Iulia
Cojocariu
*abc,
Silvia
Carlotto
*de,
Daniel
Baranowski
b,
Matteo
Jugovac
bc,
Jan
Dreiser
f,
Luca
Schio
g,
Luca
Floreano
g,
Maurizio
Casarin
d,
Vitaliy
Feyer
*bh and
Claus M.
Schneider
bhi
aDipartimento di Fisica, Università degli Studi di Trieste, via A. Valerio 2, 34127 Trieste, Italy
bPeter Grünberg Institute (PGI-6), Forschungszentrum Jülich GmbH, Leo-Brandt-Straße, 52428 Jülich, Germany. E-mail: iulia.cojocariu@units.it; v.feyer@fz-juelich.de
cElettra-Sincrotrone, S.C.p.A, S.S 14 – km 163.5, 34149 Trieste, Italy
dDipartimento di Scienze Chimiche, Università degli Studi di Padova, via F. Marzolo 1, 35131 Padova, Italy. E-mail: silvia.carlotto@unipd.it
eInstitute of Condensed Matter Chemistry and Technologies for Energy (ICMATE), National Research Council (CNR), c/o Department of Chemistry, University of Padova, via F. Marzolo 1, 35131 Padova, Italy
fSwiss Light Source, Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland
gCNR-IOM, Lab. TASC, s.s. 14 km 163,5, 34149 Trieste, Italy
hFakultät f. Physik and Center for Nanointegration Duisburg-Essen (CENIDE), Universität Duisburg-Essen, D-47048 Duisburg, Germany
iDepartment of Physics and Astronomy, UC Davis, Davis, CA 95616, USA
First published on 23rd October 2023
In this study, the magnetic anisotropy of metal complexes is explored for its crucial role in the development of molecular materials for cutting-edge applications in spintronics, memory storage, and quantum computing. The challenge of achieving maximum magnetic anisotropy for paramagnetic single nickel ion sites is addressed and realized through an on-surface thermally induced planarization reaction in tetraphenylporphyrin, which maintains the nickel species in a square planar coordination environment. At the same time, the effective ligand field reduction due to the increased π-conjugation results in a lower reactivity of the molecular species. The results herein reported showcase the synergy between magnetic anisotropy and chemical robustness in single-site magnetic materials, thus opening exciting prospects for the development of stable uniaxial anisotropy in these materials. Such a finding represents a relevant advance in the field and validates a protocol for exploring magnetic anisotropy in metal complexes.
In this regard, a well-known surface-assisted ring-closure reaction in the case of tetraphenylporphyrin (TPP) results in the formation of a planarized porphyrin molecule with an extended π-conjugated macrocycle ring.13,14 Recently, the on–surface planarization reaction has also been suggested as an effective way to alter the magnetic properties, as evidenced by the increase of the effective spin moment in iron(II)-tetrabenzoporphyrin.15
The electronic and magnetic properties of surface-supported metal porphyrin arrays may be further modified by functionalizing the CM with axial ligands.5,16–18 Small ligands such as NO, NH3, and NO2 have been shown to bind to the porphyrin's CM, inducing changes in its spin state. The presence of new bonds with the surface and additional ligands increases the CM coordination number of the porphyrin complexes from four to five or six. As such, the coordination number increase is usually accompanied by a magnetic anisotropy decrease as a consequence of the complex symmetry reduction, which lowers the energy barrier between different magnetic configurations. Additionally, the presence of axial ligands may influence the CM oxidation state, spin state, and magnetic moment thus affecting the magnetic anisotropy. Factors promoting reactivity towards axial ligands include the partial occupation of the CM-based dz2 atomic orbital (AO), a low CM oxidation state, and the complex conformational flexibility.10,11,19,20 A precise control of these factors is pivotal to preserve both the magnetic anisotropy and a preferred magnetization direction.
However, the correlation between the planarization-induced changes in the electronic and magnetic molecular properties with the axial reactivity quenching and the anisotropy preservation has not been fully explored yet.
To fill this gap, in this study, a multi-technique approach based on photoemission and absorption spectroscopy along with X-ray magnetic circular dichroism (XMCD) combined with semiempirical crystal field multiplet calculations and density functional theory (DFT) simulations within the molecular cluster approach has been exploited to delve into the energy level alignment and charge distribution in the Ni 3d AOs for NiTPP and its planarized form, both of them supported on Cu(100). Although a substantial charge transfer and hybridization between the Ni 3d states and substrate 4s states leads to a magnetically active configuration in both molecular forms, their different electronic structure results in distinct magnetic and chemical behaviour. More specifically, the findings herein reported highlight the crucial role played by planarization in quenching the reactivity towards NO2 and preserving the molecular magnetic anisotropy. Indeed, the reactivity quenching has been associated with the weakening of the porphyrin ligand field, which accompanies the π-conjugation increase due to the ring-closure reaction.
As a whole, these insights provide valuable information for functionalizing surface-supported metal porphyrins, fine-tuning electromagnetic properties, and paving the way for applications such as single-ion catalysis, magnetism, and biosensing.
C K-edge NEXAFS spectra of NiTPP deposited on Cu(100) at 500 K and RT reveal changes in the linear dichroism, indicating that the flattening of NiTPP occurs only on the hot substrate (see Fig. 1(a)). In more detail, the π* resonance (labelled as BC) well evident in the RT s-polarized C K-edge spectrum associated with the phenyl rings (Phs) almost perpendicular to the macrocycle plane, is significantly reduced when NiTPP is deposited on the hot substrate consistently with a coplanarity established between the Phs (associated transition labelled as BC) and the macrocycle fragment (associated transition labelled as AC). As observed for H2TPP on Ag(111),13 the absence of the Ph-based π* resonance in the s-polarized C K-edge spectrum is a clear indication of the flattening of adsorbed porphyrin molecules, which results from a dehydrogenation reaction accompanied by the formation of an additional C–C bond between Ph rings and the porphyrinic macrocycle. The deposition of NiTPP on a hot substrate is then not only a key factor in transforming it into the corresponding flat derivative (hereafter, dhNiTPP) but, more generally, an efficient method to trigger the ring-closure reaction when the reagent is strongly anchored to the substrate, and the post-deposition annealing of the densely packed network formed at RT proves to be ineffective.21 Diverse planarized NiTPP derivates may be formed during the cyclodehydrogenation reaction,22,23 with two of them being predominant, namely the spiral one (spi-dhNiTPP), where each Ph ring joins an adjacent pyrrole (Py) group on the same side, and the rectangular one (rect-dhNiTPP), where two adjacent Phs join the same Py, thus leaving two Pys with no Ph decoration. DFT numerical simulations carried out to optimize the geometrical parameters of gas-phase spi-dhNiTPP and rect-dhNiTPP (see Section S3 in ESI†) found the former more stable than the latter by 0.31 eV. Insights into dehydrogenation pathways have been obtained following the on-surface reaction via photoemission orbital tomography (POT). Inasmuch as the dhNiTPP/Cu(100) interface has a preferential on-surface orientation and molecular levels characterized by sharp momentum maps, POT enables the assignment of the valence band (VB) spectral features to the initial molecular orbitals (MO) through the comparison of experimental and theoretical photoelectron angular distributions.24 The results clearly state that the spi-dhNiTPP/Cu(100) interface better matches the experimental momentum maps than the rect-dhNiTPP one (see Fig. 1(c) and Fig. S4 in ESI†), thus indicating the preferential formation of the spi-dhNiTPP species. Based on this, hereafter, dhNiTPP corresponds to the spi-dhNiTPP species. Fig. 1(d) reveals that the adsorbate–substrate interaction is quite similar in NiTPP and dhNiTPP. Indeed, the HOMO−1 and the energy-degenerate LUMO/+1 (former gas-phase highest occupied and lowest unoccupied MO)25 peaks lie at the same binding energy, the only difference being the nature of the molecular orbital observed in proximity of the Fermi level, the NiTPP LUMO+3 and the dhNiTPP LUMO+2 located at 0.2 and 0.4 eV, respectively; a difference ultimately due to the charge redistribution upon dehydrogenation of NiTPP. The similar charge transfer suggests a similar adsorption geometry both in terms of adsorption site and adsorption height.
Effects induced by the increased delocalization associated with the dehydrogenation reaction are not straightforwardly revealed if we limit our attention to the interaction between the CM and the first ligand sphere (Ni–N). In fact, the Ni–N σ bond (see the Ni–N Nalewajski–Mrozek bond multiplicity indexes),27 as well as the Ni–N bond length (see Table 1), are very similar in NiTPP and dhNiTPP.
Ni–N/Å | NMINi–N | NMINi–Cu | NMINi–NO2 | |
---|---|---|---|---|
NiTPP | 1.98 | 0.47 | — | — |
dhNiTPP | 1.95 | 0.46 | — | — |
Cu–NiTPP | 2.04 | 0.31 | 0.57 | — |
Cu–dhNiTPP | 2.01 | 0.31 | 0.54 | — |
Cu–NiTPP-NO2 | 2.12 | 0.26 | 0.07 | 0.37 |
Nevertheless, more information can be gained by looking at the changes in the Ni(II) 3d-based MOs population upon moving from NiTPP to dhNiTPP. In particular, the extended π-delocalization for dhNiTPP leads to increased stabilization of all the Ni 3d-based orbitals and affects not only the energy positions of the degenerate Ni(II) 3dxz- and 3dyz-based orbitals, which mainly overlap with the ligand-based π orbitals, (⊥t2g-like MOs are upward shifted by 0.4 eV) but also their localization (⊥t2g-like MOs are less concentrated on Ni by 23%) (see Fig. 2). Additionally, the planarization stabilizes both Ni(II) eg-like MOs, the ⊥eg-like (z2) by 0.3 eV and the unoccupied ‖eg-like (x2 − y2) by 0.18 eV. In particular, the ‖eg-like MO also shows a slightly lower degree of localization on nickel (by 4%) for the planarized system.
Previously, we showed that the strong charge transfer from Cu(100) to NiTPP implies nickel reduction.28,29 The 3d9 Ni(I) species thus formed is paramagnetic, and it carries one unpaired electron in the ‖eg 3dx2−y2 AO.10 The previously discussed similarity of the adsorbate–substrate interaction in NiTPP and dhNiTPP prompts us to assume that the CM reduction is preserved for dhNiTPP/Cu(100) (see Section S2 and S6 in ESI†).
XMCD measurements recorded in the presence of an external magnetic field applied along the light propagation direction have been carried out to look into the magnetic properties of the Cu-supported systems. Magnetic dichroism signal plots obtained from the absorption spectra recorded with different light helicities across the Ni L3,2-edges are displayed in Fig. 3 for normal (θ = 0°) and grazing (θ = 70°) incidences. XMCD spectra of NiTPP and dhNiTPP are similar when it comes to the energy position of the maximum of their signal. Perfectly in agreement with the Ni 2p core level spectra, this XMCD finding is consistent with a similar charge transfer at the interface. At the same time, the analysis of the XMCD signal shapes testifies to a quite big difference in the in-plane magnetic properties of NiTPP and dhNiTPP chemisorbed on Cu(100). The sum rule analysis30,31 has been then performed to estimate the corresponding orbital and spin angular momenta (the Ni(I) 3d9 configuration implies the assumption of a single hole in the 3d shell of both systems). The results of the analysis are reported in Table 2.
Fig. 3 Normal and grazing incidence XMCD spectra at the Ni L3,2-edge for NiTPP/Cu(100) and dhNiTPP/Cu(100) along with respective crystal field multiplet calculations. |
Normal | Grazing | ||
---|---|---|---|
NiTPP/Cu(100) | m l/μB | 0.16 ± 0.02 | 0.19 ± 0.03 |
m s,eff/μB | 1.34 ± 0.08 | 0.46 ± 0.09 | |
g-factor | 2.24 | 2.83 | |
dhNiTPP/Cu(100) | m l/μB | 0.10 ± 0.02 | 0.16 ± 0.03 |
m s,eff/μB | 1.36 ± 0.08 | 0.45 ± 0.09 | |
g-factor | 2.15 | 2.71 |
It emerges that the dhNiTPP in-plane orbital angular momentum is quenched, while the out-of-plane magnetic components change to a lower extent. Such evidence implies a variation in the degree of spin–orbit coupling (SOC) in NiTPP and dhNiTPP, whose effective g-factor values are also reported in Table 2 (deviation from the free ion value ∼2 quantifies the multiplet mixing). The larger NiTPP orbital component reflects a richer multiplet structure, well evident in the grazing geometry. As far as the spin momentum asphericity is concerned, this is similar for both complexes, thus justifying the assumption of a comparable charge density at the NiTPP and dhNiTPP MC as a consequence of the similar charge transfer taking place at the interface.15 Indeed, the magnetic anisotropy of the Ni species is preserved upon planarization.
To account for all these changes in the multiplet splitting features of the Ni L3,2 XMCD spectra, semiempirical crystal field multiplet (CFM) calculations have been carried out for both NiTPP and dhNiTPP by combining the QUANTY code with the CRISPY graphical interface.32,33 More specifically, CFM calculations were performed to gain insights into the highly localized Ni(I) 2p63d9 to 2p53d10 transitions in a D4h symmetry by taking into account Coulomb and exchange interactions, SOC, and crystal field (see Methods section). As such, CFM results state that differences between NiTPP/Cu(100) and dhNiTPP/Cu(100) XMCD spectra are mainly determined by the different ligand fields, whose effects have been modeled by varying the ligand–metal hybridization charge transfer (LMCT) term in the CFM calculations, which reflects the charge transfer and hybridization between the Ni 3d levels and the ligand. In more detail, the multiplet structure characterizing the NiTPP/Cu(100) XMCD spectra is well reproduced by setting the hybridization parameters corresponding to the eg-like levels (Va1g(3d,L1) and Vb1g(3d,L1)) to 1.0 and 0.3, respectively, while a satisfactory agreement between experiment and theory is obtained in dhNiTPP just by lowering the 3dz2-related Va1g(3d,L1) parameter to 0.75 with no other changes. This ultimately reflects an increased occupation of the Ni 3d AOs upon moving from NiTPP to dhNiTPP.
The panorama provided by the CMF semiempirical calculations well agrees with that pictured by DFT numerical experiments carried out within the molecular cluster approach (see Computational details).9,11,34,35 As such, the interaction of NiTPP and dhNiTPP with Cu(100) has been herein modeled by adopting two molecular clusters, namely Cu–NiTPP and Cu–dhNiTPP (see Section S7 in ESI† for details on the model used).
It has already been mentioned for the gas-phase calculations that the planarization is accompanied by the stabilization of both Ni(II) eg-like MOs. Such an effect is preserved on the surface and enhanced when the 3dz2⊥eg-like orbital, mostly involved in the axial interaction with the substrate Cu 4s AO (see Fig. 4), is considered. Theoretical Cu–Ni bonding energies are 0.31 eV and 0.24 eV in Cu–NiTPP and Cu–dhNiTPP, respectively, while the corresponding NMINi–Cu values are 0.57 and 0.53 (see Table 1). The picture emerging from these data is that, both in NiTPP and dhNiTPP, Ni(II) is reduced to Ni(I) upon coordination to Cu; moreover, the half occupation of the Ni(I) ‖eg-like orbital (Ni–N antibonding, see the corresponding 3D CPs displayed in Fig. 2) induces in both cases similar lengthening of the Ni–N bond (from 1.98 to 2.04 Å in Cu–NiTPP and from 1.95 Å to 2.01 Å in Cu–dhNiTPP, which tinily affects the NMINi–N values (see Table 1)).
In this regard, the Ni net spin polarization, evaluated as the occupation difference between α (↑) and β (↓) spins, amounts to 1.11 (1.06) in the Cu–NiTPP (Cu–dhNiTPP) cluster, which suggests a charge transfer slightly stronger in NiTPP/Cu(100) than in dhNiTPP/Cu(100) thus providing a rationale of the destabilization of the Ni 3 dz2(↓) ⊥eg-like MO, and its energy splitting in the former case.
NO2 is a strongly oxidizing gas that showed to be highly interacting with the pristine Ni(I)TPP systems already at room temperature after a few Langmuir exposure.10,36 For a more sophisticated description of the experimental findings, a theoretical description of the reactivity turns out to be necessary. The NO2 interaction with the surface-supported NiTPP has been modeled by considering the Cu–NiTPP-NO2 molecular cluster.
The NO2 interaction with the Cu–NiTPP systems is highly favored, with a bonding energy of 0.93 eV, in good agreement with previously reported DFT periodic calculations.10 The strong interaction reflects in a NMINi–NO2 value of 0.37 for the Cu–NiTPP-NO2 cluster. Upon NO2 coordination with the Cu–NiTPP complex, the porphyrin macrocycle adopts a slightly bent conformation reflected in a lower NMINi–N index (see Table 1). The interaction is strongly driven by the surface trans-effect,34,37 as testified by the significant NMINi–Cu index decreasing in favor of the NO2 interaction. This is reflected in the relocalization of charge on copper and the reoxidation to Ni(II) upon NO2 coordination.
As a general effect on the Ni 3d pDOS after NO2 coordination, there is a widespread shift of these MOs towards more negative energies: the Ni–NO2 bond stabilizes all the Ni 3d-based MOs (see Fig. 5). Perfectly in agreement with periodic calculations,10 the Ni(I) is oxidized to a Ni(II) high spin state with electronic configuration (dxy)2 (dyz, dxz)4 (dz2)1 (dx2−y2)1 (see Fig. 5) upon NO2 coordination (an t62g-like, ⊥e1g-like, ‖e1g-like Ni(II)). The presence of two unpaired electrons and the expected magnetic moment value after coordination well match with the magnetic dichroism signal at the Ni L3,2 edges and NEXAFS analysis. As expected, the Ni(I) 3d-based MO mainly involved in the oxidation process is the ⊥eg-like 3dz2 one, which transfers one of its two electrons to the NO2 LUMO, thus decreasing the magnetic anisotropy.10
No interaction between the NO2 ligand and the Ni ion is observed after the exposure of the dhNiTPP/Cu(100) system to the same NO2 amount (see XPS spectra at the N 1s and Ni 2p core level in Fig. 5(b)). The results herein reported prompt us then to ascribe the different reactivity to electronic and structural reasons: the 3dz2⊥eg-like MO is too deep in energy to efficiently interact with the half-occupied NO2⊥π* MO, while the extended π-conjugation restricts the molecular flexibility. The inability of the surface-supported dhNiTPP to effectively interact with NO2 also mirrors in a NMINi–NO2 value which equals 0.092 for Cu–dhNiTPP-NO2. This leads to a surface-supported system unable to axially coordinate with external ligands, allowing for the preservation of the highly reactive magnetic state and the in-plane magnetic anisotropy.
The NEXAFS spectra of the molecular films were finally normalized to the same reference spectra measured on the clean substrate.
This effective model is not intended to describe the adsorption configuration (experimentally found in the hollow site, with Ni at a distance of 1.93 Å from the surface29), however it accurately describes the NiTPP/Cu(100) molecular electronic properties by setting the Cu–Ni internuclear distance to 2.1 Å (see Sections S6 and S7 of ESI†). The same distance is assumed also for the Cu–dhNiTPP cluster because of the similar charge transfer observed on the Cu(100) surface.
To favor the comprehension of the ground state (GS) theoretical outcomes, the molecular eigenvalues have been graphically displayed as the partial density of states (pDOS) by using a 0.25 eV Lorentzian broadening factor. These plots, based on Mulliken's approach46 for partitioning the overlap density, allow an easy inspection of the atomic composition of MOs over a broad energy range. Partial density of states (pDOS) are defined as
NEXAFS Ni L3,2-edges spectra have been modeled by evaluating excitation energies (EEs) and corresponding oscillator strength distributions for transitions having the Ni 2p-based MOs as initial spin orbitals with the ORCA program.47 The DFT/ROCIS method48 has been exploited, which includes spin–orbit coupling (SOC) in a molecular Russell–Saunders fashion, by adopting the B3LYP exchange–correlation (XC) functional49 and by using the def2-TZVP(-f) basis set and the resolution of identity approximation (def-TZVP/J basis).50,51 The combined use of DFT and configuration interaction needs a set of three semi-empirical parameters (c1 = 0.18, c2 = 0.20, and c3 = 0.40), which have been calibrated by Neese.47 Zeroth order regular approximation has been adopted to treat the scalar relativistic effects.52 Numerical integrations for all calculations have been carried out on a dense Lebedev grid (302 points).53
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3tc02726k |
This journal is © The Royal Society of Chemistry 2023 |