Spin-state-dependent electrical conductivity in single-walled carbon nanotubes encapsulating spin-crossover molecules

Spin crossover (SCO) molecules are promising nanoscale magnetic switches due to their ability to modify their spin state under several stimuli. However, SCO systems face several bottlenecks when downscaling into nanoscale spintronic devices: their instability at the nanoscale, their insulating character and the lack of control when positioning nanocrystals in nanodevices. Here we show the encapsulation of robust Fe-based SCO molecules within the 1D cavities of single-walled carbon nanotubes (SWCNT). We find that the SCO mechanism endures encapsulation and positioning of individual heterostructures in nanoscale transistors. The SCO switch in the guest molecules triggers a large conductance bistability through the host SWCNT. Moreover, the SCO transition shifts to higher temperatures and displays hysteresis cycles, and thus memory effect, not present in crystalline samples. Our results demonstrate how encapsulation in SWCNTs provides the backbone for the readout and positioning of SCO molecules into nanodevices, and can also help to tune their magnetic properties at the nanoscale.

Supplementary Figure 1.The SCO1 structure and magnetism. Experimental XRD pattern of SCO1 (a, pink) and calculated pattern obtained from single crystal structure (a, grey). Crystallographic data was obtained from Real et al. 1 b) IR spectrum obtained for SCO1. c) χT vs T curve of SCO1, T1/2 = 167 K.
Supplementary Figure 2. The SCO2 structure and magnetism. Experimental XRD pattern of SCO2 (a, orange) and calculated pattern obtained from single crystal structure (a, grey). Crystallographic data was obtained from Real, et al 1 b) IR spectrum obtained for SCO2. c) χT vs T curve of SCO2, T1/2 = 160 K.

Supplementary Note 2. Detailed procedure for the encapsulation of SCO in SWCNTs
1.-SWCNT pre-treatment 100 mg SWCNTs purchased from Cheap Tubes, Inc. were previously opened by thermal oxidation in air atmosphere at 600 ºC for 45 min (ca. 40% weight was lost in the process). They were then purified by two sequential acid washes: the nanotubes were suspended in 40 mL HCl 35% (1.5 mg SWCNT/mL HCl) and sonicated for 10 min. The mixture was poured in 120 mL miliQ water and filtered through a 0.2 µm poresized polycarbonate membrane. The solid obtained was washed with water until neutral pH and then dried in an oven at 150 ºC for 90 min. The cleaning procedure was repeated twice, and the metallic residue was reduced to 4.6% (according to TGA).

2.-Procedure for the encapsulation
For the encapsulation, 90 mg SCO1 or SCO2 and 25 mg opened SWCNTs (oSWCNTs) were sealed in a quartz ampoule at 10 -6 mbar and then heated in an oven at 150 ºC for 2 and 7 days. The selected temperature is close to the experimental sublimation temperature found at 10 -2 mbar (162 ºC for SCO1 and 160 ºC for SCO2). The applied low pressures ensure the sublimation of both complexes. The complex adsorbed on the SWCNT surface was eliminated by washing with tetrachloroethane and dichloromethane, applying 3 min sonication between washes. The washes were stopped when the supernatant solution was completely colourless. The results obtained for the TGA analysis are similar for both encapsulations, as expected given the structural similarity of the complexes. For SCO1@SWCNTs, the plot obtained in air shows a first weight loss that starts at around 250 ⁰C and ends up combined with the SWCNTs loss. This loss has been adscribed as the de-encapsulation and burning of the organic ligands. This curve differs from that of pure SCO1, with has three different losses, the first one starting at 200 ⁰C. In the case of SCO2@SWCNTs, the ligands weight loss is situated at 230 ⁰C, while SCO2 decomposes almost completely at 160 ⁰C. This shifts towards higher temperatures are explained by the protective shield that constitutes the SWCNTs wall. The key feature in both cases is the remaining metal particles after burning the sample until 970 ⁰C., probably already in its oxidized state. In the nitrogen spectra there is as well a displacement between the encapsulated weight losses and pristine complexes.

Supplementary Note 14. Additional electron transport measurements and AFM images of devices
Supplementary Figure 13 shows the current-voltage (IV) characteristics measured at room temperature before and after deposition of a SCO2@SWCNT hybrid by dielectrophoresis. The current before dielectrophoresis (red curve) remains in the 10 -12 A range. This is the noise level of our electronics and indicates that the gap between the electrodes is empty. In contrast, after dielectrophoresis, the current levels jump to the microampere range with a characteristic s-shape, typical of one or a few carbon nanotubes presenting Schottky barriers at the interface with the Au electrodes. Figure 13. Trapping SWCNTs. Current-voltage IV characteristic measured in a device before (red) and after (black) dielectrophoresis of a SCO2@SWCNT mixed-dimensional hybrid. The current level before dielectrophoresis points to an empty gap, whereas the current level and s-shape characteristic after dielectrophoresis is typical of one or a few carbon nanotubes presenting Schottky barriers at the interface with the Au electrodes.

Supplementary
Supplementary Figure 14 shows the temperature dependent current measured on the two samples presented in the manuscript together with four additional samples. All samples show the same thermally induced hysteresis in the conductance. The green lines are fits to the HS and LS curves using the Arrhenius law described in the main manuscript. The fitting parameters are summarized in Supplementary Table 2. In most of the cases, the HS to LS transition is associated with a drop in the activation energy U and the preexponential factor I0.  Figure 14.
The curves present an overall similar shape and THC and TLC transition temperatures. Some of the peculiarities we observe are that the THC transition (from HS to LS) is barely visible in sample 2 whereas it cannot be observed in sample 6. This is due to the relative value of THC with respect to activation energy and exponential pre-factor from the Arrhenius fit, as explain in the main text. The closer the THC to the crossing between HS and LS conductance curves the more difficult to observe the HS to LS transition. It is also interesting to observe how in sample 3 there appear to be three distinct jumps from HC (red curve) to LC (blue curve). This probably indicates that different segments or clusters of the encapsulated SCO are switching from LS to HS at slightly different temperatures. Finally, sample 4 seems shifted to slightly lower temperatures close to bulk values.
Supplementary Figure 14. Additional samples. Current I measured at a fixed bias V = 1 V as a function of the temperature in different SCO2@SWCNT (samples 1, 3 and 5) and SCO1@SWCNT (sample 2, 4 and 6) devices. The samples are initially cooled down to 90 K (blue curve) and subsequently heated up to room temperature (red curve). The conductance switches between a lowconductance LC and a high-conductance HC state describing a thermal hysteresis. The solid green lines are fits to the HC and LC states with an Arrhenius model for thermally activated transport.
It is interesting to note that although the resulting thermal cycles are roughly similar in all samples, there are a few differences between SCO1@SWCNT and SCO2@SWCNT hybrids. SCO2 gives rise to more elongated and flat hysteresis with well-defined THC and TLC transitions. In contrast, SCO1 shows more asymmetric hysteresis loops with large and sharp TLC jumps and barely visible THC transitions. The number of measurements does not allow thought to extract any solid conclusion in this respect. This will be studied in subsequent works.
We have also checked the stability and reproducibility of the hysteresis loop in sample 6. Supplementary Figure 15 shows the temperature dependence of the current measured in three consecutive thermal cycles.
The key parameters of the hysteresis in the current are roughly stable and reproducible.
Supplementary Figure 17. AFM image of a SWCNT device. AFM image and selected profiles of a nano-device containing a SWCNT.

Supplementary Note 15. Computational details
The adsorption of a single molecule on the single-walled carbon nanotube (16,8) has been studied within periodic density functional theory (DFT) using the VASP (Vienna ab initio simulation package) code 3, 4, 5, 6 employing the generalized gradient approximation (GGA) with the revised Perdew-Burke-Ernzerhof functional (rPBE) 7 and projector-augmented wave (PAW) potentials. 8,9 rPBE functional has been proven to provide a good LS-HS balance (much better than other GGA functionals such as PBE) for well-known SCO complexes containing Fe(II) and Fe(III). 10 Valence electrons are described using a planewave basis set with a cutoff of 500 eV and the Γ-point of the Brillouin zone is used. 11 The optimized lattice parameters for the nanotube are a=b=16.75 Å and c=11.35 Å. Hence, the diameter of the nanotube once optimized is of 1.675 nm.
To study the encapsulation, we have used a 1x1x2 supercell with 37Å of vacuum between nanotubes (54 Å x 54 Å x 22.7 Å) containing 448 carbon atoms, and a single Fe SCO complex. Electronic relaxation has been performed until the change in the total energy between two consecutive steps is smaller than 10 -6 eV and the ionic relaxation has been performed until the Hellmann-Feynman forces were lower than 0.025 eV/Å. As we are interested in the different magnetic solutions, the NUPDOWN option is used, which forces the difference between number of electrons in up and down spin channels, Nα-Nβ, to be equal to 0 (LS) or 4 (HS). Several starting geometries for geometry optimizations have been tested. Interactions energies between the encapsulated SCO2 complex and the nanotube, Eint, were calculated as Eint=Emolecule@SWCNT -(ESWCNT+Emolecule). A negative Eint value means that the encapsulated molecule is more stable than the free molecule.
Finally, the density of states (DOS) has been obtained from a single-point calculation on the optimized geo1 structure for each spin state with 1x1x9 k points.
The impact of an external electric field on the relative stability of the HS and LS spin states are analyzed through single-point density functional theory (DFT)-based calculations on the isolated molecule, using Gaussian 09 code. The lowest-energy geometry adopted by the molecule for each state, with the same orientation as in geo1, is employed in all these calculations. The TPSSh functional 12,13 is used with the def2SVP basis set 14 for all the atoms in the molecule. This functional has been reported to exhibit high accuracy for first-row transition metal systems, including iron SCO complexes. 10,15 The total energy of the molecule on the LS and HS states is evaluated under the effect of an electric field applied along the x, y or z directions. The strength of the electric field is expressed in atomic units, where 1 arb. unit = 5.14 10 9 V/cm.