SEARCH

SEARCH BY CITATION

Keywords:

  • interfaces;
  • organic semiconductors;
  • organic spintronics;
  • spin valves

Abstract

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Spin-polarized tunneling
  5. 3 Electronic structure of thin film, amorphous organic semiconductors
  6. 4 Electronic structure at interfaces
  7. 5 Magnetoresistance in organic spin valves
  8. 6 The role of interfacial layers
  9. 7 Hole, electron, or ambipolar transport?
  10. 8 Conclusions
  11. Acknowledgements
  12. Biographical Information

Interfaces between dissimilar materials are critical to the performance of many electronic devices. In electrical devices where the active transport is through an organic semiconductor (OSC) layer there are metallic electrodes to supply and remove the charge carriers. At metal/molecule and molecule/metal interfaces the electrical contact is usually not Ohmic. As a result, in most metal-organic semiconductors–metal devices charge injection is mainly governed by tunneling across the barriers. When the metal electrodes are ferromagnetic the density of occupied and unoccupied electronic states is spin-dependent, which presents further subtleties to the electron transfer. In this article we summarize some of the important experimental results that have advanced our understanding of spin-polarized electron transfer across ferromagnetic metal (FM)/OSC interfaces. In particular, we highlight key spectroscopic studies that have revealed insights into the nature of the electronic structure between OSCs and FMs, in particular between Alq3 and Co and Fe surfaces. We discuss the relationship between energy level diagrams for these systems and transport measurements made on spin-valve devices.


1 Introduction

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Spin-polarized tunneling
  5. 3 Electronic structure of thin film, amorphous organic semiconductors
  6. 4 Electronic structure at interfaces
  7. 5 Magnetoresistance in organic spin valves
  8. 6 The role of interfacial layers
  9. 7 Hole, electron, or ambipolar transport?
  10. 8 Conclusions
  11. Acknowledgements
  12. Biographical Information

The transfer of electrons across the interface between inorganic and organic materials is critical to the performance of organic light emitting diodes, field effect transistors, and photovoltaic devices. More recently the spin-dependent transfer of electrons between ferromagnetic metals (FMs) and organic semiconductors (OSCs) has become an active area of research, owing to its important role in determining the performance of organic spin-valve devices. In its simplest manifestation a spin-valve is a trilayer structure, where two magnetic layers are separated by a non-magnetic layer. The term spin-valve refers to a device that allows preferred transport of a given carrier spin-channel over another with respect to the magnetic configuration of the electrodes. The spin-valve effect was first observed in multilayer structures of Co/Cr 1. In this article the focus is organic spin-valves, specifically where the non-magnetic layer is an OSC. The interfaces between the FM and the organic layer are not likely to be Ohmic. For example the vacuum level offset at a Co/Alq3 interface has been found to be 1.4 eV 2. Hence, one must first consider tunneling as the primary mechanism for charge/spin transfer across such interfaces. Transport, primarily via tunneling, in structures where a single molecular layer is sandwiched between metal electrodes has been extensively studied and well documented over the past few decades 3. However, much less is known about spin-dependent tunneling mechanisms across FM/molecule interfaces. Consequently, we will begin by briefly reviewing the fundamental concepts of spin-dependent tunneling across ferromagnetic/inorganic tunnel barrier interfaces as way of an introduction to this topic.

2 Spin-polarized tunneling

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Spin-polarized tunneling
  5. 3 Electronic structure of thin film, amorphous organic semiconductors
  6. 4 Electronic structure at interfaces
  7. 5 Magnetoresistance in organic spin valves
  8. 6 The role of interfacial layers
  9. 7 Hole, electron, or ambipolar transport?
  10. 8 Conclusions
  11. Acknowledgements
  12. Biographical Information

Majority and minority spin electrons are often referred to as “up” or “down” electrons, respectively. Consider the simplified drawing of the bulk electronic band structure for Fe and Ni shown in Fig. 1. It is well known the d-bands split due to the exchange interaction 4. For Ni there is a larger density of states at the Fermi energy for the minority electrons, while Fe has a larger density of states for the majority electrons. Not shown in Fig. 1 is the sp hybridized band that crosses the Fermi energy and this has important consequences in transport studies described below. The band structure depicted in Fig. 1 suggests that the majority and minority bands for Fe are inverted with respect to Ni. However, tunneling experiments that utilize a superconducting electrode as a spin-detector reveal a different picture. Tedrow and Meservey pioneered measurement of the spin-polarization of a ferromagnet by tunneling spectroscopy. The details of the technique will not be described here, but the interested reader should consult their excellent review article 5. Briefly a superconducting film, usually Al, is held in an external magnetic field that is parallel to the surface. If the film is separated by an insulating barrier from a ferromagnetic electrode and is cooled below the critical superconducting temperature, the spin-up and spin-down density of states are separated by the Zeeman energy, i.e., 2 µBH, where µB is the Bohr magneton and H is the strength of the external magnetic field. When the conductance is measured under these conditions there will be two peaks symmetrically displaced from zero bias, due to the different energy gap for the spin-up and spin-down electrons. However, the amplitude of the peaks will not be the same due to the different density of states in the spin-up and spin-down density of states. Using a rigorous theoretical description of the superconductor it is possible to extract both the sign and magnitude of the spin-polarization 5. The results of many experiments reveal that the sign of the polarization is positive for Fe, Ni, and Co, which does not seem to support the simple band structure shown in Fig. 1. The positive spin-polarization in Fe, Ni, and Co results from the fact that the tunneling probability is higher for s electrons over d electrons because they have a higher group velocity at the Fermi energy.

thumbnail image

Figure 1. (online color at: www.pss-b.com) Schematic diagram of the bulk electronic structure for Fe, and Ni.

Download figure to PowerPoint

In 1975 Jullière measured the tunneling magnetoresistance ratio (TMR) for a Co/Ge/Fe sandwich structure and proposed a simple way to estimate the magnitude of the effect 6. The model is essentially a combination of the Mott two-current model and the Meservey–Tedrow estimate of the effective density of states measured by spin-polarized tunneling. In the Jullière model, the spin-up and spin-down electrons are thought to cross the barrier in parallel channels. If the electron's spin is conserved, then the tunneling conductance between the ferromagnetic electrodes can be written as the sum of the conductance for the two channels. Thus for parallel alignment of the electrodes magnetizations the spin-up electrons on the left side ([UPWARDS ARROW]l) are still spin-up on the right ([UPWARDS ARROW]r) after tunneling through the barrier. Conversely, the spin-down electrons on the left ([DOWNWARDS ARROW]l) remain spin-down on the right ([DOWNWARDS ARROW]r). In contrast for an anti-parallel alignment of the electrodes magnetizations, electrons that are locally spin-up find themselves in a region of opposite magnetization, which means that they are locally spin-down. Consequently, spins-down on the left become spins-up on the right. In this model it is straightforward to derive the formula for TMR. In the parallel and anti-parallel configurations of the magnetization directions the equations for DC conductance are Gp = G([UPWARDS ARROW]l,[UPWARDS ARROW]r) + G([DOWNWARDS ARROW]l,[DOWNWARDS ARROW]r) and Gap = G([UPWARDS ARROW]l,[DOWNWARDS ARROW]r) + G([UPWARDS ARROW]r,[DOWNWARDS ARROW]l), respectively. The hypothesis is that the conductance is proportional to the density of states of the left and right electrodes. The TMR is usually defined as the ratio of the change in conductance to the minimum conductance by the equation TMR = (Gp − Gap)/Gap. Furthermore if the polarization of the left and right electrodes are defined as Pl = N([UPWARDS ARROW]l) − N([DOWNWARDS ARROW]l)/N([UPWARDS ARROW]l) + N([DOWNWARDS ARROW]l) and Pr = N([UPWARDS ARROW]r) − N([DOWNWARDS ARROW]r) / N([UPWARDS ARROW]r) + N([DOWNWARDS ARROW]r), respectively, where N is the number of spin-up and spin-down electrons, the equation for the TMR becomes 2PLPR/(1 − PLPR). Thus the TMR can also be expressed in terms of the spin polarization of the left and right electrodes. Although this formula is frequently used to analyze data from TMR experiments, it must be taken with great care. First, the polarization should not be interpreted as the spin-polarization of the density of states at the Fermi energy. The reason for this is that the spin polarization of the tunneling current can only be measured when electrons tunnel between a ferromagnetic electrode and a superconducting electrode. Another indication that the TMR depends on both the barrier and electrodes is that the sign of the TMR has been observed to change when the barrier is changed without replacing the electrodes. The Jullière formula is most appropriate when comparing TMR for systems with different electrodes while keeping the same barrier material.

In contrast to the discussion above about amorphous barriers, there is a well-developed theory to describe TMR with crystalline tunnel barriers. MacLaren et al. 7 was amongst the first to illustrate the shortcomings of the Jullière model. Furthermore, using first principles calculations for epitaxial Fe(100)/MgO(100)/Fe(100) trilayers, Butler et al. 8 showed how the decay of the evanescent wavefunction into the barrier depends critically on its symmetry, which has a dramatic influence on the calculated TMR value. In contrast, in the Jullière model the TMR is dependent only on the polarization of the electrodes and not on the properties of the barrier.

3 Electronic structure of thin film, amorphous organic semiconductors

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Spin-polarized tunneling
  5. 3 Electronic structure of thin film, amorphous organic semiconductors
  6. 4 Electronic structure at interfaces
  7. 5 Magnetoresistance in organic spin valves
  8. 6 The role of interfacial layers
  9. 7 Hole, electron, or ambipolar transport?
  10. 8 Conclusions
  11. Acknowledgements
  12. Biographical Information

When molecules are incorporated into a tunnel barrier several new aspects emerge. First, the molecules may or may not form chemical bonds with the metal electrode. In the former case, i.e. physisorption, the electron cloud of the molecule will “push back” the wavefunction of the metal surface. The result is a lowering in the work function of the metal. In the latter case, i.e. chemisorption, charge transfer may take place and a strong interface dipole may form. These two bonding regimes suggest two limiting cases of electronic coupling between the metal's and molecule's wave functions. First, in the case of physisorption we expect very little overlap of the wave functions, however in the case of chemisorption there is strong electronic coupling. Second, molecules have discrete energy levels that may participate in charge transport across the barrier, i.e., resonant tunneling. There are a few examples in metal–molecule–metal junctions where this behavior has been observed 9, 10. The molecular orbitals most likely to participate in charge transport are the highest occupied and lowest unoccupied, abbreviated HOMO and LUMO, respectively. Measuring the energetic location of the HOMO with respect to the Fermi energy of the metal is rather straightforward in an ultraviolet photoelectron spectroscopy (UPS) experiment. In contrast, measuring the energetic location of the LUMO is more challenging. The most common methods to determine the LUMO of a monolayer or thin-film are inverse photoemission (IPES) 11, two-photon photoemission 12, 13, scanning tunneling microscopy 14, or measurement of the standard reduction potential in an electrochemical cell 15.

Depending on the method used to measure the HOMO and LUMO the energy difference can be large (∼1 eV). In the context of electron tunneling the energetic position of the LUMO is perhaps more important because electrons can only temporarily reside in unoccupied orbitals. Unfortunately, due to the experimental difficulties in measuring the LUMO at metal/molecule interfaces, there is an absence of data. A further complication is that at a metal/molecule interface, there may be a formation of gap states, which form if a reaction occurs. As a result, many authors report the optical gap as the HOMO–LUMO difference. If the HOMO is known then the assumption is that the optical gap can be added to locate the LUMO, but it is difficult to use optical methods to precisely pin down the LUMO position and there is considerable uncertainty in the values. Figure 2 illustrates three different ways often used to designate the HOMO–LUMO separation for Alq3. In the panel on the left, the LUMO and HOMO energies measured by IPES and UPS are indicated 11. This is commonly referred to as the “transport gap” 16. In the middle panel the LUMO energy is determined by adding the optical absorption. Finally the panel on the right shows the LUMO energy measured with respect to the vacuum level by determining the standard reduction potential in an electrochemical cell 15.

thumbnail image

Figure 2. HOMO and LUMO levels of Alq3 determined by various methods.

Download figure to PowerPoint

4 Electronic structure at interfaces

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Spin-polarized tunneling
  5. 3 Electronic structure of thin film, amorphous organic semiconductors
  6. 4 Electronic structure at interfaces
  7. 5 Magnetoresistance in organic spin valves
  8. 6 The role of interfacial layers
  9. 7 Hole, electron, or ambipolar transport?
  10. 8 Conclusions
  11. Acknowledgements
  12. Biographical Information

4.1 Molecule adsorption on ferromagnetic metals

Due to the importance of charge transfer across metal/OSC interfaces in field effect transistors and organic light emitting devices (OLEDs) there is a large body of experimental work that has been cited in several excellent review articles 17, 18. However, there is much less known about the electronic structure at FM/OSC interfaces. We begin by describing what is known about the adsorption of Alq3 and CuPc on Co, Fe, and LSMO surfaces, since these combinations have been used in several spin-valve devices.

Suzuki et al. 19, 20 performed the first measurement of the spin-polarization of metal pththalocyanines (abbreviated as M-Pc) on magnetized Fe(100) surfaces. Using spin-polarized metastable de-excitation spectroscopy they measured the electronic structure and spin-asymmetry for one monolayer of Mn, Fe, Cu, and Mg–Pc adsorbed on Fe(100). In their first experiment Suzuki et al. found a large difference in the spin-polarization of CuPc on a clean and oxygen covered Fe surface. The clean Fe surface had a positive spin-polarization at the Fermi energy, but changed to a negative polarization after adsorption of molecular oxygen. The exact origin of the difference was not well understood. Interestingly the spin-polarization of the CuPc layer was positive on both surfaces. In their second paper, Suzuki et al. observed a large difference in the spin-asymmetry between the different M-Pc's. They attributed the difference to the perturbation of the electronic structure caused by the central metal ion, which was supported by theoretical calculations of the molecular orbitals.

In 2005 Caruso et al. 21 studied the adsorption of Alq3 on polycrystalline Au and Co thin films with photoemission experiments. A significant difference in the spectra was revealed upon excitation at 32 and 72.8 eV. The latter photon energy was selected because it satisfied the resonant excitation condition of the Al 2p1/2 to 3s transition. The major finding of this study was the large difference in the photoemission intensities of the Alq3 molecular orbitals on Co and Au surfaces. In the spectra taken with the 72.8 eV photons, the orbitals derived from molecular states with strong nitrogen character were enhanced on the Au surface. In contrast the orbitals derived from states with strong oxygen character were enhanced on the Co surface. Caruso suggested the adsorption geometry of the Alq3 on Au and Co must be different to produce a perturbation in the molecular orbitals. Zhan et al. 22, 23 used UPS and X-ray photoelectron spectroscopy (XPS) to characterize the adsorption of Alq3 on LSMO and Co. They found that the work function of LSMO and Co decreased by 0.9 and 1.4 eV, respectively, upon adsorption of Alq3. When Alq3 was adsorbed on LSMO only a single N(1s) peak was observed, which indicated a very small perturbation of the electronic structure of Alq3. In contrast, a very large perturbation in the electronic structure was observed for Alq3 adsorption on Co and Fe surfaces 24. In the XPS spectra three peaks were observed near 400, 398.5, and 397 eV for Alq3 adsorption on Co and Fe. The experimental results were supported by DFT calculations that showed that the nitrogen atoms of the quinolate ligands interact strongly with the Fe surface. These results confirmed the earlier photoemission results by Caruso et al. 21.

The influence of the perturbed electronic structure at these interfaces has not been completely understood yet. There have been several other adsorbate systems where detailed information about the electronic structure has been probed, but how this affects the performance of an organic spin-valve is still largely unknown. One of the challenges that must be met in the future is to link the photoemission work, which is often carried out under high vacuum (HV) conditions on single crystal surfaces, to polycrystalline electrodes deposited under HV.

4.2 Deposition of ferromagnetic metals onto molecules

The deposition of Co onto Alq3, CuPc, and pentacene has also been studied, as well as Fe on rubrene and CuPc. As might be expected, there are clear differences when molecules are deposited onto metal surfaces versus the metal being deposited onto the molecules. We focus on the deposition of Co on Alq3 since this interface has been used in several spin valve devices. Xu et al. studied the deposition of Co onto Alq3in situ by UPS and XPS under Ultra High Vacuum conditions 25. The XPS spectra are shown in Fig. 3 and the main findings are summarized below. First there were shifts in the C(1s), O(1s), and N(1s) core level binding energies for Alq3 after deposition of Co, with the largest shift observed in the N(1s) peak. After deposition of 24 Å of Co a N(1s) binding energy feature appeared ∼1.3 eV below the main peak observed for pristine Alq3 (see Fig. 3, middle part). Similar XPS spectra have been observed for other metals like Al, Ca, Mg, Na, K, and Li deposited onto Alq3. Second, the Co(2p) binding energy was typical of metallic, not oxidized Co (see Fig. 4). In fact, X-ray magnetic circular dichroism (XMCD) spectra at 300 K showed that a 2 nm film of Co on Alq3 was ferromagnetic (see inset of Fig. 4). Furthermore, as the Co thickness was increased the N, C, and O(1s) core levels decreased exponentially suggesting that the Co layer grew homogeneously across the Alq3 surface. In fact, a cross-section TEM image verified this conclusion. In general, Dediu and co-workers reached the same conclusions, although they found evidence for some Co diffusion into the Alq3 layer 26, 27. Recently the electronic and magnetic properties of Co doped Alq3 complex has been investigated by DFT calculations 28. The results indicate a charge transfer from the Co into a quinolate ligand that induces local magnetic moments in the complex.

thumbnail image

Figure 3. (online color at: www.pss-b.com) X-ray photoelectron spectra of Alq3 core levels before and after 2, 4, 6, 12, and 24 Å deposition of Co. Taken from Ref. 25.

Download figure to PowerPoint

thumbnail image

Figure 4. (online color at: www.pss-b.com) X-ray photoelectron spectra of the Co(2p3/2) core levels before and after a 2, 4, 6, 12, and 24 Å cobalt deposition on Alq3. The inset shows the XAS and XMCD spectra recorded at 300 K for a 20 Å deposition of cobalt on Alq3. Taken from Ref. 25.

Download figure to PowerPoint

Aristov et al. 29 studied the deposition of Fe and Co onto CuPc with near-edge X-ray absorption spectroscopy and XPS. After deposition of ∼35 Å Co or Fe onto CuPc the binding energy of the Cu(2p) peak shifted from ∼936 to ∼933 eV that suggests the Cu2+ oxidation state has been reduced to Cu0. Also the appearance of a low binding energy peak in the C(1s) region was observed after the ∼35 Å deposition of metal, which suggests metal–carbide formation. Finally they found the C(1s) core level were attenuated exponentially with increased Co or Fe deposition, which suggest very little diffusion of metal into the organic film.

The examples discussed above show that a chemical reaction occurs between the top metal contact and molecule. On one hand this may help prevent the diffusion of metal into the organic layer. On the other hand how the electronic structure of the metal/molecule complex influences spin-injection is far from understood.

5 Magnetoresistance in organic spin valves

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Spin-polarized tunneling
  5. 3 Electronic structure of thin film, amorphous organic semiconductors
  6. 4 Electronic structure at interfaces
  7. 5 Magnetoresistance in organic spin valves
  8. 6 The role of interfacial layers
  9. 7 Hole, electron, or ambipolar transport?
  10. 8 Conclusions
  11. Acknowledgements
  12. Biographical Information

Since the initial report of MR in organic spin-valves in 2004 using Alq3 as the semiconductor sandwiched between LSMO and Co electrodes 30, there have been many other reports of the spin valve effect using other FM and OSCs 31. Whilst we note that spin-polarization is the key concept here, and that ferromagnetism is not a requirement for spin polarization, in the context of this article MR refers to the injection, transport, and detection of spin-polarized carriers from one ferromagnetic electrode, through the OSC layer, and into a second ferromagnetic electrode. The temperature dependent electrical transport in such a structure should strongly depend on the thickness of the semiconductor layer. In the limit of a “thin” layer (i.e., no more than a few nm's) tunneling from one FM to the other should dominate. In contrast, in the limit of a “thick” organic layer typical of OLEDs (upwards of tens of nm) the transport is best described as hopping. A priori we expect very different device behaviors in these two regimes. First, in the tunneling limit there is the expectation that current/voltage (I/V) curves will be nonlinear due to the exponential dependence of tunneling current on the barrier width and height. Specifically the differential conductance, G(V) = dI/dV, should scale quadratically with voltage. Second if the area of the device is varied (and the thickness is held constant) the resistance-area product should remain constant. If so, this is a good indication of uniform conduction and no “hot-spots” in locally thin regions of the barrier. Third, the resistance should change very little with temperature. In fact one of the most accepted criteria to support tunneling is a slight increase (∼10–20%) in the junction resistance with decreasing temperature 32. If the resistance decreases with decreasing temperature then metallic shorts are likely between the electrodes. Finally, if one of the ferromagnetic electrodes is replaced with a superconductor then an energy gap (characteristic of the superconductor) should appear in a differential conductance measurement.

The first experiments to demonstrate tunneling MR through Alq3 were published by Santos et al. 33 and Xu et al. 34. Prior to 2007 a few reports of TMR with self-assembled monolayers were published 35. In this article the focus is primarily on Alq3 since the electrical transport characteristics of two-terminal devices with non-magnetic electrodes is well documented in the literature. Santos used NiFe and Co electrodes and observed ∼6% TMR at 300 K, which increased to 10% at 4 K. In addition to the TMR measurements Santos et al. measured the spin-polarization of the top electrode using the Tedrow–Meservey spin-polarized tunneling technique at 0.5 K. The spin-polarization was 30 and 38% for Fe and Ni80Fe20 top electrodes, respectively. In contrast, when no Al2O3 barrier was present on the superconducting Al electrode the polarization of the Co top electrode dropped to 6%. Xu et al. reported a −30% TMR at 11 K with a Co top electrode, but no MR at room temperature due to the loss of spin-polarization on the LSMO electrode.

Once the semiconductor layer is greater than ∼50 nm the I/V curves still should be nonlinear and a power law of the type I ∼ Vn where n > 2 is typical 36. The I/V curves should depend strongly on temperature because of the mechanism of charge transport in thermally–assisted hopping from one molecule to the next. In addition the I/V should be a function of the layer thickness since the mobility is electric field dependent. In the space-charge limit the current density should scale as d−336. Finally the magnetic field where the device resistance changes should coincide with the coercive fields of the electrodes measured independently by magnetometry. In the limit of thick Alq3 films there has been a wide range of reports 31. However, very few experiments have been done with a systematic characterization of the magnetic and electrical properties of many devices. A very comprehensive set of measurements on LSMO/LAO/Alq3/Fe spin-valves was reported by Yoo et al. 37. These authors varied the thickness of a rubrene layer between 5 and 50 nm and were able to distinguish between the tunneling and hopping regimes discussed above. In the thin limit, the TMR was about 12% at 10 K and monotonically decreased as the temperature was increased and above 250 K no TMR was observed. This trend in the MR with increasing temperature has been observed in most spin-valves using LSMO as the bottom electrode. When the rubrene layer was increases to 20 and 30 nm, the MR at 10 K decreased to ∼6 and ∼2%, respectively. For rubrene layers thicker than 40 nm no GMR was observed, which implies the spin-diffusion length in rubrene at 10 K is 10–20 nm. This value is consistent with the 13 nm spin-diffusion length for rubrene at 0.45 K, estimated by the decay of the spin-polarization using the Tedrow–Meservey relation 38.

6 The role of interfacial layers

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Spin-polarized tunneling
  5. 3 Electronic structure of thin film, amorphous organic semiconductors
  6. 4 Electronic structure at interfaces
  7. 5 Magnetoresistance in organic spin valves
  8. 6 The role of interfacial layers
  9. 7 Hole, electron, or ambipolar transport?
  10. 8 Conclusions
  11. Acknowledgements
  12. Biographical Information

Interestingly, there still remains some controversy in the community around how organic spin valve devices function. For example, there are reports of spin valves made from nominally the same materials and structure having quite different properties. For Fe/Alq3/Co based devices, there have been reports of no MR 39, positive MR 40, and negative MR 41. As yet, there is no consensus that explains this. However, it is becoming clear that the interface is very important in spin-based organic devices, and this could offer an answer to these differences. Recently, Barraud et al. 42 demonstrated that an Alq3 based nanojunction has a positive tunneling MR of 300%, at 2 K, whereas they point out that a negative MR has been observed for larger and thicker barriers. They explain this discrepancy by using a spin-dependent transport model based on an interfacial spin hybridization induced polarized state (SHIPS). This hybridization is accompanied by a spin-dependent broadening and a spin-dependent energy shift of the density of states in the molecular monolayer. The formation of SHIPS in the first molecular layer at the electrode interface can therefore lead to an increase of the effective spin polarization of the electrodes or even a change in sign. However, it is essential to note that the sign inversion in their model is only possible locally for the molecular states brought close enough to the electrode's Fermi level, the very same states that in their model are also dominating the injection. They conclude that the local character of this SHIPS-induced polarization inversion means it should not be observed in their point-like experiments, whereas since these SHIPS states are responsible for the majority of the current, it can be observed in large area devices. However, this model may not be generally applicable, as recent large-area device measurements with Low Energy Muon Spin Rotation (LE-µSR) indicate that the spin-polarized current is not dominated by hotspots, but is present over a large area within the device 43. This is because the LE-µSR technique measures the areal average over the size of the muon beam-spot 44, which is shown in Fig. 5 to be roughly circular with a radius of a few mm. As is explicitly stated in Barraud's manuscript, the SHIPS states can only occur at extremely rare locations at the interface because of the amorphous structure of the organic layer and the large anisotropy of the OSC-FM coupling. It was recently estimated that approximately only one so-called “hot spot” exists per 400 µm245. If these hot-spots were present, then the muons would sample the large fraction of the device that is carrying little current, with the signal from the fraction that is responsible for carrying the current being lost in the error bars.

thumbnail image

Figure 5. (online color at: www.pss-b.com) (a) The muon beam profile of the LE-µSR spectrometer, showing how it matches up to experiments on large area organic spin valves 43, 56. The data was obtained by putting a 2D pixelated detector at the sample position 44. (b) A color contour plot of the muon beam profile, showing most of the muons stop within circle with a few millimeter radius. This overlaps well with the active area of the spin valve.

Download figure to PowerPoint

Clearly, it is also possible to change the FM-OSC coupling by the use of thin interfacial layers. The effect of a tunnel barrier between the metal and the OSC on the functionality of a spin-valve device has proven to be critical for spin-polarized tunneling 33, 37, 38, 46–49. Dediu and coworkers have demonstrated that the quality of the metal-organic interfaces can be improved by inserting an Al2O3 layer on top of the organic material, prior to growing the top ferromagnetic electrode 50. Furthermore, the sign of the MR of organic-based devices has also been changed by adding LiF to the interfaces 51–53. Interestingly, in all the cases where the device composition was LSMO/Alq3/X/Co where X is either absent, LiF or Al2O3, the MR was found negative. This contrasts the positive MR for devices with the composition LSMO/Y/Co where Y is a conducting polymer in place of Alq353–55.

Recent work by Schulz et al. 56 showed using LE-µSR that it is possible to engineer interfaces in large area-spin valve devices. They measured large area spin valves with a structure of NiFe/Alq3/FeCo and NiFe/LiF/Alq3/FeCo. It was shown that the composition is indeed a key player for spin-polarised charge carrier interface propagation. The insertion of a very thin LiF layer between the NiFe cathode and Alq3 reversed the sign of the spin-polarization of the charge carriers. This is explained by electrical dipoles induced by the additional LiF layer, which cause a vacuum level shift δ that is accompanied by a new alignment of the HOMO level of the organic layer with the spin-dependent sub-bands of the ferromagnetic layer, as is schematically shown in Fig. 6. The different energy level alignment switches the dominant spin sub-band for interface propagation. As holes are extracted, the probability of one particular spin state dominating the extraction is related to the spin density of states at the extraction energy. For the case of a device without LiF, where there is no vacuum level shift, this results in a minority spin electron accumulation at the cathode. Whereas for the device with LiF, where there is a vacuum level shift δ, there is an electron majority spin accumulation, as spin majority holes are extracted more efficiently (Fig. 6b and c).

thumbnail image

Figure 6. (online color at: www.pss-b.com) (a) A vacuum level shift, δ, can change the dominant spin sub-band. (b) If the dominant sub-band is spin minority, then spin minority holes can be extracted efficiently and there is an accumulation of spin minority electrons in the semiconductor. (c) On the other hand, if a vacuum level shift (due to LiF) changes the dominant sub-band to spin majority, then there is an accumulation of spin majority electrons in the organic layer. Adapted from Ref. 55.

Download figure to PowerPoint

It is clear that this mechanism, where different spin-dependent density of states are accessed depending on interface states and energy level alignment, can explain the contradictory results in the literature. In addition to any sign reversal induced by electric dipole moments at the interface, it is also possible that there is a modification to the net propagated spin polarisation without reversing its sign, hence leading to a change of the size of MR or indeed the absence of it. This behavior could be related to the thickness of the polar layer 57, or indeed a production process that could lead to a different metal-organic interface 23.

7 Hole, electron, or ambipolar transport?

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Spin-polarized tunneling
  5. 3 Electronic structure of thin film, amorphous organic semiconductors
  6. 4 Electronic structure at interfaces
  7. 5 Magnetoresistance in organic spin valves
  8. 6 The role of interfacial layers
  9. 7 Hole, electron, or ambipolar transport?
  10. 8 Conclusions
  11. Acknowledgements
  12. Biographical Information

There is a debate within the community about whether spin is transported via electrons in the LUMO or holes in the HOMO, and this is still an open question. Some authors claim that spin transport in OSCs is via electrons 23, 27, 58. The main argument for electron transport appears to be that in Alq3 hole transport is significantly worse than electron transport 51. There is also evidence that the energy level alignment of Co/Alq3 interfaces have a larger barrier for hole injection 23.

Other authors support hole transport based on the idea that the ferromagnetic electrodes in organic spin valves are inefficient electron injectors, especially at the low voltages at which the devices operate, due to their very high work functions 26, 54, 56, 59. There is evidence (based for example on internal photoemission measurements of Schottky barriers) that the size of the barriers for electron and hole injection scales with the electrode work function at metal/polymer interfaces 60. X-ray and UPS studies on the interface between LSMO and α-6T or CuPc also proved the hole-injection barrier to be significantly lower than the one for electron injection 61. Holes injected from the anode into the HOMO of the OSC have also been claimed in a LSMO/Alq3/Co spin valve at low bias voltages based on an estimate band Diagram 30. In this work the authors also noted that electroluminescence was observed in an ITO/Alq3/Co OLED at very high voltages (∼20 V) compared with the ∼2.5 mV operating voltage used for the optimum MR. This demonstrates that electron injection can occur from transition metal cathodes with a high work function cathodes with a large applied electric field, but at the very low voltages used for the MR measurements the current will be essentially hole only. A similar result was obtained in an ITO/TPD/Au device, where light emission was observed at voltages greater than 12 V but with a device efficiency of only 10−11%, compared to 0.1% with an optimised electron injection layer 62. This highlights that even gold, with a work function of 5.1 eV and which is commonly thought of as an excellent anode material, can inject electrons but only under extreme applied voltages. 30. The idea that hole transport takes place in organic spin valves because of the inefficiency of high work function ferromagnets as electron injectors is further supported by an in-depth experimental study on the efficiency of OLEDs 63. The relative luminance of ITO/Alq3/cathode OLED devices (at a constant current of 50 mA/cm2) was measured for Al, Ag, Zn, or Cu cathodes, whose work functions are 4.30, 4.32, 4.47, and 4.70 eV, respectively. The light output is significantly lower than that of devices with a cathode like Mg, Yb, Li, Ca, or Sm whose work functions are lower than 3.7 eV 63. The relative luminance is shown in Fig. 7. However, it has to be noted that the authors also report that the usage of very low work function metals does not give rise to increased device efficiency 63, 64.

thumbnail image

Figure 7. Relative luminescence of a series of ITO/Alq3/cathode devices, at a drive current of 50 mA/cm2, showing that luminescence drops off as work function increases. Adapted from Ref. 63.

Download figure to PowerPoint

To return to the question of whether the NiFe cathode is an efficient electron injector or not 56, we note that its work function is close to 5 eV. This suggests that NiFe is a poor electron injector. To confirm this, we have recently performed a similar experiment to Stössel et al. 63, on an ITO/Alq3/NiFe OLED with NiFe as the cathode. No significant luminescence was observed 65, confirming the suggestion that NiFe is a poor electron injector.

Finally, there is a clear debate on whether the observed MR in organic spin valves is due to transport within the organic layers or tunneling through thin regions 34, 39, 66. For the same arguments as used in Section 6 about “hot spot” transport, the LE-µSR measurements appear to show the former is true 43, 56.

8 Conclusions

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Spin-polarized tunneling
  5. 3 Electronic structure of thin film, amorphous organic semiconductors
  6. 4 Electronic structure at interfaces
  7. 5 Magnetoresistance in organic spin valves
  8. 6 The role of interfacial layers
  9. 7 Hole, electron, or ambipolar transport?
  10. 8 Conclusions
  11. Acknowledgements
  12. Biographical Information

A few general conclusions can be made about the role of interfaces in organic spin valves. First, there is strong evidence that inorganic barriers like amorphous Al2O3 improve the performance of magnetic tunnel junctions with thin OSC layers. Second the deposition of FMs onto OSCs tends to be reactive, although the influence of this reacted interface on spin-injection efficiency has not been quantified. In contrast, there is much more variability in the devices made with thick OSC layers. Perhaps it is the slight variation in growth conditions (e.g., base pressure, growth rate, and electrode thickness) and sample size/geometry that has caused the different observations in the literature, or exposing an incomplete device to air. Future studies should focus on a thorough electronic/magnetic characterization of the interfaces as well as spin-valve devices, prepared without breaking vacuum, to better understand the factors that govern MR phenomena in organic spin-valves.

Biographical Information

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Spin-polarized tunneling
  5. 3 Electronic structure of thin film, amorphous organic semiconductors
  6. 4 Electronic structure at interfaces
  7. 5 Magnetoresistance in organic spin valves
  8. 6 The role of interfacial layers
  9. 7 Hole, electron, or ambipolar transport?
  10. 8 Conclusions
  11. Acknowledgements
  12. Biographical Information

Alan J. Drew was educated at University of Birmingham (B.Sc. Physics 1999, M. Phil. Materials Engineering 2000) and the University of St. Andrews (PhD Physics 2004). He was then awarded a Research Fellowship of the Royal Commission for the Exhibition of 1851 (2004–2006) split between the University of St. Andrews, Paul Scherrer Institute/ETHZ Switzerland and the Rutherford Appleton Laboratory. He was then an Oberassistant at the University of Fribourg Switzerland (2006–2008) before returning to the UK as a Leverhulme Research Fellow and Lecturer at Queen Mary University of London (from 2008 to present day).

Thumbnail image of