Polarization-resolved (2 + 1) resonance-enhanced multiphoton ionization spectroscopy of CF,I (6s) Rydberg states

Transcription

1 Polarization-resolved (2 + 1) resonance-enhanced multiphoton ionization spectroscopy of CF,I (6s) Rydberg states Craig A. Taatjes, Johanna. G. Mastenbroek, Ger van den Hoek,a) Jaap G. Snijders, and Steven Stolte Laser Centrum and Vakgroep Fysiche and Theoretische Chemie, Faculteit der Scheikunde, Vrije Universiteit, de Boelelaan 1083, 1081 HV Amsterdam, The Netherlands (Received 16 September 1992; accepted 9 November 1992) The CF,I( 5pr-6s) Rydberg transitions in the energy range cm- are investigated using (2+ 1) resonance-enhanced multiphoton ionization. The polarization of the twophoton transitions is used to definitely assign the symmetries of the resonant intermediate states. The four allowed electronic transitions in the (5p~--6s) manifold have been assigned and some vibrational constants in the excited states have been determined. Hot band spectra have been obtained in a supersonic expansion of CF,I through an oven. The upper spin-orbit components (the 2E1,2 ion core states) are perturbed by a dissociative state at approximately cm-, possibly the (T-O* transition centered on the C-I bond. Density functional calculations have been performed in order to help determine the nature of the perturbing states. Vibronic interactions in the excited states are investigated, and evidence is seen for quadratic Jahn-Teller interactions for y6 in the lower (2E3,2> spin-orbit state. 1. lntroductlon The CFsI molecule has been the subject of recent study in connection with the ability to pump it via multiphoton absorption of CO, laser radiation into highly vibrationally excited states.1a The electronic spectroscopy of CF,I is of interest for several reasons. (2 + 1) resonance-enhanced multiphoton ionization (REMPI) has been investigated for use as a probe method for vibrational pumping experiments.5 The analogy of CF,I to the far more extensively studied methyl iodide ( CH31) molecule g is a probe of the effect on electronic properties of lluorine substitution. Although the importance of alkyl iodides themselves in atmospheric chemistry is small, the photochemistry of fluorinated alkyl halides has been of considerable recent interest because of their role in atmospheric ozone depletion. In addition, the electronic effects of fluorination have been studied in connection with the dissociation properties of atomic iodine laser precursors.12 The spectroscopy of CFsI was studied by one-photon absorption in the early 1960 s by Herzberg13,14 and by Sutcliffe and alsh. The first strong structured band seen after the dissociative A band in the absorption studies has an origin at cm- (air) In the early work this band was assigned to the 5pz--6s Rydberg excitation, in analogy with the already thoroughly studied B band of CH31. This assignment had been questioned by Robin, who noted the low term values obtained for the states assuming the Rydberg assignment, and also discovered another weak absorption band (the B state) at cm-, which he suggested may be the lower spin-orbit state of the 5~~6s transition.r6 However, later the trend of term values on fluorination of alkyl halides led Robin to suggest that the new B state was actually of valence character,17 and that )Department of Molecular and Laser Physics, Katholieke Universiteit Nijmegen, Toemooiveld, 6525 ED Nijmegen, The Netherlands. the electronic assignments of Sutcliffe and alsh were likely correct. The 5~~6s transition is split by the large spin-orbit couping of the I atom ( cm- ), and as in CH31, these are additionally split by the exchange interaction. The ion core states, the 2E3,2 and 2E1,2, are thus split into states with projections of the total electronic angular momentum IR of 2, 1, 0, 1, in order of increasing energy. e will label these four states as ion core states with definite total electronic angular momentum projection [n], i.e., 2E3,2 ion core state (C state) produces 2E3&l] and 2E3/2[l] states upon coupling with the Rydberg electron, and the 2E1,2 ion core state (D state) produces 2E1,2[O] and 2E1,2[ l] states. The analogous states of CH,I have been very extensively studied. The 5pg-6s Rydberg transition to the 2E3/2 ion core state, the so-called B band of methyl iodide, was studied by Mulliken and Teller,6 who demonstrated the pseudoparallel nature of the one-photon absorption and showed that the upper state was indeed the E-symmetry 6s Rydberg level. Both spin-orbit states were studied with RBMPI by Parker et al.,7 who used polarization techniques to definitely assign the various electronic origins as well as the vibrational bands. By analysis of the rotational profiles of one-photon and two-photon1g spectra, coupled with a multichannel quantum defect treatment,20 Mc- Glynn and co-workers investigated the decoupling of the spin of the Rydberg electron in these states of CH,I. Herzberg noted the presence of Jahn-Teller interactions in the excited C state of CF31 by studying the rovibrational structure of hot band transitions. l4 Infrared spectroscopy has given a relatively detailed picture of the vibrational structure in the ground state,21 but in the upper states the vibrational frequencies are incompletely known. One-photon absorption has determined frequencies for the totally symmetric modes, but the degenerate vibrations (except for the lowest-frequency mode ~~vg) have not been J. Chem. Phys. 98 (6), 15 March /93/ i 1993 American Institute of Physics 4355

2 4356 Taatjes et a/.: Spectroscopy of CF,I Rydberg states TABLE I. Polarization and rotational selection rules for two-photon absorption. C,, symmetry: Transition type Selection rules and polarization ratio P=ZdZ,l A,++Ai AK=0 AJ=O, kl, h2; P=3/2 :K=O AJ=O; P=O AI+& forbidden for identical photons A,ctE AK=*1 *2 AJ=O, *l, f2; P=3/2 EE AK=O, *I *2 AJ=O, *l, *2; P=3/2 ;\+K=O AJ=O; P=O C, symmetry: Transition type A ua AK=O, =t2 AJ=O, &l, &t2; P=3/2 :K=O AJ=O; P=O A cta AK=O, ~2 AJ=O, *l, *2; P=3/2 + AK=0 AJ=O; P=O A ++A AK=*1 AJ=O, &1,&2; P=3/2 Vhe rotational selection rules for C, symmetry are not strictly valid, since C, symmetry molecules are asymmetric tops; we assume that the reduction from C,, symmetry is small enough to allow the rotation to be approximated by symmetric top wave functions. previously measured. Fuss used pulsed COZ laser pumping and vuv absorption to probe the ground and excited state vibrational structure,22 and was able to make an estimate of the vi frequency in the C state based on hotband 5y transitions from the 2~; ground state level. In the present study we look at the 5p7r-6s Rydberg transitions in CF,I in a supersonic expansion using a massand polarization-resolved (2+ 1) REMPI detection scheme. The use of polarization to determine the symmetry of two-photon transitions was described by McClain and others,23-26 and has been applied to several molecules Expressions for the absorption from a nonsymmetric ground state has been derived by Bader and Gold;2g the relevant results for C,,, and C, symmetry groups are listed in Table I; Formost of this discussion we assume that the effective symmetry group of CF,I is C3, is both ground and excited states, but the possibility of a bent equilibrium configuration due to Jahn-Teller distortion in the degenerate electronic states is considered. The rotational selection rules, assuming symmetric-top rotational wave functions, are also given in Table I. Here we use the polarization information available in a two-photon transition to definitely assign the C and D states of CF,I to the two spin-orbit components of the 5~~6s Rydberg transition, analogous to the B and C bands in methyl iodide. In addition, we are able to detect transitions involving degenerate modes and tentatively assign frequencies for these modes in the upper states. The electronic structure of the Rydberg levels is perturbed by coupling with other electronic states, perhaps with the a-@ transitions as suggested by Robin.30 Density functional calculations are performed in order to determine the possible identity of the perturbing state(s). In addition, the vibronic effects in the excited states are shown to be more complex than can be explained by the simple linear Jahn- Teller effect which had previously been assumed to be dominant In Sec. II of this paper we describe the apparatus used in these experiments. Section III presents the results, and in Sec. IV there is a discussion of several new questions which are raised by the (2 + 1) spectroscopy of this molecule. In Sec. V some concluding statements are made concerning the electronic and vibronic structure of CF,I as seen in this work. II. EXPERIMENT The (2 + 1) REMPI spectrum of CFsI was measured in Amsterdam using a skimmed pulsed supersonic molec- I n hollow cathode lamp U SHG I I \ n gated Integrator/ (_>,I A/D v boxca; avemrage; ~~ molecular beam/ TOF detector beam dump /power meter FIG. 1. Schematic diagram of the experimental apparatus.

3 Taatjes et al.: Spectroscopy of CF,I Rydberg states 4357 ular beam of CF,I seeded in a He carrier. A schematic of the experimental apparatus is shown in Fig. 1. The frequency-doubled, focused output of a Nd:YAG;pumped dye laser crosses the molecular beam, and ions produced via a (2+ 1) REMPI process are detected and mass analyzed by a time-of-flight mass spectrometer. The improved signal-to-noise ratio over the earlier preliminary experimerits is mostly due to increased laser power, narrower frequency width, and slightly improved rotational cooling in the Amsterdam machine. The molecular beam is generated by a pulsed solenoid valve with a nozzle diameter of 0.3 mm, and passes through a 5 mm diameter skimmer after a distance of 7 cm. The opening time of the valve was varied to produce gas pulses of between 100 and 500,us FHM. A further 5 cm downstream from the skimmer, the beam crosses through the extraction region of a 20 cm flight path time-of-flight mass spectrometer. The CF,I is seeded as a 2% or 10% mixture in helium, and backing pressures used ranged between l-2.5 bar. Most of the spectra reported here were a) I I I I I I I I E,,, Ill 0, 1 Es/z 1210,0 tx 50 (4 b) Two-photon energy (cm- ) I I I I I 1 I r Two-photon energy (cm- ) FIG. 2. (2+ 1) REMPI spectrum of CFJI, taken between and cm-, monitoring CF,I+. The scaling is changed from (a) to (b) to (c) in approximately the ratio 1:2:4, in order to more clearly show the structure at the higher energies.

4 4358 Taatjes et al.: Spectroscopy of CF,I Rydberg states I I I I I I I I- G?.z (d i 5840 I I I Two-photon energy (cm- 1 FIG. 2. (Continued.) taken with a 2% mixture at a stagnation pressure of 1.4 bar. The base pressure in the detection chamber was 1 x 10B6 mbar, and pressures in the detection chamber during an experiment were between 2 and 7~ 1O-6 mbar, depending on the duration of the gas pulse. To produce the necessary uv light of about 350 and 320 nm an injection-seeded pulsed Nd:YAG with a 10 Hz repetition rate is used to pump a dye laser. The dye laser output (wavelength around 700 and 640 nm, LDS698 and DCM dye, respectively) is frequency doubled in a KDP crystal. A homebuilt autotracker system angle tunes the KDP crystal while the dye laser is scanned. The uv power is monitored by a power meter after exiting the vacuum chamber, and by a photodiode which intercepts the reflection from the front window of the chamber. Pulse energies used in the experiment could be varied between 0.5 and 15 mj in a 5 ns pulse. The ( 1 cm diam. ) laser beam was focused into the interaction region by a 30 cm focal length fused silica lens. The wavelength of the dye laser was determined by calibrating the laser grating to several neon lines in a hollow cathode discharge lamp, using scattered light about 10 cm from the beam. The uncertainty in the absolute wavelength obtained by this method is ~0.5 cm- and is determined mostly by day-to-day variations of temperature and atmospheric pressure. The relative wavelength uncertainty within a single scan is much smaller, as measured by the distance between Ne lines in the discharge lamp. This error is less than the error introduced by uncertainties in the peak positions due to the unresolved rotational structure. A second source of absolute wavelength calibration is offered by the appearance of several multiphoton resonance lines of atomic iodine in the CFsI spectrum. Comparison of these lines with the known energies of the iodine transi- tions3* agreed with the calibrated grating to within 0.5 cm - in all cases. The dye laser bandwidth is specified at 0.07 cm-t for operation with Rhodamine dye; we determined the bandwidth in the wavelength region of this experiment to be ~0.1 cm-, corresponding to a 0.2 cm- FHM in the 2 uv photon energy. The molecular beam and the laser beam cross in the extraction region of a standard design time-of-flight spectrometer. An ion detector (particle multiplier) is positioned at the end of a 20 cm drift tube. The mass resolution of the system (-20) is more than sufficient for a clear distinction among CF31f, If, and CFZ, the ions of interest in this study. The particle multiplier signal is amplified by 10 or 100 times by a fast current amplifier and is fed into a gated integrator/boxcar averager. The amplified particle multiplier signal is processed by between one and four boxcar amplifiers to monitor several masses and polarization signals simultaneously. The boxcar outputs are transferred, along with etalon or power measurements, to a microcomputer using a 1Zbit A/D converter and are stored for analysis. In order to measure the polarization behavior of the (2+ 1) spectrum, the polarization of the uv light was changed from linear to circular with an electro-optic modulator (EOM; a Pockels cell from a Nd:YAG laser Q-switch was used). The retardance through the Pockels cell can be varied by changing the voltage applied to the cell s birefringent crystal (KD*P). The /2/2 voltage for a given wavelength is determined by minimizing the reflection of a beam off of a piece of quartz tilted at Brewsters angle. Because of the linearity of the Pockels effect, the i1/4 voltage can then be directly calculated. The high voltage to the EOM is triggered at half the laser repetition rate,

5 Taatjes et al: Spectroscopy of CFBl Rydberg states 4359 TABLE II. Vibronic transitions to the *Ej12 ion core states of CFsI. TABLE II. (Continued.) Frequency (relative Relative to *Es&] origin) Assignment8 strengthb Comments : %,M@ $ ~';zplfjb 456 5&4,) : 2 3:6f 6: 3: G 3;6; 3;6; $&rz13b41 316; 6: 3;6;+3; 6: 61M2Jm +3t GM) 5: 0: 3: 3;6;? 3:( +3;6;?) 3;6; ( +3:?) 3:5; 60,) 3: C*Q2PlLi or 2%) ; 2;6: 3;6; 3;6,$?) {or 6&3:6:} 3: 5:(E) { E,,[2] 1636 or 2;3;6;) 2;3$ 2:3;6f 2;3;+2;6; V V V V m m VS S S V V V V m m w-m 656 2;6; ti 5i ;3f6; 1;6f( 1;3f) 1;3f( 1;6;) 1;6: I Al Cl63:6:1 V vs V- V w-m vs V V Oven spectrum Oven spectrum Oven spectrum Oven spectrum Oven spectrum Oven spectrum Frequency (relative Relative to 2E3,1[ l] origin) Assignmenta strengthb Comments w-m 1282 {&) I w-m 1287 C4:(A,)l w-m FHM- 1 cm-, linearly polarized 1516 V 1638 gk5;1 00 Unless otherwise indicated, transitions are to the Ejj2[ l] electronic state. Assignments within brackets {} are uncertain; many of these are discussed in the text. bintensities relative to the 2E3,2[1] origin intensity. Intensity scale as a fraction of origin intensity: vs.> 0.2; s>o.l; m> lo-*; w> 10m3; V < Oven spectrum Broad ( > 10 cm- ) Oven spectrum thereby switching the polarization from linear to circular Broad on alternate laser shots. The change in signal between the two polarizations is directly obtained by a differential gated integrator/boxcar unit at the same time as another boxcar unit acquires the Oven spectrum average signal. The polarization ratio of interest is the ratio Oven spectrum of the signal for circular polarization (a) to that for linear Broad polarization (11 ). The signals obtained are the difference Oven spectrum (a- 11 ) and the average ((T+ [I J/2. However, in the present experiments we are mainly interested in the Broad changes in polarization behavior among different lines in Oven spectrum FHM- 1 cm-, the spectrum, and the difference signal (in fact, just its linearly polarized sign) is sufficient for our purposes. The linear polarization of the uv beam was purified before entering the EOM by passing it through a prism polarizer of extinction ratio > 10m3. The quality of the emerging polarization was seen to be extremely sensitive to the alignment of the EOM in the laser beam. Typically the EOM was aligned by setting the voltage at its /2/2 value Broad Higher power, and checking that the polarization remained linear on expossibly spurious iting, and in some cases by directly maximizing the polar- FHM- 1 cm-, ization difference signal on a CF,I transition. However, the linearly polarized purity of the circular polarization is difficult to measure, and several sources of error, chief among them beam walk and alignment of the EOM, diminish the likely purity of Reversed rotational line shape the circular polarization. The Nijmegen apparatus used in the measurement of hot band transitions will be described in more detail else- Very broad where;32 it is similar to the Amsterdam machine, except (>20 cm- ) that the expansion takes place through an oven in order to Higher power, possibly spurious populate higher vibrational levels of the electronic ground state. The oven was built after a design of Colson and co-workers,33 and consists of a small aluminum oxide pipe with an inner diameter of 1 mm around which is wound a. tungsten wire. The wire is resistively heated, allowing tem- Broad peratures of up to about 1100 C to be reached. The tube is attached to the outlet of a pulsed nozzle, separated by a cooling element to avoid burning the nozzle. The gas pulse entering the small heated tube via the pulsed nozzle is heated during the few ps transit time. In the subsequent expansion into the vacuum, collisional cooling takes place. FHM- 1 cm-, linearly polarized J. Chem. Phys., Vol. 96, No. 6, 15 March 1993 Since cross sections for rotational relaxation are much larger than for vibrational relaxation, the rotational cool-

6 4360 Taatjes et al.: Spectroscopy of CF,I Rydberg states * -* pvzq...l ioo 57ioo 5i ioo Two-photon Energy (cm- ) FIG. 3. REMPI spectrum taken with oven expansion. The power through the heating coils is indicated for each spectrum. Spectra have been displaced vertically for clariiy. ing is much more effective than the vibrational cooling, and rotationally cold spectra of vibrational hotband transitions can be examined. For the hot spectra, a stagnation pressure of 1250 Torr of 5% CF,I seeded in argon is maintained behind the nozzle. The uv power used for detection is 3 mj per pulse and is focused by a spherical lens with a focal length of 25 cm. The ions are detected at the end of a time of flight mass spectrometer by two cascaded microchannel plates, and the signal is collected by a gated integrator/boxcar averager. III. RESULTS A. The 2E3,2 ion core states The (2+ 1) REMPI spectrum of this band was first reported by van den Hoek et al., who measured a number of vibrational transitions, mostly involving the totally symmetric vibrations.5 No attempt was made to resolve the electronic assignment question and the band was simply referred to as the C band in accordance with numeration conventions. e have re-examined the REMPI spectrum in the cm- region using polarization analysis to definitely assign the character of the resonant intermediate state. e have thereby been able to assign transi-* tions arising from excitation of nontotally symmetric modes, as well as transitions to the lower lying fin=2 component, the 2E,,,[2] electronic state. 1. The 2E3,d1] state Figure 2 shows the (2+ 1) REMPI spectrum of CF,I taken in a supersonic expansion at two-photon energies between and cm-. Most of the vibrational progressions agree with those measured by van den Hoek et a1.,5 although several new lines have been noted. The observed transitions, with suggested assignments, are listed in Table II. The width of the transitions ( -5 cm- ) is believed to be mainly due to unresolved rotational structure which is present even at the low rotational temperatures of the supersonic expansion. Several much narrower features ( cm- ) are also observed (marked with an asterisk in Fig. 2). These features are due to parallel-type (symmetry A,-At) transitions involving excitation of degenerate vibrational modes in the resonant intermediate state and are discussed further below. The beginnings of progressions in the totally symmetric modes are measured, which correspond to the earlier measurements of Sutcliffe and alsh. The oven spectra of this state are shown in Fig. 3; lines from these spectra have been used to establish assignments in the cold spectrum and also allow determination of new lines involving higher vibrational excitation in both ground and excited states. The intensities of the various transitions are of course sensitive to experimental conditions such as beam alignment and focussing and laser powers; for this reason the relative intensities in Table II are given only in a somewhat qualitative sense, and the intensities for the oven spectra are omitted altogether. The polarization analysis of many of the lines in the spectrum has been carried out. The polarization ratio for the origin band of an AI-E transition is expected to be P=Id1ll =3/2, whereas for Al-Al transition a Qr branch is expected with enhanced linear polarization (111 > I,).24,25 The polarization ratios we measure are slightly different than the theoretical predictions. Figure 4 shows

7 Taatjes et al.: Spectroscopy of CFBl Rydberg states 4361 I I I I I I I Relative two-photon energy (cm- ) FIG. 4. Polarization behavior of the *&[l] origin. The dots are the polarization difference signal, (1,--I,, ), times three; the solid line is the average signal, (I,+111 )/2. The size of the difference signal relative to the average corresponds to a polarization ratio of P=7/5. the polarization-resolved scan over the origin of the 2,!?S,2[1] state. A ratio of P=3/2 for circular vs linear excitation would correspond to a ratio of 2f5 for the difference to the average polarization signal; as the figure shows the measured ratio is nearly 2/5, but slightly below ( - l/3). The ratio is sensitive to the purity of the circular polarization and also to any possible polarization dependence of the ionization step. The polarization behavior of the narrower features is shown in Fig. 5. The striking reversed polarization dependence (11, > 1,) indicates an AI-A, transition, i.e., excitation to a totally symmetric intermediate state. In the E symmetry electronic state this is only possible with excitation of a degenerate vibrational mode. The degenerate vibrations in CFJ are of e symmetry in the C,, point group: vibronic coupling within the E symmetry electronic state produces states of E@e=A, +A2+E symmetries. Excitation to the A, component of this state produces a with preferred linear polarization. The linear to circular ratio of this excitation can be used to determine the nature of the virtual state involved in the two-photon excitation step.34,35 e have not attempted a quantitative measure of the 6 5 i0 i5 i0 i5 Relative two-photon energy (cm- ) FIG. 5. Polarization behavior of the narrow lines in the *L&[l] spectrum. The bold line is the negative of the polarization difference signal, i.e. (11 --I,), scaled so that the peak intensity at the narrow transition matches the average signal. Transitions with ambiguous assignments are simply labeled by their energy relative to the 2E3,2[1] origin. virtual state branching ratio, because such measurements are extremely sensitive to imperfections in the polarization state of the interrogating light. An estimate of the polarization ratio P can also be made based on the strength of branch relative to the other branches in the linearly polarized spectrum.34 This method is somewhat less susceptible to experimental uncertainty in the polarization, since the linear polarization state is very well characterized. The spectra in Fig. 5 have been scaled so that the peak heights of the sharp parallel transitions coincide. One can see that in all cases except the line at 1287 cm- there appears to be a small component of the excitation which is not completely linearly polarized, as shown by the slightly broader profile of the total signal as compared to the polarization difference signal. Part of this broader profile could be due to excitation to an E symmetry level at close to the same energy. A completely linearly polarized transition, exhibiting only would indicate a totally symmetric virtual level. The strong linear polarization of the parallel transitions, coupled with the small contribu- i0

8 4362 Taatjes et al: Spectroscopy of CFBl Rydberg states Relative two-photon energy (cm- ) Relative two-photon energy (cm? FIG. 6. (a) Fragmentation spectrum of the 2Ej,2[1]3h and 6: transitions. The fragmentation ratios. as measured bv I+/CFJ+, are 3fi-0.4 and 6&0.1. (b) Fragmentatibn spectrum of the 3; and6!.hotbaids, 52 and 40 cm- red of the 2E3,2[1] origin. The I+CF,I+ ratio of both transitions is approximately that of the 6; line. tion of AJ#O branches suggests that the two-photon transition takes place mainly via a symmetric virtual state. The narrowness of the reverse polarized lines is due to the 11 vsl character of the excitation rather than to any change in the lifetime of the intermediate state. The asymmetric shape of the broader lines indicates unresolved rotational structure which is not washed out by whatever lifetime broadening is present in these states. The degenerate vibrations in CF$ are expected to exhibit a Jahn-Teller-type vibronic splitting in the E-symmetry states. Evidence for this was seen by Herzberg in the form of a splitting of hot bands involving one quantum of ve in the excited state He observed three lines at -40, -48, and -56 cm- from the origin, which were assigned to the 6:(E), 6i(A1+A,), and 3; transitions, respectively. However, in the (2 + 1) jet-cooled spectrum no evidence was seen for the splitting; lines at -40 and -52 cm- were assigned to the 6: and the 3: transitions.5 The mass-resolved fragmentation spectra allow a more thorough investigation of this discrepancy. Figure 6 (a) shows the fragmentation spectrum of the 6: and 3; bands in the 2E3/2[l] state. There is a clear difference between the fragmentation behavior for one quantum of v3 (C-I stretch) and that for one quantum of v,j (F,-C-I bend), even though the two intermediate levels lie at very nearly the same total energy. e cannot determine with certainty whether this difference is due to increased (mode-specific) vibrational predissociation of either the intermediate neutral or the ion, or due to a change in Franck-Condon factors for absorption of a fourth photon to a dissociative ion state, or to some other effect. It does seem likely that the dissociation takes place from the ion (see Sec. IV B). From Fig. 6(a) it can be seen that the I+/CF31+ ratio of the 3: transition is four times that of the 6; transition. The fragmentation ratios (measured by CF$/C!F,I+ or I /CF,I+ signals) of the other parallel (A 1-A 1 ) transitions do not exhibit any systematic deviation from those of the broader (Al-E) lines at similar final state energies. Figure 6(b) shows the fragmentation spectrum of the hot bands 40 and 52 cm- to lower energy from the 2E3/2[l] origin. Both bands exhibit a fragmentation ratio (I+/CF31+) approximately equal to that observed for the 6: band. This is an indication that these two bands are indeed two components of a Jahn-Teller split 6: hot band. The - 52 cm- band has a slightly ( -20%) higher fragmentation ratio than the -40 cm- band, suggesting that the 3; excitation is overlapped by the (stronger) 6; transition. No sharp Griation in fragmentation could be detected, to within experimental error ( f lo%), across the -52 cm- band, so we cannot resolve the frequencies of the individual 3: and 6; components. The frequency of mode 5 (nonsymmetric C-F bending) in the excited state was estimated by Fuss to be 458 f 15 cm- on the basis of hotband transitions which he assigned as arising from the 2~: ( Zk =2) level of the ground electronic state.22 (H ere Zk is the vibrational angular momentum quantum number in the ground state.13) The onephoton selection rules for j, the vibronic angular momentum in the excited state, are j,-i,= f 1/2;13 it should be noted that in the case of significant Jahn-Teller activity in vi, the ( jk=5/2) Esymmetry component of 2~; will lie considerably lower in energy than twice the vi fundamental. Therefore, transitions to this level do not necessarily give a reliable indication of the fundamental frequency, certainly not without additional information. e observe a linearly polarized feature at 456 cm- which could be assigned to the A, fundamental 5;. However, this line could just as easily be assigned to 6i(A,) or 3:6A(A,). One of the lines at 522 and 527 cm- could possibly be assigned as the E component of 5h higher energy transitions at 739 and 750 cm- could possibly be combinations with vg or v3. The ground state frequency of mode 5 is 540 cm-, so vi- 525 cm- would give rise to hotbands shifted about 15 cm - to the red. Such lines are indeed observed in the oven spectrum built both on the origin and the 3; peaks (see Fig. 3 ). These lines had also been observed in absorption by Sutcliffe and alsh, who declined to assign them to 5; transitions because of their relatively high intensity. l5 However, an alternative explanation of the bands as rota-

9 Taatjes et a/: Spectroscopy of CF,I Rydberg states 4363 h / I 1 I / I Relative two-photon energy (cm- ) FIG. 7. Detail of the peaks 522 and 527 cm- from the 2.E3,2[1] origin. Although all the vibronic transitions in the 2&2[1] spectrum exhibit a rotational profile which rises more sharply on the high-energy side, the 527 cm- line rises more sharply on the low-energy side and is shaded to the blue. The 2E3,[2] lines all show a slight blue shading, but the 527 cm- line is exceptionally dramatic. tional structure is in our judgement less likely, because of rotational cooling in the oven expansion and a lack of similar structures on sequence transitions. e therefore assign one of the 522 and 527 cm- lines to the E-symmetry fundamental 5;. The other is possibly a transition built on the 2E3,2[2] state, which may interact via a Fermi resonance with the Iv, level of the 2E3,2[1] state. The Jahn- Teller activity of v5 and possible interactions with levels arising from the 2E3,2[2] state are discussed below (Sec. III A 2) and in Sec. IV. The frequency of mode 4 (asymmetric C-F stretch) is somewhat confusing. A line at 1287 cm- from the origin of the 2E3,2[1] state exhibits the narrow rotational profile and the linear polarization characteristic of an excitation involving a degenerate vibrational mode. The neighboring peak at 1282 cm- was misassigned in previous work to g, but is clearly the E symmetry counterpart to the 1287 cm- line. The 2: feature expected at about 1350 cm-, which has been observed in one-photon absorption 3 5 as part of a progression stretching to 2:, is not observed in the (2f 1) REMPI spectrum, for unknown reasons. Both these lines (1282 and 1287 cm- ) must involve some excitation of v,+. In the ground-state, the frequency of mode 4 is only 1187 cm- ; if we assign the 1287 cm- feature to the (A1)v4 fundamental we are thus faced with the unsettling situation of having one frequency ( v4) increase in the upper state while all the other frequencies decrease. In particular, the symmetric C-F stretch (vl) decreases by more than 100 cm-. It is possible that the fundamental excitation of mode 4 does not appear in the spectrum, or is buried under another peak, and the 1287 cm- peak is a combination band involving v4 2. The 2E3,d2] state Several lines appear to the red of the 2E3/2[l] origin which cannot be assigned to any reasonable sequence bands arising from the 2E3/2[l] state. The lowest energy of these lines appears at cm- from the 2E3,2[ 1] origin. Polarizatiofi analysis shows that this state is preferentially excited by circular polarization, indicating E symmetry of the resonant intermediate state. Experiments involving CO, laser pumping of CF31 have shown that this transition is depleted by CO, radiation tuned to vibrational transitions out of the CF,I ground state.32 The oven spectrum shows no increase of the transition with temperature, providing further evidence against assignment as a hot band. This leads us to assign this line as the origin of the transition to the 2E3&] state. e also assign a 6; line built on this origin with a frequency vi=223 *4 cm-, similar to the frequency in the 2E3/2[l] state. At 239 cm- higher energy than the 2E3,.J2] 6: transition, there is another feature, which we assign as 3;6& Other lines which resist assignment to 2E3/2[l] transitions are perhaps also due to excitations out of this electronic state. In addition, the lines at 522 and 527 cm- relative to the 2E3/2[l] origin exhibit an unusual shape which may indicate interaction between two accidentally neardegenerate levels, one from each of the two components of the 2E3/2 ion core excitation. Figure 7 shows these two lines in more detail. All the 2E3,2[ 1] transitions have a rotational envelope which rises sharply on the blue side and is shaded to the red, but the 527 cm- line shows the opposite behavior, rising more sharply on the red side. The rotational envelopes of the 2E3,&Z] transitions appear nearly symmetric, with only a slight shading to the blue. The exaggerated

10 4364 Taatjes et a/.: Spectroscopy of CF31 Rydberg states behavior of the 527 cm- line could indicate that it is perturbed by a nearby state, and that the 522 and 527 cm - lines are a Fermi-resonant pair. Both lines are preferentially excited by circular polarization, indicating that the final levels have the same symmetry. The information about the 2E3,2[2] state available in the (2+ 1) spectrum is insufficient to provide more than a guess at vibrational frequencies, but a level with 1156 cm- vibrational energy (the difference between the 527 cm- line and the 2E3,2[2] origin) could, for example, be (lv;+lv;) or (l~~+lvj+lv~). These assignments can be supported by noting that the difference between the line at 288 cm- (917 cm- from the 2Es,2[2] origin) and the 527 cm- line matches the 239 cm- energy difference between the 2E,,2[2] 6: and the next 2Es,2[2] transition (2-f33,2Pl 3&i). B. The *E,,* ion core states The 2E1,2 ion core states of CF,I appear at around cm- and they were first investigated by Sutcliffe and alsh-l5 They noted an increased broadening of the lines relative to the cm- bands, and assigned some vibrational progressions and sequence bands. The presence of both spin components of the state was suggested, but no definite assignments could be made. The earlier (2+ 1) REMPI work on CFsI by van den Hoek et al. noted two bands at and cm- I, which were assigned as the origin and 3: transitions of the D state,5 but because of low signal to noise (S/N) these states were not investigated further. The vibrational structure of the transitions to these states is far less extensively resolvable than that of the lower spin-orbit state, mainly because of the broadening and lower S/N of the transitions. The fragmentation ratio in this state is far higher than that in the lower spinorbit state as well; the If and CFZ channels dominate over the parent ion channel by a factor of For this reason the polarization spectra for the 2E1,2 ion core states were acquired on the CFZ mass channel. In the spectrum of CH,I, the dominant (2-t 1) REMPI transition is to the 2El,2[O] state,7 which is a relatively weak transition in one-photon absorption. This state has A, electronic symmetry, and the origin band could be expected to exhibit branch with preferred linear polarization. In the (2+ 1) REMPI spectrum of CF,I the 2E1,2[O] excitation is weaker than that to the 2E1,2[1] state. In the present investigation we use the polarization behavior to ascertain the identity of the lines in the 2E1,2 ion core manifold. The application of the twophoton polarization expression to (2+ 1) REMPI is not necessarily straightforward because of the possibility of additional polarization dependence in the ionization step.27y36 The detection of ions other than the parent ion, which can be necessary in cases where fragmentation is extreme, is further complicated by the additional absorption of several photons, sometimes in a process of unknown order, all of which can contribute polarization effects to the measured signal. Usually, one relies on reorientation of the molecule on a rapid timescale (i.e., between two-photon absorption and the ionizing step) or saturation of the ionizing transi- 2- l- 52 *.gl I ' l e.* 0 l * ln(power(mi/pulse)) FIG. 8. Power dependence of the CF$ signal from the 2E3,2[l] origin. In the intermediate power region the dependence is approximately quadratic. tion, either of which will remove the polarization dependence of the final step. In a supersonic expansion, the collisional reorientation of the molecule is essentially nonexistent. Saturation of the ionizing step (but not the two-photon step) will give a quadratic dependence on power. The power dependence of the CF$ signal from the 2E1,2 ion core states is not well fit by a simple power dependence. Figure 8 shows the power dependence of the 2E,,2[1] origin; the transition begins at low power with an approximately cubic dependence on the incident laser energy, then gradually turns over to a quadratic dependence, before everything saturates at very high power. The polarization measurements for both spin-orbit states are all taken in the region where the power dependence is approximately quadratic, and it is assumed that the polarization dependence reflects the character of the two-photon resonant intermediate levels. The (2+ 1) spectrum of the 2E,,2 ion core states is shown in Fig. 9. The lower 2El,2[O] state spectrum in Fig. 9(a) was taken on the CF3f mass channel, but the stronger spectrum of the 2E1,2[1] [Fig. 9 (b)] was taken using the parent ion mass; note the difference in signal to noise. The spectra taken on CFf and CF,I+ channels show identical features. Polarization measurements clearly show the presence of two electronic states with vibrational transitions built on them. Observed lines and suggested assignments are given in Table III. The lines in the 2E1,2 ion core state spectra are clearly much broader ( cm- FHM) than in the lower spin-orbit state (Fig. 2), and the linearly polarized lines are, in contrast to the 2E3,2 ion core states, not any narrower than the circularly polarized perpendicular bands. It is therefore probable that the broadening is due to coupling to some dissociative state at the resonant intermediate level. Such an explanation is reinforced by the presence of the large blob at about cm- which may mark the crossing area to the dissociative curve. Speculations concerning the nature of this dissociative state are discussed in Sec. IV. Vibrational transitions involving the totally symmetric modes are seen, allowing determination of vibrational fre-

11 Taatjes et al.: Spectroscopy of CF,I Rydberg states 4365 a) CT.% 5 d f-i H 2.E! Ei 61.5 m Two-photon energy (cm- ) b) I I I I I I Two-photon energy (cm ) FIG. 9. (a) Spectrum of the 2E1,2[O] state up to and including the origin of the 2E1,2[1] state, taken on the CF$ mass channel. (b) Continuation of the spectrum, taken on the CFJI+ mass channel. The signal to noise is reduced from (a) because the extensive fragmentation of the CFJ yields more signal on the fragment ion. Spectra taken on both mass channels exhibit similar features. The three-photon energy lies above the dissociation energy of CF,I+-+CF$-+I. quencies Y; and V; in both states, and V$ in the 2E,,2[O] state. The frequencies differ between the two components and both differ from the 2E3,2 states. No fundamental transitions to nontotally symmetric modes in the excited state are measured; however, hot band transitions to the red of the 2E1,2[ l] origin allow an estimate to be made of vh in the excited state. Suggested vibrational frequencies in the CF,I states investigated are reported in Table IV. The circularly polarized line at cm- offers some problems in assigning. It is clearly unlikely to be a hot band built on the 2E1i2[1] state because of the large ( cm- ) redshift from the 2E1,2[1] origin. It is thus built on the 2E1,2[O] state, but the 666 cm- difference does not match with any degenerate vibrational frequency. It is therefore most likely some combination involving a degenerate mode, and is assigned to the $6; transition. More details about the vi-

12 4366 Taatjes et ab: Spectroscopy of CF,I Rydberg states TABLE III. Vibronic transitions to the 2E,,, ion core states. Frequency (cm- ) Assignment E1,2[0] origin %dv~ E&&% C2-%2D12A *~d~ 6 b &2[113; C2%2[llCI E,,2[1] origin %2[01 1: %~[11% %2Pl 1:3; [?I E,,2[ I 1% %2P11:, [?I16 Linear=P< 1; circular=p-3/2. bduschinsky-effect allowed. Narrow lines (FHM-5 cm- ). Polarization* Linear Linear Linear Linear Circular Circular Circular Circular Linear Circular Linear Circular Linear Circular bronic couplings in these states are discussed in Sec. IV. In addition, ions were observed at the mass of 12 under some expansion conditions. It was found that by reducing the backing pressure in the expansion that this signal could be completely eliminated. ith higher backing pressures, the 12 and the If signals would grow dramatically, while the CFZ signal remained nearly unchanged. Possible contributions from I, impurities were ruled out, indicating that the source of these ions is probably a CF,I cluster species. Neutral I2 has been observed by Donaldson and co-workers from dissociative states of CH31 dimers,37 and it is likely that the CF,I dimer species is the major contributor to the 1; signal observed in our experiments. The reaction of I+ with CF31 can also produce I:, but this channel is energetically less favorable than the creation of CF,f and has been shown to be a minor pathway at 298 K.38 No detailed investigation of the cluster signal was made in the current work, and the expansion conditions were chosen in all experiments to minimize these signals. IV. DISCUSSION A. Electronic structure The electronic transitions reported here can be definitely assigned, on the basis of the symmetries and splittings of the states, to be the components of the 5p+6s Rydberg excitation. The splitting of the ion core states, cm-, is quite close to the splitting in the bare ioni The Rydberg series assignments dating back to Sutcliffe and alsh are supported. There are also transitions not previously observed in this molecule, which seem to indicate mixing with another dissociative state near the 2E,,2 ion core states. In order to gain insight into the nature of the relevant electronic states of CF,I, density functional (Hartree- Fock-Slater) calculations were carried out on this molecule using the Amsterdam Molecular Structure Program ( AMOL).39 Relativistic effects, which are extremely important in iodine-containing compounds, were taken into account by incorporating the scalar relativistic corrections [mass-velocity, Darwin, and relativistic (core) potential operators] into the SCF calculation, while the spin-orbit interaction was added in the end through perturbation theory The geometry of the molecule was taken from Ref. 42. A fairly large valence basis was used (quadruple zeta STO s plus polarization functions), while the I[Kr] and C,F[He] core was kept frozen. The multiplet energies of the various excitations were calculated using a diagonal sum procedure which relates the multiplets to the energies of individual determinantal wave functions. The highest occupied orbital in CF31 was found to be the iodine lone pair (n) orbital which is highly spin-orbit split (calculated As,= 5300 cm-, experimentally about 5030 cm- ), followed by the C-I bonding orbital (a). The C-F bonding orbitals and the fluorine lone pairs were all at much lower energies. The lowest unoccupied orbital has mainly C-I antibonding character ((T*), while the next unoccupied orbital is indeed a diffuse one of mostly iodine 6s character. The lowest-energy excitations arise from a promotion of a single electron from the two highest occu- TABLE IV. Vibrational frequencies for CF,I. Electronic state VI v2 v3 V4 VS v6 X( A,Y &,2[21 {917}b {69Qb 239b 223 2E3,2[ ,1287= 525 U-0, 223~~ C456ld 211 (A,,E?) 2E1,2Pl 945 {630*2O)e 666-v, 666-v, 2E,,2[ {218}g Reference 21. bfrequency determination depends on assignment of combination band, see Sec. III A 2. See Sets. III A 1 and IV B 2. dambiguous assignment; see Sets. IV B 1 and IV B 2. Broad feature at -610 cm- assigned as fundamental; combination band 1;2: at 1594 cm- indicates higher frequency, ~~-650 cm-. Modes appear only in combination band; see Sec. IV B 3. staken from hot-band transition.

13 Taatjes et al: Spectroscopy of CF,I Rydberg states 4367 TABLE V. Calculated transition energies for CF31 (cm- ). Transition type Excited state n-d E&l J4/2[11 E,,2[01 -%2Ul A,[O+ 11 A,101 n-.6s %2Pl -%zul -%2[11 Theory (present work) References 16 and 30. bcenter of broad A band containing all four n-ti Present work. dbroad feature. transitions. Experiment a.b Csd c pied orbitals to these two empty orbitals; the corresponding vertical excitation energies are collected in Table V. The first excitation arises from an n + ti transition and is split into four components by the spin-orbit and exchange interactions. The resultant excited states are dissociative along the C-I stretch coordinate due to the occupation of the C-I antibonding ti and can be identified as the broad dissociative A state found experimentally. The next excitation arises from the (T-+ df: transition. Since both cr and 8 are localized in the C-I bonding region, a large exchange splitting is found between the resulting 3A1 and ial states. Although the 3A1 can split into 3A1[1] and 3Al[O] components, this is a second-order spin-orbit effect, which has not been calculated and is expected to be small. The 3A1 state, calculated at cm-, can be identified with the B state observed by Robin. l6 Its essentially triplet character explains why it is very weak and has not been seen in the RFMPI spectrum.5 The Ai component of the u+o* transition is much higher in energy and we will see that this dissociative state might perturb other states in the region. The n+6s transition gives rise to four excited states, which are split into two groups by the large ion core spinorbit splitting; each spin-orbit component is further split by a relatively small exchange splitting. These states can clearly be identified with the C and D states that form the main subject of this paper. It is seen in Table V that the calculated character of these four states completely agrees with those derived from the polarization measurements in the REMPI spectrum, and there is also near-quantitative agreement for the spin-orbit and exchange splittings. The relative order of the components of the C and D states can in fact be rationalized if one realizes that to a first approximation (no second-order spin-orbit interaction) the 2E3&!] and the 2E1,2[O] are pure 3E states, while the 2E3,2[11 and 2&(2[11 are equal mixtures (via the spin-orbit interaction) of E and E. Since the exchange interaction tends to stabilize triplets with respect to singlets, the order of the exchange components is the expected one. The dissociative o+tilal state is calculated to be rather close to the 2E1j2 ion core states of the n+6s tran- sition and might easily cross these states along the C-I stretch coordinate. It is therefore a prime candidate to explain the perturbations in the REMPI spectrum of the 2E1,2 ion core transitions. hile it has the right symmetry to perturb the 2E1,2[O], interaction with the 2E1,2[1] component is symmetry forbidden. The fact that the 2E1,2[1] state is nevertheless seen to be perturbed as well can possibly be rationalized by assuming a Jahn-Teller distortion of the nuclear framework which is large enough to make the perturbation allowed and appreciable. Alternatively, the 2E1,2[1] state may be perturbed by another unknown state; however, the a-+6s and the n +6p excitations, the most obvious candidates, lie much higher in energy. The possible role of CT+~ transitions in the region of Rydberg excitations has been discussed by Robin.30 The valence o* states can mix with nearby Rydberg levels of the same symmetry. The o+ ti transition on the C-I bond is a transition between states of Al symmetry in the C3, group. As a consequence, the two-photon transition to this state would be expected to exhibit a linear > circular polarization dependence, as the broad feature in the REMPI spectrum indeed does. However, as was mentioned above, the C,, symmetry is only approximate in the degenerate electronic states because of the Jahn-Teller vibronic interaction. For strong interaction, the equilibrium position will be significantly away from the C3, geometry, and the symmetry will be lowered toward C, Polarization ratios reduced from the theoretical maximum have been observed in two-photon laser-induced fluorescence for molecules where vibronic mixing between states of differing symmetries is important.28 Similar vibronic mixing may explain the polarization behavior of the Rydberg levels of CF,I studied in the present work, i.e., mixing of the E symmetry levels with levels of A symmetry via nontotally symmetric vibrations. The polarization ratios measured for the origins of the three (nominally) E symmetry bands are nearly the theoretical 3/2 value ( 1.3 f 0.2)) but the inability to obtain measurements which reflected the maximum could indicate band mixing. -The fragmentation patterns of the various vibronic transitions in the 2E3,2[ l] state, in particular the difference between Iv, and Iv3 excitation, shed some light on the intersection of dissociative states in the ion. The energy of three photons of around 350 nm, used in the (2+ 1) REMPI spectroscopy of the 2E3,2 ion core Rydberg states is not quite sufficient to produce CF,I+ above its dissociation threshold.43 If it is assumed that the dissociation is due to absorption of a fourth photon by the parent ion, then the difference in the fragmentation ratios must arise from a changed Franck-Condon overlap between the ionic ground state and the (pre) dissociative excited ion state. Differing I/I* branching ratios for dissociation on the CX31 neutral dissociative A surface have been noted for different vibrational because of a change in the coupling at a conical intersection.4s The mechanisms of fragmentation in the present experiment are somewhat ill-characterized due to the number of competing pathways available in the REMPI process. Fragment ions can be produced by essentially two different

14 4368 Taatjes et a/.: Spectroscopy of CF,I Rydberg states processes; fragmentation of parent ions ( ladder mechanism) or ionization of neutral fragments ( ladder switching ). The two processes are distinguishable, for example, by their power dependences. Ionization of neutral I atoms produced from dissociation of the resonant intermediate level requires three additional photons, or a total of five vs the four needed for fragmentation of the ion. However, the power dependences are complicated by partial saturation of any or all of the various photon absorption steps involved, and can be very difficult to unambiguously interpret for multiple pathways. In the 2Es,2[1] state, the fragmentation ratio is independent of power, to within experimental error, and the polarization behavior of the fragments and the parent ion signals is the same. ork by Szaflarski and El-Sayed on the multiphoton ionization dissociation of CHsI at 266 nm (resonant with the dissociative A state) indicates that the ladder mechanism is more important as the excitation pulselength decreases; they observed a predominance of the ladder mechanism for picosecond-pulse excitation.46 The lifetime-topulsewidth ratio in the CF,I 2Es,2[1] state REMPI is similar to that of the El-Sayed experiments, if a lower limit on the lifetime of the resonant intermediate state of 50 ps is estimated based on the width of the parallel transitions. In addition, the kinetic energy release has been measured for CF$; produced via (2+n) REMPI through the 2Es,2[1] state 1; transition,47 using a 2-dimensional ion imaging technique.48 The difference between the energy of the 2E3,2[ l] state and the asymptotic I + CF, energy is nearly 5 ev, but no CF$ was seen corresponding to a kinetic energy release over about 2 ev. The images show two distributions of CFf; some CFZ is produced via a very small kinetic energy release ( ~0.2 ev) and some in an approximately cos2 8 distribution from a higher kinetic energy release (-2 ev). The energy of the 1; transition is nearly twothirds the reported appearance potential for CF$ + I[2P3,2],43 and it is likely that the low kinetic energy CFt is produced by three photon absorption to just above the CFt +I limit. The higher energy component fits well with the energy difference between the four-photon energy and the threshold for production of CF$ +I[2P,,2], indicating that the fragmentation in this transition is mostly due to fragmentation of the parent ion. It is likely that the situation in the lower vibrational levels of the 2E3/2[l] state is similar, and that the difference in fragmentation between the 6; and 3; transitions in the present experiment arises from changes in the probability of absorption of a fourth photon by the parent ion to reach a (pre)dissociative level. B. Vibronic coupling ionic The interactions between degenerate vibrational and electronic states have a long history of study, especially in connection with the E o e Jahn-Teller effect. The 2E3,2[ l] state has been previously pointed to as a particularly good example because of the observation of splitting in the 6f band and the observation of fundamental 6; transitions. 3p14*49 The analogous bands in CH31 have been studied in considerable detai16-g 50 and offer a good point of departure for the present study. In the methyl iodide studies, the intensity of the fundamental transitions in v6 were attributed to a Herzberg-Teller intensity borrowing mech- anism, combined with Jahn-Teller intrastate vibronic mixing.6p5y The assumed dominance of Jahn-Teller interactions in the case of CF31 (Refs. 13 and 14) is apparently mainly due to the observation of the splitting of the 6: hot band transitions, and the assignment of this splitting as due to the E-A1,A2 linear Jahn-Teller interaction. It has been suggested that the role of Herzberg-Teller interstate vibronic mechanism has been unfairly underestimated in these transitions,51 and in the present investigation we have gained new information on the symmetries of the vibronic levels which shows that the situation in CF31 is more complicated than a simple linear Jahn-Teller picture would allow. Linear Jahn-Teller interaction of an e symmetry vibration in an E symmetry electronic state splits the E and Al,A2 vibronic components (in C,,), with the energy of the doubly degenerate E level raised relative to the degenerate A,, A2 pair. 52 The energy levels for a relatively small linear Jahn-Teller interaction exhibit splittings of 4Dwo, where D is the dimensionless Jahn-Teller interaction parameter, and w. is the unperturbed vibrational frequency.i3 Secondorder terms in the vibronic interaction Hamiltonian act to split the Al and A, components; these terms are not necessarily small with respect to the linear Jahn-Teller interaction terms.53 In addition, other quadratic terms can serve to couple different modes of the same (Duschinsky effect) 54 or different symmetries.55 The cross-quadratic terms which link modes of different symmetries can induce vibronic intensity and can split vibronic levels even in the absence of Jahn-Teller distortion.56 In this section we will discuss the observed vibronic interactions in CF31, concentrating in turn on the v6 interactions in the 2E3/2[1] State, the interactions of the other nonsymmetric modes in the 2E3,-JJ] state, and finally on vibronic effects in the nondegenerate 2E,,2[O] state. 1. The 2E3,d1] 8 ve rnteractron The linear Jahn-Teller interactions along v6 in the 2E3,2[1] state have been investigated by Herzberg in onephoton absoiption The one-photon absorption spectrum differs from the (2+ 1) REMPI spectrum in several respects. First, in the one-photon spectrum of the 6: band, the E vibronic component was seen, but we see only the A, component. Second, the positions of the split 61, hot band transitions are different. Finally, in the (2+1) REMPI spectrum the Al component, as assigned by polarization measurements, is seen to lie above the other component of the vg= 1 level, which is inconsistent with the view that the splitting arises from linear Jahn-Teller interactions. Interactions of y3 and vg via cross-quadratic coupling terms could account for,the splitting of the v6 = 1 level. The coupling of e and a modes in C,, can be described similarly to the treatment of D6h symmetry given by Scharf.56 In this case the components of the e mode ( vg) which correspond to E symmetry will interact with the y3= 1 level and be shifted away from it, while the A, and A2 components will

15 Taatjes et al: Spectroscopy of CF,I Rydberg states 4369 not interact. Since v3 > vg, the A components ill appear higher in energy than the E components. This explanation would require that the v3= 1 level also shift, to higher energy. Since the progression in v3 appears very harmonic, the energy shift of the v3= 1 level is certainly quite small, indicating that the cross-quadratic terms are probably not the dominant mechanism for the splitting of the vg= 1 level. If on the other hand the largest splitting is the secondorder splitting of A, and A, and the linear Jahn-Teller splitting is roughly half the AI-A2 distance, then the E component could be accidentally degenerate with the A,. The lack of a fundamental transition to the other (v=211 cm- >, A, symmetry component could be explained, since an AI-A2 transition is two-photon forbidden for photons of the same frequency. The linear Jahn-Teller splitting of approximately 6 cm- given by this explanation is only slightly smaller than the 8 cm- estimated from the roomtemperature observations, even though the nature of the splitting was not completely known in the earlier work. The 6;(E) transition was reported by Herzberg, l4 who noted the same characteristic intensity alternation and wide spacing in the K structure which Mulliken and Teller had used to establish the presence of Jahn-Teller interactions in the analogous state of CH31.6 This transition was seen in the region where the 6A(A,) transition appears in the (2+ 1) REMPI spectrum; however, in the current work we see little evidence for the E component at this energy, indicating that its intensity is much less than that of the A, component. The component of the 6: transition which is not completely linearly polarized (see Fig. 5 and the discussion in Sec.111 A 1) could be partly due to the E component of the excitation. This difference in the intensities of the one-photon and (2+ 1) spectra is not completely understood; at the very least, it raises doubts about the direct applicability to the two-photon spectrum of the calculations of the intensities of vibrational progressions which have been carried out for Jahn-Teller active systems Such intensity differences possibly contribute to the shifting of the position of the 6: hotband peaks between the one-photon and two-photon spectra. Of course, the rotational structure present in the room temperature experiments could also shift the intensity maxima for the overlapping transitions. For a small Jahn-Teller interaction of D-0.01, as is suggested for vg in CF,I, the intensity of the 6, transitions is expected to drop very sharply with increasing n, so that only the 6: should be observed. e have remarked above that the transition at 456 cm- could be assigned with nearly equal ease to 5:, 3;6& or 6;. Considerations based on the predicted relative intensities of the members of the vibrational progression would tend to reduce the likelihood of the final assignment, but given the discrepancies between one- and two-photon intensities observed for the 6; transitions, we do not preclude assignment of the 456 cm- line as 6:. The combination of linear and quadratic Jahn-Teller interactions described in the preceding paragraphs is entirely consistent with placing the energy of the 6i(A,) com- ponent at 456 cm-. 2. The 2E3,Jl] 8 vs, v4 interactions The argument for assigning the 456 cm- transition to the Al component of the 5: transition is based in large part on the observed V; of 458 f 15 cm- reported by Fuss.~~ The possible Jahn-Teller activity of y5 had been given as an explanation for broadening observed in hot bands involving v5.22 As described above the 456 cm- line could also easily be assigned to another transition, and the (A,) component of 5: taken to be absent. This would indicate that the Al intensity borrowing mechanism which is responsible for the strong parallel 6: transitions is considerably less important in the case of v? Even in the case that the 456 cm- band is indeed the 5; (A, ) transition, we must address the question of the Jahn-Teller activity of the mode in order to reconcile our observations with those of Fuss. A linear Jahn-Teller parameter of D-O.04 can accommodate both the E symmetry component of lv{ at -525 cm -l, and the E-symmetry 2v5(jk= 5/2) level at -920 cm -I as the measurements of Fuss suggest. If anharmonicity is assumed to reduce the frequency of the 2v5, we can postulate that the 456 cm- line is the A, component of the 5: transition and that the 522 cm- is the E component. This case would give a slightly smaller linear Jahn-Teller activity for v5, D=O.O3. The approximate expressions for the energies of Jahn-Teller split vibrational levels given by Longuet-Higgins et al52 are valid for D < 0.05,57 and we can use these to calculate expected energies for other vibrational levels. The effects of anharmonicity are included in an ad hoc manner, by simply lowering the energy of the nv5 manifold while maintaining the -69 cm- Jahn- Teller splitting. Such a treatment would predict the 2v5(jk = l/2) (E) level at about 990 cm-, and the 2v5(jk= 3/2) (A,) level at cm-. It is tempting to assign the lines at 1046 and 1516 cm- to the 5i(A1) and 5:(E) transitions based on this crude calculation. However, although the 1046 and 1516 cm- lines are too weak to allow reliable polarization analysis, the width of the 1046 cm- feature seems to preclude its being a 11 Al-Al transition. e note in passing that we do not see evidence for a Fermi resonance between 2v5 and lvl, as exists in the ground electronic state. In any case, accommodation of the Fuss measurement with the observed E symmetry v5 fundamental in the present work suggests that the linear Jahn-Teller interaction is some three times stronger for v5 than for vg. The problem of the highest frequency degenerate vibrational mode, v4, is even more vexing. The pair of transitions at 1282 and 1287 cm- from the origin cannot be fit to reasonable combinations of the other five modes and must be attributed to excitations involving v4 (the 1282 cm -* peak had been incorrectly assigned as 2: in a previous report5). As in the case of vg, the A, symmetry component is higher in energy, implying that the linear Jahn- Teller effect is not sufficient to explain the vibronic activity of this mode. In this case the splitting could not be due to cross-quadratic coupling with v, (the nearest a, mode), because the interaction with a lower frequency mode would shift the E symmetry component higher in energy, just as linear Jahn-Teller coupling. Uncovering the true

16 4370 Taatjes et al.: Spectroscopy of CFsI Rydberg states vibronic interaction once again requires a decision between two possible assignments for the observed features. Assigning this pair of lines to 4; would force us to accept a frequency Y: which is nearly 10% higher than Y;, while all the other vibrational frequencies decrease considerably upon electronic excitation. An alternative explanation is to postulate a very small vibronic interaction for mode 4 and assign the 1282 and 1287 cm- lines as Duschinsky-effect allowed 4;6: combination bands. In some sense this merely shifts the questions from the strange frequency of mode 4 in the excited state to the splitting of the combination level. It is not clear what splittings to expect for a combination of degenerate modes, at least one of which is linear +quadratic Jahn-Teller active. e have chosen the apparently simpler explanation of assigning the line to the 4: fundamental split by combined linear and quadratic Jahn- Teller interactions. Final resolution of the problem would be possible with some independent measurement of v4 in the excited state. 3. Duschinsky rotation in the *ElIdO] state The nondegenerate 2E1,2[O] state is not subject to any Jahn-Teller interactions. However, in electronic states where coupling to another electronic state is possible via more than one vibrational mode of a given symmetry, the normal coordinates are not necessarily the same as the ground state coordinates, but are related by a linear transformation (Duschinsky rotation).54 One consequence of this is that the overlap between the ground state and combination levels involving two modes of the same symmetry is no longer zero.58 The circularly polarized feature 666 cm- from the 2E,,,[O] state origin, at cm-, must be due to excitation of a degenerate vibration, but the energy difference does not match a reasonable guess at the fundamental frequency of any nonsymmetric mode. However, the $6: combination could be expected to appear at a similar energy [in the 2E3,2[1] state the sum of the At fundamentals of v5 and vg is 677 cm-, if one accepts the 456 cm- transition as 5A(A,)]. e tentatively assign the 666 cm- feature as this Duschinsky-allowed combination band. An alternative explanation as a combination involving v3 and vg seems less likely because of the absence of a vg fundamental transition. The quadratic interactions which are responsible for the Duschinsky effect are related to the cross-quadratic interaction terms discussed above for modes of different symmetry V. CONCLUSIONS The (2+ 1) REMPI spectrum of CF,I 5pr-6s Rydberg transitions has been measured. Using the polarization dependence of the two-photon resonant excitation we have been able to assign all four electronic origins in the manifold. Crossing of the Rydberg states with a dissociative curve occurs at about cm-. Density functional calculations indicate that the perturbing state is likely a a-8 transition localized on the C-I bond. The perturbation of both exchange components of the 2E1,2 ion core state, which have different symmetries, points either to a second interacting state in the region or to a reduction from C3, symmetry in the excited 2E,,,[1] state via Jahn-Teller interactions. The vibronic interactions in these states are more complicated than previously thought. Linear and quadratic Jahn-Teller interactions are evident in the splitting of the lv6 mode in the 2E3,2[l] state, with the A&, splitting apparently about twice the A-E linear Jahn-Teller splitting. The mechanism of intensity borrowing operating in these states is still unclear. The Al symmetry component of the 6; band in the 2E3,2[1] state is far stronger than Jahn- Teller induced excitation to the E symmetry component. Assimilation of the observed 2E312[ l] (2 + 1) spectrum with previously reported hot-band transitions requires a Jahn- Teller activity of mode 5 which is about a factor of three larger than that of the lowest-frequency mode, vg. In addition, the frequency of the highest nonsymmetric mode, v4, is still the subject of some uncertainty. Evidence for Duschinsky rotation is seen in the spectrum of the 2E1,2[0] state. Finally, the dynamics of the dissociation of the CF,I+ ion following (2+ 1) REMPI offers some intriguing prospects for further study. Excitation via the 2E3/2[l] 3: transition shows a I+/CF,I+ branching ratio which is four times that for excitation via the near-degenerate 2E3,2[ 116; transition. Indirect evidence, from dissociation via the 2E3/2[l]lh band and from comparison with CHsI dissociations, suggests that the dissociation takes place via absorption of a fourth photon by the ion. Resolution of the kinetic energy and angular distributions of the fragment ions produced by REMPI at these two transitions could shed light on interesting dynamical effects in the electronic states of the CF,I cation. ACKNOLEDGMENTS The authors would like to acknowledge Dr. M. H. M. Janssen for helpful discussions and constructive comments on this manuscript, and J. S. de Keijzer and R. Mooyman for technical assistance. Spectra-Physics B. V. is gratefully acknowledged for the loan of the EOM used in these experiments. e also thank Professor. Hogervorst for providing some of the vacuum equipment necessary for these studies. This work is part of the research programs of the Stichting Scheikundig Onderzoek in Nederland (SON) and the Stichting voor Fundamenteel Onderzoek der Materie (FOM), with financial support from the Nederlandse Organisatie voor etenschappelijk Onderzoek (NO). V. N. Bagiatashvili, V. S. Letokhov, A. A. Makarov, and E. A. Ryabov, MultipIe Photon Infrared Laser Photophysics and Photochemistry (Harwood Academic, London, 1985). *R. M. Robertson, D. M. Golden, and M. J. Rossi, J. Chem. Phys. 89, 2925 (1988). M. Quack, Infrared Phys. 29,441 (1989); M. Quack, R. Schwartz, and G. Seyfang, J. Chem. Phys. 96, 8727 (1992). 4B. Abel, H. Hippler, and J. Troe, 3. Chem. Phys. 96, 8863 (1992); 96, 8872 (1992). G. van den Hoek, J.. Thoman, Jr., D.. Chandler, and S. Stolte, Chem. Phys. Lett. 188, 413 (1992). 6R. S. Mulliken and E. Teller, Phys. Rev. 61, 283 (1942). D. H. Parker, R. Pandolfi, P. R. Stannard, and M. A. El-Sayed, Chem.

17 Taatjes et al.: Spectroscopy of CF31 Rydberg states 4371 Phys. 45,27 (1979); D. H. Parker and R. B. Bernstein, J. Phys. Chem. 86, 60 (1982). *D. J. Donaldson, V. Vaida, and R. Naaman, J. Chem. Phys. 87, 2522 (1987); S. P. Sapers, V. Vaida, and R. Naaman, ibid. 88, 3638 (1988); S. P. Sapers and D. J. Donaldson, Chem. Phys. Lett. 173, 257 (1990). 9 J. A. Syage and J. Steadman, J. Phys. Chem. 94, 7343 (1990); P. G. ang and L. D. Ziegler, J. Chem. Phys. 90,4115 (1989); P. G. ang, Y. P. Zhang, C. J. Ruggles, and L. D. Ziegler, ibid. 92, 2806 (1990). lo. L. Chameides and D. D. Davis, J. Geophys. Res. 85, 7383 (1980); R. B. Chatfield and P. J. Crutzen, J. Geophys. Res. D 95, ( 1990). J. C. Farman, B. G. Gardiner, and J. D. Shanklin, Nature (London) 315, 207 (1985); M. H. Proffitt, J. J. Margitan, K. K. Kelly, M. Loewenstein, J. R. Podolske, and K. R. Chan, ibid. 347, 31 ( 1990). 12K. Hohla and K. L. Kompa, in Handbook of Chemical Lasers, edited by R.. F. Gross and J. F. Bott (iley, New York, 1976). )G. Herzberg, Molecular Spectra and Molecular Structure, u. IIIz Electronic Spectra of Polyatomic Molecules (Van Nostrand Reinhold, New York, 1966). G. Herzberg, Discuss. Faraday Sot. 35, 7 (1963). L. H. Sutcliffe and A. D. alsh, Trans. Faraday Sot. 57, 873 (1961). 16M B. Robin, Higher Excited States of Polyatomic MolecuIes, Vol. I (Academic, New York, 1974). 7M. B. Robin, Higher Excited States of PoIyatomic Molecules, Vol. II (Academic, New York, 1975). *J. A. Dagata, M. A. Scott, and S. P. McGlynn, J. Chem. Phys. 85, 5401 (1986). 19J. A. Dagata,. S. Felps, and S. P. McGlynn, I. Chem. Phys. 85, 2483 (1986). J. A. Dagata, M. A. Scott, and S. P. McGlynn, J. Chem. Phys. 88, 9 (1988). * H. Biirger, K. Bruczyk, H. Hollenstein, and M. Quack, Mol. Phys. 55, 255 (1989). 22. Fuss, J. Phys. Chem. 86, 731 (1982). *. M. McClain, J. Chem. Phys. 55, 2789 (1971).. M. McClain and R. A. Harris, in Excited States, edited by E. C. Lim, Vol. 3 (Academic, New York, 1977), pp. l-56. * K.-M. Chen and E. S. Yeung, J. Chem. Phys. 69, 43 (1978). 26R G. Bray and R. M. Hochstrasser, Mol. Phys. 31, 1199 (1976). 27M. B. Robin, Appl. Opt. 19, 3941 (1980). 2sD. M. Sammeth, S. S. Siewert, L. H. Spangler, and P. R. Callis, Chem. Phys. Lett. 193, 532 (1992); A. A. Rehms and P. R. Callis, ibid. 140, 83 (1987); R. D. Jones and P. R. Callis, ibid. 144, 158 (1988). 29T. R. Bader and A. Gold, Phys. Rev. 171, 997 (1968). MM. B. Robin, Can. J. Chem. 63, 2032 (1985). C E. Moore, Atomic Energy Levels, Vol. III, Natl. Bur. Stand. Circ. 467 (NBS, ashington, D.C., 1958). 32G. van den Hoek, C. A. Taatjes, M. Jonker, D. H. Parker, and S. Stolte (to be published). P. Chen, S. D. Colson,. A. Chupka, and J. A. Berson, J. Phys. Chem. 90, 2319 (1986). 34T. F. Hansico, C. Yan, and A. C. Kummel, J. Phys. Chem. 96, 2982 (1992). 35T. F. Hansico and A. C. Kummel, J. Chem. Phys. 95, 8565 (1991). 36D. H. Parker, J. 0. Berg, and M. A. El-Sayed, in Aduances in Laser Chemistry, edited by A. H. Zewail (Springer, Berlin, 1978). 37Y. B. Fan and D. J. Donaldson, J. Chem. Phys. 97, 189 (1992). *R. A. Morris, J. M. VanDoren, A. A. Viggiano, and J. F. Paulson, J. Chem. Phys. 97, 173 (1992). 39E. J. Baerends and P. Ros, Int. J. Quantum Chem. S 12, 169 (1978). 4oJ. G. Snijders, E. J. Baerends, and P. Ros, Mol. Phys. 38, 1909 (1979). 4 T. Ziegler, E. J. Baerends, J. G. Snijders, and. Ravenek, J. Chem. Phys. 93, 3056 (1983). 42B.. Yates, K. H. Tan, G. M. Bancroft, and J. S. Tse, J. Chem. Phys. 85, 3840 (1986) Powis, 0. Dutuit, M. Richard-Viard, and P. M. Guyon, J. Chem. Phys. 92, 1643 for example the 298 K measurements of CH,I fragmentation at 193 nm by. P. Hess, R. Naaman, and S. R. Leone, J. Chem. Phys. 91, 6085 (1987) with molecular beam measurements of G. N. A. van Veen, T. Baller, and A. E. de Vries, Chem. Phys. 97, 179 (1985). 45M D Person, P.. Kash, and L. J. Butler, J. Chem. Phys. 94, 2557 (1991 ). 46D. M. Szaflarski and M. A. El-Sayed, J. Chem. Phys. 92, 2234 (1988). 47G. van den Hoek and D.. Chandler (unpublished data). 48J. Thoman Jr., D.. Chandler, D. H. Parker, and M. H. M. Janssen, Laser Chem. 9, 27 (1988). 49F. T. Chau and L. Karlsson, Phys. Scripta 16, 248 (1977). 50H. Sponer and E. Teller, Rev. Mod. Phys. 13, 75 (1941). E. Perets and B. E. Scharf, Chem. Phys. Lett. 167, 490 (1990). 52H. C. Longuet-Higgins, U. Gpik, M. H. L. Pryce, and R. A. Sack, Proc. R. Sot. A 244, 1 (1958). s3j. T. Hougen, J. Chem. Phys. 38, 1167 (1963). 54F. Duschinsky, Acta Physicochim. USSR 1, 551 (1937); G. J. Small, J. Chem. Phys. 54,330O (1971); B. Sharf, ibid. 55, 1379 (1971); B. Sharf and B. Honig, Chem. Phys. Lett. 7, 132 (1970);. Siebrand and M. Z. Zgierski, ibid. 62, 3 (1979) 55B. Scharf, Chem. Phys. Lett. 96, 89 (1983); R. Meiswinkel and H. Koppel, Chem. Phys. 129, 463 (1989). 56B. Scharf, Chem. Phys. Lett. 98, 81 (1983). 57M. S. Child, J. Mol. Spectrosc. 10, 357 (1963). R A. Kennedy, T. A. Miller, and B. Scharf, J. Chem. Phys. 85, 1336 (i986).

18