Substrate driven Spin Crossover in a Fe(II) scorpionate complex.

: A new spin crossover complex based on a heteroscorpionate ligand was synthesized and characterized. Thin films were grown by sublimation in ultra-high vacuum on highly oriented pyrolytic graphite (HOPG) and on gold single crystal Au (111), and spectroscopically characterized through X-ray absorption and by X-ray photoemission. Temperature-dependent experiments on sub-nanometric deposits demonstrated that the thermally induced spin-crossover is preserved at a sub-monolayer (0.7 ML) coverage on HOPG, while deposits with similar thickness lose the switching behaviour on Au(111) surface. The system was unresponsive to light stimuli at low temperature indepently of the used substrate.


Introduction
Spin crossover (SCO) compounds are a class of magnetic molecules 1 where the electronic configuration of a metal ion can be trigged by external stimuli 2-3-4-5 . In particular, Fe(II) based SCO complexes showing a diamagnetic and a paramagnetic state, have been proposed to be employed in molecular electronic and in multifunctional spintronic devices. Indeed, SCO complexes have an enormous potential in this field because of the possibility to obtain active materials capable of interacting with charges (or spin) and not merely transport them 6-7-8-9 . Furthermore, using synthetic strategies it is possible to introduce new functionalities tailoring the properties of molecular materials [10][11] . The assembly of SCO molecules on a solid support is the fundamental prerequisite for molecular-device fabrication, and high-vacuum sublimation deposition in particular is the cleanest method to obtain high quality nanometric films suitable for that purposes, with a good control of the final thickness 12 . Unfortunately, only a few SCO complexes can be heated up to the sublimation temperature without thermal degradation 13,14 . Among them are the scorpionate 15,16 complexes [FeTp2] and [FeTp*2] (Tp = HB(pz)3, pz = 1H-pyrazol-1-yl, Tp* = HB(dmpz)3, dmpz = 3,5dimethyl-1H-pyrazol-1-yl). Both compounds show hysteretic SCO behaviour, with a conversion at high temperature for the first (300-460K) 17 and below room temperature (240-150K) 18,19 for the latter. Following the work of Buchen and Gütlich on this scorpionate family, our goal was to tune the ligand structure to obtain a SCO complex with a room temperature conversion. Here we report the synthesis, structural, magnetic characterization, and nanostructuration of a heteroscorpionate SCO complex [Fe(HB(pz)2(dmpz))2] 15,20 (hereafter called 1), the deposition of which has been attempted on Highly Ordered Pyrolytic Graphite (HOPG) and gold single crystal (Au(111)). Deposition of the complex was performed via ultra-high vacuum sublimation (UHV) and a full spectroscopic characterization has been performed: Xray Photoelectron Spectroscopy (XPS) to confirm the integrity of sublimated 1 (on HOPG and Au(111)) and X-ray Absorption Spectroscopy (XAS) to study the magnetic properties at the nanoscale (on both substrates). Indeed, literature reports already suggested that small changes in the structure of SCO complexes affects their switching properties 21 and a full characterization of their magnetic behaviour in nanostructures is mandatory. In nanostructured SCO complexes, the interaction with the surface can cause a total or a partial loss of the SCO properties of the system, due to a strong interaction with the substrate 22 or due to a partial degradation of the molecular structure 23 . However, the surface doesn't always have a negative role: an interesting modulation of the SCO properties, mediated by the interaction with the surface has been observed 24 for [Fe(H2B(pz)2)2bipy] (bipy = 2,2'-bipyridine) deposited on dielectric surfaces such as SiO2 and Al2O3. Indeed, a pinning of the LS state well above the bulk transition temperature has been observed for these nanostructured molecules onto these surfaces. Unlocking of the spin state, giving rise to a HS state, has been detected utilizing Xray irradiation or a mild heating as transition stimuli. In order to avoid such surface-induced pinning effects, the use of nonmetallic, weakly interacting substrates like HOPG might be a winning strategy, and indeed SCO complexes in direct contact with carbon-based substrates such as HOPG or graphene have been reported to retain switching properties. [25][26][27][28][29] We have thus studied the deposition of a submonolayer of 1 on these two surfaces, Au(111) and HOPG, that can promote different interactions with this molecule. We evaluated the substrate effect on the thermally induced spin crossover through the analysis of XPS and XAS spectra recorded as a function of temperature.

Synthesis and properties in bulk
Complex 1 may be compared to its two homoscorpionates analogues, [FeTp*2] and [FeTp2]. 30 [FeTp*2] is a classic example of Fe(II) SCO complex showing HS to LS conversion upon cooling. It has regained attention since the 2010's as one of the few sublimable SCO molecules, and was studied in bulk and thin films on various substrates. This complex has an asymmetric and hysteretic conversion, 31 and shows both soft X-ray induced spin state trapping (SOXIESST) [32][33][34] and remarkable light-induced spin state trapping (LIESST) 4,33,35 effects. An epitaxial relationship with Au(111) and Cu(111) was also demonstrated for submonolayers deposited on single crystals. 35,36 [FeTp2] has also attracted attention as it is sublimable and shows a LS to HS conversion above room temperature. Contrary to its methylated analogue that crystallizes in the P-1 space group 19 , [FeTp2] has a rich polymorphism that leads to unconventional SCO behaviour. Despite their similar structures on the molecular scale, steric hindrance and packing effects lead to very different properties for these two complexes. Hence the idea of modulating the properties by synthesizing a complex with intermediate composition, the heteroscorpionate 1. Preparing scorpionates ligands with a boron center bearing different azolyl moieties is not a new idea in itself. 16 However, the very low (<20) number of publications about these specific heteroscorpionates, from 1967 up to now, is indicative of the difficulty of such synthesis. The first attempt, reported in 1995, is the preparation of 1 and its analogue, [Fe(HB(pz)(dmpz)2)2], by Buchen and Gütlich. 20 Their work is based on Trofimenko's classical procedure, a condensation of excess azole with sodium or potassium borohydride. The reaction takes place in melted azole, and the degree of substitution can be controlled by varying the reaction temperature. Thus, a stepwise condensation with different azoles had been realized (see Figure 1a), with hardly any characterization of the obtained ligand. Wołowiec et al. 37 had shown later that this approach always leads to a mixture of ligands having the general formula HB(pz)x(pz')(3-x)with x = 0-3. In the 2000's, they published a series of articles presenting different heteroscorpionates ligands, [38][39][40][41][42][43] and developed a new one-pot synthesis in solution ( Figure 1b). While still leading to a mixture of ligands, this approach allowed for higher yields of the targeted heteroscorpionate. Purification was achieved after conversion of the crude salts to their Co(II) complexes. Preparation of pure heteroscorpionate ligands were reported around the same time by Connelly et al. 44,45 and Desrochers et al. 46 Both teams adopted a similar procedure called heterocycle metathesis (Figure 1c). The reaction proceeds by heating under vacuum a homoleptic tris(pyrazolyl)borate and the other heterocycle (in their cases, triazole derivatives), and is driven by the sublimation of the released pyrazole. This procedure was used for the quantitative substitution of pyrazole by one or two nitropyrazoles, 47 and the key parameters to control the reaction outcomes rationalized recently. 48 In our case, attempts of heterocycle metathesis proved unsuccessful (see further details in ESI). This is likely due to the similarity between pyrazole and dimethylpyrazole, both in terms of steric hindrance and HSAB properties (pKa values: dmpz = 4.06, pz = 2.49, 4-NO2pz = -2.0). 49 Consequently, we turned our attention to the other pathways, and obtained as expected mixtures of homo-and heteroscorpionate salts from paths A and B ( Figure 1). The presence of HB(pz)2(dmpz)was evidenced by mass spectrometry (MS) on the crude products, with HB(pz)3as the main component of each mixture, and a higher fraction of the desired heteroscorpionate from path B (Figures S2 and S3). When using ligand B for complexation without further purification, Fe(II) complexes could be isolated as microcrystalline violet powder composed of aggregated tiny needles. Elemental analysis seemed to be in reasonable agreement with complex 1 composition, but MS analysis revealed a mixture of complexes with the general formula [FeH2B2(pz)6-n(dmpz)n, with n= 1-5 ( Figure S7). It is worth noticing that, even though KTp predominates in the mixture of ligands, [FeTp2] is not observed in the crystals, due to a higher solubility of this complex in methanol. Much improved results are obtained by simply washing the ligands mixture with diethyl ether, which increases significantly the relative amount of target ligand to the point of making it the main component of the mix ( Figure S3). Then, crystals of pure complex 1 can be successfully obtained by slow diffusion crystallization in methanol, as seen by MS ( Figure S6). The product obtained is often contaminated by inorganic salts (complexation by-products) that precipitate in methanol, and are easily purified by some manual triage followed by sublimation on a cold finger (5 10 -6 mbar, 140 °C). A single crystal of good quality was selected and allowed us to determine the structure of the complex with excellent accuracy. At 120 K, we found that the compound crystallizes in the triclinic P-1 spacegroup, with two half-molecules of complex 1 in the asymmetric unit with the iron atoms lying on special positions (see all relevant details in Table S3, and inset in Figure 2). No significant residual peaks in the Fourier electron density difference map are seen that could suggest any situational disorder of the methyl groups on the pyrazole rings, a situation we observed in a previous crystallization batch starting from an impure ligand (see Table S3). The structure confirmed thus that we obtained the desired heteroscorpionate homoleptic complex. It is noteworthy that only the trans configuration for the two scorpionate ligands is obtained, as necessary with the special position of the iron atom on an inversion centre, when both cis and trans configurations are theoretically permitted. The presence of two independent Fe(II) complexes in the asymmetric unit differs from the unique molecule observed for the parent compounds [FeTp*2] 19 and [FeTp2] (for what is considered to be the most stable polymorph, in the monoclinic P21/n spacegroup). 50 Nevertheless, a similar case can be found in the first described polymorph of [Fe(HB(pz)3)2] (in the monoclinic P21/c spacegroup) 51 Figure S16 to S18). The average bond length between Fe(II) and the pyrazole ligands, at 1.984 and 1.979 Å for the two independent complexes, support a fully low-spin (LS) state for both molecules. This is also supported by considering the distortion parameters as calculated by the OCTADIST software 53 (see Table S4), and the Continuous Symmetry 54,55 and Shape 56 Measures as calculated by the online CoSym tool (see Table S5). Indeed, the calculated values show that at 120 K the two molecules of complex 1 are overall very close to C2 symmetry, with a FeN6 coordination sphere very little distorted respective to a perfect octahedron. Those values are intermediate between the strongly symmetric LS structure of [Fe(HB(dmpz)3)2] at 100 K, and the slightly more distorted LS structure of [Fe(HB(pz)3)2] at 180 K (see Tables S4 and S5). We collected data sets increasing the temperature up to 380 K then back to 300 K. Throughout these measurements, the crystalline symmetry did not change, with the triclinic P-1 spacegroup (see Table S3), and the crystal diffraction quality was preserved. The evolution of the cell volume V (Table S3 and Figure S19), average Fe-N bond length, distortion parameters  and  (Table S4 and Figure S20) and Continuous Shape Measure (Table S5 and Figure S20) clearly support that gradual spin crossover occurs, but starting only between 250 and 300 K. The values of those parameters at 380 K are found to be intermediate between those for fully high-spin (HS) [Fe(HB(dmpz)3)2] at RT and partially HS [Fe(HB(pz)3)2] at 420 K, supporting thus that the spin crossover is incomplete at the highest temperature reached. Cooling down, full reversibility was observed. No difference was observed between the two independent complexes in the unit cell. Bulk properties of the purified powder of complex 1 were assessed by Mössbauer spectroscopy at room temperature, and with a Vibrating Sample Magnetometer (VSM) up to 450 K ( Figure  2). At 295 K, in agreement with the behaviour observed on single crystals, the Mössbauer spectra shows the occurrence of two doublet resonances ( Figure S21). The first one shows an isotopic shift  of 0.41(2) mms -1 and a quadrupolar splitting  of 0.15 (5) 57 The second doublet shows an isotopic shift  of 1.07(4) mms -1 and a quadrupolar splitting  of 3.20(9) mms -1 . Those values are again typical but for HS Fe(II), and absolutely comparable to values found for HS [FeTp*2] (1.03/3.67 mms -1 at 300 K), or the small HS fraction surprisingly induced on [FeTp2] by very high pressure (0.86/3.23 mms -1 at RT under 7.8 GPa). 57 The usual approximation is to assume similar Lamb-Mössbauer factor for the HS and LS chromophores, which yields from the ratio of areas a HS fraction (nHS) of 25±5%. The error bar on this value is significantly high, due on one hand to the moderate signal/noise ratio of the collected spectrum, and on the other hand to very different linewidths for the two signals (0.53(4) and 0.82(9) mms -1 respectively), pointing to a likely dissimilarity of the Lamb-Mössbauer factors for the two states. Magnetic susceptibility  shows the expected behaviour for a spin crossover complex ( Figure 2). The T product at 180 K is 1.5 10 -6 m 3 Kmol -1 , indicating an almost fully LS complex. It increases very gradually up to 40.6 10 -6 m 3 Kmol -1 at 480 K, just above the 37.7 10 -6 m 3 Kmol -1 spin-only value for a fully HS S=2 Fe(II) complex. Considering then we have a fully HS state at that temperature, we calculated the nHS(T) curve and obtain a T1/2 of 312(2) K. Fitting with a simple Boltzmann dependence 58 yields H = 19.5(5) kJmol -1 and S = 60.8(2) JK -1 mol -1 ( Figure S22). The latter values are in line with usual values for SCO complexes. 58

Characterization of thin films
As reported in the synthesis section of complex 1, it is easy to sublimate it. In order to chemically characterise sublimated deposits, a sub-monolayer on both the surfaces has been studied by Variable Temperature XPS (VT-XPS) and structurally studied by STM at 25K. It is worth noticing that a low sticking coefficient has been revealed for 1 by XPS and XAS data after the first layer is deposited. Even if the initial deposition rate estimated in our apparatus was 2nm/h, it was not possible to grown more than 1 ML, even exposing the surfaces for very long time (4 hours) to the molecular flux. (see Table S6). STM at 25K reveals a coverage of a monolayer both on HOPG and Au(111). By STM it was possible to observe a molecular lattice on both the surfaces ( Figure S23), revealing the presence of almost spherical object with a dimension of ca. 0.5 nm, in agreement with average molecular dimensions taken from the crystal structure. The stoichiometry of these deposits has been estimated by semiquantitative analysis of the XPS data and compared with the one detected on the powder sample. This elemental analysis (see Table S7) has been performed in order to evaluate the molecular integrity of the sublimed films. For the bulk powder and for all the monolayers, the N1s region features one main component at 399.9 eV ( Figure S24) attributable to the nitrogen atoms bonded with boron and to the nitrogen atoms involved in iron coordination, 59 plus a small shake up located at 401.4 eV. The C1s for the powder and for the monolayer on Au(111), has been fitted using three components: one related to aliphatic carbon at 284.4 eV (ca 42%) plus another one given by aromatic carbon and C-N placed at 285.2 eV (58%) plus its shake-up placed at 286.6 eV. The scenario behind the Fe2p region results a bit more complex, due to the presence of a fine structure indicative of the valence and spin state of the iron as it has been already reported by some of us for other SCO systems. [10][11][12] The Fe2p region has been deconvoluted with five components for both 2p1/2 and 2p3/2 peaks ( ). Their energy and the presence of some shake-up features is typical of a paramagnetic Fe(II) ion 60 and they have been found both in the bulk powder and the nanometric films, coherently with mixed spin states of the complex at room temperature. The Fe/N ratio has been identified as the best parameter to evaluate molecular integrity, since iron and nitrogen are not overlapped by other elements (e.g. C1s of 1 is totally covered by the contribution from HOPG crystal). The N/Fe ratio thus obtained is 12.9 for the bulk powder, 11.9 for the film on HOPG and 12.1 for Au(111), all in agreement with the theoretical value of 12.0. From this ratio it is possible to assume that the stoichiometry of the SCO complex is well retained after the sublimation, both on HOPG and on Au (111). In Table S6 we report the complete stoichiometry analysis both on bulk and on the sub-monolayer coverages on HOPG and Au(111). The spectra variation with temperature reported in Figure  3reveal that qualitatively spin crossover occurs for complex 1 in the bulk as expected but also for the sub-monolayer coverages. Nevertheless, significant changes respective to the bulk are observed depending on the substrate. The Fe2p XPS temperature dependence on HOPG follow the trend for the SCO expected from the T plot (Figure 2), as seen by plotting the sum of the A and A' components as a function of the temperature ( Figure S24). For the deposit on Au(111) the temperature variation seems to be less efficient, as it's possible to see from the A+A' ratio vs temperature reported in Figure S24. Moreover it has been observed as well that the reversibility after the heating procedure is not retained on Au(111), and it presents a pinning of the switchability as is possible to see also from Figure S25. This scenario can be followed also by Ultraviolet Photoelectron Spectroscopy (UPS), 7,10,61 that reveals a non-complete reversibility of the molecular layer once adsorbed on Au (111) at variance with what observed on HOPG (see figure S26).
To shed some light on this behaviour we performed a temperature dependent XAS experiment at the ESRF ID32 beamline 62 (Figure 4). Variable temperature spectra are related to a sub-monolayer deposit on HOPG (Figure 4a bottom) and on an Au(111) single crystal (Figure 4a top). To verify the trend found by XPS, the XAS spectra were acquired cooling down, starting from 320 K. For comparison, Figure 4a reports XAS spectra at the lowest and the highest temperature for 0.7ML of 1 on HOPG and on Au(111). In order to be able to obtain a quantitative HS fraction of the two systems, experimental Fe L3 edge spectra have been linearly interpolated employing a weighted sum of reference spectra, as described in the method section (see details in Figure S27). From the results of the interpolation, we have extracted the HS fraction nHS for each temperature and for each sample, yielding the plot reported in Figure 4b. Thanks to this approach, it is possible to compare the SCO behaviour of the three nanostructured systems and the bulk material. The line shape of the spectra recorded on the 0.7ML nm deposit on HOPG in the range of temperature from 2 K to 320 K appears compatible with an Fe(II) system in an octahedral environment. 63 From 2K to 200K complex 1 appears to be in the LS configuration, as expected from magnetometry and crystallography results on the bulk material ( Figure 2 and Figure S20). Increasing the temperature above 200 K, the line shape of the Fe L3 edge evolves in line with an increase of the HS contribution, up to 320 K (maximum operation temperature of the set-up available on ID32). This is in line with our hypothesis that HOPG is an "innocent" surface for this compound and does not have a major influence on the thermally driven spin conversion, as seen from VT-XPS on HOPG as already observed by some of us 61 and reported in literature. 14,25,64-66 On the other hand, the XAS characterizations carried out on a similar deposit of 0.7 ML on Au (111) (Figure 4a), from 2 K up to 320K, reveal a different scenario: the sample seems to show already a high contribution of the HS species even at cryogenic temperature, where only LS species should be present. To simplify the comparison the high spin fraction for the bulk material is also reported in Figure 4b, evaluating it directly from the magnetic susceptibility data 67 . This comparison shows that the magnetic behaviour of molecules of 1 on HOPG is comparable with the bulk, even in the case of 0.7 ML where the investigated molecules are in direct contact with the surface. Like mentioned above, the spectra of complex 1 on Au(111) show a coexistence of both HS and LS states even at cryogenic temperature, suggesting a modification of the coordination environment respective to intact molecules, induced by the gold surface. On the other hand, the spectra show a small evolution upon increasing the temperature and this may be attributed to the presence of a fraction of intact molecules preserving SCO behaviour. These effects could be ascribed also to a hybridization with surface states, leading to a broadening and chemical shift of the Fe edge. 64 It is noteworthy that similar spin pinning effects are reported in the literature for sub-monolayers and thin films of [Fe(Tp*)2] on Au(111) 33,36,68 and Cu(111) 69,70 . An important parameter that can be extracted from the XAS spectra is the evolution of the branching ratio as a function of temperature, that can be used to evaluate the spin transition. The branching ratio is defined as the ratio between the intensity of the L3 iron edge and the total edges (L3+L2). The obtained branching ratios value are 0.63 and 0.74 (2 and 320 K respectively) for the 0.7ML film on HOPG, and 0.75 and 0.72 (2 and 320 K respectively) for the submonolayer on Au(111). The two values on HOPG are in line with values reported in literature on another SCO complex 71 but in the two spin states. The gradual decrease of the branching ratio from high to low temperature is typical of a HS-LS crossover, due to the vanishing of the spin orbit coupling in the initial state. In these conditions the hole has randomly oriented spin and angular momenta and the integrated intensities of the L3 and L2 iron edges depend exclusively on the final states statistics. 72 77 has been detected in the 0.7ML film on HOPG (see Figure S28 in Supplementary Information). We checked by performing a photomagnetic experiment on bulk microcrystalline powder within a SQUID magnetometer equipped with a dedicated setup. 78,79 Some very limited photoconversion (a few %, see Figure S29 in Supplementary Information) was observed using either 532 or 650 nm laser light, but relaxation was observed to be quite fast, with a T(LIESST) about 14K. No significant reverse-LIESST effect was observed when irradiating at 830 nm. This quite fast relaxation of the photoinduced state could be expected on the basis of the relatively high SCO T1/2 of 312 K, while the limited photoconversion is certainly due to the strong purple colour of the complex in the LS state. While absorption effects are absent for the submonolayers, the total absence of photoinduced effect may point to the surface inducing an even faster relaxation, making it so that the metastable photoexcited state has shorter lifetime than the XAS measurement.

Conclusion
Although SCO complexes have been studied in an enormous variety of nanostructures, only few examples of molecules sublimable in high vacuum and retaining their switching behaviour at the monolayer level have been reported so far. In light of the very interesting properties shown by [Fe(Tp*)2], with remarkable SOXIESST, [32][33][34]33,35 properties, and epitaxial growth on Au(111) and Cu(111) single crystals, 35,36 we engineered the synthesis of a mixed heteroscorpionate ligand which yielded complex 1. Complex 1 showed gradual spin crossover above room temperature. Its isolation as a pure species allowed us to show that it can be thermally sublimated in high-vacuum environment into monolayer films.
In this study we demonstrated that the SCO complex 1 can be deposited on a surface while conserving its chemical integrity, that the thermally induced SCO is retained and does not change significantly, at the monolayer level, when compared to the reference crystalline bulk sample once the SCO has been sublimated on HOPG.
This study confirms that HOPG surfaces are weakly interacting substrates suitable for SCO molecules deposition. On the other hand, Au(111) surfaces inhibit the SCO behaviour of the same compound. Therefore, the crucial role played by the interface between molecular SCO materials and inorganic materials has been confirmed. Such understanding paves the way towards molecular hybrid, multifunctional electronic and spintronic devices based on SCO complexes. Thanks to the temperature range of the investigated SCO enclosing room temperature, experiments aimed at investigating the electrical conductance of nanometric films of complex 1 are currently underway in our laboratories.

Synthesis of KHB(pz)2(dmpz):
The procedure was adapted from ref 42  In order to separate the compound from the potassium salts that are precipitating during the diffusion, the reaction products are suspended in ethanol and filtered, then washed with small amounts of water until the liquid phase remains colourless. It is then rinsed with ethanol and dried in air. The complex is obtained as dark violet microcrystals (37.6 mg, 21% yield). Powder X-ray diffraction patterns were acquired on a PANalytical X'Pert MPD PRO diffractometer with Bragg-Brentano geometry, Cu K radiation ( = 1.54184 Å) and a secondary graphite 370 monochromator. It was measured on an aluminium sample holder, on a 8-80° 2 angular range over 2096s. Data were analyzed with Jana2006 software. 80 Single-crystal X-ray diffraction experiments were performed on a Bruker-Nonius -CCD diffractometer with Mo K radiation (0.71073Å). Data collection was performed between 120 and 380 K, using an Oxford CryoSystem 700 installed on the diffractometer, with the crystal glued to a glass fiber using nail varnish. Data was collected using sets of  scans at different  or 2 angles, reduced and scaled using the DENZO and Scalepack softwares. 81 The structural determination by intrinsic phasing and the refinement of atomic parameters based on full-matrix least squares on F 2 were performed using the SHELXT 82 and SHELXL-2018 83 programs within the Olex2 package. 84 All refinement details are given in Table S3  Lorentzian linewidth and relative areas) were refined using both homemade programs and the WinNormos® software (Wissenschaftliche Elektronik GmbH). Vacuum sublimation has been carried out using a home-made Knudsen cell, the sublimation temperature was 410K measured by a K thermocouple with a base pressure of 10 -8 mbar. The same Knudsen cell has been employed to sublimate in-situ with a base pressure of 10 -9 /10 -10 mbar at ID32 beamline at ESRF 62 . For inhouse experiment and for Large Scale Facilities experiments the nominal thickness was estimated using a quartz crystal microbalance (QCM). XPS data were acquired using monochromatic Al Ka radiation (h = 1486.6 eV, SPECS mod. XR-MS focus 600) operating at a power of 100 W (13 kV and 7.7 mA) and a SPECS Phoibos 150 1DLD electron analyser mounted at 54.441 to the X-ray source. The XPS spectra were collected at normal emission with the fixed pass energy set to 40 eV. The spectra were analysed using the CasaXPS software. All the spectra were calibrated at the Fermi value of the substrate. The background in the spectra was subtracted using a linear background, and the deconvolution of the XPS spectra was carried out as a combination of Gaussian and Lorentzian functions (70/30). Crossing sections values for semiquantitative analysis have been provided from Elettra synchrotron (https://vuo.elettra.eu/services/elements/WebElements.html). XPS spectra related to N1s, C1s, B1s and Fe2p have been deconvoluted employing the same Voigt components for both bulk powder and deposited film. The variable temperature experiment was performed by using a liquid nitrogen-based cryostat connected to the XPS sample holder. Every spectrum represented herein results from averaging 16 spectra collected after 1 h of thermalisation at a specific temperature. STM measurements were carried out using an Omicron VT-STM at 25 K for the characterization of molecular deposits. All the images were acquired using an electrochemically etched W tip. Images have been analysed with the Gwyddion software. 86 HOPG (ZYA grade, purchased from NT-MDT) was freshly exfoliated and once introduced in UHV warmed up to 420K immediately before the deposition that was then performed when this substrate was back at room temperature. Au (111) single crystal was freshly prepared using a standard procedure involving a sputtering and annealing procedure. XAS characterization has been performed at ID32 at the European Synchrotron Radiation Facility (ESRF) on a UHV compatible pumped 4 He cryo-magnet 62 . XAS spectra were measured in Total Electron Yield (TEY) detection mode to guarantee the optimal surface detection sensitivity. All the characterisations were performed using a low density of photons in order to avoid radiation damage, as checked by the absence of evolution of RT spectra. Estimation of the temperature dependence of the nHS-Fe(II) molar fraction of the submonolayer coverage was performed through least-squares interpolation of normalized L3 XAS spectra with two reference spectra of 100% LS and 100% HS as we have already done in 61,63,87 . These reference spectra have been obtained from experimental data acquired during this experiment: 100% LS reference is the spectrum related to 0.7ML film on HOPG measured at 2 K, while 100% HS reference has been obtained by subtracting the LS reference multiplied by a coefficient from the spectrum measured at 320 K. Irradiation experiments were performed using a ThorLabs CLD101 compact laser diode controller, and 516 and 660 nm single mode fiber-pigtailed laser diodes. Laser light was directed inside the UHV cryo-magnetic chamber through a glass view port using a fibered reflective collimator, at a distance of ca. 36 cm of the sample, with a laser spot about 1 cm width and fluences on the same order of magnitude than our photomagnetic experiments.

Conflicts of interest
There are no conflicts to declare.