Fresenius J Anal Chem (1999) 364 : 219–223

© Springer-Verlag 1999

O R I G I N A L PA P E R

Mikael T. Harju · Peter Haglund

Determination of the rotational energy barriers of atropisomeric polychlorinated biphenyls

Received: 19 October 1998 / Revised: 26 November 1998 / Accepted: 28 December 1998

Abstract The rotational energy barriers were determined for twelve of the nineteen environmentally stable atropisomeric polychlorinated biphenyls (PCBs), viz. PCB 84, 131, 132, 135, 136, 144, 149, 174, 175, 176, 183, and 196, by thermal racemization of enantiomerically pure PCBs. The rate of racemization was primarily determined using off-line gas chromatography (GC) or high-performance liquid chromatography (HPLC), with permethylated cyclodextrin (PMCD) as the chiral selector. GC was used for PCB 84, 132, 136, 149, 174, and 176, while PCB 131, 175, and 196 were analyzed using HPLC. The remaining PCBs, i.e. PCB 144 and 183, were separated by HPLC using a Chiralpak OP(+) polymethacrylate column. Gibbs free energy of activation (∆‡G) was 176.6 to 184.8 kJ/ mole for tri-ortho PCBs. For tetra-ortho PCBs the ∆‡G was estimated to be ~246 kJ/mole. A buttressing effect of 6.4 kJ/mole was observed for tri-ortho PCBs that have one buttressing chlorine.

1 Introduction Seventy eight of the 209 polychlorinated biphenyl (PCB) congeners display axial chirality and it is predicted that 19 exist as stable atropisomers at physiological temperatures due to restricted rotation about the central C-C bond [1]. Atropisomers are a type of conformational isomers, which are defined as molecules with axial chirality in which rotation about the axis is hindered by steric congestion [2]. In biphenyls the restricted rotation is sustained by bulky atoms in the ortho positions of the phenyl rings. Oki predicted the existence of atropisomerism, where isomers can be isolated, for compounds that have a racemization halflife t1/2 > 1000 s or 93.3 kJ/mole at 300 K [3]. Only the 19 PCBs that contain three or four ortho chlorine atoms

M. T. Harju (Y) · Peter Haglund Umeå University, Institute of Environmental Chemistry, SE-901 87 Umeå, Sweden e-mail: [email protected]

fulfil this criterion. Thermal racemization experiments have shown the rotational energy barrier for one of the triortho PCBs, viz. 2,2′,3,3′,4,6′-hexachlorobiphenyl (PCB132 [4, 5]), to be 180 kJ/mole [6]. Tetra-ortho PCBs are expected to have even higher rotational barriers since these molecules cannot relax by bending the central C-C bond as it passes the planar transition state. Similarly, certain atropisomeric PCBs with adjacent ortho and meta substitutents have higher barriers than PCBs that only have ortho substitutents. In these congeners the meta chlorine substitutents impede the outward bending of ortho chlorine substitutents in transition state, and raises the rotational energy barrier. This type of steric congestion is referred to as the “buttressing effect”. Of the conformationally stable atropisomeric PCBs nine are present in commercial PCB mixtures above 1% concentration, and are therefore expected to be released to the environment [1]. High concentrations of PCB 132 and 149 have been found in human milk [7] and blue mussels [8], respectively. Many toxic effects have been associated with the PCBs, such as: carcinogenicity, endocrine effects, and disturbances on the reproduction and immune system [9]. PCBs are also potent inducers of drug metabolizing enzyme systems [10], and the activity of the different congeners depends on the chlorine substitution pattern, especially on the number of ortho chlorine substitutents. It is assumed that the coplanar conformation is less populated or even unpopulated if a PCB contains several bulky substitutents in the ortho positions. Generally, PCBs with more than two ortho substitutents, e.g. 2,2′,4,4′-substituted PCBs, induces phenobarbital (PB) type of responses, while the nonortho PCBs are methylcholanthrene (MC) type inducers, and mono-ortho PCBs are “mixed type” inducers [10]. Enantioselective biological activity has been reported for some atropisomeric PCBs, i.e. PCB 88, 139 and 197 [10, 11]. The tested enantiomers exhibit enantioselective effects on cytochrome P-450 induction, and etoxyresorufin-O-deethylase (EROD), benzphetamine N-demethylase (BPDM), aldrin epoxidase, and aminopyrine N-demethylase activity [11]. However, the reported differences in induction strength do not necessarily reflect differences in

220

potency since pharmacokinetic parameters, such as enantioselective metabolism or transport might influence the observed activity. Enantioselective transformation processes in biota result in an enantiomeric ratio (ER) different from racemic, a definition of ER is found in [12]. An ER of 1.2 was observed for PCB149 in blue mussel samples from the German Bight [8]. Human milk samples have also been investigated, with similar results, excess (ER = 1.2 to 2.2) of one of the PCB132 enantiomers [7]. The aim of the present study was to determine the enantiomeric rotational energy barriers of 12 of the 19 triand tetra-ortho atropisomeric PCBs by determining the rate of racemization at elevated temperatures. The results will confirm that the PCB atropisomers are stable enough to resist enantiomerization at physiological temperatures, and will also indicate whether or not precautions are needed to prevent racemization during gas chromatographic (GC) analyses of environmental samples.

2 Materials and methods 2.1 Chemicals All PCB congeners used, viz. PCBs 84, 131, 132, 135, 136, 144, 149, 174, 175, 176, 183, and 196, were obtained as racemic compounds (> 98% purity) from Accustandard (New Haven, CT, USA). A mixture of the technical PCB products Arochlor 1242, 1248, 1254, and 1260 (1 : 1 : 1 : 1 w/w/w/w; A1111) was prepared according to Schwartz et al. [13]. All solvents used were of high quality: n-docosane and n-hexatriacontane were of 99% purity obtained from Aldrich-Chemie GmbH (Steinheim, Germany). Dichloromethane and n-hexane of HPLC grade were supplied by Labscan Ltd. (Dublin, Ireland), glass distilled grade toluene was from Burdick and Jackson (Muskegon, MI, USA), and analytical grade methanol was from Mallinckrodt Baker B.V. (Deventer, Holland). Water was purified using a Milli-Q apparatus (Millipore, Bedford, MA, USA). 2.2 HPLC isolation of atropisomeric PCBs Chiral separation and isolation of pure PCB enantiomers were performed by reversed-phase high-performance liquid chromatography (RP-HPLC). Two different columns were used for enantiomer

Table 1 PCB optical rotation properties (+/–), separation technique, resolution (R), and selectivity factors (α)

In all cases, the second eluting enantiomer was used in the experiments

separation: a 4.6 × 250 mm Nucleodex β-PM column (MachereyNagel, Düren, Germany) packed with a permethylated β-cyclodextrin stationary phase (PMCD), and a 4.6 × 250 mm Chiralpak OP(+) column (Daicel Chemical Industries, Ltd) packed with a (+)-poly(diphenyl-2-pyridylmethyl methacrylate). The PCBs 84, 131, 132, 135, 136, 174, 175, 176, and 196 were previously isolated using the PMCD column, as described elsewhere [14], whilst the OP(+) column was used during the isolation of PCBs 144, 149, and 183. Saturated methanolic solutions of these congeners (5 mg/mL in methanol) were separated using 90/ 10 (v/v) methanol/water at a flow rate of 0.50 mL/min, and a column temperature of 0 °C. Aliquots of 5 or 10 µL were injected, fractions were collected manually, and the appropriate fractions were pooled. Totally 350–400 µg of each PCB was isolated for use in the subsequent experiments. The optical purity of all PCBs used was better than 98%. 2.3 Thermal racemization experiments The rotational energy barriers were determined for one of the enantiomers of each of the atropisomeric PCBs 84, 131, 132, 135, 136, 144, 149, 174, 175, 176, 183, and 196. The method used was based on an enantiomerization kinetic study on PCB 132 performed by Schurig et al. [6]. In our experiments we used n-docosane (C22H46, Bp 369 °C) as solvent for the tri-ortho PCBs, and n-hexatriacontane (C36H74, Bp 450 °C) as solvent for the tetra-ortho PCBs. The second eluting enantiomers, on the chromatographic systems used, were selected for the studies of the enantiomerization kinetics due to the relative ease of area determination of the first eluting enantiomer, which has build-up during the experiment. About 200 ng, or 2000 ng if the enantiomers had to be analyzed by HPLC, were dissolved in 250 µL of high boiling alkane, and were transferred to 1.2 mL ampoules. The ampoules were argon flushed and flame sealed. The enantiomerization kinetics of the tri-ortho PCBs was studied by heating the ampoules at different temperatures (270 to 300 °C), and length of time (32 to 128 min). Due to the very high enantiomerization barrier for the tetra-ortho PCBs (PCB 136 and 176) these were only treated at the maximum achievable temperature (425 °C) for 21 h, to check if any conversion at all occurs at this temperature. The heat-treated samples were dissolved in 1 mL n-hexane. The n-docosane, or n-hexatriacontane was removed by high-resolution gel permeation chromatography (HR-GPC), to allow splitless GC injection and avoid chromatographic interferences. The separation was obtained using two serially connected 300 × 7.8 mm HR-GPC columns (5 µm polystyrene-divinylbenzene copolymer, 50 Å pore size, Polymer Laboratories, Church Stretton, UK) eluted with 50/50 (v/v) n-hexane/ dichloromethane, at a flow rate of 0.7 mL/min. The PCBs were recovered by collecting the eluent between 18 and 25 min.

IUPAC

Enantiomer Separation technique opt. rot.

R

α

84 132 135 136 149 174 176 131 144 175 183 196

(+) (+) (–) (+) (+) (–) (+) (+) (+) (+) (+) (+)

1.12 1.81 0.97 1.16 1.65 0.86 0.85 0.99 1.27 1.35 1.46 1.46

1.02 1.01 1.01 1.02 1.03 1.01 1.01 1.08 1.20 1.10 1.23 1.11

GC, Chirasil-Dex, isothermal 145 °C (65 min) GC, Chirasil-Dex, 130 °C-1 °C/min-185 °C GC, Chirasil-Dex, isothermal 150 °C (60 min) GC, Chirasil-Dex, isothermal 150 °C (60 min) GC, Chirasil-Dex, isothermal 150 °C (70 min) GC, Chirasil-Dex, isothermal 160 °C (90 min) GC, Chirasil-Dex, isothermal 160 °C (70 min) LC, Nucleodex β-PM LC, Chiralpak OP(+) LC, Nucleodex β-PM LC, Chiralpak OP(+) LC, Nucleodex β-PM

221 2.4 Enantioselective GC and HPLC analysis The products of the enantiomerization experiments were analyzed by either GC with an electron capture detector (ECD), or HPLC with UV detection (λ = 210 nm), c.f. Table 1. For the GC analyses a 25 m × 0.25 mm CP Chirasil-Dex CB fused silica column, surface modified with covalently bond PMCD (Chrompack, NL), was used with hydrogen as the carrier gas. The GC oven temperature program was optimized for each of the PCB enantiomers, and the best separation was generally obtained with a long period of isothermal operation at 145 °C to 160 °C. For the HPLC analyses the same equipment and experimental conditions were used as for the isolation of the pure enantiomers, see above, except for a slightly higher flow rate, 0.40 and 0.80 mL/min for the PMCD and OP(+) analyses, respectively. 2.5 GC/mass spectrometry analysis of thermolysis products GC/mass spectrometry was used to screen the heat-treated tetra-ortho PCB samples for thermal degradation products. The analyses were performed using a GC- quadrupole MS system (Fisons GC8000/ MD800, Manchester, UK). The GC was equipped with a 60m × 0.32 mm Rtx-5 capillary column, 0.25 µm film thickness, obtained from Restek Corporation (Bellefonte, USA). Analyses were performed in the single ion recording (SIR) mode under electron ionization (EI) conditions (70 eV). The two most intense ions of the molecular ion isotope distribution clusters were monitored for each homologue (tri- to nonachloro CB). The oven was temperature programmed, as follows: 80 °C (2 min), 15 °C/min to 205 °C, 2 °C/ min to 275 °C, 15 °C/min to 300 °C, isothermal for 10 min. Injections of 1 µL aliquots were performed in the split-less mode, and helium was used as carrier gas at a head pressure of 11 psi. The test solution A1111 was used for peak identification.

Fig. 1 Partial GC chromatograms of (+)-PCB 149 treated for different times (0–128 min) at 290 °C

All of the atropisomeric PCBs were successfully separated by either enantioselective RP-HPLC or GC. The OP(+) HPLC column resolved the enantiomers of PCB 144 and 183 (α = 1.2, R = 1.2), and the PMCD column separated the enantiomers of PCB 131, 175, and 196 (α = 1.1, R = 1). The remaining atropisomers, PCB 84, 132, 135, 136, 149, and 176, were separated by GC on the Chirasil-Dex PMCD column (α = 1.0, R = 0.8 to 1.8), c.f. Table 1.

sulted in racemization half-lives between 7300 and 370 s, c.f. Table 2 and Fig. 1. The reaction rate constants range between 4.73 × 10–5 and 9.26 × 10–4 (s–1). The racemization half-lives of the tetra-ortho PCBs 136 and 176, at 425 °C, was 56000 s (16 h) and 70000 s (19.6 h), with corresponding reaction rate constants of 6.14 × 10–6 and 4.91 × 10–6 (s–1), c.f. Table 2. The regression coefficients (r2) for the tri-ortho PCBs ranged between 0.82 and 1.00. The ∆‡G values for tri-ortho PCBs were in the interval 176.6 to 184.9 kJ/mole with a variance (p = 0.05) of 0.8 to 3.5 kJ/mole, and the values for the tetra-ortho PCBs were ~246 kJ/mole, see Table 2. Interestingly, our value for PCB132 (184.9 kJ/mole) is in close agreement with the value reported by Schurig [6] (183.7 kJ/mole). Further, in all cases the rotational energy barrier is far above what is required for environmental stability, i.e. 93 kJ/mole [3]. Many groups have tried to predict the rotational energy barriers of tri-ortho PCBs. The first to predict the energy barrier was Kaiser [1], who estimated the barrier to be 105 to 121 kJ/mole. Similar energy estimates have been achieved using AM1 semi-empirical calculations (109 to 117 kJ/mole) [16]. However, another investigation utilizing the same technique suggested a significantly higher interconversion barrier of 189.7 to 206.8 kJ/mole [17]. Semi-empirical calculations performed with the INDO method resulted in an intermediate value of 132.2 kJ/mole [18]. Thus, the theoretically calculated rotational energy barriers for tri-ortho PCBs are sometimes close to the experimental values obtained in this work, but usually quite far off. A compilation of experimentally determined and theoretically calculated data is shown in Table 3. A few attempts have been done to predict the rotational energy barriers of tetra-ortho PCBs. Tang et al. [19] estimated an energy barrier of 164 kJ/mole using the AM1 method, whilst values close to the experimental results (246 kJ/ mole) have been suggested by Kaiser (243 kJ/mole)[1], or INDO calculated by Cullen et al. (232 kJ/mole) [18].

3.2 Thermal racemerization experiments

3.3 Buttressing effect

The thermal racemization experiments performed on the tri-ortho PCBs in the temperature range 270 to 300 °C re-

Some atropisomeric tri- and tetra-ortho PCBs are expected to exhibit a buttressing effect, e.g. PCB 84, and some

2.6 Calculations The rotational energy barriers of PCB enantiomers were determined by investigating the temperature dependence of the racemization rate constant of enantiomerically pure compounds. The enantiomerization kinetics can be expressed as a reversible firstorder reaction where the initial enantiomer concentration is [A0], and [A] is the enantiomer concentration at time t (s). The first-order rate constant k (s–1) is obtained from a plot of ln{[A0/([A]02[A])} vs. t. The slope of such a plot is 2k (the rate of racemization) [2]. Using the Arrhenius expression [15] the Gibbs free energy of activation (∆‡G) could be calculated from the rate constant k and the racemization temperature.

3 Results and discussion 3.1 Enantioselective RP-HPLC and GC analysis

222 Table 2 Enantiomerization rate constants, racemization half-life (t1/2), Gibbs free energy of activation (∆‡G) and the regression coefficient (r2) for each PCB and temperature

IUPAC

Tempera- Rate constant ture ( °C) (s–1)

r2

t1/2 (s)

∆‡G (kJ/mole)

Average ∆‡G ± confidence limit (p = 0.05) (kJ/mole)

84

280 290 300

6.24 × 10–5 1.24 × 10–4 2.07 × 10–4

0.980 0.962 0.999

5550 2800 1670

182.8 183.0 183.9

183.3 ± 1.4

131

280 290 300

6.22 × 10–5 1.15 × 10–4 2.15 × 10–4

0.979 0.995 0.995

5570 3010 1610

182.9 183.4 183.7

183.3 ± 1.1

132

280 290 300

4.73 × 10–5 7.70 × 10–5 1.61 × 10–4

0.977 0.995 1.000

7330 4500 2150

184.1 185.1 185.1

184.9 ± 1.6

135

280 290 300

4.81 × 10–5 9.32 × 10–5 1.39 × 10–4

0.985 – 0.910

2710 3720 2490

184.1 184.4 185.8

184.7 ± 2.3

144

270 280 290 300

1.10 × 10–4 2.05 × 10–4 3.66 × 10–4 7.32 × 10–4

0.990 0.949 0.963 –

3150 1690 947 474

176.9 177.4 178.0 177.9

177.5 ± 0.8

149

270 280 290 300

1.20 × 10–4 2.28 × 10–4 5.26 × 10–4 7.69 × 10–4

0.999 0.998 0.993 –

2890 1520 659 451

176.5 176.9 176.3 177.7

176.8 ± 0.8

174

280 290 300

5.45 × 10–5 1.53 × 10–4 1.71 × 10–4

0.999 0.949 0.820

6360 2270 2030

183.5 182.1 184.9

183.5 ± 3.5

175

280 290 300

7.55 × 10–5 1.70 × 10–4 2.38 × 10–4

0.940 0.953 0.970

4590 2040 1460

182.0 181.5 183.3

182.3 ± 2.2

183

270 280 290 300

1.34 × 10–4 2.74 × 10–4 4.00 × 10–4 9.26 × 10–4

0.997 0.991 0.976 –

2590 1270 866 374

176.0 176.1 177.5 176.8

176.6 ± 1.2

196

280 290 300

8.88 × 10–5 1.58 × 10–4 2.86 × 10–4

0.977 0.977 0.993

3900 2190 1210

181.3 181.9 182.4

181.9 ± 1.3

136

425

6.14 × 10–6

–

56 500

245.6

176

425

4.91 × 10–6

–

70 600

246.9

are not, e.g. PCB144. By comparing the rotational barriers for the two groups an average buttressing effect for the triortho PCBs could be calculated. The average ∆‡G value of the first group (PCB 84, 131, 132, 135, 174, 175, and 196) is 183.4 ± 0.7 kJ/mole (p = 0.01), compared to 177.0 ± 0.5 kJ/mole for the second group (PCB 144, 149, and 183). Thus, the buttressing effect contributes with about 6.4 kJ/ mole to the rotational energy barriers of the first group. The difference between the groups is statistically significant at the p < 0.01 level. The importance of this effect can be illustrated by comparing the racemization half-life (t1/2). PCBs which exhibit buttressing effect have average t1/2 values of 5790 s (280 °C), 2930 s (290 °C), and 1800 s (300 °C), and atropisomers that lack buttressing chlorines have t1/2 values of 1490, 824, and 433 s, at the same tem-

peratures. Thus, a 20 °C rise in temperature is needed to achieve the same rate of racemization for a PCB that has one buttressing chlorine, compared to a PCB that has none. 3.4 Thermal degradation of PCBs Tetra-ortho PCBs have a high rotational energy barrier and therefore experiments were conducted at 425 °C for 21 h. As the reaction mixture was analyzed by GC it was realized that several thermal degradation products were formed during the experiment. GC/MS analyses verified the presence of degradation products from both PCB136 and PCB176. In the reaction mixture from PCB176 we only found hexachlorobi-

223 Table 3 Comparison of calculated and experimentally determined rotational energy barriers for atropisomeric PCBs

a Ref.

[1] [18], INDO semiempirical calculation only trans conformation considered c Ref. [17], AM1 semiempirical calculation d This work b Ref.

IUPAC

Number of Substitution chlorine pattern substitutents

Rotational barriera (kJ/mole)

Rotational barrierb (kJ/mole)

Rotational barrierc (kJ/mole)

Rotational barrierd (kJ/mole)

45 84 91 95 88 132 135 136 149 131 139 144 171 175 174 176 183 196 197

4 5

105 121 105 105 105 121 121 243 105 121 105 105 121 121 121 243 105 121 243

128.3 127.0 151.1 127.3 138.3 128.5 134.9 232.1 128.4 131.1 130.3 129.6 – – – – – – –

190.2 203.3 189.7 190.9 192.0 203.9 204.3 415.3 189.9 205.1 191.5 192.7 206.1 206.4 204.8 417.1 191.9 206.8 420.2

– 183.3 – – – 184.9 184.7 245.6 176.8 183.3 – 177.5 – 182.3 183.5 246.9 176.6 181.9 –

236 236 236 236 2346 236 236 236 236 2346 2346 2346 2346 2346 2345 2346 2346 2346 2346

6

7

8

Table 4 Percentages of parent compounds and transformation products detected in the thermolysis mixtures of PCB 136, (left) and PCB 176 (right). The products are tentatively identified using GC/MS and the A1111 mixture IUPAC

Percentage of total PCB

IUPAC

Percentage of total PCB

136 96 95 84 54 53

64.8% 22.4% 6.6% 3.7% 1.5% 1.1%

176 150 136 132 149 164 145 168 144

58.9% 10.6% 8.9% 5.3% 4.5% 4.2% 3.8% 2.2% 1.6%

phenyls (PCB 132, 136, 144, 145, 149, and 150), whilst from PCB136 both pentachlorobiphenyls (PCB 84, 95, and 96) and tetrachlorobiphenyls (PCB 53 and 54) were detected, c.f. Table 4. In both cases, the total amount of dechlorination products was comparable to the amount of the parent PCB. Thus, the rotational energy barrier is of the same magnitude as the carbon-chlorine (C-Cl) bond strength. Simple dechlorination as well as chlorine migration reactions were suggested by the structure of the pyrolysis products.

References 1. Kaiser KLE (1974) Environ Pollut 7 : 93–101 2. Eliel EL, Wilen SH, Mander LN (ed) (1994) Stereochemistry of organic compounds. John Wiley & sons, New York, pp 424– 427, 1142–1151

2′ 2′3′ 2′4′ 2′5′ 2′ 2′3′4′ 2′3′5′ 2′3′6′ 2′4′5′ 2′3′ 2′4′ 2′5′ 2′3′4′ 2′3′5′ 2′3′6′ 2′3′6′ 2′4′5′ 2′3′4′5′ 2′3′4′6′

3. Oki M (1983) Recent advances in atropisomerism. Top Stereochem 14 : 1–9 4. Ballschmiter K, Zell M (1980) Fresenius Z Anal Chem 302 : 20–31 5. Schultze E, Malisch R (1983) Fresenius Z Anal Chem 319 : 545–551 6. Schurig V, Glausch A, Fluck M (1995) Tetrahedron: Asymmetry 6 : 2161–2164 7. Glausch A, Hahn J, Schurig V (1995) Chemosphere 30 : 2079– 2085 8. Hühnerfuss H, Pfaffenberger B, Gehrcke B, Karbe L, Konig WA, Landraff O (1995) Marine Pollution Bulletin 30 : 332–340 9. Safe SH (1994) Critical Reviews in Toxicology 24 : 87–149 10. Püttmann M, Mannschreck A, Oesch F, Robertson L (1989) Biochemical Pharmacology 38 : 1345–1352 11. Rodman LE, Shedlofsky SI, Mannschreck A, Püttmann M, Swim AT, Robertson LW (1991) Biochemical Pharmacology 41 : 915–922 12. Vetter W, Schurig V (1997) J Chromatogr A 774 : 143–175 13. Schwartz TR, Campbell RD, Stalling DL, Little RL, Petty JD, Hogan JW, Kaiser EM (1984) Anal Chem 56 : 1303–1308 14. Haglund P (1996) J Chromatogr A 724 : 219–228 15. Amdur I, Hammes GG (1966) Chemical kinetics: Principles and selected topics, McGraw-Hill, New York, pp 43–58 16. Krupcik J, Majekova M, Hrouzek J, Benicka E, Onuska F, Sandra P, de Zeeuw J (1995) Fresenius J Anal Chem 352 : 696–698 17. Andersson PL, Haglund P, Tysklind M (1997) Environ Sci & Pollut Res 4 : 75–81 18. Cullen JM, Kaiser KLE (1984) Proceedings of the Workshop on Quantitative Structure-Activity Relationships (QSAR) in Environmental Toxicology held at McMaster University, Hamilton, Ontario, Canada, pp 39–66 19. Tang T-H, Nowakowska M, Guillet JE, Csizmadia IG (1991) J Molecular Structure (Theochem) 232 : 133–146