2474
R. A. BONHAM AND F. A. MOMANY
Vol. 67
THE STRUCTURE OF THIOPHENE I N THE GAS PHASE AS DETERMINED BY ELECTRON DIFFRACTION1 BY R. A. BONHAM AND F. A. MOMANY Department of Chemistry, Indiana University, Bloomington, Indiana Received June i7,1963 The molecular structure of thiophene was determined by the sector-microphotometer method of electron diffraction. Techniques were developed to obtain the experimental data missing as a result of the presence of the beam trap and a limited maximum scattering angle. These techniques should prove useful in the case of the missing small angle scattering intensity; however, it appears that the use of a damping factor is still necessary t o reduce the effects of integral termination errors owing to the lack of large angle scattering data. The parameters characterizing the structure of thiophene have been found to be T,(S-C) = 1.714, T ~ ( C - C = ~ )1.370, r,(C-C2) =: 1.419, T ~ ( C - H=) 1.092,Zs-c = 0.049, lc-C, (assumed) = 0.044, 1c-c2 (assumed) =00.046, IC-H = 0.070 A.; L C-S-C = 92.2' and L S-C-C = 111.4". The structure is planar t o better than 0.1 A. and the longer C-C bond is the one directly opposite the sulfur atom.
A determination of the molecular structure of thiophene by electron diffraction can serve as a valuable calibration of the experimental apparatus used. Accurate structural parameters obtained by microwave techniques exist for this moleculeza and thus provide a comparison with the present results. An earlier electron diffraction study was carried out in 1939 by use of the visual method. However, it is felt that the precision of present day methods should lead to substantially more reliable parameter values than those reported in the previous The structural data for thiophene were collected a t the same time that data on several other molecules were obtainedJ3-j and it was felt that this work would not only help to establish a check on the structural parameters obtained in these other studies, but also to serve as a test molecule for the investigation of new techniques in the analysis of diffraction data. The structure of thiopheiie pIays a particularly important role as a comparison structure for the analysis of the new molecule 1,2,5-thiadia~ole.~ It is the purpose of this work to report oil the results of the structural analysis of thiophene and the results of the new techniques that were employed in the analysis of the structural data. Experimental Procedure.--8 sample of reagent grade thiophene was purified by distillation immediately prior to taking the diffraction photographs. The purity was checked by boiling point and index of refraction measurements. These measurements indicated that the sample used was better than 98% pure. Electron diffraction photographs were obtained with an 9 sector using the sectorelectron diffraction camera in Professor L. s. Bartell's laboratory a t Iowa State University. Specimen pressures of 20 to 80 mm. with exposure times of 1 to 4 sec. were used at a temperature of 298°K. The electron beam current was 0.3 p and the accelerating voltage was 40 kv. Four research grade plates at each of two camera distances were obtained. The photographic plates were scanned with a microphotometer located in Professor L. 0. Brockway's laboratory a t the University of Michigan. The microphotometer traces of the electron diffraction photographs were read a t 0.25-mm. intervals with the aid of a Bausch and Lomb 7-power measuring magnifier. The photographic emulsion was calibrated by a procedure due to Bartell and Brockway.6 ( I ) (a) Contribution Kumber 1151 from the Chemical Laboratories of Indiana University: (b) the authors wish t o t.hank the National Science Foundation and the U.S. Atomic Energy Commission for financial support of this work. (2) (a) B. Bak, D. Christersen, L. I-Iansen-Nypaard, and J. RastrupAnderson, J. M o l . Spectry., 7, 58 (1961); (b) V. Schomaker and L. Pauling, J . Am. Chem. Soc., 61, 1769 (1839). (3) R. A. Bonham and F. A. Momany, ibid., 83, 4476 (1961). (4) T. Ukaji and R. A. Bonham, ibid.. 84, 3627 (1962). (5) T. Ukaji and R. A. Bonham, ibid., 84,3631 (1962).
The index of resolution for the short camera distance data (10.66 em.) was found t o be 1.00 and the index of resolution for the long camera distance (21.35 em.) was found to be 0.75. The low value of the index of resolution for the long camera distance is probably due to delocalized gas scattering and for this reason, the short camera distance data (20 < p < 90) was felt to be more reliable than the data for the longer camera distance (10 < p < 60). Structure Analysis.-Three different techniques were employed in the analysis of the experimental data. The first approach utilizes a scheme described elsewhere.? Bond lengths and vibrational amplitudes8 were all obtained by analysis of the data using least squares procedures. Corrections to the data for electronelectron scattering, asymmetry of vibration for bonded distances, and Fourier integral termination errors were applied. Corrections for the failure of the Born approximation were made by characterizing the radial distribution curve, corrected for asymmetry effects, with a model which included corrections for failure of the Born appr~ximation.~It was found necessary to use a damping factor of the form e-aq2 where a = 0.0009, in order to prevent the introduction of errors into the analysis due to the fact that the analytical representations of the phases az(p) are given incorrectly past p = 180. The calculated experimental radial distribution function made use of experimental intensity data, Ie(p), a t integral values of p (q = 40 sin e/A) over the range p = 12 to p = 90. Data for the region p = 0 to p = 12 were calculated using several theoretical models and corresponding radial distribution functions were obtained. The background function, B(p), was corrected by use of Karle's nonnegative area criterion.10 The radial distribution function was then analyzed by a least squares procedure and a new theoretical model was computed on the basis of the results of this analysis. This process was repeated until the results became consistent with those from the previous cycle. The final valuea for the molecular parameters obtained with this procedure were also checked by the use of the correlation method with the experimental intensity curve. I n this way, the planarity of the ring was tested and it was found that the ring atoms were planar to within better than AO.1 A*
The second procedure for analyzing the intensity data is similar to the first in all respects except for the handling of the missing experimental data in the region 0 < p < 15. The technique employed here is an analytical scheme due to Karle11 which attempts t o reconstruct the missing data in the small angle scattering region by making use of the nonnegative area criterion and imposing the added restriction of a smooth background. A radial distribution curve is obtained from the intensity curye, with electron-
s. Bartell and L. 0. Brockway, J . A p p l . Phys.. 24,656 (1953). (7) R. A. Bonham and L. S. Bartell, J . C h e n . Phya., 81, 709 (1959). ( 8 ) The term vibrational amplitude as used i n this paper corresponds closely to the term 1, defined by several authors in previous publications. See for instance: K. Kuohitsu and L. 8. Bartell, zbid., 35, 1845 (1961); R. A. Bonham and J. L. Peaoher, ibid., 38, 2319 (1963). (9) R. A. Bonham and T. Ukaji, zbid., 86, 72 (1962). Note that the second exponential term of eq. 6 should read e-0.069g03, 5 (n/2) is missing in front of eq. 13. Also in eq. 13, there is a A in front of the b i j in the X =k term, and the expression for A shoud read: A = 0.5 arc tan [(ZACsj)/lij21 4(6) L.
Aaij.
(10) I. L. Karle and J. Krtrle, ibid., 17,1052 (1949). (11) J. Karle, private communication.
Nov., 1963
MOLECULAR STRUCTURE OF THIOPHENE BY ELECTRON DIFFRACTION
electron scattering corrections and a trial background, without any added theoretical data. The positive areas of this radial distribution curve, in the region where interatomic distances should occur, are then reinverted and the resulting intensity curve in the region 0 < (7 < 15 is then used to replace the missing data in this region. A gaussian damping factor is used in this procedure so that termination errors from the missing data at q > 90 would not affect the results. The procedure may then be iterated until no important changes in the generated intensity data are observed. In general, convergence is very slow, so an additional modification t o the procedure was made. The lim M ( q ) can be shown to be approximately ~ I M & T + Z , / ~ ( Z , ~
THIOPHENE
A
3.O
2475
+
q-+O
63
Zk), where &It> is a multiplicity factor and the Z values are the atomic numbers. This quantity is independent of the structure of the molecule and depends only on a knowledge of the molecular formula. The corresponding experimental quantity M(0)expt can be shown tot be given approximately aa the first moment of the positive part of the radial distribution functionf+(r) or TmgX dr
Ln
rf*(r). The convergence rate of the data generation procedure can then be greatly accelerated by multiplyingthegenerated missing data by the ratio M ( 0 ) / r l l ( O ) e Xa~tt the end of eachcycle. I n this way, the missing data can be constructed in three to four iterations of the procedure. It was found that this technique was a helpful check on assessing the reliability of bond length determinations from minor peaks in the radial distribution curve. For instance, if the major contribution to a particular bond length occurred in the range of missing data, the use of a theoretical model will often give this peak a normal area in the radial distribution curve, while the generation of missing data technique will usually give an area that is considerably smaller 1han that predicted by theory. I n this case, care must be exercised i n the interpretation of the peak in question. A notable example of this effect occurred in the case of the 2.1-A. nonbonded C-H distance in t-butyl chloride,12 where the generation of missing data technique for the region 0 < q < 15 gave almost no peak for this distance i n the radial distribution curve. It it3 of interest to note that the earlier work of Bastiansen,13 which employed a graphical technique for obtaining the missing data and is similar in many ways to the analytical method described here, also showed a near absence of the same peak from the radial distribution curve. It should be noted, however, that such effects may be largely eliminated by obtaining experimental data into 9 = 5 or less for most molecules. It is also important to point out that this foregoing procedure fulfills an additional function in that it provides a guide for the drawing of the background function in the region of small q. If a trial background is drawn which is not compatible, a t the point where the generated M ( s ) function joins the experimental data, with the background indicated by the generated M ( s ) data, then the resulting f(r) function will have slowly varying anomalous positive and negative fluctuations. The shape of the f(r) function can thus be used as a guide to obtain the proper background function for small values of q. The analytical procedure just described was also employed to generate the missing data a t large scattering angles. However, the rate of convergence and dependence on the background made the process impractical for routine data analysis. It was also discovered that in most cases where the missing data extended over the region 0 < q < 15, it was possible to successfully reproduce the missing data. When the range of missing data was extended to p = 20, it was discovered that the procedure often failed to converge. The other technique utilized in this work was the use of a sharpened radial distribution function obtained from the experimental intensity data without the use of the missing data in the region 0 < q < 12. Distances were obtained by correlating this sharpened function with a theoretically computed function using data over the same range in q. The results of this analysis are reported in Table I. The use of an antidamping factor has the advantage that changes in the lt, and rij values of 1 0 . 0 0 2 i . make detectable differences in the correlation between the experimental and theoretical curves. Also the procedure places no reliance on a theoretical model for the missing data in the (12) F. A. WIomeny, M. 1,. Druelinger, and R. A. Bonham, J . Am. Chena. Soc., in press. (13) 0.Bastiansen end L. Smedvik, Acta Chem. Scand. 7 , 653 (1903).
Fig. 1.-Radial dbtribution curves for thiophene: the upper curve isf(r) with theoretical data added; the middle curve isj(r) with the generated data added; and the lower curve isf(r) without any data added in the missing region of the Me(p) curve. The ordinate scale refers to the absolute value of f ( r ) .
THEORETICAL GENERATED
---
.-
-- _ _ _ _ _ -_--- - 0
9.
10
15
Fig. %-Generated and theoretical M e ( q ) functions over the region of missing data 0 5 q 5 15. The generated data is the result of six iterations and the line beyond q = 15 represents cxperimental data: the ordinate scale refers to the absolute value of M ( q ) = (Io(p)/B(q) 1) corrected for electron-electron scrtttering. The ordinate scale refers to the absolute value of M(y).
-
n
I
.3I
.29
.27
.25
-2.0
30 9,
40
Fig. 3.-Long-camera-distance intensity data, l ~ ( q ) and , the background function, B(q) (smooth curve): the ordinate values refer to an arbitrary relative intensity scale.
R. A. BONHAM AND F. A. MOMANY
2476
Vol. 67
Results
t1
I .74 I
I
I
I
40
Table I shows a comparison of the final least squares analysis of the radial distribution curve for thiophene with the results obtained by BakZausing microwave tec,hniques. Also included are the observed distances found from the sharpened radial distribution function. The approximate standard deviatioiis of the least squares parameters were calculated from equations 12a and 12b of ref. 7. These estimates are admittedly somewhat crude in the case of badly overlapping peaks, but seem to furnish realistic estimates in other cases.16 To obtain a total estimate of the precision for the various measured values, it was necessary to include a scale factor error of two parts per thousand.
I
80
60
IO0
9. Fig. 4.-Short-camera-distance intensity data, lo(y), and the background function, B ( q ) (smooth curve): the ordinate values refer to an arbitrary relative intensity scale.
TABLE I STRUCTURAL PARAMETERS DERIVEDFROX RADIALDISTRIBUTION CURVESAND CALCULATED PRECISIOKS C-H S-C Cl-Ce Cz-Ca LC-9-C LS-C-C
-
.05
0,
DIFFERENCE CURVE -.
. -.. I
I
I
I
-.~-
I
1
20 40 q 60 80 100 Fig. 5.-Experimental and theoretical reduced molecular intensity curves and their difference: the ordinate scale refers to the absolute value of M ( q ) . THIOPHENE
30
I
10
20
r. ,A"
30
40
6r 0.008
1.714 1.370 1.419 = 92.2 = 111.4
I 0.070
,004 ,049 ,006 .044d ,007 , 04Bd 0.20 f 0.2'
*
rm(l)'
61
0.010 .005
... ...
1.07Sb 1.os1 1.714 1.370 1.423 92O 10' 111' 28'
1.08
6r 0.01
1.71 .Ol 1 . 3 9 ~ " ~ ..01
...
...
a See ref. 1. The shorter distance is the C-H nearest the sulfur, and the longer C-H distance is farthest from the sulfur. AsResults from the sharpened radial distribution function. sumed values.
EXPE R I MEN TAL
',=I- I I
rt(o)
1.092
I
Fig. 6.-Sharpened radial distribution function: the upper curve is experimental and the lom-er curve is theoretical. The ordinate scale refers to the absolute value of the modified f ( r ) function. region 0 < q < 12 and overlapping peaks are resolved to a greater extent. An excellent detailed discussion of the theory and use of damping and antidamping factors has been given previously by Waser and S ~ h 0 m a k e r . l ~
Figure 1 shows the comparison of radial distribution curves obtained with assumed theoretical data in the region 0 < y < 15, with the missing data generated from the experimental data according to the procedure described in the last section and the radial distribution curve without the use of any data over the range 0 < y < 15. I n Fig. 2, the M ( s ) function obtained by generation of the missing data as well as the M ( s ) function for the best theoretical model in the region 0 5 q 5 15 are shown. Kote that the differences between the t'wo curves are quite small. In Fig. 3 and 4, the intensity data, l o ( q ) ,are shown with the experimental background functions used to obtain the best radial distribution function. It should be noted that in Fig. 3 and 4 not all of the overlap region has been shown and that the short distance data extend out to q = 100 although in the actual analysis only the data out to y = 90 were used. The agreement between theory and experiment is shown in Fig. 5. Note that the difference curve indicates an average deviation between the theory and experiment that is somewhat better than 5% of the theoretical M ( s ) curve. The sharpened radial distribution functions for both the best theoretical model and the experimental data are shown in Fig. 6. The structure of thiophene is of interest chemically because of the effect a sulfur atom has on a conjugated system when it is used to replace an ethylene group. It is clear from the study of thiophene by Bak2"and also from this study that the introduction of a sulfur atom into a benzene ring by replacing an ethylene group decreases the aromaticity. This can be readily seen in the unequal C-C distances of 1.42 and 1.37 A.reported (14) J. Waser and V. Schomaker, Rev. M o d . Phys., 26, 671 (1953). (15) Compare, for instance, the errors given by a more sophisticated treatment [O. Bastiansen, L. Hedberg, and K. Hedberg, J . Chem. Phys., 27, 1811 (1957) 1 with those given by the fornlulas used here for a so111ewlmt silnilur case [L. 9 . Uarteli and R. A. Bonimm, ibid.. 31,400 (1959)l.
Nov., 1963
ELECTRONIC SPECTRA OF BIS-(CYCLOPEXTADIENYL)-METAL COMPOUKDS
in this work. These values may be compared with the C-C bond length values 1.337 and 1.483 A.16for the nonaromatic molecule butadiene and the completely aromatic benzene system where the C-C distances are probably all equal to 1.397 A.17 (16) A. Almenningen, 0. Bastiansen, and N. Traetteberg, Acta Chem. Scand., 12, 1221 (1958). (17) (a) I. 12. Karle, J . C k e n . Phys., 20, 65 (1952); (b) A. Almenningen, 0. Bastiansen, and L. Ferrtholt, KQZ. Norske Videnskab. Selskabs Skrzfter (1958).
2477
Acknowledgments.-The authors wish to thank Professor L. S. Bartell for the use of his electron diffraction unit and Professor L. 0. Brockway for the use of his microphotometer. We also wish to thank the Indiana University Research Computing Center for use of their equipment, Mrs. Connie Williams for her help in the preparation of the manuscript, and Mr. A. J. Atkinson for help in reading traces.
ELECTRONIC SPECTRA OF SOME BIS- (CYCLOPENTADIENYL)-METAL COMPOUNDS BY JAMES C. W. CHIEX The Hercules Research Center, Wilmington, Delaware Received J u n e 36, 1963 The electronic spectra of (C&)zTiC12, ( CSH5)2TiBrzJ ( CjHj)ZTiIa, and (C,€€~),VCLare described. The absorptionn are assigned to electronic transitions based on a LCAO-NO treatment. The metal-ligand bondings involve two bonding molecular orbitals per ligand; they are strongest in the chloride and weakest in the iodide. Two new forbidden bands have been found in the spectrum of (CiH&Fe. The band positions agree with Yamazaki’s SCF calculations
Introduction The electronic structure of bis-(cyclopentadieny1)iron has been a subject of continuing interest. Earlier treatmeiits1,2showed that the metal and the ligands are bonded by dr-pa bonds. Linnett3 theorized that there are three equivalent bonding orbitals between iron and eaclh cyclopentadienyl ring. The molecular orbital energy levels were calculated approximately by Dunitz and 01-gel.~ Recently results of more sophisticated calculations have Csing slightly different sets of basic functions but similar methods of calculation, these authors arrived at molecular orbital energy levels which differ in both the absolute values and the order of stabilities. The purpose of this communication is to present the electronic spectra of (C,H&TiX,, where X = C1, Br, I, and t o report weak new absorption bands in the spectrum of (C6H&Fe. The absorption bands have been assigned to appropriate electronic transitions. The relative merits of the (C6H6)*Femolecular orbital calculations are briefly discussed. Experimental The compounds used in this work were prepared by Dr. W. P. Long of these Laboratories. Bis-(cyclopentadieny1)-titaniumand -vanadium dihalides were prepared by the method of Wilkinson and Birmingham .* (CgM&Fewas prepared according to the procedure of Kealy and P a u ~ o n . ~ Analyses of all compounds were correct with the exception of (C5H&Ti12, which appeared to be contaminated with (C6H&Ti(OH)I. The electronic spectra were obtained in the absence of oxygen. A Cary Model 14 spectrophotometer was used in this work. (1) W. Moffitt, J . Am. Chem. Soe., 76, 3386 (1954). (2) A. D. Liehr and C . J. Ballhausen. Acta Chem. Scand., 11, 207 (19.57). (3) J. W-.Linnett, Trans. Faraday Soc., 52, 904 (1956). (4) J. D. Dunitz and L. E. Orgel, J . Chem. Phyu., 28, 964 (3865). (5) M. Pamaaaki, ibid., 24, 1260 (1956). (6) E. XI. Shustorovich and M. E. Dyatkina. Doh?,. A k a d . A’auk SSSR, 128,1234 (1959);Z h . Strukt. Khin., 3,345 (1962). (7) J. P. Dah1 and C. J Ballhausen, M a t . Fye. M e d d . Dan. Via. Selsk., 33,
I(1961).
(8) G . Wilkirisori and J . M. Birminghain, J . Ani. Chem. Suc., 76, 4 2 8 1 [ 1R54).
(9) T. J. Kealy arid P. L. Pauson, Xature. 168, 1039 (1951).
Results and Discussion Electronic Spectra of Bis- (cyclopentadieny1)-titanium and -vanadium Dihalides-In their study of (C&&Fe, Kaplan, Rester, and Katz’O found absorption a t 32.6 kK. in carbon tetrachloride. This absorption is absent in hexane, ethanol, and methanol. Brand and Snedden11 showed that the 32.6-kK. absorption is obtained in all halogenated solvents, and they attributed this absorption to the dissociative charge-transfer process hu
(C6H&Fe
+ CCh I_ (C&&F”+
-I- C1-
+ CC13 (1)
The possibility of dissociative charge-transfer absorptioil of (CsH6)2TiC12 in methylene chloridewasinvestigated by determining it,s spectra in methylene chloride, tetrahydrofuran, and diethyl ether (Fig. 1). The first two werefound to be identical. Apparently, dissociative chargetransfer processes of the type mentioned are unimportant for (C6H&TiC12. A slight decrease in short wave length absorption intensities was observed in diethyl ether. I n addition, there exists a long wave length tail which gives a pinkish tint to the diethyl ether solution. The nature of the long wave length tail absorption in diethyl ether is not understood. Another point which concerned us is the possibility of dimerization. Long12 furnished spectroscopic evidence for complex formation between (C5H&TiC1, and aluminum chloride and alkylaluminum chloride. ChienI3 also postulated these complexes as the active species in the low pressure polymerization of ethylene. We have determined the molecular weight of (C6H6)2TiCI2in ethylene dichloride and found it to be 250, 252. Furthermore, the methylene chloride solutions of (C6H6),TiC12follow Beer’s law from to M. (10) L. Kaplan,
W-.L. Kester, and J. J. Kate, J . Am. C h e n . Soc., 74, 5531
( 1952).
i l l ) J. C. D. Brandand W. Snedden, Trans. Faraday Soc., 53,894 (19571. (12) R-. P. Long, .1.A m . Chem. Soc., 81, 5312 (1959). (13) J. C. W. Chien, i b d . 8 1 , 86 (1959).