Electron injection into organic semiconductor devices from high work function cathodes

Abstract

We show that polymer light-emitting diodes with high work-function cathodes and conjugated polyelectrolyte injection/transport layers exhibit excellent efficiencies despite large electron-injection barriers. Correlation of device response times with structure provides evidence that the electron-injection mechanism involves redistribution of the ions within the polyelectrolyte electron-transport layer and hole accumulation at the interface between the emissive and electron-transport layers. Both processes lead to screening of the internal electric field and a lowering of the electron-injection barrier. The hole and electron currents are therefore diffusion currents rather than drift currents. The response time and the device performance are influenced by the type of counterion used.

Light-emitting diodes and thin-film transistors fabricated with semiconducting (conjugated) polymers are examples of an emerging technology with potential impact in low-cost displays and solid-state lighting (1). Balanced charge injection (holes into the π-band from the anode and electrons into the π*-band from the cathode) is a basic requirement of high-efficiency polymer light-emitting diodes (PLEDs) (2). In the absence of interfacial effects, the barrier for electron injection is determined by the difference between the energy of the bottom of the π*-band (lowest occupied molecular orbit, LUMO) of the polymer and the Fermi energy of the metal used as the cathode; similarly, the barrier for hole injection is determined by the difference between the energy of the top of the π-band (highest occupied molecular orbit, HOMO) and the Fermi energy of the anode. Because the charge injection is described (in first approximation) by a combination of Fowler–Nordheim tunneling and thermionic emission mechanisms, these barriers limit the device performance (3). Large and unequal barriers reduce power and optical output efficiencies by increasing the turn-on voltages and creating unbalanced injection of charge carriers. Electron injection continues to be an important problem because low work-function metals such as Ca or Ba, with Fermi energies that match the π*-bands of organic semiconductors, are unstable and decrease device operational lifetimes.

Inserting injection/transport layers (TLs) between the emissive layer (EL) and the electrodes can improve charge injection into organic LEDs via different mechanisms. For example, a favorable dipole can be introduced that shifts the vacuum level at the electrode/TL interface (4). Charge-carrier blocking and accumulation at the EL/TL interface can also lead to improved injection by redistributing the field toward the TL and thereby reducing the charge tunneling distance (5–8). Efficient electron injection from stable metals into PLEDs incorporating a conjugated polyelectrolyte (CPE) electron TL (ETL) was recently demonstrated. CPE materials are characterized by a π-delocalized backbone with pendant groups bearing ionic functionalities (9). The efficient injection was initially attributed to the introduction of a permanent interfacial dipole (10–15). However, recent studies show that the current density varies on a multisecond time scale, implying that ion migration within the CPE layer is involved in the electron-injection mechanism (16).

We present here a model and supporting experimental evidence for electron injection from high work function metals using a conjugated polyelectrolyte ETL. The combination of ion rearrangement and associated screening within the ETL, together with interfacial charge accumulation (hole blocking) at the EL/ETL interface, redistributes the internal field to enable efficient electron injection. Time-dependent measurements of current density (J) and luminance (L) as a function of the PLED architecture confirm the basic features of the model.

Results and Discussion

The device test structure is shown in Fig. 1a. Poly[2-methoxy-5-(2-ethylhexyloxy)-1,4-phenylenevinylene] (MEH-PPV) is used for the EL layer. Poly[9,9′-bis[6″-(N,N,N-trimethylammonium) hexyl]fluorene-alt-co-phenylene] (17) with tetrakis(imidazolyl) borate counterions (18) (PFN-BIm₄) is used for the ETL (hole blocking) layer (Fig. 1b). Multilayer fabrication is facilitated by the orthogonal solubility of MEH-PPV and PFN-BIm₄ (19, 20). Gold or aluminum was used as the cathode.

Fig. 1.

Schematic of structures. (a) Device structure. (b) Molecular structures of MEH-PPV and PFN-BIm_4. (c) Energy-level diagram.

The energy-level diagram of the various materials used in these devices is shown in Fig. 1c (3, 6, 14, 21–24). The close match between the work function of ITO/PEDOT and MEH-PPV results in a negligible hole-injection barrier. In the absence of an interfacial shift in vacuum level, the mismatch between the work functions of Al or Au and the LUMO of MEH-PPV results in electron barriers of 1.5 eV or 2.3 eV, respectively. In devices with a PFN-BIm₄ ETL, the electron-injection barriers are 3 eV for Au and 2.2 eV for Al; too large for efficient performance in the absence of additional effects.

Despite these obviously large barriers for electron injection, PLEDs containing the PFN-BIm₄ ETL layer exhibit remarkable performance, as demonstrated in Fig. 2. The insertion of a 12-nm PFN-BIm₄ ETL into PLEDs with Al cathodes reduces the turn-on-voltage (V_turn-on = V when luminance, L, = 0.1 cd/m²) from 4 V to 2 V and increases the maximum efficiency from 0.006 cd/A at 6.5 V to 1.3 cd/A at 4.4 V (Fig. 2a). A 0.05-V/sec scan rate was used. The maximum brightness increases from L = 29 cd/m² at 10.9 V to L = 8,000 cd/m² at 6.1 V. Significantly, the PFN-BIm₄/Al device shows a higher maximum efficiency and higher maximum L than a device without the ETL but with the low work function metal Ba as the cathode (0.9 cd/A at 2.9 V and 4,900 cd/m² at 6.5 V). Fig. 2a also shows that insertion of PFN-BIm₄ also greatly improves electron injection from Au. The ITO/PEDOT/MEH-PPV/Au device shows no luminance; hole current thus dominates. With PFN-BIm₄/Au, electron injection is improved with V_turn-on = 3.2 V, a maximum L = 2,000 cd/m² at 6.7 V and a maximum efficiency of 0.24 cd/A at 6.9 V.

Fig. 2.

Electrical characteristics of devices. (a) J-V and L-V curves for devices with structures ITO/PEDOT/MEH-PPV/Ba/Al (red), ITO/PEDOT/MEH-PPV/Al (blue), ITO/PEDOT/MEH-PPV/Au (black), ITO/PEDOT/MEH-PPV/PFN-BIm₄/Al (green), and ITO/PEDOT/MEH-PPV/PFN-BIm₄/Au (orange). The PFN-BIm₄ thickness is 12 nm. (b) Time response of J and L for devices with structures ITO/PEDOT/MEH-PPV/Ba/Al (red), ITO/PEDOT/MEH-PPV/Al (blue), ITO/PEDOT/MEH-PPV/Au (black), ITO/PEDOT/MEH-PPV/PFN-BIm₄/Al (green), and ITO/PEDOT/MEH-PPV/PFN-BIm₄/Au (orange). The PFN-BIm₄ thickness is 12 nm, and applied bias is 5 V. (Inset) Repeatability of ion redistribution shown for ITO/PEDOT/MEH-PPV/PFN-BIm₄/Al device with 18-nm PFN-BIm₄ thickness and 5-V applied bias with first scan (red) and second scan (blue) of the same device 17 h later.

Our model for the improved charge injection using CPE ETLs involves a combination of hole blocking at the EL/ETL interface and ionic redistribution across the ETL. To illustrate this concept, Fig. 3 shows the electric fields in an ITO/PEDOT/MEH-PPV(80 nm)/PFN-BIm₄(20 nm)/Al device with 5-V applied bias. As holes are injected, they accumulate at the EL/ETL interface and screen the electric field in the EL, as shown in the transition from Fig. 3a to Fig. 3b. The electric field is thereby redistributed so that voltage drops predominantly across the ETL, producing a large internal field in that layer. This large internal field leads to anion motion toward the EL/ETL interface, leaving an excess of cations adjacent to the cathode. The redistribution of mobile ions screens the field within the ETL. Experimental evidence discussed subsequently shows that anions do not penetrate into the MEH-PPV layer. The electric field across the ETL is therefore redistributed into two double layers at the EL/ETL interface and at the ETL/cathode interface, as in Fig. 3c. The end result is that the electric fields in both layers are screened with nearly all of the applied voltage localized across the double layers. Consequently, although the electron-injection barrier (ΔE_e) remains the same, efficient electron injection occurs via tunneling through the ultrathin double layer, and radiative recombination of electrons and holes occurs in the MEH-PPV near the EL/ETL interface. Because the internal electric field is screened and nearly zero in both layers, the hole current and the electron current are diffusion currents rather than drift currents.

Fig. 3.

Schematic response of an ITO/PEDOT/MEH-PPV/PFN-BIm₄/Al device under 5-V applied bias. (a) The electric field (slope of the energy levels) is evenly distributed across the device. (b) Holes (open circles) accumulate at the MEH-PPV/PFN-BIm₄ interface, screening the electric field to the PFN-BIm₄ layer. (c) Ions (charges within the PFN-BIm₄ layer) then redistribute to screen the electric field to the two ETL interfaces.

Fig. 2b shows the time response of current density, J, and L of PLEDs with and without PFN-BIm₄ at an applied bias of 5 V. At long times, the steady-state currents follow the same trends observed in Fig. 2a. Looking at the response times, defined as the time for J to reach 50% of its maximum value, one observes a substantially slower response when PFN-BIm₄ is present. For example, the response times of PFN-BIm₄ (12 nm)/Al and PFN-BIm₄ (12 nm)/Au devices are 2.9 and 28 sec, respectively. These temporal responses are much longer than in conventional PLEDs, which are typically on the nanosecond or microsecond time scales (25, 26) and are consistent with ion redistribution within the PFN-BIm₄ layer. Given that hole injection is facile, the correlated increase in L and J indicates that the increase in current is due to increased electron injection. Ion redistribution is not permanent, as indicated by the Inset of Fig. 2b, which compares the J responses of an ITO/PEDOT/MEH-PPV/PFN-BIm₄ (18 nm)/Al after allowing the return to nonbiased equilibrium for 17 h. The relaxation is not immediate and also depends on the ion mobility.

The temporal response due to ion redistribution is similar to that seen in polymer light-emitting electrochemical cells (LECs), which incorporate a blend of mobile ions in an ion-transporting polymer into the conjugated polymer EL (27, 28). When a bias less than the band gap of the polymer is applied in an LEC, the ions migrate, screening the electric field and creating double layers at the polymer/electrode interfaces. At voltages greater than the polymer band gap, in situ doping occurs, and a p–i–n junction is formed (27–29), although some debate remains (30). Over potential effects caused by phase separation with the ions in the ion-transport polymer (rather than in the luminescent semiconducting polymer) can inhibit the onset of redox doping. Detailed experimental evidence has demonstrated the formation of a p–i–n junction (31–35). However, unlike an LEC, the ion redistribution in the multilayer devices shown in Fig. 1a causes screening of the internal field in the ETL. Because there is no pathway for ionic transport across the entire device, redox doping and p–i–n junction formation cannot occur. That there is a fundamental difference in operation is verified by the better performance of the Al cathode relative to the Au counterpart in Fig. 2. If the mobile anions traveled across the EL layer, the device would behave like an LEC, which is insensitive to electrode work function. Confining the mobile ions to a hole-blocking ETL has multiple potential advantageous over LECs, including a reduced distance the ions must redistribute over (28), improved stability, and recombination moved away from the electrodes (34).

The time response of the ITO/PEDOT/MEH-PPV/Al device is shorter than the ITO/PEDOT/MEH-PPV/Au device because the Al cathode has better electron injection than the Au cathode before ion redistribution (Fig. 2b). Similar dependence of electron-injection barriers on time response can be seen in LECs (36, 37).

Hole build up at the interface, as shown in Fig. 3b, occurs before the first measurement in Fig. 2b. Electroluminescence in conventional multilayer organic LEDs is slightly delayed, but the response time is still in the microsecond time scale (38). Therefore, Fig. 2b records the time dependence of J and L as the ions redistribute the field in the ETL and after hole accumulation at the EL/ETL interface. The low initial J and L values before ionic redistribution for the devices with a PFN-BIm₄ are significant, because they rule out an important contribution by a permanent and static interfacial dipole at the cathode interface.

Extrapolation of the current characteristics to very short times indicates that before ionic redistribution, the current densities are not only lower than their steady-state values but are also lower than for devices without the PFN-BIm₄ layer. This feature is more readily observed when using Au. Furthermore, ITO/PEDOT/MEH-PPV/Al and ITO/PEDOT/MEH-PPV/Au devices have lower luminances than their ITO/PEDOT/MEH-PPV/PFN-BIm₄/Al and ITO/PEDOT/MEH-PPV/PFN-BIm₄/Al counterparts, despite their higher initial current densities. These observations provide evidence that PFN-BIm₄ functions as a hole-blocking layer that improves charge recombination by accumulating holes at the EL/ETL interface.

The critical role of hole accumulation in lowering the electron-injection barrier is shown in Fig. 4b, which shows the field redistribution expected by ion migration alone. Comparison with Fig. 4a shows a substantial difference in the height and thickness of the electron injection barrier. In Fig. 4b, which assumes no hole accumulation, the effective electron-injection barrier (ΔE_s) is reduced by the ion redistribution (from the original electron-injection barrier, ΔE_e in Fig. 3a), described by:

where d_ETL is the thickness of the ETL, d_total is the total device thickness (ETL+EL), V_app is the applied voltage, and V_bi is the built in voltage arising from the difference in work function of the two electrodes. The much smaller effective electron injection barrier (ΔE_eff) seen in Fig. 4a is described by:

where ψ_ha is the band bending in the hole accumulation layer. Both Eqs. 1 and 2 assume that the double layer is sufficiently thin to allow efficient tunneling, and the voltage drop across the ETL is equally distributed between the two interfacial double layers.

Fig. 4.

Schematic of the band diagram in a device under 5-V applied bias with ion screening of the electric field in the ETL. (a) Device with hole accumulation at the EL/ETL interface. (b) Device without hole accumulation.

Electroluminance spectra show emission only from the EL layer [similar results have been seen previously for devices containing CPE ETLs (10, 39, 40)]. Therefore, holes do not tunnel into the PFN-BIm₄ layer; all recombination occurs in the EL near the EL/ETL interface. The image charge potential, double-layer thicknesses, and injection-tunneling distances at the two ETL interfaces are not symmetric or holes would be expected to be injected from the EL into the ETL layer. The asymmetry results from the fundamental differences between the two interfaces; there is a metal cathode on one side of the ETL and semiconducting MEH-PPV of the other side of the ETL. The Thomas–Fermi screening length in the metal (< 0.1 nm) is much smaller than the width of the hole-accumulation layer (space charge layer) in the MEH-PPV, as sketched in Fig. 3c (41–43). Differences in the spatial distribution of excess ions can also arise because anions are much more mobile than the polymer-bound cationic centers.

Fig. 5a shows the influence of PFN-BIm₄ thickness on the response times. For ITO/PEDOT/MEH-PPV/PFN-BIm₄/Al devices at 5 V, the response times as a function of thickness are 2.9 sec (12 nm), 14 sec (18 nm), and 80 sec (22 nm). With 3-nm PFN-BIm₄, the J of the first measurement at 80 ms (130 mA/cm²) is greater than 50% of the maximum value, and with 45-nm PFN-BIm₄ no maximum value was reached after 300 sec. Differences are greater at 4 V, with the response time changing from 4.4 to 25 sec as the thickness is changed from 12 to 18 nm. The differences in response times for ITO/PEDOT/MEH-PPV/PFN-BIm₄/Au devices at 5 V are larger than for ITO/PEDOT/MEH-PPV/PFN-BIm₄/Al: 2.8 sec/3 nm, 28 sec/12 nm and 70 sec/18 nm. The response time must increase with thickness because ions must traverse farther to completely screen the electric field. Additionally, at a given bias, the electric field decreases with increasing thickness, implying a weaker driving force for ion distribution and therefore a longer time to approach the steady-state condition in which the internal field is screened to zero. It is relevant here that the response times of LECs (28, 44) also have a thickness dependence because the distance that ions must move to screen the field is increased.

Fig. 5.

Layer-thickness dependence of electrical characteristics. (a) Time response of J and L for ITO/PEDOT/MEH-PPV/PFN-BIm₄/Al devices with 5-V applied bias and PFN-BIm₄ thickness of 3 nm (red), 12 nm (blue), 18 nm (black), 22 nm (green), and 45 nm (orange). (b) J-V curves for the same devices. (c) J-V curves for ITO/PEDOT/MEH-PPV/PFN-BIm₄/Al devices with 12-nm PFN-BIm₄ thickness and MEH-PPV thickness of 80 nm (red), 88 nm (blue), 100 nm (black), 124 nm (green), and 140 nm (orange). Insets show J vs. 1/d_ETL in b and J vs. 1/d_EL in c at 4 V. The thickness error bars correspond to variations in roughness and thickness across the samples and from device to device.

The response time was not appreciably altered under 5 V when the MEH-PPV thickness was varied from 80 to 140 nm while keeping the PFN-BIm₄ layer at 12 nm. Ions thus do not traverse the EL layer. These observations are also consistent with hole accumulation redistributing the electric field from the EL to the ETL, otherwise the electric field across the ETL would decrease with increasing MEH-PPV thickness with a concomitant increase in response time.

From Fig. 5a, one observes that the steady-state current density is also reduced as the thickness of the PFN-BIm₄ layer increases. Fig. 5b shows the J-V curves as the PFN-BIm₄ thickness is increased, and Fig. 5c shows the J-V curves as the MEH-PPV thickness is increased for devices with Al cathodes. The changes in current are not due to changes in the Fowler–Nordheim tunneling distances because the curves do not overlap when plotted against electric field rather than applied voltage. The current is clearly more sensitive to the ETL thickness than the MEH-PPV thickness. Devices with Au cathodes exhibit even a larger dependence on ETL thickness.

Double logarithmic plots of the current density against the inverse of the ETL thickness (d_ETL) and the EL (d_EL) are provided as Insets in Fig. 5 b and c. In both cases, as well as for devices with Au cathode, the slope is ≈3. Therefore:

This thickness dependence is typical for space charge-limited current (SCLC). SCLC occurs when charge accumulation restricts further charge injection, and although diffusion has often been ignored in works on SCLC, it has the same thickness dependence when included (45, 46). Interestingly, that the 1/d³ dependence for J includes the ETL or EL layers and not total device thickness indicates that the EL layer carries only holes and that the ETL carries only electrons (47, 48). Recombination must therefore take place near the EL/ETL interface. As in an LEC (28, 44), the response times of these devices also depend on the applied voltage (16).

Because ion motion has a major role in the model proposed here, it was anticipated that the choice of counterion would influence device performance. We thus prepared PFN-F, where the large BIm₄⁻ was exchanged for fluoride (F⁻) and used this new material to fabricate similar devices. Fig. 6 compares the response of ITO/PEDOT/MEH-PPV/PFN-BIm₄ (18 nm)/Al to ITO/PEDOT/MEH-PPV/PFN-F (17 nm)/Al at a 4-V applied bias. The response time of the PFN-F device (12 sec) is shorter than the PFN-BIm₄ device (25 sec). Even with a PFN-F thickness of 25 nm, the response time is still only 21 sec. The shorter response time with PFN-F may be due to the smaller size of the F⁻ relative to BIm₄⁻. It can also be seen from Fig. 6 that J and L values are higher for PFN-F than PFN-BIm₄, consistent with the fact that different counterions will change interchain packing (49). The smaller fluoride counterion may lead to improved interchain packing and electron mobility in the ETL, increasing both J and L.

Fig. 6.

Time response for different counterions. Time response of J and L for an ITO/PEDOT/MEH-PPV/PFN-BIm₄(18 nm)/Al device (red) and an ITO/PEDOT/MEH-PPV/PFN-F(17 nm)/Al device (blue) with a 4-V applied bias.

In an earlier publication, we showed improved and efficient electron injection due to ion movement with a similar device but with the larger band gap poly(9,9-dioctyl-fluorene) (PFO) as the EL (16). For these devices, the situation is slightly more complicated. Because of the larger band gap, the PFO/PEDOT interface also has a large hole-injection barrier. However, it has been shown for PFO devices without ETLs that electrons can be trapped at the PFO/PEDOT interface, screening the field through the device and increasing the field at the anode, enhancing hole injection (5, 50). Another complicating factor is that, unlike the MEH-PPV/PFN-BIm₄ interface, the PFO/PFN-BIm₄ interface would be expected to have a small and potentially ineffective hole-blocking energy barrier. Nonetheless, interfacial traps can serve to screen the field, much like at the PFO/PEDOT interface (51, 52).

Multilayer PLEDs with conjugated polyelectrolyte interfaces are attractive because of the implied low fabrication cost and efficient electron injection from stable metal cathodes. However, the mechanism has not previously been well understood. Here, we provide a more complete model showing that the improved electron injection results from the combined effects of hole accumulation at the EL/ETL interface and electric field screening within the ETL by the mobile counterions. Device response time can be improved by reducing the thickness of the ETL and using a smaller, more mobile counterion, an important consideration for PLEDs targeted for video applications.

Materials and Methods

Devices with and without ETLs were fabricated by spin casting a toluene solution of MEH-PPV onto an ITO electrode that was wet-cleaned, UV/ozone treated, and then passivated with 80 nm of poly(3,4-ethylenedioxythiophene)-poly(styrenesulfonate) (PEDOT:PSS, Baytron P VP AL 4083). The resulting MEH-PPV thickness was ≈80 nm, as determined by atomic-force microscopy (AFM) measurements. After keeping the device under vacuum at 10⁻⁴ torr for 2 h to remove residual solvents, a layer of PFN-BIm₄ was deposited from methanol to serve as the ETL. This counterion was previously shown to give excellent performance in related PLEDs (18). The synthesis of PFN-BIm₄ is described in refs. 17 and 18. The fluoride anion exchange was done according to the previous report (49). The PFN-BIm₄ thickness was controlled by varying the spin rate and concentration and was measured by using AFM trenches cut in both single PFN-BIm₄ layers on ITO and comparison of the multilayer device thicknesses with a control PEDOT/MEH-PPV layer thickness. Before applying the cathode by evaporation at 10⁻⁶ torr, the devices were kept under vacuum at 10⁻⁴ torr for 12 h. The Al or Au cathodes were deposited at 0.5 and 0.1 Å/sec, respectively, for 10 nm and then deposited at a rate of 1 Å/sec for a total thickness of ≈100 nm. The Ba cathodes were deposited at 0.4 Å/sec for 5 nm and then capped with 100 nm of Al. Testing was done with a Keithley 2602 source-measure unit and a photodiode connected to a HP 3278A multimeter. For J-V and L-V curves, a scan rates of 0.05 V/sec was used. All fabrication and testing were carried out inside a N₂ atmosphere dry box.