SUPPLEMENTARY INFORMATION

Transcription

1 SUPPLEMENTARY INFORMATION doi: /nmat797 Spin injection/detection via an organic-based magnetic semiconductor Jung-Woo Yoo 1,, Chia-Yi Chen 3, H. W. Jang 4, C. W. Bark 4, V. N. Prigodin 1, C. B. Eom 4, A. J. Epstein 1, 1 Department of Physics, The Ohio State University, Columbus, OH , USA Department of Chemistry, The Ohio State University, Columbus, OH , USA 3 Chemical Physics Program, The Ohio State University, Columbus, OH, USA 4 Department of Materials Science and Engineering, University of Wisconsin-Madison, Madison, Wisconsin , USA *To whom correspondence should be addressed. epstein@mps.ohio-state.edu Materials and Methods The hybrid magnetic tunnel junction were fabricated to achieve electrical detection of spin polarization of the organic-based magnetic semiconductor, V(TCNE: tetracyanoethylene) x. We employed La /3 Sr 1/3 MnO 3 (LSMO) film as a counter magnetic layer with thin organic small molecule film as a tunnel barrier. The epitaxial LSMO (80 nm) thin films followed by three unit cells of LaAlO 3 (LAO) (1. nm) layers were grown on lattice matching pseudo-cubic (001) NdGaO 3 (NGO) single crystal substrates by pulse laser deposition with in situ high pressure RHEED. The device areas were defined by the deposition of SiO (500 nm) for the 1 mm 1 mm junctions. Fig. S1 shows actual device structure on NGO substrate. The SiO enclosing layer was deposited by e-beam with a shadow mask. Sublimed grade rubrene (C 4 H 8, from Aldrich) was deposited at a source temperature of 105 C with a constant deposition rate of ~.5 nm/min under the base pressure of 10-7 torr. The typical surface morphology of 5 nm rubrene layer grown on the nature materials 1

2 supplementary information doi: /nmat797 (001) surfaces of LAO displays rms roughness ~ 1 nm determined by atomic force microscopy (AFM) as shown in Fig. S. After deposition of the rubrene layers, the samples were transferred to a CVD chamber through an Ar filled glove box (O < 0.3 ppm; H O < 1.0 ppm). The solvent-free V(TCNE) x films were grown via chemical vapour deposition (CVD) with deposition rate ~ 5.5 nm/min in the Ar-filled glove box following the literature S1. The total thickness of V(TCNE) x films are ~ 500 nm. The samples were transferred through the Ar filled glove box to a vacuum chamber for the deposition of the top Al contact. 50 nm of Al layers were deposited via e-beam. Two terminal contacts were used for the magnetotransport measurements. Contacts were prepared in the glove box. Though the solvent-free CVD grown samples are still sensitive to the oxygen, the devices capped with Al electrodes show nearly stable resistance in the ambient air for several hours. After a brief exposure to the air (~ 1 min.), devices were transferred to Quantum Design physical property measurement system (PPMS), where all magnetotransport studies were performed with an external Keithley 400 sourcemeter. Magnetizations of the individual magnetic layers were recorded with a superconducting quantum interference device (SQUID) magnetometer. Rubrene/V(TCNE) x (5 nm/500 nm) Al electrode LSMO/LAO (80 nm/3 u.c.) NGO substrate SiO (500 nm) I nature MATERIALS

3 doi: /nmat797 supplementary information Figure S1: Geometry of hybrid magnetic tunnel junction devices. Figure S: AFM image (scan size = 500 nm 500 nm) and roughness analysis (rms roughness = 0.6 nm; peak to peak distance = 4.3 nm) of 5 nm rubrene layer deposited on (001) LAO surface. nature materials 3

4 supplementary information Variation of resistance and magnetoresistance from device to device doi: /nmat797 4 Magnetic tunnel junctions (MTJ) were fabricated for three different type of barriers, single rubrene Figure S: AFM image (scan size = 500 nm 500 nm) and roughness analysis (rms roughness = (5 nm), single LAO (3 u. c.), and hybrid (rubrene (5 nm)/lao (3 u. c.)) barriers. Several devices for 0.6 nm; peak to peak distance = 4.3 nm) of 5 nm rubrene layer deposited on (001) LAO surface. each single LAO and hybrid barrier were successfully fabricated and the resistance of the devices varies from device to device, which is typical behaviour for tunnel junctions. Fig. S3a shows the variation of device resistance and MR measured at 100 K and for different tunnel Variation barriers. The of resistance device resistance and magnetoresistance varies nearly one from order device of magnitude to device for each type of barrier. The Magnetic devices with tunnel hybrid junctions barrier (MTJ) typically were fabricated show higher for resistance three different and type higher of MR. barriers, The single noise rubrene level of (5 device nm), single resistance LAO also (3 u. varies. c.), and Variations hybrid (rubrene of magnetoresistance (5 nm)/lao (3 u. curves c.)) barriers. from device Several to devices device for are each displayed single in LAO Fig. and S3c-f hybrid for hybrid barrier barriers, were successfully Fig. S3g-j for fabricated single LAO and barriers, the resistance and Fig. of S3k the devices and l for varies rubrene from barriers. device We to device, were not which successful is typical to behaviour fabricate for MTJs tunnel with junctions. single rubrene Fig. S3a barriers. shows the The variation absence of MR device is likely resistance due to and the MR presence measured of small at 100 area K contacts and Vbecause b = 0.5 of V rough for different morphology tunnel of barriers. our rubrene The device layers. resistance The other varies reason nearly may be one the order dipole of magnitude barriers at the for each metal/osc type of contact, barrier. which The devices interfere with with hybrid spin conserved barrier typically tunnelling show as higher previously resistance suggested and higher S, S3. Typical MR. The magnetoresistance noise level of device curves resistance for single also rubrene varies. barriers Variations show bumps of magnetoresistance (Fig. 3Sk and l) curves around from zero device field, which to device extend are to displayed the coercive in Fig. field S3c-f of LSMO for hybrid films, barriers, instead Fig. of S3g-j separated for single MR peaks LAO for barriers, positive and and Fig. negative S3k and scans. l for rubrene We believe barriers. that We this were originates not successful from magnetic to fabricate coupling MTJs between with single two magnetic rubrene barriers. layers due The to absence imperfect of MR coverage is likely of due thin to rubrene the presence layers. of Employing small area double contacts barrier because (LAO of rough and morphology rubrene) should of our help rubrene to minimize layers. the The effect other of reason pinholes may because the the dipole pinholes barriers in both at the layers metal/osc do not contact, align easily which and interfere with spin conserved tunnelling as previously suggested S, S3. Typical magnetoresistance curves for single rubrene barriers show bumps (Fig. 3Sk and l) around zero field, which extend to the coercive field of LSMO films, instead of separated MR peaks for positive and negative scans. We believe that this originates from magnetic coupling between two magnetic layers due to imperfect coverage of thin rubrene layers. Employing double barrier (LAO and rubrene) should help to minimize the effect of pinholes because the pinholes in both layers do not align easily and 4 nature MATERIALS

5 doi: /nmat797 supplementary information the rubrene molecules typically tend to sit on defect/pinholes covering defects/pinholes of insulating LAO barrier during the deposition. The measurements of MR for devices with low noise are highly reproducible for the repeated scans. Fig. S3b displays 4 different magnetoresistance curves for a single device with hybrid barrier, which were collected over two days confirming reproducibility of our measurements. Slight shifts of device resistance occurred after the temperature scans were performed. We also noticed that the noise level of device resistance decreased when the constant bias was applied for a long time. nature materials 5

6 supplementary information doi: /nmat797 a c d Device A-1 Rubrene (5 nm) 3.80 LAO/rubrene LAO (3 u.c.) 3.78 LAO(3 u.c.)/rubrene (5 nm) V b =0.5 V MR (%) Resistance (Ohm) b 3.84 Device A-1 Scan1, day1 3.8 LAO/rubrene Scan, day1 Scan3, day Scan4, day i 3.80 V b = 500 mv e g k Device B LAO barrier Device B LAO barrier DeviceC-1 Rubrene barrier Device A LAO/rubrene Device A Device A-4 LAO/rubrene LAO/rubrene f h j l Device B LAO barrier Device B-4 LAO barrier Device C- Rubrene barrier Figure S3: a, Variation of device resistance and MR from device to device. MR is determined from resistance at around peaks (anti-parallel) and resistance at H = 500 Oe (parallel). The error bars represent standard deviations based on data in each magnetoresistance curve. b, Reproducibility of magnetoresistance curves for the device with hybrid barrier collected over days. c-f, Magnetoresistance curves for devices with hybrid (LAO (3 u.c.)/rubrene (5 nm)) barrier. g-j, Magnetoresistance curves for devices with single LAO (3 u.c.) barrier. k and l, Magnetoresistance curves for rubrene (5 nm) barrier. 6 nature MATERIALS

7 doi: /nmat797 supplementary information Device A Device A-1 Device B-1 LAO barrier 110 K 0.35 Device B-1 LAO barrier 0.6 V.5.0 V b = 0.5 V V 0.40 V b = 0.5 V 110 K V 10 K V K K V Figure S4: a, Temperature and bias dependence of real scale magnetoresistance curves for device A-1 and B-1, which were displayed in the manuscript. nature materials 7

8 supplementary information doi: /nmat797 Temperature dependence of device resistance and bulk resistance of V(TCNE) x Fig. S4 compares the temperature dependence of normalized device resistance with hybrid barrier and bulk resistance of V(TCNE) x film. The bulk resistance of V(TCNE) x film was measured by two terminal 1 mm width gold electrodes separated by 00 µm. In this lateral configuration with large separation of electrodes, the contact resistance of Au/V(TCNE) x can be ignored as confirmed in -terminal and 4-terminal comparison S4. The resistances were collected with high impedance Keithley 617 electrometer and typically were measurable down to around 100 K. The obtained temperature dependent resistance for hybrid magnetic tunnel junction device and bulk resistance of V(TCNE) x is normalized as Ohm cm and displayed in the same plot in Fig. S5. The typical room temperature resistivity of V(TCNE) x film is ~ 10 Ohm cm. As seen in Fig. S5, the interface resistance dominates at high T. The red line indicates fitting with injection model of thermionic field emission. The phonon-assisted tunneling model developed by Kiveris and Pipinys S5,S6 were employed to fit T dependent device resistance, similar to the approach performed for the organic semiconductor spaced spin valve devices S7. The phonon assisted tunneling rate of electron under the electric field at the metal/semiconductor interface is as follow S5,S6 W T ee = (8m ε ) 1/ T [(1 + γ ) 1/ 4 (m ) exp 3 eeh 1/ ε 3 / T γ ] 1/ [(1 + γ ) [1 + γ ] 1/ 1/ 4 γ ] [(1 + γ ) 1/ 1 + γ ], where 1/ (m ) Γ γ = 8ehEε 1/ T Here ε T is the energetic depth of defect states from the conduction band of semiconductor, E = V b /d is the applied electric field, Γ = 8a( hω) (n + 1) is the width of the defect states broadened by the interaction with optical phonons, n = 1/[exp( h ω / k T ) 1], and a is the electron-phonon coupling B 8 nature MATERIALS

9 doi: /nmat797 supplementary information constant ( a = Γ 0 / 8( hω) ). Qualitatively good fitting can be obtained at high T region with parameters of a = 1.6, h ω = 0.03 ev, m* = 1. m e and ε T = 0.37 ev. Larger deviation between injection model and device resistance can be found as T decreases below 100 K. Below 100 K, the bulk resistance of V(TCNE) x film starts to prevail the total device resistance. The crossover between injection-limited and V(TCNE) x bulk-limited should be depend on the applied bias as seen in Fig. d. It should be also mentioned that the bulk resistance of V(TCNE) x film as well as the total resistance of hybrid magnetic tunnel junctions slightly varies from sample to samples. However, the crossover between injection limited to bulk limited device current should typically occur below 100 K, where the interface resistance starts to be weakly dependent on the temperature. This crossover indicates that the sudden decrease of MR below 100 K at 0.5 V (Fig. 3d) could originate from the bulk resistance of V(TCNE) x, which continues to increase exponentially as T is lowered. This interpretation is in consistent with the substantial drop of giant magnetoresistance with increasing thickness of rubrene space layers S7. Normalized resistance (Ohm cm) device at 10 5 V(TCNE) x bulk 10 4 Injection model 10 3 with barrier height E a = 0.37 ev /T (K -1 ) Figure S5: Temperature dependent normalized resistance for hybrid barrier device and V(TCNE) x bulk. nature materials 9

10 supplementary information doi: /nmat797 LSMO Thickness Dependence of the Magnetoresistance Curves The shapes of magnetoresistance curves in our hybrid tunnel junctions are strongly dependent on the thickness of LSMO films, as the magnetic domain structures of the LSMO films depend on the film thickness due to competition of domain wall energy and strain-induced uniaxial anisotropy. Fig. S6a shows the magnetization curves for three different thickness (60, 80, 100 nm) of LSMO films grown in pseudo-cubic (001) NGO substrates. In particular, the 80 nm thick LSMO film displays the sharpest magnetization steps as a function of magnetic field. Fig. S6b displays magnetoresistance curves of hybrid junction (LSMO(60 nm)/lao(1. nm)/rubrene(5 nm)/v(tcne) x (500 nm)) recorded at 100 K and a applied bias of 100 mv. The shape of magnetization is well reflected in the shape of magnetoresistance curves of our hybrid magnetic tunnel junction as our devices are large scale multi-domain devices. We did not observe significant change of the MR magnitude by varying LSMO thickness in our hybrid magnetic tunnel junction. Reflection of magnetization shape on the magnetoresistance curves undoubtedly shows that the magnetoresistance in our devices originates from spin dependent tunneling between the conventional magnetic film (LSMO) and the organic-based magnet V(TCNE) x. a 1.0 b 4 V b = 100 mv V(TCNE) x 1 M (arb. units) nm LSMO 80 nm LSMO 100 nm LSMO MR (%) LSMO(60 nm) 0-1 M (arb. units) Figure S6: a, Hysteresis curves of LSMO films on NGO substrates for the thickness of 60, 80, 100 nm. b, Magnetoresistance curves of hybrid magnetic tunnel junctions (V(TCNE) x /rubrene/lao/lsmo) for the 60 nm thick LSMO film. 10 nature MATERIALS

11 doi: /nmat797 supplementary information Comparison of temperature dependence of MR in different hybrid spin valves exploiting LSMO electrodes The spin polarizations of ferromagnetic layers in magnetic tunnel junctions often are estimated by using Jullière s formula S8 (MR = P 1 P /(1 P 1 P ), P 1 and P are the spin polarizations of the two ferromagnetic layers). However, the temperature dependence of surface spin polarization in LSMO, the spin-flip scattering at the interfaces, and the imperfection of the large scale barrier significantly reduce the MR in our device so that it is difficult to extract the spin polarization of V(TCNE) x film from the observed MR in our hybrid tunnel junctions. Fig. S7 compares the temperature dependence of MR in our hybrid device with that of previously reported MR for hybrid spin valves using LSMO electrodes. The black line and symbols are the temperature dependence of MR reported by Hueso et al. S9 in LSMO/carbon nanotube/lsmo spin valve measured at 5 mv. The red line and symbols are for the reported MR by Xiong et al. S10 in LSMO/Alq 3 /Co heterojunction measured at.5 mv. The green line and symbols are for the reported MR by Yoo et al. S7 in micron-size LSMO/LAO(1. nm)/rubrene(5 nm)/co heterojunction measured at 10 mv. The temperature dependence of MR measured at 0.5 V in our hybrid magnetic tunnel junction (V(TCNE) x (500 nm)/rubrene(5 nm)/lao(1. nm)/lsmo(80 nm)) is depicted with blue line and symbols (after Fig. 3d). This comparison also implies that the substantial decrease of MR at high T in our device is induced by temperature dependent surface spin polarization of the LSMO films. All previous reports show significant decrease of MR with increasing bias. Due to high noise level at low bias in our device, we applied relatively high bias to obtain temperature dependent MR. The temperature dependent MR at high bias in our device is still comparable to the MR in previous literature using LSMO layer. nature materials 11

12 supplementary information doi: /nmat797 MR (%) LSMO/carbon nanotube/lsmo at V b = 5 mv (afte Ref. S9) LSMO/Alq 3 /Co at V b =.5 mv (after Ref. S10) LSMO/LAO/rubrene/Fe, at V b = 10 mv (after Ref. S7) LSMO/LAO/rubrene/V(TCNE) x at V b = 0.5 V T (K) Figure S7: Temperature dependence of MR in organic-inorganic hybrid spin valves using LSMO electrodes. The data for black and red symbols are taken from the literatures as indicated in the Figure. Blue symbols indicate temperature dependent MR (at ) for the hybrid magnetic tunnel junction using organic-based magnet V(TCNE) x as a ferromagnetic layer (after Fig. 3d). References S1. Pokhodnya, K. I., Epstein, A. J. & Miller, J. S. Thin-film V[TCNE] x magnets. Adv. Mater. 1, (000). S. Santos, T. S. et al. Room-temperature tunnel magnetoresistance and spin-polarized tunneling through an organic semiconductor barrier. Phys. Rev. Lett. 98, (007). S3. Shim, J. H. et al. Large spin diffusion length in an amorphous organic semiconductor. Phys. 1 nature MATERIALS

13 doi: /nmat797 supplementary information Rev. Lett. 100, 6603 (008). S4. Yoo, J.-W. et al. Supplemental Information S in Multiple photonic responses in films of organic-based magnetic semiconductor V(TCNE) x, x ~. Phys. Rev. Lett. 97, 4705 (006). S5. Kiveris, A., Kudžmauskas, Š. & Pipinys, P. Release of electrons from traps by an electricfield with phonon participation. Phys. Status Solidi A (1976). S6. Pipinys, P. & Kiveris, A. Phonon-assisted tunnelling as a process determining current dependence on field and temperature in MEH-PPV diodes. J. Phys.: Condens. Matter (005). S7. Yoo, J.-W. et al. Giant Magnetoresistance in Ferromagnet/Organic Semiconductor /Ferromagnet Heterojunctions. Phys. Rev. B (009). S8. Julliere, M. Tunneling between ferromagnetic films. Phys. Lett. 54A, 5-6 (1975). S9. Hueso, L. E. et al. Transformation of spin information into large electrical signals using carbon nanotubes. Nature 445, (007). S10. Xiong, Z. H., Wu, D., Vardeny, Z. V. & Shi, J. Giant magnetoresistance in organic spinvalves. Nature 47, (004). nature materials 13