Enhanced Cooperativity in Supported Spin-Crossover Metal-Organic Frameworks.

The impact of surface deposition on cooperativity is explored in Au(111)-supported self-assembled metal-organic frameworks (MOFs) based on Fe(II) ions. Using a thermodynamic model, we first demonstrate that dimensionality reduction combined with deposition on a metal surface is likely to deeply enhance the spin-crossover cooperativity, going from γ3D = 16 K for the bulk material to γ2Dsupp = 386 K for its 2D supported derivative. On the basis of density functional theory, we then elucidate the electronic structure of a promising Fe-based MOF. A chemical strategy is proposed to turn a weakly interacting magnetic system into a strongly cooperative spin-crossover monolayer with γMOFAu(111) = 83 K. These results open a promising route to the fabrication of cooperative materials based on SCO Fe(II) platforms.

ture of a promising Fe-based MOF. A chemical strategy is proposed to turn a weakly interacting magnetic system into a strongly cooperative spin-crossover monolayer with

TOC ENTRY REQUIRED
Molecular electronics originally proposed in the 1950s and reinvigorated by the founding proposition of Aviram and Ratner 1 has led to intense activity on the study of electronic transport through single molecules, supported by the advent of scanning tunneling microscopy (STM) and break junction setups. 2,3 More recently, molecules entered the field of spintronics, i.e. the control and manipulation of spins, 4 and gave birth to molecular spintronics. [5][6][7][8] Whereas single molecule magnets (SMM) display bistability at rather low temperatures, 9,10 higher temperature regimes are reached for spin-crossover (SCO) materials. 11 In order to bridge the gap between functional molecules and practical devices, one has to deposit the objects on a surface while retaining the bistability property. With respect to the intense activity in the field of adsorbed SMM, 8,[12][13][14][15] SCO supported compounds have received much less attention. It is only recently that supported examples have emerged, starting from 3D materials with thin-films grown on gold. 11,16,17 0D systems followed based on single SCO molecules, [18][19][20][21][22][23][24] or molecular junctions. 25 However, 2D monolayer organized networks exhibiting SCO behavior have not been reported so far. Not only should lowdimensional SCO patterns offer new insights into the spin transition phenomenon, but such organized networks may also offer the opportunity to control the intermolecular interactions dictating the collective behavior of transiting centers (a.k.a. cooperativity). 26 With this goal in mind, a promising route might come from the use of self-assembled MOF. 27 Indeed, exchange couplings 28 were reported in several 2D MOF based on 3d ions such as Fe, 29,30 Ni, 31 Mn 32,33 or Cu. 34 Recently, Umbach et al. have investigated a Au(111)supported MOF based on Fe(II) ions and 2,4,6-tris(4-pyridyl)-1,3,5-triazine (T4PT) ligand. 30 Interestingly, in addition to the observation of a weakly ferromagnetic behaviour, the speculated structure of the system consists in FeN 6 units, an archetype arrangement of SCO systems. 35 In this letter, we use a model based on Fe(II) transiting unit to investigate the effect of dimensionality reduction and surface deposition. Our thermodynamic model shows that a metallic substrate is likely to enhance cooperativity by an order of magnitude. In the light of these findings, we then perform calculations based on density functional theory (DFT) to establish the structure of the synthetic Fe-T4PT MOF supported on a Au (111) surface. 30 We show that the most stable structure consists of FeN 3 Au 3 units rather than on the expected FeN 6 building blocks. Beyond the interpretation of experimental results, a chemical modification of the ligand is proposed to generate a FeN 6 candidate for spincrossover behavior. Finally, we demonstrate that the Au(111) surface is likely to drive the original weakly-coupled system into a highly cooperative material.
From a thermodynamic point of view, the low-spin (LS) to high-spin (HS) transition displays an overall entropy increase resulting from (i) the spin change (from LS, S = 0 to HS, S = 2 in the case of Fe(II) ions), and (ii) the weakening of the metal-ligand bond strength (vibrational contribution). As a consequence, for the transition to occur the enthalpy change must be positive as well. This quantity can be approximated by the electronic adiabatic energy difference ∆E adia , i.e. the energetic difference between the HS and LS states each taken at its optimized geometry. It is accessible from quantum chemistry calculations. Naturally, weak fields that favor HS over LS as the ground-state are not suitable for SCO. In the same way, too strong fields will lead to a blocked LS state. For Fe(II)-based compounds, ∆E adia is traditionally of the order of 100 meV when room-temperature transition is observed.
The steepness of the transition phenomenon is characteristic of the extent of cooperativity, a much more difficult quantity to retrieve from calculations. Indeed, interactions between transiting units can be mediated by phonon coupling, electron-phonon coupling 36,37 and driven by the Madelung field modulation in the crystal. 26,38 In the following, we intend to inspect changes in charge distribution resulting from dimensionality reduction and surface deposition. Thus, let us briefly recall the so-called polarization contribution. 38 In a mean-field scheme, the molar Gibbs free energy G of a mixture of spin transiting centers as a function of the HS molar fraction x reads where G ni = xG HS + (1 − x)G LS is the free energy of non-interacting sites. S mix is the entropy of mixing −R(x ln(x)+(1−x) ln(1−x)). G pol arises from the electrostatic potentials generated by the rest of the transiting sites and possibly the metallic surface at the position of (i) Fe ions (V LS , V HS ), and (ii) ligands (v LS , v HS ). 26,38 This contribution G pol to the Gibbs free energy reads : where ∆Q = Q HS − Q LS is the charge variation on the Fe center. δV HS and δV LS are the potential differences between metal and ligand positions in the HS and LS state, respectively.
In order to modulate the dimensionality of the materials, let us consider a model system consisting in neutral Fe(II) units in a typical N 6 environment. Fe(NCH) 4 (NCS) 2 units are formed from a formally Fe 2+ ion surrounded by four neutral NCH and two thiocyanate NCS − ligands. Charges are extracted from ab initio calculations, 26 and summarized in Table 1 along with the geometrical parameters defining the hypothetical 3D cubic crystal structure. we considered a 3D network, then a free-standing network, and finally a 2D network supported on a metallic surface. ∆E adia and ∆S are chosen to establish a critical temperature T 1/2 equal to 200 K. Let us stress that these parameters naturally affect the shape of the transition but not the γ value. From the values given in Table 1, the 3D network exhibits a weak cooperativity with γ 3D = 16 K and a smooth transition is observed ( Figure 1). Figure 1: Simulated temperature-induced spin-crossover for the model FeN 6 units organized in (i) a 3D network, (ii) a free-standing 2D network, and (iii) a 2D network supported on a metallic surface generating mirror charges. ∆E adia is 140 meV and ∆S is set to 0.70 meV/K leading to a critical temperature T 1/2 of 200 K. Deposition on a surface leads to a large increase of the polarization-induced cooperativity and, thus, to a more abrupt transition.
Upon 3D to 2D dimension reduction, the cooperativity is moderately increased by a factor of 3.5 with γ free 2D = 57 K. Formally, half of the point charges generating the Madelung field were set to zero when moving to this free standing 2D network. The presence of a metallic surface is likely to change this state of affairs. From elementary electrostatics, each charge above the equipotential metallic surface generates a mirror charge within the bulk. Quite remarkably, the cooperativity parameter is enhanced to γ supp 2D = 386 K. As a main conclusion, the polarization induced by the proximity of a metallic surface drives the system into a highly cooperative regime, featured by an amplification factor γ supp 2D /γ 3D for cooperativity larger than 20. Naturally, our model is likely to exaggerate polarization effects. Indeed, the limited size of the ligands concentrates the charge variation, with respect to more realistic ligands, and therefore might artificially amplify the role of the surface. To support these findings on a simplified model, we then assessed its robustness with a synthetic Fe(II)-based MOF self-assembled on Au(111). 30 First, structural and electronic information are explored from DFT calculations. Out of atomically resolved STM topographic images, Umbach et al. have  Figure S1 and Table S1 in Supporting Information), while the Fe-N distances with the second triazine layer are larger than 4Å. Therefore, the system does not present itself as an expected octahedral FeN 6 coordination sphere but rather as a trigonal prismatic FeAu 3 N 3 one. Trigonal prismatic geometry is well known in Fe(II) complexes. In particular, it is often observed when a geometric constraint is imposed, e.g. by macrocycle ligands. 39,40 Here, the constraints is imposed by the periodicity of the hybrid network (Fe-T4PT and Au (111)). This optimized structure leads to STM simulated images that compare well with the experimental ones (Figure 2-e,f). 30 A low bias constant current image highlights the first layer, while a higher bias reveals the position of the second layer.
The computed Bader charge for Fe (+1.76 e − ) is compatible with a +II formal oxidation state. This is confirmed by the analysis of the density of states (DOS, Figure 2-g). The iron levels are typical of a HS Fe(II) ion with five electrons occupying the spin-up 3d levels, while a single one lies in the spin-down 3d levels. This picture is also supported by X-ray absorption spectroscopy (XAS) and X-ray magnetic circular dichroism (XMCD) measurements. 30  Here, we restrict our study to two tridentate N-terminated ligands. As a first guess, we consider the 1,4,7-trimethyl-1,4,7-triazacyclononane (tacn) ligand, known to generate Fe(II) compounds exhibiting temperature-induced SCO. 41 As hopped, the iron center leaves the gold hollow site and a N 6 coordination sphere is restored after DFT geometry optimization. A similar picture holds for tris(pyrazolyl)-borate (p-IC 6 H 4 )B(pz) 3 ligand (Figure 3-a), a well-known negatively-charged ligand in SCO chemistry. 42,43 In this first step, we have conserved the original periodicity of the network imposed by the commensuration with the Au(111) surface. In that configuration, Fe-T4PT distances (≈ 2.15Å) are too large to stabilize the LS state. However, since the Fe ion is not bound to the Au hollow site anymore and that the T4PT is weakly adsorbed to the substrate, there is no reason for the MOF to conserve the same periodicity. From a computational perspective, it is not affordable to vary the periodicity of the MOF network while keeping the gold substrate fixed. Thus, we considered the free-standing Fe(T4PT)(tacn) and Fe(T4PT)(p-IC 6 H 4 )B(pz) 3 2D structures and optimized the atomic positions as well as the unit cell vectors (Figure 3-b) for the HS and LS states.
Unfortunately, a negative adiabatic energy difference ∆E adia = −340 meV is calculated for the tacn ligand. Thus, the HS state remains the ground-state and no SCO behavior can be anticipated. A contrasted scenario is observed when T4PT is substituted by (p-IC 6 H 4 )B(pz) 3 . The adiabatic energy difference is much reduced and remains positive, ∆E adia =140 meV ( Figure 3-c). Meanwhile, the equilibrium geometries are characterized by a Fe-N average distance that varies from 2.04Å (LS) to 2.21Å (HS) along with a 0.31Å elongation of the Fe-Fe distance (12.45 to 12.76Å). All these values are compatible with a typical SCO behavior and the parameters used in the model previously presented (see Table 1).
In order to evaluate the charge redistribution, the network must be placed on top of the Au(111) surface. As the periodicities of both patterns now differ, we used a single unit consisting of one Fe center and its coordination sphere, i.e. three T4PT and one The dipole induced by the adsorption of molecules on the gold surface is treated with the dipole correction as implemented in SIESTA. 51 The lack of electron correlation is accounted for through the DFT+U approach following Dudarev's scheme. 52 Here, we use a value of U eff = 2.0 eV for Fe atom 3d orbitals. This value is compatible with previous inspection of SCO materials. 53 We checked that the amplitude of this energy difference is only slightly modified (less than 0.1 eV) when the the U eff parameter in DFT calculations is varied from 1.0 eV up to 4.0 eV. Core electrons are described with Troullier-Martins pseudopotentials. 54 The valence wavefunction is developed over a double-ζ polarized basis set of finite-range numerical pseudoatomic orbitals. 55 We rationalize the experimental structure observed by STM images, 30