Report

Spin-Polarized Light-Emitting Diode Based on an Organic Bipolar Spin Valve

See allHide authors and affiliations

Science  13 Jul 2012:
Vol. 337, Issue 6091, pp. 204-209
DOI: 10.1126/science.1223444

Spin-Dependent Light Emission

Spintronic devices exploit electronic currents that are spin polarized, which have an excess of one spin current over the other. One way to detect this polarization would be to create a light-emitting diode that is sensitive to spin polarization. Along these lines, Nguyen et al. (p. 204) constructed a bipolar device in which an organic semiconductor was sandwiched between two ferromagnetic contacts whose relative polarization could be controlled by an applied magnetic field. Magneto-electroluminescence of the order of ∼1% was observed at a bias voltage of ∼3.5 V. The use of a deuterated organic polymer interlayer improved spin transport relative to polymers with hydrogen side groups, and a thin LiF buffer layer on the ferromagnetic cathode improved electron injection efficiency.

Abstract

The spin-polarized organic light-emitting diode (spin-OLED) has been a long-sought device within the field of organic spintronics. We designed, fabricated, and studied a spin-OLED with ferromagnetic electrodes that acts as a bipolar organic spin valve (OSV), based on a deuterated derivative of poly(phenylene-vinylene) with small hyperfine interaction. In the double-injection limit, the device shows ~1% spin valve magneto-electroluminescence (MEL) response, which follows the ferromagnetic electrode coercive fields and originates from the bipolar spin-polarized space charge–limited current. In stark contrast to the response properties of homopolar OSV devices, the MEL response in the double-injection device is practically independent of bias voltage, and its temperature dependence follows that of the ferromagnetic electrode magnetization. Our findings provide a pathway for organic displays controlled by external magnetic fields.

The quest for a spin-polarized organic light-emitting diode (spin-OLED) (13), in which the electroluminescence (EL) intensity is sensitive to the spin polarization of the injected carriers, has been a goal in the field of organic spintronics since the successful implementation of an organic spin valve (OSV) based on the small molecule aluminum tris(8-hydroxyquinoline) (Alq3) (4). Despite several attempts at spin-OLEDs (1, 5) in which Alq3 was used as the organic interlayer between two ferromagnetic (FM) electrodes in a vertical configuration, this goal has not been achieved as yet. The main obstacle in realizing such a device has been the relatively high bias voltage Vb needed for reaching substantive EL efficiency in the device at low temperatures. For example, Vb > 10 V is needed for EL of Alq3 with FM electrodes at temperature T = 10 K, but the OSV performance sharply deteriorates with Vb and is limited to <1 V for practical operation (4, 6, 7).

We report the realization of a spin-OLED based on a bipolar OSV device that exhibits magneto-electroluminescence (MEL) on the order of ~1% at Vb ≈ 3.5 V, with emission intensity modulation that follows the coercive fields of the FM electrodes. Two important technical advances enabled this achievement. First, our devices are based on a deuterated organic polymer interlayer with superior spin transport properties that has a smaller hyperfine interaction than polymers based on hydrogen side groups (7). Second, we deposited a thin LiF buffer layer in front of the FM cathode to improve the electron injection efficiency (8). The bipolar OSV response has substantially different voltage, temperature, and thickness dependencies relative to the response in homopolar OSV based on the same organic interlayer. These differences are caused in part by the spin-aligned space charge–limited current (SCLC) operation upon reaching double-injection conditions during bipolar operation.

The device operation scheme (Fig. 1A) shows the injected electrons and holes initially forming polaron pairs (PPs) at the appropriate Vb needed for bipolar injection. These species are precursor excitations to singlet excitons (SEs) that may recombine radiatively and emit EL. With non-FM electrodes (Fig. 1A, panel 1), the net electron-hole bimolecular rate coefficient b for forming PPs did not depend on the magnetic field. Under the assumption of SCLC operation, the fraction of current from electron-hole recombination was inversely proportional to the rate b (9). When the OLED device was driven with FM electrodes that inject spin-aligned carriers, the rate b became field-dependent (Fig. 1A, panels 2 and 3) because the external magnetic field changed the mutual magnetization directions of the spin-injecting FM electrodes. Thus, the PP formation rate, EL intensity (MEL), and current density (magnetoconductivity, MC) all become field-dependent. This operation scenario of spin-OLED is more realistic than the simple model described in (2, 6) because the intermediate step of PP formation, as well as the spin mixing among its spin singlet (PPS) and spin triplet (PPT) configurations, is explicitly considered (10, 11). In fact, the spin-mixing channel is responsible for a variety of effects in OLED devices with non-FM electrodes (such as monotonic MC and MEL responses) that we term “intrinsic” MC and MEL responses (7, 11, 12), as well as EL quantum efficiency that is not limited to 25% (13).

Fig. 1

(A) Spin-OLED device operation under the condition of unbalanced electron-hole space charge limited current (SCLC): (1) OLED with non-FM electrodes; the “recombination” current δI is inversely related to the efficiency of PP formation via the bimolecular recombination coefficient b; also, EL ∝ δI. (2 and 3) OLED with FM electrodes: b becomes magnetic field–dependent via the spin injection of the FM electrodes, giving rise to spin-dependent current and EL. (B) The spin-OLED device structure, where the D-DOO-PPV organic layer thickness is ~25 nm and LiF buffer layer thickness is ~1.5 nm. Here the in-plane magnetic field (black arrow) causes the FM magnetizations (red arrows) to align parallel to each other. The EL emission (wavy red line) is collected through the Co/Al thin electrode. (C) The device I-V and EL-V characteristics; the EL onset is at Vo ≈ 3.5 V. Inset: D-DOO-PPV polymer chemical structure.

The spin-OSV device was designed to achieve efficient EL emission at relatively low Vb Fig. 1C, with sizable spin injection capability from the FM electrodes and with large spin diffusion length in the organic interlayer. We show the spin-OLED device structure in Fig. 1B. For the anode, we used the half-metal FM La0.7Sr0.3MnO3 (LSMO), which has a coercive field Bc ≈ 5 mT at cryogenic temperatures (Fig. 2D); the cathode was a FM Co thin film (Bc ≈ 35 mT at cryogenic temperatures; Fig. 2D) capped with an Al layer for corrosion protection. The organic interlayer film, with thickness d ranging from 18 to 50 nm, was based on deuterated poly(dioctyloxy)phenyl vinylene (D-DOO-PPV), a π-conjugated polymer in which all the hydrogen atoms closest to the backbone chain were replaced by deuterium (Fig. 1C, inset). It was previously shown (7) that the hyperfine interaction in D-DOO-PPV is considerably reduced, thus increasing the spin diffusion length λS to ~45 nm; this is about 3 times the value of λS in H-DOO-PPV polymer. In addition, a thin LiF layer (thickness d′ ranging from 0.8 to 1.5 nm) was deposited as a buffer layer between the organic layer and Co electrode to improve electron injection (14) and to block the formation of Co inclusions (4, 15).

Fig. 2

Magneto-electroluminescence (MEL) response of a spin-OLED device. (A) Obtained MELEX(B) response for up (red) and down (blue) B-sweeps, measured at Vb = 4.5 V and T = 10 K, for device A (d = 25 nm, d′ = 1.5 nm). The black dashed line describes the nonhysteretic, intrinsic MEL background response for an up-sweep. The horizontal arrows mark the relative electrode magnetization directions. (B) The net MELSV(B) response after subtraction of the background MEL from the measured MEL response shown in (A). (C) The bias voltage dependence of the maximum MELSV value. (D) Magneto-optic Kerr effect (MOKE) measurements of the LSMO and Co/LiF electrodes at 10 K that show coercive fields Bc(FM1) ≈ 5 mT and Bc(FM2) ≈ 35 mT, respectively.

The turn-on voltage Vo for sizable EL emission at the double-injection condition was reached at Vo ≈ 3.5 V (Fig. 1C); this value was Vo ≈ 10 V without the LiF layer (8). Because Vo is still relatively high in the spin-OLED device, we conjecture that hole injection is more efficient than electron injection. This difference led to unbalanced charge injection; most of the current density was carried by the holes, whereas the EL intensity was limited by the minority electron injection from the Co/LiF cathode. Under these conditions, the “intrinsic” MEL and MC responses (7), those unrelated to the spin valve, were small (figs. S1 to S3), and this allowed us to readily study the spin valve–related MEL response.

At cryogenic temperatures, the FM LSMO (FM1) and Co (FM2) electrodes in the spin-OLED had nominal spin injection degrees of polarization of P1 ≈ 95% and P2 ≈ 30% [which may depend on the environment; see (16)]. However, P2 substantially dropped because of the LiF buffer layer (8, 17). Because Bc(FM1) ≠ Bc(FM2), we could switch their relative magnetization directions between parallel (↑↑) and antiparallel (↑↓) relative alignments by sweeping the external magnetic field B (horizontal arrows in Fig. 2A), whereby the device resistance, conductance, and EL intensity depended on the relative magnetization orientations of the FM electrodes. We thus measured MEL(B) and MC(B) at various bias voltages, temperatures, and device thicknesses.

A typical EL(B) response of a D-DOO-PPV spin-OLED measured at 10 K is plotted as MELEX(B) ≡ [EL(B) – EL(↑↑)]/EL(↑↑) in Fig. 2A for a device with d = 25 nm and LiF d′ = 1.5 nm. The EL(B) response had two components: (i) a hysteretic negative MELSV component and (ii) a nonhysteretic positive MELLSMO component (black dashed line in Fig. 2A). The MELSV response component consisted of a downward sharp jump of ~0.4% in the antiparallel magnetization configuration between 4 and 30 mT that followed the electrodes’ coercive fields (Fig. 2D). The MELLSMO response was caused by the magnetic properties of the LSMO electrode (4) combined with the “intrinsic” MEL response (7); it was a monotonic function of |B| and symmetric with respect to B = 0 (fig. S2). A similar MEL component was measured before in FM-OLED devices based on Alq3 at room temperature (1) and was ascribed to the non–spin valve MEL response of the organic interlayer. In that case, the sudden change in the EL(B) response at the electrodes’ respective Bc values was positive with increasing B, and was thus interpreted as having been caused by the stray field BS that arises from the proximity of the FM electrodes to the organic interlayer. We measured BS of the LSMO and Co/LiF electrodes in our device (figs. S2 and S3). For devices with one FM electrode, we found BS(LSMO) ≈ 0.7 mT (fig. S2) and BS(Co) ≈ 3.5 mT (fig. S3) at cryogenic temperatures. However, the average BS increased when two FM electrodes were deposited; in this case, we measured BS ≈ 4 mT (fig. S4), which is somewhat greater than in devices with one FM electrode but is too small for explaining the MELSV sharp response in our devices, given that the intrinsic MEL response is weak (fig. S1). In addition, the MELSV response was negative, in contrast to the positive MEL jump related to the stray field (1) (fig. S4).

Moreover, the MEL was isotope-dependent. We measured the MEL response in devices with different DOO-PPV isotopes (7). The spin diffusion length was isotope-dependent, and the spin valve–related MEL response indeed depended on the polymer isotope (fig. S5). Thus, the MEL response cannot be interpreted as arising from the stray fields that influence the intrinsic MEL response, as in (1). We thus conjecture that the obtained MELSV response in the bipolar OSV is a genuine spin valve effect.

To facilitate data analysis, we subtracted the nonhysteretic MELLSMO response [component (ii)] from the MELEX(B) response (Fig. 2A) to obtain the “net” spin valve–related response [component (i)], MELSV(B) ≡ MELEX – MELLSMO (Fig. 2B). MELSV(B) displayed the typical hysteretic spin valve characteristic response with sharp jumps at the LSMO and Co coercive fields. Moreover, one of the most prominent features of the MELSV(B) response is the very weak dependence of its maximum value, MELmax ≡ max(|MELSV(B)|), on Vb (Fig. 2C). This response substantially differs from the strong decrease of the magnetoresistance MRmax with Vb in homopolar OSV devices (4, 18, 19). It is thus clear that the performance of the bipolar OSV device degrades less with Vb relative to a homopolar OSV based on the same organic layer (see below).

We measured the OSV “figure of merit” MELmax at 10 K and V = 4.5 V for various device thicknesses d and LiF buffer layer thicknesses d′ (Fig. 3A). We found that MELmax decreases as d and d′ increase. The decreased performance with increasing LiF d′ may be readily explained as arising from the decrease of the cathode spin polarization P2 with the LiF buffer layer thickness (8). The decreased performance with increasing organic layer d may be caused by a finite “effective” spin diffusion length λS at the bipolar injection condition reached here. From the device thickness dependence shown in Fig. 3A we estimate λS ≈ 25 nm, which is different from λS = 45 nm obtained at small bias voltage (7). The best device performance, a MELmax value of 1.1% (Fig. 3B), was obtained for a bipolar OSV device having d = 18 nm and d′ = 0.8 nm (Fig. 3A). Further decreases of d and d′ caused the OLED devices to become unstable.

Fig. 3

(A) The maximum MELSV response of spin-OLED devices at various polymer thicknesses d and LiF buffer layer thicknesses d′ of 0.8 nm (red squares) and 1.5 nm (blue squares), measured at T = 10 K and Vb = 4.5 V. (B) The optimum MELSV(B) response of ~1.1% measured for a device with d = 18 nm and d′ = 0.8 nm. (C) The maximum MELSV(T) response at Vb = 5 V (red squares) for a spin-OLED device with d = 25 nm and d′ = 1.5 nm; the LSMO bulk magnetization versus T measured by superconducting quantum interference device (SQUID) (blue stars); and its fit using the Brillouin function BJ(T/Tc) with J = 5/2 and Tc = 307 K (blue line). (D to G) MEL(B) response at selected temperatures.

In Fig. 3, D to G, we show MELSV(B) response at various temperatures, and summarize MELmax versus temperature relative to the measured LSMO bulk magnetization, M(T) in Fig. 3C. The MELmax(T) values almost perfectly follow the M(T) response. This behavior is in stark contrast to MRmax(T) in homopolar OSV devices, where a much steeper temperature dependence was observed (2023) and was explained (21, 22) as having been caused by the LSMO surface magnetization decreasing with T (23).

To better compare the homopolar and bipolar OSV devices, we show in Fig. 4 the effect of the LiF buffer layer on the device magnetoconductance MC(B) response. The measured response, MCEX(B) ≡ [I(B) – I(↑↑)]/I(↑↑), shows a nonhysteretic background that is similar to that observed in the MELEX(B) response in Fig. 2A. We again subtracted this background MC response to obtain the net response MCSV(B), which is shown in Fig. 4, A and B, for the bipolar (LiF/Co cathode) OSV device and in Fig. 4, D and E, for the homopolar (Co cathode) OSV device. The opposite sign of the two MC response sets demonstrates that the LiF layer reverses the cathode spin polarization in agreement with (8). In Fig. 4, C and F, we show MCmax ≡ max(|MCSV(B)|) as a function of Vb for the homopolar and bipolar OSV devices. Surprisingly, we see that although the MCmax(Vb) dependence of the bipolar OSV device sharply decreased for Vb < 3.5 V, it abruptly leveled off at Vo and became practically independent of bias voltage. This outstanding property of the bipolar OSV device facilitates the realization of spin-OLED at Vb > V0.

Fig. 4

Magnetoconductance (MC) response of bipolar and homopolar OSV devices based on D-DOO-PPV and measured at 10 K (d = 25 nm, d′ = 1.5 nm). (A and B) MC(B) response of bipolar OSV device measured at Vb = 0.6 V and 5 V, respectively, at positive (red) and negative (blue) B-sweeps. (C) Maximum MCSV value versus Vb for the bipolar device. (D and E) MC(B) response of homopolar OSV device measured at Vb = 1 V and 3.5 V, respectively, at positive (red) and negative (blue) B-sweeps. (F) Maximum MCSV value versus Vb for the homopolar device.

In the following, we analyze the spin-OLED device response under conditions of unbalanced bipolar current density J where the electron current density Je is injection-limited and Je << J. Under these conditions, most of the device current density is carried by the hole current Jh along with an additional small “recombination current” JR caused by the electron-hole “recombination” that leads to PP formation (Fig. 1A). Jh, which is the sole current through the device for Vb < Vo, gives rise to the bias voltage–dependent MCSV that is usually observed in homopolar OSV devices (7). The homopolar MCSV appears to follow a Jullière-type behavior (4, 17): MCSV(Vb < Vo) ∝ 2P1P2/(1 + P1P2), where P1 and P2 are the cathode and anode spin polarizations, respectively. JR, however, turns on at VbVo and is responsible for the voltage-independent MELSV and MCSV responses. These latter responses appear to follow a novel “recombination-modified” Jullière-type behavior: Both MCSV(Vb > Vo) and MELSV are proportional to P1P2Δb, where Δb = buubud, and buu and bud are the spin-dependent bimolecular recombination rate constants for up-up and up-down electron-hole relative spin directions, respectively. Note that although both MCSV(Vb < Vo) and MCSV(Vb > Vo) are proportional to P1P2, only MCSV(Vb < Vo) is voltage-dependent. We thus conclude that the homopolar MCSV voltage dependence cannot be caused by the FM electrode polarization as originally postulated (4), but rather originates within the device volume by a mechanism that does not affect the recombination current JR. The electron-hole recombination products that are the singlet (PPS) and triplet (PPT) polaron pairs intermix through an intersystem crossing enabled via a variety of spin mixing interactions, such as the hyperfine, exchange, and spin-orbit interactions. Both same-spin polarized and opposite-spin polarized electron-hole “recombination” contribute, albeit not equally, to the steady-state PPS density and eventually to EL (7).

To understand the obtained bipolar OSV properties, we extend the classical bipolar SCLC Parmenter-Ruppel (PR) model (9) to include FM electrodes under the condition of unbalanced current density without the effect of traps (24). In this case the J-V relation is given byEmbedded Image (1)where ε is the dielectric constant; μh, μe, and μR = εε0b/2e are the hole, electron, and recombination mobilities, respectively; b is the bimolecular recombination coefficient in the reaction rate RPP = bnp in which electrons of density n and holes of density p generate weakly coupled PP species (see supplementary text); Jh (>> Je) is the hole majority SCLC density; and JR is the recombination current density. Although JR was originally ignored by PR because JR << Jh, here we keep this term because it is the only term that leads to EL emission.

For FM electrodes, the fraction of spin-polarized electrons injected by the cathode FM1 and collected by FM2 is (1 ± P1P2)/2 for ↑↑ and ↑↓ electrode magnetization directions, respectively, and the same is true for the fraction of spin-polarized holes that is injected by the anode FM2 and collected by FM1; here we assumed for simplicity that the spin diffusion length λS >> d. In this case the spin-sensitive bimolecular recombination coefficients buu and bud cause JR to depend on the mutual magnetization directions of the FM electrodes. The electrode magnetization–dependent SCLC can then be written asEmbedded Image (2)where Embedded Image are the recombination mobilities for parallel and antiparallel electrode magnetizations, respectively (see supplementary text). Thus, the magnetoconductance, defined as MC = (J↑↑J↑↓)/ J↑↑ is composed of two components, MCh, from the majority hole current, and MCR from the recombination current. When JR << JJh, these two components are

Embedded Image (3)

where Jh and JR are given in Eq. 1 with μR = (μR↑↑ + μR↑↓)/2 and b = (buu + bud)/2. Note that MCh has the form of the Jullière formula (17) for a homopolar OSV, which is derived here for the case of SCLC, whereas the new term MCR is related to both electrode polarizations as well as the difference, Δb. For LSMO (P1 ≈ 1) and Co/LiF [P2 ≈ 0.04 at small Vb (Fig. 4F)], (P1P2)2 ≈ 10−3 << 1 and thus MCh ≈ 2P1P2, whereas MCR ≈ 2P1P2(JR/Jh)(Δb/b). We conclude that both MCh and MCR are proportional to P1P2 and thus disappear in an OLED with non-FM electrodes.

The EL emission results from the radiative recombination of singlet excitons that emerge from their PPS precursor. Thus, the EL intensity is directly proportional to the steady-state PPS density NPPS. The intermixing of PPS ↔ PPT means that NPPS is determined by both singlet and triplet channels, Embedded Image (4)whereEmbedded Image(5)is the singlet (triplet) channel “recombination” (or PP formation) rate, and κS(T) designates the effective singlet (triplet) channel reaction rate, which is spin- and magnetization-independent. Using a rate equation approach to calculate NPPS, we find Embedded Image (6)(see supplementary text). All spin-independent rates cancel out from the MEL expression. When comparing Eqs. 3 and 6, for bipolar OSV, MC and MEL have the same sign, and MEL is greater than MC by the factor Jh/JR (>>1).

Figure 4C shows two regimes in the MCSV(Vb) response for the bipolar OSV. For Vb < V0 (i.e., the hole-only injection regime), MCSV decreases by a factor of ~50 between Vb ≈ 0 and Vb = 3.5 V, similar to the homopolar OSV based on D-DOO-PPV (Fig. 4F). However, for Vb > V0 (i.e., the bipolar injection regime), MCSV(Vb) is practically voltage-independent, unlike MCSV(Vb) of the homopolar device (Fig. 4F). Note that ELSV is also voltage-independent (Fig. 2C). We thus conclude that homopolar OSV devices become less efficient at large Vb, but less so for bipolar operation. Our SCLC model separates MCSV into two different components: namely, the “homopolar MC” component (MCh in Eq. 3) and the “recombination MC” component (MCR in Eq. 3). We conjecture that the homopolar MC component decreases with Vb, whereas the recombination MC component does not depend on Vb. For Vb < Vo, the bipolar MC(Vb) response is dominated by the hole-only OSV that monotonically decreases with Vb. However, as bipolar injection sets in at Vo, the voltage-independent MCR takes over and the MC(Vb) response becomes Vb-independent. Simultaneously, MEL is given by Eq. 6 and thus is also independent of bias voltage. MC and MEL also have the same sign for Vb > Vo, as predicted by Eqs. 3 and 6. In addition, the obtained ratio MELSV/MCSV ≈ 25 measured at Vb > 4 V (Figs. 2C and 4C) is in agreement with the larger MEL predicted by our model, where MEL/MC ≈ J/JR >> 1.

The performance of homopolar OSV devices severely degrades with Vb (7, 16, 19). Two possible mechanisms might explain this behavior: (i) decrease of the spin injection efficiency of the electrodes with increasing Vb via the term P1P2, and (ii) voltage-dependent processes that occur in the organic layer. Because both MCh and MCR are proportional to P1P2 (Eq. 3), but only MCh degrades with Vb, we conjecture that the observed MCSV(Vb) decrease cannot originate from a decrease of P1P2 dependence on Vb. By adding the screened Frenkel effect to the homopolar SCLC operation, the MCSV(Vb) decrease was recently explained as arising from the magnetic field–dependent “screening length” λsc (25). Such a mechanism would not affect the “recombination current” in a bipolar OSV for electron-hole distances r < λsc, and this may explain the voltage-independent response of the spin-OLED.

Our results provide a pathway toward organic displays controlled by external magnetic fields, but such applications would require a larger MEL and room-temperature operation. These requirements might be achieved by choosing different FM electrodes and/or organic interlayers. Finally, we note the possibility of manipulating the EL emission colors in spin-OLEDs by an external magnetic field, unlike inorganic spin-LEDs.

Supplementary Materials

References and Notes

  1. Acknowledgments: Supported by NSF grant DMR-1104495 and MRSEC, DMR-1121252 program at the UoU (T.D.N. and Z.V.V.), Israel Science Foundation grant ISF 472/11 (E.E.), and Israel-USA BSF grant 2010135 (Z.V.V. and E.E.). The D-DOO-PPV polymer synthesis was supported by U.S. Department of Energy grant DE-FG02-04ER46109. We thank X.-G. Li (USTC) for providing the LSMO substrates. The authors declare no conflict of interest associated with this work. A patent disclosure related to the spin-OLED invention was recently filed with the University of Utah, disclosure no. 5249, which has been filed as a provisional patent application.
View Abstract

Stay Connected to Science

Navigate This Article