J . Phys. Chem. 1989, 93, 749-753 parameters and corresponding quantities in the model.
This investigation was made possible by a grant to W. J. by the Fulbright Commission and by Grant C H E 810055 to A.T.W. from the National Science Foundation (Chemical Dynamics). We are grateful to Erzsebet Lugosi for the vectorized equation solver and to John Tyson for collaboration
749
in Tucson, January 1987, when the first Oregonator rotors finally worked, and for subsequent improvement of initial conditions and discussions of the Oregonator throughout the preparation of this paper. Registry No. Br03-, 15541-45-4; malonic acid, 141-82-2; ferroin, 14708-99-7.
Stability and Decomposition of Gaseous Allylcyclopentadienylpalladium G. T. Stauf, P. A. Dowben,* Laboratory for Solid State Science and Technology, Syracuse University, Syracuse, New York 13244-1 130
K. Emrich, S . Barfuss, W. Hirschwald, Institut fur Physikalische Chemie, Freie Universitat Berlin, Takustrasse 3, IO00 Berlin 33, Federal Republic of Germany
and N. M. Boag Department of Chemistry and Applied Chemistry, Salford University, Salford, England, M5-4 W T (Received: May 31, 1988)
The ionic and neutral decomposition of gaseous allylcyclopentadienylpalladium Pd(q5-C5H5)(q3-C3Hs) has been investigated via photon- and electron-induced ionization mass spectroscopy. The thermodynamic cycles incorporating both ionic and some neutral fragments have been constructed from these results. The thermodynamic cycles illuminate the decomposition process of allylcyclopentadienylpalladium.This has potential application in the production of metallic Pd coatings both by pyrolysis and by patterned photolysis using an UV laser.
I. Introduction A careful analysis of the thermodynamics of molecular decomposition can be very useful in determining whether or not a particular organometallic compound has any potential as a source compound for organometallic vapor-phase epitaxy (OMVPE or OMCVD). Nonetheless, only a few detailed studies have been undertaken that investigate the decomposition thermodynamics of candidate organometallic complexes.'** This is unfortunate, because patterned chemical vapor deposition (CVD) (often laser-assisted) of metal films is a useful way of "direct" writing interconnects, and repairing damaged masks and The study of a thermodynamic cycle describing the neutral decomposition of a molecule is important in understanding both photolytic deposition and pyrolysis. Pyrolysis is a common way to deposit a coating by CVD, while photolytic and photoassisted deposition are being heavily investigated on a more fundamental level, due to this promise of patterned deposition. The photoassisted deposition of palladium, using gaseous allycyclopentadienylpalladium as a source material, has recently been demonstrated! We have, therefore, in this study investigated the bond energies within this promising compound, Pd(Cp)(C3H5), using both electron impact and photoinduced ionization to determine the energies needed to break the bonds and decompose the molecule (throughout this paper, C p = cyclopentadiene = q5-CSHS). The information was then used to construct a thermodynamic cycle, which illustrates possible decomposition pathways. For plasma-assisted deposition processes, so common in the semiconductor industry, an understanding of the plasma chemistry would seem to be crucial to modeling of plasma decomposition and predicting deposition products. In a plasma, positive ions are created from the ambient gas as a result of the impact of electrons into neutral gas molecule^.^ Similar processes, resulting in the creation of positive ions, occur in the ion source of a mass spec-
* Author to whom correspondence should be addressed. 0022-365418912093-0749$01.50/0
trometer, although the reactions of ions in a mass spectrometer do not always parallel those reactions found in commercial plasmas or electron- and ion-assisted deposition processes (which have much higher pressures and a wider range of impact energies). Incident electrons will, if sufficiently energetic, create positive ions, which may then be mass selected and detected. By characterization of the ionic decomposition products and associated energetics with a mass spectrometer, some processes that take place in a plasma can be investigated and an ionic decomposition side of the thermodynamic cycle built UP.^-'^ While Pd(Cp)(C3H5)+has been previously studied by mass spectroscopy, only the types of fragments were examined, with no determination of the energies necessary to produce 11. Experimental Section
The Pd(Cp)(C3Hs) complex was prepared as described previously.12 The photolytic coatings discussed herein were made over periods from 1 to 2 h by using the 337-nm (3.68 eV) line of a nitrogen laser operating at 10 pulsesls, 0.4 M W per pulse, (1) Dowben, P. A,; Spencer, J. T.; Stauf, G. T., to be submitted for pub-
lication.
(2) Stauf, G. T.; Driscoll, D. C.; Dowben, P. A.; Barfuss, S.; Grade, M. Thin Solid Films 1987, 153, 421. (3) Ehrlich, D. J.; Tsao, J. Y . J. Vac. Sci. Technol., E 1983, 1, 969. (4) Baum, T. H.; Marinero, E. E.; Jones, C. R. Appl. Phys. Lett. 1986, 49, 1213. (5) Blonder, G. E.; Higashi, G. S.; Fleming, C. H. Appl. Phys. Lett. 1987, 50, 766. (6) Stauf, G. T.; Dowben, P. A. Thin Solid Films 1988, 156, L31-L36. (7) Rand, M. J. J . Vac. Sci. Technol. 1979, 16, 420. (8) Driscoll, D. C.;Bishop, J. A.; Sturm, B. J.; Dowben, P.A.; Olsen, C. J. J . Vac. Sci. Technol., A 1986, 4, 823. (9) Kime, Y. J.; Driscoll, D. C.; Dowben, P. A. J . Chem. Soc., Faraday Trans. 2 1987, 83, 403. (10) Zakurin, N. V.; Gubin, S. P.; Bochin, V. P. J . Organomet. Chem. 1970, 23, 535. (11) King, R. B. Appl. Spectrosc. 1969, 23(2), 148. (12) Shriver, D. F. Inorganic Syntheses; Wiley: New York, 1979; Vol. 19, p 220.
0 1989 American Chemical Society
750 The Journal of Physical Chemistry, Vol. 93, No. 2, 1989
giving a power density at the focused spot on the substrate of not more then 0.434 W/cm2. This laser was shined through a nickel wire square mesh, which was clamped over the substrate at a distance of around 60-300 wm (0.06-0.3 mm), on the basis of calculations using diffraction patterns discussed later. The coatings were made by using a continuous flow of vapor from the organometallic source, in a glass vacuum system capable of achieving a base pressure of =lo" Torr. In order to collect the XES spectra, we used a KEVEX 5500 spectrometer fitted onto an IS1 Super I1 SEM, with 25-keV incident electron energy. More details can be found in ref 6. The electron impact mass spectroscopy experiments were undertaken by using a molecular beam of sample vapor generated in an alumina Knudsen cell. This beam was directed into the ionization region of the electron impact ion source of a Varian MAT single-sector magnetic field mass spectrometer. Ionization efficiency curves (IEC), i.e. plots of ion intensity versus electron impact energy, were recorded under isothermal vaporization conditions in steps of 0.025 eV between 1 - and 26-eV incident electron energy. Calibration, data reduction, evaluation procedure, and analysis of the fine structure of the IEC's were undertaken with procedures outlined e l ~ e w h e r e . ' ~ , ' ~ The alumina Knudsen cell, used as a source for gaseous parent species, was encased in a tantalum mantle. The sample gas was generated from the Knudsen cell, which contained equilibrium gas phase above the pure solid organometallic at temperatures between 300 and 350 K. A constant temperature in the Knudsen cell was used for each experiment. Experiments were repeated at several different temperatures and different corresponding effusion rates to confirm that results remained independent of pressure. The fragmentation of Pd(Cp)(C3H5)was also studied inside a high-temperature photoionization system using synchrotron radiation in the photon energy region 6-24 eV. The synchrotron radiation source was the electron storage ring BESSY (Berliner Elektronenspeicherring Gesellschaft fuer Synchrotronstrahlung mbH) in Berlin, BRD. The light was dispersed by a Wadsworth monochromator with a 1200 lines/mm MgF2-coated AI parabolic grating and reflecting mirror. The ions were detected with a Balzers QMG 5 11 quadrupole mass spectrometer. The Auger electron spectroscopy results were taken with a Perkin-Elmer CMC, following Ar+ ion bombardment to remove surface contamination. Approximately 100 A of the thin film was removed prior to acquiring representative AES results. 111. Results
The following parent and major fragment species were identified during electron impact mass spectroscopy: Pd(Cp)(C3H5)+ (parent), Pd(Cp)+, Pd(C3H5)+,Pd', C5H5+(Cp), C5H6+,C3H5+ (allyl) and C3H3+.The ionization (IP) and appearance potentials (AP) for the major species are summarized in Table I. Good agreement for the IP and AP data obtained with both electron and photon impact mass spectroscopy was found, with the largest difference (0.4 eV) between the results of the two methods occurring in the measured appearance potential of the Pd+ fragment. A comparison of the two techniques for the parent ionization potentials and first appearance potentials of fragment ions is included in Table I. Both ionization and appearance potentials for the parent and fragments were obtained from the ionization efficiency curves (IEC's). The IEC is, effectively, the molecular or fragment ion intensity plotted against the electron (or photon) impact energy employed to ionize the gaseous species, as shown in Figure 1. With increasing impact energy, a greater number of molecular orbital excitations become accessible, causing increases in the parent IEC intensity vs energy slope. In addition, of course, these higher impact energies cause breakage of bonds within the (13) Grade, M.; Wienecke, J.; Rosinger, W.; Hirschwald, W. Ber. Bunsen-Ges. Phys. Chem. 1983, 87, 355. (14) Rosinger, W.; Grade, M.; Hirschwald, W. Ber. Bunsen-Ges. Phys. Chem. 1983, 87, 5 3 6 .
Stauf et al. TABLE I ionization and Appearance Potentials Based on the Ionization Efficiency Curves of Pd(Cp)(CIHS)"
re1 increase (1) or decrease (1) species PdCpCIHct _ -
PdCpt
PdC3HSt
Pd+
CSH,+
C5H,+ (Cp)
AP, eV 7.7 (electron) 7.7 (photo) 7.7 (avg) 8.5 8.9 9.3 10.4 12.2 12.8 13.9 10.8 (electron) 10.8 (photo) 10.8 (avg) 11.6 12.0 12.5 13.4 14.1 14.9 16.9 12.0 (electron) 12.1 (photo) 12.0 (avg) 13.7 15.1 17.3 19.5 14.0 (electron) 13.6 (photo) 13.8 (avg) 15.6 16.5 18.6 21.0 8.9 (electron) 9.6 10.1 11.3 8.8 (electron) 8.8 (photo) 8.8 (avg) 9.7 10.2 11.1 11.7 13.2 14.7 8.2 (electron) 8.2 (photo) 8.2 (avg) 9.0 12.4 13.1 14.6
of slope in parent IEC
m/e
(based on
Pd'") 212
t t t t 1 i 1 t 171
t t t t t t 1 1
147
t t t 1 1 106
t t t t t t t t t
66
65
t t t t t 1 1 41
t t 1 1 1
"AII.OWS designate relative increases ( 7 ) and decreases (4) of slope in the parent IEC's. Initial AP's are given both for electron and photoionization data, which are then averaged, while higher AP's are from electron impact data only. Values are given in electronvolts and rounded to tenths, as estimated error is *O.l eV.
molecule and downward slope changes in the IEC. Thus, from the IEC, higher A P s may be determined. It should be noted that while photon impact data provides very accurate IP's of the parent and first A P s of the fragment ions due to its high degree of monochromaticity, it is not as useful for obtaining higher appearance potentials (slope changes) from IEC plots. This can be seen in Figure 1, in IEC's derived from electron impact and photon impact of Pd(Cp)(C3H5). The considerably better signal to noise ratio (due to much higher cross section for impact) of the electron technique versus the photon technique is apparent. In addition, the photoionization IEC has much more
The Journal of Physical Chemistry, Vol. 93, No. 2, 1989 751
Decomposition of Pd(q5-CSH5)(q3-C3H5) 1001
1
TABLE 11: Ionic Fragmentation Processesu breaks in
probable parent fragment fragmentation parent IEC, eV AP, eV fragment pathway PdCpCpH5 8.5 (t) 8.2 C3H5+ predissociation PdCpC3HS - CjHS 8.9 (t) 8.8 Cp+ predissociation PdCpC3HS - Cp 10.4 (1) 10.8 PdCp+ direct PdCPC3HS - CpH5 12.2 (1) 12.0 PdCpH5+ direct PdCpC3H5 - Cp 13.9 (t) 13.8 Pd+
Electron Impact Energy (eV)
t
I? The fragmentation processes suggested by our data are indicated. The energies of fragmentation are found from considering both the fragment appearance potential and the appropriate slope changes (breaks) in the parent ion IEC.
d
B
a
8:
Photon Energy (eV)
Figure 1. (A) Ionization efficiency curve (IEC) from electron impact
ionization for Pd(Cp)(C3H5). Straight lines are drawn through appropriate segments of the plot, with slope changes indicated at arrows. (B) IEC from photoionization for Pd(Cp)(C3H5).
ou, , -
,A+-;,
, , , , , , , , , , , ,
I
8 9 IO II 12 13 14 I5 16 17 18 19P21 2223242526 Electron Impact Energy (sV)
Figure 2. Breakdown diagrams for Pd(Cp)(C3H5)derived from electron impact IECs. Relative intensities are plotted as a function of electron impact energy, with an intensity of 100%implying that this is the only ion, (0; observed fragment. Main plot: (*) parent Pd(C5HS)(C3H5)+ C3H5+ion, (A)C3H3+ion. Corner inset plot: ( 0 ) Pd+ ion, (0)PdC5HS ion, (0) PdC3H5+ion, (A) C5HS+ion, (V) CSH6+ion.
variation in its slope, going from positive to negative to flat to positive again, due to the complexity of the light absorption cross section of organometallic molecules such as Pd(Cp)(C,H,). The disadvantage of electron impact ionization, on the other hand, is that it lacks (compared to photoionization) high monochromaticity, causing an exponential tail in electron-impact-derived IEC's. This exponential (rather than the desired linear) rise a t the onset of ionization can be difficult to factor out during data analysis. Fortunately, a combination of the two ionization techniques acts to eliminate the uncertainties associated with either one alone. The relative intensities of the parent and fragment ions are seen plotted in Figure 2. They are based on electron impact data, because the magnetic sector mass spectrometer used for the electron impact studies was capable of better mass resolution than the quadrupole mass spectrometer used for photoionization and was able to easily distinguish C5HS+from C5H6+,a I-amu separation.
IV. Discussion The ionization potential and appearance potentials of the parent and fragment ions, listed in Table I, are seen to be internally consistent. Not only do the photon and electron impact measurements agree, but the higher appearance potentials of the parent IEC can be related to the AP's of the fragment ions. Thus several fragmentation processes, summarized in Table 11, can be assessed from a comparison of the parent ion and fragment ion appearance potentials. A decrease in the slope of the parent Pd(Cp)(C3H5)+ion IEC is observed from electron impact measurements at 10.4 eV, in agreement with the 10.8 eV A P of the Pd(Cp)+ fragment. This downward slope change in the parent ion IEC suggests that at this energy, 10.4-10.8 eV, a pathway for direct fragmentation of the parent ion, leading to production of Pd(Cp)+, becomes available as described in Table 11. Similarly, another slope decrease occurs at 12.2 eV in the parent ion electron impact IEC, which is close to the 12.0-eV energy at which Pd(C3H5)+appears. This again suggests direct bond breakage. At 8.5 eV, however, an increase in the slope is observed in the parent IEC. At this energy we also found the A P of the C3H5+fragment ion. The upward slope change in the electron impact parent IEC is not typical of direct cleavage of the bond, but rather suggestive of a fragmentation mechanism that involves a predissociation excitation, Le. an orbital excitation to an antibonding orbital. The partial filling of an antibonding orbital in the parent ion must increase the probability for creating the C3H5+ion. This upward slope change is also seen in the formation of the Cp+ ion fragment with an A P at 8.8 eV, in agreement with the higher appearance potential of 8.9 eV in the parent IEC, and again illustrative of predissociation. Unfortunately, not enough is known about the source of the Pd+ fragment in the mass spectra to be sure of its origin. From the relative abundances shown in Figure 2, we observe that the allyl and the Cp fragments appear at close to the same energy, indicating that their bonds to the Pd are of approximately equal strength. The rearranged allyl fragment, C3H3+,appears at an energy of about 1 eV larger than the C3H5+and Cp+ ion A P s , indicating that the allyl tends to break off as C3H5+and will undergo rearrangement only if sufficiently excited. It will also be noted that the C5H5+and the C5H6+fragments are roughly equal in abundance at every energy above the 8.8-eV AP of Cp+. This would indicate that the fragmentation route eliminating Cp from Pd(Cp)(C3H5)includes two routes, in one case creating C5H6+ (abstracting H). As we show in Figure 2, the probability of creating fragment ions apart from Pd+, ligands, and (Pd ligand)+, even at the comparatively high electron impact energy of 25 eV, is small. Additional fragments we observed at the highest energy (= 26 eV) available to us were C7H7+,C6H7+,C&+, C6H5+,C5H4+,C5H3+,C3H6+,and C3Hz+.The most abundant of these minor fragments, C7H7+,was never observed with greater than a 5% relative abundance, and all these fragments added together never exceeded 33% abundance. The fragment CsHlo+, resulting from combination of C p and allyl, may also have been
+
152 The Journal ofphysical Chemistry. Vol. 93. No. 2. 1989
Stauf et al.
F i m 3. Decomposition thermodynamic cycles for neutral and ionic fragmentationof Pd(Cp)(C3HI),constructed from the AP/IP infnmation in Table 1. The average AP and IP values listed (average between electron and photoionization data) were used. All numbers are in mils of eletronvolts. Superscript a indicates values from ref IS, and b indicates values calculated by combining these values with ones obtained from completing the thermodynamic cycle.
present in small amounts but was dilficult todistinguish from the PdlM isotope, which appears at the same m / e ratio. Previous studies"," have obtained similar results showing fairly minor amounts of rearrangement even at energies of about 70 eV. This 'clean" cleavage of ligands bodes well for hopes of making uncontaminated coatings, possibly even via plasma deposition. Using A P s and IP's taken from Table 1. the ionic bond dissociation energy between the Pd and the C,Hs ligand in the parent ion can be calculated as D(PdCp+-C,Hs) = AP(PdCp+) - IP(PdCpC3HS) = 10.8 - 7.7 = 3.1 eV Similarly, the bond energy betwem the Pd and the Cp ring in the parent ion can be found as D(PdC,Hs+Cp) = AP(PdCIH,*)
- IP(PdCpC3Hs) = 12.0 - 7.7 = 4.3 eV
Using the IPenergy for Pd taken from the literature (IP(Pd) = 8.3 eV"), we can also calculate the neutral dissociation energy for the process Pd(Cp)(c~Hs) (Pd) + (Cp) + (C,Hs) as D(Pd-CpC,HS)' = AP(Pd+ Cp + CIH,) - IP(Pd) = 13.8 - 8.3 = 5.5 eV From this extraction of bond strengths we have constructed partial thermodynamic cycles for Pd(Cp)(C,H5) decomposition. as illustrated in Figure 3. This thermodynamic cycle does not exclude the possibility of C p and allyl recohbination or rearrangement during the abstraction prcces. Such behavior has bem seen with a variety of metaIlocenes.l6 One may also conclude from the Cp+ and C,HC AP's that the creation of Cp+ from Pd(Cp)(C,Hs)* is D(PdCiHsCp+) = AP(Cp)+ - IP(PdCpC1Hs) = 8.8 - 7.7 = 1.1 eV and creation of C,Hf from Pd(Cp)(CIHs)* is
-
+
D(PdCpCIHs+) = AP(CIHs)+ - IP(PdCpCIH,) = 8.2 - 7.7 = 0.5 eV Using additional literature values for the ionization of C p and C,H% (IP(Cp) = 8.6 eV; IP(C,H,) = 8.1 eVO). we can get further neutral decomposition energies D(PdCpCIH,)O = IP(PdCpC3HS) D(PdCp-C,Hj+) IP(C,H,) = 7.7 + 0.5 - 8.1 = 0.1 eV
+
and D(PdC3HsCp)' = IP(PdCpCiH5) D(PdC3HrCp') IP(Cp) = 7.7 + 1.1 - 8.6 = 0.2 eV
+
Thcse energies are seen to be much smaller than the energy needed in each case to remove the second ligand from the core (15) Rosenslock. H.Enrgnics of Garcow Ions; American Institute of Physics: New York. 1977; J . Phys. Chcm. Ref. Daro. Suppl. 6. No, 1. (16) BarfwS.;Gradc. M.;Hirschrald. W.;Rosinger. W.;bag. N. M.; Driscnll. D.C.;Dorkn. P. A. 1. Voc. Sci. Techno/.. A 1987. 5. 1451.
Figme4. (a) Scanning clearan microscope (SEM) picture of patterned photoinduced Pd deposition on a Si substrate (magnification I2SOX). with (b) an XES elemental map showing the locations of Pd (bright a m ) in the picture above, made by using the Pd XES line groupat 2.84. 2.98. 3.16. and 3.32 keV. (c) Another SEM picture. 18OOX magnification, of a Pd square on a cleaved (smooth) Si surface showing the finely structured diffraction-induced deposition paltern more clearly. Each square is 20 X 20 Mm. The coating was made by depositing over I '12 h as described previously.
Pdo (over 5 eV for both). This would suggest that these AP values are not representative of thermodynamic bond energies. The predissociation mechanism makcs assignment of the AP to the bond-breaking excitation very difficult. and the bond dissociation energies mentioned above for D(PdCp-C,H,+) and D(PdC,H+p+) demonstrate the dangers of relying on the apparent APs to calculate when predissociation excitation mechanism exist. In fact, these measurements for Cp+ and ClH5+ are clearly indicative of the Cp Cp' and C,Hs C3H,* ionization p r u s s e r reported in the literature." Recently: we have demonstrated that photolysis of Pd(Cp)(C,H,) using an N2laser (3.68-eV photon energy) leads to deposition of Pd thin films. Both photolysis and pyrolysis of
-
-
J. Phys. Chem. 1989, 93, 753-758 Pd(Cp)(C3H,) lead to the deposition of Pd. This fact is quite clear from the X-ray electron spectroscopy (XES or EDAX) spectra we have collected previously.6 From Auger electron spectroscopy no appreciable oxygen contamination was found, while carbon contamination after Ar+ sputtering to remove surface impurities was less than 10%. With use of close contact masks, as outlined elsewhere,6 photolysis of Pd(Cp)(C3H,) can be used to deposit Pd in submicron features as shown in Figure 4. While the actual squares in the figure are only 20 X 20 pm, ultimate resolution should be better, if use is made of the diffraction patterns seen. As previously mentioned, the mask-substrate distance in Figure 4a was calculated to be around 60-300 pm on the basis of the spacing of the fringes. If these patterns resulted from physical blockage of surface sites by the mask, it would then be too close for the diffraction pattern in Figure 4a to be produced. The diffraction subpatterns thus support our conclusion that the decomposition of this moolecule takes place on the surface and is truly photolytic in nature rather than pyrolytic. If pyrolysis were the driving force, laser-heated areas would surely not remain so distinct during a several hour deposition (see ref 6 for more detailed arguments on the subject). Since the neutral decomposition of Pd(Cp)(C3H5)to Pdo(g) requires 5.5 eV, as illustrated in Figure 3, the complete photolysis of Pd(Cp)(C3H5)cannot be a gaseous reaction. Such a process would have to be a two-photon process at 3.68-eV photon energy, and at a photon flux of around 6 photons/(A*/ns) this two-photon process is unlikely (similar flux per femtosecond would be necessary). Therefore, we can conclude that the deposition of Pd from the photolysis of Pd(Cp)(C3H,) occurs either from incomplete dissociation of the ligands or from the fact that the surface species are sufficiently long-lived to permit two-photon dissociation processes to occur. A third possibility is that the surface acts as a catalyst, lowering the bond energies in the molecule. Regardless of which of these processes is taking place, it is certain that photolysis of surface species must provide better pattern resolution than gaseous decomposition.
753
The photolysis of Pt(Cp)(C3H5)with 308-nm laser radiation has recently been demonstrated to lead to the deposition of Pt on quartz," with a thin-film composition of 24% C, 76% Pt. Auger electron spectroscopy of our coatings, in marked contrast, shows almost no carbon (much less than 5 1 0 % ) . This suggests either that photolysis of the allylcyclopentadienylplatinum is incomplete or that decomposition of the ligands occurs. The energetics of decomposition for Pt(Cp)(C3H,) is currently under investigation by methods similar to those reported herein. V. Conclusions We have constructed the thermodynamic decomposition cycle for gaseous allylcyclopentadienylpalladium, Pd(Cp)(C3H5). Both electron-induced ionization and photoionization were used to obtain appearance potentials with a high degree of confidence. With the use of literature data, both the ionic and some neutral parts of the cycle could be derived. This cycle and relative abundance information were used to draw the conclusion that the decomposition of Pd(Cp)(C3H5)would be relatively "clean" for photon energies above 5.5 eV. Photodeposition leading to thin-film formation is more likely to be a surface-nucleated reaction with the proper choice of photon energy (less than 5.5 eV). This surface selectivity may be very useful for obtaining the ultrafine line resolution in such demand in the semiconductor and other industries, and it is observed to result in the deposition of very clean Pd films.
Acknowledgment. This work was funded by the US.DOE through Grant No. DE-FG-02-87-ER-453 19, the Deutsche Forschungsgemeinschaft/Sonderforschung Bereicht 6 (DFG/SFB 6), the Syracuse University Senate, and Engelhard Corp. We thank G. 0. Ramsayer for his assistance in obtaining the AES results. Registry No. Pd(Cp)(C,H,), 1271-03-0. (17) Rooney, D.; Negrotti, D.; Byassee, T.; Macero, D.; Chaiken, J.; Vastag, B., submitted for publication in J . Electrochem. SOC.
Importance of Molecular Size on the Dynamics of Solvent Relaxatlon Shyh-Gang Su and John D. Simon*,+ Department of Chemistry B-041, Institute of Nonlinear Studies, University of California at San Diego, La Jolla, California 92093 (Received: June 27, 1988)
The time evolution of the Stokes shift of the twisted intramolecular charge-transfer emission from (dimethy1amino)benzonitrile (DMABN) and (diethy1amino)benzonitrile (DEABN) is examined in 1-propanol solution. Over the temperature range from -10 to -50 OC the Stokes shift correlation function, C(t),is nonexponential with an average relaxation time, ( T J , different from that predicted by theories that model the solvent as a dielectric continuum. In addition, ( T ~ for ) DMABN is faster than for DEABN, clearly showing that the solvent relaxation measured by C(t) is dependent on the size of the solute. The data are compared with recent mean spherical approximation (MSA) models for ion and dipole solvation that take into account the relative sizes of the solvent and solute. The ion-MSA theory provides a reasonable fit to the data; however, the dipole-MSA theory significantly underestimates the time scale of the solvent relaxation. Comparison with related studies on the solvation ) in the order C153 > DEABN > DMABN, opposite to of Coumarin 153 (C153) in 1-propanol reveals that ( T ~ decreases that predicted by the MSA models.
Introduction Dielectric continuum models are commonly used to gauge solvent fluctuation time scales in polar solvents.'** In particular, with a continuum model, the relevant time scale in electrontransfer processes is given by the longitudinal relaxation time, TL ( 7 = ~ ~ ~ c = , / t ~ ) . ~Whether -~ dielectric continuum theory is able to accurately gauge solvent relaxation times has prompted several 'National Science Foundation Presidential Young Investigator 1985-1990, Alfred P. Sloan Fellow 1988-1992.
0022-365418912093-0753$01 SO10
experimental studies aimed at measuring the microscopic dynamics of solvent relaxation. These studies have focused on measuring the time-dependent Stokes shift of a probe molecule dissolved in the polar solvent of i n t e r e ~ t . ~ - 'The ~ time-dependent spectral (1) (2) 1949. (3) (4) (5)
Kivelson, D.; Madden, P. A. Annu. Rev. Phys. Chem. 1980, 31, 523. Frohlich, H. Theory of Dielectrics;Oxford Univ. Press: Oxford, U.K., Sumi, H.; Marcus, R. A. J. Chem. Phys. 1986,84, 4859. Rips, I.; Jortner, J. J. Chem. Phys. 1987, 87, 2090. Sparpaglione, M.; Mukamel, S . J . Chem. Phys. 1988.88, 3263.
0 1989 American Chemical Society
