3.1.

Electromodulation Measurements

Figure 1 shows the EM spectrum of a single layer Alq₃ with the structure ITO/(100 nm)Alq₃/Al. The dc bias was −2 V and the reference amplitude was 200 mV rms. The spectrum is normalized to the sample transmission and closely approximates the change in optical density due to the electric field. The experimental data are shown as symbols and the solid line was calculated from a multi-Gaussian fit to the experimental data before normalization. The spectrum shows a classic oscillatory behavior with a maxima at λ = 450 nm. For the purposes of this work, the intensity of the EA signal is proportional to the average electric field in the Alq₃ layer. The EA signal had a linear dependence on the applied bias with a zero level crossing at 2.8 V, the built-in potential (V_bi). This is much higher that that expected from the work function difference between ITO (5 eV) and aluminum (4.1), indicating the presence of interface dipoles and/or significant chemistry at the ITO/Alq₃ and Alq₃/Al interfaces.³³ This was primarily a test structure to identify the position of the EA peak of Alq₃. The EA spectrum of a single layer α-NPD device is shown in Fig. 1 for comparison.²⁹ This work was performed by Rohlfing on a device with a 100 nm thick layer of α-NPD and under a field of 21 kV/cm. The EA spectrum of α-NPD has a maxima at λ = 410 nm and does not contribute for λ > 450 nm.

Fig. 1

EM spectrum of (a) ITO/(100 nm)Alq₃/Al (F = X kV/cm) and (b) ITO/(100 nm)a-NPD/Al devices (F = 21 kV/cm). Symbols show the experimental data in (a) and the line shows a multi-Gaussian fit to ΔT. The data for Fig. 1 was measured by Rohlfing (Ref. 24).

Figure 2 shows the EM spectrum of a bi-layer device with a polymer HIL, measured with the dc bias set to –0.7 V and the reference amplitude to 0.5 V rms. The spectrum has a maxima at λ = 460 nm and minima at λ ≈ 400 nm and λ = 520 nm. The maxima is attributed to EA of Alq₃ as it matches the peak in the single layer Alq₃ device. The minima at λ ≈ 400 nm lies at a point where there is a minima from Alq₃ as well as a maxima from α-NPD, assuming that the electric field has the same polarity in both layers. In devices measured by Martin, contributions from α-NPD were the dominant feature at this wavelength, resulting in a maxima at 400 nm rather than a minima.³⁰ This was largely due to the EA signal of α-NPD being 8 times larger than that of Alq₃. In contrast, the ratio of these two peaks measured here [Figs. 1, 2] is fairly similar for both structures, roughly 4:3. This would suggest that even in reverse bias, the electric field strength in the α-NPD layer is roughly an order of magnitude weaker than that in the Alq₃ layer; an effect that can only be attributed to the incorporation of an HIL. The minima at λ = 520 nm was observed by neither Rohlfing²⁹ nor Martin³⁰ and thus is a direct consequence of incorporation of a hole injection layer.

Fig. 2

EM spectrum of an ITO/HIL/(40 nm)α-NPD/(40 nm)Alq₃/LiF/Al device, measured at –0.7 V dc bias and with a 0.5 V rms reference bias modulated at 1 kHz. In-phase: filled symbols. Quadrature: open symbols. The line is shown as a guide for the eye.

Figure 3 shows the dependence of the Alq₃ signal, measured at a fixed wavelength of 460 nm as the dc bias was swept from –5 to +2 V with a reference bias amplitude of 0.5 V rms. The signal depends linearly upon the applied bias between – and –1 V and then abruptly changes slope. The zero-level crossing occurs at 0.3 V forward bias and the signal increases magnitude, reaching a maximum negative amplitude at ∼2 V forward bias. This behavior lies in sharp contrast to that of the single layer device or bi-layer devices previously studied.^{22, 24} The change in slope of Alq₃ EA was correlated with an increase in the intensity of the EM band, shown in Figure 4. The greater sample transmission below the optical gap permitted a lower reference bias, 0.2 V rms. The excited state absorption ESA signal is roughly independent of the dc bias up to 0.2 V reverse bias and then sharply increases, reaching a maximum value at 3 V forward bias. In addition to a sharp increase in the magnitude of the EM signal at 0 V, there is also a slight phase shift as can be seen by a non-zero quadrature signal.

Fig. 3

Dependence of the (a) Alq₃ EA signal (λ = 460 nm) and (b) EM signal (λ = 520 nm) on the applied bias. The reference voltage was 0.2 V rms, modulated at 1 kHz for both measurements.

Fig. 4

EM spectra of the bi-layer OLED measured at (a) 1.7 V dc bias with a reference bias of 0.5 V rms and (b) 3 V dc bias with a reference bias of 200 mV rms. The reference bias was modulated at 1 kHz for both scans In-phase: filled symbols. Quadrature: open symbols. The line is shown as a guide for the eyes for the near-UV portion of the spectrum.

The correlation between the Alq₃ EA signal and the EM band is made clear by the EM spectra of the OLED in forward bias both above and below the turn-on voltage (2.5 V). Figure 4 shows the EM spectrum of the device at a dc bias of 1.7 V, where the EM signal is changing most rapidly with the dc bias, and at a dc bias of 3 V, at its maximum value. A larger reference bias could be used for the spectrum at 1.7 V as the bias dependence is linear between 1 and 3 V. This permitted a better signal-to-noise ratio to be obtained in the portions of the spectrum where the spectrometer throughput was weak (< 450 nm). The measurement above the turn-on voltage was performed in balanced mode and the lock-in phase was set to that of the EL to eliminate any phase shifts associated with the electrical circuit. Both spectra are dominated by the EM band as one might anticipate by the sharp increase in the bias sweeps. Unlike the spectrum measured in reverse bias, there is a slight shift in the phase of the signal, resulting in a non-zero quadrature signal with the same lineshape as that of the in-phase signal and much weaker amplitude.

3.2.

Field-Modulated Excited State Absorption

We consider the origin of the EM band with a minima at λ = 520 nm. This feature cannot arise from EA of the transport layers as it lies beyond the cut-off wavelength (see Fig. 1) nor can it originate from EA of the hole injection layer as it cannot support a strong internal electric field. We note that this signal was detected below the turn-on voltage and thus neutral species such as singlet and triplet excitons are precluded. Furthermore, it was not detected in studies of bi-layer α-NPD/Alq₃ OLEDs without the HIL.^{29, 30} These observations lead us to focus on the HIL/α-NPD interface as the origin of this feature with the possibility that the HIL may be partially oxidizing the α-NPD layer. Upon removing a charge from a π-conjugated molecule, structural relaxation shifts the LUMO and what was the HOMO into the gap.³² Adding charge to a π-conjugated molecule results in similar relaxation. In consequence, induced absorption bands form a below the absorption edge and in the infrared arise and these could account for a subgap EM signal.

There are three possibilities which could account for the EM band: α-NPD cations, Alq₃ anions, or charges in the HIL. We consider each in turn. Alq₃ has previously been observed to react with the cathode or an electron injection layer such as LiF. Choong studied the interface between calcium and Alq₃ by ultraviolet photoemission spectroscopy.³⁷ Doping-induced absorption was seen between 0.8 and 2.0 eV (λ > 600 nm) and the maxima of subgap absorption occurs at 720 nm. The EM band lies at a significantly shorter wavelength and therefore cannot originate from Alq₃ anions. Jiang measured the absorption spectrum of PTT:PFFSA and showed that the absorption peak from doped states occurs at around 1500 nm.³⁵ Furthermore, the absorption cut-off of the HIL lies below 500 nm and therefore this feature cannot be related to de-doping of the HIL. We are therefore left with the possibility that the field modulation of α-NPD cations could be the source of the EM band.

In order to prove this assignment, we measured changes in the absorption spectrum of α-NPD following oxidation by FeCl₃, a Lewis acid. Solution of α-NPD and FeCl₃ in chloroform were prepared at a concentration of 0.017 M (10 mg/ml for α-NPD). Figure 5 shows the absorption spectra as the α-NPD solution was progressively doped by addition of FeCl₃ (roughly 5% addition by volume each). An optical transition with a maxima around 480 to 500 nm appears immediately upon addition of FeCl₃ and strengthens with increasing amounts of FeCl₃. A second lower-energy transition is seen in the long wavelength portion of these spectra. The lineshape of the doping induced absorption spectrum was independent of the dopant concentration and we did not observe a second oxidation step due to dication formation. Figure 5 compares the EM band with doping-induced changes in α-NPD absorption. The EM band occurs at a slightly longer wavelength than solution doping, due to solvatochromic effects. These results are consistent with those observed by Endo, who studied co-evaporated films of α-NPD and FeCl₃.³⁶ Given the close resemblance between the EM band and α-NPD⁺ absorption and the dissimilarities with charged species of Alq₃ and PTT:PFFSA, we therefore conclude that this band is due to ESA by field-modulated α-NPD cations.

Fig. 5

(a) Absorption spectrum of a 17 mM α-NPD solution as it is progressively doped by FeCl₃. Inset: a Jablonski diagram of the optical transitions of α-NPD and α-NPD⁺. (b) Dashed line: doping-induced absorption spectrum of α-NPD. Solid line: difference between the EA spectrum of a single layer Alq₃ device and that of the bi-layer device with a PTT:PFFSA hole injection layer.

Further insight into the nature of the species responsible for ESA is obtained by considering its sign, phase, and amplitude. For EA of a device, increased absorption occurs when the reference electric field has the same polarity as the dc field. For a device in reverse bias with respect to the built-in potential, the internal field is greatest when sin Ωt < 0. The ESA has the opposite sign as that of the EA signal and therefore follows the absolute polarity of the reference field, not its polarity relative to the built-in potential. Thus, the density of α-NPD cations increases with increasing positive bias and is independent of the polarity of the dc field.

We next consider whether the ESA signal is due to injected carriers or trapped charge. These species can be distinguished by the dependence of the signal upon the modulation frequency. The dependence of the ESA signal on modulation frequency and excitation lifetime is given by:

The doping-induced and electromodulation spectra permit us to calculate the magnitude of field-dependent doping. The transmitted probe light I(d) through the sample is given by:

2.3.

Internal Electric Field Distribution

We finally consider the effect of the HIL on the electric field distribution within the device and attempt to determine the built-in potential (V_bi). The calculation of V_bi in our devices is complicated by field-dependent modulation of the trapped charge density at the α-NPD/HIL interface as well as overlap of the EA spectrum of Alq₃ and the ESA spectrum of α-NPD⁺. Bodrozic previously considered how to determine V_bi in devices with one of several different polyfluorene derivatives as the active layer and an ITO/PEDOT:PSS anode. Careful analysis of the dependence of the EA signal upon wavelength, frequency, and applied bias enabled an accurate determination of V_bi. It was important to measure the bias-dependence of the EA peak at high frequency (Ωτ ≪ 1) where the frequency dependence of the ESA minimizes its contribution [see Eqs. 3, 4]. The ESA signal in this device is considerably stronger relative to the EA signal seen in studies of polymer LEDs.²⁸ For example, the EA signal of poly(9,9-dioctyl-fluorene-alt-N-(4-butylphenyl)-diphenylamine) was 2 orders of magnitude larger than that of the ESA signal. Hence, the ESA signal was a moderate perturbation that was amenable to analysis. The ESA signal in forward bias is roughly an order of magnitude stronger than the EA signal in the spectra of our device.

We first consider how field-modulated doping of α-NPD impacts the bias response of the EA signal. Consider the bias dependence of the EA signal shown in Fig. 3. The EA signal is linear with dc bias when the device is in reverse bias, but the slope is much weaker than one might expect. The extrapolated zero-level crossing occurs at 7.0 V, an unphysical built-in potential. Couple this observation with the relatively weak electric field in the α-NPD layer in comparison to that observed by Rohlfing²⁹ and Martin³⁰ and it becomes evident that the HIL is doing more than simply assisting injection of holes. At roughly zero dc bias, there is a sharp increase in the magnitude of the ESA signal and a concommitant increase in the slope of the EA signal. The zero-level crossing occurs at 0.3 V forward bias, well below the expected built-in potential, and is indicative of a sharply reduced field strength in the Alq₃ layer as well as overlap between the Alq₃ EA and α-NPD⁺ ESA spectra. Couple these observations with the relatively weak electric field in the α-NPD layer in comparison to that observed by Rohlfing²⁹ and Martin³⁰ and it becomes evident that the HIL is doing more than simply assisting injection of holes. Rather, the electric field appears to be concentrated in the Alq₃ layer for a device in reverse bias and the field distribution dramatically changes in forward bias.

Although the forward bias EM spectra are dominated by the α-NPD⁺ band, there are slight differences between the two spectra that are due to electroabsorption and allow us to infer the built-in potential. Figure 6 compares the two spectra, scaled to the α-NPD⁺ minima at 520 nm. There is a clear oscillatory deviation between the two spectra below 480 nm. This difference, shown in Fig. 6, approximates the EA portion of the spectrum shown in Fig. 1. It is difficult to pinpoint the exact built-in potential as the EA signal of Alq₃ for V > V_bi will be negative and difficult to differentiate from the much stronger negative signal of the α-NPD⁺ ESA band. It is clear that the spectrum measured with V_dc = 1.7 V lies below the built-in potential as the EA peak is still positive and rides on top of the negative ESA band of α-NPD⁺. The difference spectra gives a peak magnitude of [TeX:] ${{ - \Delta T} \mathord{\left/ {\vphantom {{ - \Delta T} T}} \right. \kern-\nulldelimiterspace} T} = 2 \times 10^{ - 5}$$-\Delta T\left./\phantom{-\Delta TT}\right)T=2\times {10}^{-5}$. Given the EA response of Alq₃ [see Fig. 1], this corresponds to a potential difference of roughly 0.5 V and a built-in potential of 2.3 ± 0.1 V.

Fig. 6

(a) EM spectra of the bi-layer OLED from Fig. 5, scaled to the maxima of the α-NPD⁺ peak. Filled symbols: measured with 1.7 V dc bias. Open symbols: measured with 3 V dc bias. (b) The difference spectrum between these two EM spectra.

The observation of modulated charge injection into α-NPD has significant implications for device operation. We observed a suppression of the electric field in the α-NPD layer in reverse bias, possibly due to accumulation of holes at the α-NPD/Alq₃ interface would concentrate the electric field in the Alq₃ layer. Small molecule OLEDs have been produced on similar principles, using a doped layer of the hole transport material for enhanced hole injection. Endo used a composite α-NPD:FeCl₃ layer to enhance hole injection in bi-layer OLEDs.³⁴ Leo and co-workers have demonstrated highly efficient phosphorescent OLEDs using doped hole transport and electron transport layers. One potential limitation of this approach is that ions can migrate through the organic layer, reducing device efficiency and lifetime. Our results suggest that a polymer HIL has a similar effect with the added advantage that doping is field dependent rather than on a concentration gradient and proceeds on a time scale of order microseconds.