M. Stener, S. Furlan and P. Decleva
Dipartimento di Scienze Chimiche, Uniersità di Trieste, Via L. Giorgieri 1, I-34127, Trieste, Italy
First published on 5th December 2000
The cross section and the asymmetry parameter profiles of C6H6 have been calculated for all the one-electron states, from the outer valence to the core C 1s. The theoretical method is based on the density functional theory, and consists in solving the Kohn–Sham equations for the explicit continuum wavefunction, within a single center basis set of B-splines functions. The employment of the LB94 exchange-correlation potential with the correct asymptotic behaviour gives an improvement of the results with respect to previous calculations. High energy features discovered in a recent experiment have been successfully reproduced by the theory. The core cross section is well reproduced by the theory and the observed features are ascribed to shape resonances and an assignment is proposed.
In this work we are interested in the following aspects of photoionization: the absolute intensity (cross section) and the angular distribution (asymmetry parameter) for the emitted photoelectrons of the main line peaks in the photoelectron spectrum as functions of the photon energy. The theoretical calculation of these two observables needs explicit evaluation of the unbound photoelectron wavefunction, since the intensity is related to the transition moment between the initial and final state wavefunction, and the latter must describe properly the emitted photoelectron. The asymmetry parameter is related to the unbound wavefunction phase-shifts with respect to the asymptotic solutions. The calculation of the unbound wavefunction for molecules is still far from being a routine task in quantum chemistry, due to the special boundary condition obeyed by such wavefunctions as well as to the basis set requirements, which are different from conventional bound-state calculations.
So far, the most commonly employed methods for calculating such properties on medium-sized systems are the Stieltjes theory (ST) imaging method1 and the continuum multiple scattering method (CMSM).2,3 In the ST approach, the problem of calculating the unbound orbital is actually circumvented, since the cross section is extracted from a discrete variational pseudo-spectrum which can be obtained with conventional well established quantum chemistry methods. The advantages are straightforward implementation and computational economy, and the possibility to include electron correlation in the computational scheme used to produce the discrete spectrum. The disadvantages are the low resolution on the energy scale, which prevents description of sharp or multiple structures, and the total preclusion of calculating the angular distribution. In particular, the low resolution is a very strong limitation, since the importance of the cross section profiles stems mainly from the presence of sharp features like the shape-resonances. The CMSM approach is a full-continuum treatment, but the effective potential employed is built with the muffin-tin (MT) approximation, which is rather crude and has been completely abandoned for bound states calculations. Despite this limitation, the CMSM approach is the only practicable way to study very large systems, and is still widely employed for the description of X-ray absorption spectra.4
A very promising new treatment of the molecular continuum is represented by the Lobatto technique,5–8 which joins the flexibility of localized Gaussian-type orbitals (GTO) with a single-centre expansion whose radial components are developed with Lobatto-shape functions, which are particularly suited to the proper treatment of the continuum boundary conditions. The Lobatto method is very powerful since it allows calculations on rather extended systems, like for example Ni(C5H5)NO9 without drastic approximation of the potential, as in the MT CMSM methods.
In this work we employ the B-spline one-centre expansion (OCE) density functional method,10 which has proven very efficient from small molecules,11,12 up to large systems, like C6013 and M@C60.14,15 This method has been recently updated, by employing the LB94 exchange-correlation potential,16 which displays the correct asymptotic Coulomb behaviour. In fact it has been recently pointed out that the correct asymptotic behaviour of the potential improves the results for some properties which are sensitive to the regions of space at large distances, like for example the cross section itself,17 or the dynamic polarizability18 calculated with the time-dependent version of the theory (TD-DFT). It has been shown17 that for the cross section calculations the LB94 potential in the ground state (GS) configuration is, in general, a better choice than the previously employed Vosko–Wilk–Nusair (VWN)19 potential with the transition state (TS) configuration, which corresponds to half an electron removed from the ionized orbital. The arguments in favour of the LB94-GS choice are essentially two: better accuracy of the results and computational economy. This latter is due to the GS configuration choice, which is obviously the same irrespective of the orbital which is ionized, so that the same Hamiltonian is used in all the calculations with noticeable computational economy.
For this reason in this work we have decided to reconsider the C6H6 molecule with the LB94-GS potential: the profiles have been recalculated and extended to higher photon energy than in the previous study,10 in order to compare with more recent experimental data.20 Moreover, due to the more economic computational scheme, we have explored the ionization of all the orbitals, from the valence to the core C 1s, while in our previous work only the seven uppermost valence levels were considered.10
It is important to underline that in the present method the electronic structure is described in terms of a one-electron picture, according to the Kohn–Sham (KS) DFT approach.21 For this reason the phenomena related to many-body effects, like for example the intensity redistribution from the main line to satellites, cannot be described. However, these effects are not prominent in benzene, and are expected to be really important only for the inner valence ionizations, so we believe that this limitation is not relevant for the present study.
Finally, experimental data are rather complete for C6H620,22–24 so this theoretical study may help give better understanding of the spectroscopic measurements, and from the observed discrepancies some suggestions to improve the present method are gained.
The basis set consists of a OCE, factorized in a radial set of B-splines and an angular set of real spherical harmonics (Yl,mR). A given function may be expanded, if the basis is complete, according to:
(1) |
and the radial expansions Rl,m(r) are actually calculated by means of:
(2) |
Rl,m(r) is expanded in one-dimensional B-splines,25 while a further transformation26 adapts the real harmonics to the point group symmetry.
In the Hamiltonian matrix, the kinetic energy and the nuclear attraction terms are easily calculated, while electron–electron interaction is treated with a KS approach.21 The self-consistent field (SCF) electron density is obtained by a previous KS-LCAO calculation. The SCF density is then expanded ia eqn. (1) and (2), then the Hartree potential is computed by solving the Poisson equation. The exchange correlation potential is a function of the density and its gradient (LB94),16 so it is calculated from the SCF density and expanded with eqn. (1) and (2).
The bound states are calculated by means of a generalized diagonalization of the KS Hamiltonian matrix HKS, while the inverse iteration procedure27 is employed to get the continuum states. The cross section and the asymmetry parameter are finally obtained ia dipole transition moments and phase shifts.3
LB94 GS | ||||||
---|---|---|---|---|---|---|
OCE | ||||||
Orbital | ADF | LMAX = 30 | LMAX = 50 | LMAX = 70 | Exp. IP | ADF VWN TS |
a Ref. 22.b Ref. 30.c Result of a single, symmetry lowered, calculation to allow core hole localization. | ||||||
1e1g | −11.64 | −11.69 | −11.75 | −11.82 | 9.2a | −10.00 |
3e2g | −13.13 | −13.06 | −13.13 | −13.16 | 11.5a | −11.84 |
1a2u | −14.23 | −14.35 | −14.41 | −14.47 | 12.3a | −12.69 |
3e1u | −15.17 | −14.97 | −15.11 | −15.15 | 13.9a | −13.86 |
1b2u | −15.70 | −15.70 | −15.77 | −15.80 | 14.8a | −14.97 |
2b1u | −16.14 | −15.13 | −15.63 | −15.78 | 15.5a | −14.68 |
3a1g | −17.89 | −17.81 | −17.88 | −17.90 | 16.9a | −16.53 |
2e2g | −19.63 | −18.22 | −18.94 | −19.14 | 19.2a | −18.47 |
2e1u | −23.15 | −20.64 | −21.70 | −22.30 | 22.5a | −22.18 |
2a1g | −25.92 | −23.49 | −24.73 | −25.05 | 25.9a | −25.29 |
1b1u | −290.55 | −242.40 | −267.45 | −273.90 | 290.38b | −294.40c |
1e2g | −290.55 | −244.24 | −268.13 | −274.05 | 290.38b | −294.40c |
1e1u | −290.55 | −242.00 | −267.70 | −273.88 | 290.38b | −294.40c |
1a1g | −290.55 | −245.04 | −268.22 | −274.01 | 290.38b | −294.40c |
The comparison between the ADF and the OCE results is useful to assess the convergence of the OCE description, since the latter is strongly affected by the maximum angular momentum (LMAX) employed in the calculation. The convergence of the eigenvalues with LMAX is fairly good for the outer valence, with differences lying between 0.1 and 0.2 eV going from LMAX = 30 up to 70. For the inner valence states with C 2s atomic character (2b1u, 2e2g, 2e1u and 2a1g), the convergence is of course slower, owing to the pronounced localization of these states; the differences being less than 1.3 eV going from LMAX = 30 to 50, and less than 0.6 from LMAX = 50 to 70. It is interesting that all the orbitals with C 2s atomic character lie between the core C 1s states and the other outer valence states, with the exceptions of the 2b1u orbital which is pushed above the 3a1g level. For the core C 1s levels, the convergence is very slow, due to the extreme localization of the orbitals, and the eigenvalues decrease by more than 20 eV from LMAX = 30 to 50 and by about 6 eV from LMAX = 50 to 70. From this eigenvalues analysis it seems that only outer valence states are described with reasonable accuracy. However, in our previous work33 we have pointed out that the convergence of the cross section and asymmetry parameter may be much faster than that of the eigenvalues, so we have calculated the cross section and the asymmetry parameter profiles of the 3a1g and 1a1g levels with LMAX = 30, 50 and 70, and the results are reported in Fig. 1. It is quite apparent that the results obtained with LMAX = 30 are already convergent, at least for the purposes of the present work. Very small differences are observed between the results obtained with LMAX = 30 and 50, while the further basis enlargement to LMAX = 70 no longer changes the results appreciably and complete convergence is gained. In practice the largest deviation detected and attributable to the slow OCE convergence is the maximum of the shape resonance in the 3a1g cross section profile, which increases from 12.8 up to 14.2 and 14.5 Mb for LMAX = 30, 50 and 70 respectively.
Fig. 1 The cross section and the asymmetry parameter profiles of the 3a1g and 1a1g orbitals of C6H6: (——) LMAX = 30, (– – –) LMAX = 50, (···) LMAX = 70. |
Moreover, the core profiles display the same fast convergence as the valence ones, despite the much worse performances as regards the eigenvalues. For this reason we decided to employ LMAX = 30 in all further calculations, for which we expect the same convergence as found for the 3a1g and 1a1g levels.
Fig. 2 The cross section and the asymmetry parameter profiles of the 1e1g and 3e2g orbitals of C6H6: (——) present LB94-GS, (···) previous VWN-TS.10 Experimental data: (○) ref. 22, (♦) ref. 20. |
Fig. 3 The cross section and the asymmetry parameter profiles of the 1a2u and 3e1u orbitals of C6H6: (——) present LB94-GS, (···) previous VWN-TS.10 Experimental data: (○) ref. 22, (♦) ref. 20. |
Fig. 4 The cross section and the asymmetry parameter profiles of the 1b2u and 2b1u orbitals of C6H6: (——) present LB94-GS, (···) previous VWN-TS.10 Experimental data: (○) ref. 22, (♦) ref. 20. |
Fig. 5 The cross section and the asymmetry parameter profiles of the 3a1g and 2e2g orbitals of C6H6: (——) present LB94-GS, (···) previous VWN-TS.10 Experimental data: (○) ref. 22, (♦) ref. 20. |
Fig. 6 The cross section and the asymmetry parameter profiles of the 2e1u and 2a1g orbitals of C6H6: (——) present LB94-GS, (···) previous VWN-TS.10 Experimental data: (○) ref. 22, (♦) ref. 20. |
The cross section of the HOMO orbital 1e1g (Fig. 2) displays better correlation with experiment than the previous VWN-TS result. The difference is actually only significant within 3 eV above the threshold, while at higher energy the two profiles are almost indistinguishable. The observed difference between the two calculations is attributed to the more attractive character of the LB94-GS potential, which brings below the threshold the narrow shape resonance which was present in the VWN-TS calculation. Although the results are slightly improved, the discrepancy with respect to experiment is still important: in the experimental result the shape resonance lies to higher energy by about 6 eV and the profile is broader. It is reasonable to attribute this deviation to neglect of the screening effect due to the response of the system to the external field. In fact these effects can be considered by using TD-DFT:34,35 TD-DFT calculations with LB94-GS potential have recently been performed on N2 and PH3, giving excellent agreement with experiment.36 In particular, it has been observed that the main effect of inclusion of the response in the formalism is to shift the shape-resonances to higher energy with a concomitant broadening of the profile, which is the same discrepancy between the calculated and the experimental cross section observed here.
In contrast to the cross section, the calculated 1e1g asymmetry parameter profile shows a noticeable improvement with respect to the previous VWN-TS calculation: although the shape is very similar the agreement with experiment is now more quantitative, only near the threshold is the theoretical slope still too high. It will be noticed also in the analysis of the further profiles that the asymmetry parameter profile is more sensitive than the cross section to the potential choice, at least for energy higher than a few tens of eV. This is probably due to the fact that the cross section decays rather quickly with the photon energy and therefore the differences soon become too small to be noted.
For the 3e2g orbital, the LB94-GS cross section is too intense near the threshold, and the shape resonance is calculated too sharp, again here it is a typical discrepancy due to neglect of screening. Comparison with the VWN-TS calculation suggests that only a slight shift to lower energy due to the more attractive LB94 potential is actually observed. For the asymmetry parameter, the situation is more interesting, since the potential choice changes the shape of the profile more noticeably, which with the LB94-GS scheme performs better with respect to experiment.
The next orbital is the 1a2u, and for this state the cross section results do not compare favourably with experiment: the measured feature is completely absent since it is shifted below the threshold; the previous VWN-TS calculations gave a very similar low intensity. For the asymmetry parameter the situation is less negative: although the agreement is not quantitative, the shape resembles the experimental data. We are led to assign this deterioration also to screening effects, since these may redistribute the intensity above the threshold. There are a series of reasons which corroborate the attribution of the observed discrepancies between the theory and the experiment to screening effects: first, the screening effects are expected to be important only at low photon energy (few tens of eV), and this is in agreement with the good results obtained with the present method at higher energy (see Fig. 8, later). Moreover, the response of the system should influence the results mainly in the outer valence, in fact as we analyze the one-electron spectrum starting with the HOMO and going to lower levels, the screening effects should become less and less important. This finding is also consistent with the present results, in fact only the three highest orbitals shows important deviations with respect to experiment, while the inner ones perform much better. The next orbital 3e1u is the fourth one and the calculated cross section is in good agreement with experiment, in accord with the previous arguments about the minor importance of screening for the inner states. The previous VWN-TS curve is too intense, and it is interesting that the sharp shape resonance at 23 eV due to the ka2g continuum (compatible at least with angular momentum l = 4) and due to a valence virtual σ*(C–C) molecular orbital,10 is now shifted to lower energy (21 eV) and broader, while the two shape resonances just above the threshold are now unresolved but are calculated at the same position for both the potential choices. The asymmetry parameter profile compares equally favourably with experiment, with a decided improvement with respect to the previous VWN-TS calculation, especially for photon energy higher than 30 eV.
The next orbital 1b2u is in very good agreement with experiment (Fig. 4), both for the cross section and the asymmetry parameter. The LB94-GS potential improves the results in the energy region near the threshold in both curves.
The results concerning the 2b1u orbital are also displayed in Fig. 4. The present theoretical cross section profile is overestimated with respect to experiment in the 5 eV wide region above the threshold, while for higher energies the results match the experiment well. In the VWN-TS curve the overestimate is much larger, and refers to the same region near the threshold. The asymmetry parameter is nicely reproduced by the theory, also here the threshold region is more difficult to describe, but above 23 eV the agreement is good.
The results concerning the 3a1g ionization are reported in Fig. 5: the cross section is too sharp and shifted too near the threshold with respect to experiment. This discrepancy is very similar to that found for the three highest valence orbitals, and is similarly ascribed to screening effects, it is rather curious that this shows up again in this rather inner state. In any case, with the present potential the intensity at the threshold is reduced. The calculated asymmetry parameter reproduces the experimental pattern of the broad maximum around 35 eV, although it is calculated too high and a little shifted towards the threshold.
The next orbitals investigated belong to the inner valence ionizations (2e2g, 2e1u and 2a1g): they are characterized to have high C 2s contribution, this is reflected in the weak and slowly decreasing absolute cross section and in flat asymmetry parameter profiles (Fig. 5 and 6). The comparison with experiment is rather good in general, but it is important to consider that the presence of many satellites in the photoelectron spectrum20 makes this comparison less easy, since the intensity is redistributed over many states which cannot be described at the present level of theory, which is limited to a one-electron model.
We conclude this section by giving a general observation about the valence ionization profiles, which appear to be in good agreement with the experimental data. The most important discrepancies are found for the higher orbitals and have been ascribed to neglect of screening effects in the present computational scheme, apart from possible experimental inaccuracies.
This finding suggests that the efforts to extend the capabilities of the TD-DFT method to more extended systems are meaningful. On the other hand, it is worth noting that screening effects are beyond the single-particle picture, and therefore are less easily rationalized with simplified and straightforward models. For the present case, we are led to consider that the presence of many π electrons may enhance the polarizability due to the high electron mobility.
Fig. 7 The cross section and the asymmetry parameter profiles relative to the two unresolved bands 3e2g + 1a2u and 3e1u + 1b2u + 2b1u of C6H6: (——) present LB94-GS, (···) previous VWN-TS.10 Experimental data: (○) ref. 22. |
The cross section profile of the second band (3e1u + 1b2u + 2b1u) is in very good agreement with experiment, in this case it is also improved with respect to the previous VWN-TS calculation. For the asymmetry parameter, it seems that the region at about 20 eV is properly described, for higher energy values the increasing trend is also properly found, although a little overestimated.
Fig. 8 The asymmetry parameter profiles relative to all valence orbitals of C6H6 up to 120 eV in the photon energy scale: (——) present LB94-GS. Experimental data: (♦) ref. 20. |
It is worth noting that, apart from these three very structured profiles just discussed, there are some minor features which are present in the asymmetry parameter of the other orbitals and which are correctly reproduced by the theory. In the 1e1g orbital, the experimental data follow an increasing curve, but the slope decreases with two discontinuities at about 30 and 45 eV. The theoretical results are in good agreement with the experimental findings. Further interesting behaviour well predicted by the theory is the slope change at 80 eV found in the 2b1u ionization.
The 1a2u orbital is the only one for which the theory furnishes results which are in disagreement with experiment, and it is reasonable to attribute this deterioration to the fact that the peak is actually unresolved in the photoelectron spectrum, and therefore the experimental data may suffer from a large error due to the arbitrary deconvolution.
The presence of important features at high photon energy values (more than 40 eV) is very interesting, since it is widely believed that the asymmetry parameter profile reaches an asympototic value rather quickly (in a few tens of eV). Such features appear to be well described by the present theoretical approach, indeed a comparison on a wider energy scale reveals much closer agreement than can be expected on the basis of the near-threshold region alone. So it is expected that significant new information can be gained by exploring a wide energy range, now accessible with improved experimental facilities.
However, it is still very difficult to identify a simple physical mechanism which may originate these high energy features. As already pointed out in ref. 20, the resonances observed experimentally in the asymmetry parameter profiles can be ascribed to autoionization resonances or to shape resonances. We can exclude the presence of autoionization processes at high energy because the present calculation would not have found them, since it cannot describe this kind of process, and therefore the observed features must be ascribed to shape resonances. These latter are usually described in terms of excitations to virtual valence orbitals lying above the threshold, whose position on the energy scale may be roughly given by a minimal basis calculation. We performed a minimal basis calculation on C6H6 employing the LB94-GS potential, and the highest valence virtual state is calculated to be only 6.7 eV above the threshold, a value which excludes the presence of such a mechanism as the physical origin for the features at high photon energy. Therefore it is reasonable to attribute the high energy features to spatial anisotropy of the molecular effective potential, which may contain shallow minima and walls, where the continuum states may localize themselves to some extent within a given energy interval. It is worth noting that at high photoelectron energy, it is rather difficult to characterize the continuum states, in fact while for low energy the resonances may be attributed for example to waves with well defined angular momentum, at high energy the continuum is distributed uniformly over a rather wide set of angular momentum contribution, and therefore the physical origin of the resonance is somehow hidden as a result of very many contributions.
Although a precise assessment of these high energy features is still difficult, it is possible at least to identify whether they are connected to the nature of the ionized orbitals. It is quite easy to identify three important patterns in the calculated profiles.
First, the 1e1g and 1a2u profiles do not display important structures at the threshold, they increase rather rapidly within 30 eV and then they keep an almost constant value; this behaviour may be ascribed to the π nature of these orbitals. It is interesting that very similar behaviour has been observed experimentally for both π orbitals of butadiene.37 Moreover, this is actually another argument in favour of the theoretical 1a2u profile with respect to the experiment, this latter probably being affected by deconvolution errors. It is worth noting that this behaviour, typical of π ionizations, has been found also in C6H5Cl, C6H5Br and C6H5I.38–40
Second, the most structured profiles are 3e2g , 3e1u, 1b2u and 3a1g , which are attributed to C–C and C–H σ bonds, with mainly C 2p contribution. The close resemblance of the profiles relative to these ionizations with the corresponding ones in C6H5Cl, C6H5Br and C6H5I has already been pointed out in ref. 38. In butadiene37 the structures are not very evident, but at least the pronounced minimum at low energy is much wider than in the π orbitals. So we are led to attribute the features of C6H6 to the C 2p contribution of σ type.
Third, the remaining orbitals (2b1u, 2e2g, 2e1u and 2a1g) are essentially of C 2s nature, and their asymmetry parameter profiles are less structured. Their C 2s nature is also reflected by the rather low absolute cross section (Fig. 4–6).
In summary, the high energy region, different from what is usually assumed, is rather rich in features, as has already been found experimentally, and these features can be satisfactorily reproduced by the present theoretical model. Moreover, the features seem to be connected to the nature of the orbital with σ C 2p character.
It is important to stress that, at high photon energy, the asymmetry parameter profile is more suitable than the cross section to analyze the features, in fact the cross section falls down rather rapidly with increasing energy and it is difficult to identify tiny structures over a very weak background. However, to overcome this problem, it is possible to analyze the ratio between different profiles: this excludes normalization error and may enhance the features at high energy. This approach has been successfully employed, for example, to identify some oscillations in the ratio between the HOMO and the successive two inner bands in C60 , both in experiment41,42 and in the theory.13 We tried to perform the same analysis on C6H6; so in Fig. 9 the ratio σHOMO/σi is reported, where σHOMO is the HOMO (1e1g) cross section and σi is the cross section of the ith orbital. It is quite apparent that some features are now visible, although the differences are not so neat as for the asymmetry parameter, and therefore the results are not so interesting, for example, to assist in an assignment problem. The most original features are present in the 3a1g and 2a1g ionizations, with satisfactory agreement with experiment. Some oscillations are visible, especially in the theoretical 1a2u profile, but the wavelength is rather high (about 80 eV) so the presence of behaviour similar to C60 would need corroboration by an analysis to even higher photon energy.
Fig. 9 The ratio σHOMO/σi relative to all valence orbitals of C6H6 up to 120 eV in the photon energy scale: (——) present LB94-GS, (···) experimental data ref. 20. |
Fig. 10 The cross section and the asymmetry parameter profiles relative to the C 1s ionization of C6H6: (——) total C 1s contribution. Partial contributions from different initial states are: (···) 1a1g, (– – –) 1e1u, (–·–·–) 1e2g, (–··–··–) 1b1u. In the lower panel the energy scale is expanded and only the 1a1g contribution is shown. |
Fig. 11 Upper panel: total C 1s cross section in C6H6, (– – –) present calculated profile, (——) absorption experimental data,24 (○) and (●) main line photoemission experimental data,24 (■) main line + satellites photoemission experimental data.24 Lower panel: C 1s asymmetry parameter in C6H6, (——) present calculated profile, (○) main line experimental data,24 (+) main line experimental data,23 (●) satellite experimental data.24 |
Fig. 12 Final partial channel cross section profiles relative to each initial state which contributes to the C 1s ionization of C6H6: (——) total final contribution for each initial state. 1a1g initial state partial final contributions: (···) ka2u, (– – –) ke1u. 1e1u initial state partial final contributions: (···) ka1g, (– – –) ka2g, (–·–·–) ke1g, (–··–··–) ke2g. 1e2g initial state partial final contributions: (···) kb1u, (– – –) kb2u, (–·–·–) ke1u, (–··–··–) ke2u. 1b1u initial state partial final contributions: (···) kb2g, (– – –) ke2g. |
The last interesting feature in the cross section is the weak structure at 325 eV, which cannot be related to any valence virtual state, due to its very high energy position. This is in excellent agreement with the experimental observation, as regards both energy position and absolute intensity. Even a second weak feature is apparent around 360 eV. These features are the first terms of a series of EXAFS oscillations, which have been determined experimentally:24 the analysis of the EXAFS signal gave a very good estimate of the C–C bond distance. From Fig. 12 it is possible to attribute the features at 325 and 360 eV to shape resonances of type 1a1g → ke1u and 1b1u → ke2g respectively. It is worth noting that the features related to EXAFS oscillations have been already calculated with the OCE VWN-TS method in N233 and with its LCAO extension in Cl2,44 ClF and ClF3,45 for both Cl 1s and Cl 2p core ionization.
Even the absolute agreement with the absorption data reported in Table 2 is quite satisfactory, although there is a clear overestimate close to the threshold which reduces as the energy increases. This can also be a minor manifestation of the neglect of screening, which tends to redistribute the oscillator strength towards higher energies. On the whole the agreement with the experimental absorption spectrum is very satisfactory. The surprisingly sharp resonance close to the threshold comes out very clearly, as do the following structures. It has to be noted that in the latest experimental study,24 comparison with photoemission spectra led to assignment of the following resonances to two-electron processes giving rise to satellite states. It is not straightforward whether the present theoretical model, which takes into account only the primary hole final states, should be better compared to the photoabsorption (sum of all channels which originate from the primary hole) or to the single-hole photoemission channel. Comparison with experimental data clearly favours the former. What is however clear is that the presently calculated structures are one-electron processes (shape resonances) and are sufficient to give a full reproduction of the experimental features of the absorption spectrum.
Cross section/Mb | ||
---|---|---|
Photon energy/eV | Theory | Experiment |
300 | 12.794 | 9.592 |
325 | 5.687 | 5.015 |
350 | 4.317 | 3.843 |
375 | 3.718 | 3.447 |
400 | 3.002 | 2.989 |
The C 1s asymmetry parameter profile does not display important features, but is in general good agreement with experiment, apart from the region very near to the threshold, where the experimental data increase regularly and the calculated one is a little bit too high and shows weak structures. These latter are probably the counterpart of the shape-resonances observed in the cross section profile.
In the low photoelectron energy region of the valence ionizations, the present results are in better agreement with experiment than previous ones obtained at the VWN-TS level. Some discrepancies which are still present are ascribed to screening effects, which are not considered in the present theoretical scheme but are included in TD-DFT.34,35 The recent OCE implementation of the TD-DFT36 should soon be efficient enough to treat medium-sized systems like C6H6 and therefore to confirm such an hypothesis and cure the present discrepancies.
The high energy region of valence ionization has been calculated for the first time, and the agreement with experiment is quite good, with the only exception being the 1a2u orbital. The latter, however, is not completely resolved in the photoelectron spectrum, and therefore it is not clear if this is a fault of the theory or is due to an arbitrary deconvolution of the unresolved band. The high energy prominent features observed in the experimental asymmetry parameter profiles of some orbitals have been confirmed by the present calculation, this finding is interesting since it is generally believed that the asymmetry parameter should soon reach an asymptotic value. It is difficult to identify a precise physical origin for these features, since they are far away from the expected position of virtual valence orbitals obtained with minimal basis calculation. A comparison with previous experimental data on butadiene and monohalogenobenzenes seems to suggest that the high energy features might be related to σ C–C bonding orbitals with C 2p character.
The cross section of core C 1s ionization is in very good agreement with experimental data, and the experimental features have been ascribed to shape-resonances, with a different assignment with respect to a previous MS-Xα calculation. The first calculated term of the series of EXAFS oscillations has been found to be in excellent agreement with experiment.
In summary the LB94-GS exchange-correlation has proven, up to now, the best choice within DFT to give cross section and asymmetry parameter profiles which are in good accordance with experiment, for both valence and core ionization and over a very wide photoelectron energy range.
This journal is © the Owner Societies 2001 |