Exchange-induced spin polarization in a single magnetic molecule junction

Many spintronic devices rely on the presence of spin-polarized currents at zero magnetic field. This is often obtained by spin exchange-bias, where an element with long-range magnetic order creates magnetized states and displaces the hysteresis loop. Here we demonstrate that exchange-split spin states are observable and usable in the smallest conceivable unit: a single magnetic molecule. We use a redox-active porphyrin as a transport channel, coordinating a dysprosium-based single-molecule-magnet inside a graphene nano-gap. Single-molecule transport in magnetic field reveals the existence of exchange-split channels with different spin-polarizations that depend strongly on the field orientation, and comparison with the diamagnetic isostructural compound and milikelvin torque magnetometry unravels the role of the single-molecule anisotropy and the molecular orientation. These results open a path to using spin-exchange in molecular electronics, and offer a method to quantify the internal spin structure of single molecules in multiple oxidation states.


General Synthetic Details
Reagents were purchased from commercial sources. Solvents were used as supplied (analytical/HPLC-grade from Fisher or Sigma-Aldrich) or if dry solvents were required, taken from a solvent drying system (MBraun MB-SPS-5-Bench Top) under nitrogen. Petrol ether (PE) over a boiling point range of 40-60 °C was used. Eluent mixtures are reported in volume: volume. Column chromatography was carried out using Merck Geduran silica gel 60 under N2 pressure. TLC was carried out on Merck silica gel 60 F254 Al plates. MALDI-TOF-MS was carried out in positive reflectron mode using a Bruker MALDI microflex instrument with dithranol as a matrix. NMR spectroscopy measurements were recorded using a Bruker AVII400 instrument. All peaks were referenced to the residual solvent peak. UV-vis spectra were recorded using a PerkinElmer Lambda 25 instrument. Size exclusion chromatography (SEC) was carried out using Bio-Beads S-X1, 200-400 mesh (Bio Rad). Recycling gel permeation chromatography (GPC) was carried out on Shimadzu system equipped with a set of JAIGEL 3H (20 × 600 mm) and JAIGEL 4H (20 × 600 mm) columns with a flow rate of 3.5 mL/min. Analytical GPC was performed on a JAIGEL H-P pre-column, a JAIGEL 3H-A (8 mm × 500 mm) and a JAIGEL 4H-A column (8 mm × 500 mm) in series with tetrahydrofuran/1% pyridine as eluent.

Synthesis of H2YP:
The free-base porphyrin FBP (95 mg, 60 µmol), YCl3.6H2O (116 mg, 0.60 mmol), benzimidazole (1.0 g), imidazole (2.0 g) and diphenyl ether (2.0 g) were added to a dry Schlenk tube. The reaction mixture was heated to 200 °C overnight while stirring under argon. The mixture was allowed to cool and CHCl3 (30 mL) was added. The solution was washed with water (3 × 30 mL), filtered over MgSO4 and dried. The diphenyl ether was removed under high vacuum using a Hickmann distillation head. The crude product was stirred in CHCl3 (15 mL) with NaLOEt (300 mg, 0.30 mmol) for 1 hour. A silica column was used to purify the product, with a solvent gradient (100:0 to 6:1 PE:ethylacetate). The solvent was removed under reduced pressure to yield H2YP as a pink solid (22 mg, 17 %).

Synthesis of Br2YP:
A solution of freshly recrystallized Nbromosuccinimide (NBS, 3.5 mg, 20 µmol) in dichloromethane (1.5 mL) was added dropwise to a solution of the porphyrin H2YP (20 mg, 9.0 µmol) in dichloromethane (3 mL) with stirring. The reaction was monitored by TLC, and after completion the mixture was passed over a silica plug, eluting with PE:CH2Cl2, 3:1. The solvent was removed to yield the product as a purple solid (21 mg, 100%).

Synthesis of H2DyP:
The free-base porphyrin FBP (85 mg, 53 µmol), DyCl3.6H2O (189 mg, 0.50 mmol), benzimidazole (1.0 g), imidazole (1.2 g) and diphenyl ether (1.8 g) were added to a dry Schlenk tube. The reaction mixture was heated to 210 °C overnight while stirring under argon. The mixture was allowed to cool and CHCl3 (30 mL) was added. The solution was washed with water (3 × 50 mL), filtered over MgSO4 and dried. The diphenyl ether was removed under high vacuum using a Hickmann distillation head. The crude product was stirred in CHCl3 (15 mL) with NaLOEt (65 mg, 65 µmol) for 1 hour. A silica column was used to purify the product, with a solvent gradient (100:0 to 6:1 PE:ethylacetate). The solvent was removed under reduced pressure to yield H2DyP as a pink solid (85 mg, 70 %).

Synthesis of DyP:
The bromoporphyrin Br2DyP (20 mg, 8.3 µmol) was placed in a dry Schlenk tube along with Pd(PPh3)4 (2.0 mg, 1.7 µmol), CuI (0.5 mg, 2.5 µmol) and 1,3,6-tris(dodecyloxy)-8-ethynylpyrene (19.2 mg, 25 µmol) under argon. Dry THF (0.8 mL) and dry di-iso-propylamine (0.8 mL) were added and the mixture was immediately frozen, and subjected to three freeze-pump-thaw cycles. After 2 hours at 50 °C the reaction mixture was passed over a short plug in PE:CH2Cl2:PE (3:1). The mixture was passed over a silica column using an eluent gradient from 100:1 to 6:1 PE:ethylacetate. The mixture was further purified on a SEC column, eluting with toluene. Recycling GPC using toluene:pyridine 99:1 was used as a final purification step. The solvent was removed to yield the final product DyP as a brown solid

Assignment of Molecular Oxidation States
The molecular oxidation states in different Coulomb-blocked regions of a stability diagram can be assigned by measuring the magnitude of current in the sequential tunneling region that separates them. 5 The molecule-electrode coupling need to be sufficiently asymmetric, i.e., ΓS > ΓD or ΓD > ΓS. In addition, the frontier orbitals involved in transport should be only spin-degenerate. As outlined in Ref 5 , this leads to the four corners of the sequential tunneling region to have different magnitudes of current. The region where the magnitude of current is highest (given by the yellow triangle in Supplementary Fig. 28) is on the side (in terms of VG) of the transition where the frontier orbital involved in transport is in the singly occupied state. For example, if the transition is MP + /MP, the highest current will be on the MP + side (at more negative VG than the transition). If the transition is MP/MPthe highest current corner will be on the MPside (at more positive VG than the transition). We note that in the spin degeneracy in the HOMO of DyP + is lifted by exchange interactions with the Dy(III) ion, as is discussed in great detail in the main text. However as long as the current is measured at |VSD| much greater than the level splitting (0.8 meV), the assignment is valid.
The transitions from the 3 devices we study in this paper are assigned on this basis. For the DyP device presented in full in the main text, the highest-current corner is top left (Supplementary Fig.  28a), this indicates a DyP + /DyP transition and ΓD > ΓS. For the second DyP device, shown in Supplementary Fig. 26, the highest-current corner is again top left ( Supplementary Fig. 28b), so it is a DyP + /DyP transition and ΓD > ΓS. For the YP device the highest current corner is bottom left (Supplementary Fig. 28c), so it is a YP + /YP and ΓS > ΓD.
For this analysis it was assumed that the closest transition to the Fermi level of the graphene (i.e. the closest transition to VG = 0) was either DyP + /DyP or DyP/DyP-. It is possible that purely from the analysis above that the transitions in Supplementary Fig. 28 could be any odd/even transition such as DyP 3+ /DyP 2+ or DyP -/DyP 2-. However this would require the molecule to transfer at least two electrons to or from the substrate upon adsorption onto the graphene nanogap at VG=0, which is unlikely. Furthermore, as the stability diagrams in Supplementary Fig. 24, 25 and 26 show, there is a large Coulomb-blocked region across the central portion of the full stability diagram for all the devices. This large Coulomb blocked region is due to a large addition energy for the N (MP) state. If we were studying higher oxidation states we would expect the neighboring transitions to be relatively close in potential on either side, however these are not observed experimentally. Therefore we are confident in the assignment of the molecular oxidation states as given on the stability diagrams.

Transport Simulation
In the following section the case of the + / transitions that we observe experimentally are modelled within a rate-equation framework. 6 The three populations (P) that need to be considered for the + / transition are the neutral and the ground and excited state of the cation,

(Supplementary Equation 2)
Where → is the rate of electron transfer from state i to j.
where Γ is the electronic coupling to the source and drain electrodes, ( ) is the Fermi-Dirac distribution of electron energies in the leads, → ( ) and is the energy-dependent molecular densities of states. In the absence of electron-vibrational coupling the form of the molecular densities of states is simply a Lorentzian centred on the potential of the → transition ( ) and lifetime broadened by the molecule-electrode coupling:

(Supplementary Equation 4)
The transition potentials, → depend on the source-drain and gate voltage, scaled by the coupling to the electrode: can be fitted with two Lorentzians that are split by the intramolecular exchange splits so they can be fitted and integrated separated. Integration over these two Lorentzian gives ↑↓ and ↑↑ . The ratio vs VG is calculated from these and displayed in Supplementary Fig. 30c, but only when the signal-to-noise ratio ( / ) is greater than 1. is calculated as the root-mean-squared (RMS) value of the current in a Coulomb-blocked region of the same dataset.
At B = 0.08 T we calculate as 0.94 ± 0.35 at VG = -4.029 V. The range over which the spin polarisation ratio is over 0.90 with an error below 0.40 is ~ 4 meV. As BY increases to 5 T the spin polarisation of the current is similar within the error, as 0.97 ± 0.28 at VG = -4.027 V, however the range over which the value is over 0.90 with an error of less than 0.40 is 8 meV, due to Zeeman splitting further separating the energies of the transitions.
The error, Δ , on is calculated as follows, where Δ ↑↓ = Δ ↑↑ = : ↑↑ + / channels are modelled in red and blue respectively, (each one is the integral over a Lorentzian for when the state enters the bias window). The sum is given in green, the shaded green region is the error on the current. The ratio of is plotted in (c) and (d) in the regions where SNR > 1 (Inoise = 0.12 pA), otherwise it is set to zero. The shaded purple area is the VG range over which the ratio > 0.9 and Δ < 0.3, and widens with increasing BY. . The inclusion of additional fitting parameters (i.e. C5 onwards) did not decrease the fitting errors.