Mixed-Sandwich Titanium(III) Qubits on Au(111): Electron Delocalization Ruled by Molecular Packing

Organometallic sandwich complexes are versatile molecular systems that have been recently employed for single-molecule manipulation and spin sensing experiments. Among related organometallic compounds, the mixed-sandwich S = 1/2 complex (η8-cyclooctatetraene)(η5-cyclopentadienyl)titanium, here [CpTi(cot)], has attracted interest as a spin qubit because of the long coherence time. Here the structural and chemical properties of [CpTi(cot)] on Au(111) are investigated at the monolayer level by experimental and computational methods. Scanning tunneling microscopy suggests that adsorption occurs in two molecular orientations, lying and standing, with a 3:1 ratio. XPS data evidence that a fraction of the molecules undergo partial electron transfer to gold, while our computational analysis suggests that only the standing molecules experience charge delocalization toward the surface. Such a phenomenon depends on intermolecular interactions that stabilize the molecular packing in the monolayer. This orientation-dependent molecule–surface hybridization opens exciting perspectives for selective control of the molecule–substrate spin delocalization in hybrid interfaces.


-Experimental Methods
Synthesis. [CpTi(cot)] was synthetized and characterized according to the procedure reported by de Camargo et al., 1 based on the reaction of [Cp2TiCl2] and cyclooctatetraene with n-butyl lithium in tetrahydrofuran. To obtain high purity crystalline samples for surface studies, the product was sublimed at ca 130 °C/10 -3 mm Hg and then recrystallized from saturated toluene solutions at -20°C. Due to the high molecular sensitivity to the air, the crystals were isolated and handled under inert atmosphere and stored in sealed glass tubes. To perform deposition in ultra-high vacuum (UHV), the crucible for molecular sublimation was filled in a dry N2 glove box, and then transferred in static vacuum to the deposition chamber.   All the photoelectron experiments were performed at 150 K to avoid desorption of the molecular layer.

-Theoretical Methods
Periodic DFT simulations. The CP2K package was used 2-4 along with rVV10 non-local empirical dispersion corrections. 5 Norm-conserving Goedecker-Tetter-Hutter pseudopotentials 6 and a double zeta basis set with polarization functions (DZVP-MOLOPT-SR) were employed for all the atoms.
The cell parameters were kept fixed throughout the optimizations. The plane-wave cut-off value was set to 400 Ry. The wavefunction convergence (EPS_SCF) was set to 1.0x10 - Single-point DFT+U calculations were also computed on optimized structures within the Dudarev 7 implementation. The effective parameter Ueff = U -J was applied only to the titanium atom, where Ueff is the effective repulsion between electrons localized on the same site; a value of 3 eV for the S5 3d orbitals was used according to literature. 8,9 Ueff was introduced for obtaining more accurate Ti 3d energies since pure DFT functionals tend to over-delocalize the electron density.
STM images were simulated at the pDFT+U level on the optimized pDFT structure of the isolated [CpTiCot]@Au and on the monolayer array of [CpTiCot]@Au according to the Tersoff-Hamann approximation 10 as implemented in CP2K. The computed bias ranged from -2.0 V to +3 V.
Non-periodic all-electron DFT calculations on isolated molecules and clusters. These sets of calculations were performed with the ORCA 4.0 package of programs. def2-TZVP basis sets 11 were employed for Ti, C, and H, while ma-def2-TZVP with def2-ECP replacing 60 core electrons was chosen for the Au atoms. 12 The cluster models (see Figure S9) were obtained by a tailored cut of the previously optimized structures at the periodic DFT level. A fourth slab was added to reproduce better the surface Fermi states. 13         The electronic structure of the molecular film was studied experimentally by UPS and rationalized in comparison with the DOS obtained by pDFT+U on the [CpTi(cot)]@Au monolayer ( Figure S8).

-UPS analysis
A detailed description of the method employed in these calculations is reported in Ref. 17  and c) confirms that its presence is due to Ti III in the molecular layer (marked with asterisk in Figure S8b), in good agreement with the DOS of other Ti III systems. 18-20

-Simulation of the XPS results
The

S19
The Ti 3d orbitals of the lying molecules mainly contribute to the PDOS at negative biases, while those from the standing molecules appear at positive ones. The major contributions of the Ti 3d states are observed at higher energies (close to the Fermi level) than the negative biases experimentally used in STM experiments. Noticeably, while both at -2 eV and -1 eV minor 3d components are observed for the lying molecules, those related to the standing molecules are seen only at -2 eV. This agrees with the simulations presented in Figure 3, where the standing molecules show brighter spots at the bias of -2 V than at -1 V. For positive bias, only contributions from the standing molecules are observed up to +2 V. In the simulated STM images, however, rows 2 are not bright for biases below +2 V. This is because, in the standing orientation, the carbon electron density of the ligand shields the 3d orbitals from the STM tip. As a consequence, their contributions are not as relevant as those from the lying molecules, where the metal ion is directly accessible to the tip.
The C 2p PDOS support the above statements: a non-negligible spin polarization is present for all types of molecules and, despite an overall similarity with the 3d component, the standing molecules do not show any significant contribution from -1.5 to +1.5 eV, at variance from the standing Ti 3d components. Conversely, the lying molecules cover the energy window from -1.5 to the Fermi energy, while the empty states are available only above +2 eV. The availability of empty 2p states at different energies for the standing and lying molecules explains why we observe an upshift in the calculated STM profiles with respect to the experimental positive bias.
We can therefore state that, due to the different orbitals exposed to the STM tip as a function of the molecular orientation, i) both the Ti 3d and C 2p orbitals contribute to the on/off bias dependence shown by the lying molecules; ii) only the C 2p orbitals contribute to the observed bias dependence presented by the standing molecules.