENDOR spectrum; (c) general TRIPLE resonance spectrum, irradiating on the line marked with an arrow; (d) special TRIPLE resonance spectrum. general TRIPLE resonance assuming that the largest absol- ute value is negative. Fig. 1 shows the EPR, ENDOR and TRIPLE resonance spectra of 1'-K+ in DME, observed at 193 K. The values of the proton and deuteron hfs are col- lected in Table 1. The proton hyperfine splitting pattern is not very sensitive to changes in the counter-cation, the solvent Table 1 Experimental hfs of the radical anions l'-, Id-and ld2'-obtained by reduction with different alkali metals" hyperfine splitting constant/mT 1'-b -0.196 +0.037 -0.409 +0.119 -0.434 0.119 (2H) -0.433 (1H) Id--0.198 +0.036 -0.408 +0.119 0.064 (1D) 0.119 (2H) ld2*--0.196 +0.037 -0.409 +0.119 -0.434 0.017 (2D) For the assignments see the Discussion section.Identical hyperfine patterns were obtained with Li, Na,K or Cs in MTHF, THF, DME and in mixtures of DME and HMPT. 3275 or the temperature. Assignment of the largest hfs to protons at the 10,ll-positions were based on the effects of selective deuteriation of one of these positions. The effects of labelling on the EPR and ENDOR spectra were a change of the multi- plicity of the proton hfs (0.434 mT) from two to one and the appearance of an hfs of 0.064 mT, corresponding to one deu- teron.The effect of deuteriation of the methylene group was to change the multiplicity of the proton hfs (0.119 mT) from four to two, and for the deuteron hfs of 0.017 mT, with a multiplicity of two, to appear. This observation allowed two protons with an hfs of 0.119 mT to be assigned to the methyl- ene group. The assignment of the remaining hfs was based on theoretical calculations as discussed later. Hyperfine inter- actions due to the counter-cation were not observed. The g value of 1'-K+ in DME at room temperature was found to be 2.0028 fO.OOO1.Radical Trianions Exhaustive reductions of 1, Id and Id, upon prolonged contact of their ethereal solutions with the alkali metal led to the disappearance of the EPR spectra of the radical anions and subsequent observation of new well resolved EPR spectra identified as the radical trianions 1'3-, ldo3-and ld203-. These species were very persistent and could be studied by EPR and ENDOR spectroscopy in the tem-perature range 183-298 K. Contrary to observations for the radical anions, the ENDOR spectra of the lithium and sodium radical trianion gave evidence for the hyperfine inter- action of two counter-cations with two different alkali-metal hfs. On the other hand, the potassium and caesium radical trianions exhibited two identical hfs from the alkali-metal cations.The simulation of the corresponding EPR spectra ruled out any interaction with a third counter-cation despite the expected strong association of the radical trianions with three positively charged counter-cations. When 1'-was reduced in mixtures of ethereal solvents and HMPA, the spectrum indicated the presence of the above species together with a second radical. After some hours, this new species was the only one to be detected and had a lifetime of up to several months. As will be shown below, this corresponds to a radical trianion probably with a different structural arrange- ment. In the absence of HMPT reduction of 1'-also led to this new spectrum after several days which remained unchanged for long periods of time.Table 2 gives the hyperfine data for the two radical tri- anions (hereafter referred to as A and B, respectively) with different counter-cations and solvents. The relative signs and multiplicities of the hfs were determined by general and special TRIPLE resonance, respectively. The EPR and ENDOR spectra of lo3-3Cs+ in DME at 193 K and lo3-3K+ in DME-HMPA at 193 K are shown in Fig. 2 and 3, respectively. The proton hfs vary only slightly with the counter-cation (Table 2) as well as with the solvent and the temperature. However, the alkali-metal hfs are strongly dependent on the environmental conditions. As for the radical anion, from the analysis of the EPR and ENDOR spectra of the monodeuteriated ldo3-and the dideuteriated ld203-, respectively, the largest hfs were assign- ed to protons in the C( 10) and C(11) positions and the smal- lest one to the methylene protons for both radical trianions.The assignment of the remaining hfs was based on theoretical spin density calculations as described in the Discussion. No alternating linewidth effects were observed in the tem- perature range used for the EPR experiments which might give evidence for the existence of any dynamic equilibrium. The g value of l.3-3K+ in DME (species B) at room tem- perature is 2.0040 & 0.0oO1. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Table 2 Experimental hfs for the radical trianions la3-,id3-and Id,''-obtained by reduction with different alkali metals in ethereal solvents' hyperfine splitting constant/mT radical trianion C(1, 9) C(2, 8) C(3, 7) C74, 6) C(l0, 11) C(12) (2H) aM1 aM* lS3--DME Li A B -0.171 -0.23 1 +0.026 +0.059 -0.37 1 -0.417 +0.097 +0.099 -0.463 -0.434 0.010 +0.027 -0.073 0.073 ca.0.002 0.013 Na A B -0.167 -0.235 +0.024 +0.058 -0.367 -0.42 1 +0.092 +0.098 -0.447 -0.45 1 0.010 +0.026 -0.135 0.102 0.024 0.051 K A B -0.158 -0.230 +0.029 +0.059 -0.367 -0.418 +0.096 +0.098 -0.444 -0.437 0.007 +0.026 0.053 0.027 0.053 0.027 cs A B -0.153 -0.232 +0.028 +0.059 -0.366 -0.404 +0.097 +0.097 -0.438 -0.428 +0.015 +0.027 0.280 0.280 0.280 0.280 1-3--DME-HMPT Na A B -0.167 -0.235 +0.026 +0.056 -0.368 -0.419 +0.100 +0.102 -0.445 -0.429 +0.010 +0.026 0.133 0.032 0.036 0.01 3 K A B -0.158 -0.232 +0.030 +0.055 -0.376 -0.419 +0.095 +0.100 -0.443 -0.434 +0.011 +0.027 0.021 0.015 0.021 0.015 cs A B -0.156 -0.232 +0.027 +0.057 -0.367 -0.422 +0.102 +0.102 -0.43 7 -0.437 +0.008 +0.027 0.280 0.280 0.280 0.280 la3--THF Na A B -0.170 -0.235 +0.024 +0.059 -0.370 -0.425 +0.102 +0.104 -0.464 -0.454 0.01 1 0.027 0.088 0.1 10 0.066 0.027 la3--MTHF Na A B -0.174 -0.238 +0.059 +0.023 -0.373 -0.430 +0.095 +0.107 -0.455 -0.456 +0.010 +0.033 0.062 0.105 0.011 -0.050 ld' --D ME K A -0.158 +0.029 -0.367 +0.096 -0.444 (1H) 0.007 0.053 0.053 B -0.230 +0.059 -0.419 +0.098 -0.437 (1H) 0.071 (1D) +0.026 0.027 0.027 0.070 (1D) 1dza3--DME K A B -0.158 -0.232 +0.029 +0.055 -0.367 -0.419 +0.096 +0.096 -0.444 -0.437 -- 0.053 0.027 0.053 0.027 For assignments see Discussion.The identification of the radical trianions was corroborated by three further experiments. Reoxidation of the samples with dry oxygen yielded the starting compound as the sole product at any stage of the reduction. This finding demon- strated that the molecular framework of the highly reduced species remained intact excluding the occurrence of any chemical reaction on the primary radical anion. By reoxidation with two equivalents of the neutral com- pound the EPR spectrum of either radical trianion disap- peared, being replaced by the EPR spectrum of the radical anion, according to the comproportionation reaction 1.3-+ 21 -+31'-(1) This finding gave further evidence for the reversibility of the formation of the radical trianion.Quenching experiments performed by adding electrophilic reagents such as methanol or excess dry methyl iodide allowed the number of negative charges in the highest reduced species to be determined. These reactions were per- formed with solutions exhibiting either EPR spectrum of the radical trianions A and B. Di- and tetra-hydro or di- and tetra-methyl adducts were accordingly obtained, and were identified by mass spectrometry. This result is compatible with the presence of a radical trianion in either case. In effect, in highly reduced n systems such as these, intermolecular electron-exchange processes compete with electrophilic addi- tion reactions according to the following disproportionation equilibrium 21'3--12-+ 14-(2)A even when the anion is slowly added to an excess of the elec- trophile.Radical Cation The radical cations l*+,Id*+and Id," generated in CH,Cl, containing AlCl, were studied by EPR, ENDOR and TRIPLE resonance spectroscopy in the temperature range between 213 K and room temperature. The relative signs and multiplicities of the hfs were determined by general and special TRIPLE resonance, respectively. The value of the proton and deuteron hfs are included in the Table 3. In order to assign the hfs to protons in individual postions of lo+,the effect of selective deuteriation at the C(10) and C(12) posi- tions was investigated whilst the remaining hfs were assigned on the basis of theoretical calculations, which reproduced satisfactorily the observed signs of hfs, as discussed later.The undeuteriated radical cation 1'+has two hfs (a = -0.770 and -0.680 mT) with a multiplicity of one, whereas the remain- ing hfs have multiplicities of four (0.276 mT) and two (0.757, Table 3 Experimental hfs of the radical cations l'+,Id+ and Id,'+ in CH,CI, at 223 K" ~ ~ hyperfine splitting constant/mT 1'+ 0.276 0.115 0.276 0.142 0.680 (1H) 0.757 (2H) 0.770 (1H) ld+ 0.276 0.115 0.276 0.142 0.674 (1H) 0.766 (2H) 0.102 (lD) Id2'+ 0.276 0.115 0.276 0.142 0.683 (1H) 0.109 (2D) 0.803 (1H) a For assignments, see Discussion. J. CHEM.SOC. FARADAY TRANS., 1994, VOL. 90 (c) Fig. 2 EPR, ENDOR and TRIPLE resonance spectra of la3-3Cs+ in DME at 193 K (species A): (a) EPR spectrum; (b) ENDOR spec- trum; (c) general TRIPLE resonance spectrum irradiating on the line marked with an arrow 0.142 and 0.115 mT). This set of hfs denoted a reduced sym- metry of the paramagnetic species, as was clearly shown by its EPR spectrum without a central line (Fig. 4). The two hfs with a multiplicity of one are attributed to the protons in the C(10)-and C(11)-positions while a = 0.757 mT is attributed to the methylene protons. This assignment was made by analysis of the EPR and ENDOR spectra of the radical cations generated from Id and Id, under the same condi- tions. Discussion Radical Anions and Trianioos The observed invariance of the proton hfs of the primary radical anion with the counter-cation, the solvent and the temperature indicated that throughout the temperature range studied these radical anions behaved as solvent-separated ion pairs or even as free-radical anions.The final paramagnetic species obtained upon prolonged reduction with the alkali metals must be identified as the radical trianion. In effect the absence of any structural change resulting from the occurrence of a chemical reaction of the radical anion, together with the reversibility found for the reduction pro- cesses, as well as the results of the quenching experiments, supported its assignment to the radical trianion, 1'3-3M'. Further evidence was given by the observation of two alkali- metal hfs even at low temperatures, e.g.183 K. The observed stability of these radical trianions may result from a strong association with the three positively charged counter-cations. (c1 Fig. 3 EPR, ENDOR and TRIPLE resonance spectra of lS3-3K+ in DME-HMPA at 193 K (species B): (a) EPR spectrum; (b)ENDOR spectrum; (c)general TRIPLE resonance spectrum irradiat-ing on the line marked with an arrow Even in strongly cation-solvating mixtures of HMPT and ethereal solvents such as DME, the interactions of the radical trianion with the counter-ions is very significant. This is sup- ported by the observation of hfs from two Na' ions when the reduction was carried out in this medium (Table 2). The cycloheptatriene ring with an exocyclic carbon double bond must be the relevant structural feature on the anion moiety favouring conversion into the radical trianion.This arrange- ment may be responsible for rendering the energies of the LUMO and of the NLUMO sufficiently close. ZNDO Calculations on the Free-radical Anion 1'-Information about the molecular geometry of the radical species is necessary for the theoretical analysis of the experi- mental hfs. However, while unambiguous structural determi- nation of these paramagnetic species in solution is not possible, the only information has been derived from the theoretical spin-density calculations performed in order to fit the experimental results. Knowledge of hfs including their signs as well as the assignment of the hfs corresponding to C(lo), C(11) and C( 12) positions provided the experimental support for discussing the molecular geometry.The INDO method seems to be the most suitable for spin density calcu- lations on these non-planar systems. Even for the parent neutral compound 1, there is no single-crystal X-ray structure determination reported in the literature. For that reason we began the calculations with the X-ray structural data published for the related compound 4-(5-rnethylene-W- dibenzo[a,dJcycloheptenyl)-1-methylpiperidine. ' Accord- J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 "H I I' (d ) Fig. 4 EPR, ENDOR and TRIPLE resonance spectra of 1'+ in CH,Cl, containing AlCl, at 223 K: (a) experimental EPR spectrum; (b) simulated EPR spectrum with linewidth 0.015 mT and the hfs given in Table 3; (c) ENDOR spectrum; (d)special TRIPLE reson- ance spectrum ingly, the central seven-membered ring adopted a boat con- formation with dihedral angles, 8, and e,, of 26.9" and 50.6", respectively, and a dihedral angle, 8,, between the plane of the two benzene rings of 124.1" (Fig.5). Both the order and the signs of the INDO calculated hfs for the free-radical anion, assuming a geometry analogous to that of the above related compound, showed poor correlation with experimental results (Table 4). For that reason we have tried to improve the INDO calculations by making some Fig. 5 Configuration of the central seven-membered ring for 4-45 methylene-5H-dibenzo[a,~cycloheptenyl)-1-methyl-piperidine' adjustments to the structure, particularly by varying the dihe- dral angles el, 8, and 8, .t The range of variation of 8, and 8, was restricted owing to the steric interaction between the protons bound to C(12) and C(4) or C(6) (e.g., for 8, =30", 8, (140" in order to have distances greater than 2 A between the sterically interacting hydrogen atoms).8,,19, and 8, were varied in the ranges 30-lo", 60-10" and 160-120", respec-tively. A much better agreement between the observed and calculated hfs was obtained for a more flattened structure with 8, =20", 8, =30" and 8, =140". This structure is illus- trated in Fig. qa). An adjustment was also made to the C(lO)-C(ll) bond length in order to improve the results [C( lo)--( 11) =1.366 A].The calculated hfs assuming such a structure are shown in Table 4. INDO Calculations on the Contact Radical Trianion lo,-3Li' The molecular geometry of the two radical trianions was much more difficult to establish owing to the uncertainty of the location of the three associated counter-cations. There- fore we only adjusted the calculations for the radical trianion (a) (b ) Fig. 6 Molecular structure assumed in the INDO calculations: (a) radical anion; (b) radical trianion fGeometry optimization by MIND0/3 was also tried, but it was restricted to the ionic R system neglecting both ion pairing and solva- tion effects. The INDO calculated hfs based on such a structure were substantially improved, but the magnitude of the hfs still showed large errors (Table 4).Table 4 INDO calculated hfs for the radical anions 1'-and lS3-assuming different structures hyperfine splitting constant/mT ~~ ~ 1'-l2 +0.214 -0.3 10 +0.065 -0.028 +0.843 1'-b -0.355 +0.241 -0.429 +0.284 -0.347 1'--0.241 +0.151 -0.337 +0.181 -0.397 1'3 --0.29 1 +0.164 -0.309 +0.214 -0.428 l2 Geometry based on the X-ray structure published for the related compound cited in the text.I7 Geometry adjusted as described in the text [Fig. qa)]. ~ ~~ -0.554 ---0.045 --0.099 -0.03 1 Li( 1) -0.01 Li(2) -0.02 Li( 3) -0.036 ~ ~ ~~____________ Optimized geometry by MIND0/3. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 referred to as A.As a starting point, any change in the geometry of the anion moiety which might occur on attrac- tion of the counter-cations was neglected. Note that the cal- culations performed were restricted to the lithium radical trianion owing to the limitations of the available QCPE INDO version which includes only first-row atoms. Note, also, that every satisfactory structure we find is only from a theoretical approach. The location of the counter-cations was tentatively deduced from simple models of electrostatic inter- action and symmetry arguments. Hence, they must occupy positions in the mirror plane of the molecule since no asym- metric spin density distribution was observed. However, they are expected to be situated in the proximity of the highest negative charge-density carbon atoms.Fig. 7 shows the excess negative charge distribution in the radical trianion calculated by the INDO method, excluding the influence of any counter-cation. The largest calculated excess negative charge lies on the set of the exocyclic carbon atom C(12) and its vicinal hydrogen atoms; the next largest excess negative charge lies on the pairs of carbon atoms of the central seven-membered ring C(4a)-C(5a) and C( 10)- C(11) with its bonded hydrogen atoms. Some positions of the benzo rings such as C(2) and C(3) [or C(7) and C(8)] and neighbouring hydrogen atoms are also expected to have considerable excess negative charge density. On the other hand, the folded structure of the mol- ecule (as illustrated in Fig.6) has two sterically different sides relative to the counter-cation attraction, the concave side of the central seven-membered ring being less sterically hindered than the convex one. Therefore, we considered that one Li ion [Li(l)] is more associated with the C(12) atom, whereas the other two [Li(2), Li(3)] are polarized by the remaining excess negative charge centres. We have looked for positions of the Li ions which yielded the best overall agreement of all hfs with experiment. From the countless structures attempt- ed, the one that gave the best fit with experimental results is represented in Fig. qb). In this structure the Li ion associated with C(12) and the neighbouring hydrogen atoms was con- sidered to be situated in the concave side and considerably deviated away from the bulky molecule.The most suitable distances between Li ions and the neighbouring atoms were found to be Li(1)-C(12) = 2.05 A; Li(2)-C(10) = 2.47 8,; Li(3)-C(4a) = 2.52 A whereas Li(1)-Li(2) = 3.13 A. In this structure some adjustments to C(5)-C(12) (1.44 A) and C(lO)-C(ll) (1.41 A) bond lengths were finally made. The calculated hfs are included in the Table 4. MO Calculations The hyperfine data determined experimentally for 1'-and 1'3 -reflect, respectively, the shapes of the LUMO and NLUMO of 1. These MOs were calculated by the Huckel- McLachlan's procedure so that a satisfactory correlation was obtained between the theoretical and experimental hfs. The resonance parameters, ycc , used in these calculations were based on the molecular geometry referred to above as giving the best fit on the INDO calculated hfs.They were corrected -0.142.--0.098 Fig. 7 Excess negative charge distribution in the radical trianion both for the bond length and the twist angle between the two p-orbitals of the bond, according to Streitwieser." A positive Coulomb parameter on the exocyclic carbon atom C(12) was used in order to take into account the strong electron- donating influence of the cycloheptatrienyl ring. A better fit was obtained by further adjustment of the Coulomb param- eter on the remaining carbon atoms according to the w-technique'* with o = 0.9. The resulting LUMO and NLUMO of 1, neglecting the effect of counter-cation associ- ation, are schematically depicted in Fig.8(a). The areas of the circles are proportional to the squares of the LCAO coeff- cients. Blank and filled circles symbolize the different signs of the coefficients. Whereas the hfs of the protons in 1.-, listed in the Table 1, are consistent with the single occupancy of the LUMO, as shown in the Table 5, those for the radical tri- anion lS3-,included in Table 2, disagree with the ones calcu- lated for the NLUMO. Nevertheless, a good correlation was obtained when the effect of the tighter ionic association in the radical trianion was considered in the MO calculations. The cation association was introduced as usual by further adjust- ment of the Coulomb parameters with positive corrections on the neighbouring carbon atoms, in order to account for the +I effect of the positive charge.As illustrated in Fig. 8(b), the calculated energetic sequence of the MO is interchanged. Two reasons for this orbital exchange can be advanced: (i) the LUMO and NLUMO in 1 are almost degenerate, being separated by only a small energy gap (0.025 b) and (ii) the LUMO has two nodes on the C(5) and C(12) positions whereas these carbon atoms have high orbital density in the NLUMO. As discussed previously, those positions are expected to be the most perturbed by ion pairing. Therefore, while the NLUMO of 1 is strongly stabilized owing to the attraction between the oppositely charged ions, the energy of the LUMO is almost unaffected, the cation effect being SUE-cient to shift the Huckel energy of the NLUMO below that of the LUMO as shown in Fig.8. The good fit obtained in the hfs thus calculated for the radical trianion la3-3M+, included in Table 5, allowed the assignment of its SOMO to the NLUMO of Fig. 8(b). The ion-pairing effect decreases the energy necessary to transfer the second and third electrons in the successive reduction steps of the radical anion. Radical Cations Reduced symmetry was observed in the EPR and ENDOR spectra of 1'+ relative to the parent compound. Analysis of the hyperfine pattern of Id' allowed the identification of the NLuMo* @E = a-0.736p E = a-0.6448 LUMo m m E = a-0.71 18 E = a-0.36Sg (a1 (b1 Fig. 8 Schematic presentation of the LUMO and the NLUMO of 1 calculated by the Huckel model (the parameters used in these calcu- lations are included as footnotes to Table 5): (a) not taking into account the counter-cation association ; (b) taking into account counter-cation association J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Table 5 Calculated hfs by the Huckel-McLachlan procedure, assuming different singly occupied MOs (SOMO) hyperfine splitting constant/mT HOMO" -0.085 +0.013 -0.228 +0.127 -0.566 0.906 HOMO*-' -0.1 34d + 0.M6d -0.230" +0.120" -0.625(10) -0.759 consistent with SOMO of 1'+ -0.798(11) LUMO~.~ -0.201 + 0.067 -0.364 + 0.061 -0.375 +0.029 consistent with SOMO of 1'-NLUMOB + 0.066 -0.303 +0.335 -0.3 10 + 0.303 -1.739 NLUMO~.' -0.189 + 0.058 -0.338 + 0.067 -0.436 +0.010 consistent with SOMO of 1'3-~~ "HMO parameters: Y,,~= Y~,~=~S,~=YS.,~= ~5.44=~5.5a=O-79;1-03; ~2.3=~3,4=~6.7=~7.8=~44,iia=~sa.9a=0~97~ ~io,9a= yll, = 0.84; ylo, ,, = 0, 9; y5.12 = 1.1; a,, = 0.22; 6,, = a,, = 0.15; QzH = -3.0 mT. HMO parameters as " except a,, = -0.05 and 6, = 0.1. aH(l'+)satisfactorily correlate with these calculated values. Two very close hfs were calculated for the two positions. These are the corresponding mean values. HMO parameters: yl, = 78, = 1.03; Y2, = Y3.4 = 76, = Y7,8 = 0.97; Y4.Q = Y6, 5" = 0.968; 75. = Y5.50 = 0-76; Y&, 11" = 75". 9a --0.975; ylo, 9a = y,,, = 0.865; ylo, ,, = 1.04; y5, 12 = 1.08; = 0.014; a,, = S,, = -0.276 (calculated by the o-satisfactorily correlates with these calculated values. HMO parameters as for '.HMOtechnique'8 with w = 0.9); Q:H = -2.3 mT. aH(lo-) parameters as for e except a,, = 1.25; a,, = a,, = -0.176; y5, 12 = 0.9; ylo, ,, = 1.0. ' aH(lS3-)satisfactorily correlates with these calculated values. non-equivalence of the two vicinal carbon atoms C(10) and already used for the radical anion and calculating the C(1 1) (Table 3). These two positions have simultaneously Coulomb parameters by the o-techniqueI8 with o = 0.9 as high spin density and the highest excess positive charge well. The loss of symmetry referred to above can be ascribed density. to some specific interaction with the medium, namely with a Fig. 9(a) shows the HOMO of the radical cation 1" calcu-counter-anion such as AlCI,-, the coordination being prefer- lated by the Hiickel-Mclachlan approximation.The areas of entially oriented to one of the highest positively charged the circles are proportional to the squares of the LCAO coef-C(l0) or C(11) positions. The effect of this perturbation was ficients. Blank and filled circles symbolize the different signs simulated using the HMO approximation by an additional of these coefficients. The excess positive charge distribution is negative Coulomb parameter defining this carbon atom. The indicated in Fig. 9(b) and is obviously symmetric. These cal- HOMO thus calculated is shown in Fig. 9(c). The calculated culations were performed assuming the resonance parameters hfs, assuming the single occupancy of this HOMO, correlated well with the observed hyperfine pattern for 1'+ as shown in Table 5 which substantiates the arguments used.References 1 F. Gerson, R. Heckendorn, D. 0. Cowan, A. M. Kini and M. Maxfield, J. Am. Chem. SOC., 1983,105,7017. 2 W. Huber, Tetrahedron Lett., 1983,24, 3595. 3 W. Huber, Helv. Chim. Acta, 1983,66, 2582. 4 F. Gerson and W. Huber, Acc. Chem. Res., 1987,20, 85. +O. 149 5 M. Hirayama, A. Seki, Y. Yamashita, T. Suzuki and T. Miyashi, Chem. Lett., 1988,67 and 769. 6 M. Hirayama, A. Seki, Y. Yamashita, T. Suzuki and T. Miyashi, J. Chem. SOC.,Chem. Commun., 1988,490. 7 F. Gerson, G. Gescheidt, K. Hafner, N. Nimmerfroh and B. Stowasser, Helv. Chim. Acta, 1988, 71, 101 1.+0.027 8 A. de Meijere, F. Gerson, B. Konig, 0. Reiser and T. Wellauer, +0.040 J. Am. Chem. SOC., 1990,112,6827. 9 H. Fujita and H. Ohya-Nishiguchi, J. Chem. SOC., Chem. Commun., 1989,1091. 10 A. Sawluk and J. Voss, Angew. Chem., Int. Ed. Engl., 1989, 28, 906. 11 M. Szwarc, Ions and Ion Pairs in Organic Reactions, 1974, vol. 2, Wiley, New York, ch. 1. 12 A. C. Cope and S. W. Fenton, J. Org. Chem., 1951, 1673. 13 W. Treibs and H-J. Klinkhammer, Chem. Ber., 1951,84, 671. 14 F. Hoffmann-La Roche and Co., A.G.; Belg. Pat. 659 599, 1965; Swiss Appl., 1964; Chem. Abs., 64, P 5023g. (c1 15 A. Terahara, H. Ohya-Nishiguchi and N. Hirota, J. Phys. Chem., 1986,90,1564. Fig. 9 Schematic presentation of the HOMO of 1 and excess posi- 16 R. R. Gagne, C. A. Koval and G. C. Lisensky, Inorg. Chem. tive charge distribution in 1" calculated by the Huckel model (the 1980,19,2854. parameters used in the calculations are included as footnotes to 17 B. Birknes, Acta Crystallogr., Sect. B, 1977,33, 687. Table 5): (a) calculated HOMO, not taking into account any 18 A. Streitwieser Jr., Molecular Orbital Theory for Organic Chem- counter-anion association; (b) calculated excess positive charge dis-ists, Wiley, New York, 1961, ch. 4. tribution, not taking into account any counter-anion association; (c) calculated HOMO taking into account counter-anion association Paper 4/02896A; Received 16th May, 1994