VUV excitation and electronic decay of rubidiumhalide molecules
A Calo1, E Kukk2, M Huttula1, E Nommiste3, V Kisand3, S Osmekhin1,H Aksela1 and S Aksela1
1 Department of Physical Sciences, PO Box 3000, 90014 University of Oulu, Finland2 Department of Physics, University of Turku, Turku, FIN-20014, Finland3 Institute of Physics, University of Tartu, Riia 142, 51014 Tartu, Estonia
Received 7 February 2007, in final form 8 May 2007Published 5 June 2007Online at stacks.iop.org/JPhysB/40/2261
AbstractPreviously we reported an interionic Auger decay following the VUV resonantexcitations of the metal atom in the alkali halide molecules. Opening andclosing of the spectator Auger decay channel was found to strongly influencethe lineshape of the participator decay transitions, an effect attributed to thechanges in the lifetime of the VUV-excited state. In this work, the VUVexcitation and the following electronic decay of the resonant states of rubidiumhalides is studied. A series of electron spectra have been measured at photonenergies around the 4p → nl resonance excitation region. Experimental resultsand the theoretical modelling of the excitation and decay spectra are presented.
1. Introduction
The spectroscopic studies of alkali halide molecules have been of great interest ([1, 2] andreferences therein) because of the strong ionic character of their molecular bond, allowing theapplication of simple theoretical models. The development of modern synchrotron radiationsources and high resolution detectors has opened new possibilities for more systematic analysis.The capability of selecting the excitation energy with high level of accuracy in a broad rangeallows us to investigate in detail how the electronic transition and the molecular fragmentationprocesses change with the excitation energy. In earlier studies, it was found that KCl [3]and CsCl [4, 5] monomers differ in cross-ionic electronic decay creating a peculiar molecularfragmentation pattern for different excitation energies. Similar studies have also been dedicatedto dimers ([6] and references therein) showing how the relaxation process changes across theexcitation energy. In previous publications, we reported that a cross-ionic type of Auger decayfollows the VUV resonant excitations of the metal atom in the alkali halide molecules [7], aprocess known also as interatomic or (intermolecular) Coulombic decay [8]. In the case ofCsCl, the valence photoelectron spectrum showed strong changes in its structure at photon
energies corresponding to the inner valence excitation resonance maxima. At these energies,the valence photoelectron spectrum has contributions from two channels—direct ionizationand the participator decay (autoionization) of the resonantly excited state. The lineshapechanges seen in CsCl spectra were interpreted as due to the participator channel, stronglymodified by the lifetime vibrational interference (LVI) effect. Therefore, the lifetime of theexcited state, as compared to the vibrational period of the molecule, plays a crucial role. Forexample, similar changes as for CsCl were not observed in KCl, which was attributed to theshorter lifetime of the resonance, causing the participator spectrum to resemble closely thedirect ionization lineshape.
In this paper, as a continuation of our previous work, we present a study of the electronicdecay of VUV excited resonant states of rubidium halides. A series of electron spectrawere measured at resonant and non-resonant energies around the Rb 4p → nl resonanceexcitation region in RbX (X = F, Cl, Br, I) molecules, with two main purposes: (i) to obtaina better insight into the character of the VUV-resonances from the comparison with ab initiocalculations and (ii) to obtain quantitative estimates on the lifetime of the resonant states andstudy its dependence on the photon energy. The rubidium halide series was chosen for thisstudy, since preliminary theoretical estimates predicted strong variations within this group inthe lifetime between the individual compounds and with the changes of the photon energy.
2. Experimental setup
Measurements were performed at the beamline 52 at MAX-I storage ring [9] in Lund,Sweden. In this beamline, the synchrotron radiation is dispersed by a 1 m normal incidencemonochromator working in the 5–30 eV energy range. The monochromatic light is refocusedthrough a differential pumping section by a toroidal mirror. Photoelectron spectra wererecorded at the magic angle of 54.7◦ relative to the polarization vector of the synchrotronradiation. For a detailed description of the experimental setup see [10]. Briefly, a commercialhemispherical electron energy analyser Gammadata Scienta SES-100 was used together witha resistively heated oven built at the University of Oulu. The spectrometer was operated at20 eV pass energy for a corresponding analyser contribution of approximately 70 meV.Samples were evaporated at the 720–820 K temperature range. The calibration of the electronspectra was performed using the Xe 5p photoelectron lines at 13.436 and 12.130 eV bindingenergy [11].
3. Results and discussion
3.1. Photoabsorption and Auger electron spectra
The absorption spectra were measured as partial-electron-yield (PEY) spectra of outer valencephotoelectron line (autoionization). The absorption spectra are presented as total-ion-yields(TIY) in [12]. The PEY spectra were normalized for the photon flux, but the intensity ratios ofthe peaks at different photon energies still remain distorted due to the changes in the electronanalyser’s transmission as the kinetic energy window follows the valence photoelectron lines.The experimental PEY results are presented in figure 1. The photoabsorption features aremainly related to the RbX monomers [12], as contributions from dimers is expected to besmall (i.e. less than 10%). The series of resonant peaks are assigned to the Rb+4p6 → 4p5nl
excitation [12]. A more detailed assignment for these structures is not possible since ourcalculations indicated an overlap of several excited states. A series of resonant Auger electronspectra (RAS) were then measured with photon energies tuned on and off the main resonances
VUV excitation and electronic decay of rubidium halide molecules 2263
(a) RbF
(c) RbBr (d) RbBI
(b) RbCI
Figure 1. Measured absorption spectra for (a) RbF, (b) RbCl, (c) RbBr and (d) RbI. The arrowsindicate the photon energy values for the RAS shown in figures 2–5.
throughout the photon energy range. A representative selection of the measured RAS is shownin figures 2–5. The shape of the resonance peaks measured throughout the photon energyrange presents remarkable differences. For RbF the off-resonance spectra (figure 2(d)) showonly a single relatively broad photoelectron line, which is followed by a rapidly rising lowkinetic energy background. Moving on the resonant energies (figures 2(a)–(c)), the singlepeak shape is replaced by structures with either two main maxima separated by about 1 eV oreven more complex structures up to almost 2 eV energy region. In the case of RbCl and RbBr,the differences between photoelectron spectra were less obvious. At low photon energies,the RbCl spectrum presents a pronounced asymmetry partially overlapping with the (RbCl)2
dimer valence peak (∼9.2 eV [13]). A careful analysis revealed that the asymmetry decreasesfor higher photoexcitation energies. In the case of RbBr, the two main peaks are partiallyoverlapping with the (RbBr)2 dimer valence peak (∼9.1 eV [13]), that appears on the higherbinding energy side (figure 4(a)). The shoulder disappears for photon energies higher thanapproximately 16 eV (see figures 4(b)−(d)). No noticeable differences were spotted in theRAS of RbI throughout the studied region (figures 5(a)−(d)).
3.2. Photoexcitation and Auger decay processes
The differences between RAS at different resonances described in the previous section andthe different behaviour of each molecule can be understood to result in the interplay betweenthe vibrational motion period of the excited neutral molecules and their lifetime. The lifetimeof a resonant state is directly connected with the possible decay pathways following the core
2264 A Calo et al
(a)
(b)
(c)
(d)
Figure 2. RbF RAS measured on resonance ((a), (b), (c)) and off resonance (d).
excitation (see figure 6). In the simple ionic model, we can describe the rubidium halidemolecules as a positive Rb+ ion and a negative halogen ion: RbX (Rb+4p6, X−nXp6). In thismodel, the resonant photoexcitation process can be schematically described as
VUV excitation and electronic decay of rubidium halide molecules 2265
(a)
(b)
(c)
(d)
Figure 3. RbCl RAS measured on resonance ((a), (b), (c)) and off resonance (d).
where an electron from the Rb 4p-like orbital is promoted to a higher Rb-like Rydberg orbital.The non-radiative electronic decay is energetically forbidden in Rb. The excited neutralmolecule can then decay only if the decay involves electron on the halide side of the molecule.The cross-ionic decay may take place through two possible pathways, participator Auger decay(autoionization) or spectator Auger decay. In the autoionization case, the decaying processcan schematically be described as follows:
Figure 4. RbBr RAS measured on resonance ((a), (b), (c)) and off resonance (d).
This process involves a cross-ionic electronic decay, where the core hole of the Rb 4p-likeorbital is filled with an electron from the more external halogen-like orbital. The initiallyexcited electron is emitted as Auger electron. In this process, the final state is the same as inthe direct valence photoionization:
VUV excitation and electronic decay of rubidium halide molecules 2267
(a)
(b)
(c)
(d)
Figure 5. RbI RAS measured on resonance ((a), (b), (c)) and off resonance (d).
Compared to the autoionization decay, this process also involves a cross-ionic decay, themain difference being in the emission of the Auger electron from the halogen atom. As shownin figure 6, the spectator decay leads to a final ionic state whose energy is higher comparedto the final state of the autoionization. Consequently, the neutral molecule has to be excitedto a higher-energy state, in order to make the spectator Auger decay possible. The final(valence-ionized) state energy, on the other hand, is largely determined by the halide atom,since the valence orbital is of halide np character. It is then reasonable to expect the activation
2268 A Calo et al
Figure 6. Energy level diagram (at ground state geometries) schematically representingphotoexcitation, direct photoionization, spectator and participator Auger processes.
energy for the spectator process to be strongly influenced by the halide atom substitutions inthe rubidium salts. As will be shown in the next section, the opening of the second decaypathway can have remarkable consequences on the shape of the resulting electron spectra andthe changes in the line shape correlate with the activation energies for the different molecules.
We note that both the participator and spectator decays lead to a final ionic states thatare likely to dissociate producing detectable ions. Considering equations (2) and (4) onewould expect emission of Rb+ ions and positive halogen ions as dissociation product of theparticipator and the spectator decay, respectively. This indeed was proven to be a usefultool to observe and separate the two processes. The experimental evidence of the final statesdissociation products was observed in [12] where the same RbX series was studied usinga Wiley–McLaren type TOF mass spectrometer [14, 15]. In that occasion, the analysis ofthe partial ion yield revealed the appearance of positive halogen ions (i.e. activation of thespectator decay channel) starting at different photon energies for different rubidium halidecompounds.
3.3. Lifetime and interference
If the photoabsorption spectrum shows that the peaks corresponding to different vibrationalstates strongly overlap, considering the core excitation and the following Auger decay as twoseparated events does not provide an accurate description of the process. In these cases,a treatment based on a single-step model (excitation and Auger decay) is required and theso-called lifetime vibrational interference (LVI) [16] effect needs to be considered. Thetheoretical background of the LVI effect has been thoroughly discussed in previous studies[16–18]. In this work, we give a brief summary as a background for the discussion on the nextsection.
If the photoexcitation process would be immediately followed by the Auger decay (i.e.the lifetime of the intermediate core excited state would be much shorter than the period ofthe nuclear vibrational motion), then all processes could be properly described as a singlescattering event [19]. If this is the case, the intermediate excited state would have virtually no
VUV excitation and electronic decay of rubidium halide molecules 2269
Figure 7. Diagram of RbCl PECs involved in the autoionization process. Population densityfunctions are also plotted for the ground and the resonant excited states.
influence on the lineshape of the RAS, which would be identical to the direct photoionizationline. If this is not the case, and the lifetime of the intermediate excited state is comparable tothe period of the nuclear vibrational motion, then the LVI effect would come into play (seefigure 7). The neutral core excited molecule would have time to relax and the Auger decaywould occur at different internuclear distances, resulting in a distortion of the lineshape, apossible shift of the peak position and a different intensity distribution among the vibrationalpeaks in the resonant Auger spectra. In the case-limit in which the intermediate state lifetimeis much longer than the nuclear vibrational motion, the two processes would be effectivelydecoupled (i.e. the neutral core excited molecule would have time to undergo several vibrationsbefore decaying). In this case, a theoretical approach based on the two-step model and theFranck–Condon factors calculated on the time independent nuclear wavefunctions would beable to properly describe the RAS lineshape with the resulting spectrum reflecting the nodalstructure of the vibrational excited state wavefunction.
In this framework one anticipates that the increase of the photoexcitation energy, andthe consequent opening of the spectator Auger decay channel, can have a direct effect on theshape of the RAS lines. With a second decay pathway available, one expects the intermediate
2270 A Calo et al
excited state lifetime to be shorter and consequently the LVI modification of the lineshapecould be detectable. As already suggested by Thomas and Carroll [20] and later discussedby Sorensen et al [21], the LVI effect can then be used for the estimation of the intermediateexcited state lifetime.
3.4. Calculation details
Ab initio self-consistent field (SCF) [22, 23] calculations of the potential energy curves(PECs) of the ground, excited and final ionic states have been performed using the quantumchemistry software package GAMESS [24, 25]. The aim was to theoretically describe thecomplex interplay between the electronic excitation and the vibrational motion in the coreexcited states and the effects of the opening of the spectator decay channel on their lifetime.Correlation consistent Aug-cc-pVTZ basis set was used for F [26, 27], Cl [28] and Br [29],the valence triple zeta with polarization 6-311G** basis set was used for I [30] and SadlejpVTZ basis set was used for Rb [31]. The neutral ground state was calculated using restrictedHartree–Fock (RHF) SCF calculations [32]. Configuration interaction-single excitation (CIS)calculations [33, 34] were used for the core excited states of the neutral molecule. The RHFcalculations of the ground states were used as a reference starting point for the single excitedconfigurations. The choice to exclude higher excited configurations in the configurationinteraction (CI) calculations was mainly due to the combined effect of two factors: the firstwas a more strictly computational problem related to the large number of determinants presentin the full CI calculations (i.e. convergence, computational time and memory requirements).The second problem was that multiple excitations would have led to a further relaxation of theexcited states, making in turn even harder to reach the desired core–hole states. Analogouscompromises had to be done for the calculations of the ionic states. In this case, restrictedopen-shell Hartree–Fock (ROHF) calculations [35] were performed. For the evaluation ofthe spectator Auger decay final state energies, occupation restricted multiple active spacecalculations (ORMAS) were performed [36, 37]. The advantage of this method is essentiallyin the reduced number of determinants present in a full CI calculation due to a partitioningof the active space and the arbitrary choice of the occupational condition for each partition.In this way, it was possible to include single excited configurations in the CI calculations forthe determinations of the spectator Auger decay energies. The state energies for differentinternuclear distances were fitted using a Morse potential energy curve. PECs of irregularshape because of the non-crossing effect have been discarded. PECs for dissociative stateshave been fitted using Morse potential with large equilibrium internuclear distances. Thenecessary vibrational excited states energies and wavefunctions were then evaluated. Furthertheoretical details on LVI calculations can be found in [16–18].
3.5. Results of calculations
3.5.1. Absorption spectra. The experimental results from this work and from [12] areillustrated in table 1 together with the theoretical results. The photon energy region of theobserved resonances is in general agreement with the calculated ranges, although the lattertend to be broader compared to the experimental values. The reason for this is related to thecovalent character of the highly excited MO and with the complete active space (CAS) [38]calculations that made the number of contributions rather large. Thus limiting the calculationsto include only excitation in the Rb side was not a straightforward task. An example is shownin figures 8(a)−(b). The two graphs differ by the fact that the first calculated spectrum includesall the single excitations from Rb 4p-like MO, as in the second the contributions from the
VUV excitation and electronic decay of rubidium halide molecules 2271
(a) (b)
Figure 8. Measured and calculated RbCl absorption spectra: (a) all the Rb 4p excited statesincluded; (b) only the Rb+4p → Rb+nl excited states included.
Table 1. Comparison between experimental and theoretical results (values are in eV). Theresonance region data refer to the observed/calculated energies for Rb 4p → nl resonances in RbX.The participator and spectator Auger energies refer to the observed/calculated photoexcitationenergies needed to activate the corresponding decay processes.
excitations to chlorine Rydberg orbitals were removed. Note that, as mentioned before, theionic picture seems to be no longer adequate for the highly excited states, and the terminologysuch as alkali or halide type of MO might be confusing. In our calculations, the MO of theinitial and final states of excitations were described as a contribution of atomic orbitals (AO).In this work, the MO was classified as halogen or alkali type if the corresponding AO madeat least 70% of the contributions. Our calculations showed also that the excitations to thehalide type of MO seem to lead to a highly dissociative state where the two atoms lose theirionic character and the molecule is likely to dissociate without producing any charged particle.The decay of the excited state would then take place throughout radiative process in one orboth the neutral atoms. If this is the case the product would then not be visible in neither theelectron nor the ion TOF spectrum. As shown in figure 8(b) the ‘corrected’ spectrum turnedout to much better represent the experimental results. The same effect was observed in all thestudied molecules, with the correction being stronger for heavier halide atoms.
3.5.2. Auger decays and activation energies. Comparing the measured and the calculatedabsorption spectra we selected the areas where we had better correspondence and where wecould identify a series of theoretical on-resonance photoexcitation energies. For a givenexcited state, it was then possible to calculate a series of RAS whose shape would depend onthe value for the lifetime of the intermediate state used in the computation. Comparing thecalculated and the measured spectra for different photon energy, we were then able to follow
2272 A Calo et al
(a)
(b)
Figure 9. Calculated RbF RAS. See text for details.
Table 2. The intermediate state lifetime widths which give best fits between simulated andexperimental Auger electron spectra. All values in eV.
the trend of the theoretical lifetime value. Results are summarized in table 2. Two examplesof the calculated RAS for RbF and RbCl are shown in figures 9 and 10, respectively. Thechoice of the photon energies has been made to avoid energy states whose PECs shapes wereaffected by non-crossing effect.
In order to reproduce the shape for RbF at the resonant energy of 15.85 eV (figure 2(a))the lifetime width value of 0.055 eV was used (figure 9(a)). A very similar value, 0.05 eV(figure 9(b)), was used in order to reproduce the shape of the RAS measured at the resonantenergy of 19.60 eV (figure 2(c)). For RbF, it was found that the shape of the measured RAS wasbetter reproduced by values of the intermediate state lifetime that did not change considerablywithin the studied photon energy region.
In the simulations for RbCl a different behaviour was noticed. For the RAS measured atphoton energies lower than 18 eV (i.e. figure 3(a)), the intermediate state lifetime widthvalues that better reproduced the experimental spectra were found to be approximately0.03 eV (figure 10(a)). For higher excitation energies the RAS (i.e. figure 3(c)) were found
VUV excitation and electronic decay of rubidium halide molecules 2273
(a)
(b)
Figure 10. Calculated RbCl RAS. See text for details.
to be better represented by shorter intermediate states lifetime width values of approximately0.06 eV (figure 10(b)).
For RbBr, the analysis revealed a behaviour similar to the RbCl, with the simulated spectraproviding an estimation of 0.03 eV for the intermediate state lifetime width for lower photonenergies and the broadening of the lifetime width value at approximately 0.08 eV for higherphoton energies. The simulations indicated that the change occurs around 16 eV. For RbI thesimulations indicated a behaviour similar to RbF, with the intermediate state lifetime widthvalues of approximately 0.05 eV across the studied photon energy.
The complex structures of the RbF on the resonance spectra are a clear signal that thereis no obvious lifetime shortening to a value much smaller than the vibrational period throughthe photon energy range and typical peak shapes from LVI effects can be seen. In the caseof RbCl the dimer signal unfortunately partially overlaps with the tail towards lower kineticenergies due to the LVI distortion of the peak shape. The effect of the opening of the spectatorchannel decay is nevertheless evident from our calculation results with a clear shortening of theintermediate state lifetime for the spectra measured above 18 eV. For RbBr, the effect is weakbut visible. Although the double-peak structure consequence of the Stark effect [13] and thepresence of the close-lying dimer complicates the analysis of the measured RAS, neverthelessa shoulder of the peak in the lower binding energy side before the dimer peak was detected.The shoulder disappears for higher photon energies, approximately above 16 eV, as expected.For RbI we have a situation similar to the RbBr but no change was found in the peak shapes.This in agreement with the fact that no spectator channel is available in the considered energyrange and consequently no shortening of the intermediate excited state lifetime should occur.
2274 A Calo et al
4. Conclusions
The experimental absorption spectra for the RbX(X = F, Cl, Br, I) series have been describedby theoretical calculations that confirmed the dominantly atomic rubidium nature of theobserved excitations. The line shapes in the electron spectra, following these resonantexcitations, were measured and compared with theoretical spectra. The latter shapeswere calculated using theoretical potential energy curves and vibrational spacing, with theintermediate state lifetime as an adjustable parameter. The comparison indicates clearshortening of the lifetime above the indicated activation energy of the spectator Auger decayand allowed us to give quantitative estimates of the lifetime. This work allowed us to testthe limit of the ionic model for this kind of molecules also in combination with the usedcomputational method. It also showed that the theoretical model for the LVI effect is able todescribe the major features of the experimental evidence and can be used for evaluating thechanges of the excited states lifetime, which is not directly observable from the experimentalline shapes dominated by the vibrational broadening.
Acknowledgments
This work has been supported by the Research Council for the Natural Science of the FinnishAcademy, the Estonian Science Foundation (grant 6536), the Tauno Toning foundation, andthe European Community—Research Infrastructure Action under the FP6 ‘Structuring theEuropean Research Area’ Program (through the Integrated Infrastructure Initiative ‘IntegratingActivity on Synchrotron and Free Electron Laser Science’). The staff of MAX-lab areacknowledged for assistance during the measurements.
References
[1] Davidovits P and McFadden D L 1979 Alkali Halide Vapors: Structure, Spectra and Reaction Dynamics(New York: Academic)
[2] Price W C 1979 Electron Spectroscopy: Theory, Techniques and Applications ed C R Brundle and A D Baker(New York: Academic) p 151
[3] Kukk E, Huttula M, Aksela H, Aksela S, Nommiste E and Kikas A 2003 J. Phys. B: At. Mol. Opt. Phys. 36 L85[4] Kisand V, Kukk E, Huttula M, Koivukangas A, Aksela H, Nommiste E and Aksela S 2003 J. Phys. B: At. Mol.
Opt. Phys. 36 3909[5] Kisand V, Kukk E, Huttula M, Koivukangas A, Aksela H, Nommiste E and Aksela S 2004 J. Phys. B: At. Mol.
Opt. Phys. 37 3011[6] Kukk E, Huttula M, Rius J, Riu I, Aksela H and Aksela S 2004 J. Phys. B: At. Mol. Opt. Phys. 37 2739[7] Pennanen V, Huttula M, Aksela H, Nommiste E and Aksela S 2001 J. Electron. Spectrosc. Relat. Phenom.
114–116 169[8] Averbukh V and Cederbaum L S 2005 J. Chem. Phys. 123 204107[9] Sorensen S L, Olsson B J, Widlund O, Huldt S, Johansson S-E, Kallne E, Nilsson A E, Hutton R, Litzen U and
Svensson A 1990 Nucl. Instrum. Methods Phys. Res. A 297 296[10] Huttula M, Heinasmaki S, Aksela H, Kukk E and Aksela S 2007 J. Electron. Spectrosc. Relat. Phenom. at press[11] Cardona M and Ley L 1978 Photoemission in Solids: I. General Principles (Berlin: Springer)[12] Kisand V, Nommiste E, Kukk E, Calo A, Aksela H and Aksela S 2005 J. Electron. Spectrosc. Relat. Phenom.
144–147 175[13] Potts A W, Williams T A and Price W C 1974 Proc. R. Soc. London, Ser. A 341 147[14] Wiley W C and McLaren I H 1955 Rev. Sci. Instrum. 26 1150[15] Huttula M, Harkoma M, Nommiste E and Aksela S 2001 Nucl. Instrum. Methods Phys. Res., Sect. A
467–468 1514[16] Cesar A and Agren H 1992 Phys. Rev. A 45 2833[17] Kukk E, Bozek J D, Cheng W-T, Fink R F, Willis A A and Berrah N 1999 J. Chem. Phys. 111 9642[18] Kukk E, Snell G, Bozek J D, Cheng W-T and Berrah N 2001 Phys. Rev. A 63 062702
VUV excitation and electronic decay of rubidium halide molecules 2275
[19] Aberg T 1967 Phys. Rev. 156 35[20] Thomas T D and Carroll T X 1991 Chem. Phys. Lett. 185 31[21] Sorensen S L, Fink R, Feifel R, Piancastelli M N, Bassler M, Miron C, Wang H, Hjelte I, Bjorneholm O and
Svensson S 1982 Phys. Rev. A 64 012719[22] Almof J, Faegri K and Korsell K 1982 J. Comput. Chem. 3 385[23] Haser M and Ahlrichs R 1989 J. Comput. Chem. 10 104[24] Schmidt M W et al 1993 J. Comput. Chem. 14 1347[25] Gordon M S and Schmidt M W 2005 Advances in electronic structure theory: GAMESS a decade later Theory
and Applications of Computational Chemistry: The First Forty Years ed G E Scuseria (Amsterdam: Elsevier)[26] Dunning T H Jr 1989 J. Chem. Phys. 90 1007[27] Kendall R A, Dunning T H Jr and Harrison R J 1992 J. Chem. Phys. 96 6796[28] Woon D E and Dunning T H Jr 1993 J. Chem. Phys. 98 1358[29] Wilson A K, Woon D E, Peterson K A and Dunning T H Jr 1999 J. Chem. Phys. 110 7667[30] Glukhovtsev M N, Pross A, McGrath M P and Radom L 1995 J. Chem. Phys. 103 1878[31] Sadlej A J 1988 Collect. Czech. Chem. Commun. 53 1995[32] Roothan C C J 1951 Rev. Mod. Phys. 23 69[33] Foresman J B, Head-Gordon M, Pople J A and Frisch M J 1992 J. Chem. Phys. 96 135[34] Shroll R M and Edwards W D 1997 Int. J. Quantum Chem. 63 1037[35] Bobrowicz F W and Goddard W A 1977 Modern Theoretical Chemistry vol 3 ed H F Schaefer III (New York:
Plenum) chapter 4[36] Ivanic J 2003 J. Chem. Phys. 119 9364[37] Ivanic J 2003 J. Chem. Phys. 119 9377[38] Roos B O 1987 Adv. Chem. Phys. 69 339
