This document is the unedited Author’s version of a Submitted Work that was subsequently accepted for publication in Journal of Physical Chemistry C, copyright © American Chemical Society after peer review. To access the final edited and published work see: https://dx.doi.org/10.1021/acs.jpcc.0c01988. J. Physical Chemistry C Tuning the Magnetic Moment of FePc/Ag(110) Monolayer Phases through Oxygen Dosing E. Bartolomé1, J. Bartolomé2, F. Sedona3, J. Lobo-Checa2, D. Forrer3, J. Herrero-Albillos4, M. Piantek5, J. Herrero-Martín6, D. Betto7, E. Velez-Fort7, L.M. García2, M. Panighel8, A. Mugarza9,10, M. Sambi3, F. Bartolomé2  1 Escola Universitària Salesiana de Sarrià (EUSS), Passeig Sant Joan Bosco 74, 08017 Barcelona, Spain 2 Instituto de Ciencia de Materiales de Aragón and Departamento de Física de la Materia Condensada, CSIC-Universidad de Zaragoza, Pedro Cerbuna 12, 50009 Zaragoza, Spain. 3 CNR-ICMATE, Department of Chemical Sciences, University of Padova, via Marzolo 1, 35131 Padova, Italy 4 Centro Universitario de la Defensa. Academia General Militar, Zaragoza, Spain 5 Instituto de Nanociencia de Aragón, Universidad de Zaragoza, Zaragoza, Spain 6 ALBA synchrotron, Bellaterra, 08193-Barcelona, Spain 7 ESRF-The European Synchrotron Radiation Facility, Grenoble, France 8 CNR-IOM Laboratorio TASC, S.S. 14 km 163, Basovizza, I-34149 Trieste, Italy 9 Catalan Institute of Nanoscience and Nanotechnology (ICN2), CSIC and Barcelona Institute of Science and Technology, Campus UAB, Bellaterra, 08193 Barcelona, Spain 10 ICREA - Institució Catalana de Recerca i Estudis Avançats, 08010 Barcelona, Spain Received: xx, 2020; Revised Manuscript Received: xx, 2020 Iron-phtalocyanines (FePc) adsorbed onto a Ag(110) substrate self-assemble into different monolayer phases going from rectangular to different oblique phases, with increasing molecular density. We have investigated the oxygen uptake capability of the different phases and their associated magneto-structural changes. Our study combines scanning tunneling microscopy and spectroscopy (STM/STS), X-ray magnetic circular dichroism (XMCD), and density functional theory (DFT) calculations. STM measurements reveal that the oxidation reaction of the FePc/Ag(110) generally involves a displacement and a rotation of the molecules, which affects the electronic state of the Fe centers. The oxygen intercalation between FePc and the substrate is greatly obstructed by the steric hindrance in the high-density phases, to the point that a fraction of oblique phase molecules cannot change their position after oxydizing. Depending on the oxidation state and adsoption geometry, the STS spectra evidence clear differences in the Fe local density of states, which are mirrored in the XAS and XMCD experiments. Particularly, XMCD spectra of the oxidized phases reflect the distribution of FePc species (non-oxygenated, oxygenated-rotated and oxygenated-unrotated) in the different cases. Sum rule analysis yields the effective spin (mseff) and orbital (mL) magnetic moments of Fe in the different FePc species. Upon oxidation, the magnetic moment of FePc molecules increases about an order of magnitude, reaching mTOT~2.2 B per Fe atom. 1. INTRODUCTION between 410 and 520 K. The R1 phase has a centered rectangular unit cell, as shown in Fig. 1. In this phase, Fe centers ocupy sites Molecular overlayers on crystalline substrates have a broad field of on-top of a silver atom (OT sites) and all the molecules are oriented application in catalysis, sensing, molecular electronics, light-to- alike, with the molecular axis bisecting the phenyl rings forming a energy conversion, etc. In particular, bio-inspired oxygen-binding 45o angle with the substrate [11̅0] direction (OT-45 positions, see metalated macrocycles, such as iron-phtalocyanines (FePc), are Fig. 2a). The R2 phase is formed by the ordered and alternated investigated as viable substitutes for precious metals in catalysis of sequence of singly oriented rows parallel to the [001] direction of the Oxygen Reduction Reaction in low-temperature fuel cells. the Ag substrate (see Fig. 1). In this phase, Fe centers occupy short- Recently, the spontaneous assembly and the electronic and bridge sites between two silver atoms (SB sites), and the FePc magnetic properties of FePc have been studied on a variety of phenyl rings are rotatet 30o (either clockwise or anticlockwise) with substrates: Au(111)1–3, Au(001)4, Au(100)5, TiO2(110)6, respect to the substrate [11̅0] direction (SB-30 positions, see Fig. InAs(100)7, InSb(100)7, Cu(110)8, NaCl/Cu(111)9, Ag(111)10,11, 2b). graphite12 and grahene/Cu(111), Ni(111)13. In particular, At higher molecular density (≈0.449 molecules per nm2) the FePc/Ag(110) has been intensively investigated to shed light into thermal treatment stabilizes an oblique phase (OB1), as reported by the self-assembly, interaction mechanisms, structural Sedona et al.16. The molecular adsorption site and possible local transformations upon thermal treatments, and its capacity to orientations in the OB1 phase with respect to the substrate are those accommodate small molecules like O2. described for the R2 phase: in the OB1 phase the Fe centers occupy The properties of FePc sublimated onto Ag(110) were first SB sites and all of them have the same orientation, either +30o or - investigated by Palmgren et al.14 by low energy electron diffraction 30o within the same ordered domain. Cai et al.17 did observe two (LEED). Casarin et al.15 reported coverage-dependent structural additional “oblique” arrangements of FePc on Ag(110), which we and self-assembly properties of FePc based on STM and DFT have not found reproducibly as stable phases. simulations. They found that in the monolayer regime, FePc Interestingly, low-density R1 and R2 phases have been shown molecules lie flat on the Ag(110) surface. At molecular densities to be catalytically active16. Upon oxidation of these two phases, below 0.449 molecules per nm2, a mixture of two structurally oxygen atoms intercalate between the FePc and the substrate, distinct, equally dense phases coexist, namely c(10x4) = R1, and forming the FePc-(2-O2)-Ag coordination complex. The oxygen p(10x4) = R2 phases. Both phases stabilize upon annealing in UHV uptake in R2 phase involves shifting and rotating the molecules 1 Rectangular1 R1 Rectangular2 R2 Oblique1 OB1 Rectangular3 R3 Oblique2 OB2 0.421 nm-2 0.421 nm-2 0.449 nm-2 0.469 nm-2 0.527 nm-2 density Figure 1. FePc/Ag(110) sub-ML phases with increasing molecular density (from left to right). Each column identifies a phase containing an STM image acquired at room temperature, the unit cell model, the matrix notation and the LEED pattern with the corresponding simulation of the phase. STM imaging parameters for the top (75x55 Å2) images: R1 (0.4 V, 0.3 nA), R2 (0.4 V, 0.7 nA), OB1 (3.2 V, 9.0 nA), R3 (-0.9 V, 0.8 nA), OB2 (0.6 V, 0.4 nA). LEED parameters: R1(28 eV), R2 (30 eV), OB1 (28 eV), R3 (35 eV), OB2 (30 eV). from its original SB-30 to an OX(OT-45) position, and adsorbing Notably, the molecular arrangement and the Fe magnetic moment two oxygen atoms at the neighboring SB sites, as shown in Fig. 2c is reversible upon oxidation-annealing cycles18. (in the pristine R1 phase the FePc molecule is already positioned on However, the oxidation of high-density FePc phases on OT-45). Remarkably, this process decouples the molecular Ag(110) is still unreported. It is known that in the related case of electronic states from the metal surface resulting in a concomitant high-density FePc on Ag(100)19, oxygen is adsorbed (as di- and increase of one order of magnitude of the Fe magnetic moment mono-oxygen) on-top of the Fe centers instead of being intercalated (from mT≈0.26 B to mTOT ≈2.1 B), as demonstrated by XMCD. between the molecular layer and the substrate20. In this case, the FePc reactivity to oxidation seems to be mediated by the presence of Ag adatoms between the molecules, therefore involving a On top 45° Short Bridge 30° completely different reaction mechanism with respect to Ag(110). A temptative explanation for this difference assumes, on one hand, that Ag(100) is atomically "flat" compared to the corrugated No Oxygen Ag(110), which hinders the dissociation and transport of oxygen underneath the FePc molecules, and on the other hand that on Ag(110) there is no evidence of the presence of Ag adatomns in- between the FePc molecules. In this paper we report the formation of new high-density (OT-45) (SB-30) FePc/Ag(110) phases in the sub-ML to ML regime, which, despite (a) (b) their steric limitations, still present catalytic activity, at variance with the saturated Fe/Ag(100) monolayer. The oxygen intercalation capability and associated magnetization changes of these high- density phases are reported by correlating high-resolution STM images, STS spectra, XMCD measurements and DFT calculations, Oxygen and systematically compared with the analogous results obtained on the low-density phases, thereby obtaining a more detailed description of this complex and intringuing system. OX(OT-45) OX(SB-30) (c) (d) 2. METHODS Figure 2. Different geometries of as deposited and oxidized Samples. FePc molecules were evaporated on a Ag(110) single molecules in FePc/Ag(110) phases: top and lateral views. crystal substrate in several ultrahigh vacuum (UHV) chambers (base pressures in the low 10-10 mbar range), containing equipment 2 for sample sputtering, thermal annealing, FePc deposition, and (BOREAS beamline26) and ESRF (ID-3227,28) synchrotron connected to the XAS & XMCD experimental chamber. The radiation facilities. The sample was oriented with the Ag[11̅0] Ag(110) single-crystal was cleaned by repeated cycles of 1keV Ar+ direction perpendicular to the beam plane of incidence. sputtering and annealing at 550oC (10 min). The surface quality and Experiments were performed with the X-ray beam forming an angle order was monitored by LEED. of incidence 0 <  < 70o with the normal to the substrate. The A few mg of FePc (Alfa Aesar GmbH, 95% purity) were loaded detection mode was total electron yield (TEY). into a pyrolytic boron nitride (PBN) crucible connected to the In X-ray linear polarized absorption (XLPA) measurements, preparation chamber. The FePc sample was carefully degassed the polarization of incident X-ray is denoted as horizontal (H) and while monitoring the UHV base pressure. The crucible temperature vertical (V) for parallel and perpendicular, respectively, to the was raised to 540 K in approximately 30 h, held at this temperature synchrotron ring plane. for 25 h and increased to 560-570 K for additional 6 h. Depositions X-ray magnetic circular dichroism (XMCD) spectra were were performed with the crucible temperature at 550 K. The Ag measured at ALBA (BOREAS, B=6 T, T=3.4 K) and ESRF (ID8, substrate was held at room temperature (RT) during molecular B=4 T, T=6.0 K) endstations, respectively, to guarantee magnetic deposition. Different sub-ML phases of FePc/Ag(110) were saturation. To rule out experimental artifacts and reduce drift prepared, and the structure controlled in situ by LEED. phenomena, XMCD was measured by either changing the light The oxidation of the FePc deposited on the Ag substrate was helicity or the field direction. To increase statistics, the XMCD achieved by exposing the sample surface to an oxygen pressure P, spectra were determined from up to 16 XAS spectra with right- and for a time t (P x t Langmuir) in the closed chamber at room left-handed circular polarizations. temperature. The exposures for the samples used in this paper are To determine the FePc signal, the background contribution collected in Table S1. from the Ag substrate has to be carefully subtracted. X-ray absorption spectroscopy (XAS) and XLPA experiments on pristine STM. The structural evolution of FePc phases up to the full ML Ag substrates were performed under the same experimental with the oxygen dosing was monitored by Scanning Tunnelling conditions used for FePc/Ag(110) samples. The background- Microscopy (STM). The experiments were performed using two subtracted XAS and XLPA spectra were subsequently normalized different instruments: a room-temperature setup (Omicron VT- to the atomic continuum signal well above the Fe L2,3 absorption STM) at the University of Padova, and a low temperature edges. equipment (Omicron LT-STM) operated at 4.6 K at the Laboratorio de Microscopías Avanzadas (LMA, Zaragoza). 3. RESULTS STS. Differential conductance (dI/dV) curves at the Fe centres were 3.1 Modification of FePc/Ag(110) phases upon oxidation. obtained using the Omicron LT-STM at 4.6 K in two different At room temperature, FePc on Ag(110) self-assembles up to the ML operating modes: with open feedback loop at constant tip height in several ordered phases, all of them characterised by flat-lying (CH) and in constant current (CC) mode. While the CH mode is molecular overlayers. We detect five FePc phases when increasing generally used to measure the local density of states (LDOS) around the surface density, with occupation densities ranging from 0.421 the Fermi level, we found more convenient to close the feedback- to 0.527 molecules per nm2. Fig. 1 summarizes these arrangements loop (CC mode) for extended voltage ranges in order to avoid by showing their STM topography, LEED pattern, structural model structural damage of tip and sample by excessive electric fields and matrix notation with respect to the Ag(110) substrate. and/or electron injection. Here we combined the two modes in order At low densities, the surface is characterised by a mixture of the to obtain the spectra over the entire range of interest. For better rectangular R1 and R2 phases, that show the same centred unit cell visualisation, the CC data is multiplied by a constant factor in order dimensions. By increasing the FePc density, the oblique OB1 phase to stich the experimental curves. It should be noted, that we refrain is stabilized with 0.449 molecule/nm2. At even higher densities, two from comparing different phases’ spectra measured in separate other phases appear (with a unit cell almost squared but slightly experimental runs since the tip apex termination was found to be distorted): R3 phase, and a new oblique phase OB2. Among the five altered. Given that the surface was saturated with molecules, we did phases reported in Fig. 1, only in R1 the Fe is found in OT-45 not have full control over the tip nanostructuring that ultimately configuration, whereas in all the other phases the Fe is in SB-30 influence the dI/dV spectral lineshape (the observed signal is a positions. The R3 and OB2 phases often coexist, likely because of convolution of both the sample and tip LDOS). their similar structure (see also Fig. S1). Due to this similarity, in the subsequent discussion regarding oxidation they are treated as DFT. DFT+U calculations were performed on FePc adsorbed on equivalent (R3OX/OB2OX). Ag(110) before and after the oxygen dosage, using the Quantum- Fig. 3 shows STM images exhibiting the effect of exposing to ESPRESSO suite of codes21. As we need to deal with two different oxygen the different phases. Before oxidation (Fig. 3, top panel), adsorption configurations, i.e. the SB-30 and rotated OT-45, the FePc molecules characteristically show a bright central spot simulations were run on a p(6x4) rectangular cell, which can corresponding to the Fe atom. Oxygen uptake has the effect of allocate both geometries. The PBE22 exchange-correlation “dimming” this central point (Fig. 3, mid panels). This phenomenon functional was adopted and wavefunctions were expanded on a was earlier reported for low-density phases R1/R2. The formation plane-wave basis set with a cut-off of 27 Ry, while the cut-off on of the FePc-(2-O2)-Ag complex is accompanied by a downward the density was 250 Ry and a 2x2 mesh of k-points was used. The pull of the Fe atom, which is well below the macrocycle plane (as interaction between valence electrons and ion cores was described visible in Fig. 2c, 2d side views). The oxidized Fe atoms present through ultrasoft pseudopotentials23. The Hubbard U was used on also a different electronic configuration, and appear as dark spots in the Fe site only and was set to 3.9 eV24. STM images were simulated the STM images. The high-resolution STM image Fig. 3g shows within the Tersoff-Hamann approach25, while STS curves were the R1/R2OX phase contains a distribution of non-oxidized compared to the density of states projected on the Fe 3d z2 atomic molecules and oxidized molecules, OX(SB-30), see green box. orbital. Atomic charges, magnetizations and holes were computed We have found that OB1, R3 and OB2 high-density phases are within a Löwdin approach as in our previous work18. also catalytically active and behave similarly to R1 and R2, as evidenced in the middle and bottom rows of Fig. 3. XLPA, XAS, XMCD. Soft X-ray absorption and magnetic dichroism Fig. 3i displays a high-resolution STM image of a highly experiments at the Fe L2,3 and N K edges were carried out at ALBA oxygen-dosed phase R3OX, where approximately 80% of the This journal is © The Royal Society of Chemistry 20xx J. Phys. Chem. C 2020, xx, xx-xx R1 /R2 OB1 R3/OB2 (a) (b) (c) 10 nm 10 nm 10 nm (d) ox ox ox (e) (f) 10 nm 10 nm 10 nm (g) (h) (i) 2 nm 2 nm 2 nm Figure 3. Catalytic activity of the different FePc/Ag(110) phases. Top panels: STM images of pristine phases: (a) R1/R2 (V = 0.5 V, I =0.1 nA), (b) OB1 (0.5 V, 0.1 nA), (c) R3/OB2 (-1 V, 2.5 nA); Middle panels: images of partially oxidized phases: (d) R1OX /R2OX (0.5 mV, 0.2 nA), (e) OB1OX (0.4 V, 0.1 nA), (f) R3 OX/OB2OX (-1 V, 5.7 nA). Non-oxidized molecules are recognizable by their bright centers, while oxidized molecules have dark centers; Bottom panels: high-resolution STM images of oxidized phases. (g) R1OX/R2OX (0.5 V, 0.15 nA), (h) OB1OX (0.5 V, 0.2 nA), (i) R3OX (0.5 V, 0.05 nA). The R1OX/R2OX and R3OX/OB2OX phases contain a distribution of non-oxidized molecules (SB-30) and oxidized, rotated molecules OX(OT-45). In contrast, the OB1OX phase contains a distribution of non-oxidized FePc at SB-30 and two species of oxidized FePc: OX(OT-45) (green box) and OX(OT-30) (blue box). molecules were oxidized. In comparison with previously reported oxidized species. It is not surprising that the OB order is preserved oxidized R1OX and R2OX, the high-density oxidized R3OX phase by the SB-30 species, while grain boundaries are generated around shows considerable disorder. Indeed, numerous vacancies appear FePc’s shifting into OT-45 sites upon oxidation. It is evident that in and occasionally phenyl rings of neighboring oxidized molecules the OB1OX phase the FePc molecules struggle to accommodate overlap. Fig. S2 (top) shows a zoom into a non-oxidized molecule oxygen underneath, as seen by the STM image of oxidized (SB-30) and an oxidized one, OX(OT-45). The oxidation molecules and their corresponding profiles along the pyrrole rings mechanism in the R3OX phase therefore resembles the one found in (Fig. S2, bottom). Even if all non-oxidized molecules for the high- R2OX phase, although the increased molecular density strongly density phases start from SB-30 configuration, the oxidizing distorts the initial arrangement by introducing disorder due to steric activity of this OB phase is significantly lower compared to the R3 hindrance. and OB2 phases, even when these two present a higher molecular In the case of the oblique phase after oxidation (OB1OX) shown density. Therefore, the local molecular environment might be the in Fig. 3h, the system largely maintains its initial arrangement. As key for a higher reaction efficiency, as it can probably control the extracted from this figure, we can distinguish mainly three FePc ultimate molecular degrees of freedom. species: i) a majority of non-oxidized molecules, which remain at their original SB-30 positions (bright centers); ii) oxidized, rotated 3.2 Spectroscopy of molecular species in selected phases. molecules in OX(OT-45) positions (green box in Fig. 3h), STS experiments and DFT calculations have been performed to resembling the oxidized molecules in R2OX and R3OX phases, and follow the changes upon oxidation in the local density of states of iii) oxidized, non-rotated molecules maintaining their original the Fe centers. The square planar symmetry around Fe in the non- positions, OX(SB-30), blue box in Fig. 3h. This last species had oxidized molecules is modified when oxidized, since the Fe is not been previously reported. Ιn Fig. 3i we find 70% of non- shifted downwards and has 4 N and 2 O atoms as nearest neighbors, oxidized molecules, 22% of OX(SB-30) and 8% of OX(OT-45) with local C2 symmetry29. One may expect therefore a strong 4 change in the electronic states. We assume that the Fe ions are oxidized molecules are unable to rotate. In the unoccupied region, practically unaffected by adjacent molecules, so only three both oxidized species experimentally show dominant peaks (e) and molecular species have to be considered: non-oxidized SB-30, (e’) close to +1.8 V, which find its match in the calculations even if oxidized OX(OT-45) and OX(SB-30). These molecular species can slight energy deviations are observed. be identified on partially oxidized R2OX and OB1OX samples. Their In essence, the STS main experimental features are well STS spectra are measured with the tip of the microscope on top of reproduced by our theoretical calculations. Importantly, the HOMO the Fe atoms, either oxidized and non-oxidized. The energy quenching (peak (b)) and concomitant appearance of the dominant positions of the spectral features are summarized in Table S2. unoccupied states ((e) and (e’)) evidences a strong reconfiguration Fig. 4a shows the STS spectra corresponding to the partially of the electronic density of metallic d- states upon oxidation that oxidized R2OX low-density phase. The non-oxidized SB-30 simply depend on the adsorption position (determined by the molecules are found at SB-30 positions either clockwise or counter- molecule rotation) of the individual oxidized molecules. To clockwise rotated. They are found to exhibit identical conductance understand the extent of the electron transfer of the central Fe-ion spectra. Their spectrum (red line) exhibits resonances at -1.15 V (a) upon oxygen adsorption we complement our findings by studying and -0.32 V (b) below the Fermi level. Weaker features these systems using surface averaging XMCD. corresponding to unoccupied states can be found above the Fermi level: peak (c) at +0.80 V and an incomplete peak (d) in the proximity of +2 V. In contrast, the spectrum at the Fe center of the oxidized molecule OX(OT-45) (green line) shows only one dominant feature (e) above the Fermi level at +1.52 V, and another below the Fermi level (a) at -1.14 V. For the oxidized high-density OB1OX phase (Fig. 4b), we acquired the STS of the three different species. In the case of the (a) R2OX SB-30 pristine (non-oxidized) FePc (red line), despite the different tip OX(OT-45) e dI/dV (arb.units) LDOS, we find a close resemblance to the R2-phase spectrum, a b which is expected since in both cases molecules have the Fe centers at SB-30 adsorption sites. In contrast, the oxidized species can a d correspond either to rotated molecules OX(OT-45), green line, or c to unrotated FePc species, OX(SB-30), blue line. The spectrum of Fe in OX(SB-30) molecules shows a new resonance at -0.80 V, peak (f), absent in the rotated molecule CC CH CC OX(OT-45), but no other peak is found in the occupied PDOS region (note we assign the small features between -0.5 V and the -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 Fermi energy to tip contributions). In the region above the Fermi V (V) level, the dominant peaks (e’) and (e) are found at +1.58 eV for the (b) OB1OX SB-30 OX(SB-30) and at +1.78 eV for OX(OT-45), respectively. OX(SB-30) Moreover, the spectrum of OX(OT-45), blue line, exhibits a broad OX(OT-45) e dI/dV (arb.units) resonance (c) close to +1.0 eV. Importantly, these energy e' differences allow us to spectroscopically distinguish the two d oxidized species. f b c These experimental results can be compared to the PDOS a predictions for the different d orbitals, calculated for the three c molecular species depicted in Fig. 2. The experimental STS appears CC CH to follow closely the dz2 orbital PDOS predictions, as displayed in CC Fig. 4c, and to be less sensitive to the other calculated orbitals that -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 V (V) are shown in Fig. S3. A reason for this orbital selection relates to (c) PDOS the instrumental response since the tip is prone to interact SB-30 b e predominantly with vertically protruding wavefunctions, which OX(SB-30) PDOS, dz2 (arb.units) OX(OT-45) fulfills with the dz2 orbitals geometry. e' a In the case of the pristine FePc in SB-30 (red line in Fig. 4c), the dominant peak (b) in the occupied region corresponds to the HOMO orbital of the molecule, which energetically agrees well with the experimental spectra observed for the R2 and OB1 phases (Table S2). The occupied peak (a), which is clearly observed in R2 f but barely visible in OB1 at a lower energy, is neatly present in the c d a calculation, in good agreement with the experimental result in R2. Above the Fermi level, peaks (c) and (d) can also be correlated with -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 the experiments, albeit these peaks are weak and show slight energy V (V) shifts. Noteworthy is that in the occupied region we find experimentally that the oxidized rotated species OX(OT-45) in the Figure 4. STS spectra at the center of the molecule in the oxidized R2OX exhibits the state (a) below the Fermi level, but lack the (a) R2OX phase and (b) OB1OX phase (dotted lines mark the regions HOMO peak (b). The absence of this peak in the OX(OT-45) species of both phases could be interpreted as a quench of the measured in either CH or CC modes); (c) DFT-simulated PDOS HOMO state when the molecules oxidise and rotate. Contrarily, the dz2 component for the SB-30, OX(SB-30) and OX(OT-45) existence of peak (f) in OB1OX (SB-30) in proximity to (b) suggests molecular species (see also Fig. S3). a gentle electronic reconfiguration of the HOMO state whenever the This journal is © The Royal Society of Chemistry 20xx J. Phys. Chem. C 2020, xx, xx-xx 3.3 Probing oxidation of a high-density phase using XLPA. Qualitatively, a similar behavior was observed upon oxidation XLPA measurements at the N K-edge were first performed to of the R2 phase18, although in that case the D peak was more reassess the planar arrangement of the FePc molecules on the pronounced, and the process was completely reversible after studied samples. In each of the samples, the vertical (V) and annealing. In the R2 phase, the FePc molecules have, in principle, horizontal (H) N K-edge spectra in grazing incidence (=70o) were space enough to change from SB-30 to OT-45 and back upon clearly different, exhibiting the characteristic peaks corresponding oxidation and annealing respectively without much steric hindrance to * and * resonances which allows us to confirm the planar (although collisions between molecules shifted in opposite adsorption of the FePc molecules at the Ag(110) surface29 once directions may occur). We associate the intensity transfer from peak placed in the XAS sample holder. B to D upon oxidation with a ~2 eV peak shift, compatible with an Fig. 5a displays the XLPA at the Fe L3 edge for the freshly increase in the hole density above the Fermi Energy, EF, which is deposited highest density phase OB2. At normal incidence (γ=0º), compatible with the LDOS changes observed in STS on the Fe sites V and H spectra are coincident (as expected), while in grazing (cf. Fig. 4). It is feasible that that the formation of the FePc-(2-O2)- incidence (γ=70º), the linear absorption is strongly dichroic. In Ag complex has associated a charge transfer from Fe towards particular, the intense peak B in the H configuration is practically oxygen, increasing the hole density. quenched in the V one. This behavior was previously observed in Fig. 5c shows the H polarization XLPA spectrum of the R1 and R2 phases, where the B peak was identified with the oxidized sample OB2OX2 (black thick line). The experimental excitation of the Fe 2p3/2 electrons to the empty antibonding dz2(up) spectrum can be qualitatively approximated by a weighted sum of state hybridized with the N-s and N-pxy states of the next neighbor the spectra obtained for the pristine sample, where non-oxidized nitrogen atoms18. Features A-E, which were also found in R1 and FePc molecules are on SB-30 positions (green line), and that of a R2, are less pronounced in the OB2 phase. fully oxidized R2OX sample made of 100% oxidized, rotated Fig. 5b shows the evolution of the H-V XLPA spectra obtained OX(OT-45) molecules18. By this procedure, we could estimate a in grazing incidence (=70o) for four different stages in an yield of 20% of oxidized molecules for OB2OX1 and 50% in oxidation-annealing cycle: OB2→OX1→OX2→ANN, where the OB2OX2. We conclude that the oxidation of the denser phase OB2 O2 dosing of oxidized samples are given in Table S.1. The evolution is less effective than for the low-density R2 (comparing these observed in the XLPA spectra reflects the evolution of the oxidation results with those reported in ref. 18 for similar O2 dosing level, see reaction, i.e. OB2OX1 sample was less oxidized than OB2OX2: as the Fig. 10a). We note that we could reach up to 80% of oxidized OB2 phase gets oxidized, peak B fades while an intense peak D species when the sample was exposed in a different chamber to a emerges. The reverse process occurs upon annealing, although the higher oxygen dosing, as shown by the STM image in Fig. 3i. initial state was not completely recovered. The strong structural disorder observed in OB2 upon oxidation by STM (see Fig. 3i) probably justifies the incomplete recover of the pristine sample after annealing. 6 (a) 6 (b) (c) A B C D E OB2 (H) A B C D E OB2 XLPA (Normalized) XLPA (Normalized) o =70 OB2OX1 H =70 o OB2OX2 4 4 EH  OB2ANN n V EV 2 2 0 6 o 0 0 =70 =0 o (V) OB2 5 (H) OB2OX2 XLPA (Normalized) XLPA (normalized) XLPA (Normalized) =70o OB2OX1 4 50% SB-30 4 4 OB2OX2 +50% OX(OT-45) OB2ANN 3 50% OX(OT-45) 50% SB-30 2 2 2 1 0 0 0 706 708 710 712 714 716 706 708 710 712 714 716 706 708 710 712 714 716 E (eV) E (eV) E (eV) Figure 5. Normalized XLPA of phase OB2 at the Fe L3 edge before and after oxidation. (a) Non-oxidized grazing incidence (=70o) and normal incidence (=0o) spectra of two light polarizations; (b) Normalized XLPA in grazing incidence (=70o), H and V polarizations, of as deposited phase OB2, oxidized phases OB2OX1 and OB2OX2 and annealed OB2ANN; (c) The normalized XLPA (H polarization) in grazing incidence (=70o) of the oxidized phase R3OX2 is well predicted by a linear combination of 50% the XLPA of the pristine phase OB2 (containing FePc on SB-30 positions) and 50% the spectrum of the fully oxidized R2ox phase, where all molecules are OX(OT-45). Top inset: schematic view of sample FePc deposited on a Ag substrate. The incoming beam at grazing incidence is shown with the E in the H and V polarization modes. 6 3.4 Magnetic properties of oxidized high-density phases. peak at E=710 eV is uniquely strong in comparison with all the The XAS and XMCD spectra were measured on OB2 and OB1 other samples, which is indicative of the presence of a fraction of pristine and oxidized samples at the Fe L2,3 edges, under different Fe species with a different electronic structure. This observation incident angles from normal to grazing incidence, including the so correlates with the spectral differences obtained in STS of the called “magic angle” to allow the separate determination of Fe spin different oxidized species previously discussed. magnetic moment. Fig. 7 (left) compares the XMCD spectra along the cycle Fig. 6 shows the XAS & XMCD spectra obtained for pristine OB2→OX1→OX2→ANN obtained at γ=0º and γ=70º. The and oxidized OB2 and OB1 phases (upper and lower panels XMCD of pristine FePc on Ag(110) has a L2,3 lineshape very respectively). First, it has to be noted that Fe in the oxidized FePc- similar to that of metallic Fe, suggesting the strong hybridization of (2-O2)-Ag(110) complex presents a much larger XMCD signal Fe in FePc with the metal substrate. Similar results were obtained than the pristine FePc-Ag(110), in agreement with the experimental for other investigated incidence angles (see Fig. S5). The analysis results in R1 and R2 phases18. For example, O2 dosing produces an of the spectra evidences an evolution from the as-deposited phase increase in the XMCD signal in OB1 phase of a factor ~7 at =55o. to the most oxidized configuration. The XMCD obtained at the Determination of the fraction of oxidized molecules on each sample OB2OX1 and OB2OX2 phases present a progressively larger intensity and sum rule analysis allows us to properly quantify the orbital and and a shape evolving towards that observed for non-metallic Fe(III) spin magnetic moment of each configuration. Second, it is in a crystal field. After the annealing process, the XMCD recovers interesting to remark that although pristine OB2 and OB1 XMCD the metallic aspect, although the line shape is not exactly coincident spectra are essentially equal, and very similar to the XMCD of the with that of the as-deposited sample. Qualitatively, the same R1 and R2 samples, the lineshape of XMCD spectra in the oxidized behavior was observed along the oxidation-reduction reaction of the OB1OX phase is clearly different to that observed in samples R1OX, R1 and R2 phases18. R2OX (Fig. S.7) and OB2OX (Figs. S5 and 6): the intensity of the 8 8 =0o (a) OB2 (b) OB2OX2 =0o XAS (Normalized) XAS (Normalized) 6 =55o o 6 =55 =70o o =70 4 4 2 2 0 0 XMCD (Normalized) 0.0 0.0 XMCD (Normalized) -0.5 -0.5 -1.0 -1.0 -1.5 -2.0 -1.5 -2.5 700 710 720 730 700 710 720 730 E (eV) E (eV) 6 (c) OB1 6 (d) OB1OX XAS (Normalized) XAS (Normalized) 5 5 4 4 3 3 2 2 1 1 0 0 0.0 0 XMCD (Normalized) XMCD (Normalized) -0.2 -1 -0.4 -2 -3 -0.6 =0o =0 o =55 o -4 =55 o -0.8 o o =70 -5 =70 -1.0 700 710 720 730 700 710 720 730 E (eV) E (eV) Figure 6. Normalized XAS & XMCD spectra at the Fe L2,3 edge measured at different beam incident angles  for the (a) pristine R3 and (b) oxygenated OB2OX2 samples, at B=6 T, T=5.0 K, and (c) pristine OB1 and (d) oxygenated OB1OX samples, at B=4 T, T=2.8 K.  This journal is © The Royal Society of Chemistry 20xx J. Phys. Chem. C 2020, xx, xx-xx 0.4 OB2 (c) R2ox XMCD (normalised) 0 OB2OX1 0.3 mseff/nh OB2OX2 -1 0.2 OB2OX2 OB2OX1 OB2ANN -2 OB2ANN 0.1 OB2 R2 (a) =0o -3 0.0 0 0.02 (d) OB2 XMCD (normalised) -1 OB2OX1 mL/nh -2 20% OX(OT-45) + 80% SB-30 OB2OX2 -3 0.01 50% OX(OT-45) + 50% SB-30 -4 OB2ANN -5 (b) =70o 0.00 700 710 720 730 740 0 30 60 90 E (eV) Incidence angle,  (o) Figure 7. Left: Normalized XMCD spectra of the OB2, OB2OX1, OB2OX2, OB2ANN samples at the Fe L2,3 edge, measured at B=6 T, T=5.0 K at an incident angle (a) =0o, (b) =70o. The thin lines in (b) correspond to the XMCD calculated as the linear combination with (%OX- ROT) percentage of the fully-oxidized LDOX and (%NON-ROT) of the as-evaporated phase; Right: Effective spin moment (c) and orbital moments (per hole) (d) as determined with the sum rules for OB2, OB2OX1, OB2OX2 and OB2ANN. The lines in (c) and (d) show the fits using Eqs. [1] and [2], as well as curves earlier reported for low-density phases R2 and R2OX, ref. 16. The sum rules allow us to determine the orbital and effective At the magic angle (*=54.7o), the dipolar term cancels, and so spin magnetic moment per hole in the 3d band as a function of the mseff provides directly m s 30. incident angle. The right side of Fig. 7 yields the angular dependence of the orbital and effective spin magnetic moments per The sum rules were applied to obtain the angular dependencies eff 3d hole, m / n and mL / nh ( ) , determined for the OB2, m ( ) / nh and mL ( ) / nh (see Fig. 7c and Fig. S6) and the fitting eff s s h OB2OX1, OB2 and OB2ANN phases. For the sake of comparison, OX2 of the experimental data using Eqs. [1] and [2] allowed us to z xy z the previously determined dependencies for pristine and oxidized determine ms / nh , mL / nh , mL / nh and mT / nh for Fe in pristine, low-density R2 phases are also shown. oxidized and annealed OB2 and OB1 samples, which are At a given angle of incidence, the orbital moment per hole can summarized in Table 1. be expressed as a function of its components in the direction All phases display planar anisotropy, in agreement with the z xy perpendicular ( m L ) and parallel ( mL ) to the FePc molecular plane previous findings for FePc in diverse environments. However, the as: highly unquenched orbital magnetic moment of FePc when the molecule is not in contact with a metallic surface (mL/mS ≈ 1)29,31 is fully lost in FePc/Ag(110) phases, both in the pristine (or annealed) mL ( ) / nh  mLxy / nh  sin 2   mLz / nh  cos 2  . [1] phases as in the oxidized ones. On the other hand, the effective spin moment is mLz / nh mLxy / nh ms / nh mTz / nh Ref. mseff ( ) / nh  ms ( ) / nh  7 mT ( ) / nh , where ms is the isotropic spin moment and mT the angle-dependent intra-atomic spin dipole R2 0 6.5x10-3 7.1x10-2 4.3x10-3 18 moment, which is non-negligible in this case. mT expresses the R2DOX 1.2x10-2 2.2x10-2 3.7x10-1 2.9x10-2 18 inhomogeneous spatial distribution of the spin density over the R2ANN 4.0x10-3 1.5x10-2 7.1x10-2 5.9x10-3 18 atomic unit cell, due to the anisotropic charge distribution arising OB2 1.4x10-3 8.1x10-3 7.2x10-2 1.8x10-3 * from strongly directional bonds or crystal field. In the C4 symmetry OB2OX1 8.3x10-3 1.7x10-2 1.3x10-1 4.5x10-3 * of the FePc molecule, the in-plane ( mT ) and out-of-plane ( mT ) xy z OB2OX2 1.1x10-2 2.1x10-2 1.8x10-1 7.5x10-3 * OB2ANN 8.8x10-3 1.8x10-2 1.2x10-1 2.8x10-3 * components of the dipolar-moment are related by mTz  2mTxy  0 ; OB1 1.5x10-3 1.2x10-2 5.2x10-2 5.6x10-3 * thus, the effective moment angular dependence ms ( ) / nh can be eff OB1OX 2.0x10-3 4.2x10-2 4.9x10-1 6.2x10-2 * written as: Table 1. Magnetic moment parameters (in units of B/hole) for the different FePc/Ag(110) samples. * This work. mseff ( ) / nh  ms / nh  7 mTz / nh (cos 2   sin 2  / 2). [2] 8 3.5 Discussion Fig. 8 shows a comparison of the XMCD at =55o of the two oxidized high density phases OB2OX2 and OB1OX. Oxygen 0.0 XMCD (normalized) exposure produces in both cases an increase in the dichroic signal OB1ox compared to the pristine cases; however, differences in the spectra 10% SB-30 41% OX(OT-45) are observed, which should be correlated with the different 49% OX(SB-30) -0.2 molecular species detected by STM. The XMCD of the oxidized OB2OX phases can be well fitted using a linear combination of the experimental signals obtained from the pristine OB2 sample and the fully oxidized sample, i.e. a -0.4 Fe3+ linear combination of non-oxidized FePc on SB-30 and oxidized Fe2+ OX(OT-45) molecules, as shown in Fig. 8a. For the OB2OX1 and LC OB2OX2 samples the percentages of pristine vs. oxidized molecules -0.6 are 50-50 and 80-20, respectively, matching the values found for 710 715 720 725 730 the XLPA fitting (Fig. 5c). E (eV) Qualitatively, the XMCD of the OB2OX oxidized phases resemble that of the previously reported rectangular R2OX oxidized phase, which contained a major fraction of oxidized OX(OT-45) Figure 9. Above: XMCD of OB1OX phase (T=5.0 K, B=6 T, =0o): molecules and some pristine molecules occupying SB-30 sites. It is it can be described as a linear combination of the spectra of the interesting to note that the spectrum of the OX(OT-45) molecules SB-30, OX(OT-45) and OX(OT-30) molecules; below (vertically very closely corresponds to Fe3+ oxidation state in an octahedral shifted for comparison), spectra of Fe3+ and Fe2+ simulated with site, as previously shown by Ligand Field Multiplet (LFM) CTM4XAS 5.0 program32 including spin-orbit coupling, crystal calculations18 (Fig. S7). field (CF) effects and reduction of the Slater integrals to include the interatomic configuration interaction. The parameters used for the simulations were close to those of ref.12: the CF parameters were 10Dq=1.86 eV, 1.0 eV (Dt=Ds=0) for Fe3+Oh, Fe2+Oh (Oh: 0.5 (a) OB2OX2 octahedral site) respectively; a Slater reduction of 70%, 80%, 75% was used to F(dd), F(pd) and G(pd). The results were convoluted 0.0 by a Lorentzian of =0.3(0.4) eV for the L3(L2) edge to account XMCD (normalized) for intrinsic core-hole lifetime broadening, and by a Gaussian of -0.5 =0.2 eV to account for instrumental broadening. -1.0 In contrast, the XMCD of the OB1OX oxidized sample cannot OB2OXO2 be described considering only these two species. In view of the -1.5 50% SB-30 (OB2) STM images obtained in this sample (Fig. 3h), we require an 50% OX(OT-45)=fully oxidized R2ox additional component for the OX(SB-30) molecules that in the LC: 50% (SB-30) + 50% OX(OT-45) -2.0 XMCD accounts for the dichroic signal contribution observed at 705 710 715 720 725 730 735 710.6 eV. E (eV) This contribution can be empirically determined, as shown in Fig. 8b: to obtain the XMCD of the oxidized OX(OT-30) molecules (b) OB1OX (blue line), we subtract from the experimental XMCD of the 0.5 0.0 (SB-30o) OX(SB-30o) OX(OT-45o) XMCD (normalized) -0.5 -1.0 OB1OX -1.5 SB-30 OX(OT-45) OX(SB-30) -2.0 nhS+ 0.7 0.3 0.1 705 710 715 720 725 730 735 nhS- 2.7 3.6 4.0 E (eV) nh 3.4 3.4 4.1 Fe net charge 0.1 0.4 0.6 mTDFT (B) 2.1 3.3 4.0 Figure 8. Comparison of XMCD (=55o) of oxidized phases: (a) OB2OX2 (B=6 T, T=3.4 K): the measured spectrum can be Table 2. DFT simulations of STM images in CC mode at V=1 V expressed as a linear combination of the spectra of non-oxidized S S (top), and (bottom) the Fe 3d S+ holes ( nh ) and S- holes ( nh ), SB-30 molecules, and oxidized, rotated OX(OT-45) molecules; S S (b) OB2OX (B=4 T, T=6.0 K): the measured spectrum is a linear the Fe net charge, the number of holes ( nh nh  nh  nh ) and the combination of the spectra of three molecule species: SB-30, DFT OX(OT-45) and OX(SB-30). magnetic moment ( mTOT ), for the non-oxidized (left), oxidized- rotated (center) and oxidized non-rotated (right) species. This journal is © The Royal Society of Chemistry 20xx J. Phys. Chem. C 2020, xx, xx-xx Phase FePc mLxy (  B ) ms (  B ) mTOT (  B ) Factor R2OX SB-30 2.2x10-2 0.241 0.26 OX(OT-45) 1.1x10-1 2.062 2.17 8.3 R3OX SB-30 2.7x10-2 0.244 0.27 OX(OT-45) 2.1x10-1 1.59 1.80 6.7 OBOX SB-30 4.1x10-2 0.178 0.22 OX(OT-45) 1.1x10-1 2.062 2.17 9.9 OX(SB-30) 3.6x10-1 1.87 2.23 10.1 Table 3. Magnetic moments determined for the different types of molecules found in the oxidized FePc/Ag(110) phases. oxidized sample OB1OX a weighted sum of the XMCD from non- oxidized (SB-30) molecules (red line) and from oxidized OX(OT- 45) molecules (green line), see S9 for details. The residue XMCD peak at the energy of 710.6 eV in the spectrum of oxidized unrotated molecules, OX(SB-30), is quite different from the energy of the main peak (711.3 eV) obtained for the spectrum of rotated molecules, OX(OT-45). This suggests different Fe oxidation states in these two species. In order to test this hypothesis, we calculated the XMCD spectra of Fe3+ and Fe2+ spectra within a LFM model using CTM4XAS 5.032 . This version of the program allows only to simulate the XMCD spectra in normal incidence conditions. The results of our calculations are shown in Fig. 9: a linear combination of the pristine OB1 (weighed by 10%) plus the calculated Fe2+ and Fe3+ spectra (with weights of 49% and 41%, respectively) nicely Figure 10. (a) Percentage of oxygenated molecules, and (b) describe the experimental XMCD (at =0o) of the OB1OX phase. average magnetic moment per Fe atom in a ML sample, as a This calculation strongly supports our assignment of the oxidized function of the oxygen dose, for the three studied phases: the low- FePc at OX(OT-45) as Fe3+ and OX(SB-30) as Fe2+, justifying the density R2 and high-density R3/OB2 and OB1 phases. Lines are local spectroscopic difference of the two species and showing the guides for the eye. strength of the combination of STM, STS and XMCD techniques. To complete our study, we performed DFT calculations on the three different types of FePc species relevant for our work, namely, moment of the oxidized samples (  mTOT  per Fe atom) can be the pristine FePc on SB-30 positions, and oxidized molecules OX(OT-45) and OX(SB-30). The calculation yields the local tuned between ~ 0.2-2 B by controlling the O2 dose (see Fig. 10b). density of states, allowing to obtain the 3d spin-polarized number Conversely, this property might be potentially tailored to device a of holes nh , the Fe net charge, and the magnetic moment for each magnetic oxygen sensor. species, as reported in Table 2. In agreement with our XMCD and STS experimental results, 4. CONCLUSIONS DFT calculations show that oxidized unrotated OX(SB-30) In this work we have investigated the oxygen-induced magnetic molecules have a smaller Fe net charge (0.4 e-) than the rotated moment changes of FePc molecules of high-density sub-ML and OX(OT-45), with (0.6 e-). ML phases on the Ag(110) substrate. The oxidation rate of the We used the values in Table 1 together with the number of holes dense, quasi-squared R3/OB2 phases is comparable to the in the 3d band from DFT to obtain absolute values for the magnetic previously reported rectangular low-density (R1 and R2) phases. moments of the different molecular species. The results are listed in Owing to steric impediments, the oblique phase (OB1) is less Table 3. Upon oxidation, the total moment increases from mT=0.27 effective in the oxidation-reduction reaction. B to 1.2-1.6 B in the OB2 phase, which is a factor ≈6.5, and from In the R3/OB2 phases, the oxidation takes place accompanied by a lateral shift of the molecule, which moves the Fe centre from mT=0.18 B to 2.2 B in the OB1 phase, which is a factor ≈10, for SB to OT position, accompanied by a rotation of ≈15º, giving rise both species, i.e. OX(OT-45) and OX(SB-30). We note that DFT- to the FePc-(2-O2)-Ag complex. This is accompanied by an calculated total magnetic moment values are larger than the experimental ones, as previously reported18. Notably, although increase by a factor ~6.5 of the Fe total magnetic moment. Upon steric hindrance makes the accommodation of O2 in dense phases annealing, the magnetic moment decreases to a value close to that for the non-oxidized phase. However, full reversibility (as found in more difficult, we find that the increase in the magnetic moment of R1 and R2 phases) is prevented by the large disorder induced by the oxidized FePc molecules is of the same order as in the low- density phases (see Table 3). oxidation. In conclusion, Figure 10 summarizes the oxygenation and For the OB1 phase, oxygen diffusion becomes even more difficult. In order to accommodate O2, a fraction of molecules do magnetic moment capabilities found for the three studied rotate and move to OT-45 sites, while the rest is able to FePc/Ag(110) phases upon O2 dosing. As shown in Fig. 10a, the percentage of oxidized molecules in the sample can be controlled accommodate oxygen while keeping their initial SB-30 positions, by O2 dosing. The oxygenation process is most efficient in the low- due to the highly restricted molecular movement. The oxidation mode leading to the two different types of density R1/R2 phases, followed by the high-density R3/OB2 oxidized FePc species affects markedly the electronic density of phases, and is more difficult in the more distorted OB1 phase. Interestingly, by virtue of the upturn in the magnetic moment states, as proven by our STS and XMCD experiments, and DFT and produced in the oxidized FePc species, the average magnetic Ligand Field Multiplet calculations: the XMCD spectrum corresponding to the oxidized unrotated OX(SB-30) molecule 10 shows a characteristic of the Fe2+ oxidation state, while that of 104 (8), 083704. rotated OX(OT-45) is reminiscent to that of Fe3+. The Fe magnetic (5) Mugarza, A.; Robles, R.; Krull, C.; Korytár, R.; Lorente, N.; moment remains in all cases with planar anisotropy. Interestingly, Gambardella, P. Electronic and Magnetic Properties of both oxidized species show an increased magnetic moment by a Molecule-Metal Interfaces: Transition-Metal factor ~10. Phthalocyanines Adsorbed on Ag(100). Phys. Rev. B - Even though the molecular density of the R3/OB2 phases is Condens. Matter Mater. Phys. 2012, 85, 155437. larger than that of the OB1 phase, we find that the oxidation of https://doi.org/10.1103/PhysRevB.85.155437. R3/OB2 is easier, and the species of oxidized FePc resemble those (6) Palmgren, P.; Yu, S.; Hennies, F.; Nilson, K.; Åkermark, B.; found in the rectangular phases. Thus, the unit cell configuration of Göthelid, M. Changing Adsorption Mode of FePc on the ML assembly determines their oxygen capture properties. Our systematic study shows that despite the steric limitations TiO2(110) by Surface Modification with Bipyridine. J. FePc molecules in high-density FePc/Ag(110) phases are able to Chem. Phys. 2008, 129 (7), 074707. intercalate oxygen through different mechanisms. Therefore, (7) Åhlund, J.; Nilson, K.; Palmgren, P.; Göthelid, E.; control of the magnetic moment of the system in a range between Schiessling, J.; Göthelid, M.; Mårtensson, N.; Puglia, C. 0.2 to 2.2 B can be easily exerted through oxygen dosing. Molecular Growth Determined by Surface Domain Patterns. J. Phys. Chem. C 2008, 112 (17), 6887–6890. (8) Hu, F.; Mao, H.; Zhang, H.; Wu, K.; Cai, Y.; He, P. Electronic ASSOCIATED CONTENT and Structural Properties at the Interface between Iron- Phthalocyanine and Cu(110). J. Chem. Phys. 2014, 140, S1. Structural transition from R3 to OB2 phases; S2. Oxygen dosages; S3. Zoology of molecules in oxidized high-density phases; 094704. S4. DFT PDOS calculations; S5. Phase R3, N K-edge; S6. Angular (9) Scarfato, A.; Chang, S. H.; Kuck, S.; Brede, J.; Hoffmann, XAS & XMCD spectra for OB2OX1 and OB2ANN; S7. Angular G.; Wiesendanger, R. Scanning Tunneling Microscope dependence of magnetic moments for OB1; S8. R2OX XMCD and Study of Iron(II) Phthalocyanine Growth on Metals and Ligand Field Multiplet (LFM) calculations; S9. Percentages of Insulating Surfaces. Surf. Sci. 2008, 602 (3), 677–683. molecular species in OBOX oxidized phase. (10) Ohta, N.; Arafune, R.; Tsukahara, N.; Takagi, N.; Kawai, M. Adsorbed States of Iron(II) Phthalocyanine on Ag(111)  AUTHOR INFORMATION Studied by High-Resolution Electron Energy Loss Corresponding Author Spectroscopy. In Surf.&Interf.Anal.; 2014; pp 1253–1256. *Phone: +34 932805524. (11) Takami, T.; Carrizales, C.; Hipps, K. W. Commensurate E-mail: ebartolome@euss.es Ordering of Iron Phthalocyanine on Ag(111) Surface. Surf. Sci. 2009, 603 (21), 3201–3204. ACKNOWLEDGEMENTS (12) Åhlund, J.; Schnadt, J.; Nilson, K.; Göthelid, E.; Schiessling, We acknowledge financial support from the DWARFs project J.; Besenbacher, F.; Mårtensson, N.; Puglia, C. The MAT2017-83468-R and from the European Regional Development Adsorption of Iron Phthalocyanine on Graphite: A Fund (ERDF) under the program Interreg V-A España-Francia- Scanning Tunnelling Microscopy Study. Surf. Sci. 2007, Andorra (Contract No. EFA 194/16 TNSI). M.P. and A.M. are 601 (17), 3661–3667. funded by the CERCA Program/ Generalitat de Catalunya, and the (13) Dou, W.; Yang, Q.; Lee, C.-S. Anisotropic Film Growth of Severo Ochoa program from Spanish MINECO (Grant No. SEV- Iron-Phthalocyanine on Graphene on a Ni(111) Substrate: 2017-0706). The XMCD experiments were performed at the Roles of Molecule-Substrate and Intermolecular BOREAS beamline of the ALBA Synchrotron Light Facility Interaction. Appl. Phys. Lett. 2013, 102, 131606. (Barcelona, Spain) and at the ID32 beamline at the European (14) Palmgren, P.; Angot, T.; Nlebedim, C. I.; Layet, J. M.; Le Synchrotron Radiation Facility-ESRF (Grenoble, France). We are Lay, G.; Göthelid, M. Ordered Phthalocyanine grateful to the beamline staff for providing assistance in using both Superstructures on Ag(110). J. Chem. Phys. 2008, 128 (6). beamlines. (15) Casarin, M.; Di Marino, M.; Forrer, D.; Sambi, M.; Sedona, F.; Tondello, E.; Vittadini, A.; Barone, V.; Pavone, M. Coverage-Dependent Architectures of Iron  REFERENCES Phthalocyanine on Ag(110): A Comprehensive STM/DFT (1) Anand, M.; Siahrostami, S.; Norskov, J. K. Exploring the Study. J. Phys. Chem. C 2010, 114 (5), 2144–2153. Effect of Gold Support on the Oxygen Reduction Reaction (16) Sedona, F.; Di Marino, M.; Forrer, D.; Vittadini, A.; Casarin, Activity of Metal Porphycenes. Chem.Cat.Chem. 2018, 23, M.; Cossaro, A.; Floreano, L.; Verdini, A.; Sambi, M. Tuning 5505–5510. the Catalytic Activity of Ag(110)-Supported Fe (2) Bartolome, F.; Bunau, O.; Garcia, L. M.; Natoli, C. R.; Phthalocyanine in the Oxygen Reduction Reaction. Nat. Piantek, M.; Pascual, J. I.; Schuller, I. K.; Gredig, T.; Mater. 2012, 11 (11), 970–977. Wilhelm, F.; Rogalev, A.; et al. Molecular Tilting and (17) Cai, Y. .; Rehman, R. A.; Ke, W.; Zhang, H.-J.; He, P.; Bao, S. Columnar Stacking of Fe Phthalocyanine Thin Films on The Transition Behavior of FePc on Ag(110). Chem. Phys. Au(111). J. Appl. Phys. 2015, 117 (17), 17A735. Lett. 2013, 582, 90–94. (3) Stepanow, S.; Miedema, P. S.; Mugarza, A.; Ceballos, G.; (18) Bartolomé, J.; Bartolomé, F.; Brookes, N. B.; Sedona, F.; Moras, P.; Cezar, J. C.; Carbone, C.; De Groot, F. M. F.; Basagni, A.; Forrer, D.; Sambi, M. Reversible Fe Magnetic Gambardella, P. Mixed-Valence Behavior and Strong Moment Switching in Catalytic Oxygen Reduction Correlation Effects of Metal Phthalocyanines Adsorbed on Reaction of Fe-Phthalocyanine Adsorbed on Ag(110). J. Metals. Phys. Rev. B - Condens. Matter Mater. Phys. 2011. Phys. Chem. C 2015, 119 (22), 12488–12495. https://doi.org/10.1103/PhysRevB.83.220401. (19) Sedona, F.; Lo Cicero, M.; Carlotto, S.; Basagni, A.; (4) Molodtsova, O. V; Knupfer, M.; Ossipyan, Y. A.; Aristov, V. Fakhrabadi; Fakhrabadi, M. M. S.; Casarin, M.; Sambi, M. Y. Molecular Orientation and Ordering in CoPc and FePc Substrate Involvement in Dioxygen Bond Dissociation Thin Films Grown on Au(001)-5 X 20. J. Appl. Phys. 2008, Catalysed by Iron Phthalocyanine Supported on Ag(100). This journal is © The Royal Society of Chemistry 20xx J. Phys. Chem. C 2020, xx, xx-xx Chem. Commun. 2018, 54 (6). (20) Jones, E. .; Rocha, T. C. R.; Knop-Gericke, A.; Stampfl, C.; Schlöglb, R.; Piccinina, S. Jones. Phys.Chem.Chem.Phys 2015, 17, 9288. (21) Giannozzi, P.; Baroni, S.; Bonini, N.; Calandra, M.; Car, R.; Cavazzoni, C.; Ceresoli, D.; Chiarotti, G. L.; Cococcioni, M.; Dabo I., et al. QUANTUM ESPRESSO: A Modular and Open-Source Software Project for Quantum Simulations of Materials. J. Phys. Condens. Matter 2009, 21, 395502– 395520. (22) Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 1996, 77, 3865–3868. (23) Vanderbilt, D. Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue Formalism. Phys. Rev. B 1990, 41, 7892−7895. (24) Scherlis, D. A.; Cococcioni, M.; Sit, P.; Marzari, N. Simulation of Heme Using DFT + U: A Step toward Accurate Spin-State Energetics. J. Phys. Chem. B 2007, 111, 7384−7391. (25) Tersoff, J.; Hamann, D. R. Theory and Application for the Scanning Tunneling Microscope. Phys. Rev. Lett. 1983, 50, 1998. (26) Barla, A. A.; Nicolás, J.; Cocco, D.; Valvidares, S. M.; Herrero-Martín, J.; Gargiani, P.; Moldes, J.; Ruget, C.; Pellegrin, E.; Ferrer, S. Design and Performance of BOREAS, the Beamline for Resonant X-Ray Absorption and Scattering Experiments at the ALBA Synchrotron Light Source. J. Synchrotron Rad. 2016, 23, 1507–1517. (27) Kummer, K.; Fondacaro, A.; Jimenez, E.; Velez-Fort, E.; Amorese, A.; Aspbury, M.; Yakhou-Harris, F.; van der Linden, P.; Brookes, N. B. The High-Field Magnet Endstation for X-Ray Magnetic Dichroism Experiments at ESRF Soft X-Ray Beamline ID32. J. Synchrotron Rad. 2016, 23, 464–473. (28) Brookes, N. B.; Yakhou-Harris, F.; Kummer, K.; Fondacaro, A.; Cezar, J. C.; Betto, D.; Velez-Fort, E.; Amorese, A.; Ghiringhellib, G.; Braicovicha, L.; et al. The Beamline ID32 at the ESRF for Soft X-Ray High Energy Resolution Resonant Inelastic X-Ray Scattering and Polarisation Dependent X-Ray Absorption Spectroscopy. In Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment,903, 21; 2018; pp 175–192. (29) Bartolomé, J.; Bartolomé, F.; García, L. M.; Filoti, G.; Gredig, T.; Colesniuc, C. N.; Schuller, I. K.; Cezar, J. C. Highly Unquenched Orbital Moment in Textured Fe- Phthalocyanine Thin Films. Phys. Rev. B - Condens. Matter Mater. Phys. 2010, 81 (19), 195405. (30) Stöhr, J.; König, H. Determination of Spin- and Orbital- Moment Anisotropies in Transition Metals by Angle- Dependent X-Ray Magnetic Circular Dichroism. Phys. Rev. Lett. 1995, 75 (20), 3748–3751. (31) Bartolomé, J.; Bartolomé, F.; García, L. M.; Gredig, T.; Schuller, I. K.; Cezar, J. C. Magnetic Anisotropy in Fe Phthalocyanine Film Deposited on Si(110) Substrate: Standing Configuration. Low Temp. Phys. 2017, 43 (8), 955–959. (32) Stavitski, E.; de Groot, F. M. F. The CTM4XAS Program for EELS and XAS Spectral Shape Analysis of Transition Metal L Edges. Micron 2010, 41 (7), 687–694. 12 TOC image FePc/Ag(110) OB1 R3/OB2 0 XMCD (norm.) 0 XMCD (norm.) Non-oxidized -1 -1 Oxidized -2 -2 705 710 715 720 725 730 735 705 710 715 720 725 730 735 E (eV) E (eV) 2 nm This journal is © The Royal Society of Chemistry 20xx J. Phys. Chem. C 2020, xx, xx-xx