- Email: [email protected]
graphite
Surface Science 414 (1998) 118–130
Core level spectroscopy study of N adsorbed on (2×2)K/graphite 2 C. Puglia a,*, P. Bennich a, J. Hasselstro¨m a, C. Ribbing a, P.A. Bru¨hwiler a, A. Nilsson a, Z.Y. Li b, N. Ma˚rtensson a a Department of Physics, Uppsala University, Box 530, 75121 Uppsala, Sweden b Cavendish Laboratory, University of Cambridge, Madingley Road, Cambridge, CB3 0HE, UK Received 19 February 1998; accepted for publication 22 June 1998
Abstract We report a study of N /(2×2)K/graphite at 25 K using X-ray photoemission, X-ray absorption ( XAS), ultraviolet photoemission, 2 autoionization and Auger spectroscopies. At this temperature we found that N physisorbs. Comparisons with the physisorbed 2 system, N /graphite, reveal the role played by the alkali. The molecular coverage corresponding to saturation of the first monolayer 2 is reduced drastically by the presence of potassium. Another dramatic effect induced by the potassium is the very high binding energy of the N1s level due to a potassium-induced change of the work function and to a less efficient image charge screening. On the other hand, XAS, autoionization and Auger data show that in the core excited state the molecule increases its interaction with the substrate to the extent that the adsorbed potassium facilitates charge transfer to neutral core-excited N . This is manifested by 2 the appearance of a new type of final state in the autoionization. © 1998 Elsevier Science B.V. All rights reserved. Keywords: Graphite; Nitrogen; Photoelectron spectroscopy; Photon absorption spectroscopy; Potassium; Synchrotron radiation photoelectron spectroscopy; X-ray absorption spectroscopy; X-ray photoelectron spectroscopy
1. Introduction Due to the importance of nitrogen in many catalytic reactions and in the formation of different compounds, the adsorption of N has been studied 2 on different metallic substrates. The technological importance of the catalytic synthesis of ammonia has been the basis for many studies dedicated to the adsorption of nitrogen on promoted metal surfaces [1,2]. Alkali atoms, pre-adsorbed on a metal, can act as promoters in the dissociation of molecular nitrogen [3,4]. According to these studies, the enhancement of the electron density of the substrate in the presence of an alkali over* Corresponding author. E-mail: [email protected]
layer induces a more pronounced back-bonding contribution (charge transfer) to the molecular antibonding 1p orbitals. The effect of this is an g increase in the adsorption energy for the N mole2 cules and a lowering of the activation barrier to dissociation [1,3,4]. Besides their technological importance, these types of coadsorption systems offer the possibility to study fundamental aspects of the adsorbate–adsorbate and the adsorbate– substrate interactions. In this paper we present a study of the adsorption of N on a monolayer (ML, the so-called 2 (2×2) phase [5,6 ]) of K/graphite. This system shows characteristics of both physisorption and chemisorption. We used X-ray photoelectron ( XPS), X-ray absorption ( XAS ), ultraviolet photoelectron ( UPS ), Auger and autoionization
0039-6028/98/$ – see front matter © 1998 Elsevier Science B.V. All rights reserved. PII: S0 0 39 - 6 0 28 ( 98 ) 0 05 0 2 -0
C. Puglia et al. / Surface Science 414 (1998) 118–130
spectroscopies to investigate the properties of the system. A comparison with the previously studied [7–10] adsorption of N on graphite makes it 2 possible to identify the effects of the alkali layer. Furthermore, by comparing the results for this system with those from O /(2×2)K/graphite [11] 2 it is possible to see how differences in the electronic structure of the molecule influence the interaction with the alkali/graphite substrate. Both oxygen and nitrogen physisorb on clean graphite [7–10,12]. In the presence of potassium, oxygen shows a variety of coverage-dependent phases. The open valence shell of oxygen provides means for a strong interaction between the electronegative molecules and the electropositive alkali atoms. This interaction produces different ionic phases, depending on the charge added to the oxygen valence 1p orbital. Superoxide (O− ), perg 2 oxide (O2− ) and dissociated oxygen have been 2 identified in different alkali systems [13–19]. The chemical interaction between the adsorbed molecules and the potassium strongly modifies the electronic structure of the (2×2) layer. In addition the charge which is initially donated from the potassium to the graphite is gradually withdrawn as the oxygen coverage is increased and it is donated to the oxygen instead. The valence band, the C1s and the K2p and 3p lines show large changes as a result of the oxygen–potassium interaction [11]. An electrostatic model has been introduced in order to discuss the charge redistribution from the alkali/graphite to the coadsorbed molecules [19] for the case of isolated K atoms in the dispersed phase. Depending on the electron affinity of the adsorbed molecules, it is then seen that the total energy of the system can be lowered by charge donation from the alkali towards the coadsorbate rather than to the graphite. This is the case, for example, when oxygen is adsorbed on a dispersed phase of potassium on graphite. In the case of CO molecules, on the other hand, the model predicts that the energy of the system cannot be lowered by charge redistribution from the substrate to the adsorbate due to the negative electron affinity of CO. This has also been verified experimentally with EELS ( Electron Energy Loss Spectroscopy) which shows that the K electronic charge is not
119
removed from the substrate upon adsorption of CO on a low-coverage (dispersed) phase of alkali on graphite [19]. Considering the fact that N is 2 isoelectronic with CO one might expect only a very weak interaction (or physisorption) of nitrogen on (2×2)K/graphite, due to the close outer shell of N . 2 We verify this expectation, finding that N on 2 (2×2)K/graphite retains many of its gas phase properties. However, we have detected a new type of screened final state in the 1s autoionization decay of the adsorbed molecule. This indicates that in the core-excited state there is a substantial interaction between the adsorbate and the substrate orbitals consistent with the increased electronegativity of the (Z+1) molecule, NO [20].
2. Experimental The sample, highly oriented pyrolytic graphite (HOPG), was mounted on a copper block in direct contact with a cold-finger of a cryostat. This allowed us to reach very low temperatures (25 K ) by a continuous flow of liquid helium during the preparation of the physisorbed phase. All experiments that we show here were performed at ca 25 K. The sample was cleaned by resistive heating, running a high current through the crystal perpendicular to the planes of the graphite (due to the higher resistance in this direction). The temperature was measured by a chromel–alumel thermocouple fixed in the front part of the sample holder. The alkali ( K ), from a SAES getter source, was evaporated onto the graphite at a temperature of about 90 K in order to get an ordered (2×2) layer [21]. After the evaporation the sample was quickly cooled down to 25 K in order to avoid any intercalation. The ML dose was calibrated using the line profile of the K3p and C1s line [6,21]. The ML dose of N was calibrated by dosing the gas at a 2 temperature of ca 25 K and monitoring with XPS the desorption of the multilayers while increasing the substrate temperature. The ML dose was also defined by observing the change in the line profile of the N1s line. In fact, due to different N1s XPS binding energies for the molecules adsorbed in the first or second layer the dose when addition broad-
120
C. Puglia et al. / Surface Science 414 (1998) 118–130
ening occurs can be related to the ML coverage [10]. The X-ray and ultraviolet ( UV ) photoemission experiments were carried out in Uppsala. The system consists of two ultra high vacuum ( UHV ) chambers, one for the preparation and basic characterization of the surfaces and one for the spectroscopic measurements. The photoelectron spectrometer consists of an Al Ka X-ray source with a water-cooled rotating anode and a monochromator on a Rowland circle arrangement. The photoelectrons were analysed in a hemispherical electrostatic analyser with a 36 cm mean radius and were detected by a multichannel detection system. The overall resolution in the present XPS experiments was 0.4 eV. The angle of the incoming light and outgoing electrons was ca 45° from the surface normal. The system contains a discharge lamp and a toroidal grating monochromator for HeI (21.2 eV ) and HeII (40.8 eV ) radiation. The overall resolution in our HeII-excited spectra was 0.4 eV. Since graphite is a semimetal with a very low density of states (DOS ) around the Fermi level (E ), the electron spectra were calibrated F relative to the Fermi edge of a Pt crystal mounted in electrical contact with the graphite. XAS, autoionization, Auger and further XPS experiments were performed at MAX-LAB, the Swedish synchrotron radiation facility in Lund. The beam line uses a plane grating monochromator of modified SX-700 type and a hemispherical electron energy analyser of modified Scienta type with a 20 cm mean radius [22]. The XP spectra for the C1s and K2p lines were taken at normal emission of the photoelectrons, with a photon energy of 350 eV and a resolution of ca 0.3 eV. The XA spectra were measured by recording the secondary electrons collected in the partial yield mode by a pulse counting channeltron detector. The photon energy resolution in the XAS measurements was ca 0.25 eV. The autoionization and Auger spectra were recorded with a photon energy resolution of 0.3 eV and an electron resolution of ca 0.6 eV. The autoionization spectra were recorded at the 1s-to-1p absorption resonance, g while a photon energy of 500 eV was used for the Auger measurements.
3. Results In Fig. 1 the N1s photoelectron line for 1 ML of N on a (2×2) layer of K on graphite is shown. 2 The binding energy is 406.8 eV and the full width at half maximum (fwhm) is ca 1.3 eV. The N1s line for 1 ML of N /graphite is also shown for 2 comparison. In that case the N1s line is centred at 403.9 eV and its fwhm is 0.83 eV [10]. The lower intensity of the N1s line for N / 2 (2×2)K/graphite indicates that the ML coverage is drastically reduced in presence of the alkali adlayer. We were not able to quantify the reduction due to normalization uncertainties. Fig. 2 shows that the adsorption of N does not 2 induce any significant changes in the K2p or C1s lines. UPS data excited by HeII radiation (40.8 eV ) are shown in Fig. 3. The peaks at ca 18.5 eV are due to the K3p levels. The intensity at E indicates F no change in metallic character upon N coadsorp2 tion. The broad structures at ca 21 and 8.3 eV are due to the valence electron states of graphite. For
Fig. 1. N1s XPS lines for 1 ML N (2×2)K/graphite and 1 ML 2 N /graphite (as in Ref. [10]). The intensity of the spectrum for 2 N /graphite has been divided by 8. The spectra are excited by 2 Al Ka radiation (1486.6 eV ).
C. Puglia et al. / Surface Science 414 (1998) 118–130
121
Fig. 2. C1s and K2p photoemission spectra lines taken at a photon energy of 350 eV for the indicated preparations.
Fig. 3. UPS specta excited by HeII radiation (40.8 eV ) for the indicated preparations. The constant intensity at the Fermi edge indicates that the metallic character of the alkali adlayer is preserved upon nitrogen adsorption.
the N overlayer we can clearly identify the nitro2 gen valence electron features at 14.9 eV (2s ), u 13.2 eV (1p ) and 12 eV (3s ). Fig. 4 shows a u g comparison between the valence photoemission spectra for N (2×2)K/graphite and N /graphite. 2 2 The peaks for N (2×2)K/graphite are rigidly 2 shifted by ca 2.4 eV to higher binding energy with respect to the corresponding peaks for N phy2 sisorbed on graphite. Fig. 5 presents the XAS data recorded in normal and grazing incidence of the exciting light. In this spectroscopy, a core electron is excited to an initially empty orbital. XAS is therefore a technique that gives information about the empty DOS of the excited molecule, i.e., in the presence of the core hole. The XA process is dominated by dipole transitions. Hence, using polarized synchrotron radiation, it is possible to separately map states with different symmetry [23]. In 1s absorption spectra for a linear molecule, s orbitals are reached
when the E-vector of the incident radiation is oriented along the molecular axis and p orbitals are reached when the E-vector is perpendicular to it. Furthermore, if the degeneracy of the p orbitals g is broken in the presence of a surface the two orthogonal p -orbitals can be selected by varying g the geometry of the incoming light, that is, the p g orbital parallel to the surface (p ) and the p orbital d g orthogonal to the surface (p ) can be separately ) mapped. If the symmetry properties of the XAS transitions have been identified one can then use angle-dependent XAS measurements to determine the orientation of the molecules on the surface. For N (2×2)K/graphite the p resonance, which 2 corresponds to a 1s–1p excitation, is centred at g 401 eV and has a fwhm of ca 0.7 eV both in grazing and normal incidence [7,24]. This indicates that there is no strong breaking of the degeneracy of the p and the p orbitals. In the region between d ) 405 and 410 eV we find excitations to the molecular
122
C. Puglia et al. / Surface Science 414 (1998) 118–130
Fig. 4. Detailed valence spectra for the indicated preparations.
Rydberg states while at higher energy we have contributions due to multielectron excitations. The s resonance, which corresponds to the excitation of a 1s electron to the 3s orbital, is centred at ca u 420 eV [7,24]. Comparing the spectra in normal and grazing incidence (with the E vector parallel and normal to the surface, respectively) we can see that we have somewhat more intensity in the s region in the normal incidence spectrum. This indicates that the molecules tend to orient themselves parallel to the surface. However, as it will be discussed later, by a comparison with the N /graphite data ( Figs. 6 and 7) the N molecules 2 2 on K covered surface may be more disorder or somewhat tilted. As for other physisorbed systems, the N1s XPS binding energy (marked in the figure by a bar) is higher than the energy of the p resonance. Since the core-hole in physisorbed systems is screened only by formation of an image potential in the substrate, the XPS final state is not fully screened and it does not correspond to the lowest core-hole state generally seen in chemisorbed systems [25]. Figs. 6 and 7 compare XA
Fig. 5. X-ray absorption spectra in normal incidence ( · · · ) and in grazing incidence (——) for N (2×2)K/graphite. The 2 bar marks the N1s binding energy.
data for N (2×2)K/graphite and N /graphite, in 2 2 normal and grazing incidence, respectively. The fine structure seen in Fig. 8 is due to different vibrational states excited in the XAS process. In order to observe such a vibrational progression the lifetime broadening of the excited state has to be of the same order of magnitude or smaller than the vibrational spacing. In cases of strong bonding (i.e. chemisorption) the vibrational fine-structure is also usually quenched. The fact that we observe the vibrational progression for N (2×2)K/graphite indicates weak bonding (phy2 sisorption). This behaviour is the same as observed for N /graphite [7]. However, the splitting and the 2 vibrational intensity distribution are somewhat different in the two systems. For N (2×2)K/ 2 graphite the vibrational slitting is ca 0.20 eV whereas for N /graphite it is of ca 0.23 eV. 2 Core-hole decay spectroscopy (i.e. autoionization and Auger spectroscopies) gives important
C. Puglia et al. / Surface Science 414 (1998) 118–130
Fig. 6. X-ray absorption spectra in normal incidence for the indicated samples. The bars mark the N1s XPS binding energies relative to E for the two systems. In the figure the estimated F ILs are also indicated.
information about the core hole decay dynamics and about the local electronic structure within the adsorbate–substrate system, that is, around the excited atom [26,27]. An autoionization spectrum corresponds to the decay of a neutral core-excited XAS final state. In the present case the autoionization spectra have been recorded at the 1p XAS g resonance. We usually distinguish between two types of autoionization processes. Within the single particle picture there can occur participator transitions in which the excited electron takes part in the core hole decay. The final state in this case is a one-hole final state as in valence photoemission spectroscopy (PES ). If the excited electron remains as a spectator during the decay, two-hole oneparticle final states will be observed. These final states correspond to shake-up satellites in PES. In Auger spectroscopy we consider instead the decay
123
Fig. 7. X-ray absorption spectra in grazing incidence for the indicated samples.
of the core-ionized state, that is, the XPS final state. This leads to two-hole final states. Fig. 9 shows the Auger spectrum and the autoionization spectra for N (2×2)K/graphite and 2 N /graphite. The structure centred at ca 387.4 eV 2 in the autoionization spectrum for N (2×2)K/ 2 graphite is due to participator decay to final states with holes in the 1p or 3s orbitals, as can be u g understood by a comparison with gas phase and electron valence photoemission data. At lower kinetic energies we find peaks which correspond to spectator transitions to final states of (2s 1p 3s )−21p type. All these transitions are u u g g also found in autoionization of N /graphite. The 2 peak centred at ca 392.2 eV is most interesting since it represents a final state which is lower in binding energy than the HOMO (highest occupied molecular orbital ). We attribute this peak to a screened valence hole state of (3s−1 )p or g g (1p−1 )1p type, confirmed by a comparison with u g X-ray emission spectroscopy ( XES ) data for gas
124
C. Puglia et al. / Surface Science 414 (1998) 118–130
Fig. 8. High resolution N1s-to-1p X-ray absorption spectra for g the indicated samples.
phase N , as it will be discussed in Section 4. This 2 represents a type of participator decay which, to our knowledge, has not been observed before. Several other possible origins of this feature have been considered. Measurements on different sample preparations, producing the same autoionization spectrum, lead us to rule out contamination. Furthermore, it can be excluded that it is due to a N1s XPS feature excited by second order radiation. Its intensity follows the intensity of the other autoionization features over the resonance and it does not show the photon energy dependence of a second order photoemission feature. We therefore conclude that it is an intrinsic part of the autoionization spectrum. The detailed assignment will be discussed next. The Auger spectrum is much broader and indicates the presence of spectral features which are typical for both autoionization and Auger decay. The structures at lower kinetic energies are due to regular Auger processes. The structures between 380 and 390 eV are due to one-hole final states which indicate that decay occurs also from neutral intermediate states. The same type of effect has
Fig. 9. Auger spectrum taken with a photon energy of 500 eV (top) and autoionization spectrum taken with a photon energy of 401 eV for N (2×2)K/graphite and for N /graphite 2 2 (———).
previously been seen for N /graphite [9]. The 2 implication of this will be discussed in Section 4.
4. Discussion 4.1. Physisorption of N on (2×2)K/graphite 2 The XPS data together with UPS and XAS data give us important and direct information about the adsorption of N on (2×2)K/graphite. The 2 identical appearance of the C1s and K2p lines with and without N ( Fig. 2) indicates that the 2 alkali–graphite system is minimally influenced by the presence of the N molecules. The lack of a 2 shift in the graphite and potassium valence band structures ( Fig. 3) also suggests that no significant charge redistribution between the substrate and adsorbate takes place. The lack of noticeable changes in the region of the Fermi level shows
C. Puglia et al. / Surface Science 414 (1998) 118–130
that the metallic character of the alkali layer is preserved upon N adsorption. 2 The energy separations of the valence orbitals (seen in Figs. 3 and 4) correspond to those for free N and for N /graphite ( Fig. 4). This suggests 2 2 that the electronic configuration of the molecule is hardly perturbed by the adsorption since, in the case of chemisorption, different orbitals may be affected differently. Furthermore, charge transfer to or from the molecule is likely to affect the molecular bond lengths and thereby the relative positions of the valence orbitals. Similar effects are expected in XAS. An increased charge in the initially empty molecular 1p orbital would cause g an increased bond length and hence a shift of the s resonance [22,23]. For N (2×2)K/graphite the 2 energy positions of the resonances in the XA spectrum do not change with respect to the free molecule or to N /graphite. This, together with 2 the UPS data, shows that the molecular bond length is not significantly altered. Furthermore, the fact that the XA p resonance occurs at a photon energy which is less than the 1s binding energy indicates that XPS does not produce the lowest core-hole state. This is characteristic of a weakly-interacting (physisorbed) system due to the lack of metallic screening of the XPS final state. A physisorbed molecule interacts only weakly (via van der Waals forces) with the substrate, preserving many of the characteristic properties of the free molecule. However, even in the case of physisorption there are important differences between the free and adsorbed molecules. The substrate has an essential role in orienting the molecules and it also provides a medium which can be polarized in order to screen excitation and de-excitation processes such as those discussed here. For free molecules the relaxation and screening of the core-hole final state occurs via a rearrangement of the molecular orbitals in a way that can be called ‘‘internal screening’’. For adsorbed molecules ‘‘external screening’’ (i.e. the creation of an image charge in the substrate and by the polarization of the surrounding molecules [28]) is also possible which will lower the core hole binding energy of a physisorbed species compared with that of the free molecules. Since for physisorbed
125
systems the overlap between the molecular and the substrate orbitals is very small there is no screening of the core hole involving charge transfer from the substrate to the adsorbate, as usually observed in chemisorption. Fig. 5 shows the XAS p resonance for N (2×2)K/graphite. The N1s core level bind2 ing energy is marked by a bar in the figure. The large energy separation between these two states immediately shows that there is no significant charge transfer screening in the present case. The shift on binding energy of the XPS line for physisorbed systems has been discussed in connection with many previous studies [10,21,28–33]. Two different types of model have been introduced in order to describe the binding energy shift: the work function model (DW) [31,33]; and the final state screening model (DE ) [29,32]. According to R the work function model, the binding energy is constant if referred to the vacuum level. The apparent shift is thus only due to the change of the work function upon adsorption [31,33]. In the final state screening model, on the other hand, the shift in the XPS lines is due to the lowering of the core ionization energies introduced by the polarization of the substrate (image charge) and of the surrounding molecules [28,29], in addition to DW that is, it takes other effects into account. Fig. 1 compares the N1s lines for nitrogen physisorbed on K/graphite (at 406.8 eV ) and on graphite (at 403.9 eV ) [10]. In addition to the large shift to higher binding energy (2.9 eV ) we also observe that the presence of the (2×2)K overlayer leads to a much lower intensity and a broadening of the N1s line. The binding energies in Fig. 1 are referenced to the Fermi level. Also when relating the N1s binding energies for N on 2 (2×2)K/graphite and for N /graphite to the 2 vacuum level, we obtain a shift although of smaller magnitude. Using the work function values (W) of 4.7 eV for clean graphite and 2.9 eV for (2×2)K/graphite (DW=1.8 eV [21]), the vacuumlevel referenced N1s lines are at 408.6 and 409.7 eV respectively, that is, the N1s level is shifted by 1.1 eV towards higher binding energies in the presence of the (2×2)K layer. It is, however, important to note that we have not taken into account any changes in the work function caused by the adsorption of N . On the other hand, we expect 2
126
C. Puglia et al. / Surface Science 414 (1998) 118–130
these changes for a weakly interacting system to be much less than 1.1 eV [2,10,28,29,32]. Comparing to the gas phase N1s binding energy of 409.9 eV we obtain a shift of 1.3 eV for N on 2 the clean graphite substrate and of only 0.2 eV for N on (2×2)K/graphite substrate. This latter 2 very low value indicates that there are competing contributions which partially cancel the image potential screening [2,34]. As a confirmation of this, we can calculate the image plane positions in the two cases assuming that the shifts relative to the gas phase are entirely due to image potential screening. We obtained for N /graphite that the 2 ˚ , whereas for image plane should be at ca 2.77 A N (2×2)K/graphite we find an unreasonable 2 ˚ . The change in the DOS of the value of 17 A substrate in the presence of the (2×2)K layer is furthermore expected to lead to an increased image charge screening due to the increased charge density in the substrate [6 ]. We can assume that the lowest screening value would be given by an image plane located precisely in the middle of the alkali overlayer, that is, neglecting the fact that a metallic K layer would place the image plane closer to the adsorbed N [35]. In this way we obtain an image 2 charge screening of ca 1.4 eV. Since polarization of the neighbouring N molecules within the 2 adsorbed layer can give a maximum contribution of ca 0.25 eV [28,36 ], we could expect a total polarization screening of ca 1.65 eV. Thus, other effects are responsible for the observed very small screening. Different measurements have been performed by varying the alkali coverages in order to try to understand the adsorption dynamics (or geometry) of the N ML. Increasing the K coverage from a 2 dispersed phase [6 ] to a complete (2×2) ML, the N1s line corresponding to 1 ML of N decreases 2 in intensity and progressively shifts towards higher binding energy. The lower intensity can then be considered as consistent with the N molecules 2 adsorbed in/over the hollow site formed by the (2×2)K layer. This means that the lower molecular coverage can be interpreted as the alkali atoms block the adsorption sites for N . An electrostatic 2 repulsion between a slightly polarized alkali overlayer and the adsorbed ionized molecules (as in the final state reached in a XPS process) can
increase the final state energy. The significant reduction of molecular N coverage can be related 2 to a repulsion of the molecule with alkali. Similar effects have been seen for N coadsorbed with 2 potassium on Rh(111) [2] and Ru(001) [34]. In these studies, the lower uptake of nitrogen was attributed to a repulsive interaction between the alkali and the nitrogen molecules, in part due to the low polarizability of N [2,34]. In our case, 2 however, we should take in mind that the (2×2)K phase is a metallic overlayer. The N valence orbitals also shift due to the 2 presence of potassium as seen from the UP spectra in Fig. 4. The relative energy separations, however, are unaffected. The shift with respect to N /graphite is 2.4 eV for the valence levels as 2 compared to 2.9 eV for the core. We suggest that this is due to the fact that the charge distribution of a valence hole is symmetrically distributed over the molecule whereas the core hole final state leads to a more asymmetric charge distribution. This leads to different screening effects [8,10]. The co-adsorption with potassium introduces a considerable broadening of the N1s line. In Fig. 1 we find a fwhm of ca 1.3 eV for 1 ML N (2×2)K/graphite compared to 0.83 eV for 2 1 ML N /graphite [10]. Possible mechanisms for 2 this have been discussed in connection with other systems [7,8,10]. The population of closely-spaced vibronic states in the final state can describe the broadening of the core lines for physisorbed systems with respect to the corresponding gas phase lines. Closely-packed states arise from low frequency rotational and translational modes that become frustrated upon adsorption. One should also consider the possibility that the broadening of the line depends on the adsorption geometry since the nitrogen atoms in the molecule can be inequivalent with respect to the interaction with the alkali adatoms. The XAS data provide more information about the physisorption state and about the geometrical arrangement of the molecules. In Fig. 5 we see that the intensity in the s resonance region (ca 420 eV ) is higher in normal incidence for N 2 adsorbed on (2×2)K/graphite. However, in Fig. 7 ( XAS grazing incidence) we can see that for N (2×2)K/graphite the s resonance region has 2
C. Puglia et al. / Surface Science 414 (1998) 118–130
slightly more intensity while the p resonance is less intense than for N /graphite. This implies that for 2 N on the (2×2) layer of K/graphite the molecular 2 axis is somewhat tilted from the orientation parallel to the surface. Thus, from the XPS, UPS and XAS data we find that N is physisorbed on (2×2)K/graphite. 2 The adsorption on this new surface is characterized by a lower uptake of nitrogen, a broadening and higher binding energy of all nitrogen photoemission lines with respect to N /graphite. 2 4.2. Bonding of core-excited N to 2 (2×2)K/graphite The XA spectra in Figs. 6–8 provide qualitative insight into the interaction of the core-excited state with the substrate. In Fig. 8 a comparison between the XA p resonances for N /graphite and 2 N (2×2)K/graphite reveals a difference between 2 the vibrational splitting for the two systems which can be attributed to different final states reached in the absorption process. The Morse potential curve of the corresponding ( Z+1) state, corrected by the different reduced mass, is used to describe the XAS final state. The vibrational splitting in the corrected NO ground state is 0.23 eV and for NO− 0.15 eV [37]. The decrease in vibration splitting for NO− is attributed to an extra electron in the antibonding 2p* orbital. For N (2×2)K/ 2 graphite the vibrational splitting lies between the values for NO and NO−. This indicates that we have a fractional extra population of the 2p* level through hybridization with the alkali/graphite substrate. Using a local picture we can describe this hybridization as a configuration interaction of NO and NO−. Hence, using a notation which is derived by the ground state of N , we propose that the 2 XA final state is due to a mixture of (1s−11p1 ) g and (1s−11p2 ) character. Thus we propose that g the attractive core-hole potential brings an empty adsorbate level closer to the Fermi level [25,38,39], which leads to this fractional population. We discuss this in more detail elsewhere [20]. The increased adsorbate–substrate interaction is also seen for the core-excited states (see Figs. 6 and 7). The XAS energy region between 405 and 410 eV shows transitions to what are sometimes
127
referred to as Rydberg-derived states. The region just above (up to 416 eV ) contains multielectron excitations involving 1p - and 3s -to-1p transu g g itions, accompanying the 1s-to-1p excitation. For g the free molecule, the Rydberg states are bound states of p or s symmetry. The ionization limit (IL), that is, the vacuum level (E ), for the phyv sisorbed case is indicated in Fig. 6. The fact that the structure at 406 eV persists after a 2 eV shift of the vacuum level (E ) suggests that this level is v much better described within a molecular orbital picture, rather than as part of a hydrogenic Rydberg series [12]. The lower intensity, broadening and shift in energy (with respect to the IL) of the features near E are a further sign that for v N (2×2)K/graphite we have more interaction 2 (hybridization) in the core hole state between the molecular and the substrate orbitals compared to the N /graphite system. However, the lack of any 2 energy shift and broadening of the p resonance with respect to the gas phase or N /graphite, 2 combined with the excellent correspondence between the main features in both cases, indicates that the hybridization is weak. We now examine the core-hole decay dynamics. Fig. 9 shows the N1s autoionization spectrum taken at a photon energy which corresponds to the 1s to 1p excitation together with the Auger g spectrum taken at a photon energy of 500 eV (i.e. well above the IL) corresponding to an ionic initial state. In the autoionization spectrum, the intense peak at ca 387.4 eV kinetic energy is due to participator decay with 1p−1 or 3s−1 final states. At u g lower kinetic energy we find features corresponding to spectator decay with final states of the type (2s 1p 3s )−21p . u u g g The results in Fig. 9 show that there are contributions from ionic and neutral initial states in the Auger spectrum as well, that is, it is possible to distinguish structures due to both autoionization (charge-transfer-screened ) and Auger (non-chargetransfer-screened ) processes, as already seen in other systems [7,10,40]. As already shown for N /graphite [7,9], integrating the relative inten2 sities of these neutral and ionic decay processes in the Auger spectrum, we can determine the characteristic time of the charge transfer screening. We obtain a time of ca 6×10−15 s for
128
C. Puglia et al. / Surface Science 414 (1998) 118–130
N (2×2)K/graphite, that is, a charge transfer 2 time slightly shorter than for N /graphite 2 (9×10−15 s). Though these estimates are determined with error bars of ±25% the relative change is consistent with the speculation from the XA data that there is an increased overlap in the molecular core hole state between the N orbitals 2 and the substrate (alkali layer) valence states. We can use the obtained charge transfer time (6×10−15 s) in order to estimate the intensity of a corresponding screened final state of (1s−11p1 ) g type in the XP spectrum. According to a model introduced by Gunnarsson and Scho¨nhammer [41], a charge transfer time of 6×10−15 s would correspond to an interaction width of 0.1 eV. The intensity ratio between the screened and unscreened states in the XPS can then be determined by this interaction width and the energy of the screening level in the ionic and neutral states. In the present case we obtained a screened fraction of ca 0.003, well below the experimental detection limit of 0.01. This is a further indication that the adsorbate–substrate hybridization is very weak. In the autoionization spectrum we see a broad and relatively weak feature at ca 392.2 eV, which has no correspondence in the spectrum for N /graphite (Fig. 9). Due to the high kinetic 2 energy of this feature we assign it to a decay to a charge-transfer-screened valence hole final state of (3s−1 1p ) and/or (1p−1 1p ) type. The energy of g g u g the transition from a (1s−11p ) initial state to a g (3s−1 1p ) final state has been measured by XES g g ( X-ray emission spectroscopy) for N in the gas 2 phase [42]. This transition energy is 392.2 eV, that is, identical to the E -referenced kinetic energy F which we observe in the autoionization spectrum. Other possible final states of the same type (not seen in XES due to the dipole selection rules relevant there) should have energies relatively close to this one. From this we can conclude that the proposed assignment is consistent with the observed energy. The absence of any structure in the autoionization spectrum for N /graphite ( Fig. 9) correspond2 ing to the low-binding energy peak observed for N (2×2)K/graphite indicates that it represents a 2 unique result related to the presence of the alkali layer. It can be viewed as a direct consequence of
the increased interaction already revealed by the XAS data. This new final state observed for N (2×2)K/ 2 graphite can, within the (Z+1) approximation, be related to the description of the NO on K/graphite adsorption system as we discuss elsewhere [20]. Replacing the excited NN* species with an NO molecule this new structure in the autoionization spectrum can be considered as a first spectroscopic evidence of an NO affinity-derived state. According to a model initially introduced by Nørskov et al. [43] and then reviewed by Kasemo et al. [44] the electron or photon emission occurring during the chemisorption process of NO molecules on alkali promoted surfaces, is due to the population of this state which is lowered in energy due to image charge forces and/or chemical effects as the molecule approaches the surface. The new chargetransfer-screened state seen in the autoionization spectrum can be viewed as a consequence of an important covalent component to the bonding with the surface. It would then correspond to the partial population of the affinity-derived state of ‘‘NO’’ which is lowered in energy to the vicinity of E upon adsorption [20]. F Looking at the electronic structure of the nitrogen molecule and regarding the new peak in the autoionization spectrum as due to final states of (3s−1 1p ) and/or (1p−1 1p ) type, we could expect g g u g a structure in the spectrum due to a final state corresponding to the ground state of the molecule, which is not observed. the energy position of this feature should be at ca 8 eV higher kinetic energy than the structure at 392.2 eV, as it can be estimated considering the 3s -to-1p excitation energy g g [45]. An estimation of the intensity of such a feature relative to the 392.2 eV structure can be obtained simply by considering the number of ways in which the two final states can be created. We then view both processes to be due to the (1s−11p2 ) admixture in the intermediate state. In g this way we find this ratio to be only a few per cent which explains why this feature would be difficult to resolve from the background noise. In order to compare the line shapes and intensities of the decay spectra for the two systems ( Fig. 9), it is important to consider the dynamics of the intermediate state and the nuclear motion
C. Puglia et al. / Surface Science 414 (1998) 118–130
that can occur on a timescale comparable with the core hole decay [39]. The higher degree of interaction that we can see for the core hole final state in N adsorbed on (2×2)K/graphite can produce 2 a significant vibrational broadening.
5. Conclusions We have studied the adsorption of N on 2 (2×2)K/graphite at a temperature of ca 25 K. By a comparison with N /graphite, we can distinguish 2 differences in the interaction with the substrate for the two cases within the physisorption framework. The presence of the potassium has very complex and important effects. First of all, we see a drastic decrease in the ML saturation coverage of the molecules. Photoionized states of N are less effec2 tively screened than the corresponding states for N /graphite due to a combination of different 2 effects including the change in work function. We attribute the broadening of the XP line to different adsorption sites and/or different vibrational states. The valence nitrogen lines are also broadened and shifted, but the relative energy separation is preserved. Thus, all the photoemission data indicate that the ground state of the molecules is very little perturbed upon adsorption, as expected for physisorption. The XAS, autoionization and Auger data point to interesting K-induced dynamics of the core hole state of the molecules. The broadening of the higher bound states in the XA spectrum together with the broadening of the features in the decay spectra (autoionization and Auger) indicate a stronger interaction in the core hole and twovalence-hole states between the adsorbate and the substrate. Among the new dynamics, we find a new type of charge-transfer-screened final states of (3s−1 1p ) or (1p−1 1p )-type in the 1s autoionig g u g zation decay.
Acknowledgements The assistance of the staff at MAX-LAB is gratefully acknowledged. We want to acknowledge O. Bjo¨rneholm and S. Svensson for valuable dis-
129
cussions. A.W. Moore, who supplied the graphite samples, is gratefully acknowledged. This research has been supported by the Human Capital and Mobility Programme of the European Community, HCM network Contract No. CHRX-CT94-0580, and by the Swedish Materials Research Consortium on Clusters and Ultrafine Particles, which is funded by the Swedish National Board for Industrial and Technical Development (NUTEK ) and the Swedish National Science Research Council (NFR). Z.Y. Li would also like to acknowledge supports from the UK Engineering and Physical Sciences Research Council.
References [1] G. Ertl, S.B. Lee, M. Weiss, Surf. Sci. 114 (1982) 527. [2] L. Bugyi, F. Solymosi, Surf. Sci. 258 (1991) 55. [3] J.K. Nørskov, S. Holloway, N.D. Lang, Surf. Sci. 137 (1984) 65. [4] N.D. Lang, S. Holloway, J.K. Nørskov, Surf. Sci. 150 (1985) 24. [5] Z.Y. Li, K.M. Hock, R.E. Palmer, Phys. Rev. Lett. 67 (1991) 1562. [6 ] P. Bennich, C. Puglia, P.A. Bru¨hwiler, A. Nilsson, A.J. Maxwell, A. Sandell, N. Ma˚rtensson, P. Rudolf, to be published. [7] A. Nilsson, O. Bjo¨rneholm, T. Tillborg, R.J. Guest, A. Sandell, R.E. Palmer, N. Ma˚rtensson, Surf. Sci. 287/288 (1992) 758. [8] A. Nilsson, R.E. Palmer, H. Tillborg, B. Hernna¨s, R.J. Guest, N. Ma˚rtensson, Phys. Rev. Lett. 68 (1992) 982. [9] O. Bjo¨rneholm, A. Nilsson, A. Sandell, B. Hernna¨s, N. Ma˚rtensson, Surf. Sci. 68 (1992) 1892. [10] H. Tillborg, A. Nilsson, B. Hernna¨s, N. Ma˚rtensson, Surf. Sci. 295 (1993) 1. [11] C. Puglia, P. Bennich, J. Hasselstro¨m, P.A. Bru¨hwiler, A. Nilsson, A.J. Maxwell, N. Ma˚rtensson, P. Rudolf, Surf. Sci. 383 (1997) 149. [12] O. Bjo¨rneholm, A. Nilsson, E.O.F. Zansky, A. Sandell, H. Tillborg, J.N. Andersen, N. Ma˚rtensson, Phys. Rev. B 47 (1993) 2308. [13] C.Y. Su, I. Lindau, P.W. Chye, S. Oh, W.E. Spicer, J. Electron Spectrosc. Relat. Phenom. 31 (1983) 221. [14] R.A. de Paola, F.M. Hoffman, D. Heskett, E.W. Plummer, J. Chem. Phys. 87 (1987) 1361. [15] L. Surnev, in: H.P. Bonzel ( Ed.), Physics and Chemstry of Alkali Metal Adsorption, vol. 173, Elsevier, New York, 1989. [16 ] H. Shi, K. Jacobi, Surf. Sci. 276 (1992) 12. [17] J. Jupille, P. Dolle, M. Besanc¸on, Surf. Sci. 260 (1992) 271. [18] P. Dolle, S. Drissi, M. Besanc¸on, J. Jupille, Surf. Sci. 269/ 270 (1992) 687.
130
C. Puglia et al. / Surface Science 414 (1998) 118–130
[19] K.M. Hock, J.C. Barnard, R.E. Palmer, H. Ishida, Phys. Rev. Lett. 71 (1993) 641. [20] C. Puglia, P.A. Bru¨hwiler, J. Hasselstro¨m, A. Nilsson, J. Ma˚rtensson, J. Chem. Phys. 109 (1998) 1209. [21] K.M. Hock, R.E. Palmer, Surf. Sci. 284 (1993) 349. [22] J.N. Andersen, O. Bjo¨rneholm, A. Sandell, R. Nyholm, J. Forsell, L. Tha˚nell, A. Nilsson, N. Ma˚rtensson, Synchr. Rad. News 4 (4) (1991) 14. [23] J. Sto¨hr, NEXAFS Spectroscopy, Springer Series in Surface Science vol. 25, Springer, Berlin, 1992. [24] C.T. Chen, Y. Ma, F. Sette, Phys. Rev. A 40 (1989) 6737. [25] A. Nilsson, O. Bjo¨rneholm, E.O.F. Zdansky, H. Tillborg, N. Ma˚rtensson, J.N. Andersen, R. Nyholm, Chem. Phys. Lett. 197 (1992) 12. [26 ] R. Manne, J. Chem. Phys. 52 (1970) 5733. ˚ gren, J. Nordgren, Theor. Chim. Acta (Berlin) 58 [27] H. A (1981) 111. [28] T.-C. Chiang, G. Kaindl, T. Mandel, Phys. Rev. B 33 (1986) 695. [29] G. Kaindl, T. Chiang, D.E. Eastman, F.J. Himpsel, Phys. Rev. Lett. 45 (1980) 1808. [30] W. Eberhardt, E.W. Plummer, Phys. Rev. Lett. 47 (1981) 1476. [31] K. Jacobi, H.H. Rotermund, Surf. Sci. 116 (1982) 435. [32] K. Jacobi, Phys. Rev. B 38 (1988) 6291.
[33] K. Wandelt, Springer Series Surf. Sci. 22 (1990) 289. [34] R.A. de Paola, F.M. Hoffmann, Chem. Phys. Lett. 128 (1986) 343. [35] H. Ishida, A. Liebsch, Phys. Rev. B 42 (1990) 5505. [36 ] A. Sandell, O. Hjortstam, A. Nilsson, P.A. Bru¨hwiler, O. Eriksson, P. Bennich, P. Rudolf, J.M. Wills, B. Johansson, N. Ma˚rtensson, Phys. Rev. Lett. 78 (1997) 4994. [37] G.J. Schulz, Rev. Mod. Phys. 45 (1973) 464. [38] M. Ohno, Phys. Rev. B 50 (1994) 2566. [39] W. Wurth, D. Menzel, in: W. Eberhardt (Ed.), Application of Synchrotron Radiation, Springer Series in Surface Science vol. 35, Springer, Berlin/Heidelberg, 1995. [40] C. Puglia, A. Nilsson, B. Hernna¨s, O. Karis, P. Bennich, N. Ma˚rtensson, Surf. Sci. 342 (1995) 119. [41] O. Gunnarsson, S. Scho¨nhammer, Phys. Rev. B 22 (1980) 3710. [42] P. Glans, P. Skytt, K. Gunnelin, J. Guo, J. Nordgren, J. Electron Spectrosc. Relat. Phenom. 82 (1996) 193. [43] J.K. Nørskov, D.M. Newns, B.I. Lundqvist, Surf. Sci. 80 (1979) 179. [44] B. Kasemo, E. To¨rnqvist, J.K. Nørskov, B.I. Lundqvist, Surf. Sci. 89 (1979) 554. [45] K.P. Huber, G. Herzberg, in: Molecular Spectra and Molecular Structure IV, Constants of Diatomic Molecules, van Nostrand Reinhold, New York, 1978.
Core level spectroscopy study of N adsorbed on (2×2)K/graphite 2 C. Puglia a,*, P. Bennich a, J. Hasselstro¨m a, C. Ribbing a, P.A. Bru¨hwiler a, A. Nilsson a, Z.Y. Li b, N. Ma˚rtensson a a Department of Physics, Uppsala University, Box 530, 75121 Uppsala, Sweden b Cavendish Laboratory, University of Cambridge, Madingley Road, Cambridge, CB3 0HE, UK Received 19 February 1998; accepted for publication 22 June 1998
Abstract We report a study of N /(2×2)K/graphite at 25 K using X-ray photoemission, X-ray absorption ( XAS), ultraviolet photoemission, 2 autoionization and Auger spectroscopies. At this temperature we found that N physisorbs. Comparisons with the physisorbed 2 system, N /graphite, reveal the role played by the alkali. The molecular coverage corresponding to saturation of the first monolayer 2 is reduced drastically by the presence of potassium. Another dramatic effect induced by the potassium is the very high binding energy of the N1s level due to a potassium-induced change of the work function and to a less efficient image charge screening. On the other hand, XAS, autoionization and Auger data show that in the core excited state the molecule increases its interaction with the substrate to the extent that the adsorbed potassium facilitates charge transfer to neutral core-excited N . This is manifested by 2 the appearance of a new type of final state in the autoionization. © 1998 Elsevier Science B.V. All rights reserved. Keywords: Graphite; Nitrogen; Photoelectron spectroscopy; Photon absorption spectroscopy; Potassium; Synchrotron radiation photoelectron spectroscopy; X-ray absorption spectroscopy; X-ray photoelectron spectroscopy
1. Introduction Due to the importance of nitrogen in many catalytic reactions and in the formation of different compounds, the adsorption of N has been studied 2 on different metallic substrates. The technological importance of the catalytic synthesis of ammonia has been the basis for many studies dedicated to the adsorption of nitrogen on promoted metal surfaces [1,2]. Alkali atoms, pre-adsorbed on a metal, can act as promoters in the dissociation of molecular nitrogen [3,4]. According to these studies, the enhancement of the electron density of the substrate in the presence of an alkali over* Corresponding author. E-mail: [email protected]
layer induces a more pronounced back-bonding contribution (charge transfer) to the molecular antibonding 1p orbitals. The effect of this is an g increase in the adsorption energy for the N mole2 cules and a lowering of the activation barrier to dissociation [1,3,4]. Besides their technological importance, these types of coadsorption systems offer the possibility to study fundamental aspects of the adsorbate–adsorbate and the adsorbate– substrate interactions. In this paper we present a study of the adsorption of N on a monolayer (ML, the so-called 2 (2×2) phase [5,6 ]) of K/graphite. This system shows characteristics of both physisorption and chemisorption. We used X-ray photoelectron ( XPS), X-ray absorption ( XAS ), ultraviolet photoelectron ( UPS ), Auger and autoionization
0039-6028/98/$ – see front matter © 1998 Elsevier Science B.V. All rights reserved. PII: S0 0 39 - 6 0 28 ( 98 ) 0 05 0 2 -0
C. Puglia et al. / Surface Science 414 (1998) 118–130
spectroscopies to investigate the properties of the system. A comparison with the previously studied [7–10] adsorption of N on graphite makes it 2 possible to identify the effects of the alkali layer. Furthermore, by comparing the results for this system with those from O /(2×2)K/graphite [11] 2 it is possible to see how differences in the electronic structure of the molecule influence the interaction with the alkali/graphite substrate. Both oxygen and nitrogen physisorb on clean graphite [7–10,12]. In the presence of potassium, oxygen shows a variety of coverage-dependent phases. The open valence shell of oxygen provides means for a strong interaction between the electronegative molecules and the electropositive alkali atoms. This interaction produces different ionic phases, depending on the charge added to the oxygen valence 1p orbital. Superoxide (O− ), perg 2 oxide (O2− ) and dissociated oxygen have been 2 identified in different alkali systems [13–19]. The chemical interaction between the adsorbed molecules and the potassium strongly modifies the electronic structure of the (2×2) layer. In addition the charge which is initially donated from the potassium to the graphite is gradually withdrawn as the oxygen coverage is increased and it is donated to the oxygen instead. The valence band, the C1s and the K2p and 3p lines show large changes as a result of the oxygen–potassium interaction [11]. An electrostatic model has been introduced in order to discuss the charge redistribution from the alkali/graphite to the coadsorbed molecules [19] for the case of isolated K atoms in the dispersed phase. Depending on the electron affinity of the adsorbed molecules, it is then seen that the total energy of the system can be lowered by charge donation from the alkali towards the coadsorbate rather than to the graphite. This is the case, for example, when oxygen is adsorbed on a dispersed phase of potassium on graphite. In the case of CO molecules, on the other hand, the model predicts that the energy of the system cannot be lowered by charge redistribution from the substrate to the adsorbate due to the negative electron affinity of CO. This has also been verified experimentally with EELS ( Electron Energy Loss Spectroscopy) which shows that the K electronic charge is not
119
removed from the substrate upon adsorption of CO on a low-coverage (dispersed) phase of alkali on graphite [19]. Considering the fact that N is 2 isoelectronic with CO one might expect only a very weak interaction (or physisorption) of nitrogen on (2×2)K/graphite, due to the close outer shell of N . 2 We verify this expectation, finding that N on 2 (2×2)K/graphite retains many of its gas phase properties. However, we have detected a new type of screened final state in the 1s autoionization decay of the adsorbed molecule. This indicates that in the core-excited state there is a substantial interaction between the adsorbate and the substrate orbitals consistent with the increased electronegativity of the (Z+1) molecule, NO [20].
2. Experimental The sample, highly oriented pyrolytic graphite (HOPG), was mounted on a copper block in direct contact with a cold-finger of a cryostat. This allowed us to reach very low temperatures (25 K ) by a continuous flow of liquid helium during the preparation of the physisorbed phase. All experiments that we show here were performed at ca 25 K. The sample was cleaned by resistive heating, running a high current through the crystal perpendicular to the planes of the graphite (due to the higher resistance in this direction). The temperature was measured by a chromel–alumel thermocouple fixed in the front part of the sample holder. The alkali ( K ), from a SAES getter source, was evaporated onto the graphite at a temperature of about 90 K in order to get an ordered (2×2) layer [21]. After the evaporation the sample was quickly cooled down to 25 K in order to avoid any intercalation. The ML dose was calibrated using the line profile of the K3p and C1s line [6,21]. The ML dose of N was calibrated by dosing the gas at a 2 temperature of ca 25 K and monitoring with XPS the desorption of the multilayers while increasing the substrate temperature. The ML dose was also defined by observing the change in the line profile of the N1s line. In fact, due to different N1s XPS binding energies for the molecules adsorbed in the first or second layer the dose when addition broad-
120
C. Puglia et al. / Surface Science 414 (1998) 118–130
ening occurs can be related to the ML coverage [10]. The X-ray and ultraviolet ( UV ) photoemission experiments were carried out in Uppsala. The system consists of two ultra high vacuum ( UHV ) chambers, one for the preparation and basic characterization of the surfaces and one for the spectroscopic measurements. The photoelectron spectrometer consists of an Al Ka X-ray source with a water-cooled rotating anode and a monochromator on a Rowland circle arrangement. The photoelectrons were analysed in a hemispherical electrostatic analyser with a 36 cm mean radius and were detected by a multichannel detection system. The overall resolution in the present XPS experiments was 0.4 eV. The angle of the incoming light and outgoing electrons was ca 45° from the surface normal. The system contains a discharge lamp and a toroidal grating monochromator for HeI (21.2 eV ) and HeII (40.8 eV ) radiation. The overall resolution in our HeII-excited spectra was 0.4 eV. Since graphite is a semimetal with a very low density of states (DOS ) around the Fermi level (E ), the electron spectra were calibrated F relative to the Fermi edge of a Pt crystal mounted in electrical contact with the graphite. XAS, autoionization, Auger and further XPS experiments were performed at MAX-LAB, the Swedish synchrotron radiation facility in Lund. The beam line uses a plane grating monochromator of modified SX-700 type and a hemispherical electron energy analyser of modified Scienta type with a 20 cm mean radius [22]. The XP spectra for the C1s and K2p lines were taken at normal emission of the photoelectrons, with a photon energy of 350 eV and a resolution of ca 0.3 eV. The XA spectra were measured by recording the secondary electrons collected in the partial yield mode by a pulse counting channeltron detector. The photon energy resolution in the XAS measurements was ca 0.25 eV. The autoionization and Auger spectra were recorded with a photon energy resolution of 0.3 eV and an electron resolution of ca 0.6 eV. The autoionization spectra were recorded at the 1s-to-1p absorption resonance, g while a photon energy of 500 eV was used for the Auger measurements.
3. Results In Fig. 1 the N1s photoelectron line for 1 ML of N on a (2×2) layer of K on graphite is shown. 2 The binding energy is 406.8 eV and the full width at half maximum (fwhm) is ca 1.3 eV. The N1s line for 1 ML of N /graphite is also shown for 2 comparison. In that case the N1s line is centred at 403.9 eV and its fwhm is 0.83 eV [10]. The lower intensity of the N1s line for N / 2 (2×2)K/graphite indicates that the ML coverage is drastically reduced in presence of the alkali adlayer. We were not able to quantify the reduction due to normalization uncertainties. Fig. 2 shows that the adsorption of N does not 2 induce any significant changes in the K2p or C1s lines. UPS data excited by HeII radiation (40.8 eV ) are shown in Fig. 3. The peaks at ca 18.5 eV are due to the K3p levels. The intensity at E indicates F no change in metallic character upon N coadsorp2 tion. The broad structures at ca 21 and 8.3 eV are due to the valence electron states of graphite. For
Fig. 1. N1s XPS lines for 1 ML N (2×2)K/graphite and 1 ML 2 N /graphite (as in Ref. [10]). The intensity of the spectrum for 2 N /graphite has been divided by 8. The spectra are excited by 2 Al Ka radiation (1486.6 eV ).
C. Puglia et al. / Surface Science 414 (1998) 118–130
121
Fig. 2. C1s and K2p photoemission spectra lines taken at a photon energy of 350 eV for the indicated preparations.
Fig. 3. UPS specta excited by HeII radiation (40.8 eV ) for the indicated preparations. The constant intensity at the Fermi edge indicates that the metallic character of the alkali adlayer is preserved upon nitrogen adsorption.
the N overlayer we can clearly identify the nitro2 gen valence electron features at 14.9 eV (2s ), u 13.2 eV (1p ) and 12 eV (3s ). Fig. 4 shows a u g comparison between the valence photoemission spectra for N (2×2)K/graphite and N /graphite. 2 2 The peaks for N (2×2)K/graphite are rigidly 2 shifted by ca 2.4 eV to higher binding energy with respect to the corresponding peaks for N phy2 sisorbed on graphite. Fig. 5 presents the XAS data recorded in normal and grazing incidence of the exciting light. In this spectroscopy, a core electron is excited to an initially empty orbital. XAS is therefore a technique that gives information about the empty DOS of the excited molecule, i.e., in the presence of the core hole. The XA process is dominated by dipole transitions. Hence, using polarized synchrotron radiation, it is possible to separately map states with different symmetry [23]. In 1s absorption spectra for a linear molecule, s orbitals are reached
when the E-vector of the incident radiation is oriented along the molecular axis and p orbitals are reached when the E-vector is perpendicular to it. Furthermore, if the degeneracy of the p orbitals g is broken in the presence of a surface the two orthogonal p -orbitals can be selected by varying g the geometry of the incoming light, that is, the p g orbital parallel to the surface (p ) and the p orbital d g orthogonal to the surface (p ) can be separately ) mapped. If the symmetry properties of the XAS transitions have been identified one can then use angle-dependent XAS measurements to determine the orientation of the molecules on the surface. For N (2×2)K/graphite the p resonance, which 2 corresponds to a 1s–1p excitation, is centred at g 401 eV and has a fwhm of ca 0.7 eV both in grazing and normal incidence [7,24]. This indicates that there is no strong breaking of the degeneracy of the p and the p orbitals. In the region between d ) 405 and 410 eV we find excitations to the molecular
122
C. Puglia et al. / Surface Science 414 (1998) 118–130
Fig. 4. Detailed valence spectra for the indicated preparations.
Rydberg states while at higher energy we have contributions due to multielectron excitations. The s resonance, which corresponds to the excitation of a 1s electron to the 3s orbital, is centred at ca u 420 eV [7,24]. Comparing the spectra in normal and grazing incidence (with the E vector parallel and normal to the surface, respectively) we can see that we have somewhat more intensity in the s region in the normal incidence spectrum. This indicates that the molecules tend to orient themselves parallel to the surface. However, as it will be discussed later, by a comparison with the N /graphite data ( Figs. 6 and 7) the N molecules 2 2 on K covered surface may be more disorder or somewhat tilted. As for other physisorbed systems, the N1s XPS binding energy (marked in the figure by a bar) is higher than the energy of the p resonance. Since the core-hole in physisorbed systems is screened only by formation of an image potential in the substrate, the XPS final state is not fully screened and it does not correspond to the lowest core-hole state generally seen in chemisorbed systems [25]. Figs. 6 and 7 compare XA
Fig. 5. X-ray absorption spectra in normal incidence ( · · · ) and in grazing incidence (——) for N (2×2)K/graphite. The 2 bar marks the N1s binding energy.
data for N (2×2)K/graphite and N /graphite, in 2 2 normal and grazing incidence, respectively. The fine structure seen in Fig. 8 is due to different vibrational states excited in the XAS process. In order to observe such a vibrational progression the lifetime broadening of the excited state has to be of the same order of magnitude or smaller than the vibrational spacing. In cases of strong bonding (i.e. chemisorption) the vibrational fine-structure is also usually quenched. The fact that we observe the vibrational progression for N (2×2)K/graphite indicates weak bonding (phy2 sisorption). This behaviour is the same as observed for N /graphite [7]. However, the splitting and the 2 vibrational intensity distribution are somewhat different in the two systems. For N (2×2)K/ 2 graphite the vibrational slitting is ca 0.20 eV whereas for N /graphite it is of ca 0.23 eV. 2 Core-hole decay spectroscopy (i.e. autoionization and Auger spectroscopies) gives important
C. Puglia et al. / Surface Science 414 (1998) 118–130
Fig. 6. X-ray absorption spectra in normal incidence for the indicated samples. The bars mark the N1s XPS binding energies relative to E for the two systems. In the figure the estimated F ILs are also indicated.
information about the core hole decay dynamics and about the local electronic structure within the adsorbate–substrate system, that is, around the excited atom [26,27]. An autoionization spectrum corresponds to the decay of a neutral core-excited XAS final state. In the present case the autoionization spectra have been recorded at the 1p XAS g resonance. We usually distinguish between two types of autoionization processes. Within the single particle picture there can occur participator transitions in which the excited electron takes part in the core hole decay. The final state in this case is a one-hole final state as in valence photoemission spectroscopy (PES ). If the excited electron remains as a spectator during the decay, two-hole oneparticle final states will be observed. These final states correspond to shake-up satellites in PES. In Auger spectroscopy we consider instead the decay
123
Fig. 7. X-ray absorption spectra in grazing incidence for the indicated samples.
of the core-ionized state, that is, the XPS final state. This leads to two-hole final states. Fig. 9 shows the Auger spectrum and the autoionization spectra for N (2×2)K/graphite and 2 N /graphite. The structure centred at ca 387.4 eV 2 in the autoionization spectrum for N (2×2)K/ 2 graphite is due to participator decay to final states with holes in the 1p or 3s orbitals, as can be u g understood by a comparison with gas phase and electron valence photoemission data. At lower kinetic energies we find peaks which correspond to spectator transitions to final states of (2s 1p 3s )−21p type. All these transitions are u u g g also found in autoionization of N /graphite. The 2 peak centred at ca 392.2 eV is most interesting since it represents a final state which is lower in binding energy than the HOMO (highest occupied molecular orbital ). We attribute this peak to a screened valence hole state of (3s−1 )p or g g (1p−1 )1p type, confirmed by a comparison with u g X-ray emission spectroscopy ( XES ) data for gas
124
C. Puglia et al. / Surface Science 414 (1998) 118–130
Fig. 8. High resolution N1s-to-1p X-ray absorption spectra for g the indicated samples.
phase N , as it will be discussed in Section 4. This 2 represents a type of participator decay which, to our knowledge, has not been observed before. Several other possible origins of this feature have been considered. Measurements on different sample preparations, producing the same autoionization spectrum, lead us to rule out contamination. Furthermore, it can be excluded that it is due to a N1s XPS feature excited by second order radiation. Its intensity follows the intensity of the other autoionization features over the resonance and it does not show the photon energy dependence of a second order photoemission feature. We therefore conclude that it is an intrinsic part of the autoionization spectrum. The detailed assignment will be discussed next. The Auger spectrum is much broader and indicates the presence of spectral features which are typical for both autoionization and Auger decay. The structures at lower kinetic energies are due to regular Auger processes. The structures between 380 and 390 eV are due to one-hole final states which indicate that decay occurs also from neutral intermediate states. The same type of effect has
Fig. 9. Auger spectrum taken with a photon energy of 500 eV (top) and autoionization spectrum taken with a photon energy of 401 eV for N (2×2)K/graphite and for N /graphite 2 2 (———).
previously been seen for N /graphite [9]. The 2 implication of this will be discussed in Section 4.
4. Discussion 4.1. Physisorption of N on (2×2)K/graphite 2 The XPS data together with UPS and XAS data give us important and direct information about the adsorption of N on (2×2)K/graphite. The 2 identical appearance of the C1s and K2p lines with and without N ( Fig. 2) indicates that the 2 alkali–graphite system is minimally influenced by the presence of the N molecules. The lack of a 2 shift in the graphite and potassium valence band structures ( Fig. 3) also suggests that no significant charge redistribution between the substrate and adsorbate takes place. The lack of noticeable changes in the region of the Fermi level shows
C. Puglia et al. / Surface Science 414 (1998) 118–130
that the metallic character of the alkali layer is preserved upon N adsorption. 2 The energy separations of the valence orbitals (seen in Figs. 3 and 4) correspond to those for free N and for N /graphite ( Fig. 4). This suggests 2 2 that the electronic configuration of the molecule is hardly perturbed by the adsorption since, in the case of chemisorption, different orbitals may be affected differently. Furthermore, charge transfer to or from the molecule is likely to affect the molecular bond lengths and thereby the relative positions of the valence orbitals. Similar effects are expected in XAS. An increased charge in the initially empty molecular 1p orbital would cause g an increased bond length and hence a shift of the s resonance [22,23]. For N (2×2)K/graphite the 2 energy positions of the resonances in the XA spectrum do not change with respect to the free molecule or to N /graphite. This, together with 2 the UPS data, shows that the molecular bond length is not significantly altered. Furthermore, the fact that the XA p resonance occurs at a photon energy which is less than the 1s binding energy indicates that XPS does not produce the lowest core-hole state. This is characteristic of a weakly-interacting (physisorbed) system due to the lack of metallic screening of the XPS final state. A physisorbed molecule interacts only weakly (via van der Waals forces) with the substrate, preserving many of the characteristic properties of the free molecule. However, even in the case of physisorption there are important differences between the free and adsorbed molecules. The substrate has an essential role in orienting the molecules and it also provides a medium which can be polarized in order to screen excitation and de-excitation processes such as those discussed here. For free molecules the relaxation and screening of the core-hole final state occurs via a rearrangement of the molecular orbitals in a way that can be called ‘‘internal screening’’. For adsorbed molecules ‘‘external screening’’ (i.e. the creation of an image charge in the substrate and by the polarization of the surrounding molecules [28]) is also possible which will lower the core hole binding energy of a physisorbed species compared with that of the free molecules. Since for physisorbed
125
systems the overlap between the molecular and the substrate orbitals is very small there is no screening of the core hole involving charge transfer from the substrate to the adsorbate, as usually observed in chemisorption. Fig. 5 shows the XAS p resonance for N (2×2)K/graphite. The N1s core level bind2 ing energy is marked by a bar in the figure. The large energy separation between these two states immediately shows that there is no significant charge transfer screening in the present case. The shift on binding energy of the XPS line for physisorbed systems has been discussed in connection with many previous studies [10,21,28–33]. Two different types of model have been introduced in order to describe the binding energy shift: the work function model (DW) [31,33]; and the final state screening model (DE ) [29,32]. According to R the work function model, the binding energy is constant if referred to the vacuum level. The apparent shift is thus only due to the change of the work function upon adsorption [31,33]. In the final state screening model, on the other hand, the shift in the XPS lines is due to the lowering of the core ionization energies introduced by the polarization of the substrate (image charge) and of the surrounding molecules [28,29], in addition to DW that is, it takes other effects into account. Fig. 1 compares the N1s lines for nitrogen physisorbed on K/graphite (at 406.8 eV ) and on graphite (at 403.9 eV ) [10]. In addition to the large shift to higher binding energy (2.9 eV ) we also observe that the presence of the (2×2)K overlayer leads to a much lower intensity and a broadening of the N1s line. The binding energies in Fig. 1 are referenced to the Fermi level. Also when relating the N1s binding energies for N on 2 (2×2)K/graphite and for N /graphite to the 2 vacuum level, we obtain a shift although of smaller magnitude. Using the work function values (W) of 4.7 eV for clean graphite and 2.9 eV for (2×2)K/graphite (DW=1.8 eV [21]), the vacuumlevel referenced N1s lines are at 408.6 and 409.7 eV respectively, that is, the N1s level is shifted by 1.1 eV towards higher binding energies in the presence of the (2×2)K layer. It is, however, important to note that we have not taken into account any changes in the work function caused by the adsorption of N . On the other hand, we expect 2
126
C. Puglia et al. / Surface Science 414 (1998) 118–130
these changes for a weakly interacting system to be much less than 1.1 eV [2,10,28,29,32]. Comparing to the gas phase N1s binding energy of 409.9 eV we obtain a shift of 1.3 eV for N on 2 the clean graphite substrate and of only 0.2 eV for N on (2×2)K/graphite substrate. This latter 2 very low value indicates that there are competing contributions which partially cancel the image potential screening [2,34]. As a confirmation of this, we can calculate the image plane positions in the two cases assuming that the shifts relative to the gas phase are entirely due to image potential screening. We obtained for N /graphite that the 2 ˚ , whereas for image plane should be at ca 2.77 A N (2×2)K/graphite we find an unreasonable 2 ˚ . The change in the DOS of the value of 17 A substrate in the presence of the (2×2)K layer is furthermore expected to lead to an increased image charge screening due to the increased charge density in the substrate [6 ]. We can assume that the lowest screening value would be given by an image plane located precisely in the middle of the alkali overlayer, that is, neglecting the fact that a metallic K layer would place the image plane closer to the adsorbed N [35]. In this way we obtain an image 2 charge screening of ca 1.4 eV. Since polarization of the neighbouring N molecules within the 2 adsorbed layer can give a maximum contribution of ca 0.25 eV [28,36 ], we could expect a total polarization screening of ca 1.65 eV. Thus, other effects are responsible for the observed very small screening. Different measurements have been performed by varying the alkali coverages in order to try to understand the adsorption dynamics (or geometry) of the N ML. Increasing the K coverage from a 2 dispersed phase [6 ] to a complete (2×2) ML, the N1s line corresponding to 1 ML of N decreases 2 in intensity and progressively shifts towards higher binding energy. The lower intensity can then be considered as consistent with the N molecules 2 adsorbed in/over the hollow site formed by the (2×2)K layer. This means that the lower molecular coverage can be interpreted as the alkali atoms block the adsorption sites for N . An electrostatic 2 repulsion between a slightly polarized alkali overlayer and the adsorbed ionized molecules (as in the final state reached in a XPS process) can
increase the final state energy. The significant reduction of molecular N coverage can be related 2 to a repulsion of the molecule with alkali. Similar effects have been seen for N coadsorbed with 2 potassium on Rh(111) [2] and Ru(001) [34]. In these studies, the lower uptake of nitrogen was attributed to a repulsive interaction between the alkali and the nitrogen molecules, in part due to the low polarizability of N [2,34]. In our case, 2 however, we should take in mind that the (2×2)K phase is a metallic overlayer. The N valence orbitals also shift due to the 2 presence of potassium as seen from the UP spectra in Fig. 4. The relative energy separations, however, are unaffected. The shift with respect to N /graphite is 2.4 eV for the valence levels as 2 compared to 2.9 eV for the core. We suggest that this is due to the fact that the charge distribution of a valence hole is symmetrically distributed over the molecule whereas the core hole final state leads to a more asymmetric charge distribution. This leads to different screening effects [8,10]. The co-adsorption with potassium introduces a considerable broadening of the N1s line. In Fig. 1 we find a fwhm of ca 1.3 eV for 1 ML N (2×2)K/graphite compared to 0.83 eV for 2 1 ML N /graphite [10]. Possible mechanisms for 2 this have been discussed in connection with other systems [7,8,10]. The population of closely-spaced vibronic states in the final state can describe the broadening of the core lines for physisorbed systems with respect to the corresponding gas phase lines. Closely-packed states arise from low frequency rotational and translational modes that become frustrated upon adsorption. One should also consider the possibility that the broadening of the line depends on the adsorption geometry since the nitrogen atoms in the molecule can be inequivalent with respect to the interaction with the alkali adatoms. The XAS data provide more information about the physisorption state and about the geometrical arrangement of the molecules. In Fig. 5 we see that the intensity in the s resonance region (ca 420 eV ) is higher in normal incidence for N 2 adsorbed on (2×2)K/graphite. However, in Fig. 7 ( XAS grazing incidence) we can see that for N (2×2)K/graphite the s resonance region has 2
C. Puglia et al. / Surface Science 414 (1998) 118–130
slightly more intensity while the p resonance is less intense than for N /graphite. This implies that for 2 N on the (2×2) layer of K/graphite the molecular 2 axis is somewhat tilted from the orientation parallel to the surface. Thus, from the XPS, UPS and XAS data we find that N is physisorbed on (2×2)K/graphite. 2 The adsorption on this new surface is characterized by a lower uptake of nitrogen, a broadening and higher binding energy of all nitrogen photoemission lines with respect to N /graphite. 2 4.2. Bonding of core-excited N to 2 (2×2)K/graphite The XA spectra in Figs. 6–8 provide qualitative insight into the interaction of the core-excited state with the substrate. In Fig. 8 a comparison between the XA p resonances for N /graphite and 2 N (2×2)K/graphite reveals a difference between 2 the vibrational splitting for the two systems which can be attributed to different final states reached in the absorption process. The Morse potential curve of the corresponding ( Z+1) state, corrected by the different reduced mass, is used to describe the XAS final state. The vibrational splitting in the corrected NO ground state is 0.23 eV and for NO− 0.15 eV [37]. The decrease in vibration splitting for NO− is attributed to an extra electron in the antibonding 2p* orbital. For N (2×2)K/ 2 graphite the vibrational splitting lies between the values for NO and NO−. This indicates that we have a fractional extra population of the 2p* level through hybridization with the alkali/graphite substrate. Using a local picture we can describe this hybridization as a configuration interaction of NO and NO−. Hence, using a notation which is derived by the ground state of N , we propose that the 2 XA final state is due to a mixture of (1s−11p1 ) g and (1s−11p2 ) character. Thus we propose that g the attractive core-hole potential brings an empty adsorbate level closer to the Fermi level [25,38,39], which leads to this fractional population. We discuss this in more detail elsewhere [20]. The increased adsorbate–substrate interaction is also seen for the core-excited states (see Figs. 6 and 7). The XAS energy region between 405 and 410 eV shows transitions to what are sometimes
127
referred to as Rydberg-derived states. The region just above (up to 416 eV ) contains multielectron excitations involving 1p - and 3s -to-1p transu g g itions, accompanying the 1s-to-1p excitation. For g the free molecule, the Rydberg states are bound states of p or s symmetry. The ionization limit (IL), that is, the vacuum level (E ), for the phyv sisorbed case is indicated in Fig. 6. The fact that the structure at 406 eV persists after a 2 eV shift of the vacuum level (E ) suggests that this level is v much better described within a molecular orbital picture, rather than as part of a hydrogenic Rydberg series [12]. The lower intensity, broadening and shift in energy (with respect to the IL) of the features near E are a further sign that for v N (2×2)K/graphite we have more interaction 2 (hybridization) in the core hole state between the molecular and the substrate orbitals compared to the N /graphite system. However, the lack of any 2 energy shift and broadening of the p resonance with respect to the gas phase or N /graphite, 2 combined with the excellent correspondence between the main features in both cases, indicates that the hybridization is weak. We now examine the core-hole decay dynamics. Fig. 9 shows the N1s autoionization spectrum taken at a photon energy which corresponds to the 1s to 1p excitation together with the Auger g spectrum taken at a photon energy of 500 eV (i.e. well above the IL) corresponding to an ionic initial state. In the autoionization spectrum, the intense peak at ca 387.4 eV kinetic energy is due to participator decay with 1p−1 or 3s−1 final states. At u g lower kinetic energy we find features corresponding to spectator decay with final states of the type (2s 1p 3s )−21p . u u g g The results in Fig. 9 show that there are contributions from ionic and neutral initial states in the Auger spectrum as well, that is, it is possible to distinguish structures due to both autoionization (charge-transfer-screened ) and Auger (non-chargetransfer-screened ) processes, as already seen in other systems [7,10,40]. As already shown for N /graphite [7,9], integrating the relative inten2 sities of these neutral and ionic decay processes in the Auger spectrum, we can determine the characteristic time of the charge transfer screening. We obtain a time of ca 6×10−15 s for
128
C. Puglia et al. / Surface Science 414 (1998) 118–130
N (2×2)K/graphite, that is, a charge transfer 2 time slightly shorter than for N /graphite 2 (9×10−15 s). Though these estimates are determined with error bars of ±25% the relative change is consistent with the speculation from the XA data that there is an increased overlap in the molecular core hole state between the N orbitals 2 and the substrate (alkali layer) valence states. We can use the obtained charge transfer time (6×10−15 s) in order to estimate the intensity of a corresponding screened final state of (1s−11p1 ) g type in the XP spectrum. According to a model introduced by Gunnarsson and Scho¨nhammer [41], a charge transfer time of 6×10−15 s would correspond to an interaction width of 0.1 eV. The intensity ratio between the screened and unscreened states in the XPS can then be determined by this interaction width and the energy of the screening level in the ionic and neutral states. In the present case we obtained a screened fraction of ca 0.003, well below the experimental detection limit of 0.01. This is a further indication that the adsorbate–substrate hybridization is very weak. In the autoionization spectrum we see a broad and relatively weak feature at ca 392.2 eV, which has no correspondence in the spectrum for N /graphite (Fig. 9). Due to the high kinetic 2 energy of this feature we assign it to a decay to a charge-transfer-screened valence hole final state of (3s−1 1p ) and/or (1p−1 1p ) type. The energy of g g u g the transition from a (1s−11p ) initial state to a g (3s−1 1p ) final state has been measured by XES g g ( X-ray emission spectroscopy) for N in the gas 2 phase [42]. This transition energy is 392.2 eV, that is, identical to the E -referenced kinetic energy F which we observe in the autoionization spectrum. Other possible final states of the same type (not seen in XES due to the dipole selection rules relevant there) should have energies relatively close to this one. From this we can conclude that the proposed assignment is consistent with the observed energy. The absence of any structure in the autoionization spectrum for N /graphite ( Fig. 9) correspond2 ing to the low-binding energy peak observed for N (2×2)K/graphite indicates that it represents a 2 unique result related to the presence of the alkali layer. It can be viewed as a direct consequence of
the increased interaction already revealed by the XAS data. This new final state observed for N (2×2)K/ 2 graphite can, within the (Z+1) approximation, be related to the description of the NO on K/graphite adsorption system as we discuss elsewhere [20]. Replacing the excited NN* species with an NO molecule this new structure in the autoionization spectrum can be considered as a first spectroscopic evidence of an NO affinity-derived state. According to a model initially introduced by Nørskov et al. [43] and then reviewed by Kasemo et al. [44] the electron or photon emission occurring during the chemisorption process of NO molecules on alkali promoted surfaces, is due to the population of this state which is lowered in energy due to image charge forces and/or chemical effects as the molecule approaches the surface. The new chargetransfer-screened state seen in the autoionization spectrum can be viewed as a consequence of an important covalent component to the bonding with the surface. It would then correspond to the partial population of the affinity-derived state of ‘‘NO’’ which is lowered in energy to the vicinity of E upon adsorption [20]. F Looking at the electronic structure of the nitrogen molecule and regarding the new peak in the autoionization spectrum as due to final states of (3s−1 1p ) and/or (1p−1 1p ) type, we could expect g g u g a structure in the spectrum due to a final state corresponding to the ground state of the molecule, which is not observed. the energy position of this feature should be at ca 8 eV higher kinetic energy than the structure at 392.2 eV, as it can be estimated considering the 3s -to-1p excitation energy g g [45]. An estimation of the intensity of such a feature relative to the 392.2 eV structure can be obtained simply by considering the number of ways in which the two final states can be created. We then view both processes to be due to the (1s−11p2 ) admixture in the intermediate state. In g this way we find this ratio to be only a few per cent which explains why this feature would be difficult to resolve from the background noise. In order to compare the line shapes and intensities of the decay spectra for the two systems ( Fig. 9), it is important to consider the dynamics of the intermediate state and the nuclear motion
C. Puglia et al. / Surface Science 414 (1998) 118–130
that can occur on a timescale comparable with the core hole decay [39]. The higher degree of interaction that we can see for the core hole final state in N adsorbed on (2×2)K/graphite can produce 2 a significant vibrational broadening.
5. Conclusions We have studied the adsorption of N on 2 (2×2)K/graphite at a temperature of ca 25 K. By a comparison with N /graphite, we can distinguish 2 differences in the interaction with the substrate for the two cases within the physisorption framework. The presence of the potassium has very complex and important effects. First of all, we see a drastic decrease in the ML saturation coverage of the molecules. Photoionized states of N are less effec2 tively screened than the corresponding states for N /graphite due to a combination of different 2 effects including the change in work function. We attribute the broadening of the XP line to different adsorption sites and/or different vibrational states. The valence nitrogen lines are also broadened and shifted, but the relative energy separation is preserved. Thus, all the photoemission data indicate that the ground state of the molecules is very little perturbed upon adsorption, as expected for physisorption. The XAS, autoionization and Auger data point to interesting K-induced dynamics of the core hole state of the molecules. The broadening of the higher bound states in the XA spectrum together with the broadening of the features in the decay spectra (autoionization and Auger) indicate a stronger interaction in the core hole and twovalence-hole states between the adsorbate and the substrate. Among the new dynamics, we find a new type of charge-transfer-screened final states of (3s−1 1p ) or (1p−1 1p )-type in the 1s autoionig g u g zation decay.
Acknowledgements The assistance of the staff at MAX-LAB is gratefully acknowledged. We want to acknowledge O. Bjo¨rneholm and S. Svensson for valuable dis-
129
cussions. A.W. Moore, who supplied the graphite samples, is gratefully acknowledged. This research has been supported by the Human Capital and Mobility Programme of the European Community, HCM network Contract No. CHRX-CT94-0580, and by the Swedish Materials Research Consortium on Clusters and Ultrafine Particles, which is funded by the Swedish National Board for Industrial and Technical Development (NUTEK ) and the Swedish National Science Research Council (NFR). Z.Y. Li would also like to acknowledge supports from the UK Engineering and Physical Sciences Research Council.
References [1] G. Ertl, S.B. Lee, M. Weiss, Surf. Sci. 114 (1982) 527. [2] L. Bugyi, F. Solymosi, Surf. Sci. 258 (1991) 55. [3] J.K. Nørskov, S. Holloway, N.D. Lang, Surf. Sci. 137 (1984) 65. [4] N.D. Lang, S. Holloway, J.K. Nørskov, Surf. Sci. 150 (1985) 24. [5] Z.Y. Li, K.M. Hock, R.E. Palmer, Phys. Rev. Lett. 67 (1991) 1562. [6 ] P. Bennich, C. Puglia, P.A. Bru¨hwiler, A. Nilsson, A.J. Maxwell, A. Sandell, N. Ma˚rtensson, P. Rudolf, to be published. [7] A. Nilsson, O. Bjo¨rneholm, T. Tillborg, R.J. Guest, A. Sandell, R.E. Palmer, N. Ma˚rtensson, Surf. Sci. 287/288 (1992) 758. [8] A. Nilsson, R.E. Palmer, H. Tillborg, B. Hernna¨s, R.J. Guest, N. Ma˚rtensson, Phys. Rev. Lett. 68 (1992) 982. [9] O. Bjo¨rneholm, A. Nilsson, A. Sandell, B. Hernna¨s, N. Ma˚rtensson, Surf. Sci. 68 (1992) 1892. [10] H. Tillborg, A. Nilsson, B. Hernna¨s, N. Ma˚rtensson, Surf. Sci. 295 (1993) 1. [11] C. Puglia, P. Bennich, J. Hasselstro¨m, P.A. Bru¨hwiler, A. Nilsson, A.J. Maxwell, N. Ma˚rtensson, P. Rudolf, Surf. Sci. 383 (1997) 149. [12] O. Bjo¨rneholm, A. Nilsson, E.O.F. Zansky, A. Sandell, H. Tillborg, J.N. Andersen, N. Ma˚rtensson, Phys. Rev. B 47 (1993) 2308. [13] C.Y. Su, I. Lindau, P.W. Chye, S. Oh, W.E. Spicer, J. Electron Spectrosc. Relat. Phenom. 31 (1983) 221. [14] R.A. de Paola, F.M. Hoffman, D. Heskett, E.W. Plummer, J. Chem. Phys. 87 (1987) 1361. [15] L. Surnev, in: H.P. Bonzel ( Ed.), Physics and Chemstry of Alkali Metal Adsorption, vol. 173, Elsevier, New York, 1989. [16 ] H. Shi, K. Jacobi, Surf. Sci. 276 (1992) 12. [17] J. Jupille, P. Dolle, M. Besanc¸on, Surf. Sci. 260 (1992) 271. [18] P. Dolle, S. Drissi, M. Besanc¸on, J. Jupille, Surf. Sci. 269/ 270 (1992) 687.
130
C. Puglia et al. / Surface Science 414 (1998) 118–130
[19] K.M. Hock, J.C. Barnard, R.E. Palmer, H. Ishida, Phys. Rev. Lett. 71 (1993) 641. [20] C. Puglia, P.A. Bru¨hwiler, J. Hasselstro¨m, A. Nilsson, J. Ma˚rtensson, J. Chem. Phys. 109 (1998) 1209. [21] K.M. Hock, R.E. Palmer, Surf. Sci. 284 (1993) 349. [22] J.N. Andersen, O. Bjo¨rneholm, A. Sandell, R. Nyholm, J. Forsell, L. Tha˚nell, A. Nilsson, N. Ma˚rtensson, Synchr. Rad. News 4 (4) (1991) 14. [23] J. Sto¨hr, NEXAFS Spectroscopy, Springer Series in Surface Science vol. 25, Springer, Berlin, 1992. [24] C.T. Chen, Y. Ma, F. Sette, Phys. Rev. A 40 (1989) 6737. [25] A. Nilsson, O. Bjo¨rneholm, E.O.F. Zdansky, H. Tillborg, N. Ma˚rtensson, J.N. Andersen, R. Nyholm, Chem. Phys. Lett. 197 (1992) 12. [26 ] R. Manne, J. Chem. Phys. 52 (1970) 5733. ˚ gren, J. Nordgren, Theor. Chim. Acta (Berlin) 58 [27] H. A (1981) 111. [28] T.-C. Chiang, G. Kaindl, T. Mandel, Phys. Rev. B 33 (1986) 695. [29] G. Kaindl, T. Chiang, D.E. Eastman, F.J. Himpsel, Phys. Rev. Lett. 45 (1980) 1808. [30] W. Eberhardt, E.W. Plummer, Phys. Rev. Lett. 47 (1981) 1476. [31] K. Jacobi, H.H. Rotermund, Surf. Sci. 116 (1982) 435. [32] K. Jacobi, Phys. Rev. B 38 (1988) 6291.
[33] K. Wandelt, Springer Series Surf. Sci. 22 (1990) 289. [34] R.A. de Paola, F.M. Hoffmann, Chem. Phys. Lett. 128 (1986) 343. [35] H. Ishida, A. Liebsch, Phys. Rev. B 42 (1990) 5505. [36 ] A. Sandell, O. Hjortstam, A. Nilsson, P.A. Bru¨hwiler, O. Eriksson, P. Bennich, P. Rudolf, J.M. Wills, B. Johansson, N. Ma˚rtensson, Phys. Rev. Lett. 78 (1997) 4994. [37] G.J. Schulz, Rev. Mod. Phys. 45 (1973) 464. [38] M. Ohno, Phys. Rev. B 50 (1994) 2566. [39] W. Wurth, D. Menzel, in: W. Eberhardt (Ed.), Application of Synchrotron Radiation, Springer Series in Surface Science vol. 35, Springer, Berlin/Heidelberg, 1995. [40] C. Puglia, A. Nilsson, B. Hernna¨s, O. Karis, P. Bennich, N. Ma˚rtensson, Surf. Sci. 342 (1995) 119. [41] O. Gunnarsson, S. Scho¨nhammer, Phys. Rev. B 22 (1980) 3710. [42] P. Glans, P. Skytt, K. Gunnelin, J. Guo, J. Nordgren, J. Electron Spectrosc. Relat. Phenom. 82 (1996) 193. [43] J.K. Nørskov, D.M. Newns, B.I. Lundqvist, Surf. Sci. 80 (1979) 179. [44] B. Kasemo, E. To¨rnqvist, J.K. Nørskov, B.I. Lundqvist, Surf. Sci. 89 (1979) 554. [45] K.P. Huber, G. Herzberg, in: Molecular Spectra and Molecular Structure IV, Constants of Diatomic Molecules, van Nostrand Reinhold, New York, 1978.