J. CHEM. SOC. FARADAY TRANS., 1994, 90(5), 697-701 Internal Rotation in Auramine 0 Pennathur Gautam and Anthony Harriman" Center for Fast Kinetics Research, The University of Texas at Austin, Austin, TX 78712,USA The fluorescence quantum yield and excited singlet state lifetime measured for Auramine 0 in different alcohols increase with increasing microviscosity of the solvent. The results can be explained quantitatively in terms of current theoretical models in which internal rotation of the N,N-dimethyaniline groups is an activationless process controlled entirely by viscous flow. Auramine 0 binds to deoxyribonucleic acid (DNA) and horse liver alcohol dehydrogenase (HLADH) in neutral aqueous solution. Binding to DNA is accompanied by modest increases in fluorescence yield and lifetime of the dye whereas binding to HLADH facilitates a dramatic increase in fluorescence.There appear to be two binding sites for Auramine 0 on the protein, and in one particular site the conformation of the dye is essentially frozen in a form unfavourable for rotational diffusion. Internal rotation in small molecules can provide an impor- tant route for non-radiative deactivation of the first excited singlet state and, in certain cases, may result in formation of a twisted intramolecular charge-transfer state (TICT). Accord- ing to the size of the rotor and the amplitude of its motion, the rate of rotation may depend on frictional forces with adjacent solvent molecules. However, if excited state deacti- vation involves a substantial change in dipole moment it may be necessary to distinguish between 'polarity' and 'viscosity' effects induced by changes in solvent and/or temperat~re.'-~ On considering only viscosity effects it is common practice to express the rate of rotation in terms of Kramers' the~ry.~ Although few systems give an exact correspondence with the original theory, when subjected to close scrutiny, there have been numerous modifications and extensions which have sig- nificantly improved the agreement between experiment and theory.'-'' One particular improvement has involved the use of microviscosity in place of bulk or shear viscosity as a means of expressing specific solute-solvent interactions.' '-I4 We now show that the concept of microviscosity allows a good description of internal rotation in Auramine 0.77 N+ " ''1'''' CH3I CH3 -CH, CH 3 Auramine 0, a cationic dye, is used as a stain for DNA and HLADH.I5 The dye is essentially non-fluorescent in water but fluoresces upon binding to a biological substrate. Oster and NishijimaI6 observed that the fluorescence yield of Auramine 0 in glycerol was markedly dependent on tem- perature. This was explained in terms of internal rotation of the N,N-dimethylaniline groups providing an effective means for non-radiative decay to the ground state. It was further claimed that fluorescence quenching occurred if the group rotated by more than 2" during the period of excitation.I6 In fact, the N,N-dimethylaniline group is an important constitu- ent of many TICT molecules" and information regarding its rotational diffusion is of considerable current interest.We have, therefore, re-examined the photophysics of Auramine 0 in a series of normal alcohols at various temperatures. Our results indicate that rotation of the N,N-dimethylaniline group is an activationless process which depends on the vis- cosity of the surrounding medium without an apparent polarity effect. Binding the dye to a biological substrate can cause a substantial decrease in the rate of internal rotation, in certain cases the effect is far more pronounced than that which can be achieved by a viscous solvent. Experimental Auramine 0 [4,4'-(imidocarbonyl)bis(N,N-dimethylaniline) monohydrochloride] was obtained from Aldrich Chemicals and purified by repeated recrystallization from warm 0.02 mol dm-3 NaCl solution before being dried over P205.I5 The purified material gave a linear Beer's law plot in water (pH 7) containing 0.5 mol dm-3 KCl at concentrations below 5 x low3mol dm-3, from which the molar absorption coefi- cient at 431 nm was determined to be 43200 dm3 mol-' cm-'.'H NMR (D,O): S = 2.89 (s, 6 H); 6.59 (d, 2 H, J = 9.1 Hz); 7.26 (d, 2 H,J = 9.0 Hz). Elemental analysis: Calculated for C,,H,,N,Cl-H20: C = 63.44; H = 7.52; N = 13.05%. Found: C = 63.32; H = 7.60; N = 13.01%. Sol- vents, obtained from Aldrich Chemicals, were of the highest available purity and were used as received. DNA (calf thymus) and single-stranded DNA (calf thymus) were obtained from Sigma Chemicals and were used as received; concentrations were assessed by absorption spectroscopy." HLADH was obtained from Sigma Chemicals and was dia- lysed against 0.1 mol dm-3 phosphate buffer (pH 7.4), its concentration was determined by absorption spectroscopy using a molar absorption coefficient at 280 nm at 35 400 dm3 rno1-l cm-'.l5 Absorption spectra were recorded with a Hitachi U3210 spectrophotometer and fluorescence spectra were recorded with a fully corrected Perkin-Elmer LS5 spectrofluorimeter. For fluorescence studies the solution was adjusted to possess an absorbance of ca. 0.05 at the excitation wavelength of 370 nm. Solutions were purged thoroughly with nitrogen and the sample cell was thermostatted with a constant-flow circulat- ing water bath.Temperatures were measured with a thermo- couple in direct contact with the solution and were accurate to within & 0.2"C. Fluorescence lifetimes were measured by time-correlated, single-photon counting using a frequency-doubled, mode- locked, synchronously pumped, cavity dumped Styryl-9 dye laser. Fluorescence was isolated from scattered laser light using a high radiance monochromator. The excitation wave- length was 417 nm and fluorescence was detected at 540 nm via a pinhole attachment and a Hamamatsu R2809U micro-channel plate phototube. Approximately 40 O00 counts were collected in the peak channel and, for each measurement, four different time bases were used.After deconvolution of the instrument response function, the ultimate time resolution of this instrument was cu. 20 ps. Titrations of Auramine 0 with DNA or HLADH were made in aqueous solution containing 5 x mol dm-3 sodium sulfate and 1 x mol dm-3 phosphate pH 7 buffer or 0.1 mol dm-3 phosphate pH 7.4 buffer, respectively. Each titration was made by adding successive aliquots of dye to fixed concentrations of substrate and by adding increasing amounts of substrate to a fixed concentration of dye. The course of the reaction was followed by absorption and fluo- rescence spectroscopy. Time-resolved fluorescence measure- ments were made at various ratios of dye to substrate. Values for the shear viscosity (q)of the solvent at a particu-lar temperature were extrapolated from data available in the literature.’ Microscoviscosity (qp) values were calculated by the semi-empirical method of Spernol and Wirtz,” as detailed by Sun and Saltiel.14 Thus, microviscosity is defined as =fq (1) where the microviscosity factorfis expres sed as: f= (0.16 + 0.4r/r,)(0.9 + 0.4T,, -0.25T,) (2) Here, r and rL are the solute and solvent molecular radii, respectively, and the terms rLand T, refer, respectively, to the reduced solvent and solute temperatures.T -T,KL = -(3)Tb -Tm In this expression, T is the experimental temperature, and Tm are the boiling and melting points of the solvent for TL and solute for T,. Molecular radii for the various alcohols were estimated from molar volumes assuming a space-filling factor of 0.74.All calculations were made on the basis of the solute being N,N-dimethylaniline, for which the molar volume is equal to 126.4 A3 according to X-ray crystallo- graphic determination. Further details on the calculation and multifarious applications of microviscosity are provided by Sun and Saltiel.14 Results and Discussion Fluorescence in Alcohols Absorption and fluorescence spectra of Auramine 0 in decan-1-01 are shown in Fig. 1. There is a relatively large Stokes’ shift (2800 an-’)and poor mirror symmetry between the lowest-energy absorption band and the fluorescence band. This suggests to us that there might be a significant change in geometry upon promotion to the first excited singlet state.Even so, the corrected excitation spectrum gave a good match to the absorption spectrum across the entire visible region, including the absorption peak at 370 nm. The fluorescence quantum yield (a,) in decan-1-01, measured rela- tive to Rhodamine 101 in ethanol” and corrected for changes in refractive index,” was found to be 0.0045 & 0.0005 while the fluorescence lifetime (2,) was mea- sured to be 92 3 ps. The derived radiative lifetime (zo = z$QDf = 20.4 ns) was in agreement with that calculated from the Strickler-Berg expression (zo = 18.6 n~).’~ The absorption and fluorescence spectra of Auramine 0 did not change with solvent and, in particular, there was no indication of a polarity effect. It was observed, however, that (D, was markedly dependent on the nature of the solvent for a series of alcohols at 23 “C,after correction for changes in the J. CHEM.SOC. FARADAY TRANS., 1994, VOL. 90 A/nm Fig. 1 Absorption and fluorescence spectra recorded for Auramine 0 in decan-1-01 refractive index of the solvent (Table 1). For solvents in which fluorescence could be resolved from the instrumental response function, there was good agreement between z, and QD, (Table 1) and the fluorescence decay profiles could be analysed satisfactorily in terms of a single exponential com- ponent. Measurements made in pentan-1-01 and nonan-1-01 showed that Qf increased with decreasing temperature (Table 1). Additional studies carried out under isoviscosity condi- tions using different alcohols at varying temperatures, selec- ted to give a constant viscosity of 5 cP,~ indicated that both a, and z, were essentially independent of temperature over the range investigated (-41 < T/”C< 26).In general, both QD, and z, increase with increasing solvent viscosity (q), as observed first by Oster and Nishijima.16 However, as shown in Fig. 2, a plot of (l/QDf)us. (T/q)deviates from linearity. A linear plot might be expected if the rate of internal conversion is controlled by rotational diffusion of one of the N,N-dimethyaniline groups. Thus, the reciprocal of the fluorescence quantum yield can be related to the solvent Table 1 Effect of (micro)viscosity on the fluorescence quantum yield and lifetime of Auramine 0 in alcohols solvent T/K tl/cP VP/cP l/qa 7sa /ps CH30H 296.0 0.58 0.48 600 <20 C,H,OH 296.0 1.16 0.87 330 <20 C3H,0H 296.0 2.09 1.43 200 24 C,H90H 296.0 2.85 1.78 168 27 C5H110H 284.9 5.61 3.27 97 52 289.8 4.74 2.78 113 38 294.1 4.09 2.40 130 32 296.0 4.00 2.34 132 35 298.5 3.59 2.11 143 32 303.9 3.04 1.79 166 30 308.6 2.62 1.55 191 25 313.5 2.28 1.35 215 22 322.6 1.78 1.06 276 <20 296.0 4.93 2.67 108 45 296.0 6.60 3.38 98 50 296.0 8.17 3.89 78 64 282.5 16.24 7.41 36 136 288.2 13.97 6.41 42 115 292.4 11.77 5.41 51 95 296.0 10.40 4.73 62 83 297.6 9.70 4.47 65 78 303.0 8.24 3.80 72 67 308.6 6.90 3.19 86 58 312.5 5.94 2.75 100 49 320.5 4.75 2.21 123 40 296.0 13.47 5.96 53 92 a f5%.J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 0,' o II 1I I 1I I I I m2oo3oo4oo5oomm (TI'1)IK cp-' Fig. 2 1/Qf us. (0)T/q or (A)T/q, viscosity via eqn. (4): 1/@r= 1 + (Tokc) + Ca7o(T/v)l (4) Here, kisc refers to the rate constant for intersystem crossing to the triplet manifold and a is a proportionality constant. The form of Fig. 2 suggests that shear viscosity does not give an accurate measure of frictional forces with adjacent solvent molecules for this system, especially at low viscosity. Indeed, replacing shear viscosity with microviscosity (q,) resulted in a substantial improvement in the correlation between the experimental data and eqn.(4) (Fig. 2). From the intercept of this latter plot, we estimate that kisc has a value of (4 f1) x lo8 s-' and the slope (uT,) is numerically equal to 4.03 CPK-I. The intercept can be employed to estimate that in the absence of internal rotation @, would have a value of cu. 0.12. Assuming that internal conversion from the first excited singlet state occurs exclusively by way of internal rotation, the rate constant (k,J can be expressed in terms of the Stokes-Einstein relati~nship.~~ Here, k, is. the Boltzmann constant and V is the hydrody-namic volume of the rotating group. Combining eqn. (4) and (5) indicates that the proportionality constant a can be equated to kB/V and, therefore, the hydrodynamic volume of the rotor has a value of 166 A3.This derived value for the hydrodynamic volume seems quite reasonable since, from X-ray crystallographic, data, the molar volume of N,N-dimethylaniline is 126.4 A3. The photophysical properties of Auramine 0,therefore, appear to be controlled by rotational diffusion of one of the N,N-dimethylaniline groups. The diffu-sion coefficient of the rotor (D = RT/6mq,, where T is the radius of the rotor) has a value of 1 x lo-'' m2 s-' in decan-1-01at 296 K, where the microviscosity is 6 cP. On the basis of an isomicroviscosity plot made at )I, = 2.8 CP (Fig. 3), there is a small negative activation energy for internal rotation (EA= -3.8 kJ mol-I). The results obtained in different solvents at fixed temperature can be expressed in terms of Kramers' theory assuming the Smolukowski limit and using microviscosity in place of shear vi~cosity'~[Fig.Wl. krot = A/qi (6) Least-squares analysis of the experimental data gave ct = (0.97 f0.02), which confirms that the rate of rotation is controlled almost exclusively by viscous flow. Results obtained in the same solvent at different temperatures can be expressed in terms of the 'medium-enhanced barrier model' introduced by Fischer and co-workers2' [Fig. 4(b)].Again, 1 1 I I I23.70 ! I r I 3.10 320 3.3 3.40 3.50 103 KIT Fig. 3 Isomicroviscosity(2.8 cP) plot of log k,,, us. 1/T linear plots were observed with slopes close to unity for both pentan-1-01 and nonan-1-01.Individual k,,, values used for these plots were derived from the measured 7s values (k,,, x l/rJ where possible and, in all other cases, were calculated from the fluorescence quantum yields (k,,, cc l&). In the medium-enhanced barrier model, k,,, is related to microviscosity according to In k,,, = In k, + ASJR + b In B,, -b In q,, (7) where k, is the limiting rate constant for rotation at zero microviscosity, AS, is the medium-imposed entropy change, Y ACL -c 24.5 24.0 23.5 23.0 ! -m I 1 I I-3 om 1 Ias0 I I too ILsa UK] In '1, 25.0 I I 1 I 1 24.5 24.0 Y 0 aL -c 23.5 22.0 am 0.50 100 150 200 250 In ftp Fig. 4 (a)Kramers' plot for the experimental data collected in dif-ferent alcohols at 296 K.(b) Medium-enhanced barrier model plot for the experimental data collected in (m)pentan-1-01 and (A)nonan-1-01 at different temperatures. and B,, is the microviscosity at 0 K. The constant b refers specifically to the fraction of the activation energy of viscous flow that gives the enthalpy increment by which the medium augments the intrinsic barrier for solute rearrangement. l4 From Fig. qb), b = 1.00 & 0.05 for both pentan-1-01 and nonan-1-01. This finding is consistent with an activationless structural change that is controlled exclusively by viscous flow. The slight change in intercept between the two solvents can be related to changes in the entropy term due to the difference in molar volume of the solvent.In a given solvent, k,,, can also be expressed in terms of transition state theory In (k,,JT) = ln(ick,/h)exp(ASS/R) -(AHS/RT) (8) where AH' is the enthalpy of activation change, AS$ the entropy of activation and IC an adiabaticity factor. From the corresponding plots for results obtained in pentan-1-01 and nonan-1-01 (Fig. 5), AH' values of 18.4 and 22.1 kJ mol-', respectively, were found for pentan- l-ol and nonan- l-ol. Indi- vidual AH' values can be expressed in the form AH' = AH: + AHv (9) where AH: refers to the activation enthalpy change for twist- ing of the molecule and AHv refers to the medium-imposed enthalpy change. For this system, AHv is equal to the activa- tion energy for viscous flow of the solvent, as obtained from the Andrade expre~sion,'~ since b = 1, and has values of 23.0 and 27.2 kJ mol- ',respectively, for pentan-1-01 and nonan-1- 01.' Consequently, the activation enthalpy change for twist- ing of the molecule has a value of ca.-5 kJ mol-'. This derived value is in excellent accord with the activation energy obtained from the isomicroviscosity plot (Fig. 3). Fluorescence of Auramine 0 Bound to Biological Substrates Fluorescence from Auramine 0 in water was difficult to detect and time-resolved fluorescence studies indicated that z, was less than 20 ps. On addition of DNA to a neutral aqueous solution of Auramine 0 containing Na2S04 (5 x lop3 mol drnp3) the fluorescence intensity increased markedly and the peak maximum was shifted from 485 to 533 nm (Fig.6). The absorption peak maximum was red- shifted by only ca. 4 nm, such that upon binding to the poly- nucleotide the Stokes' shift increases to 4225 cm-'. The fluorescence spectral changes gave a good fit to the Scatchard with a binding constant of (1.4 & 0.2) x lo6 dm3 mol-' and a saturation number 0.9 +_ 0.1 molecules of dye per phosphate group. Similar properties were observed for binding to single-stranded DNA. It seems likely, therefore, E92, II I I I I I 4 I J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 m 0.80 0 (D2 0.60-s 0.40 --0.20 0.m 300350400450500550m650 Llnm Fig. 6 Absorption and fluorescence spectra recorded for Auramine 0 in the presence of excess DNA (phosphate :dye ratio = 60) that binding to the polynucleotide involves primarily electro- static interactions with the phosphate groups without inter- calation between base-pairs.At a phosphate: dye ratio of 60, under conditions where ca. 99% of the dye is bound to DNA, the fluorescence quantum yield was determined to be 0.0014 +_ 0.0005. This corresponds to a 15-fold increase in the fluorescence quantum yield upon binding to DNA relative to aqueous solution. Time-resolved fluorescence studies indicated that the decay profile could be analysed satisfactorily in terms of a single exponential component of lifetime 60 & 4 ps. Thus, binding to the polynucleotide surface hinders rotation about the N,N-dimethyaniline groups and thereby decreases the rate of internal conversion.The bound molecule still under- goes rapid internal conversion, however, and it is clear that such surface binding does not present a major barrier to internal rotation. Previous work has established that Auramine 0 binds selectively to HLADH in neutral aqueous sol~tion.'~ Addi-tion of HLADH to an aqueous solution of the dye containing phosphate buffer (0.1 mol dm-3, pH 7.4) had little effect on the absorption spectrum but enhanced the fluorescence inten- sity (Fig. 7). The fluorescence peak for bound dye was located at 523 nm. Analysis of the titration data according to the Scatchard gave an association constant of (5.6 _+ 0.5) x lo6 dm3 mol- ' and a saturation number corre- sponding to a maximum loading of 1.5 dye molecules per 16.8 ' "40 4h & &I & 6;o I I 1 1 1 I 3ba 3.b 3h 3.30 3.k 3io 3L L/nm Fig.7 Fluorescence spectra recorded for (a) Auramine 0 in waterlo3 KIT containing 0.1 mol dm-3 phosphate pH 7.4 buffer and (b) after addi-Fig. 5 Plot to the transition state theory for experimental data col- tion of HLADH (protein :dye molar ratio = 60).The spectrum lected in (a)pentan-1-01 and (m) nonan-1-01 at different tem- recorded in water is shown at x 20 amplification and the sharp peak peratures centred at 427 nm is a Raman excitation band. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 molecule of protein. Similar parameters were derived by fitting the titration data to the Hill expression2’ and earlier work by Conrad et a1.” also observed a non-integer satura- tion number (n = 1.6-1.7).Since HLADH exists in solution in the form of a dimer composed of two identical subunits28 it appears that three molecules of Auramine 0 bind to a single dimer molecule. The binding sites appear to exhibit comparable association constants without displaying c~operativity~~and it should be noted that previous studies showed that binding of dye does not inhibit the normal enzy- matic functions of the protein.’ Time-resolved fluorescence studies made over a range of dye : protein molar ratios required analysis in terms of three exponential components. A very short-lived component (7 z 20 ps) was dominant at high dye : protein ratios but its fractional contribution to the total signal decreased progres- sively with increasing moIe fraction of protein.This com- ponent is attributed to Auramine 0 free in solution; its apparent lifetime of 20 ps is, in fact, broadened by the instru- ment response. The other two components had lifetimes of 400 & 50 ps and 1.0 & 0.1 ns and occurred at a fixed ratio of 2 : 1, with the shorter-lived component predominating, throughout the titration. Both lifetimes are attributed to dye molecules bound to the protein. The shorter lifetime is assign- ed to dye bound in a relatively flexible site, and the longer lifetime is assigned to dye molecules localized at a tight binding site where rotation is severely hindered. It is clear that binding of Auramine 0 to HLADH causes a dramatic reduction in the rate of internal conversion and, on the assumption that this process involves rotational diffusion, the protein is able to lock the conformation in a geometry unfavourable for internal conversion.There are two binding sites on the protein, distinguishable by the fluorescence life- time of bound Auramine 0, but even dye molecules located at the less tightly bound site are in a rigid environment. Thus, the fluorescence lifetime of Auramine 0 localized at such sites is 400 f50 ps while the measured fluorescence lifetime in glycerol at 22°C (q = 12 P) was 300 f20 ps. Dye molecules located at the more tightly bound site are effectively frozen since the observed fluorescence lifetime is 1.0 k0.1 ns. Our work does not permit identification of either binding site but it indicates that at least one of the sites provides a cavity whose dimensions almost exactly match those of Auramine 0.Concluding Remarks This work has considered two methods for restricting inter- nal rotation in Auramine 0; namely, increasing solvent vis- cosity and binding to a biological macromolecule. The effects of solvent viscosity are readily interpreted in terms of current theoretical models since the rate of internal rotation is con- trolled uniquely by viscous flow. The effects of binding to a macromolecule are both more pronounced and more difficult to interpret. Binding to the surface of DNA via electrostatic forces can be related to a decrease in the rotational diffusion coefficient and the observed fluorescence spectral shifts imply structural distortion in the excited singlet state manifold. Binding to HLADH causes a dramatic reduction in the rate 70 1 of internal rotation which is equivalent to freezing the mol-ecule into a set conformation. The rotational diffusion coeffi- cients for bound forms of the dye appear to be extremely low.However, according to the medium-enhanced barrier model [eqn. (7)] and transition state theory [eqn. (S)], a low k,,, can also be caused by a large negative entropy change. Thus, a dye molecule bound to HLADH may be considered to reside in a conformation unfavourable for rapid internal rotation. Twisting to a more favourable conformation, which has to occur within the excited state lifetime, may involve a signifi-cant decrease in entropy.This work was supported by the National Institutes of Health (CA53619). The CFKR is supported by The Uni- versity of Texas at Austin. References 1 J. Hicks, M. T. Vandersall, E. V. Sitzmann and K. B. Eisenthal, Chem. Phys. Lett., 1987, 135, 413. 2 J. D. Simon and S-G. Su J. Phys. Chem., 1990,94,3656. 3 A. Harriman, J. Photochem. Photobiol. A: Chem., 1992,65,79. 4 J. E. I. Korppi-Tommola, A. Hakkarainen, T. Hukka and J. Subbi, J. Phys. Chem., 1991,95,8482. 5 R. F. Grote and J. T. Hynes, J. Chem. Phys., 1980,73,2715. 6 J. L. Skinner and P. G. Wolynes, J. Chem. Phys., 1978,69,2143. 7 B. Bagchi and D. W. Oxtoby, J. Chem. Phys., 1983,78,2735. 8 S. P. Velsko, D. H. Waldeck and G.R. Fleming, J. Chem. Phys., 1983, 78, 249. 9 D. Ben-Amotz and J. M. Drake, J. Chem. Phys., 1988,89,1019. 10 G. Ponterini and F. Momicchioli, Chem. Phys., 1991,151, 111. 11 J. L. Dote, D. Kivelson and R. N. Schwartz, J. Phys. Chem., 1981,85,2169. 12 A. Gierer and K. Wirtz, 2. Naturforsch, Teil A, 1953,8, 532. 13 Y-P. Sun, J. Saltiel, N. S. Park, E. A. Hoburg, and D. H. Waldeck, J. Phys. Chem., 1991,%, 10336. 14 Y-P.Sun and J. Saltiel, J. Phys. Chem., 1989,93, 8310. 15 R. H. Conrad, J. R. Heitz and L. Brand, Biochemistry, 1970, 9, 1540. 16 G. Oster and Y. Nishijima, J. Am. Chem. SOC.,1956,78, 1581. 17 E. W. Lippert, W. Rettig, V. Bonacic-Koutecky, F. Heisel and J. A. Miehe, Adv. Chem. Phys., 1987,68, 1. 18 S. J. Atherton and A. Harriman, J. Am. Chem. Soc., 1993, 115, 1816. 19 Landolt-Bornstein, Zahlenwerte und Functionen., Secheste Auflage, Band 11, Teil 5, Springer Verlag, Berlin, 1969, pp. 208-230. 20 A. Spernol and K. Wirtz, Naturforsch., Teil A, 1953,8,522. 21 S. R. Meech and D. Phillips, J. Photochem, 1983,23, 193. 22 N. Nakashima, S. R. Meech, A. R. Auty, A. C. Jones and D. Phillips, J. Photochem., 1985,30,207. 23 S. J. Strickler and R. A. Berg, J. Chem. Phys., 1962,37,814. 24 E. Akesson, A. Hakkarainen, E. Laitinen, V. Helenius, T. Gillbro, J. Korppi-Tommola and V. Sundstrom, J. Chern.Phys., 1991,95,6508. 25 D. Gegiou, K. A. Muszkat and E. Fischer, J. Am. Chem. SOC., 1968,90,12. 26 I. Klotz, Science, 1982,217,1247. 27 T. G. Traylor, M. J. Mitchell, J. P. Ciccone and S. Melson, J. Am. Chem. SOC., 1982,104,4986. 28 C-I. Branden, H. Ekland, B. Nordstrom, Y. Boiwe, G. Soder- lund, E. Zeppezauer, I. Ohlsson and A. Akeson, Proc. Natl. Acad. Sci. USA, 1973,70,2439. Paper 3/06039J; Received 11th October, 1993