Photoelectron Spectroscopy

1 Introduction

Photoelectron spectroscopy remains an active field of research at both the fundamental spectroscopic level and also as regards its applications to problems in chemistry and physics. The scope of the technique as applied to a variety of chemical problems has been discussed in some review articles. These cover both gas-phase studies using U.V. exciting radiation ('u.v. molecular photoelectron spectroscopy') and solid state work using soft X-radiation ('X-ray photoelectron spectroscopy').

The proceedings of a major conference on electron spectroscopy in general, held at Asilomar, California, in September 1971, have recently been published. The contents provide a most illuminating perspective of the subject. Also now generally available are the proceedings of a conference on molecular spectroscopy (including photoelectron spectroscopy), held at Brighton, also in September 1971. These two publications will be fully reviewed in next year's Report, though we shall below make reference to any work germane to the 1971 literature.

Special mention should also be made of a recent review by Brundle on 'The Application of Electron Spectroscopy to Surface Studies'. A particularly valuable feature of this article for our purposes is the discussion of various branches of electron spectroscopy, such as Auger spectroscopy, which are not dealt with in the present Report.

2 Ultraviolet Molecular Photoelectron Spectroscopy

Instrumental. Few major advances have been reported this year, although there have been a number of interesting developments. High-quality work continues to appear from many types of analyser: however, it is worthy of note that in practice, equally high resolution may be obtained from the simpler analysers as from the more complex designs. For example, Price et al., using a conventional 10 cm radius 127° cylindrical analyser, have resolved the spin-orbit splitting in the first p.e. band of O2, first reported by Edqvist et al. in 1970; the latter used a variable retarding field with a 10 cm spherical analyser. A very simple and inexpensive spectrometer design, combining a variable retarding field with a very small deflection analyser, has been described. The results from this instrument, which is capable of 25 meV resolution, suggest that for many applications a relatively unsophisticated system is quite adequate. The use of a multi-channel analyser with a simple spherical-grid retarding spectrometer has been described by Delwiche et al. Although the resolution, ca. 30meV at best, is remarkable for this type of electron analyser, the sensitivity of the equipment described is very much less than is readily attainable using a conventional deflection analyser and counting system at much lower total cost.

The application of a type of analyser new to u.v.–p.e. work, the cylindrical mirror, has been described by Berkowitz, who also incorporated a high-temperature molecular beam generator in his instrument. The system in this mode, however, gives a performance which seems inferior to that of the new Perkin-Elmer heated inlet system on a more traditional instrument (see below). Studies on HF and DF also failed to establish any significant advantage for this analyser over the conventional type previously used by Brundle to study the same molecules. Moreover, the combination of a focussing lens system feeding a spherical analyser has been shown by Heddle to be superior theoretically (in terms of étendue) to the cylindrical mirror, although it has to be admitted that such an arrangement is inherently much more complex than a simple deflection analyser.

The variation of sensitivity with electron kinetic energy in deflection analysers has been discussed by Berkowitz and Guyon: they point out that the 'l/KE' factor frequently used to correct experimental intensities is not universally applicable. Moreover, in retarding analysers it now appears that whether the sensitivity is constant or declines with increasing electron kinetic energy depends on the optics of the retarding system used, and the position here is thus less straightforward than was implied in last year's Report.

Research by Perkin-Elmer Ltd. has now produced a simple and very effective method for running less-volatile materials. Using the lamp plasma as a source of heat, target chamber temperatures of up to ca. 500°C have been attained, although only 250° C is guaranteed in the current PS 18 model incorporating the device. This development should greatly extend the scope of vapour-phase u.v.–p.e. spectroscopy in the immediate future. As an alternative to a separate u.v.–p.e. instrument, the manufacturers of the established commercial X–p.e. equipment (vide infra, p. 22) all now offer accessory U.V. sources for gas-phase work. Varian, for example, claim 1OOOOO counts s-1 at 21 meV resolution (Ar 2P3/2 line) and 45 000 counts s-1 at 14 meV. No sample pressure is specified, but this performance would seem to be competitive with that of the purpose-built Perkin-Elmer PS 16–18 series as far as work at room temperature is concerned.

An increasing number of He-II spectra are now being reported. Although lamps giving virtually 100%, He-II radiation have been produced experimentally, using both windowed and windowless configurations (Figure l), the experimental difficulties continue to prevent their widespread application and many workers are still using photon fluxes of ca. 1 — 2% of the concomitant He-I output. The principal difficulties appear to centre on the problem of maintaining adequate helium purity at the very high current densities necessary for generation of the 304 Å, line in high intensity.

A Compilation of U.V. –P.E. References. — The inorganic molecules whose u.v. –p.e. spectra. have been reported or discussed in 1971 are listed below (Table 1). Some important related 1972 references have also been included, although comprehensive coverage of the 1972 literature has not been attempted. These latter references are enclosed in brackets. References to 1970 papers have been included in Table 1 only if no mention was made of them in last year's Report. Papers including He-II spectra are marked with a superscipt b: these papers usually include He-I spectra as well. A few organic molecules of particular spectroscopic or inorganic interest have been included, but we have not attempted a comprehensive survey of organic u.v.–p.e. work. Work concerned solely with angular distributions is considered separately below, and has not been included in Table 1.

It will be noted that this table is substantially shorter than last year's : nevertheless, it is our impression that there is no diminution of activity in the u.v.–p.e. field. Last year's Report was augmented by the massive compendium of Turner et al., the fruit of ten years' development work, and the publication of the Royal Society's discussion on p.e. spectroscopy. No such volumes have appeared this year, although we would like again to draw attention to the report of the conference on 'Electron Spectroscopy' held in Asilomar, California in 1971, and to the volume 'Molecular Spectroscopy (1971)', which were not in fact published until well into 1972, and will accordingly be fully discussed in next year's Report, A recent u.v.–p.e. review by Worley also merits attention.

Assignment Criteria. — A full account of the assignment criteria used to interpret u.v.–p.e. spectra was given in last year's Report, and therefore needs little elaboration here. All these criteria have continued to be extensively used during 1971, only one essentially new technique being introduced, the use of angular distribution measurements.

Angular Distribution Measurements. Measurements of the angular distribution of photoelectrons from monatomic and diatomic species are well established and have continued to appear in 1971. Some theoretical studies have also appeared, the results generally being in reasonable agreement with experiment. The relative magnitude of the p.e. intensity at an angle θ is given by

[MATHEMATICAL EQUATIONS NOT REPRODUCIBLE IN ASCII] (1)

The anisotropy parameter β depends on the orbital involved and on the energy of the emitted photoelectron. The most interesting discovery this year is that the several vibrational elements of a band sometimes have different values of β: this may be expected when autoionization is known to occur (e.g. in the Ne-I spectrum of O2), but in the 584 Å spectrum of N2, for instance, it seems to be due to a breakdown of the Born–Oppenheimer approximation. The angular distribution of photoelectrons from different rotational H2 -> H2+ transitions has been investigated by Niehaus and Ruf: they find β = 1.95 for ΔN = 0 and 0.85 for ΔN = 2. However, these results are primarily of spectroscopic interest : as an aid to assignment the essential feature is that the variation of intensity with angle is different for ionization from different orbitals. This produces a change in overall band contour with angle whenever the band in question consists of several overlapping ionization processes, and calculation of β can in principle enable the orbitals in question to be identified. The recent publication of results for benzene8s (Figure 2) demonstrates that this method can be a very powerful assignment aid for polyatomic molecules even when accurate calculations of β are not feasible. (In the present instance, the assignment given last year by Åsbrink et al., using a sophisticated vibrational analysis, is confirmed.) Further data have very recently been published by Carlson et al., and the method clearly has considerable potential.

Calculations. Many workers have again made use of approximate MO calculations in their assignment of p.e. spectra. Nevertheless, the Reporters do not feel that much reliance should be placed on them as the principal basis for the assignment. Correlations with approximate calculations seem most helpful when the main features of the spectra in question can be assigned without their aid [e.g. F2O, Cl2O (Figure 3), xenon fluorides (Figure 9, p. 16)], and the experimental results used to pinpoint deficiencies in the computational method. That calculation alone is not a satisfactory means of assignment is clearly shown by the example of borazine, Two independent studies of substituted borazines have demonstrated that several earlier calculations inverted the uppermost filled levels. The importance of d-orbital participation in the valence structure of silicon and germanium compounds is another problem which has been studied this year by (inter alia) the correlation of approximate MO calculations and u.v.–p.e. results. The reliability of such methods is, however, open to question.

Discussion of Selected Results. — Small Molecules, Radical Species, and Excited States. In 1971 the majority of workers have again concentrated on the smaller molecular species. This is perhaps not very surprising, though it does seem to the Reporters that the assignment problem is often no harder in large molecules containing atoms of widely differing electronegativities than in small molecules with low symmetry, which in any case are often of more purely spectroscopic interest than chemical interest. Reliable assignment in these latter cases is often surprisingly difficult [e.g. carbonyl halides, NO (vide infra)].

The benefits of high resolution in the study of small molecules are well demonstrated in the high-resolution spectrum of H2S, reported recently both by Frost et al. and Potts and Price (Figure 4). Both these groups resolved fine structure on the high I.E. side of the second band (2A1), and both explain it as vibrational-rotational structure, in accord with the calculations of Dixon. Delwiche et al., however, working at lower resolution, ca. 50 meV (as were Turner et al. before them), fail to see this structure and consequently assume, erroneously, that the molecule predissociates where the main progression terminates.

However, the most impressive high-resolution work this year comes, once again, from Åsbrink et al., who continue their extremely careful studies on small molecules with another paper on nitric oxide. This not only includes a He-I spectrum at ca. 10 meV resolution — sufficient to resolve the spin–orbit splitting (12 — 16 meV) (Figure 5: cf. O2 last year) — but also a He-II spectrum at 25 meV resolution which took no less than 75 hours to record. This latter spectrum (Figure 6) incidentally contains an unusually small He self-absorption peak. This improved data enables the authors to correct the assignment they previously favoured: the weak band at 20.4 eV, previously assigned to the C3Π state, is shown to be due to an unsuspected 320.4Å line [2p2(3P) -> 1s2p(3PO)] in the discharge lamp. The C3Π and B1Π states are now believed to be almost degenerate at 21.7 eV, the band at 22.5 eV being assigned to B11Σ+. In addition, the I.E. values given in the earlier work are corrected by amounts ranging up to 10 meV. (The earlier assignment had also been questioned by Lefebvre-Brion, who, using CI methods, predicted weak additional bands in the spectrum.) This work demonstrates that the danger of misinterpretation of spectra due to the unsuspected existence of weak 'satellite' lines in the light can still be underestimated even by the most careful of workers. Other spectra of impressive quality reported by this group in 1971 include a number of organic molecules containing five-membered ring systems.

With the supply of very small stable molecules almost exhausted, some workers are turning to the p.e. spectroscopy of radical species and excited states. The atomic hydrogen spectrum contains but one band at 13.61 eV, in excellent agreement with expectations. Initially, the expected 2P ion state of oxygen was not detected, although the 4S and 2D states were clearly visible." However, in a later paper the same authors reproduce a spectrum (Figure 7) showing all three expected states of O+, but three other lines in this latter spectrum remain unexplained. Other examples reported include those of electronically excited (1Δg) oxygen, SO, and of the bent triatomic species NF2. This latter spectrum was assigned with the aid of INDO calculations on both radical and ion: recent ab initio calculations on NF2 and CF2 (isoelectronic with NF2+) also support the assignment proposed. Much of this work is of predominantly spectroscopic interest, though Jonathan et al. suggest that their results indicate the usefulness of u.v.–p.e. spectroscopy as a technique for studying gas kinetics.

Other groups, realizing perhaps that corrosive material can often be handled at low pressures without catastrophic damage to spectrometers, have been investigating such chemically reactive molecules as NSF, F2, and HF. Five groups, for example, have investigated the halogens this year. The earlier low-resolution data for fluorine reported by Frost et al, is shown to be in error, owing to nitrogen impurities in the sample: such contamination can be readily identified in the high-resolution p.e. spectrum by its characteristic vibrational structure. The new results are generally in good agreement with each other, although some of the I.E. error limits suggested seem rather optimistic considering the spread of values reported. The two groups working at the highest resolution observed 'hot bands' (p.e. bands arising from ionization of vibrationally excited molecules) in the chlorine and bromine spectra (see Figure 8), and the careful determination of the adiabatic I.P. values by Dibeler et al., who recorded photoion-yield curves at a number of temperatures, may therefore give the most reliable values. Brundle et al. have continued their interesting and technically very demanding work on the fluorides of the noble gases: the spectra of XeF2, XeF4, and XeF6 are shown in Figure 9. (The recently reported KrF2 p.e. spectra indicate that its electronic structure is very similar to that of XeF2). The MOs of predominantly F and Xe character are best separated in XeF2, but the 'fluorine bands' can easily be seen to grow in intensity as the proportion of fluorine in the molecule increases. It is noteworthy that the relative cross-section of the largely fluorine MOs appears to increase substantially in the He-II spectra. The first two bands in the XeF2 spectrum, both relatively sharp, represent the multiplet components of a 2Π ion state : the π MO involved must be relatively non-bonding, corresponding fairly closely to a πu (e1u) combination of xenon 'lone pairs'. The σ+u (a1u)combination of lone pairs is presumably intimately involved in the Xe-F σ-bonding. In the case of planar XeF4 there are formally two xenon lone pairs which yield alg and a2u symmetry combinations. The first band in the XeF4 spectrum probably concerns an antibonding a1g MO with considerable central-atom character, while the second somewhat sharper band arises from ionization of a weakly antibonding a2u MO of mainly xenon 5p character. In the spectrum of the approximately octahedral molecule XeF6 the first band is now very broad indeed and probably represents ionization of a strongly antibonding a1g MO formed from the single Xe lone pair A 0 and the fluorine a1g σ combinations.