Spin-optoelectronic devices based on hybrid organic-inorganic trihalide perovskites

Recently the hybrid organic-inorganic trihalide perovskites have shown remarkable performance as active layers in photovoltaic and other optoelectronic devices. However, their spin characteristic properties have not been fully studied, although due to the relatively large spin-orbit coupling these materials may show great promise for spintronic applications. Here we demonstrate spin-polarized carrier injection into methylammonium lead bromide films from metallic ferromagnetic electrodes in two spintronic-based devices: a ‘spin light emitting diode’ that results in circularly polarized electroluminescence emission; and a ‘vertical spin valve’ that shows giant magnetoresistance. In addition, we also apply a magnetic field perpendicular to the injected spins orientation for measuring the ‘Hanle effect’, from which we obtain a relatively long spin lifetime for the electrically injected carriers. Our measurements initiate the field of hybrid perovskites spin-related optoelectronic applications.

F inding materials for spintronics applications that simultaneously possess strong spin-orbit coupling (SOC) for efficient spin manipulation, long-spin relaxation time for efficient spin transport, and strong photoluminescence (PL) emission for spin-optoelectronic applications, has been a challenge. Hybrid organic-inorganic trihalide perovskites (OITP) are emerging semiconductors which have recently attracted intense research interest for optoelectronic applications 1,2 . The OITP such as MAPbX 3 , where MA is methylammonium and X is a halogen, may combine the advantages of both organic and inorganic semiconductors for spin-optoelectronic applications. These compounds are grown using solution-based low-temperature methods, which lead to easy fabrication and flexible device engineering 3 . Moreover, they have synthetically tunable electronic and optical properties with strong PL emission. Also due to the heavy atoms in their building blocks, the OITP possess a relatively large SOC. Indeed, the recently obtained Rashba-splitting in the OITP 4,5 , the optical spin selection rules 6,7 , and magnetic field effect 8 caused by strong SOC have initiated spintronics research avenue for these compounds. So far, however no spin-related optoelectronics device application based on OITP has been reported. Thus, demonstration and studies of spintronic devices based on this class of semiconductors are intriguing.
In order to realize spin-related optoelectronic devices based on OITP, electrical spin injection from ferromagnet (FM) electrodes into OITP films should be demonstrated. Here, we successfully demonstrate spin injection from FM electrode into MAPbBr 3 , detected by both optical and electrical means. For the optical detection, we have used the circularly polarized electroluminescence (EL) emission from MAPbBr 3 film in a spin light emitting diode (spin-LED) device based on a FM electrode. Spin-LED devices, first realized using III-V semiconductors 9,10 , were considered to be strong evidence for confirming spin injection into various semiconductors [11][12][13] . The confirmed spin injection ability shows that the OITP successfully overcome the conductivity mismatch problem 14 . In addition, we also demonstrate electrical spin injection into OITP using spin valve (SV) devices based on MAPbBr 3 interlayer, where giant magneto-resistance (GMR) of up to about 25% has been achieved. Importantly, we also observed the "electrical Hanle effect" in the SV devices, from which we obtained at cryogenic temperatures a spin lifetime, τ s = 936 ± 23 ps for the injected spin 1/2 holes in MAPbBr 3 .

Results
Optical spin injection and Hanle effect in MAPbBr 3 . To demonstrate spin injection from FM electrodes, we first studied circularly polarized PL emission in MAPbBr 3 thin film at low temperature, in order to show that it maintains "optical spin alignment" 15 . We prepared polycrystalline films of MAPbBr 3 by spin casting (see Methods). Scanning electron microscopy (SEM) illustrated an average grain size of around 100 nm in the film, whereas the X-ray-diffraction (XRD) pattern was similar to that in the literature (Supplementary Fig. 1 & Supplementary Note 1) 16 . The films were excited using circularly polarized (σ + ) cw laser beam at 532 nm, and we measured the circular PL components, PL(σ + ) and PL(σ − ) (Fig. 1a) to obtain the degree of circular polarization, which is defined as P PL = [PL(σ + ) − PL(σ − )]/[PL (σ + ) + PL(σ − )] at 10 K; we obtained P PL of 3.1% (see Fig. 1b). The exciton binding energy in the orthorhombic phase of MAPbBr 3 is about 30 meV 17 so that the majority of the photoexcitations at steady state conditions at 10 K are excitons. The 1% is obtained. c Schematic set-up for measuring the optical Hanle effect. A magnetic field, B z perpendicular to the PL light propagation direction is applied, which causes spin precession that leads to diminishing circular polarization of PL. d The P PL (B z ) response measured at 10 K. The red solid line is a fit using Eq. (1), from which we obtained an effective exciton spin lifetime, T s = 491 ± 17 ps photogenerated excitons may populate four closely spaced states. These are three triplets, T1-T3, where T2 and T3 emit circularly polarized PL, and a dark (D) singlet exciton 18,19 . The non-zero P PL validates the optical spin selection rules in MAPbBr 3 , which allows optical orientation and detection of spin-polarized excitations, as in other direct gap semiconductors such as GaAs 20 . The polarization degree, P PL is determined by the two spinpolarized excitons (T2 and T3) with population N +(−) at steady state, where P PL ¼ N þ ÀN À N þ þN À . P PL is related to the exciton spin lifetime, τ s and lifetime τ by the relation: is the initial polarization ratio after excitation. Considering the long exciton lifetime (of the order of tens ns) in MAPbBr 3 2,21 (Supplementary Fig. 2 & Supplementary Note 2), we conclude from the obtained P PL value that τ s cannot be very short.
When an external magnetic field, B z is applied in the direction perpendicular to the spin orientation, the exciton spin precesses around B z direction, as shown in Fig. 1c; this leads to quenching of P PL (called optical Hanle effect). Figure 1d illustrates that P PL indeed decreases with B z , which is traditionally described by the relation 22 : where ω L = μ B g ex B z /ℏ is the Larmor frequency. We have measured the g-factor value, g ex of exciton in MAPbBr 3 by magnetic circular dichroism (Supplementary Fig. 3 τ s is the exciton effective spin lifetime. The red solid line in Fig. 1d is a fit using Eq. (1), from which we obtained T s = 491 ± 17 ps in MAPbBr 3 at 10 K. Taking τ = 15 ns 4 , we calculate from the measured T s the value τ s = 508 ± 17 ps for the exciton spin lifetime in MAPbBr 3 .
Spin-LED device based on MAPbBr 3 . Since MAPbBr 3 shows circularly polarized PL emission, we may use it to probe spin injection from FM electrode by studying circularly polarized EL emission in a spin-LED device. We have therefore fabricated MAPbBr 3 spin-LED based on the half metal FM electrode La 0.63 Sr 0.37 MnO 3 (LSMO), which has nominally 100% spin polarization at the Fermi energy 23 . A thin MAPbBr 3 film was spin coated on the LSMO substrate, which serves as an anode that injects spin-polarized holes. As a cathode we used an Al film coated on top of a thin layer of the organic small molecule, 2,2′,2″-(1,3,5-benzinetriyl)-tris (1-phenyl-1-H-benzimidazole) (TPBi), which serves as the electron transport layer. The device structure is illustrated in Fig. 2a. An external magnetic field was applied perpendicular to the electrodes, which is aligned along the propagation direction of the collected EL emission. Figure 2b shows the working principle of the spin-LED device; spinpolarized holes are injected from the ferromagnetic LSMO, while unpolarized electrons are injected from the non-magnetic Al electrode. Due to the relatively large exciton binding energy of 30 meV at 10 K, the injected e-h pairs form excitons according to the optical selection rules, then emit EL light (see Fig. 2b). Because of the imbalance in the spin populations of the injected holes, EL emission shows circular polarization.
Typical electrical characteristics response of the fabricated MAPbBr 3 spin-LED device is shown in Fig. 2c. A clear rectifying behavior starting from a bias voltage, V = 5 Volt is seen. The EL-V response shows a turn-on voltage at 8 Volt; bright green light can be seen from an area of 0.5 mm × 3 mm (Fig. 2d inset). At V = 9 Volt and 10 K, we measured a strong EL emission band at  (Fig. 2d), which is consistent with the PL spectrum (Fig. 1b), and with traditional LEDs based on MAPbBr 3 2 .
The two circularly polarized EL components, EL(σ + ) and EL (σ − ) were measured when the out-of-plane magnetic field aligns the LSMO magnetization, as seen in Fig. 3a. A clear difference between EL(σ + ) and EL(σ − ) responses versus B is obtained. We note that the EL(B) response is a superposition of two components. (i) An intrinsic, symmetric magneto-EL (MEL) response, which is the same for the EL(σ + ) and EL(σ − ) responses (Fig. 3b). This response is due to a spin-mixing process within the e-h pairs 8 , and is the same as in an ordinary MAPbBr 3 -based LED (with no FM electrodes) that we also fabricated (see Fig. 3b). (ii) An asymmetric MEL(B) response which is different for the EL (σ + ) and EL(σ − ) emissions. We calculated the EL degree of circular polarization, P EL using the relation P EL ¼ ELðσ þ ÞÀELðσ À Þ ELðσ þ ÞþELðσ À Þ ; its response versus magnetic field B is plotted in Fig. 3c. A maximum P EL value of 0.8% is obtained at B larger than 150 mT. For comparison purpose, we also plot in Fig. 3c the out-of-plane magnetization M(B) response of LSMO electrode. It is clear that the obtained P EL (B) response follows the magnetization response of LSMO. This unambiguously proves that the circular polarization of the EL is caused by spin injection from the LSMO electrode, since P EL (B) is proportional to the spin polarization of the injected holes from LSMO electrode, which, in turn is related to LSMO magnetization response. A slight deviation, ΔP(B) of the P EL (B) response from the M(B) response of the LSMO electrode is seen at large B. This is caused by contribution from field induced Zeeman splitting 11 , and magnetic circular dichroism of the LSMO electrode. These responses were measured separately to be smaller than 0.1%, in agreement with the deviation response. (Supplementary Fig. 4 & Supplementary Note 4).
Similar to the circular polarization degree, P PL in MAPbBr 3 films discussed above, P EL of the spin-LED can be expressed by the relation: where P s is the spin injection efficiency of the FM electrode, η is the initial spin polarization in MAPbBr 3 . Assuming η is the same for PL and EL measurements, from the obtained P EL and P PL expressions we get P EL P PL ¼ P s ¼ 26% for the LSMO ferromagnetic electrode at 10 K. In fact, η for the EL emission should be smaller than for PL emission at resonance excitation, namely η EL < η PL , since the spin aligned injected holes in the spin-LED may lose their spin alignment before forming e-h pairs. Therefore, 26% is a lower limit for the LSMO spin injection efficiency. Nevertheless, this large value is surprising when taking into account the existence of the seemingly unsurmised conductivity mismatch between a metallic FM and a semiconductor, which mostly prevents spin injection 14 . This is especially true for the perovskite-based spin-LED in which we found that the spin injection efficiency is one order of magnitude higher than the value in GaAs-based spin-LEDs 11,24 . We speculate that the efficient spin injection into the MAPbBr 3 could be attributed to its large SOC, which is manifested by the obtained surface Rashba splitting 5 that suppresses the conductivity mismatch, and, in turn improves the spin injection at the FM/MAPbBr 3 interface 14 .
Spin valve device based on MAPbBr 3 . We have also measured spin injection into MAPbBr 3 film using electrical detection in a SV device. Figure 4a shows the SV device structure which is composed of two FM electrodes, namely a LSMO film grown on SrTiO 3 (001) substrate and an evaporated cobalt film, with MAPbBr 3 film as spacer layer deposited using the same method as for the spin-LED device discussed above. An in-plane magnetic field B was applied to manipulate the magnetization orientation of the two FM electrodes. Magnetization hysteresis loops of both LSMO and Co electrodes were recorded in situ, using the magneto-optics Kerr effect (MOKE) measured by a Sagnac interferometer 25 . As shown in Fig. 4b, at 10 K we measured the coercive field, B c1 = 5 mT for LSMO and B c2 at around 75 mT for Co. Upon sweeping B, the relative magnetization orientation of the two FM electrodes changes from parallel (P) to antiparallel (AP) configuration, and vice-versa upon changing the field sweeping direction. This leads to change in the device resistance, R, with R P < R AP ; which is called giant magnetoresistance (GMR). The obtained magnetoresistance (MR) ratio is then calculated from the relation: MR = (R AP − R P )/R P . The MR(B) response of a SV device based on MAPbBr 3 measured at 10 K is shown in Fig. 4c. We obtained a substantial maximum GMR, GMR max of 25%, indicating an effective spin injection into the MAPbBr 3 interlayer film, which is consistent with our spin-LED response. Possible artefacts such as tunneling magnetoresistance (TMR), tunneling anisotropic magnetoresistance (TAMR), and anisotropic magnetoresistance (AMR) have been ruled out by control experiments. (Supplementary Figs. 6 & 7, Supplementary Note 6 & 7).
In order to cross-check our results, we conducted the electrical Hanle effect measurements on the same SV. A perpendicular magnetic field, B z was applied to the SV device when the FM electrodes magnetization are in parallel or antiparallel configuration. Under this condition, the injected spin in the MAPbBr 3 interlayer from one FM electrode precesses with Larmor frequency: ω L = gμ B B z /ℏ while the carriers drift to the other FM electrode. This spin precession induces change in the relative angle between the spin polarization of injected carriers and the magnetization direction of the detecting FM electrode, thus causing change in device resistance. The precession angle, θ p of the spin carriers that arrive at the opposing electrode is determined by the relation: θ p = 2πt trans /t p , where t p is spin precession period, and t trans is the carriers transit time (or time of flight, TOF). In disordered materials such as the OITP films, carrier transport is dispersive as demonstrated by TOF measurements 26 , and this causes a broad distribution of t trans value. Consequently, the spin polarization of injected spin aligned carriers averages out at the detecting electrode (spin dephasing), which quenches the measured GMR of the SV device 27,28 . Figure 4d shows that GMR max value vanishes with increasing B z for both parallel and antiparallel SV configuration. The observed Hanle effect unambiguously demonstrates spin injection and transport in the OITP interlayer; consequently our measurements show in fact GMR type response. As a control experiment, after the Hanle measurements were performed, the in-plane MR-loop was repeated to confirm that the magnetization state of the electrodes was not tilted or flipped by perpendicular field.
To extract additional information about the spin lifetime in the MAPbBr 3 interlayer the electrical Hanle response, R(B z ) may be fit using the 1-D spin drift-diffusion model 28 : In Eq. (3), D is the spin diffusion coefficient, d is the MAPbBr 3 film thickness (d = 207 nm here). τ s and ω L = μ B gB z /ℏ are the spin lifetime and Larmor frequency of the spin 1/2 injected carriers. Since holes are the injected carriers in our SV device type, we use in Eq. (3) the g-factor for holes: g h = 0.33 6 . Using Eq.
(3) to fit the Hanle response, we obtain τ s = 936 ± 23ps and D = 0.21 ± 0.08 cm 2 s −1 for the holes in MAPbBr 3 at 10 K. This large spin lifetime contributes to the long spin diffusion length of around 220 nm in MAPbBr 3 ( Supplementary Fig. 8 & Supplementary Note 8). It is interesting that the obtained τ s for holes is larger than that for excitons.

Discussion
In this work, we presented the study of spintronic devices based on OITP polycrystalline films. Successful spin injection from FM metallic electrodes into OITP has been demonstrated. The spin lifetime for both excitons and holes in MAPbBr 3 is surprisingly long despite the large SOC; this indicates that the OITP semiconductor family could be promising candidates for spintronics applications.

Methods
Sample preparation. The ferromagnetic LSMO thin films were grown on SrTiO 3 (001) substrates by pulsed laser deposition (PLD) and patterned by wet-etch optical lithography. The LSMO electrodes were cleaned and re-used multiple times. The MAPbBr 3 overlayer was deposited by the spin coating method inside a N 2 filled glove box (O 2 /H 2 O < 1 ppm). The MAPbBr 3 precursor solution was obtained by mixing PbBr 2 and MABr (molar ratio = 1:1.1) in dimethyl sulfoxide (DMSO) at a concentration of 0.5 M. The perovskites solutions were stirred overnight at 50°C before use. To prepare the perovskite thin film, the solutions were spin-coated on the O 2 plasma treated substrates at 4000 rpm. To achieve pinhole-free microscopic structure for the MAPbBr 3 film, chloroform solvent was drop-casted onto the perovskite film for a nanocrystal pinning process during spin coating (as described in ref. 2 ). Finally, the perovskite films were annealed on a hot plate at 100°C for 30 min. Device fabrication. For the various devices, MAPbBr 3 thin film was spin coated onto the bottom LSMO electrode. After cooling down to ambient temperature, the samples were transferred back to the vacuum chamber for e-beam evaporation of other layers. For the spin-LED device, the bottom LSMO electrode was patterned as 3 × 5 mm strips. Following the MAPbBr 3 film deposition, a 10 nm TPBi film was deposited as the electron transport layer. Subsequently, a 100 nm aluminum film was coated as top electrode in a crossbar configuration. The typical device area was 0.5 × 3 mm. For the SV device, the bottom LSMO electrode was patterned as 0.2 × 5 mm strips. 15 nm cobalt film followed by 30 nm gold were capped as the second ferromagnetic electrode in a crossbar configuration, using a shadow mask. The typical device area in this case was 200 × 200 μm.
Device characterization. The out-of-plane magnetization curve of LSMO electrode in spin-LED device is measured by SQUID magnetometer at 10 K. On the other hand, magnetic hysteresis loop of two ferromagnetic electrodes is measured by a home-built Sagnac interferometer with a static DC Kerr rotation sensitivity of 20 nrad, and spot size of about 1 μm. The MOKE responses of both LSMO and Co electrodes were measured at 10 K in a perovskite-based SV configuration. This unique technique has enabled us to measure the magnetization properties of the spin-polarized interface in situ rather than bulk Co and LSMO electrodes outside the SV device. The EL and PL were measured by a silicon detector. The transport measurements were performed in a closed-cycle refrigerator with temperature ranging from 10 to 300 K by a standard four points method with Keithley 236 power supply and Keithley 2000 multimeter with an in-plane magnetic field B.

Data availability
The data that support the findings of this study are available from the corresponding author on reasonable request.