Journal of Electron Spectroscopy and Related Phenomena 141 (2004) 171–181
Auger electron–ion coincidence studies to probe molecular dynamics Denis C´eolina,b,c , Catalin Mirona,d,e , Marc Simona,d,f , Paul Morine,∗ a
d
Laboratoire LURE, Bˆat. 209D, BP 34, Universit´e Paris-Sud, 91898 Orsay Cedex, France b Department of Physics, Uppsala University, Box 530, S-751 21 Uppsala, Sweden c MAX-Lab, University of Lund, Box 118, S-221 00 Lund, Sweden CEA/DSM/DRECAM/SPAM and Laboratoire Francis Perrin, CNRS URA 2453, Bˆat. 522, CE Saclay, 91191 Gif-sur-Yvette Cedex, France e Synchrotron SOLEIL, L’Orme des Merisiers, Saint-Aubin, BP 48, 91192 Gif-sur-Yvette Cedex, France f Laboratoire de Chimie Physique-Matière et Rayonnement, CNRS, 11 Rue Pierre et Marie Curie, 75231 Paris Cedex 05, France Received 9 April 2004; received in revised form 4 June 2004; accepted 8 June 2004
Abstract Electron analysis combined with ion mass spectrometry is shown to be a unique tool to understand fragmentation dynamics of core-excited molecules. This article describes in detail a new setup devoted to energy and angle correlations measurements between the emitted particles resulting from inner-shell ionization or excitation. The data collection system is based on a pair of position sensitive detectors mounted behind a double toroidal electron analyzer and a short time-of-flight ion spectrometer. Because all relevant information results in time measurements, a natural synchronization in the events recording is obtained. The optimized geometry for the ion extraction allows spatial focusing for the ion trajectories by means of inhomogeneous extraction fields while preserving the time focusing. The N2 molecule has been used for full characterization of the setup whereas the CO2 molecule illustrates the role of the intermediate resonant state in controlling the final dissociation pattern. The bending mode excitation is shown to emphasize the O+ production, and the ion kinetic energy distribution is rationalized through an impulsive model. © 2004 Elsevier B.V. All rights reserved. Keywords: Double toroidal electron spectrometer; Position sensitive detection; Energy and angle correlations; Coincidence technique; Pulsed extraction field; Inhomogeneous field; Auger electron; Ion; N2 ; CO2
1. Introduction Combination of electron analysis and fragment ion mass spectrometry has been shown to bring a deeper insight into the understanding of the relaxation dynamics following coreexcitation in molecules. Indeed, after resonant excitation into the first empty molecular orbitals, the excited systems are likely to undergo strong geometrical changes allowing exploration of unusual regions of the potential energy surfaces of the molecular ions [1–3]. The goal of our studies is to probe the time evolution of the excited system in such situations. In order to be relevant for these complex physical situations, measurements have to be done in coincidence and should provide information on both the kinetic energies and ∗
Corresponding author. E-mail address:
[email protected] (P. Morin).
0368-2048/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.elspec.2004.06.014
ejection angles of the created charged particles. Indeed, the energy sharing among the various fragments is governed by the final electronic states reached through Auger decay and by the actual geometry of the resonant state. In practice, since high kinetic energies are involved when dealing with Auger electrons from simple molecules, an appropriate electron analyzer with high luminosity and good resolution at high energy has to be employed. In the present paper, we show how an optimized setup combining a double toroidal electron analyzer (DTA) and a short time-of-flight ion spectrometer, equipped with a position sensitive detector and operated with an inhomogeneous extraction field, is able to provide detailed information on the dissociation dynamics. In Section 2, we describe in detail the experimental upgrade of our setup with a special emphasis given to the ion detector optimization. We use the molecular nitrogen (N2 ) example to illustrate the ion position sensitive detector (PSD)
172
D. C´eolin et al. / Journal of Electron Spectroscopy and Related Phenomena 141 (2004) 171–181
performances. As a prototype example, the case of the linear CO2 molecule is then analyzed in great detail. Carbon Kedge excitation in CO2 molecule towards the first unoccupied molecular orbital (C 1s → ∗ ), leads to a doubly degenerated electronic state, which degeneracy is lifted through the bending motion, referred as the Renner–Teller effect. By choosing a series of narrow bandwidth excitation energies along the considered electronic transition, one is able to select “the amount” of bending or stretching motion to be excited in the resonant state. The nuclear motion in the core-excited state may thus be controlled in this way. Whereas the high resolution resonant Auger spectroscopy (RAS) highlights very rich structures reflecting these different vibrational contributions, the measurements in coincidence of an energy-selected Auger electron with an ion allow to correlate the final electronic states to the dissociation products. With this last kind of experiment, it is possible to probe in a very efficient manner the photoinduced geometry changes of the core-excited states, and their influence on the fragmentation of the simply charged ions.
2. The coincidence experimental setup (EPICEA) The results discussed in this article have been obtained over two experimental sessions. One was performed at the third generation national Swedish synchrotron radiation facility, MAX-Lab, on the undulator beamline I411 [4], and the other one at the SuperACO storage ring, LURE-Orsay, on the bending magnet SA22 beamline [5]. During the first session we have taken advantage of the higher resolution provided
by the I411 beamline to perform standard ES-RAEPICO (Energy Selected Resonant Auger Electron PhotoIon COincidences) experiments, whereas during the second experimental session we have used a new functionality added to our setup through an upgrade program, i.e. the possibility to measure, via a position sensitive detection on the ion side, the kinetic energies of selected atomic or molecular fragments in coincidence with energy selected Auger electrons. The EPICEA setup has already been described in [6] and is schematically shown in Fig. 1. Briefly, it consists in a double toroidal electron analyzer [7] coupled with a classical Wiley– McLaren 282 mm long time-of-flight mass spectrometer [8]. Differing from other Auger electron–ion coincidence setups [9–12], the particularity of our apparatus is the use of a pulsed extraction field for the ions in order to keep intact the kinetic energy resolution of the electrons before their detection. The photon beam crosses an effusive gas jet in the middle of the extraction region of the time-of-flight spectrometer. The electrons emitted into a cone of 54.7 ± 3◦ with respect to the symmetry axis of the DTA are retarded in a four element conical electrostatic lens [13] and are then energy dispersed through the deflecting plates of the analyzer. Two sets of deflectors are needed in order to obtain a focusing of the electron trajectories on a plane. The energy resolution of the electrons is about 0.4% of Ep (Ep being the pass energy of the analyzer), and may slightly vary as a function of the deceleration ratio between the initial kinetic energy and Ep . A position sensitive detector enables recording of the electron impact positions which, after appropriate data processing, provide the energy and emission angle of the collected electrons. The detection of the electron triggers an extraction pulse (rise time 20 ns,
Fig. 1. Schematics of the upgraded EPICEA setup.
D. C´eolin et al. / Journal of Electron Spectroscopy and Related Phenomena 141 (2004) 171–181
adjustable pulse height) which is applied in the first region of the time-of-flight mass spectrometer. The ion(s) are then accelerated, separated in time and resolved according to their mass/charge ratio, using the transformation t = a + b mq , where a and b are constants depending on the voltages applied to the spectrometer electrodes. In order to allow full determination of the ions momenta vectors and insure natural synchronization of the electron and ion detection, we have modified our data acquisition system so that determination of the relevant information ((X, Y) for electron impact, (X, Y, t) for the ion impact) is reduced to time measurements, for both the electron and ion detectors. Such an operation mode greatly simplifies the data acquisition process. A complete description of the new position sensitive detector used for the DTA electron spectrometer will be published elsewhere [14]. Shortly, particle localization is done with a delay line detector [15] coupled with a charge imaging technique. A spatial resolution better than 80 m and a dead time of 120 ns are achieved. This latter does not represent a limitation in the present case since measurement of electron hits which are very close in time is not required: typically, after an electron impact, the associated detector is gated during a few s corresponding to the ion time-of-flight in the ion spectrometer. However, for the ion detection the dead time issue becomes important since identical particles (same m/q ratio) are to be distinguished by their kinetic energy through their position measurement. Due to this requirement, the commercial DLD40 Roentdek detector is used instead. According to our measurements, the DLD40 provides a 40 ns dead time and a spatial resolution of about 300 m. The coincidence measurements between electrons and ion(s) are realized via a time to digital converter (model no. CTN-M4 from the IPNO) with the following characteristics: one com-
173
mon START channel, eight multi-hit STOP channels, 250 ps time resolution with a differential non-linearity (DNL) of less than 3% and a dead time of 20 ns. In order to optimize the ion position determination and to reduce the effect of the ionization source volume, we have developed a new time-of-flight mass spectrometer, based on the original Wiley–McLaren design [8], and completely adapted to the particular requirements of our measurements. The originality of the design consists in the use, for the acceleration region, of inhomogeneous fields, whose role is to focus on a same spot all of the ions having same p and m/q ratio, independently of their initial position in the source volume. A drawing of the new setup is shown in Fig. 2. The final geometry of the spectrometer has been obtained from numerical simulations as described in Section 2.1. 2.1. Numerical simulations for the ion imaging spectrometer The starting point for our numerical design process was a first-order Wiley–McLaren type spectrometer for which the focusing conditions lead to the following equation: 2d2 3/2 √ D = d1 k 1− , (1) d1 (k + k) 2 E2 where k = 1 + 2d d1 E1 , D, d1 , d2 , are the lengths of the drift tube, first and second regions, respectively, whereas E1 and E2 are the corresponding electric fields. In order to choose the parameters D, d1 , d2 , E1 and E2 for our design we have to take into account two important constraints. The first one concerns the position of the ionization region, which is fixed at 10 mm from the first element of the electrostatic lens of the electron spectrometer since
Fig. 2. Representation of the new time-of-flight mass spectrometer, with illustration of the focusing trajectories thanks to the presence of inhomogeneous fields in the second region. The ions are created in a rectangle of 3 mm side following the axis perpendicular to the spectrometer, and 2 mm along this axis.
174
D. C´eolin et al. / Journal of Electron Spectroscopy and Related Phenomena 141 (2004) 171–181
it acts as its focal point. Because a pulsed field is used to extract ions towards the flight tube, one has to account for their possible motion before the extraction field is turned on; it corresponds to a realistic value of 200 ns which accounts for both traveling time of the electron through the analyzer, and the characteristic rise time of the extraction pulse provided by the electronics. These two factors led us to choose a value of the d1 parameter of 14 mm. With such a value no discrimination against energetic ions is observed, and the source volume is located in the middle of the extraction zone. The choice of the lengths of the acceleration region and drift tube, as well as of the voltages applied, depends on the equation given above, but also on the focusing criteria in the two other space dimensions, i.e. in the plan perpendicular to the axis of the spectrometer. The use of inhomogeneous fields to focus particle trajectories issued from a diffuse source has already been described [15–17]. Let us also refer to the work of Takahashi et al. [18] who coupled two time-of-flight mass spectrometers, one used for the electron analysis and the other one for the ions, and whose electrodes have a special shape enabling to gain a factor of four in the source volume size. Another solution, proposed by Lebech et al. [19], uses an extra electrode inside the drift tube that modifies the trajectories of the particles and is particularly interesting since, besides the diminution of the source size, it also allows to √ obtain impact positions varying in a quasi linear way with Ekin , where Ekin is the kinetic energy of the ions. In our case, no inhomogeneous field is continuously applied in the first region as usually done for other setups, in order to preserve the kinetic energy resolution and the angular distributions of the electrons before their detection. We have thus chosen to adapt an electrostatic lens in the second part of the time-of-flight spectrometer. This point is of importance since we are obliged to use grids to isolate the first region from the second. As it has been shown in [16], the distortion of the particle trajectories passing through these grids leads only to a slight deterioration of the position measurement resolution. The inhomogeneous fields are induced by the size differences between the drift tube and the second electrode. The shape of the equipotential produces a focusing effect that can be directly compared to the one of an optical converging lens system. In order to control this effect, we have included an additional electrode whose potential can be independently adjusted to fit the equipotential in this second region. Space constraints in the chamber have imposed the distances d2 and D whereas the values of the electric fields E1 and E2 , for a fixed distance d1 , are selected according to Eq. (1). The last peculiarity is the use of large electrode sizes in order to avoid as much as possible the field penetrations from the second region to the interaction zone. Fig. 2 illustrates the final geometry of the ion spectrometer as resulting from numerical simulations. The chosen parameters, as resulting from the numerical simulations, are as follows: d1 = 14 mm, d2 = 30 mm, D = 282 mm, DT = 62 mm and Vext /Vtube = −0.23, where Vext and Vtube are the voltages applied to the extraction
grid and to the drift tube, respectively (E1 = Vext /d1 and E2 = Vtube /d2 since the second electrode is grounded). The trajectories of 14 amu ions shown in Fig. 2 illustrate the focusing role of the electrostatic lens: starting from different points within a rectangle of 3 mm × 2 mm, the ion’s emission angles varying from 0◦ to 360◦ by steps of 45◦ with respect to the spectrometer axis. The extraction field is switched-on 200 ns after the creation of the ions, and the different voltages are extracted from Eq. (1). The performances are evaluated by measuring the spot size in (a)–(c), corresponding to the impacts of the ions emitted at 0◦ and 180◦ (a), 45◦ and 135◦ (b) (equivalent to 225◦ and 315◦ ) and 90◦ (c) (equivalent to 270◦ ). Thanks to the focusing effect, the spot size is reduced from 3 mm (source volume size) to a size better than 300 m for point (a), to 200 m for point (b) and to 150 m for point (c). On the other hand, we have also checked that the timeof-flight peaks shape is not affected by the presence of the inhomogeneous fields over a larger amount of trajectories. The shapes and the widths of the time-of-flight peaks were found indeed to be identical to a conventional Wiley–McLaren type spectrometer which is designed to compensate for the source size (2 mm) in the direction parallel to the symmetry axis. An overall shift in time of less than 10 ns for the same set of voltages, has been revealed by the simulations. To test the performances of our upgraded setup (see Fig. 1), we used the nitrogen (N2 ) molecule excited along the transition N 1s →∗ above the ionization threshold. Fig. 3 shows the time-of-flight spectra of the N+ and N2 2+ ions (same m/q ratio) recorded in coincidence with electrons corresponding to the first electronic states of the doubly charged ion following normal Auger decay. The sharp central peak corresponds to the undissociated N2 2+ ion (no kinetic energy), whereas the extended feature corresponds to the energetic N+ fragments originating from the dissociative N+ + N+ channel. In this specific experiment, we did not use the appropriate extraction conditions imposed by the Wiley–McLaren criteria, which explains why the N2 2+ peak is not centred with respect to the broad structure corresponding to N+ as expected. By selecting only those
Fig. 3. Time-of-flight mass spectra of N+ ions (broad feature) and N2 2+ ions (narrow peak) recorded in coincidence with normal Auger electrons and corresponding to the lowest electronic states of the doubly charged N2 2+ ion.
D. C´eolin et al. / Journal of Electron Spectroscopy and Related Phenomena 141 (2004) 171–181
175
Fig. 4. Impacts on the ion PSD, selected by their time-of-flight (10 ns window). Left part: N+ and N2 2+ ions; right part: N+ only.
ions having the time-of-flight of N2 2+ , i.e. a combination of N+ and N2 2+ (due to the underlying background of the sharp N2 2+ peak), or alternatively, by selecting one of the extremities of the broad structure associated with N+ , we obtain the two detector images shown in Fig. 4. From the left panel of Fig. 4, we can completely distinguish the contribution of the N+ ions from the one of the N2 2+ ions. The explanation is found in the expressions of the time-of-flight (t) and radial impact position (r): t = t0 + K p cos θ,
and
r=
p sin θ t, m
(2)
where K is a negative constant depending on the extraction voltage, t0 is the time-of-flight of a zero kinetic energy particle, p its momentum, θ the emission angle referred to the spectrometer axis and m the fragment mass. Note that this r expression assumes a purely homogenous extraction field, parallel to the spectrometer axis. This is not totally true and a correction factor has to be applied in order to obtain more precise information (see Section 3.2). All the impacts shown in the left panel of Fig. 4 are issued from the same region of the time-of-flight peak, namely around t0 ; the only difference arises from the p cos θ term which is close to zero either because of p or cos θ. The central part of the image is associated with N2 2+ only, which carries only thermal kinetic energy (≈40 meV): t ≈ t0 and r ≈ 0 . The outer circle seen in the image is associated with energetic N+ fragments with a narrow p0 distribution, mainly oriented perpendicularly to the spectrometer axis (θ = 90◦ ): t ≈ t0 and r ≈ pm0 t0 . We are thus able to distinguish between the two components of the time-of-flight peak by simultaneously measuring the ion impact positions on the detector. The picture shown in the right panel of Fig. 4 is associated to a slice of the short time part of the time-of-flight peak (4420 ± 5 ns). Since no clearly apparent structure is visible on the image, such a shape may be associated with a distribution in p cos θ of the N+ ion where cos θ is positive as it corresponds to the short time part of the time-of-flight peak. The presence of a slight deformation of the ion images on their right side is related to the local field disturbance induced by the gas capillary which is relatively close to the interaction region in order to optimize the electron analysis conditions.
Fig. 5. Resonant Auger spectra of N2 recorded at hν = 401 eV with a pass energy of the DTA fixed at 80 eV.
Ion kinetic energy measurements have been performed using the dissociative ionization process following resonant (N 1s →∗ ) core-excitation of N2 . The kinetic energy released in the N+ fragments has been recorded as a function of the internal energy of the N2 + parent ion. Fig. 5 shows the resonant Auger spectrum of N2 following N 1s →∗ excitation, where we distinguish the participator states (below 22 eV) and the spectator states (above 22 eV). The spectrum is represented on a binding energy scale: Eb = hν − EkA , where hν is the excitation energy and EkA is the Auger electron kinetic energy. The most prominent participator states correspond to the first electronic states of the N2 + ion: the ˜2 ˜ 2 + X g formed at 15.58 eV according to [20] and the A u at 16.926 eV according to [21]. The spectator states are de˜ 2 + ˜2 + signed by the two terms C g and D g [22]. The first three + N dissociation thresholds have been determined [22] at 24.293 eV (N+ (3 P) + N(4 S)), 26.192 eV (N+ (1 D) + N(4 S)) and 26.676 eV (N+ (3 P) + N(2 D)). Figs. 6 and 7 describe the evolution of the fragmentation pattern, as well as the total kinetic energy released (KER) (i.e. the sum of the kinetic energies of all fragments), as a function of the N2 + ionic state energy. The neutral fragments N are not detectable by the method we employed, but in the case of a diatomic molecule, the measurement of the kinetic energy of the N+ ion represents half of the total KER. In Fig. 6, every time-of-flight mass spectrum corresponds to an electron energy slice of 200 meV, whereas in Fig. 7, every kinetic energy release distribution curve corresponds to a 300 meV electron energy slice. Moreover, it is important to note that in Fig. 7, the energy resolution is estimated via N+ 2 kinetic energy release curve (indicated by an arrow), which corresponds to the thermal energy of this ion. The increase in KER with increasing binding energy is clearly seen, as corresponding to the formation of N+ (3 P) and N(4 S) fragments. Since an energy dissipation by fluorescence or rotation is not expected on the considered time scale, the measured KER should increase linearly, when increasing the ionic state energy. It is exactly what is observed as shown by the dashed line which represents the expected linear behavior, and confirms a correct calibration procedure. Note that close to threshold,
176
D. C´eolin et al. / Journal of Electron Spectroscopy and Related Phenomena 141 (2004) 171–181
D∞h symmetry group and is linear in its ground electronic state. The electronic configuration is given below: (1 g )2 (1 u )2 (2 g )2 (3 g )2 (2 u )2 (4 g )2 (3 u )2 (1 u )4 (1 g )4 (2 u )0,1 + g.
Fig. 6. Coincidence spectra between energy-selected Auger electrons and ions. Such spectra allow following the evolution of the dissociation pattern of N2 + as a function of the ionic energy before dissociation (the spectator part of the resonant Auger spectrum is drawn in black).
Fig. 7. Variation of the kinetic energy released (KER) in the fragment ions as a function of the internal energy of the dissociating parent ion N2 + . In the considered energy range, the only open dissociation channel is: N+ (3 P) + N(4 S). The dashed line represents the expected linear energy dispersion.
this linear law is not valid because of the complex combination of both electron and ion energy resolutions. The appearance threshold for N+ production is obtained with a rather good accuracy as compared to the value given by [22].
3. Relaxation dynamics of core-excited CO2 As stated in Section 1, the full capabilities of our new setup have been used in a study of the CO2 molecule after resonant core excitation. The carbon dioxide molecule belongs to the
(3)
The transition probability of an electron from the 1 g orbital to the empty 2 u orbital is maximum at 291 eV, and the associated peak measured by total ion yield is broad (about 700 meV FWHM), without any strong apparent structures [23,24] . In order to understand the meaning of this width, we first have to describe the 1 u electronic state corresponding to the system with one hole in the carbon K-shell and one electron into the first empty molecular orbital. Within the D∞h group, the 1 u is doubly degenerate, but this degeneracy is lifted through a Renner–Teller effect due to the bending motion. As shown in [25,26], the 1 u state is described by three p-type orbitals centred on each atom, all of them pointing in the same direction. A bending motion in the plane of these three orbitals increases significantly their overlap, whereas a bending motion in the perpendicular plane has a less pronounced effect. The lowering of symmetry, from D∞h to C2v , splits the 1 u into two terms, 1 A1 and 1 B1 , the first being localized in the plane of the orbitals and the other one out of the plane. Within the “Z + 1” approximation, the core-equivalent molecule of the considered state, i.e. ONO, is bent, with an inter-bond angle evaluated at 134◦ [27]. This important result shows that the core-excited molecule is more stable in a bent conformation as compared to the linear one. Consequently, the 1 A1 state exhibits a minimum at lower energy than the 1 B1 . This explains the width of the C 1s →∗ resonance: the lowest part corresponds to the excitation of the bending mode, whereas the highest part is related to excitation of both symmetric elongation in the linear geometry as well as high bending modes in the bent geometry. The lack of structures in the excited states is due to the high density of vibrational levels of short lifetime having therefore an important natural broadening (Γ = 99 meV [28]). The vibrational spacings have been calculated with the help of the “Z + 1” approximation by Adachi et al. [27] and are estimated to be: ν1 = 185.2 meV for the A1 state, 156.1 meV (B1 ), ν2 = 101.7 meV (A1 ), ν3 = 239.5 meV (A1 ), and 243.2 meV (B1 ). The interpretation of the electronic decay from the bent/linear geometry is presented in [26]. A closer inspection of the high resolution resonant Auger spectra will be given in Section 3.1. In order to understand the effect of the nuclear motion in the resonant core-excited state on the fragmentation of the ion we first need to describe the dissociation limits of the first singly charged ionic states. The first electronic levels of the ion were found experimentally ˜ 2 u ), 18.077 eV ˜ 2 g ), 17.314 eV (A [29] at 13.778 eV (X 2 + 2 + ˜ ˜ (B u ) 19.394 eV (C u ). We have to compare these energies to the appearance thresholds of the various fragments O+ , CO+ and C+ . The thermodynamical values calculated by Locht et al. [30] are the following: O+ (4 S) + CO(1 + )
D. C´eolin et al. / Journal of Electron Spectroscopy and Related Phenomena 141 (2004) 171–181
Fig. 8. Low resolution resonant Auger spectra of CO2 obtained for three selected photon energies along the resonant profile together with the direct photoionization spectrum (hν = 280 eV) recorded with the EPICEA setup in coincidence mode, with the parent or fragment ions produced in the considered binding energy range.
at 19.069 eV, O(3 P) + CO+ (2 + ) at 19.465 eV, O(1 D) + CO+ (2 + ) at 21.436 eV, O(3 P) + CO+ (2 ) at 21.995 eV, + 2 ˜ 3 − 2O(3 P) + C+ (2 P) at 27.824 eV and O2 (X g ) + C ( P) at 22.708 eV. The high resolution resonant Auger decay spectra [23,26] indicate that the main contribution to the decay ˜ fiof the C 1s−1 2 u excited state is associated with the A ˜ ˜ nal state. Since the B and C final states are also populated through a direct photoionization process, their contribution has to be subtracted from the coincidence spectra, in order to emphasize the resonant contribution only. 3.1. Resonant Auger electron–ion coincidence measurements We show in Fig. 8 the resonant Auger spectra obtained in an ES-RAEPICO mode, with an energy resolution of about 300 meV, for a pass energy of the DTA fixed at 40 eV. This pass energy allows us to simultaneously record a 6 eV wide energy window, and thus to follow the evolution of the fragmentation pattern within the 15–21 eV binding energy range. ˜ The C-state energy, as extracted from the off-resonance spectrum (hν = 280 eV) shown in Fig. 8, is 19.4 eV, in good agreement with the high resolution data presented in [23,26]. It is important to note that in spite of the lower resolution of the coincidence measurements they reproduce well the overall shape of the Auger distributions measured by high resolution electron spectroscopy. The lower resolution obtained in the coincidence measurements is related to the particular design of our spectrometer which is optimized for a very high luminosity rather that for very high resolution measurements (see [7] for details). When tuning the excitation energy along the C 1s →∗ resonance profile, the main changes observed in the high resolution resonant Auger spectra are lo˜ cated in the region of the first vibrational levels of the A-state [23,26]. On the same set of high resolution data, the inten˜ 1 = 0–3, 0, 0)/C(0, ˜ sity ratio A(ν 0, 0) is much larger on the spectrum recorded at hν = 291 eV (top) than on the spectrum ˜ recorded at hν = 290.6 eV (left), the contribution of the C-
177
state being the same in both cases since it is non-resonant. This behavior is attributed to a strong vibrational excitation ˜ of A-state overlapping with the non-resonant contribution of ˜ the C-state. Finally, our coincidence measurements also reproduce the intensity change between the spectrum recorded at hν = 291.4 eV (right) and the top spectrum for the first ˜ which is observed in our spectra three ν1 quanta of A-state as a shift of the low energy side of right as compared to the two other measurements. A closer inspection of the high resolution resonant Auger spectra shows that, as it is already the case for the photoelectron spectrum [31], a large vibrational progression from ˜ ˜ the A-state mixes with the B-state towards 18 eV of binding energy. An interpretation of these vibrational distributions is given in [26] and is based on the high resolution work carried out by Baltzer et al. [31]. Briefly, for the top spectrum, the observed bands are mostly related to the progression of the symmetrical stretching vibrational mode ν1 , with a relatively weak contribution (lower than 20%) of the modes ˜ (ν1 , 2, 0) for the A-state. A strict attribution of each peak to a specific mode ν1 or ν2 is not possible with such a resolution since ν1 = 2ν2 . Following [23,26], we can assign the ˜ range between 18.5 and 19.4 eV (C-state maximum) of bind˜ ing energy to high vibrational levels of the A-state probably containing a strong symmetric stretching contribution. This point is reinforced by the fact that these levels are not populated in the off-resonance spectrum. The coincidence spectra shown in Fig. 8 have to be submitted to further analysis for at least two reasons. On the one hand, and this is valid for all coincidence measurements in general, reliable quantitative information is obtained only after subtracting the contribution of fortuitous coincidences. This contribution arises from uncorrelated events, whose intensity varies quadratically with the counting rate. On the other hand, the contribution resulting from direct photoionization events must also be subtracted since we want to focus on resonant effects only. Contribution from fortuitous coincidences has been quantified in a separate experiment where the event trigger (electron signal) is replaced by a random one (pulse generator). Such a contribution is subtracted after appropriate normalization. It has been found to account for less than 10% in all the measurements. After these treatments, the breakdown curves for a specific fragment at a given photon energy are obtained thanks to the measurement of the corresponding peak area from the coincidence mass spectra corresponding to the chosen binding energy range. In order to obtain relative values, the integral of each peak is divided by the sum of the integrals of all peaks, i.e. O+ , CO+ , CO2 + . The evolution of the branching ratios as a function of the binding energy and for the three considered excitation energies is illustrated in Fig. 9. As a test procedure of the whole data analysis process, let us concentrate on the CO+ production rate: no CO+ should be observed below its formation threshold at 19.47 eV. The increase in the production rate of CO+ at about 19.7 eV (see Fig. 9) is completely coherent with the value of the ther-
178
D. C´eolin et al. / Journal of Electron Spectroscopy and Related Phenomena 141 (2004) 171–181
Due to the improved resolution in the present measurements, the evolution of the fragmentation pattern with hν can be followed independently of the binding energy in more accurate way. We have, for instance, at 19.7 eV, the following relative production rates for O+ : 54% on the curve corresponding to the left spectrum, 48% on the top and 37% on the right. Also, let us notice that above 19.7 eV, the opening of the second fragmentation channel O(3 P) + CO+ (2 + ) induces the reduction of the production rate of the O+ ion. The explanation of these results, in the case of O+ as well as in the case of CO+ , would require a description of the potential energy ˜ surfaces around the C-state. The main ideas coming out from the experimental results above can be summarized in the following points: • The formation of the O+ ion is strongly affected by the geometry of the core-excited state. In particular, an excitation on the low energy part of the resonance (bending motion involved), facilitates the production of this ion in contrast to an excitation on the high energy part (stretching motion involved). • The formation of the CO+ ion is not sensitive to such geometry changes in the intermediate state. • The bending motion favors the dissociation process, since the CO2 + ion production rate decreases when going from high to low excitation energy. • The parent molecular ion is more stable when the stretching motion is excited, which is an unusual kind of behavior against molecular fragmentation. • Independently of the binding energy, the preparation mode (stretching versus bending motion in the core-excited state controlled by the excitation energy) has a significant effect on the dissociation of CO2 leading to a competition between the various fragmentation channels producing O+ , CO+ , and CO2 + ions.
Fig. 9. Evolution of the branching ratios as a function of the binding energy, and for the three photon energies excitation hν = 290.6, 291 and 291.4 eV.
According to the correlation diagram shown in Fig. 10, it is possible to associate the observed fragments to the first electronic states of the CO2 + ion in a linear geometry.
modynamical thresholds, and we can thus conclude that the few percents measured for energies lower than 19.47 eV are within the experimental uncertainty. According to these measurements, the relative quantity of CO+ is almost independent on the photon energy, as already stated in [23]. The linear or bent conformation of the neutral CO2 in the intermediate state has thus a minor influence on the production of this ion. The nature of the O+ production is more complex. We clearly see from Fig. 9 that the lower the photon energy (left excitation), the more this ion is abundant. By taking into account the previous remark concerning the independence of the formation of CO+ on hν, this means that the molecule dissociates more efficiently at lower excitation energy. These results are consistent with the previous measurements of Morin et al. [23] performed at lower resolution.
Fig. 10. Adiabatic correlation diagram of CO2 + ion and the corresponding dissociation channels [32].
D. C´eolin et al. / Journal of Electron Spectroscopy and Related Phenomena 141 (2004) 171–181
3.1.1. Formation of CO+ In the linear geometry, the only accessible dissociation ˜ 2 u state corresponds limit (21.43 eV) as starting from the A 1 + 2 + to the products O( D) + CO ( ), too high in energy to explain the observed production of CO+ . The lowest creation ˜ 2 g threshold of CO+ (19.47 eV) is correlated to two states, X and a˜ 4 u . The first one can not be really involved in the dissociation mechanism, since this would imply storing approximately 5 eV of ro-vibrational energy, which does not fit with the binding energy region we are exploring. Moreover, taking into account the shape of the resonant Auger spectrum in the ˜ X-state region it is obvious that only the first vibrational levels are populated. The second state, located above 21 eV (˜a4 u ), ˜ B ˜ ˜ and C-states crosses the A, but interacts via a spin–orbit interaction [33] only with the first one. This coupling (calculated by Pol´ak et al. [33] in the C2v geometry) between the corresponding states (4 B1 and 2 A1 ), is almost independent of the bending coordinate: ξ1s = 42.07 cm−1 for θOCO = 140◦ . This remark will also be necessary for the explanation of the O+ formation here below. 3.1.2. Formation of O+ As already mentioned before, the O+ production is more significant at the low excitation energy when the bending motion of the molecule in the core-excited state is preferentially excited. The explanation of this bending motion-mediated production of O+ around the first thermodynamical threshold requires a description in several stages. Initially, the only state correlated with the threshold O+ (4 S) + CO(1 + ) is a 4 − -state located approximately 4.9 eV above the C-state, ˜ g i.e. outside of the binding energy range we look at in this 4 study. In the linear geometry, this 4 − g -state crosses the u ˚ without any interaction [33]. However, due at rCO = 1.35 A, to the symmetry lowering in the Cs group caused by the molecular bending motion, these two last electronic terms are split into two 4 A able now to interact. Consequently, the crossing of the two potential curves, allowed in the C2v symmetry group, gives rise to a conical intersection in the Cs symmetry. The formation of O+ may thus be explained through the bending of the molecule in the intermediate state. Let us note that the interpretation given here takes into account the coupling between the asymmetric stretching (required for the dissociation) and bending modes, whereas in the simulated resonant Auger spectra provided by Kukk et al. [26], only the symmetric modes ν1 and ν2 were used to calculate the overlap of the vibrational wave functions in the intermediate and final states. 3.2. Kinetic energy released measurements Resonant Auger decay leads to singly ionized molecular states. By only selecting the lowest dissociative states, the KER will be particularly weak. Previous works have been devoted to KER determination from time-of-flight peak shape decomposition. However, such measurements cannot distin-
179
guish easily angular effects from KER distribution effects. Thus, the advantage of coupling the information given by the time-of-flight peak shape with the one provided by the measurement of the ions impact on the detector surface is obvious. The kinetic energy distribution of an ion of mass m may be written: Ekin
1 = 2m
mr 2 2 + [qEextr (t − t0 )] , α t
(4)
where α is a corrective factor used to compensate for the effects of the inhomogeneous fields in the acceleration region of the spectrometer, and whose value is optimized by numerical simulation, r is the radial impact coordinate on the position sensitive detector, t and t0 correspond, respectively, to the ion time-of-flight with and without kinetic energy, and Eextr is the extraction electric field. The value of the α parameter strongly depends on the kinetic energy range, but in our case the KER is very weak independently of the selected dissociation path. We have retained a value of 1.65, which gives a satisfying experiment-simulation agreement in the range from 0 to 600 meV. For the KER measurements, the binding energy of the singly charged system lies between 17.5 and 20.5 eV. The two accessible fragmentation pathways in this energy range correspond to O + CO+ and O+ + CO. Let us describe the system at the time of its dissociation, without taking into account the ionization degree of the fragments, or their final electronic state. If the dissociating molecule is linear, the available energy is found mainly in the form of kinetic energy for the oxygen atom and in the form of kinetic and vibrational energy for the carbon monoxide. On the other hand, in the case of a dissociation starting from a bent geometry, part of the energy is transferred to the CO fragment rotational motion. Taking into account the slice of binding energies we look at, we know that the vibrational excitation of CO will not be important; we can thus use the harmonic approximation and consider an equal spacing between the first vibrational levels in the CO fragment. For CO and CO+ , these spacings are estimated to be 270 and 274.5 meV, respectively. Note that electronically excited fragments are not formed, since the first limit implying an excited state (O(1 D) + CO+ (2 + )) is at 21.436 eV, above the end of our energy observation window. From these last two observations, we can evaluate the range of vibrational energy accessible for CO(1 + ): from v = 0 at 19.07 eV to v = 5 at 20.15 eV; and for CO+ (2 + ): from v = 0 at 19.47 eV to v = 3 at 20.29 eV. In order to obtain the kinetic energy distribution curves shown in Fig. 11, it was necessary to subtract the contribution of direct photoionization process (off-resonance) as well as the one of fortuitous coincidences, as explained in Section 3.1. However, the distribution in energy of the fortuitous coincidences is completely flat, and removing them will change only the baseline of the measured spectra, which is not critical for a qualitative interpretation. The KER of the O+ and CO+ ions taken in coincidence with all of the elec-
180
D. C´eolin et al. / Journal of Electron Spectroscopy and Related Phenomena 141 (2004) 171–181
Fig. 11. Kinetic energy release curves of the ions O+ and CO+ for an energy range from 17.5 to 20.5 eV in the parent CO2 + ion.
trons corresponding the binding energy range 17.5–20.5 eV and for each of the three considered photon energies, is plotted in Fig. 11. The KER resolution may be estimated through the kinetic energy distribution of the CO2 + ion, which corresponds to the thermal energy (approximately 40 meV). The first remark concerns the shape of the off-resonance energy distributions: for the O+ ion it is very similar to the distributions measured on the top as well as on both sides of the resonance. The presented curves being normalized to each other, the subtraction of the off-resonance curve from the resonant contribution is balanced by the ratio (off-resonance versus top, left or right) of the ionization cross-sections at the considered energies. The participation of the fragments resulting from direct ionization on the dissociation processes initiated via the core excitation will thus only slightly influence the observed distributions. For CO+ , the behavior is completely opposite, since, during the process of direct ionization, its kinetic energy is similar to the one of CO2 + . The explanation is found in the photoelectron spectrum recorded off-resonance and presented in Fig. 8: in the energy region located around the thermodynamical thresholds of the fragment formation O+ (4 S) (19.069 eV) and CO+ (2 + ) (19.465 eV), only the ˜ C-state is populated, in the ground vibrational level (0, 0, 0) mainly (at 19.394 eV of binding energy [29]). This state is known to be completely predissociated [34]. One can thus deduce that the O+ fragment will have a kinetic energy equal ˜ − Ethermo [O+ (4 S) + to the difference Ebe [CO2 + (0, 0, 0)C] 1 + CO(v = 0)( )], i.e. 325 meV. On the contrary, CO+ can˜ not be formed from the C-state (0, 0, 0) of CO2 + for thermodynamical reasons. Only excited vibrational levels of CO2 + (above 19.465 eV) can lead to the formation of this ion with a non-zero kinetic energy, but the probability of reaching them by a direct photoionization process remains extremely weak. The previous results indicate that after resonant coreexcitation, the kinetic energy released in O+ is independent on the value of the excitation energy. For the CO+ formation, the energy stored in the ion is more significant than in the nonresonant case, and a slightly different behavior is visible for an excitation on the right side of resonance as compared to the two other photon energies. Moreover, although the formation of the O+ and CO+ ions results from independent processes, the comparison of the observed distributions shows that the
energy carried-out by O+ is weaker than the one taken away by CO+ . Let us finally come back to the formation mechanisms of the two ions after resonant excitation. The bending of the molecule in the intermediate state is necessary to the formation of O+ , whereas the production of CO+ is weakly influenced by the geometry of this core-excited state. Since O+ has low kinetic energy, it is possible to form CO with high ro-vibrational excitation. On the other hand, CO+ seems to be formed in low vibrational levels, leaving thus all of the excess energy as translational energy. The weak difference in the distribution linked to an excitation on the right side of the resonance may be explained as follows: at this photon energy, the molecular geometry in the core-excited state is likely to be linear, and if the molecule remains linear during the decay process, the asymmetrical elongation mode allows the transfer of an additional vibration quantum into CO+ . This process would be less effective in a bent geometry, since the energy transfer would be done rather to the rotational excitation instead; the stored vibrational energy would be then weaker than in the case of a linear geometry dissociation. This model is close to the impulsive model proposed previously [35,36] to explain the energy sharing between translational, stretching and bending excitation energies following photodissociation of the ICN molecule.
4. Conclusion We showed in this article, through the example of the coreexcited carbon dioxide molecule, the significant influence of the preparation mode of the CO2 + ion, namely the coreexcited state geometry, on its dissociation. Via coincidence measurements between a kinetic energy selected Auger electron and an ion, we have been able to probe the influence of the bending mode in the intermediate state, on the dissociation of CO2 + created after Auger decay. The obtained results are unexpected since the formation of O+ , as opposed to the CO+ formation, is very much affected by the geometry of the intermediate state. Moreover, we showed that in this particular case the bending motion in the core-excited state favors the fragmentation, and also that the stretching motion stabilizes the parent ion. The results obtained with
D. C´eolin et al. / Journal of Electron Spectroscopy and Related Phenomena 141 (2004) 171–181
the help of the kinetic energy released measurements show that, for the O+ (4 S) + CO(1 + ) fragmentation pathway, the O+ ion carries very little energy without any remarkable effect of the intermediate state geometry, as opposed to the O(3 P) + CO+ (2 + ) pathway, where the ion carries a kinetic energy which is dependent on the geometry of the system with a core hole. Combination of high resolution measurements on both electron and ion analysis provides with an unique tool to investigate dissociation dynamics on short (fs) and medium (ns) time scales. Further improvements of our setup are on the way to increase performances by reducing, for instance, the source volume size, the time analysis and the sample temperature. Acknowledgements We are very grateful to LURE and MAX-Lab staffs for smooth operation of the storage rings and help during the experiments. G´erard Chaplier is warmly thanked for the design and setup of the new position sensitive detector, as well as Stacey Sorensen and Andreas Lindgren for their help during the measurements at MAX-Lab. European Community is gratefully acknowledged for financial support of the whole team for the MAX-Lab experiments within the Access to Research Infrastructure action of the Improving Human Potential Programme. References [1] C. Miron, M. Simon, P. Morin, S. Nanbu, N. Kosugi, A. Naves de Brito, M.N. Piancastelli, O. Bj¨orneholm, R. Feifel, M. B¨assler, S. Svensson, J. Chem. Phys. 115 (2001) 864. [2] C. Miron, R. Feifel, O. Bj¨orneholm, S. Svensson, A. Naves de Brito, S.L. Sorensen, M.N. Piancastelli, M. Simon, P. Morin, Chem. Phys. Lett. 359 (2002) 48. [3] K. Ueda, Surf. Rev. Lett. 9 (2002) 21. [4] M. B¨assler, A. Ausmees, M. Jurvansuu, R. Feifel, J.-O. Forsell, P. de Tarso Fonseca, A. Kivima¨aki, S. Sundin, S.L. Sorensen, R. Nyholm, O. Bjo¨orneholm, S. Aksela, S. Svensson, Nucl. Instrum. Methods Phys. Res. A 469 (2001) 382. [5] E. Delcamp, Caract´erisation du monochromateur PGM 12 m`etres SA22, LURE Internal Technical Report, 1995. [6] P. Morin, M. Simon, C. Miron, N. Leclercq, J. Electron. Spectrosc. Relat. Phenom. 93 (1998) 49.
181
[7] C. Miron, M. Simon, N. Leclercq, P. Morin, Rev. Sci. Instrum. 68 (1997) 3728. [8] W.C. Wiley, I.H. McLaren, Rev. Sci. Instrum. 26 (1955) 1150. [9] W. Eberhardt, et al., Phys. Rev. Lett. 58 (1987) 207. [10] D.M. Hanson, et al., J. Chem. Phys. 93 (1990) 9200. [11] K. Ueda, et al., Phys Rev. A 52 (1995) 1815. [12] E. Kukk, et al., Phys. Rev. A 66 (2002) 012704. [13] K. Le Guen, D. C´eolin, R. Guillemin, C. Miron, N. Leclercq, M. Bougeard, M. Simon, P. Morin, A. Mocellin, F. Burmeister, O. Bj¨orneholm, A. Naves de Brito, S.L. Sorensen, Rev. Sci. Instrum. 73 (2002) 3885. [14] D. C´eolin, et al., in preparation. [15] J.H.D. Eland, Meas. Sci. Technol. 5 (1994) 1501. [16] A.T.J.B. Eppink, D.H. Parker, Rev. Sci. Instrum. 68 (1997) 3477. [17] M. Lavollee, Rev. Sci. Instrum. 70 (1999) 2968. [18] M. Takahashi, J.P. Cave, J.H.D. Eland, Rev. Sci. Instrum. 71 (2000) 1337. [19] M. Lebech, J.C. Houver, D. Dowek, Rev. Sci. Instrum. 73 (2002) 1866. [20] T. Trickl, E.F. Cromwell, Y.T. Lee, A.H. Kung, J. Chem. Phys. 91 (1989) 6006. [21] P. Baltzer, M. Larsson, L. Karlsson, B. Wannberg, M. Karlsson G¨othe, Phys. Rev. A 46 (1992) 5545. [22] J.H.D. Eland, E.J. Duerr, Chem. Phys. 229 (1998) 13. [23] P. Morin, M. Simon, C. Miron, N. Leclercq, E. Kukk, J.D. Bozek, N. Berrah, Phys. Rev. A 61 (2000) 050701. [24] Yoshida, et al., Phys. Rev. Lett. 88 (2002) 083001. [25] Y. Muramatsu, K. Ueda, N. Saito, H. Chiba, M. Lavoll´ee, A. Czasch, T. Weber, O. Jagutzki, H. Schmidt-B¨ocking, R. Moshammer, U. Becker, K. Kubozuka, I. Koyano, Phys. Rev. Lett. 8 (2002) 133002. [26] E. Kukk, J.D. Bozek, N. Berrah, Phys. Rev. A 62 (2000) 032708. [27] J. Adachi, N. Kosugi, E. Shigemasa, A. Yagishita, J. Chem. Phys. 107 (1997) 4919. [28] T.X. Carroll, J. Hahne, T.D. Thomas, L.J. Sæthre, J.D. Bozek, N. Berrah, E. Kukk, Phys. Rev. A 61 (2000) 042503. [29] L. Wang, J.E. Reutt, Y.T. Lee, D.A. Shirley, J. Electron. Spectrosc. Relat. Phenom. 47 (1988) 167. [30] R. Locht, M. Davister, Int. J. Mass Spectrom. Ion Phys. 144 (1995) 105. [31] P. Baltzer, F.T. Chau, J.H.D. Eland, L. Karlsson, M. Lundqvist, J. Rostas, K.Y. Tam, H. Veenhuizen, B. Wannberg, J. Chem. Phys. 104 (1996) 8922. [32] M.Th. Praet, J.C. Lorquet, G. Rassev, J. Chem. Phys. 77 (1982) 4611. [33] R. Pol´ak, M. Hochlaf, M. Levinas, G. Chambaud, P. Rosmus, Spectrochim. Acta A 55 (1999) 447. [34] M. Richard Viard, O. Dutuit, A. Amarkhodja, P.M. Guyon, in: F. Lahmani (Ed.), Photophysics and Photochemistry Above 6 eV, Elsevier, Amsterdam, 1985, p. 153. [35] K. Holdy, L.C. Klots, K.R. Wilson, J. Chem. Phys. 52 (1970) 4588. [36] F.E. Heidrich, K.R. Wilson, D. Rapp, J. Chem. Phys. 54 (1971) 3885.