1636
The Journal of Physical Chemistry, Vol. 83, No. 12,
1979
M. S.Khatkale and J. P. Devlin
Infrared Spectra of the Di- and Trianion Salts of Tetracyanoethylene. Assignments and Vibronic Effects Murlldhar S. Khatkale and J. Paul Devlin" Department of Chemistry, Oklahoma State University, Stillwafer, Oklahoma 74074 (Received October 13, 1978) fubllcafion costs assisted by the National Science Foundation
The vibrational and electronic spectra for the di- and trianion salts of TCNE, prepared in vacuo by solvent-free vapor codeposition methods, have been determined. The TCNE salts resemble the corresponding salts of TCNQ in the sense that NazTCNE is colorless while Na,TCNE is yellow. However, the simplicity of the TCNE salt spectra permits a more thorough analysis of the infrared spectra of the di- and trianion salts. This analysis strongly suggests that, as for NaTCNE, the more intense features are produced by totally symmetric (aJ modes. A vibrational assignment based on this conclusion has permitted the evaluation of empirical stretching force constants for TCNE2-that are consistent with MO predictions, while an apparent strengthening of the central double bond, in TCNE3- vs. TCNE2-, was unexpected. Perhaps the most important and certainly the most novel observation is that the absorption intensity for the activated agmodes of Na3TCNE increases by orders of magnitude when the salt is cooled to liquid N2temperatures while, simultaneously,the sample becomes opaque in the far-infrared.
Introduction As one of the strongest known T acids, the highly conjugated molecule tetracyanoethylene (TCNE) has been the subject of numerous studies. It has been the acceptor molecule of choice for investigations of weak charge transfer complexes in cases too numerous to cite. TCNE is also known to form dative complexes with strong electron donors so the radical anion, TCNE-, has itself been thoroughly investigated by ESRl and optical spectroscopic method^.^-^ In fact the alkali metal salts of TCNE, MTCNE, were the first radical anion salts of a conjugated molecule to be thoroughly studied by vibrational spectroscopic method^.^-^ As such, these salts also represent the first solid ionic system for which the vibronic activation of the totally symmetric (a,) modes was recognized and incorporated into the published analysis of the infrared spectrum (although Anderson had reported orally on a similar behavior for KTCNQ somewhat earlier).6 In addition to the charge oscillation activation of totally symmetric modes, there has been a sustained interest in TCNE and its anion as a model system for testing n-MO bonding prediction^.^^' The usefulness of the molecular system for such testing has been particularly promising, since early studies of the MTCNE salts also indicated that the salts M2TCNE and M3TCNE can be prepared by vacuum vapor cocondensation method^.^ Thus, its has been presumed that empirical data for the four molecule series, TCNE, TCNE-, TCNE2-, and TCNE3-, could be obtained. Recently, in a study originated primarily with the objective of determining whether the di- and trianion salts of TCNQ have interesting electrical properties, the colorless salt NazTCNQ and the golden yellow salt Na3TCNQ were thoroughly characterized spectroscopically, with empirical valence force constants evaluated.s Although the TCNQ salts were found to be insulators, this ability to prepare di- and trianion salts of conjugated ?r acids stimulated the present investigation of NazTCNE and Na3TCNE. After a description of the spectroscopic properties of NazTCNE and Na3TCNE a comparison is made with the corresponding TCNQ systems. 0022-3654/79/2083-1636$0 1.0010
Experimental Section The thin films of the sodium salts of the TCNE anions, in vacuo, were prepared by the vapor cocondensation method as originally described by Stanley et ala3 Commercially obtained TCNE (K & K Laboratories) was sublimed several times under vacuum prior to loading in a glass Knudsen effusion cell. The sodium was loaded under liquid heptane in a separate Knudsen cell with the hydrocarbon volatilized during system pump-down. Molecular beams of TCNE and sodium, the latter generated at elevated temperatures, were crossed a t a substrate chosen for absorption or Raman spectroscopic sampling. The mono-, di-, and trianion salts of TCNE were obtained by adjusting the temperatures of the two Knudsen ovens. Sample deposition times ranged from one to several hours for samples a few microns thick. Since nonstoichiometric salt solutions form for nonstoichiometric metal-organic deposition ratios, special deposition conditions were established for each stoichiometric salt. Because each anion has a very characteristic C=N stretching mode spectrum, the salt compositions were established by monitoring that region of the infrared spectra. The electronic spectra were recorded using a Cary-14 UV-visible spectrometer, with no compensation of the evacuated sample cell. The infrared spectra were measured with a Beckman IR-7 instrument while Raman spectra were obtained from a Jarrell-Ash 25-100 dual monochromator using argon-ion laser excitation (CR-52A) and Hamner photon-counting accessories. Both the infrared and electronic spectra were recorded a t room temperature while Raman signals were obtained by the single reflection technique from films at liquid nitrogen temperatures. Spectral slitwidths were typically 3-4 cm-l for count rates of the order of 2 X lo3 s-l. Results a n d Discussion A. Na2TCNE. The Dianion Salt. The infrared spectrum of the dianion salt is shown as curve (a) of Figure 1 and the observed infrared and Raman band frequencies are listed in Table I. Prior to measurement of these NazTCNE spectra, several samples, with compositions 0 1979 American Chemical Society
The Journal of Physical Chemistry, Vol. 83,
Infrared Spectra of the Di- and Trianion Salts of TCNE
TABLE I : Observed Vibrational Frequencies (cm-') for TCNE and Its Anions TCNE neutrala NaTCNE Na,TCNE IR
Raman
IR
Raman
IR
Raman
No. 12, 1979 1637
Na3TCNE IR
Raman
119
assign men t blU
'18
270 130 156 178
140 160 170
133
ag
}Us
165
b1U b3,
v12
250 359 416
'22
b2g
'14
bl,
'8
4 18
428 443 490 510 535b
'17
452
455
blU
'24
43 2
'4
b2u ag b3g
'21
521 530 556
5 54 579
549
546 552 558
523 530
549
530
1
v 3
558
560
ag b3U
'23
v11
595 679
'1
642 712' 746' 776' 800' 841' 904' 99 5 1014 1094 1253
630 744' 776' 800' 856' 901'
958
982
970
1154
1187 1282 1569 2235
1385 1428
1380 1430
1256 1263
1265
2188 2222
21 88 2225
2086 2146
2068
1346 1417' 1498' 1980 203 2
b1u
3 b2,
675
b1u
'IO
1020
1348 1456' 1503' 206 2
'16
b2U
'20
b3g
ag
'2
2120' 2160' 2230
'15
2247
b2u
2263 2236 a
Reference 7.
Reference 5.
' Probable impurity bands.
I
NaTCNE
----
"I ,t
,I
I ' I
0
2000y600
1400
1
I
\
I
I
I 1
l200 1000 C d ~
BOO
600
Figure 1. Infrared spectra of thin films of (a) Na,TCNE and (b) Na,TCNE. The dashed curve segments are for two different films of Na,TCNE at -90 K, but all spectra are for samples 2-3-pm thick.
ranging from 1:1 to progressively greater metal organic ratios, were prepared by codeposition. The spectra of the 1:l salt were in agreement with published observations, but the measured vibrational band frequencies are recorded in Table I and the electronic spectrum in Figure 2 for comparison with the di- and trianion values. As the metal-organic ratio was increased from sample to sample the violet color changed to blue, weakened, and finally vanished. Simultaneously the new set of infrared bands
Figure 2. The UV-visible spectra of thin films of NaTCNE, Na2TCNE, and Na,TCNE. Band intensities are relative and not comparable from salt-to-salt.
that we attribute to TCNE2- (curve (a) in Figure 1) appeared, became dominant, and finally completely replaced the monoanion salt spectrum. We are not aware of previous empirical evidence that the dianion of TCNE is colorless, but this result is not surprising. Both an acetonitrile solution of TCNQ2- and the salt Na2TCNQ are known to be colorless8~9and MO calculations predict a large energy gap for TCNQ2-.10 The electronic spectrum for NazTCNE is presented in Figure 2 where it can be noted that two bands with peak positions
1638
The Journal of Physical Chemistry, Vol. 83, No. 12, 1979
of 237 (5.23 eV) and 213 nm (5.82 eV) have been observed. The NazTCNE infrared spectrum is very similar in general appearance to that of NaTCNE. Of course the intense CEN stretching bands are strongly downshifted in frequency (by an average of 89 cm-l) but this is an anticipated effect of adding a second antibonding electron. As in the case of NaTCNE, there is only a single intense absorption (1256 cm-l) in the range 1000-2000 cm-l. The corresponding NaTCNE band (1385 cm-l) has been definitely identified as the C=C ag stretching mode band produced by a vibronically based charge o~cillation.~ Similar vibronic activation in TCNQ- salts has been confirmed and related to charge exchange between neighboring TCNQ- molecules along the TCNQ- molecular stack, an effect that is particularly strong when the stack is composed of d i m e r ~ . ~ JThe - ~ ~1256-cm-l band is well positioned for the TCNE2- C=C mode which should be further reduced in frequency, relative to the TCNEO value (1569 cm-'), by the presence of the second antibonding electron. However, caution is necessary in accepting this assignment which, in the absence of severe structural distortion, requires a charge oscillation vibronic mechanism. That such charge exchange between TCNE2-units, with exclusively doubly occupied molecular orbitals, is unlikely, but charge transfer to the polarizing Na+ cation must be considered. In any case, vibronic effects in the infrared are expected to be weak in the absence of intense low energy electronic excited state absorptions. The visible-UV spectrum for Na,TCNE shows no sign of such absorptions. However, the third most intense Na2TCNE absorption, at 546 cm-l, is close to TCNEO (535 cm-') and NaTCNE (530 cm-l) bands which have been assigned to the totally symmetric C-C stretching mode,5 and is shifted in the direction expected for that mode, since the lowest unoccupied TCNE orbital is bonding for the C-C single bonds.l* Thus, the strongest infrared bands for Na2TCNE, as for NaTCNE, seem to be produced by totally symmetric modes, possible vibronically activated. For polycrystalline samples such as our deposits, the best test of this possibility is a check for coincidence, or near coincidence, of Raman bands with the infrared bands in question. The Raman spectrum for thin films of this colorless salt (not shown) was extremely weak, but as indicated in Table I the only observed Raman features (549,1265,2068cm-') were nearly coincident with the three strongest infrared absorptions. Thus there remains little doubt that the NazTCNE infrared spectrum is dominated by absorption by totally symmetric modes. This is the basis for the assignment of Table I and is the presumption followed in the evaluation of force constants (section C). B. Na3TCNE. The Trianion Salt. The infrared spectrum of the trianion salt is presented as curve (b) in Figure 1, and the measured infrared and Raman frequencies are given in Table I. As in the preparation of Na2TCNE, the Na3TCNE thin film deposits were obtained only after samples with a range of metal-organic ratios, between 2.0 and -2.8, had been prepared and studied. As the ratio was increased beyond 2:l the deposits assumed a faint yellow color that deepened as the ratio became greater. Concurrently, a new set of C=N stretching mode infrared bands appeared, increased in intensity, and finally completely replaced the bands characteristic of the TCNE2- ion. The electronic spectrum for this sample is represented by the solid curve of Figure 2. It is clear from Figure 1that the infrared spectrum for the yellow salt, for which the NazTCNE C=N bands at 2072 and 2146 cm-l are completely replaced by bands a t
M. S.Khatkale and J. P. Devlin
1980 and 2032 cm-l, is quite different from that of Na2TCNE. There is no conclusive evidence from this study that the new salt is Na,TCNE. However, the strong downshift in the C r N band position is consistent with MO predictions of the effect of a third electron on the molecular vibrations of TCNE.14J5 There is a strong possibility that the moderately strong bands at 1417 and 1498 cm-l, as well as the weaker bands at 746, 776, and 800 cm-l, result from oxidation products in the sample rather than Na3TCNE. Even the monoanion salt NaTCNE is notoriously reactive toward oxygen. The vacuum procedure has protected both the NaTCNE and Na2TCNE salts from oxidation in the present study. However, very weak broad bands near 1450 cm-l in the NazTCNE spectrum may signal a degree of oxidation in that case, and there are no obvious TCNE3- modes for which to assign the 1417 and 1498 cm-l bands in spectrum (b) of Figure 1. Further, the three bands in the 800-cm-l range are common to both curves (a) and (b) and seem to vary in relative intensity from sample to sample. To test the importance of oxidation products, the Na,TCNE thin films have been deliberately exposed to dry oxygen. The bands near 1450 cm-l do strengthen and shift slightly but new bands emerge near the Na2TCNE C=N stretching mode absorptions, bands that are totally absent in Figure 1, curve b. Thus, if these bands in curve (b) reflect oxidation that occurs during the vapor codeposition process, the oxidation product differs from that produced during exposure of the film after deposition. The question of the assignment of the observed Na3TCNE absorption bands to particular vibrational modes reduces to the same question addressed in section A for NazTCNE and previously for the di- and trianion salts of TCNQ:* namely, are the symmetric ag modes infrared activated by a vibronically based charge oscillation. This question cannot be answered authoritatively from the data of curve b, Figure 1 and Table I alone, although the near coincidences between infrared and Raman bands in the regions near 550 and 1346 cm-l are suggestive. The conclusive evidence comes from a most unusual observation the details of which will be the subject of a later paper. It has been observed that the infrared bands at 549,1346, and to a lesser extent, 1980 cm-l are enhanced to an extraordinary degree when the infrared measurement is made at 90 K rather than room temperature. This band enhancement, roughly represented by the dashed curves in Figure 1,has exceeded two orders of magnitude for the 549- and 1346-cm-' bands while the UV-visible spectrum was insensitive to temperature change. The source of the enhancement, which is reversible but with a long relaxation period, will be analyzed elsewhere. The point is that the totally symmetric modes, for well understood reasons, are the ones that have typically shown unusual vibronic intensity effects for radical anion sah5J1J2 We believe this is another example of an intense vibronic activation of the ag modes, this time for TCNE3-. As mentioned earlier vibronic activation is presumed to require a vibrationally induced oscillation in the charge exchange between neighboring TCNE"- ions that are stacked in linear assays. Of course the arrangement of TCNE" ions is not known for either NazTCNE or Na,TCNE, but, even should they form one dimensional stacks, it is surprising that charge exchange between multiply charged anions, which must repel one another electrostatically, is sufficient to produce the infrared intensity effects cited above for Na3TCNE a t low temperatures.
The Journal of Physical Chemisfry, Vol. 83, No. 12, 1979
Infrared Spectra of the Di- and Trianion Salts of TCNE
TABLE 11: Force Constants and n-Bond Ordersa for TCNE and Its Anions TCNE" TCNEforce const n-bond force n-bond force defn bond constb order constb order
c=c
Kl
C-C CkN
K2 K,
c-c= c c-c-c C-CkN c=c, c-c
H, H5 H6 F1,2
6.59 5.32 17.04 0.43 0.96 0.38
0.76 0.35 0.87
4.54 5.68 16.35
0.58 0.46 0.79
TCNEZforce const
n-bond order
TCNE3force const
3.16 5.90 14.8
0.42 0.56 0.71
4.03 6.34 12.8
1639
0.12 -0.06 Reference 5. K is the stretching force constants in mdyn/A; H , bending force constants in mdyn A / a Reference 14. (rad)2. F, stretch-stretch interaction force constants. C-C,
'2,6
CkN
TABLE 111: Observed and Calculated In-Plane Vibrational Frequencies (cm-l) of TCNE and Its Anions TCNE' a TCNE- a TCNE2TCNE 3symmetry speciesb obsd calcd obsd calcd obsd calcd obsd calcd ag
V1
'2 v3
v4 ' 5
b 3g
'19
'20 v21
v22
b1U
v9
'IO v11 '12
b2u
'15
'16 v17 V18
2235 1569 53 5 490 130d 2247 1282 510 254 2230 958 579 165 2263 1155 443 119
2235 1569 558 489 133 2257 1287 452 24 5 2231 953 558 182 2253 1152 440 81
2200' 1392' 532 464
970
1187
2201 1392 565 466 132 2231 1308 454 246 2198 967 563 181 2226 1178 440 81
2116' 1256 546 452
982
Reference 7. ' Multiplet structure averaged. This value for K. E. Rieckhoff, and E. M. Voigt, Chem. Phys. Lett., 39, 521 (1976)). a Reference 5.
C. Force Constants and Bonding for TCNE2- and TCNE3-. The arguments presented in the previous sections lead to the vibrational assignments for TCNE2- and TCNE3- given in Table I. Since the strong infrared and Raman bands are assigned to ag type modes, there remain only a few bands to assign to nontotally symmetric vibrations. Consequently, there is an insufficiency of data for an evaluation of the complete force fields for these anions. However, the assigned bands are mostly for stretching modes so the procedure followed in the earlier study of TCNE-,5 wherein the bending and interaction force constants are transferred from TCNEO while only the stretching force constants are fit to the new data, remains feasible. The valence force field stretching force constants evaluated for TCNE2- and TCNE3- are presented in Table I1 along with the complete original set for TCNQO, as well as the values reported for TCNE-. In the absence of crystallographic data for the salts the structural parameters for TCNEO were used t h r o ~ g h o u t . ' ~The corresponding calculated frequencies are compared with the assigned observed frequencies in Table 111. The fit to the observed TCNE2- frequencies, from the adjustment of three force constants to five observed frequencies, resulted in stretching force constants that extended the trend established in TCNE-. As is obvious from the observed frequencies, the rapid decrease in the C=N and C=C bond strengths is continued, while the C-C bonds gain strength becoming nearly twice as strong as the central "double" bond. The effect of the third electron is apparently quite different, however, since the TCNE3- force constants show
2116 1261 567 434 130 2157 1310 453 246 2115 972 565 181 2150 1184 438 81 v 5 has
2006' 1346 549
995
__
2007 1346 574 458 132 2066 1310 454 246 2001 981 569 181 2054 1193 438 81
been challenged (K. H. Michaelian,
a reversal in the effect on the double bond strength. From an extended Huckel MO calculation the third electron is expected to enter an orbital that is nonbonding in the C=C region, nonbonding in the single C-C bonds, and antibonding for the C=N bonds.15 The force constants suggest that the latter influence is observable but the apparent strengthening of the C-C and C=C bonds was not anticipated. D. Na4TCNE. The Tetraanion Salt. It is apparently not possible to prepare pure Na,TCNE. However, metal-rich samples of Na3TCNE display a pronounced infrared absorption band at 1860 cm-l. These same samples show the remarkable low temperature enhancement of the 549-, 1346-, and 1980-cm-' bands of TCNE3-. The tentative interpretation is that the 1860-cm-l band indicates the formation of some TCNE4- and signals the increase in the metal-organic ratio past 3:l. It has not been established whether or not a ratio greater than three is required for the extraordinary intensity enhancement, but the evidence from six deposits in the composition range of 2.8-3.1 suggests that possibility. Conclusions This study has demonstrated that the di- and trianion salts of TCNE, like those of TCNQ,* can be prepared in essentially pure form by solvent free procedures. Several properties of these new salts, relative to the parent molecule, are analogous to those of the TCNQ salts. The color sequence (colorless, violet, colorless, yellow) for the TCNE series can be compared with the similar sequence (yellow, blue, colorless, yellow) for the TCNQ series. The strong down shifting of the C=N stretching mode fre-
1640
The Journal of Physical Chemistry, Vol. 83, No. 12, 1979
quencies is comparable in the two cases and the force constants show that the general tendency of the added electrons is to strengthen single bonds (C-C) and weaken the multiple bonds (C=N, C=C). The spectroscopic evidence also strongly suggests that the TCNE and TCNQ dianion and trianion salts are insulators or, possibly, weak semiconductors, with one dramatic exception. No resistance studies have yet been made on the metal-rich Na3TCNE salts that show a low concentration of TCNE4-, but the remarkable low temperature enhancement of the a,-mode infrared intensities is accompanied by an intense broad far-infrared absorption that may signal onset of metallic behavior.16 The absence of a similar behavior for Na3TCNQ suggests some significant structural difference between the two trianion salts. The only possibility suggested by the available data is that the TCNQ trianion, judging from the large range in C r N stretching frequencies reported for Na3TCNQ (262 cm-l),8 may be much more distorted from the original parent molecule’s planar D 2 h structure. By contrast, the range of C=N stretching frequencies for TCNE3- is only 82 cm-l. Acknowledgment. This research has been sponsored by the National Science Foundation under Grant CHE-7709653.
Arieh Warshel
References and Notes W. D. Phillips, J. C. Rowell, and S. I. Weissman, J . Chem. fhys., 33, 626 (1960). (a) M. Itoh, J . Am. Chem. Soc., 92, 886 (1970); (b) Bull. Chem. SOC.Jpn., 45, 1947 (1972). J. Stanley, D. Smith, B. Latimer, and J. P. Devlin, J . fhys. Chem., 70, 2011 (1966). J. C.Moore, D. Smith, Y. Youhne, and J. P. Devlin, J . Phys. Chem., 75, 325 (1971). J. J. Hinkel and J. P. Devlin, J . Chem. fhys., 58, 4750 (1973). G. R. Anderson and R. L. McNeeb, Paper R5, Symposium on Molecular Structure and Spectroscopy, Ohio State University, June, 1963. F. A. Miller, 0. Sala, P. Devlin, J. Overend, E. Lippert, W. Luder, H. Moser, and J. Varchim, Spectrochim. Acta, 20, 1233 (1964). M. Khatkale and J. P. Devlin, J . Chem. fhys., 70, 1851 (1979). M. R. Suchanski and R. P. Van Duyne, J . Am. Chem. Soc., 98, 250 (1976). A. Bieber and J. J. Andre, Chem. Phys., 5, 166 (1974). G. R. Anderson and J. P. Devlin, J . fhys. Chem., 79, 1100 (1975). R. Bozio, A. Girlando, and C. Pecile, Chem. fhys., 21, 257 (1977). (a) R. Bozio and C. Pecile, J. Chem. fhys., 67, 3864 (1977); (b) M. J. Rice, N. 0. Lipari, and S. Strassler, fhys. Rev. Lett., 39, 1359 (1977). B. R. Penfold and W. N. Lipscomb, Acta. Crystalkgr., 14, 589 (1961). The electron enters the seventh MO of the conjugated T system which is either of b or b symmetry (J. J. Hinkle, Ph.D. Thesis, Oklahoma State Univeri&, 19?4). In either case, the orbital is antibonding CF3-4 and nonbonding otherwise. See, for example, (a) A. Brau, P. Bruesch, J. P. Farges, W. Hinz, and D. Kuse, fhys. Status Solidi, 62, 615 (1974); (b) A. A. Bright, A. F. Gariio, and A. J. Heeger, Sold State Commun., 13, 943 (1973).
Calculations of Chemical Processes in Solutions Arieh Warshel Department of Chemistry, University of Southern California, Los Angeles, California 90007 (Received September 29, 1978; Revised Manuscript Received February 27, 1979) Publicailon costs assisted by the National Institutes of Health
A new model for the calculation of solvation energies of polar solutes is presented. The model, referred to here as the “surface constrained soft sphere dipoles” (SCSSD) model, overcomes some of the problems in the theoretical treatment of solvent effects. On one hand, it avoids the problems of dielectric continuum approaches by explicitly including solvent molecules. On the other hand, it overcomes the extreme inefficiency of the supermolecule approaches by drastically simplifying the solvent molecules while retaining their main physical features. This is done by representing the solvent molecules as soft sphere point dipoles and minimizing the solute-solvent energy with respect to the orientations and positions of these dipoles while constraining the surface dipoles to have the positions of the bulk molecules. The model allows evaluation of the microscopic energy balance of charge separation in polar liquids. It is shown that the effective dielectric is more than 20, even when the charges are separated by 3.5 A. This microscopic dielectric effect is due to the fact that the solvent system has a large number of degrees of freedom and any change in the solute-solute electrostatic interactions can be balanced by a rearrangement of the solvent molecules. The SCSSD model makes it possible to perform quantum mechanical and empirical calculations for many chemical phenomena in solutions. This is demonstrated here by evaluating solvation enthalpies of ions with different charges and radii, dissociation enthalpies of acids in aqueous solution, parallel base stacking of the cytosine dimer, and electronic excitation energies. The good agreement between the calculated and experimental results indicates that the model can be used for quantitative studies of solvent effects in chemistry.
Introduction Since most quantum mechanical and empirical calculation schemes are designed to treat isolated molecules, they cannot be applied to many chemical and biochemical processes in solution. For example, the gas phase energy of ionic reactions can amount to 100-200 kcal/mol, while the corresponding energies in solution are only 10-30 kcal/mol. The difference between these energies is due
exclusively to the medium. Omitting the medium from the calculations amounts, in some cases, to dealing with only half of the energy contributions of the complete system. One possible way to overcome part of this problem is to treat the solvent as a dielectric continuum.’ Such an approach was used by Sinanoglu and co-workers,2Claverie et and more recently by Huron and C l a ~ e r i eBev,~
0022-3654/79/2083-1640$01.00/00 1979 American Chemical Society