THE JOURNAL OF CHEMICAL PHYSICS 125,

Transcription

1 THE JOURNAL OF CHEMICAL PHYSICS 125, An experimental and theoretical study of double photoionization of CF 4 using time-of-flight photoelectron-photoelectron photoion-photoion coincidence spectroscopy R. Feifel a and J. H. D. Eland Physical and Theoretical Chemistry Laboratory, Oxford University, South Parks Road, Oxford OX1 3QZ, United Kingdom L. Storchi and F. Tarantelli CNR ISTM, Dipartimento di Chimica, Università di Perugia, Via Elce di Sotto 8, Perugia, Italy Received 30 June 2006; accepted 9 October 2006; published online 21 November 2006 Single photon double ionization of CF 4 has been studied by means of a time-of-flight photoelectron-photoelectron coincidence technique, which has very recently been extended towards ion detection, with energy analysis for the electrons and mass analysis for the ions. The complete single photon double ionization electron spectrum of CF 4 up to a binding energy of 51 ev is presented and discussed, also with the aid of accurate ab initio Green s function calculations. From ion detection in coincidence with the ejected electrons, we derive fragmentation pathway-selected double ionization electron spectra of CF 4. From the same data we extract the yield of each doubly charged ion or ion pair as a function of the double ionization energy American Institute of Physics. DOI: / I. INTRODUCTION Tetrafluoromethane CF 4 plays an important role in various different technologies, for example, in the semiconductor industry as a plasma etching gas. By industrial use it has been introduced into the Earth s outer atmosphere, where it is believed to be involved in the depletion of the Earth s ozone layer. To understand its physical and chemical properties in various circumstances, knowledge of its electronic structure in different charge states is essential. The neutral ground state electronic configuration of CF 4 can be denoted as 1a 1 2 1t 2 6 2a 1 2 3a 1 2 2t 2 6 4a 1 2 3t 2 6 1e 4 4t 2 6 1t 1 6 : X 1 A 1, where the square brackets distinguish between the core F 1s, C1s, inner valence, and outer valence orbitals, respectively. The vertical energies for single ionization of the five outer valence orbitals are, according to Brundle et al., , 17.40, 18.50, 22.12, and ev, and for the inner valence orbitals are, according to Siegbahn et al. 2 and Banna et al., and 43.8 ev, respectively. More refined studies on the valence ionization of CF 4 have recently been performed with higher resolution threshold and conventional photoelectron spectrometers, which revealed essentially broad spectral features for the first three outermost electronic states X 2 T 1, Ã 2 T 2, and B 2 E, with some vibrational fine structure superimposed on the B 2 E state see, for example, Refs. 4 and 5 and references therein. Both Yencha et al. 4 and Holland et al. 5 observed regular vibrational progressions for the C 2 T 2 and D 2 A 1 states, which were also found to contain a Electronic mail: raimund.feifel@fysik.uu.se some Jahn-Teller excited modes. The inner valence spectral region around the Ẽ 2 T ev and F 2 A ev states, which appears to be broad and essentially structureless in all these studies, was proposed by Yencha et al. 4 to contain some contributions from doubly ionized CF 4. Experimental information on the CF 4 dication has been obtained by various different techniques such as Auger electron 6,7 and double charge transfer spectroscopies, 7 9 time-of-flight mass spectrometry, 10 photoion-photoion coincidence 11 PIPICO, and photoelectron-photoionphotoion coincidence PEPIPICO spectroscopy, 12,13 as well as by threshold photoelectron s coincidence TPEsCO spectroscopy. 14 In particular, the TPEsCO work of Hall et al. 14 found the lowest threshold for double ionization of CF 4 to be 37.5±0.5 ev, and the PEPIPICO work of Codling et al. 13 determined thresholds for the ion-pair dissociation of CF 2+ 4 into CF + 3 +F ev, CF + 2 +F ev, CF + +F ev, and C + +F + 62 ev, respectively, and tentatively correlated these thresholds with two-hole states of CF 4 as calculated by Larkins and Tulea. 15 More recently, Gottfried et al. 16 recalculated the double ionization spectrum of CF 4 using Green s function ADC 2 method and performed a two-hole density analysis, which led, in particular, to a deeper understanding of the general nature of the Auger spectra. In this work, we present the first complete single photon double ionization electron spectrum of CF 4 up to a binding energy of 51 ev measured with the established time-offlight photoelectron-photoelectron coincidence TOF- PEPECO spectroscopy technique, 17 which collects all electrons emitted at any kinetic energy hence the terminology complete contrary to the TPEsCO technique where the kinetic energy of at least one of the two detected electrons is /2006/ /194318/8/$ , American Institute of Physics

2 Feifel et al. J. Chem. Phys. 125, limited to a narrow range around the zero energy edge. 18,19 To fully interpret the experimental spectrum and discuss its features, we have repeated and analyzed in detail the theoretical calculations of Ref. 16. Very recently, the TOF- PEPECO technique has been further developed towards a new double time-of-flight multiparticle spectroscopy technique, 20 where any number of electrons and ions can be detected in coincidence, with energy analysis for the electrons and mass analysis for the ions. From ion information in coincidence with ejected electrons, fragmentation pathwayselected electron spectra of doubly ionized CF 4 are revealed. II. EXPERIMENT Electron-electron and electron-electron-ion-ion coincidence data of CF 4 were recorded at the photon energies of and ev, respectively, by means of time-offlight coincidence techniques 17,20 22 which are based on a magnetic bottle electron spectrometer as described by Kruit and Read. 23 Two different experimental setups were used for this study. The first one consists of a 5.5 m magnetic bottle well suited for electron-electron coincidence studies at high energy resolution. 17,21,22 The second setup consists of a similar 2.2 m magnetic bottle electron spectrometer to which a 12 cm long two-field time-of-flight ion mass spectrometer has been added for detecting the ions created in coincidence with the ejected electrons. 20 Briefly, in both setups, wavelength selected light from a pulsed low-pressure discharge helium lamp ionizes an effusive jet of the target gas in a crossed beam configuration. Electrons from ionization are directed by the inhomogeneous magnetic field of a permanent magnet to follow nominally on-axis field lines inside a long solenoid to the multichannel plate detector. In order to extract the ions in opposite direction to the electrons, a hollow ring magnet design was chosen for the 2.2 m apparatus which replaces the conically shaped permanent magnet used in the 5.5 m apparatus. A pulsed electrical field for ion extraction is applied about 200 ns after the initial light pulse in the interaction region of the 2.2 m apparatus. The weaker and less strongly divergent magnetic field of the hollow ring magnet results in an inferior electron energy resolution of the 2.2 m apparatus compared to the 5.5 m instrument. 20 The collection-detection efficiencies are experimentally found to be 40% for electrons in both instruments, independent of energy, and 10% for ions. The helium lamps work at a repetition rate of 2 10 khz and emit light pulses which are less than 10 ns long. The timing of electron signals is referenced either to an electrical pulse generated when the lamp fires or to the simultaneous visible light pulse detected by a photomultiplier. Pairs of electrons arriving within 20 s long apparatus /10 s short apparatus of each other are recognized as true coincidences. The times of flight t can be fitted with good precision to the simple form D t = E + E 0 1/2 t 0, 1 where E denotes the kinetic energy of the electrons and E 0 and t 0 are two fitting parameters for calibration. The flight length D is fixed at 9265 ns ev 1/2 long apparatus and FIG. 1. Double ionization electron spectrum of CF 4 recorded at the photon energy of ev. The signal rises significantly 38 ev binding energy. The false coincidences are comparatively few that the statistical uncertainty is essentially given by the square root in terms of standard deviation ns ev 1/2 short apparatus, respectively. Energy calibration of the spectra is done by recording the well known single electron spectrum of molecular oxygen and the rare gases before and after each run. The typical run time for these experiments is several days, and longer runs are currently not practicable because of stability problems. Commercially available CF 4 gas with a stated purity of 99% was used for the experiments. The purity of the gas has been checked carefully by recording the single electron valence band spectrum and, in the case of the 2.2 m apparatus, by examining the mass spectrum, before and after each coincidence run. III. EXPERIMENTAL RESULTS In Fig. 1 we present the complete double ionization electron spectrum of CF 4 obtained from the full coincidence data set recorded at the 5.5 m apparatus at ev photon energy. The signal rises significantly at 38 ev binding energy, which corresponds closely to the threshold value for double ionization of CF 4 of 37.5±0.5 ev reported earlier by Hall et al. 14 Below 38 ev binding energy an extended tail is visible which has an onset at around 25 ev not shown here ; a similar tail, which is likely to arise from false coincidence events, has been reported earlier for the double ionization spectrum of SF 6 cf. Ref. 24, and its relative height depends on instrumental conditions. The double ionization electron spectrum of CF 4 presented here reveals distinct substructures as indicated by the vertical dashed lines in the figure, contrary to the TPEsCO result of Hall et al. 14 which mainly shows a broad spectral feature centered around 43 ev binding energy. Whether or not the less structured TPEsCO spectrum reflects the predominance of autoionization processes, as was found in the case of doubly ionized SF 6 cf. Refs. 24 and 25, is difficult to assess, as the measurements of Hall et al. 14 were performed only at comparatively moderate energy resolution 0.5 ev photon energy step.

3 Double photoionization of CF 4 J. Chem. Phys. 125, FIG. 2. Comparison of the single electron spectrum and the double ionization electron spectrum of CF 4 measured simultaneously at ev photon energy. To further analyze the possible involvement of molecular autoionization processes in single photon double ionization of CF 4, it is helpful to compare the double ionization electron spectrum with the inner valence spectral region of the conventional photoelectron spectrum between 36 and 48 ev, which may, in a simplified picture, be associated with the formation of the 2t ev and 3a ev states of CF + 4 known to possess pronounced F 2s character cf. Ref. 5. In Fig. 2 we display the single electron spectrum upper panel recorded simultaneously in our experiment with the double ionization electron spectrum lower panel of CF 4 at ev photon energy. As we can see in both spectra, broad, overlapping features are visible in the 36 to 48 ev region. Even though the centroid of the double ionization spectrum is shifted towards higher binding energies compared to the partially unresolved inner valence part of the conventional photoelectron spectrum, which qualitatively differs from the SF 6 case, 24 a significant portion of the double ionization electron spectrum still lies below the 2t ev and 3a ev regions, which energetically allows for autoionization processes. It should be noted that, in contrast to the sharp peaks of the outer valence photoelectron spectrum, which are well reproduced by Green s function calculations, 5 the two observed inner valence bands are very broad and the calculations show a very pronounced scattering of the intensity over many states and clear difficulties in reproducing the experiment. 5 While higher order configuration mixing, not accounted for in the ADC 3 treatment of Ref. 5, may contribute to these discrepancies, inherent line broadening due to decay via electron loss may also be part of the explanation. In order to get a better understanding of single photon double ionization processes in CF 4, the electron-electron coincidence map can be studied. In Fig. 3 we display, as a gray scale color online dot plot, the complete coincidence map of CF 4 recorded at ev photon energy, where the sum of the two electron energies E 1 +E 2 is displayed on the ionization energy scale vertical axis versus the kinetic energy scale of one of the electrons E 1 horizontal axis. Aswecan FIG. 3. Complete coincidence map of CF 4 recorded at ev photon energy as a gray scale color online dot plot, where the sum of the two electron energies E 1 +E 2 is displayed on the ionization energy scale vertical axis vs the kinetic energy of one of the electrons E 1 horizontal axis. see, the intensity is higher in the low kinetic energy region of E 1 compared to the region close to the main diagonal of this plot. As in our previous investigation of SF 6, double ionization electron spectra of CF 4 can be projected from the full coincidence map in various different ways by taking either all electrons or only certain selections of them. 24 In Fig. 4 we show a comparison of double ionization electron spectra obtained in different ways. Panel a presents the spectrum where all electrons were used cf. Figs. 1 and 2 above ; panel b shows the spectrum obtained by choosing a narrow slice of coincidence events from the low kinetic energy region of Fig. 3 E 1 from 0 to 1 ev kinetic energy at 40 ev binding energy and from proportionally narrower ranges at lower electron pair energies ; panel c shows a similarly created spectrum, where an even narrower E 1 range of ev kinetic energy at 40 ev binding energy was chosen; and panel d shows the spectrum created by selecting a slice of coincidence events from the fairly uniform area close to the main diagonal of Fig. 3, where the width of the slice was chosen to vary from 2 ev at 40 ev binding energy to proportionally narrower values at lower electron pair energies. As we can see, the spectra are very different, and the changes observed show a trend reminiscent of the SF 6 case. 24 In particular, spectra b and c appear essentially structureless, in contrast to the well structured spectra a and d. As was argued in the SF 6 case, these differences in spectral appearance could be related to the different proportion of autoionization versus direct double ionization. Spectra b and c are dominated by coincidence events related to low kinetic energy electrons whose intensity seems to be comparatively high cf. Fig. 3 and which possibly originate from molecular autoionization processes involving the singly ionized inner valence electronic states associated with the 2t 2 and 3a 1 orbitals of CF 4. On the contrary, the spectrum of Fig. 4 d, where electrons with nearly equal energies are taken from the less intense region close to the diagonal, is more likely to arise as a direct CF 4 2+ spectrum cf. Ref. 24. A more detailed comparison of the spectra in Fig. 4 seems to support this conclusion, because the relative inten-

4 Feifel et al. J. Chem. Phys. 125, FIG. 4. Comparison of differently obtained double ionization electron spectra. Panel a presents the spectrum where all electrons were used cf. Figs. 1 and 2 above ; panel b shows the spectrum obtained by choosing a slice of coincidence events from the low kinetic energy region of Fig. 3 E 1 from 0 to 1 ev kinetic energy at 40 ev binding energy and from proportionally narrower ranges at lower electron pair energies ; panel c shows a similarly created spectrum, where a much narrower E 1 range of ev kinetic energy at 40 ev binding energy was chosen; and panel d shows the spectrum created by selecting a slice of coincidence events from the fairly uniform area close to the main diagonal of Fig. 3, where the width of the slice was chosen to vary from 2 ev at 40 ev binding energy to proportionally narrower values at lower electron pair energies. sity of the spectral region of double ionization lying above the higher single ionization energy 43.8 ev seems pronouncedly quenched in the low-e 1 spectra. This is demonstrated in Fig. 5, where the experimental data points of Figs. 4 a, 4 b, and 4 d have been spline smoothed and reported FIG. 5. Natural spline smoothing of the experimental data of Figs. 4 a, 4 b, and 4 d. The vertical line at 43.8 ev shows the position of the high energy inner valence photoelectron band of CF 4. FIG. 6. Upper panel: ion selected double ionization electron spectra of CF 4, intensity corrected for the ion collection-detection efficiency. The permanent arrows mark the threshold values for the ion-pair formation found by Codling et al. Ref. 13, and the dashed line/arrows indicate suggestions for more refined values. The permanent line marks the double ionization energy of 43.3 ev for which the yield of each doubly charged ion or ion pair is exemplary extracted see text. Lower panel: sum of the partial double ionization electron spectra of CF 4. on the same vertical scale. While the three spectral profiles are very similar below the ionization energy at 43.8 ev vertical line, the distinct features seen lying above this threshold in the E 1 E 2 spectrum are very much weaker in the all-electron spectrum and essentially absent in the low-e 1 spectrum. As the autoionization channel should be less efficient or even unavailable for reaching these higher energy dicationic states, our findings are consistent with the surmise that the direct one-photon double ionization mechanism prevails, in particular, for those events leading to an even energy share between the two outgoing electrons. The double ionization electron spectrum of CF 4 is inherently complex due to a large number of possible electronic states and dissociation channels. An experimental way of partially disentangling it is to detect the created ions in coincidence with the ejected electrons as can be done with our 2.2 m instrument. The upper panel of Fig. 6 gives such results for doubly ionized CF 4, showing which electronic states of CF 4 2+ lead to which product ion s formed on the mass spectrometer time scale lifetimes 10 5 s. As we can see, CF 3 2+ and CF 2 2+, which are frequently used in collision experiments for the study of bimolecular reactions see, e.g., Refs. 26 and 27 and references therein, but not CF 4 2+, are stable doubly charged ions. Furthermore, CF 4 2+ decomposes predominantly by CF 3 + +F + and CF 2 + +F + formations and a little bit of CF + +F + formation, whereas the C + +F + and F + +F + channels are not opened yet in the energy region studied

5 Double photoionization of CF 4 J. Chem. Phys. 125, FIG. 7. Schematic diagram illustrating the energetics of the dissociation channels of doubly ionized CF 4 studied in this work. here. Figure 7 summarizes, in schematic form, the energetics of the dissociation channels of doubly ionized CF 4 studied here. By correcting the relative intensities with respect to the collection-detection efficiencies of the ions, we can sum up the partial double ionization electron spectra from the upper panel of Fig. 6, which results in the so-called PEPEPI PI - COsum spectrum as displayed in the lower panel of Fig. 6.A comparison of this spectrum with the complete double ionization electron spectrum from above cf. Figs. 1 and 2, which is made in Fig. 8, demonstrates how the differently obtained spectra correspond well to each other within the statistical limitations given. More importantly, from the spectra presented in Fig. 6, the yield of each doubly charged ion or ion pair as a function of the double ionization energy can easily be extracted. For example, at the double ionization energy of 43.3 ev, which is marked by the permanent line in Fig. 6, we obtain the following yields: CF %, CF %, CF 3 + +F + 35%, CF 2 + +F + 35%, and 0% for the remaining channels. FIG. 8. Sum of the TOF-PEPEPIPICO spectra cf. Fig. 6 in comparison to the complete double ionization electron spectrum of CF 4 cf. Fig. 1. In the PEPIPICO work of Codling et al., 13 the following threshold values for ion-pair formation from CF 4 2+ were determined: CF 3 + +F ev, CF 2 + +F ev, CF + +F ev, and C + +F + 62 ev. For orientation, these values are indicated by permanent arrows in Fig. 6. As we can see, even though the threshold values of Codling et al. fall within the range of bands observed in our spectra, some corrections are needed. Due to comparatively low statistics in the present fragmentation pathway-selected electron spectra, we can give only tentative threshold values, as indicated in the figure by the dashed line and dashed arrows, respectively, until data of even better quality can be obtained: CF 3 + +F ev, CF 2 + +F ev, and CF + +F ev. In the threshold photoelectron spectroscopy work of Yencha et al., 4 it was proposed that a spectral feature observed at 37.2 ev in their electron spectrum is at least partially associated with the opening of the lowest double ionization channel, CF 4 +h CF 3 + +F + +2e. As we can see from Fig. 6, the presence of this specific channel in the threshold photoelectron spectrum may be possible but can also be expected to be comparatively weak in this energy range. Furthermore, it should also be noted that in the partial double ionization electron spectrum measured in coincidence with CF 3 2+ formation, some weak structures seem to be visible in the 37 ev region, which raises the question whether or not the lowest double ionization channel can really be attributed to a charge separation process in the form of CF 4 +h CF 3 + +F + +2e, or if perhaps a charge retaining process in the form of CF 4 +h CF F+2e is the lowest one. Further investigations are needed to answer these questions. IV. THEORETICAL ANALYSIS OF THE SPECTRUM AND DISCUSSION The entire theoretical double ionization spectrum of CF 4, extending above 100 ev, computed by Green s function ADC 2 method and analyzed by means of the two-hole density analysis was presented and discussed in detail in Ref. 16, where the carbon and fluorine Auger spectra were studied in particular. Here we shall focus our attention on the low double ionization energy region of the computed spectrum, below 51 ev, in connection with the present direct double photoionization experiment. There are some 50 relevant doubly ionized states calculated to lie in this region. The twohole population analysis of the dicationic states, discussed in Ref. 16, is a useful tool to begin interpreting the present spectrum. It showed that the states of CF 4 2+, essentially like those of other small fluorides, present a pronounced atomic localization of the two holes, which is essentially dominated by either one of two components. The two electron vacancies localize either on different fluorine atoms F 1 F 1 component or on the same fluorine F 2 component. The other components of the 2h strength involving carbon, C 1 F 1 and C 2, are generally much smaller. Furthermore, the dicationic states tend to cluster in separate groups of alternating character, according to the dominating component and the valence shells involved in the ionization. In particular, the states lying in the low energy region explored by the present

6 Feifel et al. J. Chem. Phys. 125, double ionization spectrum are all characterized as having the two electron vacancies mainly localized in the 2p shells of two different fluorine atoms, because this minimizes holehole repulsion. The other components of the two-hole density are very small 5% 20% of the total by comparison and increase in magnitude only at higher energy. The two-hole density analysis led us to a very similar description of the nature of the double ionization spectrum of SF 6, 24 where a much larger roughly double number of dicationic states were found contributing in a comparable energy range, but some differences can be pointed out. One may note, in particular, that the main features of the double ionization spectrum of CF 4 are shifted by about 2 ev towards higher energies in comparison to those of SF This reflects exactly the difference in inverse distance Coulomb repulsion between two naked charges located on two distal fluorine atoms in SF 6 and on two F atoms of CF 4. In addition, the lack in CF 4 of the inequivalent pairs of fluorine atoms vicinal and distal present in SF 6 is partly responsible for the less structured double ionization spectrum of the former system. To compare the theoretical results with the experimental data, we use the same method already adopted for the SF 6 case, 24 drawing the theoretical spectrum as a Gaussian convolution of the 2h spectroscopic factors of the states or some component thereof, each weighted by its total degeneracy. This procedure must be expected to give little more than a qualitative reproduction of the experiment, but it provides nonetheless insightful elements for the analysis and the interpretation of such complex spectra. Two such computed profiles differing as discussed below are displayed in Fig. 9 a below the experimental data of Fig. 4 d. The latter has been chosen for comparison because, as argued above, the portion of the coincidence data corresponding to an even energy distribution of the two ejected electrons is more likely to reflect a direct double photoionization mechanism. Both the theoretical spectra shown incorporate an energydependent intensity falloff factor varying linearly from 1 at the double ionization energy threshold to 0 at the photon energy threshold. Both are shifted up in energy by 1.5 ev in order to align better with the experimental profile, 25 and the full width at half maximum of the Gaussians convoluting the discrete data increases linearly from 0.7 ev at low energy to 1.2 ev. The difference between the two computed profiles is that the broken line represents a convolution of the total 2h strength of the states while, for the solid line spectrum, the 2h F 1 F 1 component, representing the localization of the two holes each on a different fluorine atom, has been subtracted. The resulting intensity, as we mentioned above, is comparatively very small and has been amplified in the figure to make the peak at 43.2 ev about equal in both theoretical spectra. Compared with the experimental double ionization profile, the theoretical spectra show some evident inaccuracies in the relative intensity of the various features but their number, position, and shape are remarkably well reproduced, which enables us to discuss the spectrum with confidence and in detail. It is interesting to note that the total 2h profile dashed line gives a particularly unsatisfactory account of the relative intensities in the low energy part of the spectrum. FIG. 9. Theoretical double ionization spectra of CF 4 shown below the experimental data of Fig. 4 d. a The dotted curve shows the total 2h spectrum, while the solid curve is obtained by subtracting the F 1 F 1 2h component. The vertical bars represent the computed discrete data for the latter spectrum. b Spectrum derived from the latter by doubling the relative intensity of dicationic states originating from double ionization of the fluorine lone pairs. For details see text. In particular, the first two bands at 40 and 41 ev are unduly intense compared to the peak at 43.2 ev, which is the most prominent in the experimental data. As we said above, the total 2h strength is dominated by the F 1 F 1 component. But this is especially true near threshold, while the proportion of the C 1 F 1 term tends to increase with energy see below. Subtracting the F 1 F 1 component solid line the relative intensity of the low energy bands improves substantially in qualitative terms. On the other hand, in the solid line profile, the relative proportions of the higher energy region of the spectrum above 44 ev are much exaggerated. It would thus appear that a suitable weighted sum of the components of the spectroscopic factors, with a low weight of the F 1 F 1 contribution, could provide a satisfactory qualitative account of the measured cross section profile. That the F 1 F 1 component should contribute little to the observed double ionization rates is plausible on the grounds that the probability should be smaller to populate states where the two holes are localized on distant atomic sites cf. also the smaller intensity observed for the distal F 1 F 1 component in the double ionization spectrum of SF 6 Ref. 24. But we also note that the relative intensity of the various features of the spectrum may be affected by the energy

7 Double photoionization of CF 4 J. Chem. Phys. 125, below 44 ev and amplifying by 50% the lines underlying the peak at 45 ev. The factors used are only a rough guess but the qualitative agreement with the experimental profile is significantly improved. V. SUMMARY FIG. 10. Gaussian convolution of the computed orbital composition of the dicationic states of CF 4. Solid line: component representing the double ionization of the 1e, 4t 2, and 1t 1 orbitals fluorine lone pairs. Dotted line: component representing the ionization of one lone pair and one bonding orbital 4a 1 or 3t 2. Filled curve: component representing double ionization of the bonding orbitals. dependence of the double ionization rate assumed linear in our simulation and by the possibly varying relative weight of the autoionization and direct mechanisms. On the basis of our earlier discussion, both these factors may be especially relevant in differentiating the relative appearance of the lines below and above the evident gap at 44 ev. Another line of analysis is precisely related to the fact that this gap at about 44 ev actually separates the spectrum quite neatly into two broad composite structures of different nature, as illustrated in Fig. 10. The region below 44 ev is related essentially to the double ionization of the fluorine p lone pairs, involving the 1e, 4t 2, and 1t 1 molecular orbitals. Above 44 ev, on the other hand, double ionization mainly involves taking one electron out of the lone pairs and one out of the bonding orbitals 4a 1 and 3t 2. The borderline band at 45 ev is an exception, in that both types of double ionization characters contribute almost equally. At high energy, near threshold, a small contribution begins to be noticeable arising from double ionization of the bonding orbitals. The presence of substructures within each region can qualitatively be related to the energy gaps between the different orbitals involved, as was pointed out in Ref. 16. It is, of course, the increasing involvement of the bonding orbitals with energy that is reflected in the increasing weight of the C 1 F 1 component of the 2h density noted above. It is plausible to assume that the double photoionization cross section is larger for the outermost fluorine lone pairs than when inner bonding orbitals are involved. Taking this into account would clearly tend to improve the relative intensity of the two halves of the spectrum incorrectly reproduced by the theoretical profile solid line of Fig. 9 a. Note, in particular, how the mixed nature of the band at 45 ev, mentioned above, is nicely consistent with its observed larger intensity relative to the bands at higher energy. As a simple illustration of these arguments, we show in Fig. 9 b the theoretical spectrum obtained from the one displayed as the solid line of Fig. 9 a by arbitrarily doubling the intensity of the lines Single photon double ionization of CF 4 has been investigated by means of time-of-flight coincidence techniques, where any number of electrons and ions can be detected in coincidence, with energy analysis for the electrons and mass analysis for the ions. The complete single photon double ionization electron spectrum of CF 4 up to a binding energy of 51 ev was presented, and spectra which are likely to reflect either mainly direct or mainly indirect double ionization of CF 4 were extracted from the coincidence map. Using ion information obtained in coincidence with the ejected electrons, the first fragmentation pathway-selected double ionization electron spectra of CF 4 were revealed, which give direct access to the yield of each doubly charged ion or ion pair formed on the mass spectrometer time scale as a function of the double ionization energy. The double ionization spectrum was analyzed and interpreted quite accurately by using Green s function ADC 2 calculations and a simple analysis of the two-hole density. The energy range of the spectrum covers the outermost valence ionization, and all its features and their relative intensity can be assigned and interpreted in terms of double ionization of the fluorine lone pairs and/or the bonding orbitals and in terms of the atomic localization pattern of the holes. The present results confirm satisfactorily the previous similar analysis of the SF 6 double ionization spectrum, in particular, for what concerns the smaller relative importance of intensity components related to the removal of electron pairs localized on distant atomic sites. ACKNOWLEDGMENTS One of the authors R.F. would like to thank the Swedish Research Council VR and the Swedish Foundation for International Cooperation in Research and Higher Education STINT for financial support of his stay at Oxford University. 1 C. R. Brundle, M. B. Robin, and H. Basch, J. Chem. Phys. 53, K. Siegbahn, C. Nordling, G. Johansson et al., ESCA Applied to Free Molecules North-Holland, Amsterdam, M. S. Banna, B. E. Mills, D. W. Davis, and D. A. Shirley, J. Chem. Phys. 61, A. J. Yencha, A. Hopkirk, A. Hiraya et al., J. Electron Spectrosc. Relat. Phenom. 70, D. M. P. Holland, A. W. Potts, A. B. Trofimov et al., Chem. Phys. 308, R. R. Rye and J. E. Houston, J. Chem. Phys. 78, W. J. Griffiths, S. Svensson, A. Naves de Brito, N. Correia, C. J. Reid, M. L. Langford, F. M. Harris, C. M. Liegener, and H. Ågren, Chem. Phys. 173, W. J. Griffiths and F. M. Harris, Int. J. Mass Spectrom. Ion Process. 85, F. M. Harris and B. C. Cooper, Int. J. Mass Spectrom. Ion Process. 97, T. Masuoka and A. Kobayashi, J. Chem. Phys. 113, D. M. Curtis and J. H. D. Eland, Int. J. Mass Spectrom. Ion Process. 63,

8 Feifel et al. J. Chem. Phys. 125, J. H. D. Eland, L. A. Coles, and H. Bountra, Int. J. Mass Spectrom. Ion Process. 89, K. Codling, L. J. Frasinski, P. A. Hatherly, M. Stankiewicz, and F. P. Larkins, J. Phys. B 24, R. I. Hall, L. Avaldi, G. Dawber, A. G. McConkey, M. A. MacDonald, and G. C. King, Chem. Phys. 187, F. P. Larkins and L. C. Tulea, J. Phys. Paris, Colloq. 48, F. O. Gottfried, L. S. Cederbaum, and F. Tarantelli, J. Chem. Phys. 104, J. H. D. Eland, O. Vieuxmaire, T. Kinugawa, P. Lablanquie, R. I. Hall, and F. Pennent, Phys. Rev. Lett. 90, R. I. Hall, A. McConkey, L. Avaldi, M. A. MacDonald, and G. C. King, J. Phys. B 25, B. Krässig and V. Schmidt, J. Phys. B 25, L J. H. D. Eland and R. Feifel, Chem. Phys. 327, J. H. D. Eland, S. S. W. Ho, and H. L. Worthington, Chem. Phys. 290, J. H. D. Eland, Chem. Phys. 294, P. Kruit and F. H. Read, J. Phys. E 16, R. Feifel, J. H. D. Eland, L. Storchi, and F. Tarantelli, J. Chem. Phys. 122, A. J. Yencha, M. C. A. Lopes, D. B. Thompson, and G. C. King, J. Phys. B 33, N. Tafador and S. D. Price, Int. J. Mass. Spectrom , S. M. Harper, S. W.-P. Hu, and S. D. Price, J. Chem. Phys. 121,