1394
J. Phys. Chem. 1988, 92, 1394-1396
simply inverting the experimental pressure derivatives, because the shifts computed from the hard-sphere model depend very strongly on this diameter.* From our iodine a w / d P we find in this way a value of us = 5.66 A, which agrees within about 1% with the diameter for methylcyclohexane derived by Jonas and co-workers from transport properties.” The Buckingham criterion may be employed in the same way to test the form of solute-solvent coupling for gas-to-liquid frequency shifts observed at 1 bar. Data are available for the fundamental and first two overtones of HC1 in solutions of CCI4 and two fluorocarbon solvent^;'^ the shifts are large and negative (wSln < cogas), indicating dominantly attractive interaction, but we find that the corresponding R , ratios are not far from unity (varying from 0.9 to 1.03). The much more extensive overtone data of Kiefer and Bernstein’O for I2 in several solvents also involve negative shifts. W e find that the R , for the first two overtones are quite close to unity (average for eight solvents is 0.995, range 0.96-1.03). This indicates that, for the fundamental and low overtones, the usual form for V,,, should be adequate for treating gas-to-liquid shifts as well as pressure-induced shifts. However, as shown in Figure 3, for higher overtones the R , ratios deviate appreciably from unity, indicating that higher powers of n enter
Cyclopentane and Cyclohexane n-Hexane and n-Pentane
0.0
-
0
2
4
6
8 1 0 1 2 1 4
Overtone Number (n-1)
Figure 3. Comparison of overtone shift ratios (Aw = wSo1, - wgas) for iodine solutions (-0.01 M) in several solvents at 1 bar, from data of ref 10. When two solvents are listed, the lower curve pertains to the first member of the pair and the upper curve to the second.
the shifts. Thus, the attractive interactions apparent in the gas-to-liquid shifts involve parts of Vinl that contain significant contributions from powers of above quadratic or require higher order perturbation theory.
(17) Perrot, M.; Turrell, G.;Huong, P. V. J . Mol. Spectrosc. 1970, 34, 47.
High-Resolution Core Level Photoelectron Spectra of Solid TCNQ: Determination of Molecular Orbltal Spatial Distribution from Localized Shake-up Features John M. Lindquistt and John C. Hemminger* Department of Chemistry and Institute for Surface and Interface Science, University of California, Irvine. California 9271 7 (Received: January 6. 1988)
High energy resolution X-ray photoelectron spectra of the N 1s and C 1s core levels of solid tetracyanoquinodimethane(TCNQ) are presented. The N 1s binding energy is determined to be 399.7 eV. A single shake-up satellite is detected 2.6 eV to higher binding energy in the N 1s spectrum. The C 1s spectrum shows two core features and a single shake-up satellite. The C 1s shake-up feature is associated with core ionization of the carbons external to the ring and allows determination of the spatial extent of the highest occupied and lowest unoccupied molecular orbitals of TCNQ. This data indicates that the orbitals involved in charge-transfer compound formation are located near the cyano ends of the molecule.
Introduction Studies of the molecule tetracyanoquinodimethane (TCNQ) have been carried out for more than a quater of a century since the report of its preparation by Acker and Hertler.’ A large number of these studies were aimed at the development of an understanding of the exceptional strength of TCNQ as an electron acceptor and its ability to form stable charge-transfer complexes with both organic and metallic donors [see, for example, ref 2-91. Many of these charge-transfer complexes were found to have quite high electrical conductivities and much interest in these works has focused on the mechanism of the charge-transfer process and its role in the conductivity of the compound. More specifically, past discussions have dealt with the extent of charge transfer in the complexes,6-8 and the valence orbital electronic structure of both TCNQ and some of its charge-transfer c o m p l e ~ e s . It ~ ,has ~ long been recognized that X-ray photoelectron spectroscopy (XPS) is useful as a probe of the various chemical states present in organic compounds [see, for example, the numerous discussions in ref 10 and 111. Detailed analysis of X P spectra of solids is often impaired by the instrumental energy resolution and/or natural Present address: Aerojet ElectroSystems, M/S 170-8244, P.O. Box 296, Azusa, CA 91702
0022-365418812092-1394$01.50/0
line width. Thus separation of peaks in a multicomponent spectrum is often very difficult. As a result, previously reported X P ~ p e c t r a ~ have * ~ - ’ been unable to greatly enhance our knowledge of TCNQ and many of its charge-transfer complexes. Many questions have been left unanswered on not only the number of chemically different atomic species in the molecule:but also the energies of the orbitals involved in the charge-transfer process and ( I ) Acker, D. S . ; Hertler, W. R. J. A m . Chem. S o t . 1962, 84, 3370. (2) Ikemoto, I.; Thomas, J. M.; Kuroda, H. Faraday Discuss. Chem. SOC. 1972, 54, 208. ( 3 ) Berlinsky, A. J.; Weiler, L. Solid State Commun. 1974, 15, 795. (4) Grobman, W. D.; Pollak, R. A,; Eastman, D. E.; Maas, Jr., E. T.; Scott, B. A. Phys. Rev. Lett. 1974, 32, 534. (5) Batra, I. P.; Bennett, B. I.; Herman, F. Phys. Rev. B 1975, 11, 4927. (6) Lin, S. F.; Spicer, W. E.; Schechtman, B. H. Phys. Rev. B 1975, 12, 4 184. (7) Grobman, W. D.; Silverman, 319.
B. D. Solid State Commun. 1976, 19,
(8) Ikemoto, I.; Sugano, T.; Kuroda, H. Chem. Phys. Lett. 1977, 49, 45. (9) Ritsko, J. J.; Epstein, A. J.; Salaneck, W. R.; Sandman, D. J. Phys. Rev. E 1978, 17, 1506. (10) Carlson, T. A. Photoelectron and Auger Plenum: New . Spectroscopy; . York, 1975; Chapter 4. ( I I ) Topics in Applied Physics, Ley, L., Cardona, M., Eds.; SpringerVerlag: Berlin, 1979; Volume 27.
0 1988 American Chemical Society
The Journal of Physical Chemistry, VoZ.92, No. 6,1988 1395
Letters
n
i"b.,
N Is
CIS
n
z
0
Ym a W
n m
k z
3
t26eV14
0
\
0
,
15
,
,
,
400
,
I
395
I
I
I
1
Figure 1. XP spectra of TCNQ over the N 1s region at low (a) and high (b) resolution. Note the base-line separation and energy difference between the two peaks in (b). their spatial distribution in the molecule. We present here core level X-ray photoelectron spectra of TCNQ taken with sufficient energy resolution to separate previously unresolved features. Under these conditions shake-up satellites provide us with information on the spatial distribution of the highest occupied molecular orbital, and a measure of the energy difference between the highest occupied and lowest unoccupied molecular orbitals of the photoion. Experimental Section All experiments were carried out using a VG Scientific ESCALAB Mk. I1 surface analysis instrument. The spectrometer was calibrated to the Ag 3d5/2line at 368.2 eV binding energy. High-resolution photoelectron spectra were taken with monochromated A1 Ka radiation with the spectrometer held at a constant pass energy of 10 eV. The source/analyzer contribution to the experimental energy resolution is approximately 0.6 eV in this configuration. Low energy resolution spectra were obtained by illuminating the sample with standard (unmonochromatized) AI Ka radiation with an analyzer pass energy of 20 eV. In this experiment an energy resolution of approximately 1 .O eV is obtained. T C N Q was obtained from the Aldrich Chemical Co. with a labeled purity of 98%. Spectra were obtained from three types of TCNQ samples, all of which were at room temperature during analysis: (1) unpurified powder pressed onto a Ag-painted sample holder; (2) recrystallized from spectroscopic grade acetonitrile producing crystals approximately 2 mm X 2 mm in size; these were then mounted to the sample holder with Ag paint; and (3) deposited as a thin film by placing a droplet of TCNQ-saturated acetonitrile on the sample holder and allowing the solvent to evaporate. Surface oxygen contamination was detected with a peak intensity of less than 3% of the carbon plus nitrogen intensities. High-resolution spectra from all three types of samples were nearly identical in shape, relative peak intensities, and binding energy positions. Thus purification and sample thickness had no effect on recorded spectra. This indicates that TCNQ is of low enough electrical resistivity (the electrical resistivity of TCNQ is 2 X 1OIo ohm.cm)IZ for this type of experiment to prevent significant sample charging which can hamper analysis of photoelectron spectra. Results and Discussion N 1s Region. Figure 1 shows N 1s core level spectra of TCNQ taken under low (spectrum la) and high (spectrum lb) energy resolution conditions. It may be noted that spectrum l a is quite similar to those published by Grobman et aL43' and Ikemoto et aL2 The binding energy of the most intense peak of Figure l a is 399.7 eV which is similar to that found by Grobman et A second peak, 2.6 eV to higher binding energy from the most
1
,
I
I
,
1
285
BINDING ENERGY ( e V )
Figure 2. XP spectra of TCNQ over the C 1s region at low (a) and high (b) resolution. Notice the two highest binding energy features are completely separated and have the same binding energy difference as the two features in Figure 1b. intense peak of Figure Ib, is also shown which has approximately 20% the intensity of the main peak. This relatively small peak is common to core level spectra of TCNQ [see, for example, ref 2 , 4 , 7 , 8 , 13, 141 and, as discussed by Ikemoto et a1.2 and Aarons et al.I4 is attributed to an intramolecular electronic excitation (shake-up) process. As discussed by CarlsonIo this process often occurs due to a sudden change in shielding felt by the valence electrons upon core photoionization. As shown by Aarons et aI.,l4 within the sudden approximation the transition probability is related to the square of the overlap integral between the initial and the final valence electron wave functions of the system where (TCNQ+)excited State refers to the valence electronic wave function for the TCNQ' photoion with appropriate core hole and a valence electron excitation. Indeed, Aarons et al.I4 have used wave functions within the INDO approximation to estimate the shake-up intensity associated with the N 1s photoionization of TCNQ and obtain a value of 20% of the main peak in excellent agreement with our results. They also note that the only transition giving significant intensity involves the highest energy doubly filled orbital of TCNQ and the lowest energy unfilled valence orbital of the ion. For this type of transition to have high probability upon N 1s photoionization the N atoms must contribute significantly to the highest lying molecular orbitals of TCNQ and the lowest unoccupied valence orbital of the ion. Thus it is logical to propose that the highest occupied molecular orbital of TCNQ has some spatial extent near the N atoms of the molecule. The energy shift of the satellite from the main peak is somewhat lower than the approximately 3.5-eV energy difference between the highest occupied (b1J and lowest unoccupied (bZg)TCNQ molecular orbitals as presented by Lin et aL6 This is expected, however, in light of the orbital relaxation which occurs due to creation of the core hole. This relaxation may lower the energy of the lowest unoccupied molecular orbital (LUMO) of the photoion more than that of the highest occupied molecular orbital (HOMO), decreasing the energy gap between these orbitals in the photoion and lowering the transition energy between them. C 1s Region. Spectra over the C 1s region of the photoelectron spectrum are shown at low and high resolution in Figure 2, a and b, respectively. Figure 2a is again similar to C 1s spectra previously published by Ikemoto et a1.2 In this lower resolution spectrum we note the presence of a relatively broad feature centered at approximately 286 eV and a low-intensity shoulder at higher binding energy. This shoulder, like the high-binding energy shoulder in Figure l a , is attributed to an intramolecular electronic excitation. It is obvious from the structure of the main (13) Tsuchiva. S.: Seno. M. Chem. Phvs. Left. 1982. 92. 359. (14j Aarons, L.J.; Barber, M.; Connor,*J. A.; Guest, M. F.; Hillier, I. H.; Ikemoto, I.; Thomas, J. M.; Kuroda, H. J . Chem. SOC.,Faraday Trans. I 1973, 69, 270. ,
(12) Afify, H . H.; Abdel-Kerim, F. M.; Aly, H. F.; Shabaka, A. A. Z . Naturforsch 1978, 33, 344.
I
290
BINDING ENERGY ( e V )
I
I
1396 The Journal of Physical Chemistry, Vol. 92, No. 6, 1988
TCNQ 4
\,c=c./
fN \ 2,c=c*
3/
/
+.09 -.lo H' n F'igure 3. Atomic structure of TCNQ. Carbon atoms in unique positions are labeled 1 through 4. Each carbon is also labeled with the calculated total charge as presented by Jonkman et al.I5
peak in Figure 2a that carbon atoms are present in more than one type of chemical environment in this compound. However, the limited resolution of the spectrum of Figure 2a prohibits determination of the number of significantly different chemical environments and the binding energies of peaks associated with them. As indicated in Figure 3, carbon atoms exist in four distinct environments in TCNQ (labeled 1, 2, 3, and 4 in Figure 3). The higher resolution spectrum, Figure 2b, is dramatically different from Figure 2a. It is apparent from Figure 2b that only two main carbon features are present which deviates from the maximum possibility of four types of carbon. In addition to resolution of the main carbon peak we have also achieved complete separation of the shake-up satellite feature and may assign it a binding energy of 289.2 eV. This binding energy value is 2.6 eV above that of the C Is peak at 286.6 eV which correlates with the 2.6-eV binding energy difference between the core and satellite features in Figure lb. If a shake-up satellite existed 2.6 eV above the 285.3-eV peak it would fall at 287.9 eV which is in the base-line-resolved valley between the 286.6-eV peak and the satellite a t 289.2 eV. Thus our data indicate there is no shake-up satellite associated with the C 1s feature at 285.3 eV. We are now left with the task of relating each feature in Figure 2b to the atomic and electronic structure of TCNQ. Neglecting for the moment the shake-up feature we move first to a discussion of the two peaks in Figure 2b centered at 286.6 and 285.3 eV. In general, the more positive the total charge on an atom the higher will be the binding energy of its electrons. Recent electronic structure calculations of TCNQ by Jonkman et al.I5 can be used to shed light on the C Is spectra of Figure 2b. Jonkman et al. report atomic charge distributions for TCNQ which indicate that the carbon atoms fall into two obvious categories. Carbon atoms 1 and 2 have net negative charge whereas atoms 3 and 4 have a net positive charge. The net charges reported by Jonkman et al. are shown in Figure 3. Thus it is reasonable that we have been able to separate the broad C 1s feature of Figure 2a into two major features in Figure 2b. This analysis would lead to an assignment of the peak a t 286.6 eV to the more positive carbons 3 and 4 and the peak at 285.3 eV to carbons 1 and 2 (the ring carbons). Indeed, this assignment is similar to that observed by Nakayama et a1.16 for the C 1s emission from condensed benzonitrile in which the cyanide carbon binding energy is observed to be 1.6 eV greater than that of the ring carbons. From the carbon count in each group we expect the two C 1s peaks to be of approximately equal intensity as each group contains six carbon atoms. However, we see in Figure 2b that the C Is peak intensities differ by approximately 12%. (In fact, as will be shown, the high binding energy peak intensity should include that from the shakeup feature which pushes this ratio even further from unity.) It is important to point out here that the 285.3-eV peak of Figure 2b was found to decrease in intensity as a function of time during the experiment. Thus it is likely the sample is X-ray sensitive and some degradation occurred over the 3 h required to obtain (15) Jonkman, H. T.; van der Velde, G. A.; Nieuwpoort, W. C . Chem. Phys. Left. 1974, 25, 62. (16) Nakayama, T.; Inamura, K.; Inoue, Y . ;Ikeda, S. Surf. Sci. 1987,179, 47.
Letters the spectrum of Figure 2b. From this it is reasonable that the above carbon-peak associations give rise to unequal intensities of the 286.6- and 285.3-eV peaks in Figure 2b. An interesting aspect of the C 1s s p e c p m of Figure 2b is the presence of two main peaks but a single shake-up feature. The mechanism of the shake-up process in association with this indicates that not only are two types of carbon present in TCNQ with different 1s binding energies but also their contributions to the H O M O of TCNQ vary greatly. As mentioned earlier, we can use the 2.6-eV difference in binding energy between the main N 1s peak and its shake-up satellite in Figure l b to link the shake-up feature in Figure 2b with the higher binding energy carbon peak as the binding energy difference is also 2.6 eV. Thus from the C 1s and N 1s high-resolution spectra we have located the H O M O of TCNQ near the cyano ends of the molecule. In fact, since some spatial overlap between the HOMO and LUMO should be important in the electronic excitation giving rise to the shake-up peak it may be true that the LUMO is also situated at the ends of the molecule. This positioning of the LUMO of TCNQ correlates well with the atomic charge distributions of TCNQcompared with TCNQ as presented by Jonkman et al.I5 In their study, it is seen that a significant amount of the additional electron density of the ion resides near the methane-type carbons at the end of the molecule. The occurrence of but a single shake-up feature with two main peaks has also been shown by Barber et al. in the N 1s spectrum of 3-methy1sydnone.l' In their report they show that two N atoms bound next to one another may have quite different 1s binding energies as well as different contributions to highest occupied and lowest unoccupied molecular orbitals. This is similar to the C 1s spectra of TCNQ reported here as well as our discussion which extends to the atomic orbital composition of the TCNQ molecular orbitals. Further evidence for this position of the L U M O of T C N Q may be seen in SEM images of AgTCNQ and AlTCNQ charge-transfer complexes as presented by Uyeda et a1.I8 In these micrographs it can be seen that the metal-TCNQ bonds occur near the cyano ends of the molecule. Thus the charge transfer occurring on compound formation, which must involve the LUMO of TCNQ, involves interaction between the metal and cyano ends of the molecule.
Conclusions We have identified the 1s binding energy positions of both C and N atoms in the X-ray photoelectron spectrum of solid TCNQ. The energy difference between the highest occupied and lowest unoccupied molecular orbitals in the TCNQ' photoion with a core hole is found to be 2.6 eV. The C 1s photoelectron spectrum at high energy resolution indicates the presence of two types of carbon in this molecule, consistent with charge distributions reported by Jonkman et al.15 Also, a shake-up feature associated with the N 1s peak and only one of the two C 1s peak indicates the HOMO and LUMO are located near the cyano ends of the molecule. In closing we should again note that the electronic excitation discussed above occurs simultaneously with core hole creation. Thus our conclusions regarding the energies and spatial distributions of the LUMO involved in this excitation are relevant in the strictest sense only for the TCNQ' ion with N 1s or C 1s core holes. However, the bonding position of metal atoms in AgTCNQ and AlTCNQ at the cyano ends of the molecule indicate that the difference in position of the LUMO before and after photoionization may be negligible. Our data indicate the same is true for the H O M O of TCNQ. Acknowledgment. This work was supported in part by the Office of Naval Research. J.M.L. acknowledges support from the IBM Corporation in the form of a Graduate Research Fellowship for 1986-1987. J.C.H. thanks the Alfred P. Sloan Foundation for support in the form of an Alfred P. Sloan Foundation Research Fellowship. (17) Barber, M.; Broadbent, S . J.; Connor, J. A.; Guest, M. F.; Hillier, I. H.; Puxley, H. J. J . Chem. SOC.,Perkin Trans. 2 1972, 1517. (18) Uyeda, N.; Kobayashi, T.; Ishizuka, K.; Fujiyoshi, Y . ;Inokuchi, H.; Saito, G. Mol. Cryst. Liq. Crysf. 1985, 125, 103.