3.1. Adsorption Configurations
We start by describing the adsorption configurations of the two selected organic moieties on graphene: Pyridine (C
H
N) interacts mainly via weak van der Waals forces with graphene and the pyridine radical chemisorbs on graphene. The most stable configuration of the pyridine/graphene case occurs when the pyridine ring is parallel to the graphene surface and is located above a C atom of graphene (similar to an AB configuration of graphite). The N atom of the molecule is at the center of a graphene ring as shown in
Figure 1a. The three inequivalent carbon atoms of the molecule are labeled as C1, C2 and C3. We also choose a C atom of graphene close to the adsorption site (labeled C
) whose purpose will be discussed later. The pyridine radical (C
H
N) chemisorbs on graphene and is oriented perpendicular to the graphene surface as shown in
Figure 1b. The covalent adsorption of the molecule with the C atom of graphene (labeled A
) distorts the graphene lattice structure, raising the A
site atom along with its nearest neighbors (labeled B
and shown in magenta) with respect to the graphene plane, as shown in
Figure 1b. These are general results for covalently-bonded species, from the simple H atom [
67] to other chemisorbed organic molecules on graphene [
25,
49]. Details about adsorption energies, bond lengths and intermediate adsorption configurations have been discussed in our previous work [
26].
3.2. X-Ray Photoemission
X-ray photoemission experiments allow one to measure the Binding Energy (BE) of the core electrons, in our case most importantly, that of the C 1s. Values of BE are sensitive to the local bonding environment and therefore rich in chemical information. Theoretically, the BE is defined as the energy difference between the system in the ground state and the one with the core electron removed, i.e., in the presence of a core hole. Absolute BEs are difficult to address numerically, but the most important information is often contained in the relative BEs of atoms of the same species, that is the Core Level Shifts (CLS) with respect to a reference value. Their determination can proceed straightforwardly by electronic structure calculations based on all-electron models, but also on more widespread pseudopotential codes. Indeed, it is sufficient to evaluate the difference in energy of the system, when a modified pseudopotential including a 1s core hole (FCH) replaces the standard one at particular atomic sites [
41]. In this work, we take the average BE of the C atoms of graphene as a reference so that the CLS
of the
i-th C atom is obtained as:
where
is the total energy of the system with the core hole at the
i-th atom and
the average of such values across the atoms of graphene in the supercell (70 in our case). To better compare with the free molecule in the forthcoming discussion, we alternatively take as a reference the average across the C atoms of pyridine,
.
We can now discuss the effect of core exciting an electron from the C 1s edge of the system and understand the effect of molecular adsorption on the binding energies of the system in terms of the core level shifts.
Figure 2a shows as a color map the C 1s CLSs of the pyridine/graphene case, calculated with respect to the average C 1s energy of graphene atoms. Values for selected atoms are tabulated in
Table 1, whereas all results can be observed in the
Supplementary Materials. The N atom of the molecule is not considered here as the N 1s ionization potential falls in a different excitation energy range. By observing the CLS from
Figure 2a, we see that the binding energies of the molecule are larger than the substrate ones by ∼0.3 eV on an average. This observation is consistent with the more efficient screening occurring within graphene than at the molecule, where the removal of the core electron is more expensive. The effect of physisorption of pyridine on graphene reduces the binding energy of the C atoms of graphene close to the adsorption site as the molecular electrons also participate in the screening. For pyridine, this appears to be especially effective around the N atom, as can be observed by looking at the CLSs of the graphene atoms close to it. The smallest CLS indeed is found for the atom of graphene that we labeled
(shown in
Figure 1a), which amounts to
eV. As for the pyridine molecule, it is instructive to compare the CLSs with respect to the results of the free species. For such a comparison, we consider in both cases the CLSs taken with respect to the molecule (see the values in parentheses in
Table 1. Noticeably, these are practically equal to those of the free molecule, testifying that the influence of the screening by graphene substrate is equally effective for a core hole at any molecule C atom that is about the same distance from the substrate. We can therefore conclude that physisorption does not significantly alter the BE of the system and only influences the substrate atoms close to the adsorption site.
The CLSs for the chemisorbed case of pyridine radical/graphene are shown in
Figure 2b. We recall that when graphene participates in chemisorption by forming a covalent bond at the A
site, the corresponding
orbital is removed from the
system in forming bonding/antibonding orbitals with the adsorbate [
25]. Consequently, this site is the one that is expected to differ significantly in its electronic properties from unperturbed graphene. Indeed, the A
site C atom of graphene has a very large and positive CLS, with the C 1s binding energy larger by ∼1 eV than average, testifying a lower capability of the system to screen a perturbation at this site. Conversely, the CLS on the B
site C atoms (see
Figure 1b) close to the adsorption site is significantly negative, although lower in magnitude (∼−0.4 eV). As one moves away from the adsorption site, one sees that the sign of the CLS of atoms belonging to the B sublattice keeps the negative sign found for the closest ones (B
) with magnitude decreasing with distance, whereas the A sublattice atoms show non-negative CLS without following a clear trend. We remark that atoms that appear farthest from the adsorption sites are however affected by the presence of periodically-repeated units as of the supercell approach. We refer to the Supporting Materials for additional information regarding the CLS values.
For the pyridine radical, the CLSs are on average positive, indicating a larger BE than the graphene atoms similar to the case of the physisorbed molecule. However, the difference with the graphene average is now smaller, which implies the core hole at the molecule being more effectively screened by graphene. In this respect, it is helpful to compare the CLSs taken with respect to the molecular average, as we move from the gas phase pyridine molecule to the chemisorbed radical (see
Table 1). In particular, the difference in the CLS of C2 and C3 upon physisorption is practically unaffected (as discussed earlier), but as the molecule is chemisorbed, such a difference increases from
eV to
eV. This can be understood since the height over graphene is approximatively equal for C2 and C3 atoms in the physisorbed case, but it is larger by 1 Å for the C3 atom of the chemisorbed radical, resulting in a less efficient screening and a larger BE. Remarkably, the C1 atom that is closest to graphene also shows an increase of BE upon chemisorption, consistent with the low capabilities of the A
site to screen the core excitation (recall the large CLS of the A
atom).
3.3. X-Ray Absorption
We now discuss the Near Edge X-ray Absorption Fine Structure (NEXAFS) spectra calculated for gas phase pyridine and further inspect the NEXAFS spectra for the organic/graphene systems to understand the effect of adsorption on the core excited spectra. The N 1s NEXAFS spectra have been shown in the
Supplementary Materials, which do not differ significantly from the C 1s shown here and therefore are not discussed in this section.
Let us first discuss the C 1s NEXAFS of gas phase pyridine, which is shown in
Figure 3a,b. As for the analysis of the CLSs, the absolute energy scale is not accessible by our method [
35], so the origin of the energy axis is fixed arbitrarily in the figure to align to the molecule/graphene case discussed later.
Figure 3a shows the total, spherically-averaged C 1s NEXAFS of gas phase pyridine decomposed into the contributions from the inequivalent C atoms (C1, C2 and C3 as shown in the inset). A first feature consisting of a double peak split by
eV can be observed. This originates from the 1s → LUMO transitions that are at different energies following the different BE of the corresponding atoms that we reported in
Table 1: in particular, at high energy, one finds the contribution from the C3 atom, which displays the highest BE. Polarized NEXAFS spectra, as accessible at synchrotron radiation facilities, further provide information on the molecular orbital symmetry and molecular orientation. In this respect,
Figure 3b shows the same C 1s NEXAFS spectra of gas phase pyridine, now resolving the different polarizations of the incoming photon field. The molecular LUMO is a
state, and as such, the first feature only appears for the photon electric field perpendicular to the pyridine ring (
z axis). Resonances resulting from transitions to states with the
character lying in the continuum of excitations can be seen at ∼10 eV higher energy (
x and
y photon field directions). These observations are in excellent agreement with previous calculations and experiments [
68].
We now discuss the effect of physisorption of the pyridine/graphene system on the C 1s NEXAFS spectra shown in
Figure 3c,d, presenting the atom-wise contributions and polarized spectra, respectively. Here, the energy for a C 1s
transition,
being the Fermi energy of the system, can be taken as a reference. The nature of the contributions is very similar to the free molecule case, indicating that the weak van der Waals coupling of the molecule with graphene does not significantly affect the NEXAFS spectra. The same applies to
Figure 3d showing the C 1s NEXAFS spectra for different photon polarizabilities. Here, the in-plane molecular axis spans the surface
x and
y directions, so that excitation to the molecular LUMO occurs in an experiment with the photon field in the
z direction orthogonal to the surface (commonly referred to as
p-polarization [
69]).
The NEXAFS spectra for the covalently-adsorbed pyridine radical show instead interesting changes with respect to the free molecule, as we report in
Figure 3e,f. The increased splitting of the CLS of C2 and C3 atoms upon chemisorption, remarkably, does not translate to an enhanced splitting of the two peaks within the LUMO feature in the NEXAFS spectrum. Conversely, that merges into a single main peak whose transition dipole moment is perpendicular to the pyridine plane showing a peak on the low-energy side, mostly as a result of the increased CLS of the C1 atom (seen in
Table 1).
We now look at the NEXAFS spectra of graphene and the influence of molecular adsorption on the latter.
Figure 4a shows the NEXAFS spectra calculated for pristine graphene. Two main structures are observed, in good agreement with previously-reported studies [
32]. The first feature at lower energy corresponds to excitation to the
electron system close to the Fermi level that occurs with the photon electric field along the graphene normal
z (
p-polarization). The second one, at about
eV, is instead due to a transition to
states that can be probed by the photon electric field in the graphene plane (
s-polarization) that we compute as the average of the
x and
y surface azimuths. One can observe by comparing these results to those in
Figure 4b that physisorption of pyridine does not significantly alter the NEXAFS spectra of graphene. There, we have chosen to report the NEXAFS spectrum calculated for the C
atom of graphene since this is the one that experiences the largest CLS change as compared to the other graphene atoms (as shown in
Figure 2a).
The effect of chemisorption of pyridine radical presents a significant effect on the NEXAFS of graphene. The A
site C atom of graphene participates in the covalent bond with the pyridine radical by donating its
orbital and forming bonding states at energies at least 4 eV below
[
25]. The corresponding antibonding orbital can be found in the unoccupied states probed by NEXAFS at
eV, as can be seen by the peak for the
p-polarized spectrum in
Figure 4c. Conversely, the intensity close to the Fermi level completely vanishes due to the chemisorption. It is also interesting to notice that the spectra in
p- and
s-polarization are very similar, consistent with the sp
hybridization at this sites. An intermediate situation occurs for the NEXAFS spectra at the B
site C atoms: these retain, to some degree, the resemblance with free standing graphene, as seen in
Figure 4d, especially as the spectra in
s-polarization are compared. The empty states corresponding to the
orbitals (
p-polarization) are perturbed by the orbital missing at the A
site, but still contribute to the spectral intensity at low energies. Finally, the presence of some states close to the Fermi level (∼1 eV) also in the
s-polarized spectra is a result of the distortion of the graphene lattice and hybridization of
and
orbitals.
3.4. Core-Excited Electronic and Magnetic Properties
We now discuss how the core-level excitation influences the electronic and magnetic properties of the two systems. For the following qualitative analysis, we concentrate on one specific case out of the possible 1s excitation sites; given the large number of inequivalent C atoms that could be considered, we select here the N atom in the molecule. It is instructive to switch on the perturbation adiabatically by performing calculations including a N 1s hole of increasing magnitude
, also considering that
corresponds to the HCH method employed for NEXAFS simulations and
to the FCH one of use for XPS analysis. The corresponding electronic Densities Of States (DOS) are shown in
Figure 5. There, spin-majority and spin-minority components are plotted separately (the latter with an inverted sign), and the projected DOS on the adsorbate and on graphene are shown. We briefly recall first the properties of the systems in the ground state [
26]. The physisorbed system is non-magnetic due to a weak van der Waals interaction of the molecule with graphene, as shown in
Figure 5a; whereas chemisorption of a molecule distorts the lattice symmetry of the substrate and graphene donates one of its
electron to participate in the covalent bond [
25]. This unbalance between the graphene sublattices results in a magnetic solution with a 1
magnetic moment carried by mid-gap states of graphene that are visible in the DOS close to the Fermi level, which have been well studied in previous works [
25,
26,
49]. The molecular orbitals are quite far away from the Fermi level in both cases, with the LUMO appearing in the DOS projected on the molecule at an energy larger than
eV, as seen in
Figure 5b.
Figure 5c,e shows the DOS of pyridine/graphene for the half core hole and full core hole final states, respectively. As a response to the creation of the core hole, the molecular orbitals are shifted to lower energies due to Coulombic interaction to screen the positive core. In
Figure 5c, where a
core charge corresponds to the transition-potential of X-ray absorption simulations, we can see that the LUMO comes closer to
, but is still unoccupied (at energy ∼
eV). The distribution of electronic charge between the two subsystems, which we compute by Löwdin population analysis, is unchanged within
e. The system is still non-magnetic as it was in the ground state. However, when the full core hole is created and the core electron is photoemitted from the system, the molecular LUMO shifts close to the Fermi level. The spin-up LUMO shifts partially below the Fermi level, and the spin-down LUMO remains above, resulting in a spin-polarized solution. We can visualize the wave functions of the system in real space by calculating the energy-Integrated Local Densities Of States (ILDOS), where the integral is performed over the energy interval of interest. This has been done for the ranges highlighted in
Figure 5e by the arrows labeled (R
) and (R
) for the two spin components. The results are reported in
Figure 6a,b for the spin-majority and spin-minority channel, confirming these peaks are dominated by the LUMO with a minor contribution by graphene (slightly larger in the spin-majority case). Compared to the ground state, we find an excess of
e on the molecule, withdrawn from graphene, which enables the LUMO to participate in screening the perturbation. As a consequence, the Dirac cone shifts to higher energies as appreciated in
Figure 5e. We remark that such a shift is strongly affected by the finite density of excited molecules and would decrease by increasing the simulation cell. Overall, the system reaches a partially-magnetic solution with a magnetic moment of
on the molecule.
For a chemisorbed system, in the HCH case, as shown in
Figure 5d, the system is magnetic due to the presence of spin-dependent mid-gap states of graphene similarly to the ground state case. The molecular LUMO shifts to lower energy values (∼
eV) and is equal for the two spin populations, but does not take part in the screening process. Nevertheless, since the mid-gap state hybridizes with molecular ones, the molecular Löwdin charges increase by
e with respect to the ground state even if this does not correspond to a filling of the LUMO. Upon creation of a full core hole, as shown in
Figure 5f, the LUMO shifts to the Fermi level. Here, it hybridizes with the mid-gap states of graphene to form a mid-gap-LUMO hybrid state in the spin up channel. This consists of two states, of which we plot the ILDOS in
Figure 6c,e, whose energy ranges are labeled as (R
) and (R
) in
Figure 5f. The first one is the bonding combination between them, and is filled; the second one is the anti-bonding combination (see a nodal plane between the molecule and graphene) and is empty, so that the system still remains magnetic, but now, the spin polarization is both on graphene (∼60%) and on the molecule (∼40%). In the spin-down channel, the LUMO does not hybridize with the mid-gap state, as can be visualized by the ILDOS shown in
Figure 6e,f whose energy ranges are appropriately marked in
Figure 5f as (R
) and (R
). Similar to the physisorbed case, the transfer of electrons to the molecule in order to screen the perturbation amounts to
e, and results in a shift of the Dirac point of graphene towards the unoccupied states. The addition of a further electron to the valence shell, as in X-ray absorption, would instead preferentially fill the minority spin population, and at self-consistency, the electronic structure changes with the extra electron being shared by the mid-gap and the spin-down LUMO forming hybrid states analogously to the spin-majority one [
26].
Before concluding this section, we remark that the shift of the Dirac cone as induced by a localized perturbation (treated by the repeated supercell approach) is a signature of electrons being attracted towards the perturbation from the surrounding graphene, hence neutralizing the system. Electron transfer from graphene to core-excited molecules has indeed been observed for bipyridine adsorbed on graphene/Ni or Au substrates [
37] where it occurs on a femtosecond timescale. It is therefore also interesting to look at the formation of the core hole for a neutral system, which is detailed in the
Supplementary Materials.