Achieving large and nonvolatile tunable magnetoresistance in organic spin valves using electronic phase separated manganites

Tailoring molecular spinterface between novel magnetic materials and organic semiconductors offers promise to achieve high spin injection efficiency. Yet it has been challenging to achieve simultaneously a high and nonvolatile control of magnetoresistance effect in organic spintronic devices. To date, the largest magnetoresistance (~300% at T = 10 K) has been reached in tris-(8-hydroxyquinoline) aluminum (Alq3)-based organic spin valves (OSVs) using La0.67Sr0.33MnO3 as a magnetic electrode. Here we demonstrate that one type of perovskite manganites, i.e., a (La2/3Pr1/3)5/8Ca3/8MnO3 thin film with pronounced electronic phase separation (EPS), can be used in Alq3-based OSVs to achieve a large magnetoresistance (MR) up to 440% at T = 10 K and a typical electrical Hanle effect as the Hallmark of the spin injection. The contactless magnetic field-controlled EPS enables us to achieve a nonvolatile tunable MR response persisting up to 120 K. Our study suggests a new route to design high performance multifunctional OSV devices using electronic phase separated manganites.

loop of the LPCMO/Al2O3/Co device under the pre-set magnetic field of 1T. MR measurements were taken at T = 10K at the voltage bias V = 0.01V. The voltage bias is chosen to be close to zero in order to achieve the maximum TMR response. under Bpre = 1T, 3T, 7T, respectively. All the MFM images were taken at a constant magnetic field (Bext=1T) after withdrawing Bpre. Magnetic force microscope (MFM) images have been normalized to directly visualize the microscopic magnetic states of the LPCMO thin film. The red color region of the MFM image represents the FMM state, while the blue and white color refers to the COI state and boundary between FMM and COI phase, respectively. The calculated area fraction of the FMM phase has been shown at the corner of each image. The MFM measurements were carried out in the PPMS System using Attocube scanning probe microscope following the same procedure for MR measurements (see Methods).
Supplementary Figure 3. Exchange bias response in the LPCMO thin film as a function of AF-COI area at different temperatures. AF-COI is controlled by applying different pre-set magnetic field (Bpre) before taking the SQUID measurement. The area fraction of the AF-COI phase is calculated from the MFM images in Figure.4 (i.e., AF-COI: fraction of blue area).        To our best knowledge, there is no literature report on the spin polarization of the phaseseparated LPCMO system. We can estimate the spin polarization of LPCMO from the TMR value of the LPCMO/Al2O3/Co devices (as shown in Supplementary Fig. 1) using the Julliè re model 9 , = 2 LPCMO Co The largest TMR that we measured is 93% at 10 K. Using PCo ~ 34% 10 , we obtain PLPCMO ~ 93% at 10 K.

Supplementary Note 2. Connection to the Julliè re model
Here we show that our modified two-current model reduces to the Julliè re model under further approximations to the tunnel junction, and that these additional approximations are not appropriate for organic spin valves with spinterfaces.
The derivation of the Julliè re model starts from assuming that the junction resistance dominates the overall device resistance and neglecting the resistance of the ferromagnetic electrodes. It also neglects the spin-flip scattering. Thus one must set = ∞, and ↑ = ↓ = ↑ = ↓ = 3 = 0. The next approximation is to set the thickness of Alq3 to zero, thus merging the two spinterfaces into one. The resistances for the parallel configuration are, The resistances for the antiparallel configuration are, The last approximation needed to derive the Julliè re model is that the spin-dependent junction resistance depends only on the spin-dependent density of states at the Fermi energy of the electrodes. Using a spin-independent constant prefactor , It is clear from this derivation that the Jullière model is a special case of Fert's twocurrent model with the additional approximations that neglect all of the ferromagnetic electrode resistances and the spinterface effects, in particular the spin filtering effect.
One can force the spin filtering effect into the Jullière formula by assuming effective spin polarization parameters that are now dependent on the spacer layer. Following the practice of previous works 1, 9 , we use Alq3/Co * and LPCMO/Alq3 * to represent the effective interfacial spin polarization arising from the Alq3/Co and LPCMO/Alq3 spinterfaces, respectively 11,12 . We apply the Jullière formula in the limit of infinite spin diffusion through the organic medium (i.e., ≪ sf , where d and sf are the thickness and spin diffusion length of the Alq3 spacer layer, respectively). Considering the fact that Alq3/Co * is a field independent constant, the change of LPCMO/Alq3 * will be solely responsible for the tunable MR response in the LPCMO-OSV device. In Supplementary Fig. 13, we plot LPCMO/Alq3 * / LPCMO/Alq3 max * as a function of / . One expects that all data to fall on a single curve. Instead, we see large differences between different temperatures, with the ratio dropping more rapidly as the temperature is increased. It suggests that the effective polarization parameter does not capture all the key physics related to the observed tunable MR.

Supplementary Note 3. Two-current model with spin-flip scattering
In the LPCMO/Alq3/Co device, we need to consider spin-flip scattering in LPCMO due to spin-orbit coupling. This is strongest at the phase boundary (and the interface) because of broken symmetry. Therefore, the spin-flip scattering rate is approximately proportional to the total area of the phase boundary. To account for the spin-flip scattering, we modify Fert's two-current model by adding an effective resistance due to spin-flip scattering, as shown in the figure.
should be inversely proportional to the area of the phase boundary. This modified two-current model is solved as the following.
Now we apply the modified two-current model to the OSV for each spin configuration. For parallel moment alignment, the resistances in the up-spin channel are 1 = LPCMO↑ (22) 2 = Co↑ + Alq3 + spinterL↑ + spinterR↑ (23) where Alq3 is the spin-independent part of the junction resistance and spinterL↑ and spinterR↑ are the left (LPCMO spinterface) and right (Co spinterface) spin-dependent resistance of the spinterfaces. The resistances in the down spin channel are, 3

= LPCMO↓
(24) 4 = Co↓ + Alq3 + spinterL↓ + spinterR↓ = ∞ (25) Here we assume that the parallel down-spin channel resistance is very large compared to all other resistance. Assuming a finite parallel down-spin resistance will not be able to reproduce the large experimental increase in the MR with a preset magnetic field is applied. For antiparallel moment alignment, the resistances are Because ↓ > 0, we must have, AP < ( R + 1) P (34) and (1 + L ) P < AP (35) Over all preset magnetic fields, At the smallest preset field, ↓ ≈ 0, so Once 's are determined, we can use them to find all other resistances under any field.
For LPCMO/Al2O3/Co devices, there are no spinterfaces. In this case, the resistances for the parallel configuration are, 1 = LPCMO↑ (42) 2 = Co↑ + Al2O3 (43) where Al2O3 is the spin-independent part of the junction resistance. 3 The extra term on the righthand side shows that in the absence of the spin filter effect, electrode resistance tends to reduce the MR. On the other hand, although we assume that all of ↓ , Co↓ , and Al2O3 scale linearly with the effective conducting cross section of the junction when the preset field is applied, ↓ should change slightly faster than the other resistances since it is at the source of the change. This in turn causes a slight reduction of MR for the Al2O3 junction with increasing Bpre, as observed experimentally.